BENEFITS FROM IMPROVEMENTS
                               IN
                    CHESAPEAKE BAY wATER QUALITY


                            Volume III
                               of
                      BENEFIT ANALYSIS USING
                 INDIRECT OR IMPUTED MARKET METHODS

                        (Budget  Period II).
DEPARTMENT OF AGRICULTURAL AND RESOURCE ECONOMICS
                         SYMONS HALL
                    UNIVERSITY  OF  MARYLAND
                        COLLEGE PARK 20742

-------
                BENEFITS FROM IMPROVEMENTS

                            IN

               CHESAPEAKE BAY WATER QUALITY


                         Volume // /

                            of

                  BENEFIT ANALYSIS USING
            INDIRECT OR  IMPUTED  MARKET METHODS

                     (Budget Period  II)
                  Prepared  • nd Edited by

Nancy  E. Bockstael,  Kenneth E. McConnell  • nd Ivar E.  Strand
                  University  of Maryland
                  Sections Contributed by

            Douglas Larson  • nd Bruce  Madariaga
                  University  of Maryland
                  Final  Draft  -  April 1988



                  Principal Investigators

        Nancy E.  Bockstael • nd Kenneth E. McConnell




             EPA Contract  No. CR-811043-O1-O
                       Project  Officer
                      Dr.  Peter  Caulkins

-------
"The information  in  this document  haa  been funded wholly or
in part by the United States Environmental  Protection Agenty
under Contract No.  811043-01-0.   It haa been subject to the
Agency*,  peer and  administrative review, and it  has  .  been
approved  for publication as an EPA  document. Mention  of
trade  names  or commercial products  does not  constitute
endorsement or  recommendation  for use. "

-------
                      Ac knowledgement s
First and  foremost,  we wish to  thank Dr.  Peter Caulkins,  our
project   officer,   who  has exhibited patience, wisdom,   and
appreciation.

Douglas  Orr and Firuzeh Arsenjani  contributed to the  analysis
in this  volume.  Special thanks  are given to Marion  Story for
her commitment  and technical  assistance.

All opinions  and remaining errors are  the  sole responsibility
of the  editors.  This  effort waa  funded by USEPA Cooperative
Agreement number CR-81 1043-01-0.

-------
                                  Forward
    This is the second of two reports  constituting the  final report for budget
period  II  of  Cooperative Agreement #811043-01-0, which was initiated   and
supported by the Environmental Protection Agency.   From  the  beginning,  this
cooperative agreement haa dealt with improving  methods of measuring the
benefits of environmental improvements.    Budget period  I  of the agreement
produced   two documents which considered theoretical, conceptual  and
methodological   issues  involved,  in  using  hedonic  models  (Vol.  I)  and
recreational demand models (Vol.  11)  evaluating environmental improvements.

    The second budget period's  work has extended  the work of  the first,
especially in the area of recreational demand models.   Volume  I of budget
period  II's final report looks at  the  theoretical issues of measuring the
benefits of quality changes,  the conceptual issues surrounding perceptions of
water quality and methodological  issues  related to  estimating  models with
sample selection problems.

    The  report which follows  is  the second part of  budget period II's final
report.   This  report provides information  on  the recreational  activities which
take place on the Bay, as well  as  the monetary values people place on these
activities.   While  not  commissioned with the intent ofhelping in the process  of
revising the Bay plans, we hope that the discussions in this report will do
just that.

-------
                             Executive Summary
    For more  than ten years, the  Chesapeake Bay has been the focus of  an
impressive amount of research and an array of environmental programs.  The
Chesapeake  Bay Program, a  cooperative effort by  the  federal government,
Maryland,  Virginia,  Pennsylvania,  and the District of Columbia, represents a
coordinated  commitment  to enhancement of the water quality of the Chesapeake
Bay.

    Large commitments  of money have been  made to  clean up the Chesapeake
Bay.    Yet there  is  little understanding of  the  nature and extent  of the
benefits which are derived  from these massive commitments.   Raising this issue
does not imply that  either  the programs  are  misguided or need to be justified
on some benefit-cost  criterion, for many believe  that  the cleanup process is  an
expression  of a fundamental morality  that  despoiling our  surroundings  is
wrong.   But  understanding more  precisely how people benefit from cleaner
water in the Bay can help in allocating  resources to clean up the water,  for
funds must be allocated temporally,  spatially and  functionally.    Perhaps
knowledge about the benefits  from water quality improvements can help with
those  decisions.

    Even though the returns from the  Program derive  from human benefits--
human use  and  human health--the specific objectives  and implementation
strategies are designed to affect chemical and biological characteristics of  the
Bay.   The connection between human benefits on the one  hand and  reductions
of nutrients and toxic  materials" on the other remains  implicit.  Perhaps  the
clearest  link  is  between human use and  fisheries and  wildlife management.
Here the vehicle for linking the strategy and the goal  is at least  under-
standable, even  if the details of  this linkage  remain indistinct.    In other
cases, however, we are left  confused as to how the policies  impact on  humans
and how we would evermeasure the success of  these  policies in terms oftheir
achievements.

    Thie report attempts  to focus  attention on the human use of the
Chesapeake Bay.   It describes something about the nature  and level of that
use.   It also considers what we know  and  what we do not  know about the
relationship  between chemical and biological characteristics of the Bay and
human use.    This  relationship  must be understood in order  to address the
more complex measurement of human benefits.

    One  objective of the  report  is  to  provide estimates  of Values of
Chesapeake  Bay recreational  activities and  willingness-to-pay estimates  of
improvements  in  water quality associated with these activities.   Available data
has been used together with what is known  about  estimating  environmental
benefits.  While Chapters 3,  4,  5  and 6 reflect our beat efforts at this task,  it
should  be kept in mind that benefit estimates have  an elusive nature.  A
number  of different studies have been  assembled,  and an array of methods
and specifications  has been used to provide as  much information as  is
currently available on the topic.


                                    iv

-------
    The benefit estimates  themselves do  not represent  this report's most
important  contribution,  however.   We  seek to describe,  model  and to some
extent explain recreational uses of   the  Chesapeake.   We may have serious
doubts about the precision of willingness-h-pay estimates,  but we still learned
a great deal about the  factors  which matter to people in using  the Bay, the
obstacles  to  their increased  enjoyment  of  the"^ Bay and the distributional
implications of  improving the Bay.

    Specific information contained in this  report  includes:

    . Maryland boaters,  beach users andsportfishermen alter their behavior
       in response to poor water quality, as  scientifically measured.

    • Demographic factors, such es  income and location of  residence,
       influence observed use of the Bay.

    . Cent  ingent  valuation experiments  (hypothetical responses)  reveal an "
       annual willingness to  pay  in increased taxes of over  $100  million for
       improvements in Chesapeake Bay water quality.

    . Observed behavior of Maryland  western shore beach users  reveal  an
       annual  willingness to pay for 20 percent  improvements in water quality
       of between $2 million and $26 million.

    •Many  of the gains from water quality  improvement are concentrated in
       the area of heaviest  use  around Annapolis, Maryland.

    The estimates give magnitudes  for the  annual benefits  to residents  of the
Baltimore-Waehington  area of improving  water quality  in the Bay  in  the range
of from  $10 million to  over $100 million.  There are numerous sources of error
and random elements in these  estimates,  and  several  activities  and populations
have  been  omitted.    But  based  on these numbers,  it  seems plausible  to
estimate that  the annual  returns to  cleaning up the Chesapeake are at least  of
this order  of magnitude.

    The long-run annual benefits will be higher than these  estimates,  however.
Firat, as people learn that  the  Bay  has become cleaner, they will  adjust their
preferences toward Bay recreation.   This  is  especially true of people  who  do
not  currently  use the  Bay and are largely excluded  from the analysis.
Second,  the population and  income ofthe area have grown since  1984, and
both are likely to  grow more,  increasing the  demand for and  value  of
improvements in water quality.   Finally, we  have ignored  the  value  (both use
and existence value)  which  households  outaide the Baltimore-Washington area
may  have  for the Bay.    The Chesapeake  Bay is  a nationally prominent
resource.   Its  improved health  is of  value to many who  will never use  it.

    In conclusion,  we hope  this volume will provide a stimulus to  decision
makers to  refocus attention on human uses of the  Bay.   Human uses and the
protection of human health have always been  the central theme of clean water
legislation, but because of difficulties in  relating  these to specific standards,
they have  often dropped from sight in the formation ofthe actual programs.
We hope to shed some light on ways in which Bay cleanup policies might  be
related  to  the behavior and preferences  of actual and  potential users  of  the  Bay.

-------
                              Table of Contents
Chapter                                                                 Page

  1.   Enviromental  Programs  for  the Chesapeske Bay	1
          The State of the Bay's Water Quality 	 1
          The Current Environmental Programs  	 2
          The Role of This  Report	4

  2.   The Role of Human Use Activities in Defining Goals and Strategies
      for the Chesapeake Bay	7
          Systematic Evidence  of the Link Between Perceptions
          and Behavior	     8
              Seme General  Attitudinal Patterns  	  10
              Two Propositions about Water  Quality and Behavior ....  12
          Results of Focus Group  Experiences  	  17
              Sumnary of Focus  Group Experience	22
          Conclusions	23

  3.   Recreational Uae of  the Bay and Willingness topsy  Estimates
      for  Improvement to  the Bay as a Whole	    ...   25
          Recreational Use of the Bay	    26
          Aggregate Willingness to  Pay	,, .  .    29
              Analysis  of Willingness  to  Pay  Responses	    30
              Results of Analysis by Subgroups  of Respondents  .  .   • •  •    32
              Regional Comparisons  	   ...    36
          Existence Value	    38
              Existence  Value  Experiment 	   ...    39
              Interpretations  of Results	• •  -    40
              Conclusions	 . *    41

  4.   Effect of Chesapeake Bay Water Quality on Beach  Use	42
          Tbe SurveyandtheData  	  44
              The Data	45
          The Varying Parameter Model  	  47
              The Varying Parameter Model  Estimates	50
              Estimated Benefit Changes	53
          Discrete/Continuous Choice Model	59
              The Choice Among Sites	59
              Estimation of Discrete Choices Among Beaches ........  64
              The Number of Trips Decision	67
              Estimated Benefit Changes 	  68
          Discussion	70

  5.   Recreational  Boating and the Benefits of Improved  Water Quality  .  71
          A Profile  of Boaters and Boat Owners	71
              The BoatOwnerSurvey	71
              Boaters and  Boat Owner Characteristic  	 71
              The Importance of Water Quality  to Boaters 	  77
          The Behavior  of  Boat Owners  Who Trailer Their  Boats  	  77
              TheGeneralModel 	  77
              The Data	80
              TheEstimatedModel  	 81

-------
        Modelling the Behavior of Boat  Owners Who Do Not Trailer
        Their Boats	83
        Calculating Estimates of the  Benefits of Water Quality .... 85
            Improvements  for  the  Trailered Boat Sample	85

6.   The Benefits for Recreational Fishing: Striped Bass	90
        A Description of the Data	92
        TheBasicModel  	 95
        Empirical Results 	 96

7.   Summary and Conclusions	   100
        Demand for Chesapeake Bay Recreational Activities 	   100
        Estimates of Benefits from Water Quality Improvements ....   102
            Caveats	   103
            Estimates	   105

    Reference	109

    Appendices
        The Random Digit Dialing Telephone Survey Procedures 	  121
        Telephone Survey  Instrument  	  125
        The User Survey and Sampling Procedure	139
        User Survey Instrument	145
                                  vii

-------
                                    Tables
Numbers                                                                   Page

 2.1    Parameter Estimates  and  Standard Errors for Weibull
        Distribution	13

 2.2    Effect  of Perceptions on Use	14

 3.1    Participation Rate in Chesapeake Bay Activities by Activity
        and Area, 1984	27

 3.2    Joint Participation Rates in Chesapeake Bay Activitiae
        By Activity and Region,  1984  .  .   •  —-  ,......,,,....   28'

 3.3    Percent of People Willing to Pay Additional  Taxes for Water
        Quality Acceptable for Swimming,  by Amount of Tax	30

 3.4    Logistic  Model  Estimates Related  to the Probability
        a Respondent  Will  Accept a Tax Increase	33

 3.5    Estimates of  Utility Parameters  by Income Group 	 34

 3.6    Expected Value of Willingness to Pay for Water Acceptable
        for  Swimming,  by  Income Group and User Group,  1984	34

 3.7    Estimated Willingness to Pay for Acceptable Water Quality
        by Participation and Racial  Composition of Household,  1984  ... 35

 3.8    Estimated Aggregate Willingness to Pay for Water Quality
        Acceptable  for Swimming,  by Classification and Scenarios, 1984  . 36

 3.9    Logistic Model Estimates Related  to the Probability  a
        Respondent Will Accept a Tax Increase to  Improve Chesapeake
        Bay Water Quality, by Geographic Area	37

 3.10   Estimated Willingness  to  Pay  for  Acceptable Water Quality
        by Region, Participation, and Racial Composition of Household,
        1984	38

 3.11   Summary  Results of Contingent Valuation Experiment
        on Existence Value	40

 4.1    Average Values of  Regression Variables for Visitors, by Beach  .  . 51

 4.2    Tobit Estimates  for  Beach Demand  Model, by Beach 	   52

 4.3    Annual Benefits per  Beach User froa a 20  Percent Decrease
        in Pollutant, by Beach,  1984	  56

 4.4    Annual Benefits per  Beach User from a 10  Percent Decrease
        in Pollutant, by Beach,  1984	  57

-------
4.5    Annual Losses per Beach User  from a 20  Percent  Increase
       in Pollutant, by Beach, 1984   	   58

4.6    Aggregate Benefits/Losses  to  Users from  Changes  in  Chesapeake
       Bay Water Quality,  by Beach,  1984	60

4.7    Logit Regression for Selection Among  Sites 	   66

4.8    Log it Analysis for Selection  Between  State Parks  and
       Local Beaches	„»,„.,,,»,,,»....   66

4.9    Ordinary Least Squares Regression for "Choice Occasions"
       or Trips	   68

4.10   Comparison of  Benefits Based on a Varying  Parameter Model
       and-Discrete/Continuous Choice	   70

5.1    Average Characteristics of  Boaters and Non-Boaters
       in Baltimore/Washington SMSA,  1984	   73

5.2    Average Characteristics of  Boat Owners and Non-Boat Owners
       in Baltimore/Washington SMSA,  1984	73

5.3    Characteristics of Boats  and Boat Owners from
       Boat Owners'  Survey,  1983	74

5.4    Numbers  of  Trailered Boats  and Boats Kept in the Water
       By Residence,  1983	76

5.5    Percent of Boaters Who Fish or Swim While  Boating,
       Boat Owners' Survey,  1983	   77

5.6    Factors Cited Most Important  in the Selection of a  Boating
       Area in Maryland   - Percent Response from 718 Boat Owners"
       WhoTrailerTheirBoats	78

5.7    Factors Cited Most Important  in the Selection of a  Boating
       Area in Maryland - Responses  from 788 Boat Owners Who Keep
       Their  Boats in Marinas	78

5.8    Estimated Tobit Demand Coefficients:   Maryland  Counties, 1983  .  .  82

5.9    Estimation Results  from Second  Stage	  83

5.10   Estimated Demand for Boating Tripe -  Boats Kept  in Marinaa  ...  84

5.11   Estimated  Demand for Boating Trips -  Boats Kept in  Marinas.
       Fishing Behavior	  85

5.12   Par  Boater Annual Benefits  from a 10% Decrease in Pollutant
       by Geographical Area	87
                                     IX

-------
5.13   Per Boater Annual Benefits from a 20% Decrease in Pollutant
       by Geographical Area	88

5.14   Per Boater Annual Losses from a 20% Increase in pollutant
       by Geographical  Area	89

6.1    Characteristics of Striped Bass Fishermen and Other
       Fishermen/Hunters  in  the  Sample  	  91

6.2    Sample Distribution Number of Fishermen, Days of Striped
       Bass Fishing in  1980,  and Catch Rate, by Regions 	 93

6.3    Tobit Estimation of the Demand  for  Striped Bass  Fishing 	 97

6.4    Aggregate Consumers'  Surplus for Striped Bass Fishing:
       Effect  of Changing Catch Rates	98

7,1    Aggregate Benefits for Three Water-related Activities
       from a "20%" Improvement in the Chesapeake Bay's
       Water Quality,  1987 dollars	106

7.2    Aggregate Benefits from Water Quality  Irrprovements-
       Contingent Valuation,  1967 dollars  	 107

-------
                                    PigUrea


 Numbers                                                                  Page

 2.1     The Sampling Region  for  the  Telephone  Survey  and  Location  of
        the  Beaches Used in the  Intercept Survey	9

,2.2     Annual  Net Change  in  Swimming Habits  1970-1964	15

 2.3     Cumulative Net Change  in Swimming Habits 1970-1964	16
                                     XI

-------
                                 Chapter 1

               Environmental Programs for the Chesapeake Bay
    Over the past ten years,  the Chesapeake  Bay has been the  focus of an
impressive  amount  of  research and the beneficiary of a  great  many
environmental programs.    Concentrated  efforts began  in 1976  when  the
Congress directed the U.  S.   Environmental Protection Agency  to  conduct a
five-year study of  the  Bay's  resources and water quality.  The  study,  which
focused on  three  major problems  of  the  Bay--nutrient  overenrichment,  toxic
substances?  and  the  decline of submerged aquatic vegetation--prompted action.
In 1983 the  three surrounding states, the District  of Columbia,  and the  federal
government signed a  pact,  the Chesapeake Bay Agreement  of1983, committing
them  to  improve and protect  water quality  of the Chesapeake  Bay  through
coordinated  environmental enhancement activities.     In late  1987 a  new
agreement was signed.


The State of  the Bay's Water Quality

    Concern  over the  Chesapeake Bay stems  from declines  in direct  and
indirect measures of the  quality of the Bay's  waters.   The most apparent
measures are related to the productivity of  the Bay.    Reduction in fish
landings, combined with an awareness  of the increasing loads ofpollutants and
their consequences,  led scientists to assess  the  Bay's water quality.

    The use of the  term  "quality"  in assessing  the Bay connotes  a set of
standards goals, or  ideals with which  the  current conditions ofthe Bay can
be compared.  The quality  of  the water depends on one's standards, and the
relevant standards  depend on  intended uses  and frame  of  reference.   For
example, if  the most desired use of  the  water were for transportation,  then
the Bay's current water quality would be quite satisfactory.    At  the other
extreme,  if  one's standards are derived relative to the state ofthe  Chesapeake
Bay three centuries ago,  its current quality is clearly too  low.

    Since the thrust of the  Chesapeake Bay  program cornea from observed
declines in  ecosystem productivity, it is  useful to summarize the nature of
those declines. Summary measures give the status of the Bay as  a whole,  but
mask considerable  differences  in  quality between the upper and lower Bay and
among the  various  river systems and inlets  of the Bay.    The following
measures  suggest the nature of the thinking that led to the conclusion that
the Chesapeake Bay waa declining in quality.

    There are two kinds of evidence of the historical decline  in  the Bay's
water quality.   First there are  scientific measures which are indicators of
impairment of the Bay as  a functioning ecosystem.   A common measure  of water
quality is  the level  of dissolved  oxygen in the  water. This  is oxygen available
to various plant  and animal life.   its absence can eliminte higher forms of  life
from ecosystems.   Studies have shown  that the  extent ofwater with little or
no dissolved oxygen has  increased by 15-fold in the  last 30 years.

-------
    Another indicator of water quality is the level of nutrients  (nitrogen  and
phosphorus, mainly).   These  nutrients, while not harmful per se,  enhance algal
growth,  whose  decay increases the  demand for  oxygen.    The increase of
nutrients in the Bay's waters is an indirect consequence of population  growth,
changing technology and  industrial and agricultural expansion  in  the area.

    The decline in submerged aquatic vegetation  (SAV)  is  another indicator of
the decline in  the Bay's water quality.   The decline  in SAV is connected with
turbidity   and   growth  of  epiphytes  and   phytoplankton,   by excessive
nutrification. The loss  of SAV means less suitable habitat for  spawning  finfish
and shellfish.    There are  ofcourse  many other indirect measures  of the
declining health of the Bay.   They all reinforce the notion that  human  factors
are destroying the traditional ecological  linkages of the Bay.

    There are other  signs of declining water quality more cogent to  the lay
public.    Landings ofwell-known  anadromous  species  such as  rockfish and
shad have dropped precipitously in the past several  decades.  Oyster harvest
and oyster reproduction have also declined  in the  past decade.   There  is some
ambiguity in the use of landings as a measure  of water quality,  of course. A
considerable increase in effort devoted to harvesting fish has happened to
coincide with the increase of effluents over  time.   Further,  natural phenomena
such  as  hurricane Agnes (1972)  induce cyclical variations in  finfish and
shellfish reproduction.   Nevertheless,  there  can be little doubt that the quality
of the Chesapeake Bay's waters has declined, both in terms of the ecological
health of the estuary and the benefits to humans of its use.


The Current Environmental  Programs
    The foundation of the Chesapeake Bay Program is  the Clean Water Act,  the
ongoing federal  environmental legislation  dealing with water.  Under the  Clean
Water Act, appropriations  have been  made available annually to the Chesapeake
Bay Program, providing both its operating budget  and its grant funds.   The
relationship between  the  federal legislation and  Chesapeake Bay activities goes
beyond funding,  ofcourse, since  the Clean Water Act  establishes the
guidelines by which  states  then set  specific  water  quality  standards.   The
Water Quality Standards Handbook is  the most recent  document which contains
the guidelines  prepared by  EPA to  assist states in implementing 1983 revisions
of the water quality  regulations.

    The Handbook defines  acceptable  approaches  by which water  quality baaed
effluent  limitations may  be determined.    Whether the pollutant  specific or
biomonitoring  approach is taken,  however,  states must adopt criteria which are
sufficient to protect the "designated uses" of a water body.   Determination of
designted uses requires an "attainability analysis,"  i.e.  physical, chemical  and
biological studies  to identify the suitable potential uses of the water and to
determine whether these  uses have been impaired.    There is,  throughout,  a
clear sense  of the central position which human use activities should  play in
the  setting  of  standards and  the overriding  obligation states  should  feel
toward the  protection of   human health where  'people are  involved in
recreational  uses of  aquatic  resources.

-------
    The first plan of  the Chesapeake Bay Commission  waa the Chesapeake  Bay
Restoration  and Protection Plan of September 1985.   This  is currently  the
central document describing the goals  of the Chesapeake  Bay Program and  the
means by which  these goals are being  achieved.   The  general  goals as stated
in  the  plan are  to

     "Improve and protect  the water quality and living  resources  of the
     Chesapeake Bay  estuarine system (in order)  to restore and maintain
     the Bay's ecological integrity, productivity,  and beneficial uses and
     to protect public health." l

The goals of  the  Restoration and Protection Plan are boad, and include both
ecological and human health, as well  as productive use by  humans.  By and
large,  however,  there  is  no clear connection  between  the  goals  of the Bay
Program which  emphasize human health  and human use  and the  means by
which humans  benefit  from  implementation.

    To accomplish the broad goals, specific objectives and implementation
strategies have been developed.   Many  of  these strategies  are  designed to
reduce  orcontrol nutrients.   Major  strategies  to control point  sources of
nutrienta include plane to provide grants to design,  construct,  operate  and
maintain sewage treatment  facilities,  and  plans  to support phosphorous removal
projects at treatment plants.   Plans to support nitrogen  removal at treatment
plants have  not been proposed,  except for an experimental  project conducted
by the State ofMaryland in the  Patuxent estuary.

    The primary  stratagy  established to control non-point  sources of nutrients
to the  Bay has been to subsidize the implementation  of "Best  Management
Practices"  (BMPs) to reduce runoff  from urban,  forested, and in particular,
agricultural lands.  Maryland, Virginia, and Pennsylvania have instituted  cost-
sharing  programs  to  promote agricultural BMPs.   Among the  agricultural  BMPs
supported through cost-sharing  have  been:   strip cropping, buffer stripping,
terrace and  diversion  construction,   animal waste system installation,  and
reduced tillage planting.   Some  of these practices are  employed to reduce
sediment,  pesticide,  and herbicide  runoff, as well  as nutrient runoff.  A
secondary strategy to control non-point source  pollution is  to control urban
runoff, in particular combined  sewer  overflows.   Tactics to control combined
sewer  overflow include  revamping  ofsewer systems  and building holding
ponds.   The  state of Maryland has also enacted legislation banning the use  of
phosphate detergents  and controlling development  along  the Bay's shoreline.

    The Chesapeake Bay Restoration and  Protection Plan  hae  enacted  a  series
of additional policies to  reduce or control the  level of  toxic materials  in  the
Bay,    Among these policies are programs  to  support pretreatment plans to
reduce  the discharge  of  metals and  organics  from sewage  treatment plants
resulting from industrial  sources,  to  fund dechlorination processes to  reduce
chlorine discharges into critical  finfish and shellfish  areas,  and to implement
oil spill response plans.
•"•Chesapeake Executive Council, Chesapeake Bay  Restoration  and  Protection
    Plan U.  S.  Environmental Protection Agency,  Sept.  1985.

-------
    Lastly, the Chesapeake Bay Restoration  and  Protection plan has instituted
a series of policies  designed directly to "provide for the restoration  and
protection  of living resources and their  habitats and ecological relationships" 2
in  the Chesapeake.    Among these policies were  programs  to  develop
comprehensive fisheries management plans,  expand oyster repletion activities,
improve  waterfowl and wildlife habitat, and re-establish  submerged aquatic
vegetation.

    There  is no  "clear   connection between the  implementation strategies
mentioned  above and the goals  of  the Chesapeake Bay  Program.   The goals of
the program are  couched  in terms  of human benefits--human health and  human
use--but the specific  objectives and implementation  strategies are designed to
affect chemical and biological characteristics  of  the  Bay.    The connection
between human benefits on  the  one hand and reductions  ofnutrients  and  toxic
materials on the  other remains implicit, unsubstantiated and  unarticulated.
Perhaps the clearest connection is between  human use and  fisheries  and
wildlife management.  Here the vehicle for linking the strategy  and the goal is
at least understandable?  even if the  details  of  this  linkage remain indistinct.
In other cases,  however, we are left confused  as to  how  the policies  impact on
humans  and how we  would ever measure  the success  of these  policies  in  terms
of their achievement of the Program's  goals.  In implementation,  the focus on
human use seems to have been lost.


The Role of This Report
    This report attempts  to  focus  attention on the  human use  of the
Chesapeake Bay.   The report  describes something about the nature  and level
of that use. It also considers what we know and  what we do not know  about
the relationship  between chemical and biological characteristic  of the Bay  and
human  use.   This  relationship must be understood in  order to  grapple with
the more complex measurement of human benefits.

    Large  commitments of  money have been made  to clean up  the Chesapeake
Bay.    Yet there is little  understanding of  the nature  and extent of  the
benefits which  are derived from these massive  commitments.   How do people
gain from  the cleanup?   Asking this question does not imply  that either  the
programs are misguided or need to be justified on some benefit-cost criterion,
for many believe that the  cleanup process  is an expression of  a fundamental
morality that despoiling our surroundings is wrong.   Whataver  the motivation
for environment, al improvement, we  believe that understanding more  precisely
how people benefit from  cleaner water in the Bay can help  in allocating
resources to clean  up  the  water.   Moral imperative are of limited usefulness
in the  tactics of cleaning up the Bay,  Even with commitments for a cleanup of
the Bay, one must allocate  those funds  temporally, spatially  and functionally.
Perhaps knowledge  about  the benefits  from water quality  improvements  can
help with  those decisions.
 Chesapeake Bay Restoration and Protection Plan, Chapter  2,  page I,

-------
    One of our objectives is to  provide some initial  estimates of values of
Chesapeake Bay recreational activities  and willingness-to-pay estimates  of
improvements  in water quality associated with these activities.   We have used
available  data together  with  what we know about estimating environmental
benefits  (see Bockstael, Hanemann and  Strand,  1985; and McConnell,  Bockstael
and Strand, 1987)  to determine  these "ball park" willingness-b-pay figures.
While Chapters 3, 4, 5 and  6 reflect  our best efforts at this  task, it should be
kept  in mind that benefit estimates  have an elusive nature. Much  has been
written about the imprecision  ofnon-market benefit estimation, and we will
have more to  say on the  question in this report.    The  usual difficulties are
compounded by generally  sketchy data.   No  new surveys were conducted  for
this  study,  and only one ofthe surveys  used in the subsequent chapters  was
designed with benefit estimation in  mind.   Nonetheless we have put  together  a
number of different  studies  and have  used an array   of methods  and
specifications  to provide as  much  information as is currently available  on  the
topic. We  provide  estimates ofaggregate benefits  for  a variety of recreational
activities.     We have also tried to provide information on the relative
magnitudes  of benefits  which are likely to accrue to  different groups of users
and to improvements in  different geographical areas.

    The benefit estimates themselves do  not represent this  report's most
important  contributions however.    We seek to  describe, model and to some
extent explain recreational uses of  the Chesapeake.    The report  represents an
attempt to  begin to understand  the preferences and behavior of individuals
toward the Bay.   Models ofbehavior are essential to benefit estimation.  Even
in the face ofhuge uncertainty over  benefit estimates,  the  underlying
behavioral models can  provide  useful and reliable  information. We may have
serious doubts about the precision of willingness-to-pay  estimates,  but  we  can
still learn a great deal about the  factors which matter to people in using the
Bay,  the  obstacles   to  their  increased enjoyment of the  Bay and  the
distributional implications of improving the Bay.

    The Restoration and  Protection Plan is  an  interim plan,  "the first iteration
of the planning effort implemented in response to this commitment.  "  As such
it is a first move in  the direction ofChesapeake  Bay improvements but it is
sub ject to  revision  and fine-tuning as  goals of environmental improvement
become clearer and information about problems, technology  and  costs becomes
more  sophisticated.

    What we hope this  volume will provide is a stimulus to  decision makers  to
refocus attention on human  uses  of the Bay, as the  goals and the  strategies
for achieving these goals  are fine-tuned  in the coming year. Human uses  and
the protection of human  health have always  been the central theme of clean
water  legislation, but because of difficulties in relating these to specific
standards,  they have often  dropped from  sight  in  the formation ofthe  actual
programs,   We hope to  shed some  light on  ways in  which Bay cleanup policies
might be  related to the behavior and preferences of actual and potential users
of the Bay.

-------
    This  report on the Chesapeake  Bay is part  of  a larger EPA Cooperative
Agreement.   The  initial agreement  dealt with improving methods of measuring
the benefits  ofenvironmental  improvement,  and did  not deal  with  the
Chesapeake Bay.    This report  provides information on the recreational
activities  which take place on the  Bay,  as well as the monetary values people
place on these activities.  While not commissioned with the intent of helping in
the process of revising the Bay  plans,  we hope that  the  discussions  in  this
report  will do  just that.

-------
                                  Chapter  2

       The  Role of  Human  Use Activities  in Defining Goals and Strategies
                           for the Chesapeake Bay


    According to  the  EPA's Water  Quality Standards Handbook.  "States  must
adopt water quality criteria sufficient  to  protect designated uses.  "   In the
process  of developing  standards, if water  body assessments  are called  for,
they  must   "characterize present uses,  uses  impaired or precluded,  and the
reasons why uses  are impaired  or precluded. "

    The  definition of  designated uses which must be protected remains  a
murky issue. Underlying much of the document is the implicit assumption  that
chemical, physical and biological parameters  can be used to define  uses.  On
the other hand  there  is  some acknowledgement of  the reality that human
activity does not necessarily align itself *with physical and chemical  water
properties:

            "The  basis of this policy is that  the States and EPA have
        an obligation  to do as much as  possible to protect the health
        of  the public  even though it may not make sense to encourage
        use of a  stream for  swimming or wading because  of  physical
        conditions.   In certain  instances,  particularly urban  areas,
        people will use  whatever water bodies are available for
        recreation.  "

    At the heart  of the dilemma is the disparity between the goals which are
couched  in terms of  human uses and the  targets of  policy actions which are
callibrated in ambient  pollution levels.    There is no one-to-one mapping
between  human use and scientific  measurement.  Failure  to  come  to  terms  with
their  relationship has lead to  something of a schizophrenia about human
activity and  scientific measurement in the Water Quality Standards.   This
schizophrenia is  not unlike that found in the  recreational demand literature
which typically seeks  to value environmental amenities by  relating behavior to
changes  in  ambient  pollution levels without  explaining how people  perceive
pollution.

    The  connection between human activity and  scientific measures  of ambient
water quality is  further obscured by the  considerable  ambiguity one finds in
both  these  discussions  about  the ways  in which individuals gain from  water
quality  improvements.   In the EPA Water Quality Standards Handbook,  we  find
repeatad reference to  protecting "uses," i.e.  recreational activities, and at the
same  time  a sense  of obligation to protect  human health.   One might argue
that these  two concepts  are coincidental,  that  we are interested in the health
of humans as they  participate  in recreational activities  in the Bay.   In terms
of pollutants which the individual can see  (or smell or  learn  about in  some
less direct way),  an individual's criteria  forusing the Bay  are likely to exceed
those minimum standards required to avoid health risks. On the other hand,
the individual is totally unaware of health  risks  stemming  from  pollutants
which  cannot  be  easily detected.   Thus many recreation decisions  are not
directly guided by quality  characteristics associated with health  standards.

                                     7

-------
    Whether modelling  recreational decisions or developing  standards, we  face
a dilemma when we  try  to  link water  quality  and human behavior.    The
obvious way  to measure water quality  is  through chemical  and physical
readings perhaps supplemented by assessments of  the  biological resources.
But water  quality improvement is  undertaken  to  enhance society's welfare
which is recognized  as deriving,  in  large part,  from human use.   Does  human
use respond to  the changes  in the  chemical  and physical composition of the
water which physical and biological sciences measure?    Are factors which
affect  their  health  the  sole factors which matter to people?    Can people
perceive changes in these measures?

    These questions not  only plague benefit measurement, they are central to
environmental  policy  making.   If gods  are  fundamentally  human oriented and
standards are scientifically  based, then the disparity between  the  two must be
resolved for environmental regulation  to achieve its potential.   In what follows
we present evidence about the human side  of  the  problem.

    First we  present descriptive information on the  variation in household
perceptions of  the Bay based on two surveys.   These  survey results do not
reveal anything  about the formation of perceptions however.    To gain some
insight into this process, we use the focus  group approach,    The  material
discussed in  an earlier  volume of this report is summarized  in  this chapter,
and insights  from our focus group experience are offered which are  specific
to water quality in the Chesapeake.   From  these various sources,  we draw
some implications for  environmental policy.


Systematic  Evidence of the Link Between Percept.ions and  Behavior

    Evidence on what people  think of the  water quality ofthe Chesapeake and
how they behave toward the Bay comes  from  two surveys:  an  on-site survey
of beach users  and a telephone survey (Figure  2.1).   Our telephone survey
was conducted May 1,  1984 to  September 1, 1984,   Research Triangle Institute
(RTI) collected  data for the  University  of Maryland on recreational  use and
perceptions of  the Chesapeake Bay using a random telephone survey.   The
telephone survey was  planned and executed jointly with an on-site survey of
beach users at  western shore  Chesapeake  beaches.  The 1,044  households with
completed interviews  were residents  ofthe  Baltimore  and  Washington SMSA's.
Demographic, attitudinal and use data were obtained.    Chapter 3 reports on
the analysis of  use patterns and activities derived from the telephone survey.
It also provides estimates of willingness to pay for Bay improvements.

    In this chapter the attitudinal  information obtained  from the telephone
survey  is examined.   This  survey allows inferences  to be made about the
impact  of  perceptions  on  decisions to use the Bay.    It  also facilitates
expansion of sample patterns  of  behavior and perceptions to  the population.

    The phone survey provides information  about broad perceptions of the
Bay, but without details about regional variation  in quality. Specific  regional
information comes from the user survey, which gathers  data about patterns of
use and perceptions for  408  users of twelve beaches on the western  shore of
the Chesapeake.   The user  survey is  described in  detail in Chapter 4 of  this
volume.

                                     8

-------
                                 Figure  2.1
The Sampling Region  for the Telephone Survey  ( jf7 ) and the Beaches  (
                       Used in the  Intercept Survey

-------
 Some General Attitudinal Patterns

     Several  important  patterns  emerged from the telephone  sample.    Forty-
 three percent  of the households responded that  they had used (31%) or
 intended  to use (12%)  the Chesapeake Bay for recreation in 1984.   For the
 users,  boating (73% of users),  sightseeing  (69%),  beach use (66%) 1, swimming
  (64%), and  fishing (56%) were quite popular activities. Hunting (5%)  and other
 uses  (20%)  were not  as prevalent  responses.   The percent  of the Bay users
who  visited eastern   shore beaches  (44%)  was nearly the same as  the
 percentage  who  visited western shore beaches  (47%). Most of the users  (81%)
 also visited ocean beaches  in 1984.

     Among those who did not use the Bay,  reasons for non-use included:

          Not interested in water-related recreation (40% ofnon-users)
          Too busy to use  (25%)
          Takes too long to get there (20%)
          Unable to use  for health reasons  (6%)
          Water quality unacceptable (5%)
          Costs too much  (5%)
          Too crowded (3%)
          Too many jellyfish  (3%)
          Other (31%)

 Personal  preferences   and the scarcity of households'  leisure  time were
 important  considerations.   Trip  costs  and poor water quality  were  not cited as
 often (5%  of the  time) but were still recognized as reasons.

     The fact that only 5 percent  of our telephone sample  stated that the
 Chesapeake  Bay water quality was responsible  for their nonparticipation may
 diminish  one's  assessment of the role of water quality to Chesapeake  citizens.
 For one thing, water quality in the Bay is not  homogeneous--it varies
 substantially and respondents in our  sample  recognized  the differences.
 Suppose respondents  living in Annapolis believe Annapolis'  water  to be
 unsuitable  for swimming  but water at Pt. Lookout  to be suitable.   These
 individuals  may  respond that time was the  prohibiting factor.   It  takes nearly
 three hours to  travel  from Annapolis  to  Pt. Lookout.     From another
 perspective, people who do not  visit the  Bay  because of  time constraints may
 know little about  the Bay's water quality  and  will not cite water  quality  as  a
 problem.

     A number  of other questions were included in both surveys to  learn more
 about  perceptions of water quality.    For  example, we  asked  telephone
 respondents,

         "Do  you consider  the water quality in  the  Chesapeake to be
         acceptable or  unacceptable for swimming and/or  other water
         activities?"

 Only 43  percent of  the telephone  respondents answered   "acceptable,"
 Alternatively stated,  57  percent  of a random sample from the Baltimore and
 Washington  SMSA's found  the Bay  water quality unacceptable  for swimming
 and/or other water activities.

                                      10

-------
    A similar question  was asked  in the user  survey concerning specific
western shore  beaches with which  the  respondent was familiar.   Even  here
there was negative reaction to  the water quality,  especially at  certain beaches.
We discovered,  in  fact, that some households found the Bay's  water  quality
unacceptable? but nonetheless used it.   There are several explanations for this
apparent inconsistency,    It is  possible that households  may find the water
unacceptable for  certain kinds  of activities (swimming)  but not  for others
 (beach use  or boating) .   Households may find the  water quality acceptable for
activities  with short duration or during certain seasons  of the year when the
Bay appears  cleaner.    Finally, as mentioned earlier, the question abstracts
entirely from  the  natural  heterogeneity  of  the Bay.   Some areas  may  be
unacceptable to just about everyone  while others  may appear clean  to  the
majority.

    Some insight was provided by  the  user survey  into the  specific factors
considered important  in visiting  a Chesapeake  public  beach.    We asked
individuals to  rank each of five factors  on  a  scale of one to five,  with five
being the most  important to  them.   The  weighted  averages  and medians were:
                                                  Mean    Median
Presence of floating debris or oil
Presence of odors
Presence of jellyfish
Presence of cloudy water
Presence of seaweed and other
aquatic plants
4.32
3.44
3.41
1.97
1.85

5
4
3
2
1

These  numbers indicate that  floating debris  and  oil is the major  quality
criterion, with odors and jellyfish being the next most important.

    The question was re-analyzed by considering  the differential  responses of
users  who came  into contact with the  water (swimmers and waders)  and those
who did not  (sunbathers, etc.).   The  contact  users cited odor as the most
important  or the  second most  important criterion 56  percent of the time,
whereas non-users cited it as highly  important only 16 percent of the time.
On the other hand, the presence of jellyfish  was  considered to rank as the
first  or second  factor for the non-contact users  84  percent of the time but
only 37 percent  for individuals who were in contact  with  the water.  These
results are  somewhat difficult to  interpret because we cannot determine cause
and effect.   Logically,  those  moat bothered by jellyfish  are likely  to refrain
from entering the water.    However,  those who go  into the water are more
likely than  those who don't to  detect unpleasant odors.

    In any  event,  of the five factors deemed  important  and  perceivable  to
beach  users, three are characteristics  which  could be linked with  water
quality.  It  is interesting,  for  the purpose ofkeeping our perspective, that a
natural factor (jellyfish )  ranks among the  unpleasant features ofthe Bay.

                                      11

-------
Two  Propositions  about Water Quality and Behavior

    To investigate the relation between current  water  quality  and  the use of
the Bay, we examined two simple propositions.

    Proposition 1:   The percentage of respondents at a particular beach who
                   find the  water quality unacceptable is related to water
                   quality as  measured by  scientific  water quality readings
                   at the site.

    Proposition 2:   An individual's use of the  Bay is  related to his/her
                   assessment of whether  the water quality is acceptable.

    Affirmation of Proposition. 1 implies a positive relationship between
individual behavior and the  objective measures  of  water quality upon which
environmental policy is based.   Proposition 2, if true, indicates  that people are
consistent in matching their  behavior to  their perceptions of  water quality.
Both propositions  are   important in  making   the   connection   between
environmental improvements and behavior-based benefit  measures.

    The Bay  is a well  studied ecosystem and  has been the  focus of  much
attention by the  U.  S.  Environmental  Protection Agency and  the states of
Maryland and  Virginia  (e.g. U. S. EPA, 1983).   Also, Maryland counties on the
western shore  sample  water at the beaches on a monthly basis in compliance
with Maryland health requirements.   These  various  sources provide objective
measures of water quality at the Chesapeake beaches,  and allow examination  of
the relationship between users'  perceptions and objective  measures  at the
beaches  visited.

    As mentioned  earlier,  the user survey instrument contained a question
asking respondents  to  judge  whether specific beaches on the  western shore
were acceptable or unacceptable  for swimming or other water related activities.
To answer, the respondent was  not required to have used the beach but only
to be familiar with it.   The water  quality at  Sandy  Point State Park was
familiar to  the largest  percentage  of people  (63 percent),  whereas only one
person  knew about the water  quality at  Camp  Merrick.    The  percentage of
those  familiar with  a beach  who found  the water quality at that beach
acceptable varied from 94 percent at Rocky Point State Park  to 12 percent at
a Baltimore  Park,  a beach used primarily by  picnickers.

    As  a guide to  the  sample's responsiveness to  water  quality,  the  percentage
of people not finding  the water quality  acceptable  (PCNA) at  a  beach was
regressed on the moot probable fecal coliform count  (FCC)  for that beach.
The fecal coliform counts were collected at  the  beaches  during the swimming
season by county officials.  Unfortunately, the FCC measurement was available
for only nine of our twelve beaches.

    One might argue that individuals  would have no way of perceiving fecal
coliform.   However,  a high  FCC might manifest itself in  odors or may  be
correlated with other factors  which cause  visible  changes  in  the  water.  Of
course,  periodically high counts could cause a beach to be occasionally closed
by the health  officials, a practice  that could  "brand"  the water  quality  at
certain  beaches.    Since there were  five  examples of  beach  closures, the

                                     12

-------
  estimation of a relationship  between PCNA and  FCC should serve  as a small
  test of the ability of individual  to perceive factors  correlated with FCC.

      To assure that obvious restrictions on the  PCNA and FCC variables were
  not violated by our  functional forms a Weibull  distribution was assumed:


(2.1)                PCNA = 1  -  exp[-(FCC/«)*),

  where  »is the shape parameter and & is the  scale parameter.    Using a
  non-linear least  squares routine, we obtained parameter estimates of 2,537 and
  .49 for 6 and • respectively with  ratios of parameter values to  standard
  errors greater than  two  in both  cases  (see Table  2.1).  These results  support
  Proposition I.  There is  an apparent  connection  between  objective  measures of
  water  quality at a beach and households' perceptions that water  quality  at the
  beach  is acceptable.
                                   Table 2.1
                    Parameter  Estimates  and Standard Errors
                             for Weibull Distribution
               Coefficient
               Estimate                       .495      2,537.

               Standard   Error                 .095        923.
      The value of  the shape parameter suggests  that the percent of beach
  users who find water quality unacceptable  is concave in  the  water quality
  variable.  (A sufficient condition for concavity is •< 1, i > 0.) To find the
  fecal coliform level  for a given level of acceptance,  equation (1) is inverted
  and estimated coefficients inserted


                    FCC = 2,537  •(-ln(PCNA))a-oa.


  For  an  acceptance  rate  of  90  percent,  the estimated maximum median fecal
  coliform count  is  in the order  of 25 fecal  coliform  per 100 ml.   At fecal
  coliform counts of200,  75 percent  of  the  users are estimated  to accept the
  quality.  At  counts of  1,200, this estimated ratio drops to 50 percent.

      The second proposition  was  tested using the telephone survey response.
  Households were asked whether  anyone  in  their household had changed
  (stopped or started) swimming patterns  in the Chesapeake because ofwater
  quality. Two hundred seven of the 1,044  telephone respondents stated they
  had  stopped, and 26  stated they had started.   Of  those who stopped,  75

                                       13

-------
percent believed the water  quality was unacceptable.     In  comparison, 53
percent  of  those  who  did  not  change   thought  the water quality was
unacceptable.    Finally,  the  water quality was  believed unacceptable by 42
percent of those who started  to  swim.

    We regressed the individual binary  response as to whether or not they
stopped swimming against their acceptance of the  water quality.  The  following
logistic probability model was estimated:
                   P = {1 + exp[a0  +

where

                        0 if the household stopped using the Bay
                   P =
                       [ 1 if the household did not stop


                 WQP " water quality percept ion
                       (1 if acceptable, 0 otherwise)
A maximum likelihood  estimation approach produced  the effects  reported in
Table 2.2.   Water  quality perception appeared to have a positive statistically
significant  impact on whether the household  continued swimming, indicating
some relationship between  users'  perceptions of water quality  and  their use of
the water.   This  result provides support for  Proposition 2 that behavior is
related to perceptions.  Nonetheless,  some people who consider water quality
unacceptable are still  observed to swim.
                                  Table 2.2
                         Effect ofPerceptions on Use

Effect                             Estimate  (al)         Standard Error

Intercept                              1.54                   .1?

Water Quality Perception                .57                   .13

Sample Size = 503
    Additional  insight into the first proposition can be  gained from an analysis
of temporal  changes  in  household habits of using  the  Chesapeake.   It is  the
consensus  among scientists  that the water  quality of  the  Chesapeake fell
substantially over the period  1950-1980  (EPA,  1983). The  living  resources  of
the  Bay have  been used  as  a primary indicator of this decline.    Submerged
aquatic  vegetation  and  anadromous  fish  stocks are  among   those  living
resources whose dramatic  decrease over this period has been cited.  If  we  can
                                      14

-------
show  that contemporaneous with the decline in objective measures  of water
quality,  individuals  were  more  likely to quit using the Bay,  we  have additional
indirect evidence of the link between behavior and water quality,

    Responses  from the telephone survey  were  used to develop a  time series
on the percentage of households  who changed  their  swimming participation.
The procedure was fairly complicated since the  size of  the population eligible
to change its behavior varied  from  year  to year.  That  is, consideration had
to be made for how  long the  household had lived  in the  Chesapeake  region
and whether  they had previously  stopped swimming.   For example, households
responding that  they stopped  swimming  in 1979 clearly were not  eligible to
stop again in 1982.

    The  time  series is shown in Figure 2.2.   Approximately  one  percent  of the
eligible  households stopped swimming each year in the  early  1970's.   This
increased to around  five percent per year  in  the  early 1980's.    The  trend is
definitely one of increasing non-participation  in swimming over  the time in
which it is believed  that  declines in water quality were  occurring.   Although
the overall  pattern is a diminishing one,   there appears to be a  possible
modification  of the trend  near  the end of the period.
                                  Figure 2.2

                       Annual Net Change in Swimming Habits
                                  1970 -1984
    Net

    Change

    in

    SwiMiin

    Habits
                   1970
1975
                                                   19RO
                                      15

-------
     If one is bold, enough to assume no  reversals in habit  for a particular
 household, the individual percentages can be  combined to show the cumulative
 effect  ofwater  quality change on a given 1970 population  Of households
 (Figure 2.3).   With this assumption,  30 percent of  households  that had been  in
 the area in  1970  would  have  had a Q ember who ceased swimming  by 1984.
                                 Figure 2.3

                   Cumulative Net Change in  Swimming Habits
 for a Popu  ation  in Residence  in 1970 and Remaining in  Resdence  until
                                                           1984
TOTAL

CHANGE

IN

SWIWING

HABITS
.1
                                  II7S
                                       mi
                                       16

-------
    The decline is interesting for a number of reasons, but mostly  because  it
shows the potential  stock of individuals who could be enticed  to return to the
Bay if water quality were improved.  Benefits from water quality improvements
will derive from increased desirability of current recreational  trips,  more (of
these higher quality) trips taken by current users, and  finally trips taken by
those who are currently  non-users  but are enticed into the activity by higher
quality.    The above  analysis suggests that  water  quality  improvements could,
over a sufficiently long time, attract  a large number  of new (or  returning)
users.   To the extent that demand analysis is based on current use and fails
to predict accurately this  potentially  large number  of  new entrants,  benefit
estimates  will be understated.
Results of Some Focus Group Experiences
    The above evidence about the relationship among  (a) use of the Bay,  (b)
subjective perceptions of the Bay's water quality,  and (c)  objective measures
of specific attributes of water quality in  the Bay  suggests  in general  that
changes in water quality  have an effect on behavior.   In subsequent chapters
we show how economists  can use  information about changes  - in behavior
induced by quality changes to assess  the  economic gains  from water quality
improvement.

    Good theory,  convincing benefit measurements,  and effective environmental
policy  do  not  require that individuals act  knowingly and mechanically  in
response to changes in ambient quality.   In  fact,  casual observations  suggest
that many  people  have only vague notions about  environmental quality, and
act unconsciously in response to changes in  quality.   However, much  can  be
gained by understanding  better  the  link between perceptions  and  behavioral
changes.   How  are perceptions  formed?    Which  aspects  of water quality,
objectively measured,  matter most to people?   These and  other  questions  about
the formation of perceptions require some insight into individual motives.

    Traditional research methods have  not  been very  helpful in obtaining
these insights.    In  contrast, focus  groups  (Reynolds and Johnson; Caldor;
Desvousges  and  Smith)  have been found to be a useful means of investigating
the existence and formation of  subjective perceptions on environmental  issues
and marketing questions.  Focus groups  are group interviews conducted in the
form of informal discussion sessions under the guidance ofa neutral
moderator.   Participants  are  encouraged to  talk at will  and describe personal
experiences,  anecdotes,  and acquired knowledge.    The  moderator  merely
encourage participation  by all, mediates arguments and spurs conversation
and  thought  through questions carefully designed to give direction to the
discussion.   By  encouraging  participants  to reveal thought processes and
levels ofawareness, their  motives begin  to emerge.

    For this study we conducted  two  focus groups of a quite different nature.
The  groups were made up of 10  to 15  individuals who had some  common
association with  one another.  Each session lasted  about one and a  half hours.
In each group there were  Chesapeake Bay users  and nonusers;  however, the
groups were chosen so as to be  heavily weighted  towards people familiar with
the Bay.


                                     17

-------
    One focus group consisted  of  students at the University of Maryland in
College Park, who were members of  a  wildlife conservation group.    Many of
these students had taken environmentally related courses.  A number of these
students  had grown up in close proximity to the Chesapeake Bay,  as active
users of the Bay.

    The second group consisted of  residents of neighborhoods  along  the
western shore of the Chesapeake  in  the vicinity of Plum Point,  Maryland.
Their participation was  solicited through an announcement in  a  local
newspaper.   Many,  although  not all,  of  the individuals  were retired, and all
lived near  the shore year  round.    Their backgrounds and education were
quite varied.

    These groups were polar in several respects.    The college  group  was
young, formally  educated in many  scientific  and environmental  matters,  and
tended to be active users of the Bay.   The Plum Point group was older,  often
retired,  not necessarily  fluent in  scientific and technical matters, and often
somewhat passive  in   their  use  of the  Bay.    The groups shared  the
characteristics ofhaving no small children and not being actively engaged in
building careers.

    In each case,  the moderator presented a formal introduction indicating the
general  purpose  of the gathering and   the underlying research.    The
introduction was notably vague so as  not to bias subsequent responses.  For
the  remainder of the session the moderator rained  questions but  did  not
attempt to  confine individuals' responses.   All  individual were asked to
respond  to moat  of the  questions  so as to avoid dominance by one or  two
people.

    Examples of  the types of questions raised were the following:

         What does water quality mean to you?
         How do you know when the water quality is poor?
        What activities do  you pursue on  the Chesapeake?
         Has  the water quality gotten  worse over  time?
         What do  you think is the moat serious cause  ofpollution in the Bay?
         Is  water quality  different in  different parts  of  the Bay?

    Initially, we had several questions in mind  which we hoped  the  focus
groups could  help us answer. These included  the  following:

    1.   What sources of information do people  use  in  forming their perceptions
        of water  quality in  the Chesapeake Bay?

    2.   What factors affect their  interpretation of  this information  (e.g. past
        experience,  attitudes), i.e.  what  is their standard based upon?

    3.   In what way does  the water quality of the Chesapeake affect people;
        i.e.,  in what  sense  do they lose when water quality deteriorates?

    4.   Do their perceptions  affect their  behavior  and how quickly  can
        behavior be expected to change  in  response to environment changes?

                                     18

-------
    Our focus group results suggest  the  following answers to these questions.


Question 1.
    Appearance  was by far the most  frequently mentioned signal of  water
quality deterioration.   Whether or not  the individual  had been  exposed to
scientific information on  the  subject, he was most  likely to report that how
the water looked, felt and smelled were his  most  important indicators.    Even
individuals who no longer used  the Bay or certain  sections  of  the Bay  based
their decisions on the water's appearance  during their last visit.

    Without exception,  individuals used  clarity  of  the water as an indicator.
The degree of transparent y of" the water was taken  for granted to be a
measure of  quality.  A few individuals,  particularly those  living  on  the Bay all
year, noted seasonal differences in clarity, but  still used this  as  their first
quality indicator.   Other factors  which  signalled  poor water  quality were the
nature of the  bottom,  discoloration of the shoreline,  froth on the water?
floating debris (including man-made)  and dead fish.  Smell was a clear signal
of poor water quality, but odors  were not  common,  and visual appearance was
used as a more discriminating indicator.

    The second most common  signal of pollution  was "guilt  by  association."
Individuals frequently stated  that  they deduced that water quality would be
poor in sections  close to  activities which they  reasoned would generate
pollution.    Such  activities included sewage treatment plants, housing and
industrial developments, marinas  and other heavy concentrations of boats, and
farms  (particularly  with  livestock) .    These  deductions took place in both
groups but were of a slightly different nature.    The college-age group was
relatively more concerned with agricultural operation and with contamination
from boat sewage.   The older  group seemed to consider development  -- with
or without sewage treatment --of greatest  concern.   Some of this  difference
can be accounted for by the spatial location and  familiarity of the  two groups.
The Plum Point group knew  local conditions  well  but were relatively immobile
and had limited experience  with the  rest of the  Bay.   In contrast, the college
students  were heavy boat users and  therefore extremely  mobile.  They  tended
to  have personal  experience along large portions  of  the  Bay and  its
tributaries.   Graphic examples ofmanure pond  overflows,  run-off from pig
farms,  etc., were offered.   Residents of  the Plum Point area  would  have  little
exposure to agricultural  runoffs more common to  the upper Bay.

    Television,  radio and newspapers were  the next most common external
source of information.     Rarely  waa specific   information  about local Bay
conditions gleaned  from  the media.   Instead,  these sources create a general
awareness of  environmental problems.   In large part  the  inferences  about
activities which  create pollution  were baaed  on information  gathered from
these  secondary sources.

    It was clear that at least some individuals  were privy  to more objective
and scientific  information than that  available in  the public media,  although the
distinction between  types  of information sources was not always made  clear.
Many of the college  students had  taken  courses  and subscribed to scientific
journals. They  were able to draw more sophisticated  deductions  about  links
                                      19

-------
 between water quality and  surrounding land uses.    Also, these  individuals
 tended to  be distrustful of information obtained from  the media.

     The final signal of pollution  also  depended on  deduction.    Individuals
 noticed changes in the amount and diversity of wildlife in and around the Bay
 and concluded that  these  changes were the result of water quality  changes.
 Specifically mentioned were crabs, turtles,  and ducks.    One  individual
associated a decline in the  diversity of finfish species with water quality
 deterioration,   Another individual argued that  increases  in fish  prices  were
 due to pollution.


 Question 2.
     Different  individuals seemed to  interpret  similar information in different
 ways.    Factors  which affected their  interpretation included their  past
 experience and general attitudes.  When individuals react to the appearance of
 something,  they must,  by  definition, be comparing  it to a standard.    More
 often  than not, the standard  used by  these groups  was past experience,
 although occasionally individuals seemed to be operating  with a  less  personal
 standard  such as pictures of  clean water  in mountain  lakes.    Frequently
 individuals compared the appearance  of the  water to  what  they (accurately or
 not) remembered experiencing as  a child.   With  the exception ofone individual
 (of the older group),  everyone remembered the Bay water being cleaner  when
 they were  children.   This  was true  of the 18-  and 19-year-olds as  well as the
 50-  and  60-year-olds.     When  questioned,  individuals admitted that  both
 maturity and publicity had raised their level ofconsciousness about water
 quality but still insisted  that water quality  was poorer now than when  they
 were children. Also, the college studenta noticed some improvement  over the
 last  fewyears  --in terms  of fewer  dead fish and  birds,  less  heavy  oil
 present,  and the cleanup of dumps along  the tributaries - although  they
 thought the water  was  dirtier now than ten  years  ago.

     Individuals'  interpretationa  of information  were also clear] y  affected by
 their  general attitudes -- level  of trust  in  political and  entrepreneurial forces
 and confidence in technology.   Among the  college students,  some indicated
 distrust for political processes   and commercial enterprises to  the  extent  that
 they believed everything was   polluted, whether or not  they could see it,
 These individuals stated that they would need  hard scientific evidence to be
 convinced that improvements had been made.   At the  other extreme, notably in
 the older group,  a few  individuals indicated a trust  in  the scientific
 community, regulator y processes and technology, fueling that the  populace
 would be protected from unsafe conditions through cleanup activities.   Some
 indicated   resignation  to the  trade-off  between    the   environment   and
 development.  In  all cases attitudes affected  how individuals  interpreted the
 same sensory and media  information.


 Question 3.

     Individuals perceived themselves to be affected  by  water quality  in a
 number of ways.   It was clear from the discussions  that both groups  were
 apprehensive about going into water they perceived to be dirty.   A distinction

                                      20

-------
was made  between wading and  swimming, as many  indicated that they  did not
dare  submerge their heads in water  they thought to be polluted.  Also many
wore  shoes when  wading to  avoid contact with the bottom.   While only a few
actually mentioned  bacteria and potential illness, most  seemed to  have health
and safety factors in mind.

    It is   difficult  to  separate   this  apprehension  from the  general
unpleasantness   associated   wit h  unattractive  water.     No  doubt  we  are
conditioned  to link clarity with cleanliness  and cleanliness with  health.
Nonetheless,  it  seems  that  even if the individuals  were convinced there were
no  health  risks, they  would consider  themselves hurt by water quality
deterioration.   Many mentioned the unpleasant fee/of the bottom and  the
sticky film that  swimmers feel on their skin  after bathing.   Others mentioned
that clear water  allowed them activities  such as seeing living organisms  in the
water, activities which were  precluded by murky  water.    Still others who
never entered the water but  only walked along  it reported the  experience
more pleasurable when  the water looked  cleaner.   In fact,  some  gave up the
activity when the water  looked dirty.

    What  is  interesting however is  that no one mentioned toxics or  heavy
metals. Some were aware of the  term nutrients,  and  most connected this with
turbidity, algal blooms and slimy bottoms.    Others emphasized  oil  spills.
Particularly in  the older group, individuals  expected that pollution could be
seen.   Those individuals seemed most  conscious  of health risks, yet  indicated
they  felt safe  going  swimming on days when  the water looked clear.    Few
indicated apprehension about health effects from unseen pollutants.

    Lasses also accrued to  individuals through perceived reductions in angler
and duck  hunting success.   Many complained of a  decline  in the quality of
fishing,  crabbing  and duck hunting.   Others  complained of the  reduced
variety of finfish available  in  the Bay.  Among the college students were some
who professed a  concern for the wildlife in situ.    That is,  some  individuals
indicated  reduced enjoyment of non-consumptive wildlife  uses.

    Individuals also indicated a fear of eating fish and shellfish caught in
polluted waters.  For many individuals low catch  rates were irrelevant because
they did not  dare eat fish caught in local waters.

    Interestingly, one  individual who did not use the Bay for  any recreational
activity  indicated  that he really did  not  care what happened to the water
quality in the Chesapeake.   His  only  concern  was the quality ofhis drinking
water.  Here  is a real world example  of the concept  of "weak complementarity."
Weak complementarily is said to characterize  an  individual's preferences if he
does not  care about  the quality  of a  resource that he does not use.


Question 4,

    Earlier  in this chapter,  survey  results  were shown to support  the
empirical relation  between  behavior and perceptions.   However,  frequently
inconsistent behavior  was  observed--some individuals perceived  the  water
quality to be poor but  continued to use it.   The  focus groups shed some light
on these  anomalies.

                                      21

-------
    With only a few exceptions,  the individuals in both focus groups were
familiar with .  the Bay and were  recreational users of one  sort  or  another.
Also,  with  only a few exceptions,  the individuals were  deeply concerned about
the water quality of the Bay.  In  most circumstances individuals had not given
up all forms of recreational use,  but they had  reduced  that use and curtailed
some activities altogether,

    We can model  an  individual's behavior as changing  in three  ways in
response to perceived changes in water quality:   he can  alter the choice of
whether  to participate in  an activity,  he  can  alter  the sites at which he
recreates, and he can alter the frequency ofrecreation.

    The two  groups revealed different types ofreactions to water quality
deterioration.   The younger,  more mobile group stopped going  to certain
places which they perceived to be worse.    The older,  less mobile, group
stopped participating in certain activities which  they perceived to be sensitive
to water quality,  particularly swimming.   They also curtailed  their  fish and
crab catch  to avoid eating contaminated fish.   In both  groups,  there  appeared
to be  a  frequency dimension  to  individuals'  reactions as well.   Many  who
found the water quality too poor for swimming generally indicated they would
go in  on especially hot days or on days when  the water looked especially
clear.   The  latter suggests  that  the  degree of intra-seasonal variation in
pollution  and other  causes   of turbidity  will  affect the  frequency of
participation in a recreational activity.   Of  course only those who live near
the shore  can assess  the water  clarity before incurring  the coats  of the
recreational  trip.

    While there  is no  firm evidence for this,  many  individuals seemed to
participate more  in recreational activities than they believed wise.   In many
cases,  it was  because they had been participating  in these activities for years
and  resisted giving  them up  and  because  they perceived no suitable
alternative.   In contrast,  some  individuals had curtailed certain activities
because of bad experiences  and indicated that it would take very convincing
evidence to bring them back.   All of this suggests  that  the response of
behavior  to perceptions may  be  significant  but may  also be  a  delayed
response.


Summary of Focus Group Experience

    In summary,  we can construct a set  of hypotheses about perception
formation.  The list would include

    10   Individual associate the quality of the water with its appearance --
        specifically its clarity  and color.

    2.   Individuals  associate   the   quality with  the  amount of   floating
        (man-made)  debris  and dead organisms.

    3.   Individuals associate quality with angler  success rates.

    4.   Individuala deduce quality from surrounding land and water uses.

                                     22

-------
    5.   Individuals  infer things about the quality  of water from  general
        publicity about  the  environment  and/or  technology change.

    6.   Individuals learn specifics  about quality from scientific publications,
        educational experiences.

The more exposure individuals had to information of  the sort included in (5)
and (6),  the more likely were they  to deduce things  about water quality from
surrounding activities.    Nonetheless,  items  (1)  and  (2)  in the above list
dominated, irrespective of age, education,  etc.

    Individuals  perceived  themselves  to  be harmed  by poor  water quality
through a number of routes:

    1.   The individual's recreational experience  is degraded  by unpleasant
        appearances floating debris, etc.

    2.   The individual fears health and safety risks.

    3.   The individual believes  poor water quality reduces catch  rates and
        variety of species.
    4.   The individual fears eating fish  from  areas with poor water quality.

    It is worthy of note  that both unpleasant appearance and poor  fishing
conditions harm individual but also serve M signals of poor water quality.
These  signals carry with them  suspicions of further losses in the  form of
health risks from contact with the water  or from eating contaminated  fish.


Conclusions

    In this chapter  we set  out  to  explore the relationship between human
activities  and  the water quality of the Bay.   This relationship is important for
the Chesapeake Bay Program for several  reasons.  First, human use  of  the Bay
is the ultimate goal  of devoting  resources to improving the quality  ofthe Bay.
Gaining some sense that  people change their use of the  Bay with  changes in
water  quality suggests  that Bay clean-up strategies can have significant
value.   Second,  economists' benefit  measures of improvements in water quality
are based primarily  on  changes in behavior.    Knowing that households have
some sense ofwater quality and  are affected by  this sense ofwater  quality
when deciding how to allocate their scarce time  and resources gives support
to the methodology of benefit measurement.

    We have explored the  relationships between perceptions  and human
activities in two ways.   From two surveys,  a  phone survey of  households and
an on-site survey of  beach users, the relationship between objective measures
of quality and perceptions of quality and  behavior  has been examined.   The
telephone survey supports the relationships in  several ways. Households that
perceive water quality as unacceptable are more likely to quit using the Bay.
The telephone survey also shows an implicit but positive correlation between
the likelihood of quitting  and  the  Bay's  "water quality.   The  user survey also
provided support for the perceptions   link.    This  survey shows positive


                                     23

-------
correlation between measures of fecal coliform at each of nine beaches and the
proportion of households  that  found  each beach unacceptable.

    The  focus groups  provide  insight into how people  judge the  quality of
water and  why they change  their behavior in response to quality changes. A
variety of sensible motives contribute to behavior  changes.   People smell, feel
and see  the water and  its surroundings.   They  react when they learn  about
changes in water quality  from  newspapers  television and other media.

    Of particular importance   to policy makers  is  the clear  signal that
individuals  suffer from  water quality deterioration in more  than one way.
Many regulation are  implicitly based on health  standards,  yet  health  effects
are only part  of the story.    Irrespective of health risks, individuals were
uniformly adament in arguing that recreation  in water  perceived to be dirty is
less enjoyable.   This dimension is not totally independent  ofhealth concerns,
however,  since dirty water was additionally considered  to be a signal  for
health risks. The word "risk"  is a key one.   Whether or not a  given state of
water quality does in fact present a health risk, the individual  suffers from
the uncertainty associated with not being able to  assess  the risk himself.
While we do  not go  into  this  problem in great depth  in  this study,  it is
important to keep a few things  in mind.    Uncertainty is ceteris  paribus
undesirable,  and there  are two sources  of uncertainty involved here.   One is
the uncertainty associated with not knowing what is  in the water and  whether
it is  potentially harmful.   The  second is  the uncertainty associated with  the
actual onset  of adverse health consequences  if indeed the water was
potentially harmful.

    The losses described  above  pertain  to  water use that involves contact.
There are  still more ways  in  which individuals  perceive themselves  to be
harmed   by poor water quality.   The enjoyment  associated with  any activity
within sight  of  the Bay is claimed  to  be diminished  if  the water appears dirty.
Finally, to  the extent that poor water  quality reduces  fish  abundance and
species variability y,  sportfishermen  see  themselves harmed.   Finally,  even if
fish catches aren't  reduced, perceived poor water quality suggests health
risks  associated with eating fish catch.

    Together  the -two sources of  information   provide   support  for the
inferences which we draw  in  the following chapters.   Individuals are aware of
water quality,  change  their behavior in response  to water quality changes,
and derive benefits when  the quality  of the Bay  is  improved.
                                     24

-------
                                  Chapter 3

         Recreational Use of  the Bay and Willingness to Pay Estimates
                    for Improvement to the Bay se a Whole


    A variety  of methods have been used to analyze  the welfare effects  of
water quality  improvements.    In the  introduction to  this  chapter,  a brief
description of  the  two basic approaches  is offered to prepare the reader for
the methods used in  this and  following chapters.  A more thorough examination
of one of  the  methods? contingent market valuation? is  offered  in  Cummings)
Brookshire and  Schulze.   Bockstael,  Hanemann  and Strand  supply a thorough
examination of  the other method, indirect market valuation,

    The indirect market approach uses individual behavior in related markets
to infer values ofnon-marksted  goods.    For the case  in question,  water
quality,  the researcher observes the demand for goods  that are related to
water quality,  such as recreational  tripe.    The usefulness of  the  approach
depends on  the responsiveness of behavior toward water  quality.  If
individuals value good water qualit y,  the y  will be drawn to goods or activities
associated with  high  quality water end will  be willing to travel farther and
incur greater  costs for  this improved experience.    Their behavior can be
observed, and  from  this, values deduced.   One drawback  to  this approach is
that assumptions regarding behavior must  be  made  in  order  to assess values.
This results in  untestable restrictions on  behavior implicitly or explicitly
imposed in the modelling process.

    Contingent  market  analysis involves the establishment,  in  the  interviewee's
mind, of a fictitious or hypothetical market circumstance.  The interviewee is
asked  to respond to the circumstance   in a hypothetical  manner.    By
establishing a  scenario to explain the respondent's answers,  the  researcher is
able  to  deduce  characteristics of  the respondents  preferences.

    The "average" willingness to  pay or sell is the predominant value  obtained
in most  contingent valuation  exercises.  A  question or series  of questions  is
designed to elicit  the respondent3 (hypothetical)  bid for or against the policy
in question.   The approach can be directed very  specifically  to the good or
quality change  to be valued,  and  thus, in theory, elicit the amount of money
needed to keep  the  individual at the same  level of satisfaction before and
after an event.  The questions can be  quite  specific and may therefore define
precisely  the event  or policy to be assessed.    The disadvantage of the
approach  is  its  hypothetical nature.   Rarely is it possible to test the validity
of the response,  through observations on behavior.   in addition, the specific
valuation  problem may be so remote  from the respondent's market valuation
experiences  as to leave him unable  to  respond reliably.

    Contingent valuation has been deemed  a useful  technique  (see Cummings,
Brookshire and  Schulze) provided it is applied to problems which are closely
associated  with common market valuation experiences.  Car-son    and   Mitchell
present  evidence  of  stable contingent  valuation  estimates for  national  benefits
of clean freshwater  in the U.  S.
                                     25

-------
    It seems  reasonable to  attempt  some contingent valuation  for the
Chesapeake Bay problem as long as caution is  exercised in the interpretation
of  the results.   For  us,  derivation of the contingent values thus obtained is
not intended as  an end unto  itself,  but rather information  to  support the
results of additional  analyses.


Recreational  Use  of the Bay

    During the summer of 1984, a telephone  survey ofover 1,000 households
in the Washington, D.  C.  and Baltimore  Statistical  Metropolitan  Sample Areas
(SMSA's)  was  conducted.    A  description  of procedures can be found in
Appendix A.   Appendix B  is a  copy of the survey  instrument.   One objective
of the survey was  to provide a complete inventory of  beach use by residents
in the Baltimore/Washington SMSA's  ( see Figure  2.1 ),  which  include the  District
of Columbia, several counties and incorporated cities in Northern Virginia and
much of central  and southern Maryland.   Restricting the geographical area in
this way biases the  sample of  individuals toward urban  residents.   However,
this area includes a  large percentage of the population surrounding the Bay.

    In the subsequent  discussion, the percentage  figures  reflect the sample
response rates corrected by  sampling weights  to  define unbiased estimates.
The projected total number of  households  purported below to  participate in
various activities are estimated as the product  of these  weighted  response
rates   and   the  approximately  two  million households   residing  in   the
Baltimore/Washington SMSA (the  1980 Census reported  1,876,144  households).

    On the  basis  of  the telephone  survey,   43 percent of the region's
households are estimated to have used  or intended to  use the Bay  for some
recreational  activity in 1984.  Participation rates varied across  the region  (see
Table 3.1) with Anne Arundel  County having the highest  percentage use (69%)
and the District  of Columbia  the lowest (21%).    Of the remaining areas,
Northern Virginia had the next  lowest participation  rate (37%)  and Montgomery
County the next highest rate of  participation  (48%).

    The households used the  Bay for a  variety of recreational  activities.
Swimming/beach use was the most popular, with  a projected 740,000 households
participating. The next  most popular activity was boating which attracted  a
projected 620,000  households.   Sightseeing  (estimated  586,000  households) and
fishing (estimated 477)000 households)  were also very popular. The projected
number of  households  who used the Bay in  conjunction with  hunting totalled
only about  45,000.   There were  an estimated 170,000 households that reported
other uses of the  Bay.

    As one might  expect, households  often  participate  in  more than one
activity.   For the major use activities of swimming, fishing and boating,  Table
3.2 shows the percentage of" respondents who  participated  in one activity or
more.   Roughly speaking,  about  one-third of the households  participated  in all
three activities,  one-third participated in two  ofthe three  activities, and the
remaining one-third participated in  a  single activity.   This  distinction has
importance for benefit estimation;  if  any one household's  participation  were
limited to only  one activity,  independent behavioral studies of each activity

                                     26

-------
                                                        Table 3.1
                                     Participation Rate" in  Chesapeake Bay Activities
                                               By Activity and Area, 1984
Northern District of Montgomery
Virginia" Colurbia county
\ Participation in
CB Activity (1984) 22 9 37
\ Participate or'
Intend to Participate 37 21 48
\ Participate CB
Fishing 12 8 16
\ Participate CB
Swiming 10 R 24
\ Participate CB
Boating 17 8 26
\ Participate CB
Hunting 1 0 1
I
Prince George's
end Charles Anne Arundel
Counties County Baltimore6 Others^
34 60 36 36
46 69 45 42
18 33 19 21
21 33 23 23
24 4B 24 33
0 523
aWeighted percentage,  representing a  random sanple of  Baltimore-Washington,  B.C.  SMSA's
^Includes Fairfax, Arlington,  Prince William and Loudon counties and the  cities of Alexandria,  Fairfax and Falls Church.
"Includes Baltimore  City  and portions of Howard  and Baltimore counties.
dlncludea Carroll  and Harford  counties and  portions of  Howard and  Baltimore counties.

-------
GO
                                                                 Table 3.2
                                          Joint Participation  Rates3 in Chesapeake Bay Activities
                                                       By Activity and Region,  1984

Overall
Northern Virginia
District of Cblunbia
Montgomery County
Prince George's and
Charles counties
Anne Arundel County
Baltimore City
and County
Other Maryland

Fishing
3
3

1
8
3
3
3
One Activity
I Swimingk I
14
15
10
22
17
3
13

Two Activities
Boating
10
11

12
7
10
9
15
Fishing
Swimming
9
6
15
8
7
7
12
4
Fishing
Boating
9
9
5
9
11
10
6
20
Swimming
Boating Fishing
20
15
15
22
19
34
20
18
Three Activities
Swiraning Boating
34
38
54
26
31
33
37
40
        tWeighted  percentages,  representing  a random ample of  the Baltimore-Washington,  B.C. SMSA's.
        '-Swiraning  includes beach use.

-------
could  be aggregated  to provide  the basis of a total benefit  estimation for
improved water quality.   Multiple participation  and the interdependence  among
activities prevents straightforward addition of  benefits calculated in demand
studies of individual activities.

    While it may be necessary eventually to undertake a comprehensive benefit
analysis of all  Bay activities? there is enough current information to shed some
light on the  value of  the recreational use of the  Bay.   Independent  studies
are useful, if  for no other reason than to establish  "conditional" relationships
between activities and  key  factors.    This may facilitate future  studies  by
isolating key  factors for which information is critical.  Moreover, by  analyzing
a series of partial systems, bounds  may be established on the total  potential
benefits.


Aggregate Willingness  to Pay

    This portion of the chapter employs the contingent valuation technique  to
value improvements in water quality  in  the Chesapeake Bay.   The hypothetical
circumstance  posed to survey resondents involves the alteration of  the Bay's
water quality from its  current  condition  to an improved condition  which,  in
the respondent's  view,  is  acceptable  for swimming.   .  Because individuals'
perceptions of water  quality are not easily linked to objective measures  (see
Chapter 2)  and because  individuals do not  easily  understand  these  scientific
measures,  the hypothetical circumstance was framed in terms  of the
respondent's acceptability.   This limits the  specific  application of the results,
since there is no  simple way to determine at what point clean-up efforts  raise
the water quality to an acceptable level for everyone.  However, the evidence
presented in  Chapter 2 offers some guidance as well as some  historical
perspective.

    The households responding to the contingent  valuation experiment are  a
subset of the telephone survey of the  Baltimore-Washington  SMSA's.   Each  of
the randomly  selected households was asked:

        "Do you consider the water  quality in  the Chesapeake to  be
        acceptable or  unacceptable  for swimming  and/or other water
        activities?"

Of the 959 respondents, over one-half  (57  percent)  found the water quality
unacceptable.   '1'hose who responded  that it was  unacceptable  were asked:

        "Would you be  willing to pay  ($A)  in  extra state or  federal
        taxes per year if the water quality  were improved so  that you
        found it acceptable  to  swim  in  the  Chesapeake?"

The  amount  of money  ($A)  was  varied randomly  from $5 to $50  over the
sample.  The  percentage of respondents  who answered "yes" is  shown in Table
3.3.
                                     29

-------
                                  Table 3.3

               Percent of People Willing to Pay Additional Taxes
         for Acceptable Water Quality  for  Swimming,  by Amount of Tax
Amount of Tax Increase  $5    $10  $15    $20   $25   $30   $35   $40    $45  $50

Percent of
Respondents Willing
To   Pay  Tax   Increase   64   66   63   70   58   46    57   47   47   53
    If sample sizes were big  enough, monotonically decreasing percentages
over the entire range would likely be revealed.   Nonetheless,  the percentages
are in general declining as  the  amount of the tax  increases.    Of those who
were presented ataxof $5,  $10,  $15,  or $20, an average  of 66 percent agreed
 (hypothetically)  to  accept the  tax burden in exchange for acceptable water
quality.    Of  those presented a  tax of $25,  $30,  $35, or $40,  the  average
percentage dropped to 52 percent.


Analysis of Willingness to  Pay Responses

    Hanemann  (1984)  describes  a  method for analyzing a central tendency  in
willingness to pay  from questions  with "yes" or  "no"  answers.    Let the
respondent derive utility from  the nonmarket  good,  water quality,  and from
money income  (y) which can be used to purchase marketed goods.  Also let a
vector  (x) of individual characteristics affect his utility.   Utility is  given  by
u i(l ,y;x)  when the water quality is acceptable and U.  (0,y;x)  when it is not.
The functions u, ,  and u.  are not known, and thus  are considered stochastic
to  the  researcher; That is

 (3.1)               Uj(j,  y;x)  = v(j,y; x) + i/j           j  = 0,1


where vj       independently and  identically distributed  random variables with
mean  zero.

    When  offered swimmable  water at a tax of  $A,  the individual will accept
the tax providing that
 (3.2)               v(l, y-A; x)  +  vl»v(0, y;x) + i/0


and decline otherwise.  In this framework,  the individual's  response becomes a


                                     30

-------
random variable with probability density


         Po = Pr{accept tax in exchange  for  swimmable water}


            = Pr{v(l, y-A; x)  - v(0, y;x)  •"<>  - v^


         pt = Pr{not accept tax  in  exchange  for swimmable water} = 1  - po.


Define    *) = v0-vl  and let ?•»)(•) be the cumulative distribution  function of 17.
Then  the probability ofaccepting the tax  in exchange  for swimmable  water
equals  FTJ(AV) where 4v is the difference between the deterministic portions  of
the utility function in the  two states (see  equation (3.2)}.

    At  this  point   any  of   a  number   of  utility functions, individual
characteristics,  and density functions   to  complete  the  analysis  could be
chosen.    Like Hanemann, we chose a logistic  cumulation distribution function.
Also, we chose  a linear  function  (see Sellar,  Chavas and  Stoll  for the
limitations of this form) for v(0),  such that


 (3.3)          v(j, y)  = aj + 9 y,             f > 0


where the arguments  of x have  been  temporarily suppressed.  The difference,
Av, is (BJ - aO  - PA,  which gives a probability model of the form


 (3.4)     F(q)  =   n F(-aO + al -  /YAi) ~  [1 -  F(-ao  + a, -  #A~)]
                its,                  itso


where So  is the  set of individuals  refusing to  pay  the tax, and S^ its the  set
accepting the tax.

    Conceptually, we  sought  the value A for which


 (3.5)          u(l, y-A; x) = u(0,  y;x).


Combining  equations  (3.2)  and  (3.3)  produces the result that when  (3.5)  holds,
A is  defined as the following


              A = [(a, . a0)(-7,) ]//»•
                                      31

-------
Since  ~n  is random,  so is A.   To evaluate A we  chose to take its expectation,
assuming «0i « i » and 0 to be constants, which yields


(3.6)          E[A]  = (at  -  «0)/0.


Thus,  («i  -   «o)/0  is  the  expected  (or average) tax  that would make an
individual  just  indifferent  to paying the tax in exchange for acceptable  water
quality and not paying the  tax  but forgoing  good water quality. Now we need
estimates of the parameters a i, «0» and ft to get  a value for  E[A],


Results ofAnalysis. BY Subgroups  of Respondents

    In developing the theory  it was admitted  that  individual  characteristics
(designated by the vector  x)  were likely to affect the utility function which  in
turn would affect  the parameters in (3.6).   Some of these characteristics are
strictly  idiosyncratic  and not worth  trying  to  model, but others  may be
associated with identifiable subgroups of the population.    Three  means of
subdividing the population  suggest  themselves--by  household income, by race
and by Bay user/nonuser.   In the sample obtained in  1984 there was sufficient
correlation between race and income  to make the  separate  treatment of these
infeasible. Additionally,  it was  difficult to subdivide the population by income
because income appears  in  the  data  set  as  a  continuous variable and
arbitrarily dividing it  into  ranges did not prove  useful.

    After some  preliminary logit analysis,  a modification  of the model shown in
(3.4)  was estimated.   One  modification entailed  making the  (*1 -  «o) depend on
whether someone  in  the  household had used or  intended  to  use  the
Chesapeake Bay  in  1984.   A variable (D  j ) was included to  reflect use.   This
approach allows us  to test whether  users  value the  change in the Bay's water
quality more than non-users,  ceteria paribus.   The  other modification involved
making the bid coefficient,  f, depend on  the racial classification.   Because
there is a wide disparity in income between whites  and non-whites  (average  of
$40,000  annually  vs.  average  of $25,000),  the marginal  utility of  income,  which
P represents,  maY be different  for the two groups.  Use  ofa binary variable
(D2> in conjunction with the tax  variable permitted an examination of the
effect  ofrace on the marginal utility of income.

    The results of the estimation are reported in Table 3.4.  The amount  of
the tax significantly  reduced  the probability that a respondent would agree  to
pay  the annual tax increase.    Also  significant  were  the use/intercept
interaction variable and  the tax/race  interaction variable.   Both users and
whitea were more likely to accept the tax increase.
                                      32

-------
                                  Table 3.4
       Logistic Model Estimates Related to  the Probability a Respondent
                          Will Accept a Tax Increase
Variable8
Constant (
-------
                                  Table  3.5

               Estimates of Utility Parameters by Income Group
Income Class
$0 - $20,000
$20,000- $50,000
$50,000 - $80,000
Over $80,000
Income not
reported
«i
Users
1.282
(1.89)"
1.652
(2.96)
1.695
(.98)
1.157
(2.81)
.533
(1.53)

Non-users
.833
(1.72)
.968
(2.04)
1.471
(2.80)
.543
(1.24)
.200
(.42)
ft
.028
(1.73)
.012
(.81)
.017
(.95)
.013
(.90)
.016
(1.11)
Sample
Size
99
200
101
22
93
Likelihood
Rat io
19.35
11.05
42.33 .
5'?.79
9.68
W-statistic in parenthesis
                                  Table 3.6

                     Expected Value of Willingness to Pay
 for Acceptable Water  Quality for Swimming by Income  Group and User Group,
                                     1984.
Income Class
o - $20,000
$20,000-$50,000
$50,000-$60,000
over $80,000
not reported
Expected
Average for All
$ 38.54
108.60
95.16
66. '44
22.26
Value of Willingness to Pay
Average for Users Average
$ 415.94
134.25
101.20
89.00
32.64
for Non-users
$29.85
78.48
88.08
41.77
12.25
                                      34

-------
                                  Table 3.7

           Estimated Willingness to Pay for Acceptable Water Quality
             by Participation and Racial Composition of Household
                                    1984.
                                    Expected  Willingness to  Pay

Participation Status                    Racial Compos  it  ion

                              White                     Non-Whit e
User
Non-User
$183.63
(55.12)a
$48.13
(10.25)
$34.16
(10.40)
$ 8.95
(2.53)
a Standard  deviation  in  parentheses
the range are less  reliable.   The results suggest  that a wider range of  tax
increases  would have yielded more confidence in the  estimates' accuracy.
Individual willingness  to pay  bids  for water quality improvements appear to
have a larger range  (i.e.  take on  larger values) than we  anticipated when
constructing  the survey.

   In Table  3.8 the  average  willingness  to pay for  each subpopulation  is
combined with estimates  of the subpopulation size to project  total  willingness
to pay figures.   The values  are based on the telephone sample estimate that
57 percent of the  population find Chesapeake Bay water quality unacceptable
and on  the sample percentages of white users  (27%),  white non-users  (35%),
non-white users  (16%)  and non-white non-users  (21%).

    Expected values as well  as  optimistic and pessimistic values are shown.
The optimistic (pessimistic)  value  is derived using the expected value of
willingness to  pay plus (minus) one standard  deviation.   On the basis  of  these
estimates, we could argue a  reasonable  range  of willingness to pay values of
$60 million to  slightly over  $100 million.    Care must be exercised when
considering the standard  deviation, as  it is computed as an  approximation  and
is not associated with the normal  distribution.   The values shown,  however,
represent  an  "order  ofmagnitude"  contingent valuation  estimate of willingness
to pay for improved water  quality.
                                      35

-------
                                  Table 3.8

     Estimated Aggregate8Willingness to Pay for Water Quality Acceptables
                 for  Swimming, by Classification and Scenarios
                                     1984
Wingness  to Pay
"Average*'
                                              Scenario
"Optimistic"0
"Pessimistic"^
                                                Thousand
User
    White

    Non-white
     55,838

      6,164
   72,595

    8,020
   39,081

    4,271
Non-user

    White

    Non-whit e


Aggregate
     83,612
  106,976
   60,275
'Baltimore-Washington SMSA population
bfiased on  expected willingness  to  pay
°Based on  expected willingness  to  pay  plus one standard deviation
^Based on expected willingness  to pay minus one standard deviation
Regional  Comparisons

    Stretching the data  somewhat  further, one  can also examine geographical
patterns of responses.   The logistic model  wae re-estimated using  sub-samples
grouped by region:  the  Southeast region (Prince George's  County,  Charles
County,  Anne Arundel  County and the District of Columbia) ,  the Western region
 (Northern Virginia,  Montgomery County) and  the Northern region  (Baltimore
                                      36

-------
City and County,  Howard County, and Harford County) .    The sub-samples
represent groups,  each of which exhibits reasonable  internal homogeneity,  for
which we  have at  least  one-hundred and  fifty responses.    Even with these
conditions, however,  the  statistical results are less significant than the  earlier
ones because ofthe smaller sample  size.

    The results suggest  regional  similarities and differences  OlcfcLe 3.9!.  Some
consistency  is  evident  as signs on all coefficients  are  the same for all regions.
Thus  an increase   in the hypothetical  tax decreases the  probability  of
acceptance of the tax  associated with water quality  improvement.   Additionally
the effect of participation and race on willingness to pay for  the improvement
is consistent  across  all  regions.
                                   Table 3.9

        Logistic Model Estimates  Related to the Probability a Respondent
     Will Accept a  Tax Increase to  Improve  Chesapeake Bay Water Quality,
                              by Geographic Area
Variable
Constant
DI " constant
Amount of Tax
D2 . Tax
Southeast •
.334
(.94)d
.78
(2.36)
-.050
(3.33)
.041
(3.15)
West b
.71
(.46)
1.02
(2.49)
-.070
(3.04)
.060
(3.00)
North0
.12
(.30)
1.67
(4.77)
-.023
(1.77)
.015
(1.36)
Chi-squared for
Likelihood ratio
36.5
37.2
48.6
aDist. of Columbia and  Counties  of Prince George's, Charles and  Anne Arundel
^Nort hem Virginia and  Montgomery County
^Baltimore City and Count  iee  of Baitimore,  Harford  and Howard
^t-ratio in parentheses
    There are, however,  systematic  differences  across regions. Users  from the
Northern region  are willing to pay on  average  substantially more  than  those
from the  southeast or western  regions.   The figures for nonusers are  less
disparate across regions, with  those  for  the  West  region  somewhat  larger.  The
estimated willingness  to  pay figures are presented in Table  3.10.
                                      37

-------
                                 Table  3.10

     Estimated  Willingness to Pay for Acceptable Water Quality by Region,
           Participation, and Racial  Composition of  Household, 1984.

Household Characteristic
White, User
Non-White, User
Non-White, Non-user
White, Non-user

Southeast
$124
22
7
3'?
Region
West
$133
25
10
55

North
$224
77
5
15
Existence Value
    In the preceding contingent  valuation experiment we present  non-zero
willingness  to pay estimates  for non-users as well  as users.   There  are a
number of  reasons why non-users  may be willing  to pay for improved  water
quality.    One  of these  reasons  has  been  labelled  existence  value  by
non-market  benefit  analysts  (Krutilla)  and stems  from early  experiences
applying  benefit cost analysis  to  water  resources  projects.   Individuals who
never use a resource either directly or  indirectly and  never intend to uae it
may still be willing to  pay to improve ita quality or assure its existence.
Formal studies  of  existence value are limited, but some  empirical  evidence
exists.    Fisher and Raucher  (1984)  suggest  that nonuse benefita (including
both option value and existence value) are some fraction of the use value of
water quality changes.  Other research (e.g.,  Walsh  et al., 1985; Schulze et al.,
1983 ) suggests that  existence value  may  be greater than use value,  and
sometimes substantially so.

    Existence value is  a frequently cited  concept in the literature, and several
studies  have attempted to  derive explicit  estimates of  existence  value
associated  with  water quality (Mitchell and Carson,  1981; Cronin, 1982; Walsh et
al., 1978; Desvousges et al., 1983).   Nonetheless,  no consensus  exists on the
models which underlie  the measurement.    Behaviorally based methods  of
welfare measurement  are unsatisfactory because,  by  definition, existence value
is unconnected with behavior.     Suspicion surrounds  contingent valuation
estimates  of existence value because  these estimates  are  even less susceptible
to proof or  disproof than contingent valuation estimates of use values.   Even
moretothe point, the  success of a contingent valuation approach depends on
well defined questions.    Without  a clear  idea of  the "motivations behind
existence value, properly  focused questions are difficult to  define.
                                     38

-------
The Existence Value  Experiment

    In this  section we present some preliminary results of an  experiment
designed to  shed  some light on the motives behind existence value for the
Chesapeake Bay.    The  sample frame  was  derived from the phone survey
described  above.  The households contacted randomly by phone were  asked if
they would complete an additional mail  survey.    Of the 1,044 contacted,  741
agreed to  fill  out and return a brief  mail  questionnaire  regarding  water
quality in the Chesapeake  Bay,  and of these  741 households,  282  actually
returned the questionnaires.    Because only  about 70 percent  of   those
contacted agreed  to  receive the mail questionnaire, and  only 38  percent of
those who  agreed actually returned these questionnaires, these results should
not be taken as representative of the population  sampled but as  useful  for
gaining preliminary y insigh ts into willingness to pay motives.

    The 282  respondents were grouped as users or non-users.    Users were
defined as all  respondents who currently use the  Bay  or  thought  they  might
do so in the future.  Respondents who felt certain that  they would not use
the Bay for recreation at any time in the future were  defined  as  non-users.
Non-users  accounted for 16.3  percent  of the  respondents.

    Respondents were  asked to consider a series ofsituations concerning
public beaches surrounding the Chesapeake Bay.   They were  asked to assume
that water quality  at  these beaches  had fallen below a level acceptable  for
swimming,    They were told  that  a project could  be undertaken  that  would
clean the  beaches so that a water quality level acceptable  for  swimming was
achieved and maintained.   The respondents were  then asked the following
question under four  scenarios:

    "Would  you  prefer that  the  clean-up  project be undertaken?"

    Scenario 1,    No additional information.

    Scenario 2.    Access to the beaches  by  the  public  is  permanently denied
                  so that even if clean,  the  beaches will not be used.

    Scenario 3.     If the project is undertaken,  taxes would be raised so much
                  that  nearly   everyone prefers   that the  project is not
                 undertaken. These taxes would be paid by individuals  other
                  than the respondent.

    Scenario 4.     If  the project  is  not  undertaken,  funds would  instead  be
                  used  to improve hospital  services  in selected communities
                  surrounding the Bay.   Of all the people who care, half want
                  the beaches  cleaned,  and half want improved hospital
                  services,  The respondent himself would never  need  to  visit
                  any of the  improved  hospitals.

The proportion of  "yes"  responses for users and non-users under each
scenario is given in  Table 3.11.
                                     39

-------
                                  Table 3.11
             Summary  Results  of Contingent Valuation Experiment
                              on Existence  Value
Scenario
Number
1
2
3
4
Proportion of
Yes Responses:
Users *
.96
.70
.71
.49
Standard
Error of
Difference

.032
.032
.035
Proportion of
Yes Responses:
Non-users^
.83
.69 .
.67
.37
Standard
Error of
Difference

.088
.088
.091
aThe number  of users in  the  sample of respondents  is 236.
blhe number  ofnonusers  in the sample ofrespondents  is 46.
GThis is the standard error  of the difference between  the proportion in
 Scenario 1 and  the proportion in the given scenario.
Interpretation of the Results
    In order to  interpret the responses  reported in Table 3.11,  it  is  necessary
first to consider the  'potential motives for existence value.  Two broad motives
may be discerned:  intrinsic and altruistic.  Existence value based  on  intrinsic
motives stems from a concern about the state of the world. Concern  about the
order of things may cause people to suffer simply by learning about pollution
incidents.   What has been called the "environmental ethic"  is closely  linked
with the  intrinsic motive.

    Of concern  here is the second  ofthe  two  motives:  altruism.   People can
gain value  from the enhanced wellbeing ofothers  (individualistic altruism).
An  extensive discussion of  these altruistic motivations can be found in
Madariaga and McConnell  (1987) .

     Responses  to the question under Scenario  1 are used as a control to be
compared with  responses under Scenarios 2  through 4, where  Scenario  1  is
purposely  ambiguous about  'project costs.   As expected, moat  respondents
preferred  that the project  be undertaken  under  Scenario  1.    Interpreting
non-user responses of  "yes" as evidence  ofexistence value, the relatively
high  number of non-users giving positive  responses  is consistent with the
results of  previous  studies that have found evidence of  existence value.
                                     40

-------
    With  access  to beaches denied under Scenario 2,  the number of "yes"
responses  to the  question predictably declined.   Since  the  number  of non-user
responses  of  l*yes" declined when access  was  denied,  it appears that existence
value, at least  to some  individuals,  is related to others' use.   Thus, altruism
may , be one motive  that underlies existence  value.   However,  even  with access
denied, most respondenta preferred  that the project be  undertaken.  This may
reflect the  presence of a number of  motivational  including  an environmental
ethic.   Finally,  it  is  interesting  to note the closeness of user and non-user
group responses  under  Scenario  2.   Since with access denied there can be no
users,   "yes"   responses  from  the  user  group will  also indicate positive
existence  value.   In this  scenario,  the proportion  ofusers and non-users
exhibiting existence  value was nearly  identical.

    Scenario 3 is similar  to Scenario  2 in  that  both  attempt to  eliminate
altruistic  motives.   In  this  scenario, the Bay can  be used  after the cleanup,
but other individuals will  be forced to  pay  more than what the improved water
quality is worth  to  them.   The similarity ofproportions  in Scenarios  2 and 3
supports  the notion that the Chesapeake resource is valued for its  own sake.
In Scenario 2,  about 70 percent  of the people support  the  project despite the
fact  that there  is no use and hence no direct use value.    In  Scenario 3,
roughly the  same  proportion  supports the project  even though there  is no net
value to the  users.

    Under Scenario  4  the  number of  "yes"  responses  fell dramatically
compared  with  the responses under  Scenario 1.   Since  less  than half of the
non-users preferred that the cleanup project be undertaken,  it appears that
the improved hospital  services  are  on average at least as valuable as clean
water in the Bay.

    The individuals  were instructed that they would not need the  services of
the hospital, themselves,   so it is  tempting  to label their value for the
improved hospital services as existence  value.   However, the entire value of
the hospital  services may be due  to altruistic  motives while individuals appear
to have motives beyond  altruism  for Chesapeake water quality  improvements.


Conclusions
    The underlying motives for existence value matter  to  the proper  design
and interpretation  of  contingent valuation experiments.     The  preliminary
results concerning  existence value  associated with  the  Chesapeake  Bay suggest
some ambiguity about its motivation.   People are  willing  to pay  for water
quality improvements  in the Bay,  but how much  they are  willing  to  pay
depends on  the specific nature of the  opportunities  foregone by doing so.
Among other  considerations,  these  suggest  attention  should be paid  to the
methods for financing the cleanup ofthe Bay.
                                     41

-------
                                  Chapter 4

           The Effect  of  Chesapeake  Bay Water Quality on Beach Use


    The  previous chapter contains  a range   of benefits  from  improved
Chesapeake Bay water quality, based on a contingent valuation  experiment.
Although there is  substantial  evidence to suggest that the responses to the
hypothetical questions were not random but rather associated with households'
use of the  Bay and  racial/income strata, problems  still  exist with the  approach.
Follow-up questioning  revealed households did not consider  alternative uses of
tax increases (e.g.  improvements in other  public goods  such  as hospitals,
roads, etc.) .   Moreover,  the subjective nature of the water quality  measure-
ment used in the contingent valuation question does not lend itself  easily  to
policy analysis,   based as it is on  objective  (scientific)  measures  of water
quality.   Finally,  the values represent aggregate values,  indistinguishable on
the basis of type ofrecreation  or geographic location  of pollution. Knowledge
of user  group and geographical impacts of  programs  can  provide a depth and
richness  of understanding important in the political process.

    The  remaining  chapters are devoted to  providing analyses of the  observed
behavior of  households based on data gathered in previous studies which are
specific  to different recreational activities.   The  analyses  use cross-sectional
information on  households to  model  beach use, boating and recreational
fishing. Once demand  functions are estimated, benefits  from  access  and from
changes in  water quality are assessed for each of  these  activities.

    This chapter contains  a cross-sectional analysis ofbeach use on the
western  shore of Maryland.  It draws  from the random telephone survey of
the Baltimore-Washington SMSA's and  a stratified random survey of twelve
public beaches on  Maryland's western  shore.    As such,  the analysis is not
comprehensive of  all beach use  in Maryland  but  rather the use  of the public
areas in one portion  of the Bay by the citizens  in  the surrounding  environs
ofthe two large metropolitan areas closest  to the  Bay.

    A number of  approaches to  estimating recreational  response to  environ-
mental quality changes have evolved.   Many of  these depend first  on the
estimation  of demand for recreational activities which are closely linked to the
environmental resource in question.  The recreational demand models currently
in use have grown out  ofthe union of neoclassical  demand theory  and the
travel cost  model proposed by  Retelling and employed extensively by recrea-
tional economists  for  the paat several decades.   The principal  contribution of
the travel cost model  is found in the  use of the travel cost to the recreational
sita as the principal  component  in constructing a "price"  for the  recreational
commodity.  The simple travel cost model can be  derived  from  a neoclassical
utility maximization framework,  as can more complex models which incorporate
added dimensions  to  the problem (see Bockstael,  Strand and Hanemann, 1986).

    One  particularly  important  modification  of  the simple  model  is  the
introduction  of quality characteristics of recreational  sites (see Volume I of
                                     42

-------
this report  for the theory).   If the recreational  demand model is to be used to
estimate  the  value ofenvironmental  quality improvements then individuals'
behavioral  responses to changes  in  quality must be  modelled.   This requires
observing behavior in the context of differing levels  of environmental  quality,
which can generally be done  only  by  examining  recreation  behavior at a given
point in time  over  a group of sites which vary in  quality.

    The procedure  can lead  to  a number of specific methods of  analysis (see
Kling,   Bockstael  and  Strand,   1985) ,  each imposing  a different set  of
restrictions/assumptions  on recreational  behavior.   While there is no  concensus
regarding the  "correct"  model,  the two most prominent,  models in the literature
can be categorized  as the  (modified) neoclassical  model and the discrete choice
model.   The neoclassical model  has the form of  the traditional  demand system,
with quantities being a  function  of  prices and water  quality.   The  model  is
modified in some way to  facilitate the  inclusion of water quality parameters
which do not tend to vary for a given site over the population,  Additionally,
demands  are generally treated independently.     A  common  approach is the
varying parameter model   (VPM), as  put  forth by Smitht Desvousges and
McGivney (1986) .    Here,  independent  single-site models of  recreational trip
demand are estimated,  and the estimates  of  the  intercept and price coefficient
are correlated with the site's water quality.   Then,  in policy analysis, changes
in water quality change the intercept and slope of the demand curve,  thereby
influencing quantity  consumed and the welfare  derived  from  recreational
activities.

    The discrete choice model  (DCM)  has  also  taken many forma  (e.  g.,  Caulkins;
Morey  and Rowe)  but the form employed by Bockatael,  Hanemann  and Strand  is
representative. In this model,  the individual is viewed as having  a number  of
choice occasions upon which to select a site.   The selection of site is discrete
in  the sense that only one   site  is chosen per choice  occasion.    Site
characteristics such as  travel  cost,  water quality and facilities are used to
explain  the choice of a  site  on any given  occasion.   A composite "value"
reflecting the desirability  of  available choices  is computed from the discrete
choice estimation,  and this is  used with other  factors to  estimate the number
of choice occasions.

    Although  both  models are  behaviorally based, there  are advantages  and
disadvantages  associated with both.  The varying  parameter  model starts from
the  assumption  that  the demand  functions  for  trips to  sites are  interior
solutions to  a utility  maximization process.    The  discrete  choice  model,
however, starts from the viewpoint that, on any given occasion, an individual
chooses among  a finite set of alternative sites.    Neither approach is perfectly
satisfactory.    In  the DC model,  the  link  between the number  of choice
occasions and  the site selection per  choice occasion  is  ad hoc.   With  the  VP
model,  the  demand  for any one  site does not adequately reflect  the  alterna-
tives.   Additionally,  the fact that individuals do  not visit all  sites  is  incon-
sistent with the  implicit theory and must be handled econometrically.   Kling
has  employed  Monte Carlo studies to  examine  the performance of these models.
Not  surprisingly,  her  results suggest that the VP  model  excels when most
recreationalists  tend to  visit almost all  alternative sites  in a  season, and  the
DC model excels when most tend  to  visit  one or  only a  few sites  in a season.


                                      43

-------
Most recreational data sets will  be characterized by something between the two
extremes, however neither  model  has an obvious advantage,  and  no tractable
model is perfectly consistent with  this situation.

    In this chapter both the varying parameter and discrete choice models are
applied to western shore beach use in Maryland.  In subsequent  chapters  only
the varying  parameters model will be  applied since neither the  boating nor
fishing  data can support  the data intensive discrete  choice model. The  data
on beach use at western shore beaches is relatively rich, however.    Both
types of models will utilize this same data set of Chesapeake Bay beach users
in the  subsequent analysis.   The results will be a  range of  values which
suggest  orders of magnitude  for  welfare  measures ofhypothetical changes in
water quality.


The Survey and the Data

    This section is devoted to a description  of the survey of Chesapeake
Beach Use  conducted in 1984.   Unlike the data used  in  analyses ofboating
and fishing in Chapters 5 and 6, the data  used  in  this  chapter were collected
during  an  earlier budget period of this  cooperative  agreement.   Great  care
was  taken  with  the  sampling frame to improve confidence  in  the results.
Because the  survey itself is important to the  project,  the content and
procedures are described  extensively  in Appendix  C.   A copy of the survey
instrument can be found in Appendix D.

    From May  26,  1984  to August  19, 1984, Research Triangle Institute  (RTI)
interviewed  individuals  on the western shore  beaches in Maryland.  The study
population  consisted of all residents  of the  Baltimore and Washington, D. C.,
SMSA'S,  age 14 or older,  that used these beaches for recreation  in 1984.  More
specifically,  the population was limited to  recreational  users  of the  following 12
beaches:
                                              Strata
      Beach                         Geographic              Size

1.   Sandy Point                       north                 large
2.   Point Lookout                     south                 large
3.   Fort Small                         north                 small
4.   Miami                             north                 small
5.   Rocky Point                       north                 small
6.   Elm' s Beach                       south                 small
7.   Bay Ridge                         south                 large
8.   Kurtz                             north                 small
9.   Breezy Point                      south                 small
10.  Rod & Reel                         south                 small
11.  Morgantown                         south                 small
12.  North Beach                       south                 small
                                     44

-------
    Four  hundred  and eight  individuals  were interviewed  at the  beach to
learn of their recreational patterns  and  perception  of  water quality at  these
beaches.   These individuals were  randomly selected  from  sample beaches and
days.   The sampling design can be described as  a  two-stage  stratified  sample
in which a probability y sample of beaches and  days was  selected  , and a  random
systematic sample  of  persons  was  interviewed at  each sample site (day-beach
combination)  .

    The User  Intercept  Survey  Questionnaire  was  designed to record and
collect the following:

         Frequency of visits made  to  beaches  on the  western  shore of the
         Chesapeake

         Activities that the respondent  (and his/her family) participated in
         when visiting beaches

         Activities not participated in and the reason why they  were not

         Cast related to a typical trip to  each  beach that the respondent had
         visited  since January 1,  1984

         The respondent's perception of  the quality ofthe beach and the
         beach  facilities at each beach  with which  he/she was familiar

         Factors  that  influenced a  respondent's decision  to  visit or not visit  a
         beach

         The respondent's willingness to  continue to  visit the sample site if
         costs  related  to the use of the  beach  were to rise.

In addition,  a  series  of demographic  questions was included  to enable analysts
to establish profiles of beach users.


The Data

Household Trips

    Respondents  were  asked, on-site,  how many trips  they had taken in 1984
prior to the interview and how many they intended to take during the rest of
1984.    Follow-up  telephone interviews  at the end of the seaeon  obtained
complete  1984  trip information for  251  ofthe  408 beach  users  interviewed.
For the remaining  households, information was obtained solely  on-site.

      To assure consistency in our  trip  measurement,  the end of the season
information was  compared with  in-season response so  that a  correction factor
could be applied to households  with  only  "in-season" information." Using data
from the largest  beach (Sand y Point), a regression of end-of-season  trips (xe  )
on reported plus  intended trips during the season  (xr) yielded:
                                      45

-------
(4.1)                X   =  -686  +  -632xr + e      Ra •  -89    (n =  148)
                     6   (1..40)  (35.00)


where the t-statistics are in parentheses.   Equation  (4.1) was  used  to predict
total trips to a site from on-site information  for  households  that  did not
receive  follow-up  inteviews.   The combination  of  a fairly small constant term
and a coefficient on Xr which is less than one suggests that households tend
to report intentions in excess  of trips later realized.


Access Costs and Time Costs
    Previous  studies  (e.g.   Bockstael,  Hanemann,  and  Strand,  1986)   have
considered  travel  costs to  include  the  person's (or household's) monetary
costs of travel as  well as the  opportunity costs of their time.   Distance to a
site,  transformed into  transportation  costs is a feature  common  to all visitors.
However,  those individuals who  forego income in order to take  time to recreate
incur  monetary expenses  in  excess of  transportation costs.     For  these
individualist these costs  can be  measured as the  foregone wage rate times the
time spent accessing the activity.

      For households without employed persona  or with  employed  persons with
fixed work  schedules,   there  is no direct  loss of income  incurred when
recreation  is  undertaken.   The opportunity cost of recreation time for these
individuals is the  value of foregone  alternative  activities.   Unfortunately,
opportunity costs will  vary over  individuals  in  ways which  are  not observable.
The only observable  factor related to  the total opportunity cost  of the
recreation experience will be the time  spent traveling and recreating.   Even
this measurement is troublesome, however, since  the  on-site  portion of this
time also measures the amount  of the recreational good consumed.   To avoid
many of  these complications  we employ only round-trip  travel  time as a
surrogate for  opportunity costs  in these cases.

      In addition to these access costs,  most western shore beaches  have an
admittance  fee which  must be  added to the  other costs of  traveling  to the
site.  Often the fee will  vary  depending on the day of the week and size of
party.


Water Quality

    The  Chessie System environmental  quality data, maintained  by  EPA's
Chesapeake Bay  program, were'  used  to  construct the water quality measures.
Turbidity,  bacteria  counts,  total  suspended  solids,  total  nitrogen, total
phosphorous,  and total chlorophyll A are among  the  potential indicators of
water  quality  to  which beach  demand might  be sensitive,    Since  these
exhibited a  high degree of collinearity, two variables were  extracted from the
data set  to use in this  analysis:  total nitrogen and total phosphorus.  A good
case can be made for using  these variables.   Studies of the Bay conducted by
the U.  S.  Environmental  Protection Agency  indicate that perhaps the most
significant problem facing Bay  restoration and protection efforts is nutrient


                                     46

-------
over-enrichment of Bay waters.   Excessive nutrient levels may be the partial
cause  of decreased  submerged  aquatic vegetation, which in turn  has adverse
effects on the food chain and on the habitat  for many  fish species.   Further,
over-enrichment leads  to  lower dissolved  oxygen levels which  have additional
adverse effects on fish stocks,  degrading the appearance of the water  as well.

    High collinearity  y between  nitrogen  and phosphorus readings  prevented
separate inclusion of both variables in  the analysis.   The  product of  nitrogen
and phosphorus was used  to avoid this  problem  and  to  capture  the interactive
nature of these nutrients.

    In  each  case, mean monthly water  quality levels from April  through
September 19?7 were calculated for areas  of twelve counties contiguous  to the
Bay,    The summer months were  chosen because they  represent  the  peak  of
recreational  activity.   Complete data over  regions of the Bay were available for
only  some years,   1977 being  the closest to the survey year.   The  relative
water quality readings across  the  Bay are  unlikely to be considerably
different between the two years, even  if the absolute  readings are  different.
Additionally, individuals * decisions are unlikely to be  related solely  to water
quality in the current year, but will  be based on  a cumulative  learning
process which  includes past  observations as well.   Consequently, there  is  no
obvious correct choice,  and  the errors associated with using data  from any
one year are unclear.


Other Variables

    Additional  factors are known  to influence recreation activity, including the
ownership ofcertain types of household capital equipment.  Boats, recreational
vehicles and swimming pools are the  types ofcapital equipment  which may
affect  the use ofbeaches on the  western shore.  Some of these beaches  have
boat-launch facilities,  some camp sites,   while others offer good swimming
possibilities.    Years living in the area,  previous recreational history,  family
size and participation are  some  other factors which may be important.


The Varying  Parameter Model

    To formalize the model of behavior,  the individual  is assumed to  maximize a
constrained utility  function which  is a  function  of  number of trips  taken to
each of nquality-differentiated  sites,  the quality  characteristics of  each site,
and a  Hicksian  good.   Thus


(4.2)                 max  u(x,  q,z)          s.t.  px + z  = y


where x is  an n-dimensional   vector  of trips   to  the n  sites, p  is  a
corresponding vector  of costs of accessing the sites, q is a matrix of  variables
qi-i,  i  = l,...,n and   i  = l,...,m,  where  qij  is the  level of  the  jth quality
characteristic  at  the f h  site, z is  the Hie  sian  good, and  y is income.  To
simplify notation in this  section, we will  assume that there is only one quality


                                     47

-------
characteristic, and thus m  =  1,  and  q represents the vector  of  values of the
one quality characteristic across sites.

    Problem  (4.2) defines  n demand functions,  each of which may be a function
of all nprices,  n quality levels,  and  income:


 (4.3)                 xi = g. (p,  q/y)              i = 1 ,...,n.


This  model cannot  be estimated with cross-section data.    Imagine  having
observations  on S individuals  who  visit site i.   Own  price  (pj)  and substitute
prices ( Pi, k * i) will typically vary across individuals if they come  to site i
from different  geographical areas.   However,  there will be no variation in the
quality characteristic  at site  i  (q ) )  across the S individuals,  nor will  there by
any variation across individual in the characteristics of other sites  (q k, k *
i).  With no variation  in the q /a across observations, their coefficients cannot
be estimated,  and nothing  can be learned about behavioral response to quality
changes.

    There are  several methods for resolving  the dilemma  presented above.
Some  of  them build on  the model presented in (4.3) (these  are  described in
Kling,  Bockstael  and Strand,  1985), while others rely  on  discrete  choice models
 (see  Bockstael,  Hanemann  and Strand,   1986) .   The method used in  this  section,
the varying parameters  model,  falls  in the former category and follows similar
methods  applied  by  Vaughan  and   Russell  (1982) ;  Smith, Desvousges  and
McGivney (1983);  and  Smith and Desvousges  (1985).

    One way  to motivate the varying  parameters model  is to consider a simple
linear form such as:


                            n
 (4.4)            x, = 00, +  £    f
                            k=l
but  to further assume  that the  parameters  in site  demand  functions  are
deterministic functions  of  the quality  characteristics.   For  example, the 0*s
might be linear  functions of the q's:
                        70  +
 (4.5)              0itk~aok  +  «ik
-------
The model in  (4.4)  and (4.5)   implies that variations  in  demand parameters
across sites (i.e., variations in the ft> i ls,  ftt ( ls,  etc. )  correspond to variations
in  own- site  attributes  (qj  )   and  subatitute-site attributes    (q^,  j  *  i) .
Specifically,  the  above model  implies a demand  for trips to site  i  of the
following form:
(4.6)       xf  ^(70  +
                                       K — 1


                           +  (6


    Even though the model can be  collapsed into  one expression as in (4.6),
the estimation procedure usually  involves two steps:   the regression  of  trips
to  each site on  prices and  income  (e.g.,  n separate regressions)  and the
regression  of the coefficients from the" first  n- regressions on  the  quality
characteristics of the sites.    The  second  step  requires the application of
generalized least  squares because of the  properties  of the error structure
implicit in the estimation of  (4.5) which must use estimated parameters  (f 's) in
place of the true  ft.

    The first -stage  estimation procedure is  further  complicated by the need to
correct  for a censored sample bias.   Moat consumer  demand problems analyzed
with household data encounter this  problem.  A  random sample of households
will reveal a.  certain  (often substantial)  number of  households  that  do not
consume the good in question and thus  have  zero as the value of their
dependent variable.   In the sample,  there will therefore be many  observations
concentrated around  zero.    Neither omitting  the  zero observations,  nor
including them  in  an OLS  regression, will produce unbiased estimates.

    Tobin analyzed  this problem in 1958 and produced the  first of  several
approaches to handling the problem.   His approach  applied to the  first stage
of our varying  parameters model  characterizes  the problem  in  the following
way:


(4.7)         x, = 0o

              x( = 0                         otherwise.


While e may be  distributed as  a normal, x will not be.   The  estimation of the
0's  therefore requires maximum likelihood  techniques, where the likelihood
function is given  by
 (4.8)               L  =
                                     49

-------
 where  zf is the right-hand  side of  (4.4),  
-------
                                                         Table  4.1



                               Average  Values  of  Regression  Variables for Visitors by Beach
vn
Beach
Tripe
Beach (#/yr)
Sandy Point
Fort Smallwood
Chesapeake Beach
Rod & Reel
Bay Ridge
Point Lookout
Rocky Point
Porter's New Beach
Miami
Morgantown
8.06
5.83
2.87
6.42
7.33
3.85
10.20
2.92
5.16
7.71
Access
coats
(S)
15.98
8.66
18.79
22.80
15.75
36.42
9.07
9.50
11.81
26.09
Access
Time
(hr)
1.29
1.11
1.45
1*58
1.14
2.73
.93
.72
.86
1.22
Substitute Beach
Access
coats
(S)
12.09
6.38
11.30
13.61
10.84
13.10
6.43
5.81
4.82
9.80
Access
Time
(hr)
.84
.69
.84
.87
.84
1.03
.55
.49
.59
1.02
Boat
(%)
15
14
17
13
12
17
13
19
19
17
Ownership
Recreation
Vehicle
(%)
18
20
28
33
28
30
12
11
17
35
Swimming
Pool
(%)
18
14
17
25
12
12
20
11
17
09

-------
                                                          Table 4.2
                                      Tobit Estimates for  Beach Demand  Model, by Beach"
Beach
Beach
Sandy Paint

Fort Smallwood

Rod & Reel

Rocky Point

Chesapeake Beach
VI
to
Porter's New
Beach

Point Lookout.

Miami

Day Ridge

Constant
8.17
(2.83)b
.16
(.05)
10.44
(--2. 18)
10.29
(2.04)
-3.96
(-1.89)


-.70
( .31)
3.49
(-2.72)
-2.20
(-1.45)
-6.96
( 1.16)
Access
Costs
-.35
(-4.07)
-.53
(-2.86)
MO
(-.84)
-.47
( 1.45)
.18
(-2. 19)


-.29
(-2.21)
-.05
(-5.62)
.09
(-1.35)
-.78
(-4.90)
Access
Time
-4.85
(-3.61)
-4.24
(-2.58)
-1.51
(-1.28)
-5.63
(-2.38)
-1.19
(-1.76)


-1.28
(-1.28)
-1.72
( 4.72)
-1.27
(-1.18)
-9.63
(-3.50)
Substitute Beach
Access Access
costs Time
.24 2.47
(2.86) (1.15)
.34
(1.14)
.29
(1.25)


.19
(1.80)


.31
(1.10)
.12 4.55
(3.35) (5.41)


.83 7.40
(3.19) (1.96)
Ownership
Rec.
Boat Veh.








3.23
(2.58)


1.54
(1.32)
2.19 2.98
(1.69) (2.50)
4.37
(2.46)
-6.19 7.55
( -1.00) (1.50)
Swim.
Pool






3.55
(1.36)




2.04
( 1.31)
-1.76
(-1.21)


-5.67
(-1.13)
a
14.85
(57.59)
9.52
(11.61)
9.72
(5.47)
12.41
(19.00)
6.16
(10.00)


3.43
(5. 15)
5.96
(15.14)
7.42
(10. OG)
18.06
(17.56)
Nonlimit/
Limit
Observations
243/139

41/198

22/201

87/66

46/272



25/118

82/262

50/121

61/292

   coefficients were  significantly  different from zero for  Morgantown site.
Dt-ratiOS  in  parentheses

-------
     The estimated coefficients  on own-price  (travel cost) were all ofthe
 expected sign)  and most  were statistically significant from zero.    Beaches  for
 which  a reasonably  large on-site  sample was obtained yielded  the most
 significant estimates.   Small sample effects  of  multicollinearity among the price
 and time variables likely caused  the  large  standard  errors for Miami,  Rod  and
 Reel, etc.   In some instances,  the multicollinearity was sufficiently  troublesome
 that only the own-price and  own-time variables were  considered.    Obtaining
results for as many  beaches as possible was  critical because  the  sample size
 in  the second-stage estimation  equals the  number of beaches  in the first
 stage.

     The results  of the  second-stage estimation,  i.e.  the estimation  of equation
  (4.5),  were obtained from a weighted  least squares procedure in which  the
 weights were l/up,  .,  the  inverse   of  the  standard error  of the  own-price
 coefficient for  each beach.  The  estimated equations  are:
  (4.10)              ftij  =  -.0308-  .00020 TNP.
                          (-.04)    (-2.22)   ]

                        =  -2.66-  .0016  TNPj
                         (-1.10)  (-.001)


 where TNP  is  the water quality variable  defined earlier,  and the values  in
 parentheses  are t-ratios.

     The results  show no significant relationship between water quality  and the
 intercepts  of the beach-use demand equations  but a significant relationship
 between water quality and the coefficients on  travel  cost.    The  poorer the
 water quality  (i.e.  the higher  the level of  TNP), the larger  the negative
 response ofbeach users to travel costs.   This results in a pivoting inward  of
 the demand  curve as water quality deteriorates  and a  pivoting outward with
 improvements.    The results are in accordance with the proposition that poor
 water quality  lowers beach  users willingness to pay for  access  to beaches.


 Estimated Benefit  Changes

     The analysis above  describes the behavioral response  of  the average
 western shore  beach  user to  the change  in water quality.     From this
 information and information on the  number  of users of the  beaches,  we are
 able to determine  some estimates of benefits  of  hypothetical  improvements  in
 water quality to  the average  user  of each beach.   We are also able to expand
 to the total population  of beach users.

     Three hypothetical changes in  the environmental variables  are  considered,
 a 10 percent  and  a 20 percent decrease  (environmental improvement)  in the
 environmental (pollution)  variables,  and a 20  percent increase (environmental
 degradation) .  Since we will want to  assess  the effects of the change, we will
 want to calculate  consumer surplus before and after the change.    The  formula


                                      53

-------
for individual i's consumer surplus from site j is given by the following when
the demand function is linear
 (4.11)              Csi;j  -


where x> • is  its demand  for trips to j and  Pj i  is the coefficient  on cost of
access  in the jfc"  site demand function.   For a given hypothetical change in
water quality at beach j,  the weighted average  change in consumer surplus
over the sample is calculated:

              N.
(4.12) ACS  = Sl(x,J(qJ))a/(-VJ1(qJ)) -(X|j(qJ))V(-20  j l (qj)) ]*k,/Nj


where k,  is the weight, qj and  qj are the  levels of water quality before and
after the change,  and  the notation xl  j (q. )  and P\ i (q^ )  implies that both
demand  and  the coefficient on  travel cost are  functions  of the level of water
quality.    Nd is the  size of the sample of  beach  users  used to  estimate the
tobit equation for beach j .    The sample  includes all 408  observations minus
those for which information about beach j was unavailable.

    Calculating consumer surplus for hypothetical environmental circumstances
 (equation 4.12)  thus requires  values for X" = x(q°), 0°(q°), x1  =  xfq1 ), and fl
= 0(ql)«  The first step in assessing the hypothetical changes  is  to use the
results  of model  (4.10)  to predict ft j  i (qj )  , that is to predict the new travel
cost coefficient given  the hypothetical  change  in water quality.     The
coefficients 9 j \ (qj) and fj t (qj) are then used to determine values  for  demand,
i.e.      xj   and    xj.

    Prediction of  the demand  for trips  is complicated because  the demand
function was initially estimated using a Tobit  procedure.    Recall the model
underlying  the Tobit,

                   x*  =  P'z  +  e             e ~  N(0,   0

and

                   x = 0                      otherwise.

Given the underlying model, the systematic  portion of (4.4) cannot be used as
the expected value IOCUIY of x,   The Tobit predicting  equation given below
adjusts  for  the censored nature of the dependent  variable:


                                     54

-------
(4.13)              B(x4j)  =  *


where z is a vector of explanatory variables and * and * are the density  and
cumulative distribution functions  for  the standard  normal,  respectively,    The
first term represents  the  conditional expectation  of trips given that  the
person  participates times the probability  that  the person participates.    The
second term corrects  for  non-normality because of potential truncation.

    There  are  two ways  of obtaining  the "before"  and "after" x's for  the
consumer surplus  functions.  One  way  is  to use the predicting  equation (4.13)
to calculate both i° and xl   values.   The second method  is  to  accept  the
observed x  as x° and then to adjust  this x by x *  -  x° to reflect  the
hypothetical  change in  water  quality to obtain  the estimated  x '.     (See
Bockstael and Strand, 1986, for details ofthe two approaches.  )

    Because there  is  no clear theoretical reason to choose  one  approach over
the other, we  calculate the results both ways.   Both methods "use formula
 (4.11) to calculate  the  change in average consumer  surplus.   However, Method
A calculates trip values as
 (4.  14)


and
                     M'ZI
(4.15)        *»j = *|-V-^ h'z«j + ^*
where the z ,  , are the explanatory variables  in  the  jth  beach's  regression  (see
Table 4.2).  Method B calculates the demand for trips in the following way:


              xf j  =  observed value of  x < j

and
where x} j and X? j  are defined in (4. 14) and (4.  15).

    Tables 4.3 -  4.5 summarize the  average beach users benefits and  losses
from  the  hypothetical  changes in the  nitrogen and phosphorus concentrations
in the Chesapeake Bay.   The first and fourth columns in each table represent
the base  line average consumer surplus over  the  entire sample of beach users


                                      55

-------
                                   Table 4.3

          Annual Benefits per  Beach User from a 20 Percent Decrease
                             in Pollutant, by Beach
                                      1984
Calculation Method Aa
Beach
Sandy Point
Fort Smallwood
Chesapeake Beach
Rod & Reel Club
Porter's New Beach
Rocky Point
Point Lookout
Bay Ridge
Miami Beach
Consumer Surplus
Before After
133.94
.82
36.32
10.32
5.95
80.38
15.86
178.18
5.38
169.03
5.17
43.88
16.19
8.45
89.53
22.61
204.76
'10.27
Benefits
35.09
4.35
7.56
5.87
2.50
9.15
6.75
26.58
4.89
Calculation Method Bb
Consumer
Before
342.04
57.69
57.89
259.81
12.20
179.65
315.27
171.64
220.68
Surplus
After Benefits
379.33
73.13
60.77
284.08
12.34
191.02
415.06
178.98
304.99
37.06
15.44
2.88
24.27
1.14
11.34
99.79
7.34
84.31
• With Method A, the average consumer surplus for a change in quality  at  beach
 j is taken over a  sample  which includes  all beech users whether or not  they
 visited beach j.

"With  Method B,  the average consumer surplus for a change in quality  at  beach
 j is  taken over a  sample which includes  only users  of beach j .

-------
                                   Table  4.4

         Annual  Benefits per Beach User from a 10 Percent Decrease
                            in Pollutant, by Beach
                                     1984
Calculation Method Aa
Beach
Sandy Point
Fort Smallwood
Chesapeake Beach
Rod & Reel Club
Porter's New Beach
Rocky Point
Point Lookout
Bay Ridge
Miami Beach
Consumer
Before
133.94
.82
36.32
10.32
5.95
80.38
15.86
178.18
5.38
Surplus
After
150.39
1.50
39.96
13.00
7.12
84.82
18.73
191.08
7.34
Benefits
16.45
.68
3.64
2.68
1.17
4.44
2.87
12.90
1.96
Calculation Method B^
Consumer
Before
342.04
57.69
57.88
259.81
12.20
179.65
315.27
171.46
220.68
Surplus
After
363.35
69.28
61.11
277.73
13.55
186.63
363.61
176.55
261.16
Benefits
21.31
11.59
3.22
17.92
1.35
6.98
48.34
5.09
40.48
awith Method A,  the  average consumer surplus for a change  in quality at beach
 j is taken over a sample which includes all beach users whether or not they
 visited beach j .
      Method B, the average consumer surplus for a change in quality at beach
 j is taken over a sample-which includes only users of beach j .
                                     57

-------
                                   Table 4.5

          Annual  Losses per Beach User  from  a  2.0  Percent Increase
                             in  Pollutant, by  Beach
                                      1984
Calculation Method Aa
Beach
Sandy Point
Fort Smallwood
Chesapeake Beach
Rod & Reel Club
Porter's New Beach
Rocky Point
Point Lookout
Bay Ridge
Miami Beach
Consumer
Before
133.94
.82
36.32
10.32
5.95
80.38
15.86
178.18
5.38
Surplus
After
106.54
.29
29.81
6.25
4.05
72.26
11.92
154.56
3.06
Losses
(27.40)
(.53)
(6.51)
(4.07)
(1.90)
(8.12)
(3.94)
(23.62)
(2.32)
Calculation Method Bb
Consumer
Before
342.04
57.69
57.88
259.81
12.20
179.65
315.27
171.64
220.68
Surplus
After
311.26
47.63
55.27
239.35
11.24
166.81
253.41
164.55
172.41
Losses
(30.78)
(10.06)
(2.62)
(20.46)
(.96)
(12.84)
(61.86)
(7.09)
(48.27)
'With Method A, the average consumer surplus for a change  in  quality  at  beach
 j is taken over a sample which  includes  all beach users whether or not they"
 visited beach j.

"with  Method  B,  the average consumer surplus for a change  in  quality  at  beach
 j is taken over a sample  which  includes  only users of beach  j.
                                      58

-------
for use  of  each beach, calculated using each of the  two methods mentioned
above.  The second  and fifth columns show the average  consumer surplus per
beach user  in the sample following a water quality change at each beach. The
third and sixth  columns represent the change  in  surplus for  the average
beach user associated with a water quality change at each beach.

    The  method of  calculation  makes a  good  deal ofdifference  for some
beaches,  especially Point Lookout,  Miami  Beach and Rod and  Reel.   Recalling
the econometric results in Table  4.2, the estimated demand equations  for these
three are price inelastic relative to other beaches; that  is, the absolute  values
of their  price coefficients are quite small.  When demand is very  inelastic, big
differences  are likely between  the mean  consumer surplus and the consumer
surplus  associated  with the mean number of tripe (see Bockstael  and Strand) .

    The  average consumer  surplus  values  are expanded  to  the entire
Baltimore-Washington SMSA's in Table 4,6.   This was  accomplished by knowing
that the 1980 number ofregional households was  1,977,000  (census of the U. S.,
1980), by determining  from  a contemporaneous phone survey that 47 percent
of the  regional population used  western shore beaches" and by  knowing the
percentage  of western shore beach users  who used each beach.    Large
aggregate benefits are  associated with  Sandy Point  (in both methods of
calculation)  because  ofthe  very large number of households  that visit that
beach.   Whereas  21  percent of  the population  used western  shore beaches,
over half used Sandy Point.  When expanding to households, Sandy Point has
nearly twice as many users as any other beach.


The Discrete/Continuous Choice Model

    The utility maximizing model  in  (4.2) and the  resulting demand functions in
(4.3) are an  apt  description ofthe individual's  decision problem only if he
chooses positive values  for all x^ (i.e., if he is at interior solutions in all the
markets).  It is not an adequate description if corner solutions arise (i.e., x i=
o).   The  discrete choice model  is appropriate  when an individual chooses one
from a finite set  of alternatives, by comparing the available alternatives. The
discrete  choice model  presented here is amended to include a component which
describes the  demand for trips as well  as  the  discrete  choice  among trips on
any choice occasion.


The Choice Among  Sites
    The first  part  of  the model  involves the estimation of the household's
choice among sites.   It will be important here  to  capture those  elements which
vary over sites.  McFadden (1976)  provides  a utility theoretic framework for
employing the multinominal logit model which is applicable to a discrete choice
problem of this sort.   For further discussion of its application to recreation
demand,  see Bockstael, Hanemann and  Strand  (1986) .
                                     59

-------
                             Table 4.6

Aggregate Benefits/Losses  to Users from Changes  in  Chesapeake Bay
                      Water Quality, by Beach
                                1984
Calculation Method A
Calculation Method B
Change Change
Improvement Degradation Improvement Degradation
Beach
Sandy Point
Fort Smallwood
Chesapeake Beach
Rod & Reel Club
Porter's New Beach
Rocky Point
Point Lookout
Bay Ridge
Miami Beach
20%

14,064
1,744
3,038
2,356
1,006
3,673
2,708
10,667
1,963
10%

6,602
275
1,462
1,075
468
1,781
1,153
5,176
788
20% 20%
. . . Thousands$
(11,001) 9,
(212) 1,
(2,612)
(1,632) 1,
(750)
(3,258)
(1,577) 12,
(9,484)
(931) 3,

967
576
680
316
52
923
484
823
975
10% ,

4,704
651
329
626
24
449
5,375
397
1,674
20%

(8,009)
(781)
(597)
(1,089)
(40)
(824)
(7,520)
(710)
(2,255)
                                60

-------
    Suppose we call \^ a latent  variable denoting  the level  of indirect  utility
associated with the ith alternative.   The observed variable Yi has the property
that
              Yi = 1    if Vf =  max(V*,VJ, .  . .,V|)

              Y  = 0    otherwise.

Indirect utility associated with the ith  alternative is  some  function of Zj_, a
vector of attributes of the ith alternative  so that V£  =  V^z^)   +  ej;.   The
random component is generally attributed  to  the systematic? but unmeasurable,
variation  in  tastes and omitted variables.  Thus, each household has a level  of
error  which,  in a   sense,  remains  with  it over  time.     If  the t's are
independently   and   identical y   distributed   with  type   I extreme   value
distribution  (Weibull),  then  it is well known  that
              Prob  (Yi=l I z) =
 (see Maddala 1983;  McFadden,  1973; Domencich  and McFadden, 1975) .    The
likelihood function for the sample is
where gi  =  1  if i is chosen,  gi = 0 otherwise.

    The multinominal logit has a property which in some circumstances  is  useful
but in others is unrealistic.    The model presented  above implicitly assumes
independence of  irrelevant  alternatives, i.e. the relative odds of choosing any
pair of alternatives remains constant  no matter  what happens  in  the  remainder
of the choice  set.    Thus,  this model allows for no specific pattern of
correlation among the errors  associated with the  alternatives;   it denies—and
in fact is violated by--any particular similarities within groups ofalternatives.

    McFadden  (1978)  has shown  that a  more general nested logit  model
specifically  incorporating varying correlations among the errors associated
with the alternatives can also be derived from a stochastic utility maximization
framework (see  also Maddala,  1983) .   If the  e's have a generalized extreme
value  distribution then  a  pattern  of correlation among  the choices  can be
allowed.   McFadden defines  a probabilistic choice model
                   P, =
                                    ev"
                                      61

-------
here G1 is the patial of G with respect to the  ith argument  and
has certain  properties which imply that
                                                                   • '
                                 =  exp{-G(e
                                         "£
is a multivariate extreme value distribution.   When G(e * ..... e N) is defined
s £evi   then  the model   reduces  to the ordinary multinomial logit (MNL)
described above.   However when
                   G(Y)
                          M
                       =  La.
                         m=l
                                ._
                                icS
where  there  are M subsets of  the N alternatives and  0 »*m < 1,  then a
general pattern  of  dependence  among the  alternatives is allowed.    The
parameters,  *m» can be interpreted as  an  index of the similarity within groups.

    Suppose we were to  classify the alternatives into these M groups where $m
denotes the  set  of alternatives  in  group m, and we were  interested in the
probability of choosing some alternative  i. Then
                   P  =
where
                               r
                               2m
                                                        if i
                                                        Otherwise
and
                                              l-«n
    The above GEV model is useful  in many applied discrete choice problems.
Frequently, alternatives group  themselves in obvious  patterns of
ability.   If they do,  it is  both convenient  and appropriate  to  estimate the GEV
model.   It is appropriate  because the results of  an ordinary  MNL will violate
                                     62

-------
the independence  of irrelevant alternatives asumption if such  a pattern
actually exists.  It is convenient because it reduces  the  number  ofalternatives
included at each stage.

    Let us make the estimation process  explicit.    In the problem at hand,
individuals are choosing  among ten beaches.    Two of these beaches  are
qualitatively different.   They  are state parks, larger and  providing more
services than the local  beaches.   Now  we can view  the choice problem as  a
two-level  nested one:  the  choice between state  park or  local beach  (m  = 1,  2)
and the choice among beaches within each group.    Consider a redefinition of
Vj
 i:
                        v.   =  »'Z.
                         im      i
where the  Z's  denote attributes  associated with all  sites  and the W's are
attributes associated with the state park and local beach  choice.   Also let us
assume that sm  is identical within  all groups  and  equal  to 6

    Now define a variable, Imi in  the following way:
                          •
Then the probabilities  above can be rewritten as
 (4. 17)
and
(4.18)                  Pm =
The variable Im  is sometimes  termed an inclusive value  (see McFadden,  1978)
and serves as an index  of  the relative value of the alternatives  included in
subgroup  m.

    As expressed  in  (4.17)  and (4.18),  the probabilities of interest  can be
estimated using   MNL proced urea.    First, the  Pi | m  are estimated with M
independent applications of the multinominal  logit.   Note  that at this stage  e is
not recoverable,  but can be estimated  only up to a scale factor of 1-4.  From

                                     63

-------
the  results  of   (4.17),   the  inclusive prices   (4.16) are   calculated and
incorporated as variables  in  the  second level of estimation (4.18).  Here  the fa
and the 6 are  estimated.

    A S outside the  unit interval is inconsistent with  the underlying utility
theoretic model and suggests misspecification  (see McFadden) .    The parameter
S is  an index of similarity  of alternatives  within  groups not present across
groups.   A value of one for 15 indicates that alternatives within a group are
perfect substitutes.  Thus,  all relevant choice involves  choice among groups.
A value  of  zero for  <*  implies there  is no special similarity of alternatives
within groups and  thus no particular gain  from using  a nested GEV model.

    Two-step estimation,  i.e.  the estimation of (4.17)  and (4.18)  independently,
is not necessarily  efficient.    Amemiya  (1973) explores this  property of the
model and presents a  correction factor.   However,  even Amemiya suggests that
the  cost in computational complexity  is  probably not worth  the gains. We
consider  McFadden's  estimation method adequate and use  it to estimate  a GEV
model in the next section.
Estimation of the  Discrete  Choice Among Beaches

    The two-tiered discrete choice model considers the  individual choosing
between two  categories of  sites  (state park  and  local beach sites) and then
choosing among  beaches within the desired category.   The state park beaches
are located at Sandy Point  (adjacent to the Chesapeake Bay Bridge)  and Point
Lookout (at the mouth  of the  Potomac),  whereas the local beaches are defined
to  include Fort  Smallwood,  Bay Ridge,  Kurtz's  Pleasure Beach,  Miami  Beach,
Morgantown Beach,  Porter's New Beach,  Rod and Reel Club Beach,  North Beach
and Chesapeake  Beach.

    In estimating  the  model,  however,  the decision among  sites within each
category is dealt  with  first.   In  assessing   the available  sites  within a
category on a choice occasion,  the household  chooses on the basis of certain
household  attributes in combination with specific site characteristics.   These
are denoted Z> ^and are defined for one model  as:

    Zn = access costs  in $ to site i,  calculated  using distance (d  ^ from  the
          household's  origin  to the site1 (Zx i=  1.088 + .049*^ - .000074d? )
          plus the entrance fee  plus  the wages  lost  from traveling  if  the
          individual had directly foregone  income to visit the site;

    Zal = access time (in minutes)  to  site  i, calculated using distance from  the
          household'a origin to  the sites(Za i= .7  + ,02d^ );

    Z3 j = water  pollution index  for site i  (see description  page 18);
1 Exact formula was determined by regressing reported  costs against distance.
2 Exact formula  was determined by  * regressing- reported  travel  time against
 distance.


                                      64

-------
    z.,  = the availability  of recreational vehicle  facilities at site i (0 if not
          available,  1 if available) times  whether the household owned  a
          recreational vehicle (0 if not owned, 1 if owned);

    Zsi = the availability of fishing   facilities at site  i  times whether the
          household owned  fishing  equipment;

    Z*, = the availability  of boat  launch facilities at  site i times whether the
          household owned  a boat.
    The first stage  of  the estimation is reported  in  Table 4.7.    The results
indicate that relatively large  monetary and  time costs negatively influence the
probability of  choosing  a beach.   Water pollution also has a negative influence
as does fishing  facilities.   Presumably,  fishing activity draws the household
members away from the  beach.   Boat  facilities and recreational  vehicle facilities
improve the  probability that  someone owning a boat or RV  will  attend beaches
with facilities  for that  equipment.

    The second  tier of  the discrete decision  involves whether individuals
select  a  state park (with many activities) or a  local  beach.    The factors
hypothesized to be important in deciding  to visit a state park were thought to
be the  years the household had visited western shore  beaches  (WI),  whether
the intercepted household had more than one family member in the party (  W2 ),
the size  of  the group  intercepted  (w, )  and the inclusive value  (y derived
from the first-stage  estimation.  People with a  larger history of beach use in
the area  would be more  likely to learn  of the smaller  beaches and  hence be
less likely to use the state parks.  On the  other hand, the  state parks usually
offer  a greater  variety  of  activities, and thus families and large parties  might
be more likely  to attend them.

    The results  of the estimations are presented in Table 4.8. The  hypotheses
about  the choice between  state parks and  local beaches were  not rejected.
Signs  of coefficients were as expected  and  coefficients  statistically significant.

    The estimated coefficient  on  the inclusive  value  term  is  .152  yielding an
estimate of  .848  for S.   This  is  significantly  different from zero suggesting
that there are  gains from using the nested  model..  There is  considerably more
similarity among beaches within the  two  categories than across the categories.
Had the  nested model  not  been   used,  the independence  ofirrelevant
alternative assumption would certainly have been violated.    The  estimate of $
is also significantly different from  one, suggesting that beaches within groups,
although similar, are not perfect substitutes,
                                      65

-------
                                   Table 4.7
                  Logit Regression for Selection Among Sites
                                         Variable
                 Access  Access    Water    Recreational Fishing     Boat
                  cost    Time   Pollution    Vehicle     Activity Facilities
Estimate          Z-^     %2i       z'(         Zl i         %i         Z6i
Coefficient
(t-statistic)
-.072
(-5.30)
-*75
(-8.63)
-.00037
(-4.00)
1.06
(1.93)
-2.09
(-5.28)
1.14
(2.10)
            =  311.2
                                   Table 4.8
      Logit Analysis for Selection Between State Parks and Local Beaches
                                            Variable
Estimate

Coefficient
(t-statistic)
Inclusive
Value
:m
.152
(9.26)
Years Attending
Western Shore Beaches
w,
-.019
(-8.34)
Family
Members
W2
.261
(4.85)
Party
Size
w,
.024
( 12. 79)
Chi-squared =  28.07
                                      66

-------
The Number of  Trips Decision

    There is no operational model that generally  treats utility maximization
with  non-negativity constraints  (Bockstael et al., 1986, Chapter 8),.   There
exists no utility  theoretic means of linking the discrete choice model  of  site
choice  on each choice  occasion to a continuous choice model  of  demand for
trips  (i.e.,  demand for choice occasions.  )

    When one decides to  use a nonclassical continuous  demand function  for the
demand  for recreational trips  (irrespective  of site), a problem  immediately
arises in determining the  appropriate choice of explanatory variables.   Since
costs  and quality  vary across   sites and since individuals are  observed
choosing more than one site in a season, which site's price and  quality should
be included?

    The approach  taken  here is to  consider the number of  trips to western
shore  beaches as a  function of a  number of explanatory variables  and an
inclusive value type variable calculated  from the  second stage  ofthe discrete
choice model.  This was  originally  suggested by Hanemann (1978) and was  used
in Bockstael  et  al.  (1986) .    Thus  when water quality  changes,  the inclusive
value changes and influences the number of trips.   In  this sense,  "the discrete
and continuous  decisions are  linked  , although not  in a utility  theoretic  way.
The discrete/continuous choice  model has  the  advantage ofemphasizing the
substitutability ofsites but does  appear to underestimate  the  response of
demand for trips  to  changes in cost and quality at  one  or more sites.

    Based on the results of the discrete  choice estimations, a new inclusive
value (In )  which  includes the factors in  the choice among sites  and the choice
between state  parks  and local beaches  is  calculated.    This  value, along  with
the individual's  income  (INC, income or full income  if at interior  in  the labor
market  ),  discretional y  time available (DT,  if at corner  in the  labor market) and
the number of  trips  to western shore beaches  in the previous year (x^ ) , is
used to estimate the 1984 total number of trips per household to western
shore beaches  (xt) .

    The higher an individual's inclusive value,  the  more attractive are his
beach  alternatives  (e.g.  good beaches are cheaper to  get  to) and the  more
trips he is likely to take.   Additionally,  beach  use  habits  (as reflected in
previous  trips)  would  likely lead to more trips.     Whether  income  and
discretionary time positively or negatively affect  the  number of trips  depends
on whether a day  trip to western shore beaches is  a normal  (positive effect)
or inferior  (negative effect)good.

    The results  ofan ordinary least  squares  regression are given in Table
4.9.  The  expected signs occur, and  the results indicate a western shore  trip
is an  inferior good,  both with respect to  income and time.    The predictive
powers  of the equation are  especially good  considering the   cross-sectional
nature of  the data.
                                      67

-------
                                  Table 4.9
      ordinary Least Squares Regression for "Choice Occasions"  or Trips
Variable
constant
Estimate
Coefficient .442
(t-statistic)
Inclusive
Value
JN
4.06
(2.92)
Lagged
Trips
xt-l
.516
(8.70)
Income
INC
-.028
(-2.79)
Discretionary
Time
DT
-.981
(-4.80)
F-value (4,253)  =  36.30
Estimated Benefit Chancres

    The ultimate purpose of the modelling effort is to estimate the benefits
associated with improvements in water  quality.   Formulas for deriving  welfare
measures in   the context  of discrete choice models of random   utility
maximization  have  been developed by  Hanemann  (1982,  1984). It is generally
the compensating and/or  equivalent  variation of the quality change which  is
taken as a useful measure of benefits.   Selecting the compensating variation
(C) ,  this measure can  be  defined by  the following expression:
v(p
                 °
y)  = v(p'
                                  y - c)
where again v is the indirect  utility function,  p  and q are  vectors of site
prices and qualities, and y is  income.

    The  compensating variation  is  now defined by


              v(p°,q°,y;O = v(p°,ql>y -C; O,

where «  is  random,  and  as a result  C is  now  a random variable.   Depending  on
how one chooses  to  take  account  of this randomness, three different measures
of compensating variation can be defined.   In  the case of  GEV  models the
median value of C  coincides with the C which equates the expected values  of
the indirect utility functions  (Hanemann,  1978) .  It  is this measure which we
calculate in the subsequent  illustration.

    In our problem,  using the  previous notation,
                                                E   e

-------
where J s  is  the set of state park sites and  J  1 is the set  of  local beach sites;
where v  = *'Z\m  +  ^'Wm; Zjm are  factors  which vary over  sites;  and Wm are
factors which  vary  between state park  and local beach sites.  Thus
                  ,        VM
              G e  , .  . . .e   =   £ t
               I           J   S=i
where
I.  =  in
                                   £ e
                                  its
                                         Is
Then the expected value of the indirect utility  function equals



              v(w°, z°, y)  =  lnG(eV?	evfi) + k,


where k is a constant.

    Now consider a change in quality  which causes w° and z° to  change to W1
and z ^ The  compensating  variation measure (C') defined above is given by
              v(W>,z<>,y)  =v(w»,z»,jrC')'
or
             in
           = In
There is no  closed-form  solution  for  compensating variation in this case,  but
Hanemann  (1982)  shows  that the compensating  variation  per choice occasion of
this change  can be approximated  as:
(14)             ACS =

                            jlj
                                      69

-------
 where the set J  includes  the two cases:   state and local beaches,  Y is  the
 element  of the • vector which serves as the price coefficient,  and vj = V w9  +
 (1-Bythe
 number of choice  occasions estimated in the continuous choice model.

     These equations were used  to estimate  the benefits from hypothetical
 water quality changes.   To be consistent with the varying parameters model
 estimates,  we considered a 20 percent reduction and a 20 percent increase in
 water  pollution.   The values  associated with the changes are $1.08 per trip
and  $4,70 per household  user of western shore beaches.    Given that 20
 percent   of the  households used  western  shore beaches  (about 401,000
 households), the total gains from a  20  percent improvement in  water  quality
 were estimated to be nearly $2 million annually.   The estimated loss for a 20
 percent degradation was approximately the same.

 Discussion

     Reiterating,  the purpose for our work was  to offer benefit estimates based
 on different methods so as  to provide a range of reasonable values.   The  two
 models  derive from two different conceptualizations  of the recreationalists1
 decisions.   The  continuous,   neoclassical model  (represented here by  the
 varying parameters model)  is   strictly  correct  only  if interior   solutions
 characterize demand for each site, with  all individuals  attending all sites,
 Another  drawback  of this model is that,  because of the econometric  functions
 estimated,  total  benefits  cannot legitimately be added across  sites.   This  sort
 of aggregation provides upwardly biased results.

     The  discrete/continuous  choice   model,  on the other  hand, begins by
 emphasizing   the corner-solution nature of the  decision on  each choice
 occasion.   Thus,  the substitutability among sites receives special attention.
 The decision about number of trips per season is not well integrated into  the
 estimation process. These models tend to provide low estimates of aggregate
 benefits because  the effect of water  quality  improvements on demand  for trips
 is not well  accounted for by the ad hoc inclusive value variables in the  trips
 equation.

     The  estimated benefit change resulting  from changes in Chesapeake  Bay
 water quality  at  the western  shore beaches is presented in Table 4.10  for  the
 two models. "  Predictably, the  varying parameter  model offers the largest
 change.

                                   Table 4.10
          Comparison of Benefits  Based on a Varying Parameter Model
                        and Discrete/Continuous  Choice

                               	Change	
 Model	20 Percent Improvement   20 Percent Degradation

 Varying Parameter              ""     ""  ' ' (in  thousands)	
   upper bound                          $26,160                - $25,839
 Discrete/Centinuous Choice              $  1,885                - $1,884
                                      70

-------
                                 Chapter 5

       Recreational Boating and the Benefits  of Improved Water Quality
    Boating is an  especially important part of any study of Chesapeake Bay
recreation.   As the  second most popular recreational activity in our telephone
sample  of  Bay users (second  only to beach use)  ,34 percent of the  area's
households  participated in  boating activities on the Chesapeake  in  1984. This
represents  nearly three-quarters  of the  houeeholdo who used  the  Bay for
recreational activities.    Its importance is  further supported by  the large
number  of registered boats in Maryland (as many as 134,000 in  1981)  and by
the fact that of the approximately 15 million Person-trips taken on boats in
Maryland waters in  1979, 90 percent  (or 13.5 million) were taken on  estuarine
waters of the  Bay or its tributaries  (Harmon and  Associates, 1983).

    In this chapter we examine the  behavior of Chesapeake boatetrs and
estimate the value to  boaters  from  improved water quality.    Since  no new
survey could be initiated  for this purpose,  the data upon  which the analysis
rests are drawn  from  a 1983 boat  owners  survey which waa made available by
the Sea Grant Program and the Department of Recreation at the University of
Maryland.


A Profile of  Boaters and Boat  Owners

The Boat Owners  Survey

    In 1983 a survey of boaters was  sponsored by the University of Maryland
Sea Grant  Program and the  Maryland  Coastal Zone Management Program.  It
consisted of a mail survey  of  2515  registered boat owners in Maryland.   The
design of the  sample provided  equal  representation to  owners of boats  kept in
slips  and owners  who trailered  their boats.   The questionnaire, which  sought  a
variety  of  information about the  household and ita boating  activities,  achieved
a response  rate of approximately 70  percent.

    The boat owners' survey  provides different but complementary information
to the  telephone  survey conducted by RTI and described  in Chapter 3,  as it
samples a different population  and uses a different sampling scheme.   The Sea
Grant survey draws  only from the population  of  registered boat owners. From
the  telephone survey, which is a random  sample  of the population, we can
identify not only those who  own boats but also those  who uae  the  Bay for
boating whether they  own a boat or  not.   The  telephone  survey provides
information  about non-boaters,  as  well.  It  does not, however,  provide  detailed
information about  boating  behavior.    Consequently,   it is  the boat owner
survey  which  will  provide  most of the data for analysis.


Boaters  and Boat Owner Characteristics

    Information from both th  boat  owner  survey and the random telephone
survey helps describe  boating in Maryland.   For  example, a  comparison  of

                                     71

-------
those households in  the  telephone survey who  reported that they  used  the
Bay  for boating with those who did not revealed no significant differences in
family size or years  lired in the area but did  suggest  considerable differences
in income  and  race  (ace  Table 5.1).    The average of boaters'  incomes  waa
significantly higher  than the  non-boaters'  average,  and  a  significantly higher
percentage  of  boaters were white.   Interestingly, this significant  difference in
income appears in the  Prince Georges/Anne Arundel/Calvert Counties  area
(southeastern  region) but  not in  Northern  Virginia/Montgomery Count y
(western region)  or in  the Baltimore area (northern region) .  Likewise,  when
broken down by subarea the difference  in  racial composition appears  evident
in  the  southeastern and western subareas  but not in the  northern subarea.

    Some information can be  extracted from  the telephone survey about boat
ownership as well.  Once  again average family size and  years  in the  area were
not  significant/ Different over  the two groups,  but average  income and  the
racial composition of the sample were  (see Table 5.2).

    The design ofthe boat owners survey permits a  distinction  to be drawn
between individuala  who  trailer their boats  and  those  who keep their  boats in
the  water during  the season (either  at  marinas,  docks or moorings) .    The
distinction between individual  in these groups is important to establish, since
the  decisions  they face  are quite  different and their behavior must be
analyzed separately.  Additionally,  we  shall see that profiles of both  boats and
owners differ somewhat between the  two groups.   The sampling design  was
stratified to contact approximately  equal  numbers from the two  groups. Of
those  who  returned questionaires, 718 trailered  their  boats and 788  kept their
boata in the water during the  season.

    Some interesting features of these two groups are presented in Table 6.3.
By  far the  most  common type of  boat in the  trailered boat  sample  is*
runabout.    For obvious  reasons,  the  sanple of trailered boats contains  very
few with cabins (4 percent) .   It also contains few  sailboats (only  6  percent),
but  this  is not  a  representative figure,  since  sailboats which do not  use
auxiliary motors are not required to register in Maryland and thus  would  not
be part of  the population sampled. Of boats  kept  in the  water,  runabouts  (at
33 percent)  remain the single moat common class and sailboats  (at  31 percent)
represent  a close second.  Once again sailboats with no auxiliary power  are
likely  to be under-repreaented  in the  sample.   However, this distortion will
affect the trailered  group more, as the boats kept  in the  water are  larger  and
more likely  to have auxiliary engines.   Combining cabin  cruisers  and cruising
sailboats, half the boats  in  the non-trailered sample are cruising  boats
presumably outfitted for more  than one-day  trips.

    Aa the difference  in  the type  of boat inthe two  groups suggests,  the
average  size  and  the average value of   boats kept  in  the  water are
significantly larger than the  averages for  trailered boats.   Additionally,  the
average income  of boat owners inthe two  groups is significantly different,
with  trailered boat owners having on average  lower  incomes.

    Returning  to the telephone  survey  which  provides data  on boaters  and  not
just boat owners, we can learn something about  the geographical distribution

-------
                                  Table 5.1

             Average Characteristics of Boaters  and Non-Boaters
                     in  BaMmore/Washington SMSA, 1984
                        from random telephone survey

Average Family Size
Average Years Lived in Area
Average Household Income8
By Area:
Northern MSL. /Montgomery Cty., Hd.
Prince George's, Anne Arundel,
Calvert Ctya. , Md.»
Baltimore County
Percent Whitea
By Area:
Northern Va . /Montgomery Cty. , Hd.*
Prince Georges, Anne Arundel,
Calvert Ctya. , Md.a
Baltimore County
Boaters
3.31
29
$46,858
$50,576
50, 2s8
41,824
88*
97%"
83*
87*
Non-boaters
3.47
26
$37,063
$41,471
28,083
38,211
74%
77*
64%
80*
aMeans of two samples are significantly different at 99% level
source : Telephone Survey, Research Triangle Institute, 1984
                                  Table 5.2
         Average  Characteristics of  Boat Owners and Non-Boat Owners
                     in Balttmore/Washington SMSA,  1984

Average E&mly Size
Average Years Lived in Area
Average Household Income3
Percent Whitea
Boat owners
3.67
28.6
$56,511
94
Non-boat Owners
3.27
28.2
$40,931
81
      of two samples are significantly different  at  99%  level.
Source: Telephone  Survey,  Research Triangle Institute,  1984.
                                      73

-------
                                  Table 5.3
                  Characteristics of Boats and Boat Owners
                       from Boat Owners'  Survey,  1983

Total number responding to
question of where boat kept
By boat type
Runabout
Cabin Cruiser
Cruising Sail
Day Sail
Workboat
Houseboat
Rowboat
Other
NA
Used boats for swimming
at least sometimes
Used boats for swimming
usually or always
Used boats for swimming
always
Used boats for fishing
at least sometimes
Used boats for fishing
usually or always
Used boats for fishing
always
Average Income of Owner'3
Average Current Boat Value'3
Average Boat Length"
Trailered Boats
718

445 (63%)
23 (3%)
4 (1%)
32 (5%)
124 (17%)
1 (<1%)
47 (7%)
35 (5%)
7
360 (51%)a
133 (19%)a
41 (6%)a
656 (94%)a
502 (72%)a
302 (43%)a
$38,000
$14,000
16 feet
Boats Kept in Water
788

251 (33%)
182 (24%)
199 (26%)
38 (5%)
38 (5%)
13 (2%)
25 (3%)
22 (3%)
20
555 (73%)a
235 (16%)a
41 (5%)a
582 (76%)a
290 (38%)a
107 (14%)'
$51,000
$25,000
23 feet
aNumbers  in  parentheses are percent of those answering question  in each
. stratified sample who gave this response.
"Means of two stratified samples  are  significantly  different at 99 percent level.


Source:   Maryland Boat Owners  Survey, 1983



                                     74

-------
of households interested in this  activity.    The  geographical distribution of
boaters  within  the state tells  us  something about  the importance  of  thie
activity to different population subgroups.   From  the telephone survey  it  is
clear that boating is an important activity to Maryland residents throughout
the state; it is  the  first  or  second moat popular recreational activity on the
Bay in each of the geographical tubareaa of the study. However, as  might be
expected, boaters are moat commonly residents of counties contiguous to the
Bay.   A good example is the large proportion of Anne Arundel  residents who
participate in boating.   Extrapolating from our  survey suggests  that almost
half  of  the households  in Anne Arundel  County had  at least one member who
went boating on  the  Bay in 1984.

    Table  5.4 crntains residence data gleaned from the boat  owners survey for
boat  owners, by trailered  and nontrailered  classes.  The values shown  in  this
table indicate  the distribution  of  residence  counties  among  those who
responded  to this  question  in  the boat owners survey.  Of those who reported
residence,  64 percent  ofeach  group lived  in counties contiguous to  the Bay.
Approximately  60 percent of the  respondents came from the  four  most
populated counties in  the state -  Baltimore, Anne  Arundel,  Montgomery and
Prince Georges.    A final interesting  feature of the sample is that  about  5
percent of the respondents lived  out-of-state even though  they registered
their boats in Maryland.

    The fact  that boating  is often the vehicle for  other Bay recreational
activities makes the  analysis of boating both important  and complex.   The
incidence of multiple activities  is critical because participation in the
complementary activities of swimming and fishing may make boaters  more
sensitive to  water quality.   Additionally,  the overlap  of activities complicates
benefit estimation  when benefits are aggregated over  activities.

    From  the boat owners' survey  we can learn something  about  the
importance ofthese multiple activity  trips.   Table  5.5 reports  frequencies  of
responses to questions regarding  these  complementary activities.  One  striking
feature  of these answers is that  fishing  is extremely  important to those
registered boat owners  who trailer  their  boats.  In  fact almosta]] (94 percent)
of the trailered  boat owners  who responded  to  the  question indicated  that
they fished at least  occasionally from  their boats, and almost half (43 percent)
claimed  to fish on every trip.   Fishing was less  important  among the
non-trailered boat  owners, although three-quarters of  them indicated  that  they
fished  from their boats at least Occasionally.   About the  same percentage of
this  group indicated they  sometimes ueed their  boats for swimming.   Fewer,
about half, of  the  trailered boat owners  sometimes  used  their  boats  for
swimming.
                                     75

-------
                                   Table 5.4
            Number of Trailered Boats  and Boats Kept  in  the Water,
                              By Residence,  1983.
Trailered Boats
Residence
Baltimore
Anne Arundel
Montgomery
Prince George's
Calvert
St. Mary's
Charles
Lower Western Shore -
Total
Cecil
Harford
Kent
Upper Bay -
Total
Dorcester
Queen Anne
Somerset
Talbot
Wicomico
Eastern Shore -
Total
Carol ine
Worcester
Carrel
Allegheny
Frederick
Garrett
Howard
Washington
Other - Total
Pennsylvania
Virginia
District of Cblvnbia
Not Identified
TOTAL





17
25
23


12
27
7


8
13
5
9
21


7
10
11
0
16
0?
19
20







162 (24%)a
99 (15%)
64 (9%)
71 (11%)
(3%)
(4%)
(3%)

65 (10%)
(2%)
(4%)
(1%)

46 (7%)
(1%)
(2X)
(1%)
(1%)
(3%)

56 (8%)
(1%)
(1%)
(2%)

(2%)

(3%)
(3%)
83 (12%)
13 (2%)
10 (1%)
6 (1%)
43
718
Boats





14
25
7


11
20
5


9
12
2
30
9


2
9
4
3
7
5
14
6






Kept in





(2%)
(3%)
(1%)


(1%)
(3%)
(1%)


(1%)
(2%)
(<1%)
(4%)
(1%)


(<1%)
(1%)
(1%)
(<1%)
(1%)
(1%)
(2%)
(1%)







144
184
60
28




46




36






62








50
67
36
18
57
788
Water

(20%)
( 25%)
(8%)
(4%)




(6%)




(5%)






(8%)








(8%)
(9%)
(5%)
(2%)


aNumbers in  parentheses represent percent  of those answering  questions  in each
 stratified sample who gave  this response.

Source:   Maryland Boat Owners Survey,  1983
                                      76

-------
                                  Table  5.5

              Percent of Boaters Who  Fish or Swim While Boating
                          Boat  Owners'Survey,  1983
                           Occasionally    Usually
Always
  At Least
Occasionally3
Trailered boats:
Fish while boating
Swim while boating
Non-trailered boats:
Fish while boating
Swim while boating

22%
33%

38%
42%

29%
13%

24%
25%

43%
6%

14%
5%

94%
52x

76%
72%
• Total of other columns
Source: Maryland Boat Owners Survey,  1883
The importance of Water Quality to  Boaters

    It is useful to consider  the  qualitative evidence that exists to support the
notion that water quality does in fact matter to boaters.   Some evidence to
this  effeot can be  found in  the  1983  boat owners survey.  Tables 5.6  and 5.7
present  a compilation of responses to  a series of questions about  factors
important  in the selection of boating areas. As can be aeon from  these  tables,
boat  owners who  trailered their boats  considered water  quality to be  the
moat  important  factor in choosing boating  areas  Water quality waaconsidered
"very  important" or  at least  "moderately important" by the  non-trailered boat
owners more often than any other  factor except  water depth.   The latter is
often a physical constraint for the  larger  boats  found in  marinas.    Comparing
the two  subgroups, i.e. those who  considered water  quality  "moderately" or
"very" important and those who did not,  it  is interesting to note that the
former had on average significantly higher incomes and more valuable boats.


The Behavior of Boat Owners Who Trailer Their boats

The   General Model

    We are interested  in modelling two types  of  decisions  that  owners  of
trailered boats  make  in a season.   One of  these  is the  commonly modelled
economic decision of  how many trips the individual  takes.   The  second haa to
do with  the location  to which the boat owner takes his boat. This subgroup
of boat  owners is far more flexible in the short  run  than those  who keep  their
boats  in  the water during the  season,  because  on a day to day basis they  can
                                     77

-------
                   Table 5.6 Factors Cited Most Important
             in the  Selection of a Boating Area in Maryland, 1983

       Percent Response from718Boat Owners Who  Trailer  Their Boats
Factor
Water
Quality
Water
Depth
Natural
Beauty
Easy
Access
Lack of
Congestion
Not
Important
11.1
16.3
15.7
13.8
19.5
Slightly
Important
9.3
13.2
17.5
20.1
16.7
Moderately
Important
33.0
29.5
34.0
35.7
32.0
Very
Important
42.2
36.4
28.1
26.7
27.4
No
Response
3.1
3.3
4.6
3.8
4.3
Source: Maryland Boat Owners Survey, 1983
Table 5.7 Factors Cited Moat Important
in the Selection of a Boating Area in Maryland, 1983
Responses from 788 Boat Owners Who Keep Their Boats in Marinas
Factor
Water
Depth
Water.
Quality
Natural
Beauty
Protected
Anchorage
Lack of
Congestion
% Not
Important
7.7
9.3
9.8
23.5
19.9
% Slightly
Important
8.9
10.0
13.3
15.1
17.4
% Moderately
Important
24.4
35.7
41.0
31.5
37.2
% Very
Important
54.9
39.7
31.2
25.1
20.4
% No
Response
4.1
5 . 3
4.7
4.8
5.1
Source: Maryland Boat  Owners Survey,  1983
                                    78

-------
alter  the  boating area by  trailering to different  launch sites.  Of course, the
farther they must trailer the boat, the more costly  (in both  time and money)
will be the trip.   Responses to the survey indicate that factors besides  costs
are important in choosing boating area.    Specifically, water quality was
considered the moat important factor  in this choice,    Thus we analyze the
demand for trips to  different sites  with  measurably different water quality.
This gives  us some basis  to deduce how  demand might change if  water quality
were to change at different sites.

    In order to accomplish  this, a model of multiple site demand is designed to
accommodate differing qualities across  sites but to  require  no more data than is
available  from the  boat owners'  survey and  Chesapeake  Bay  water quality
data. Given these  limitations  and a  desire  for consistency of  analysis  across
recreational  activities,  the  varying parameters model presented as  the first
method of analysis in  Chapter 4 is used.

    The model takes the following form.   The demand for trips  to each site j
is estimated  as a  function  of the coat to individual  i of  accessing the  site
 (c J j  ), substitute  site ace-a  ccsts (s  ^ j ),  and other  exogenous variables
associated with the individual  (s ^ ) :
 (5.1)           xjj  = fjicij,  Sij,  zj;  fj)       for all i,

where  f  j  is  the demand  function  for  the  jthsite,  fj  is the vector of
parameters in  each of the site demand  functions  to  be  estimated.

    Equation (5.1), which  is the first stage  of the varying parameters models,
can not be estimated using  ordinary least  squares methods.  The  sample upon
which the  estimation is  baaed  includes a large number  of zero values for the
dependent variable.    As described  earlier in  this volume,  ordinary least
squares applied  to censored samples will  produce biased estimators. As
described in the last chapter,  Tobit estimation  procedures are used to  correct
for the problem.

    The second stage  of the model  relates  the set of f parameters  in the site
demand  functions  to the site water  quality characteristics  (a vector w  i), In
this  way the  demand  for a site is  implicitly modelled as a function  'of the
site's  characteristics.  The fecond stage model  is of the  form:


 (5.2)                      fk. =gk(w,  )     forall  j,

where k indexes the Specific  f coefficient within the demand equations  and  j
indexes the site demand. Again the  application of ordinary least squares  is
not optimal.    The  above model likely  suffers  from heteroskedasticity  (see
Desvousges,  Smith,  and McGivney,   19S3) which will produce  Inefficient
estimators.   To correct  for the  expected heteroskedasticity,  the  entire  equation
can be  multiplied by the reciprocal of  the  standard error of the respective
estimated Coefficient   Thus,  if »k  j  is  the standard error of the  estimated
coefficient fk j from the Tobit estimation Of the  demand for site   j, then the


                                     79

-------
second stage model corrected for  heteroskedasticity  can be written


(5.3)


for each of the k=l,...,K P coefficients  from the  first stage.


The Data

    The arguments of equation (5.1) which were determined to be relevant  for
the analysis include the coat of  access  to the site,  the coat of access to  a
substitute site,  and  the value  of the boat.   The latter variable waa chosen as
a surrogate for two very important factors.  Boat  value and  income are  highly
correlated , so boat  value  serves  as a surrogate  for  income level, a variable
which is normally rather  difficult  to obtain with accuracy.  Boat value alao will
be highly correlated with boat  sise which haa an  effect on the site choice.
Since boat length and income are correlated, including  them  separately  in  the
equation  produces unresolvable  multicollinearity.

    The coat  of access  variable is  the roundtrip  cost, including time  coat.
Costs vary across boaters  for several reasons.  Cbviously the county  of origin
influences costs, for the  further the county  is from the  launch site,  the
higher the coat of travel. The cost  variables include time costs, computed as
one-half  the  individual's  average  hourly wage  (incorie/2080)  time, the  distance
traveled divided by 40 miles per hour.   Finally  the money cost of trailering  a
boat depends  on the size  of  the  individual's boat.   The  coat  per mile of
trailering a boat -was estimated,  using coat and boat length  information from
the data set, as -.78  +  .08 *(boat  length). The final  coat variable includes  the
toll for  the  Bay Bridge, when relevant.

    To calculate the  substitute  cost, the above  formula waa  applied  to  the
closest  cite not  chosen.  While a vector  of costs to all  alternative sites might
be considered preferable,  there are several practical  problems with including
such a vector in the regression, espicially severe multicollinearity.

    Not  all  observations on boat owners  who  trailered  their boats were ueed
because  of heterogeneity  of respondents and incomplete data.   Observations
were deleted if the individual did not use his boat for any trip, in  1983 or if
he did not report launch sites or  location  of residence.  Additionally,  those
who reporeted their residence to be a  distant state,  precluding day trips to
the Bay, were deleted.   Finally, to  make the sample relatively homogeneous,
sailboats were excluded from the  analysis.  The latter  accounted  ultimately  for
only seventeen  deletions.  The final sample  included  408  observations.

    The  above information, waa obtained  from the boat  owners'  survey.
However,  this data  source dose  not include  any  information about either
perceived or measurable water quality  at  various  sites.    Once again the
Chessie system  provided  the environmental data, and  the  environmental  quality
variable  used was the product of nitrogen and phosphorus,  as  in  Chapter 3.
                                      80

-------
The Estimated Model
    For the first stage  of  the model,  the parameters in the following functional
form are  estimated:


(5.4)


where x jj is  trips  to site j,      is cost  to  site j , s      cost  to  next best
alternative to  site j,
         jj is  trips  to site  j, ^  j is cost to site j ,  s^     co
       ive to site j, and b i is boat  value,  all  for individual i.
    The first  stage  results  for  each of the twelve sites are presented in Table
5.8,   The  results  are remarkably  consistent across  sites.   The own price
coefficients are all negative and  significantly different from zero at the 99%
level of confidence.  Substitute price coefficients  are universally  positive and
significantly different from zero for eight of the  twelve sites.   The coefficient
on boat value is significantly  different from zero  for seven of the sites and in
each of theaa  cases haa  a positive sign, suggesting that wealthier people
and/or people with bigger boats take more boating trips,  ceteris paribus.

    Demand is relatively inelastic  with  respect  to substitute price and boat
value (i.e.  a  1  percent change in either of these  causes a less than  1 percent
change  in the  demand for  trips),   However,  the demand for  trips to a  site  is
quite elastic with  respect  to  the cost of  accessing  the site, with own  price
elasticity ranging from -1.5  to -7.0.

    In the second stage,  there  are as many equations as there are parameter
from the first stage which we wish to allow to vary with  the environmental
factors.    Since  we have  no particular  a  priori information as to what
parameters might vary, we  can. model each as  a  function  of  the environmental
variables  and allow  the  test statistics to determine the outcome.           *

    The second stage model  is given  by


(5.5)                  JkJ  = a0  + at TNPj + vj


and  the results  are presented  in  Table 5.9.    The product of  nitrogen  and
phosphorous  serves  as  the  environmental variable.  The regression of the  own
coat  coefficients from  the linear  first  stage model  on these environmental
variables produced good results.     The coefficient is  significantly different
from  zero and negative.   The  negative sign suggests that  the demand curve
becomes less steep with increasing levels of pollutant.

    Neither the  constant  term nor  the coefficient on boat value yielded
significant second stage  results.    Even  though the coefficient  of  substitute
price waa   associated  with  a   significant   negative  coefficient  on   the
environmental variable, allowing  this  coefficient to vary  had no  appreciable
effect  on the welfare results.   As  a consequence,  in the remainder of the
analysis  we allow only the coefficient on own price  to vary  wtth  environmental
quality.   The  results  suggest that an increase in pollution  would tend to  have
the  effect  ofpivoting inward the  demand for  trips  to the site.
                                      81

-------
                                  TABLE  5.8




        Estimated Tobit  Demand Coefficients: Maryland  Counties 1983
Estimated Coefficients (Variable)3
county
Anne Arundel
Baltimore
Calvert
Cecil
Charles
Dorchester
Harford
Kent
Queen Anne's
St. Marys
Somerset
Wicomico
,aEach equation
Pi
(cost/
trip)
-.13 .
(8.75)b
-.43
(9.21)
-.14
(4.14)
-.22
(4.84)
-.34 .
(8.41)
-.09
(2.86)
-.15
(5.55)
-.25
(4.94)
-.27
(6.17)
-.11
(6.40)
-.12
(4.76)
-.15
(6. 93)
is estimated
P,
(substitute
cost/trip)
.03
(3.42)
-.02
(1.13)
.08
(13.70)
.04
(1.54)
.07
(3.77)
.08
(2.69)
.05
(2.63)
. 1
(3.57)
.07
(2.88)
.05
(3.12)
-.03
(.58)
.05
(1.58)
with the 496
/»3 do
(boat value (constant)
1000's)
1.29
(5.94)
1.78
(4.01)
1.84
(3.45)
2.12
(3.09)
2.75
(6.79)
.66
(.78)
1.51
(1.67)
1 .14
(.14)
.12
(. 19)
1.25
(2.94)
2.81
(3.13)
1.02
(1.71)
boaters who
-2.21
(1.61)
-1.94
(.77)
-27.14
(7.21)
-16.44
(3.87)
.49
(• 19)
-34.38
(6.68)
-12.21
(3.74)
-18.25
(3.45)
-3.83
(1.03)
-9.46
(3.31)
-37.20
(6.64)
-7.03
(2.02)
trailer their
non-
limit
observa-
tions
142
75
44
17
38
30
36
28
36
67
24
26
boats.
^Absolute values  of t-statistics in  parentheses

-------
                                 Table 5.9

                    Estimation  Results from Second  Stage

Regression of trip coat coefficient
Regression of substitute cost coefficient
Regression of constant
Constant
-.0887
(4.29)a
.0682
(7.47)
-19.414
(4.14)
TNP
-.000102
(3.54)
-.000016
(1.78)
.007338
(1.93)
• absolute values of t-statistics in parentheses
Modelling  the Behavior of Boat Owners Who Do Not Trailer Their Boats

    Boat owners who keep  their  boats in the  water during the season are far
more restricted  in the short run as to the quality of the water that they  can
easily enjoy.   To a/ferthe  area in which they boat they must take long trips
in their boats  or  they must change mooring arrangement.   We  do not have
access to information that  would allow us to model  either of these decisions.
None ofour data allows us  to observe these  individual  making trade-offs
between water quality and other goods ormoney.   As a  consequence we can
not  deduce from observations on  their behavior how they might value
improvements  in  water quality.

    What we can  do however  is to learn something about their demand  for
boating  and  their  use of the Chesapeake, which in itself  is useful  information
for the  policy maker.   In  this section we  estimate the demand for boating
trip,  by  boaters  who keep  their boats  in marinas.   Since we  are interested in
the short  run  decision of how many tripe to  take, the relevant cost variable  is
the  variable coat of  atrip.     Given the information available, we  can
approximate the money and  time costs of travel  to the  marina.   Explanatory
variables which  may shift  the demand function include income  and  the size  (or
value)  of the  boat."

    The boats in  the  marina  subgroup  are somewhat  heterogeneous. For  one
thing, about  half are  sail boats and half are motor boats.   Consequently we
might  wish to teat whether  the demand functions  for the  two  groups are
significantly different.

    The results of thin analysis  are reportad  in Table 5.10.  The  coefficient on
coot is negative,  as  expected, and  significantly different  from  zero.    The
coefficient on  boat  value is positive, indicating that,  all  else equal, boaters
with more expensive boats take more trips.   Income  appears not to  affect
systematically the demand for  trips.


                                     83

-------
                                 Table 5.10

                     Estimated Demand for Boating Trip.
                            Boats Kept  in Marinas
Explanation Variable
Cost
Boat Value ($1, 000)
Income ($1,000)
Sailboat Indexa
Cost Sailboat Index
Constant
R2 = .109
# of Observations = 240
Coefficient
-1.046
.357
.148
-23.596
.615
63.363


t-statistic
-3.93
1.94
.66
-2.27
2.16
6.66


•Variable equals 1 if boat is sailboat,  0  otherwise
    Both the  constant  term and the oostcoefficient  shift significantly for
individuals who own sail boats.  The  demand  function  for non-sail  boat  trips
is  given  by

            TRPS =  63.36- 1.05 coat  + .36  boat value + .15 income

where  boat value  and  income  are in  thousands  of dollars.    The  sail boat
demand for trips is given by

            TRPS =  39.77-  .43  cost + .35 boat  value +  .15 income.

At the same cost and boat value,  sailboat owner,  demand  fewer trips, and
their demand  for  trips appears to be rim-e inelastic.

    Because of the eventual need  to aggregate behavior over recreational
activities,  it would  also be useful  to know  whether those boaters who  •penal a
large  portion  of their  time fishing have significantly different  demands from
those who  do not.    These results are presented in Table 15.11.   It  appears
that the  two  groups demands are  different.   For  any  given cost  and boat
value, fishermen demand more trips and their demand tends to  be  more elastic.
Fishermen},  demand is given by
                                     84

-------
            TOES = 73.61- 1.32 coat + .35 boat value + .15 income

and demand by non-fishermen  is

            TRPS = 43.00- .51 cost + .35 boat value + .15 income.
                                 Table  5.11

            Estimated Demand  for Boating Trips,  Fishing Behavior
                           Boats Kept in Marinas
Explanatory Variable
Trip Coat
Boat Value ($1,000)
Income ($1,000)
Fishing Indexa
Coat Fishing Index
Constant
Coefficient
-.506
.350
.148
30.535
-.815
43.085
t-statistic
-3.86
1.90
.88
2.8s
-1.82
3.81
R2 = .116
• Variable equals 1 if  boater fishes on boating  tripe "usually"  or "always,1
 0 otherwise.
calculating  Estimates of the Benefits of Water Quality

Improvements for the Trailered  Boat Samle

    Because we have bean  able  to estimate the demand for boating trips by
boaters who trailer their  boats to different areas  as functions of costs  and
the water quality  in  those areas, we can estimate welfare gains  and losses
from  water quality changes to  this group.    Unfortunately no observable
behavior of the  boaters who keep their boats in the  water allows us to deduce
anything about the value  they place on  improved water quality.

    As  in the previous  chapter,  three  different changes in the environmental
variables are considered.    In one caae we  impose  a  20 percent decrease
 (environmental improvement)  in   the environmental (pollution) variables.
Subsequent  experiments include a 10 percent decrease and a 20 percent
increase (environmental  degradation).


                                    85

-------
    Additionally,  two  different calculation  procedures are provided,  identical  to
those used With  the varying parameters model  in Chapter 4. The two methods
of  calculating "before" and "after" trip demands yield two sets of  average
consumer surplus estimate, before and after  the environment change and,
consequently,  two sets of benefit (or leas) estimates due to  the  environmental
change.  (A description of these methods can be  found in Chapter  4.)  The
estimates  are reported in Tables 5.12?  5.13,  and 5.14  for the  10  percent
improvement  20  percent improvement  and 20 percent  deterioration  in the
water quality variable,  respectively.   The  average  consumer surplus  figures
are per  boater ( trailered boats, only) per season  for the designated  site.
Thus the first entry  in  Table  5.12 is an estimate of  the value of access  to
sites in Anne Arundel County per boater in the sample.

    In examining  the consumer surplus figures,  it is well to keep in mind that
these benefit estimates are affected by  the probability that a boater will go to
a particular  site, the number of trips  taken,  given that the boater  goes  to the
site, as well  as  the  size of the own-price  coefficient.   The figures in columns
3 and 6  are estimates per  boater of how the  value of access to a  site  changes
with changes in the environmental variable.   These surpluses are not additive
across sites.    That is, if we want to  consider  the effects of  a ten-percent
decrease  in  TNP  throughout  the  Chesapeake  Bay, we cannot add  surplus
changes across sites.   Each  estimate of surplus  per boater by site assumes
that the water quality  at other  sites remains  fixed.  The bias  which would be
created  by  simple aggregation  across sites depends  on price and  quality
elasticities and is  of unknown size.

    What do all  these  calculation say  about the value of reductions in the
nitrogen/phosphorus variable?   We can estimate the aggregate benefits  of
changes in TNP  from Tables 5.12,  5.13,  and  5.14 by expanding  these estimates
from  the sample to the  population of  boat owners who  trailer their boats.
Consider  St.  Mary's  County.  We have  two estimates of  the  increase in  Surplus
associated with a 10  percent  decrease in  TNP.   These suggest a range  of
between  $1 and $5  per boater  per season.  If there are  about 80,000 boaters
who trailer in Maryland   (about the number  estimated  by  Harmon and  Associates
for 1983),  we  would estimate a change in total surplus  in  the range of  $575,000
to  $1,400,000   annually for a  20 percent reduction in TNP at the sites that
Anne Arundel  County  comprises.    This calculation assumes  that the  original
sample from which the  benefit estimate,  were derived is representative of the
boater population as a whole.

    The  results  in these tables eeem at  least plausible.    For example,
surpluses  appear highest  for western shore waters,  those moat easily accessed
by the concentration of population in the state.  The  surplus  figures for any
given site are  not especially  large,  but this is to  be  expected since when
boaters have the ability to substitute  relatively cheaply among sites, very
high  surpluses  at any one site  would violate some prior  expectations on the
size of benefits.   While  the magnitudes of  returna  from changes in TNP are
not  extremely large on a  per-boater basis,  in  the aggregate they are quite
substantial
                                     86

-------
                                                        Table 5.12
                                Per  Boater  Annual  Benefits  from  a 10% Decrease in Pollutant
                                                in  each Geographical Area
CJ
Calculation Method
County with
Water Quality
Change
Anne Arundel
Baltimore
Calvert
Ceci 1
Charles
Dorcester
Harford
Kent
Queen Anne's
St. Mary's
Somerset
Wicomico
A
Calculation Method B
Consumer Surplus Consumer Surplus Consumer Surplus
Before Change After Change Benefits Before Change
$30.01
15.07
17.50
4.80
38.79
1.61
3.45
24.08
26.17
14.80
7.17
7.87
$33.30
17.38
18.60
5.43
46.05
1.78
3.89
27.09
28.99
16.03
7.60
8.60
$3.29
2.32
1.11
.63
7.26
.17
.45
3.00
2.81
1.23
.44
.73
$119.05
49.83
108.57
18.55
38.34
75.72
47.24
73.71
51.74
139.22
99.18
32.82
Consumer Surplus
After Change
$127.46
54.95
111.38
19.51
43.40
78.15
49.63
76.86
53.98
"143.71
101.95
34.53
Benefits
$8.41
5.12
2.81
.96
5.05
2.43
2.38
3.15
2.24
4.49
2.77
1.71

-------
                         Table 5.13
Per Boater Annual Benefits  from a 20% Decrease in Pollutant
                 in each Geographical Area
Calculation Method A
Count y with
Water Quality
Change
Anne Arundel
Baltimore
Calvert
Cecil
Charles
Dorcester
Harford
Kent
Queen Anne's
St. Mary's
Somerset
Wicomico
Consumer Surplus
Before Change
$30.01
15.07
17.50
4.80
38.79
1.61
3.45
24.08
26.17
14.80
7.17
7.87
Consumer Surplus
After Change
$37.17
20.33
19.79
6.17
55.54
1.98
4.41
30.51
32.18
17.40
8.08
9.46
Benefits
$7.16
5.26
2.30
1.38
16.75
.37
.97
6.42
6.00
2.61
.91
1.59
Calculation Method B
Consumer Surplus
Before Change
$119.05
49.83
108.57
18.55
38.34
75.72
47.24
73.71
51.74
139.22
99.18
32.82
Consumer Surplus
After Change
$137.05
61.20
114.35
20.60
50.13
80.74
52.27
80.38
56.51
148.54
104.88
36.44
Benefits
$18.01
11.37
5.78
2.06
11.79
5.03
5.03
6.67
4.76
9.32
5.70
3.62

-------
                                                         Table 5.14
                                 Per Boater Annual  Losses  from a 20% Increase  in  Pollutant
                                                 in each Geographical Area
oo
V0
Calculation Method A
County with
Water Quality
Change
Anne Arundel
Baltimore
Calvert
Cecil
Charles
Dorcester
Harford
Kent
Queen Anne's
St. Mary's
Somerset
Wicomico
Consumer Surplus
Before Change
$30.01
15.07
17.50
4.80
38.79
1.61
3.45
24.08
26.17
14.80
7.17
7.87
Consumer Surplus
After Change
$24.76
11.68
15.51
3.80
28.78
1.33
2.74
19.12
21.48
12.70
6.40
6.68
Losses
$5.24
3.38
1.99
1.00
10.01
.28
.71
4.96
4.70
2.09
.77
1.19
Calculation Method B
Consumer Surplus
Before Change
$119.05
49.83
108.57
18.55
38.34
75.72
47.24
73.71
51.74
139.22
99.18
32.82
Consumer Surplus
After Change
$105.07
41.96
103.98
16.93
31.51
71.30
43.12
68.34
47.93
131.10
94.08
29.88
Losses
$13.98
7.87
5.19
1.62
6.83
4.42
4.13
5.37
3.81
8.12
5.10
2.94

-------
                                 Chapter 6

             The Benefits  for  Recreational Fishing: Striped Bass


      This  chapter provides come preliminary estimates  of  the increase in
benefits to sport anglers  from increases  in water  quality.  We  use aportion  of
the 1980 National  Survey of Fishing,  Hunting and Wildlife Related Recreation,
referred toas USFWS data,  to  estimate the demand for and value  of  fishing
for striped bass in Maryland.    This  survey, while  not designed  for  these
purposes,  is the only data aet currently available which  enables  us to
investigate the  recreational fishing of the Chesapeake.   Striped bass is the
only  specie, important to the Chesapeake Bay recreational fishery forwhich
there  is sufficiently detailed-catch information to  link water quality changes to
the benefits of sportfishing.

    The link between  improved water quality and changes in recreational
fishing  demand  depends  on the ecological  connection  between water quality
and catch rates  and the behavioral  connection betwean  catch  rates and fishing
activities.  Descriptive and  analytic studies of the  Bay  have focused  on the
impacts of water  pollution on the  density  and productivity offish stocks.
Lower dissolved  oxygen,  declines in SAV,  and  increases in water toxicants all
appear to have an  impact  on  fish stocks.  Further,  where records are  kept  for
commercial fisheries, there  haa been  a  substantial decline in landings per unit
effort,  especially for  those species which spawn in the Bay or ita tributaries

    It  is plausible to expect considerable benefits to recreational fishermen
from  improvements in water  quality.   The  number of recreational anglers is
quite large, baaed on information from  the primary sources of data on
saltwater  recreational  fishing in Maryland.    Estimates  of  saltwater fishing
participation  in  Maryland  during 1980 range from 539,000 anglers over 16
years  of  age taking 4.1  million trips to somewhat over 800,000  anglers of  all
age. taking 2.7 million trips  (U. S.  Fish and Wildlife Service  and Bureau of
Census; U. S. National  Marine Fisheries Service; William  et  al.).  According to
NMFS  and State  of Maryland data, each saltwater angler took approximately
three trips, while USFWS estimates  approximately 7.6 trip, and 9.0  days fished
per angler.

    Data on  striped  bass  fishing are somewhat more difficult to obtain.
According to the Maryland Department ofNatural Resources, roughly 203,000 of
the saltwater  trips were for  stiped bass.   Our analyais of the  USFWS  data
indicates  that 239,000  anglers  (over  16 years of age)  fished for  striped bass
in Maryland and Sussex County,  Delaware, fishing for approximately  2.1  million
days,  or roughly 8.8 days  per  angler.  Estimates of    the    striped    baas
recreational catch  in Maryland range  from 211,000 to 377,000  fish,  a total
weight  of 200  to 474 metric tons.   The USFWS data are not well suited  for
estimating aggregate  catch, because  the survey used waa designed primarily
for other  purpose.,  even though  catches are self-reported by respondents  for
come saltwater  species, notably striped bass.
                                     90

-------
    Table  6.1 provides some  descriptive   information about  the sample of
anglers which  was  analyzed  in this portion of the study.    The sample was
partitioned into two groups  baaed on whether the individual fished for striped
bass or not.  Individuals in the  two subsamples  are very similar  in the amount
of  fishing  and hunting  done and in their  exprience, income, age,  education,
and other demographic makeup,   Striped baea fishermen,  on average, showed a
slightly higher propensity  to  own a  boat and to allocate moire  money to
hunting and fishing  activities, though  these differences  are not  significantly
different from zero due to  the high within-subsample variation.
                                  Table 6.1
                   Characteristics of Striped Bass Fishermen
              and Other Fishermen and/or Hunters in the  Sample
                                               Striped Baaa
                                                Fishermen
               Non-Striped  Baaa
                  Fishermen
Number of Individual  in Sample
Average Number of Days  Fishing,  Striped Baas
Average Number of Days  Fishing, All Species
Percent Who Also Hunted
Average Number of Days,  Hunting
Average Years of  Fishing  Experience
Average Age When First Fished
Percent Owning Inboard Boat
Percent Owing Outboard Boat
Percent owning  Other Boat
Average Household Income
Average Fishing/Hunting Budget in 1980
Average A9ga
Average Years of  Schooling
Percent Working  in Job  or Business
Percent fran Urban Areas
    184
     11 daya
     28 daya
     41%
     17 daya
     24 years
     10
     19%
     42%
     17%
$28,300
   $982
     38
     13 years
     70%
     44%
    576
      0 days
     2'7 daye
     37%
     15 days
     24 years
     12
      7%
     28%
     12%
$27,600
   $588
     38
     13 years
     73%
     38%
aThe sample is  for individuals 16 yeara  of  age and over.
                                      91

-------
A Description of the Data

    The  1980 National  Survey of  Fishing,  Hunting, and Wildlife-Associated
Recreation was the  source data  for analysis.   Of  the available data  sets on
Chesapeake Bay sportfishing,  the portions of this survey  relating to  saltwater
recreational fishing  in  Maryland, and by Maryland residents, offered the best
prospects for modelling the  effects of water quality improvements. This data
set contained  the  essential variables  for estimating recreational fishing demand
functions,  including  information  on  (a)  trips  taken by destination;  (b) costs
incurred by recreationists for goods and services  used in  recreation;  (c)
household income;  and   (d) catch rates  reported by anglers for certain
species.

    The survey  consisted of  two parts.   The first was a telephone screening
of households,  predominantly by telephone interviews,  to  collect  demographic
characteristics and  to  determine  the hunting,  fishing, and non-consumptive
recreation  activities  ofhousehold  members during 1980.   The second part was
a detailed questionnaire administered (typically in person)  to selected
individuals who  indicated they had hunted  or  fished in  1880,  collecting
information on activities  and expenditures.    Of  the 30,300 fishermen and
hunters and 6,000 non-consumptive users  interviewed  nationwide,   760  pursued
some or all  of these activities in  Maryland.   These 760 individuals were the
subject of this  analysis.

    Of the  760 who hunted,  fished or Participate in  non-consumptive
wildlife-related activities  in Maryland,  456 indicated  they participated in some
form of saltwater  fishing.   Catch rate estimates were only obtained for a
limited number  of  saltwater and estuarine species, with  striped baas the only
recorded species relevant to Maryland.   One hundred eighty-four individuals
indicated they fished for striped baas in 1960.

    The survey waa  designed  to provide estimates  of recreation activities and
expenditures at the state level, and states were  divided  into large  subregions
for purpose, of identifying  trip destinations.  Maryland was  divided into four
such regions, three of which border the Chesapeake and were the  location of
striped bass fishing.   Broadly defined, the four areas are:   the Southeastern
Chesapeake  region,  Northern  Chesapeake,  Southwestern Chesapeake, and
Northwestern Maryland.   Significant  numbers of Maryland residents  also fished
for  striped baaa  in Sussex  County, Delaware.    Of the  184 striped bass
fishermen in the sample,  16  reported fishing  in  Delaware, 46 indicated they
fished for striped baaa in the  Northern Chesapeake,  59 fished in the
Southeastern Chesapeake  region,  and 86 in the  Southwestern  Chesapeake
(Table 6.2).

    The data aet includes  days fished for • trtped  bass and  other species,
rather  than number of trips by  specie.,  the latter being the  preferable
measure for  travel cost models.   The survey did, however,   include the  total
number of trips  to each region.  Aggregating over  all areas to get total  trips
and all species to get total days  fished,  it was determined  that anglers took
about 4.1 million trips and fished  about 4.8 million days, yielding an average
of 1.17  days/trip.   Thus,  the  two  measures  may  not be bad  approximations  of
one another.


                                     92

-------
                                  Table 6.2

    Sample Distribution Number ofFishermen,  Days  of Striped Bass Fishing
                     in  1880, and  Catch Rate, By Regions
Region
Sussex DB
Northern Chesapeake3
Southeastern Chesapeake*3
Southwestern Chesapeake
Number
Who Visitad
16
46
59
88
Mean Striped
Baas Days
6.8
9.6
11.3
8.8
Mean Catch Rate
Fish/Day
4.4
4.9
3.3
2.8
 'Baltimore City  and  Bait  imore,  Carroll  , Cecil , Harford, Kent , and Queen Anne's
 counties.
^Carol  ine, Dorcester,  Somerset,  Talbot,  Wicomico,  and Worcester  count ies.
cAime Arundel,  Calvert,  Charles,  Howard, Montgomery,  Prince George's,  and St.
 Mary's  counties.
    While detailed information  was collected on costs of  travel,  lodging,  food,
fees,  and other  expenses incurred during recreation trip.,  these costs were
not area specific;  instead total expenditures over all saltwater fishing  trips to
all areas (regardless  of species  sought)  were  collected for  each  cost category.
The  variable cost of trips to  a single  area  could be  determined only by
prorating  total variable  costs according to  distance travelled.    The  method
ueed  in this analysis waa to determine  the  total miles  travelled by  the
individual  for all  saltwater fishing  trip, in 1980,  aa  the •m of products of
round trip miles  travelled  to (the  usual fishing location in)  each  area and the
number of trips  taken to  each area,    The  fraction  of  total  variable  fishing
expenses prorated for  eaoh  trip  to  each  site waa  the  round trip miles
travelled  to the  site divided by total  miles  travelled.    The money  coat of a
trip to each  site waa  this fraction times the  reported total  variable costs for
saltwater fishing.  Espressed  as a formula,

                                n
                                B



where MC j j is the money coat  of a  trip by individual i to area  j,  Mj_ j is the
round  trip  miles travelled by individual  i to area  j,  •< i is the number of trips
individual  i  takes to  area  j, VC ( is individual  i'.  reported saltwater fishing
variable costs, and  there  are  n  areaa.

    The  coat of  time spent in recreation  is also  an  important determinant of
demand. The • urvey data were  not  ideal for  determining this coat because no


                                      93

-------
information  was collected on the time spent in travel or at the  site on  each
trip.   However,  miles traveled is a reasonably  good  proxy for time spent  in
travel.   The procedure used here waa to assume  an average  rate of speed
during travel of 40 miles  per hour  and that  the annual  household income
divided by the number of hours  in the  average full-time work year (2,040)
was  a suitable approximation for  the wage rate.    Then, the value of  time
travelled was determined as the product  of the  amount of time spent in travel
and 40 percent of the wage  rate.  Expressed  as a formula,
TCJ -
                               y«
                             2,040
                       204,000
where TC  i j  is the time coat  for individual i travelling to  area  j, y.  is the
household of person i, and M « * is,  again, round-trip miles.   Of course,  this is
a rather arbitrary formulation for time cost based on  a series of restrictive
assumptions, but preferable ways of treating the value of time were not possi-
ble given  the available data.   The full  price of  a  trip is then calculated as
the  sum of  the time and  money  prices  for each individual 1P1 = MCj + TC^.

    In the  survey,  respondents  were asked to estimate their average catch
rate per day for selected  species.   Unfortunately,  there was a lag of up to a
year or more between  the time the  fishing trip was taken and the time the
questionnaire was answered.  There is evidence  (e.g.  Deuel,  Hiett and Worrall)
that fishermen  do not accurately remember numbers  of fish caught or  their
sizes well beyond  a period of a few  months.   A  comparison of  the  USFWS data
and  data collected by the  State of Maryland suggests that the  USFWS data
might contain an  upward  bias  in reported catch rates.    The sample and
population  average  catch  rates were  both somewhat over three striped bass
per  day,  which is considerably higher  than the State  of Maryland data which
suggests  a  catch  rate  for the came period of one striped bass per  day.   When
the sample catch  rates  were extrapolated to estimate  total  1980  catch, the
estimate was an  order ofmagnitude or more larger than the  published
estimates noted in the  introduction,  although some of  this  difference may be
attributable to difference, in estimates baaed  on  total  trips versus  total days.
The  fact  that sample catch rates do not predict aggregate catch well does not
invalidate their uae as quality indicators,  however.    As indicators  of the
quality factors  which  signal individuals'  fishing decisions,  sample  catch rates
may perform  quite  well.

    The survey data contained a categorical variable measure of household
income.   A second measure  was  also calculated:   total budget for  fishing and
hunting recreation,  the  sum of all fishing and hunting-related expenditures in
1980. If the individual has a weakly separable  utility  function and determines
first  the  total  amount of income  to allocate to hunting and fising recreation,


                                     94

-------
the  fishing and hunting budget  to a more relevant  income constraint  than
overall  household income.   The fishing/hunting budget measures, however,  is
subject  to potential  errors  Of  measurement,  both  from  faulty recall by
respondents and  from year-to-year fluctuations due  to purchase of  major
durable  goods.


The  Basic Model
    For a variety ofreasons,  the model we  estimate  for recreational fishing  is
different from  the  recreational beach uee and boating model.    In the beach
use   and  boating   estimations , trip   data  existed    for a   number  of
quality-differentiated sites.   In the fishing data,  trips are available only by
region.   There are only  four of  these regions, and  each is large so there can
easily be  as much variation within any region M there in  among the regions.
Further,  164 of the  184 striped bass anglers in the sample visited only one
region.   Rather than  estimating four demand curves,  we have estimated a
single equation where  the dependent  variable  is the sum of the  trips  to all
sites.

    The handling of the quality variables  differs  al-   For recreational
boating, w.  ueed a varying Parameter  model because the quality variable,
scientific  measures of water quality,  varied  across sites  but not  across
individuals.   The quality variable in  recreational fishting, catch rate,  varies
across individuals.    Consequently, we  need not  use  a varying  parameter
model. Instead we uee the observation on  the reported catch in arena  where
the  individual  took his trips.  The data set includes many individuals  who did
not   fish   for striped bass.   For these individual,  costs and catch rates  were
inferred.

    The  fishing model estimated waa

(6.1)          x, =  f9 + ftTCi + 0aCH, + 03IB, + 04CB1 + 0SBD1


where xi is  the number of days  taken by  the i^ individual,  TCd is the
individual', full coat  (in dollars per  trip) of striped bass fishing,  CRi is the
catch rate (fish per day), IB  ^ andOB j_are  (0,1)  variables  denoting availability
of an inboard or outboard  boat for fishing,  respectively;  and BD1 is the
individual'.  fishing/hunting budget in dollar, per year.

    No eubatitute sites were specified in the model  because  the regions  were
so broadly defined  that they might not  in fact  act  as  substitutes for  each
other.    There  is probably  extensive substitution among sites within  each
region that cannot be captured at  all given the  level  ofaggregation we  face;
and  the sample data indicates that  only  about 10 percent  of  respondents
visited  more than  one  region.     Instead,  the price and catch rate  for
Participant who visited more than one  site were calculated as the  mean of
price, and catch rate, at each region visited, weighted by the day.  fished.

    Only  slightly more than one-quarter of the respondents who  either hunted
or fished in Maryland reported  having fished for  striped bass.   This  level  of


                                    95

-------
non-participation implies a serious censored variables problem.   There are
several  ways of handling this  problem  in recreational demand models.  We
explore  these approaches in Chapter 4 of Volume  I of this  report.   For the
current  task of  estimating fisheries demand models,  we  choose the following
simple Tobit  formulation:

                       (P'Z + e     P'z + « > 0
 (6-2)              x-  lo           ft'z + « * 0

where z is the vector of  explanatory variable..   The Tobit model imposes  come
rather extreme restrictions on individual  behavior that  more general sample
selection models avoid.    But for preliminary results,  we accept these
restrictions for the sake of simplicity.

    Determining the  relevant price and catch  rate  for non-participants was
problematic.    For  these individuals,  it was not  known which  of  the four
price-quality combinations  were moat  relevant to their  decisionto  go/not  go
striped bass  fishing.   In the application  we used the minimum price to access
a striped bass "site" and its corresponding  catch  rate.

    Welfare measures are calculated, in  principle,  the  "  same way as for the
varying parameter,  model.  That  is,  the benefits  of an increase in catch  rates
are given by  the change in consumer 'a • mplua which,  for the linear model
above, is


 (6.3,
                   -2ft,      -2ft,

where t\ is the own-price coefficient,  and  x  is the individual's  trip level.


Empirical Results

    The model in  equations  (6.1)  and (6.2) was estimated using the maximum
likelihood method of LIMDBP. Table  6.3 gives  the  results which  will be  used
for preliminary  benefit  estimation, along with  the sample  means  of the
variables.  The results in  Table 6.3  are fora model  in which actual catch
rates reported were  used for participant,  and  a predicted catch rate was
used for non -participants. We  also estimated a model  in  which predicted catch
rates were used for every individual In the latter estimation, the coefficient
estimates  remained basically  unchanged, but  the  standard error on  the catch
rate  coefficient increased resulting in  a t-statistic of about 1.3.

    The coefficient  estimates all have intuitively correct signs,  and they are
different from  zero at better than the  5 percent significance level Having  an
inboard motorboat  seems to induce more striped bass  tripa than having an
outboard motorboat.  The own-price  elasticity  for  Participant is about minus
one,  while the catch  rata elasticity for participant  is about .10,
                                     96

-------
                                  Table 6.3

           Tobit Estimation of the Demand  for  Striped Bass Fishing
Explanatory
Variable
Constant (C)
Own price ('IT)
Catch rate (CR)
Inboard Motor ( IB)
Outboard Motor (OB)
Budget (FHB)
** = 18.3
N= 760
Coefficient
Estimate
-10.6
-.336
.337
12.65
6.66
1.40


t-statistic
-5.79
-7.52
2.13
4.49
3.47
3.04


Mean of
Variable
1.00
$27.2
3.2 fish/day
.10
.31
. 70($000)


    We can use the estimated coefficients  in Table  6.3  to estimate welfare
effects of increases in  catch rates.  As  in Chapters  4  and  5,  two estimates  of
consumer surplus  are provided.   Method A employs predicted trips plus
changes  in predictions whereas Method B uses  actual trips  plus changes  in
predictions.

    It is  rather eaay to expand sample results to the population, since  the
Fish  and  Wildlife  Survey  includes sample weight  or sample expansion factors,
These weights account  for the fact  that different population strata  are
sampled disproportionately.   Consumer's • rplua  for  the population is simply
the weighted sum of the  surpluses  of the sample observation:
 (6.4)
Cs =
where s is the sample size and  f ( is the expansion factor.

    Table 6.4 gives  the  estimates of aggregate surplus.   The first column is
the estimate  of the value of access to striped baas fishing  as  it was perceived
in 1880, baaed  on  1980  prices.   The actual and predicted  estimates differ
substantially, with  the  actual being more than  three  time,  larger than the
predicted.
                                     97

-------
                                  Table 6.4

           Aggregate Consumers' Surplus for  Striped Bass Fishing:
                    Effect ofChanging Catch Rates,  1960

               (Population of Maryland Hunters and  Fishermen)
                      Aggregate
                      Consumer       Surplus  Change with Change in Catch Rate
                      Surplus        	
                      for Access     20% Decrease 10%  Increase 20% Increase
                      	   Thousands of Dollars  	

Predicted (Method A)      14,652          -572            314         1,501

Actual (Method B)         54,196          -422            231           4sl
    The second, third and fourth columns  in  Table 6.4 give the net  impact of
a  20 percent  reduction,  10  percent increase and 20 percent  increase,
respectively, in the striped baa.  catch  rate compared with  the level perceived
in 1960. Here the actual and predicted results are closer, especially for the
10 percent  changes.

    The numbers that are most interesting for environmental policy on  the
Chesapeake  are  found  in the third and  fourth columns.   These figures  are
rough estimate,  of  the dollar amount people who currently fish or hunt in
Maryland might gain  annually  from  improving  striped baas  fishing.

    There are a  number of  complicating factors  which cannot be integrated
into our preliminary calculations ofbenefit  estimates. First, consider  how  long
it would take for environmental policy  to  produce a • ubdantid, sustainable
increase in catch  rate.    Reduction in effluents for one year will have only a
small effect.  To improve  ambient water quality enough to  bring about better
striped bass reproduction and survival could take many  years.

    The second question  relates to the role of expectations regarding catch.
Aaide  from the likely bias and high noise in the catch  rate estimate, what
respondents  report  is  the ox post realisation of  catch rates,  while their
decisions  regarding whether, when,  and  how frequently to  go are baaed on
expectation about the  catch  rate,  ex ante.    Consequently, while  recalled ex
post  catch rate ds the  best quality variable  we could obtain for striped baas


                                     96

-------
fishing, we  need  to be skeptical  about its  implications  for  the  relationship
between days taken and expected catch rate.

    The second question concerning  these  benefit  estimates is whether, given
sufficiently improved  ambient  water  quality,  the catch  rates  are  sustainable.
The answer is  no.  Better catch rates induce more fishing and hence more
harvest.     Since there  is  some  evidence  that overharvesting is  partly
responsible for the decline in fish populations to  begin with, it  is likely  that
healthier stocks will  induce more  harvesting.   The  long run equilibrium will
result in higher than current benefits, but smaller than the benefits which
implicitly assume  that the increase in fishing effort will have no long run
effects on fish stocks.

    Last, it is  worth remembering that the benefit  estimates are baaed on a
sample of households that hunted or  fished in Maryland in 1980.  If there are
people who  currently do not hunt or fish, but  would go striped  bass fishing
if the  fishing improved sufficiently, then  the  annual  benefit  estimates  are an
underestimate.
                                      99

-------
                                 chapter 7

                                 Conclusions


     Restorationof   Chesapeake   Bay  water qulity rewires  substantial
resources on the part  of public  agencies, private firms and households.  There
are many choices  to be made in implementing  programs to  clean up the Bay.
This report haa described acme  of  the  activities which would benefit  from  the
enhancement of Bay  water quality.   Chapters 2 and 3 described ways in which
people  think  about  the Bay  and  benefit from better water quality. Chapters 4
through 6 contain descriptions of come recreational activities  which would gain
from  improved water  quality.     All of these chapters provide estimated
willingness  to  pay from potential  improvements.

     In deriving benefits,  sometimes  we 1000  eight ofthe  informational content
of  the models behind the benefit estimates,  the  estimated demand functions
themselves.   Chapters  2  through  6 contain substantial  new  information about
the structure of demand for recreational activities associated with  the
Chesapeake Bay.    In nearly every  instance  where sufficient data  were
available,  recreators responded to travel and  time costs in a manner consistent
with  our  theoretical  model.  They were also observed to be responsive to even
the crudest of water  quality measures.   Additionally, demographic  variables
such as  income,  race,  and boat  ownership were observed  to influence behavior.
As we  turn to the benefit estimates,  the reader is reminded  not to consider
the "bottom-he" benefit figures as  the only value  of  this  report.


Demand for Chesapeake Bay Recreational Activities

    The data and  modelling  exercises  described in  Chapters  3 through 6
provide a good picture  of the recreational  use of  the  Chesapeake Bay.
Chapter 3 includes  an overall picture  of Chesapeake  recreational  activities
derived from a random sample  of  all households  in the  Baltimore/Washington
SMSA's  (BWSMSA) .   This survey revealed  that  a  full  43 percent of the BWSMSA
population used the Bay or intended to  use it  for recreation in 19S4.
Geographical  distribution of users  showed Anne Arundel County  reaidenta  (69
percent) moat likely to be Bay users and District  of  Columbia residents  (21
percent) to be least likely  (ace Table 3.1).    The moat common recreational
activities were fishing, swimming and boating,  with about a third of  the  Bay
users  participating  in all three activities.   Of  these activities,  swimming  was
enjoyed by more people than  either  of the other two,  with 77  percent  of users
participating. In the remaining  chapters,  each of these  activities was looked
at in  greater detail using specific surveys of subsamples of the population.

     In Chaptar 4,  wa provide two types  of demand models for western shore
beach uae activity.    Each draw,  on  an on-site  sample of beach users  at
western shore beaches  in  the summer  of 19S4.  The varying parameter model
is a modification of traditional  demand models where the demands  for tripe to
each cite are treated largely independently, but the difference  in parameters
across  sites are attributed in  part to  site characteristic.  The discrete choice
model  explains the  choice  among  cites  directly,  as a  function of  site
characteristics,  but does  not handle the total number of trips well.  Each type
ofmodel gives a good description  of  one  aspect  ofthe recreational decision.
                                     100

-------
     From the results in  Chapter 4 it  is char that both money and time
access  costs are extremely important in  determining demand for trips to any
given beach,  as are  the costs of  accessing alternative beaches.   Whether  or
not an  individual  owns a boat or  recreational  vehicle also affects  demand fora
subset of  beaches, those which have facilities  for these  capital goods.   Demand
functions  for  trips  to a site tend to pivot  inward,  becoming more elastic, with
declining  water quality.

     The results of the nested multinominal  logit or discrete choice model  of
beach use suggested acme  similar and  acme  additional characteristics  of
demand  for this activity.   Once again,  money  and time costs of  access were
important  this time in explaining the choice among sites.   Additionally,  the
availability of  boating and  recreational vehicle facilities increased   the
likelihood of a boat or recreational  vehicle owner  to choose  a site.  An
interesting hypothesis was tested regarding the differential substitutability
among local beaches and among state  beaches vi«-A-vi«  the substitutability
between  local  and state beaches.    Beach users seemed  to consider local
beaches  closer substitutes for one  another  than for state beaches.  Individual
with larger parties or families were more likely to attend state beaches where
a variety of activities were  available.    The  longer an  individual had attended
western shore beaches, the  more likely he was to use local rather than  state
beaches.

     Chapter 5 provides a  rather extensive profile of boaters and boat  owners
derived  from a survey of  boaters sponsored  by Maryland Sea Grant and
Maryland Coastal Zone  Management and  from the BWSMSA  telephone  survey.
The boater  survey subsample includes registered boat owners  in Maryland.
The profile includes  an  analysis of  characteristics which distinguish boat
owners  from others  and looks at these distinguishing characteristics  b y
geographical area. Average household income, for example,  is higher for boat
owners than non-owners, but this difference is only  • tatiatically  significant  in
Prince  Georges, Anne Arundel  and Calvert  counties.

     Considering the boats  themselves, a  different  profile characterizes those
which are kept  in  the water  all season  (in marinas, moored, etc.  ) than
characterizes boats which are trailered. As  would  be  expected, trailered  boats
are significantly  smaller and less valuable,  they are more  likely to  be
runabouts or workboats and their owners are likely  to  have less income than
the owners of boats kept in  the  water.    Almost all  trailered boats were used
for fishing at least Occasionally.   About three-quarters  of the non-trailered
boats  were used  for swimming at least Occasionally.

     Table 5.4  summarizes  the boat  owners survey by county  of  residence,
revealing more about the  geographical distribution of  Bay users.   Residents  of
Baltimore  and  Anne Arundel  counties accounted for 39 pecent of  the trailered
boats  and 45 percent  of  the non-trailered  boats with Prince  Georges  County
and Montgomery County residents  accounting for another  20 percent of
trailered and 12 percent of non-trailered boats.

     The last of  the descriptive information suggests the  importance of water
quality to boaters.   Water quality was considered either  moderately  or very
important  in the selection of a  boating area by 75 percent  of  the trailered
boat  owners and by 76 percent  of the  non-trailered boat owners,
                                     101

-------
     A varying parameters model similar to the one used in Chapter  4  revealed
that trailered boat owners' demand for trips from launch sites was affected by
access  costs to  the launch site and  costs of accessing alternative sites.  In
general, the demand for trips to any given site was  Positively affected by the
value of  the  boat; i.e.,  individuals with more valuable boats took more trips.
The demand function fo Bny given  site tended to pivot  inward and become
more elastic as water quality declined.

     Owners of boats kept in the water do not choose launch sites when they
take a  trip,  and consequently we  have no way of  knowing where  they  boat.
As  a  result we cannot  model  their  decisions in  responnse to varying  water
quality.    For  these  individuals?  simple demand functions were estimated.
Factors which significantly affected their demand for boating trip,  included
the coat  of a trip  (negatively) and the value of the  boat (positively).
Additionally it was determined  that  sailboat  owners tend  to take fewer trips
and their  demand for trips ds more  price  inelastic.    Finally, boat owners who
fish while boating  tend  to  demand more trips and  their demand  tend,  to be
more price elastic.

     In Chapter  6 information about •  portfishing on the Bay  is presented.
Estimates  of sportfishing activity vary by  data source and range from 539,000
to  900,000  anglers  in  19SO and from 2.7 million to 4.1 million trips for that
came year.  The two  prominent sources efinformation  on sportfishing are the
U.  S.  Fish and Wildlife Hunting and Fishing Survey  and  the U.  S.  National
Marine Fisheries Survey.

     Our analysis in this chapter concentrated on •  tripod  bass  fishing since
this was the only  species important  to  Chesapeake  recreational  fishing for
which Sufficiently detailed data existed.  One source (U. S.  Fish and  Wildlife)
reports that in  1960, 239,000  anglers fished  for  striped bass in  Maryland and
Sussex County,  Delaware  and fished  2.1 million days  in  total.     Table 6.1
presents come descriptive statistics of striped  bass fishermen and  other
Chesapeake Bay fishermen.

     In the analytical  section of  Chapter 6,  demand  for sportfishing trips was
modelled  as  a function of  the  individual'. trip costs, catch rates,  his  annual
fishing/hunting budget  and indices of types of boat  ownership. All variables
affected  the demand for  trip,  in  the  expected  direction, with  owners of
inboard  motorboat  likely to take  more trips  than those with outboard
(presumably smaller)  motorboat.

3_ "? —  of Benefits  from Water Quality Improvements

     While the  analysis of the demand  for recreational  activities  is worthwhile
in  ita own right, more information about the size of  rewards  from Bay
restoration can be obtained,     There are  several reasons for computing
aggregate  willingness   to pay rather than simply providing descriptive
measures such as  recreational use  days.   Obviously  such measures cannot be
compared to the  costs of  restoration;  they cannot even be added across
activities.   A day of fishing is different from a day  of swimming,  and changes
in water quality  have different effects on the benefits derived from the two
activities.   Further, as we observed  in Chapter 3,  there is  some willingness to
pay for clean water by  people who  do not use  the  Bay. If  we limit ourselves
to descriptive measures such  as user days, we ignore the  returns to  people
                                    102

-------
who value cleaner water  but  do not uae it.   Consequently,  we have taken  a
first step toward the  logical,  albeit venturesome,  task of estimating the
aggregate benefits of improving  the Bay's water quality.

Caveats

     The aggregation  of  benefits across activities and  for  the population at
large is venturesome because it is so filled with known difficulties. We can
take a systematic view of these potential errors by recalling the links between
environmental policies designed to  reduce effluent pollution and the benefits
of environmental improvements.   Policies  influence effluents  directly through
regulations  and indirectly  through changes in incentives.   Reductions in
effluents will  eventually improve the ambient water quality.   Improvements in
ambient quality when perceived by  individuals eventually lead to  changes in
behavior toward the  Bay,  implying  benefits.   Further, when non-users per-
ceive  improvements  in the ambient water quality, they too will be  better off.
There is potential for errors in the measurement  of  each  link  in this process.

     The analysis  of  the previous  chapters has  concentrated on the  connection
between ambient quality and economic benefits.    It rests,  however, on the
relationship  between environmental policy,  effluents, and ambient quality.  The
considerable debate regarding the connection between effluents and ambient
quality suggests  the potential  for honest  differences of opinion on the nature
of the ecological links. Similar  uncertainty over the behavioral and perception
links  exists.

     While a complete catalog  of the sources of potential error would take an
entire  chapter,  we describe  broadly what we think the major difficulties are.
If the  problems inherent  in explaining  the link between  policy and ambient
quality are ignored, the foremost uncertainty is  between ambient quality and
behavior. Recall briefly  how  this  link waa estimated. For  boating and beach
uae we used a  varying parameters model to  estimate  the  relationship between
the product  of total  phosphorus and nitrogen readings  in  1977  and  trips in
19S4. There  is clearly  substantial room for error in this relationship.

     First, since people cannot perceive nitrogen  and phosphorus,  we must
assume that  the nitrogen and phosphorus are approximate measures of the
ambient  quality.  It is not  unreasonable to expect such a relationship to  hold
in principle.   Chapter  2  describes ways  in which individuals form perceptions
of water quality.   Some  of the  deductive and media-baaed means  by which
individuals form quality perceptions may be directly related to effluent
discharges.  Others,  such as  stimulants of sensory perceptions, may  be  highly
correlated with, or even caused  by,  nitrogen and phosphorus  levels.   Previous
studies which  have  attempted  to link behavior to individual ambient water
quality   indicator.    (e.g.  Binckley  and Hanemann)   have  detected   a
correspondence.   Chapter  2  describes  acme evidence  which supports  this
hypothesized link derived  from  our telephone  survey  of  the BWSMSA and  the
field survey  of western shore  beaches.   Through the telephone a  significant
relationship waa detected between a  household perception of the water quality
in the Bay and  ita likelihood  to quit using the Bay. Additionally,  a  significant
relationship appeared between objective  measures  ofthe Bay's water quality
over time and  the proportion  of  households who atopped  using  the Bay for
recreation because  they perceived the Bay's water  quality to  be unacceptable.
Finally, the  user  (field )  survey showed a positive correlation  between
                                     103

-------
measures of  fecal coliform  at each of  nine  beaches and the  proportion  of
households  that found each  beach  unacceptable.

     A further   difficulty is   the   seven   years   which   separate   the
nitrogen/phosphorus  readings and the  recreational  behavior, since  1977 _ was
the latest year  for which  complete information was  available.    While  this is
clearly a source of  potential error,  there are a few reasons why it might not
be as  bad as  it seems.   For one thing, the relative levels across different
regions of the Bay may have  remained  approximately constant even if absolute
levels have changed.  Additionally, it is not clear what  year  or combinations of
years would  be correct in signaling the recreational behavior  stimulated by
water quality   because behavior is probably largely affected by prior
experiences.

     Since we  are really explaining choices among sites of different quality,
our behavioral models  depend more on the relative  levels  of  ambient water
quality rather than on absolute  levels; and  if relative levels have remained
fairly constant,  our  behavioral models are likely  to  be  quite good. Extracting
benefit measures from these models, however,  must be  done with  caution  since
the absolute  levels of  nitrogen  and phosphorus readings  used may not be
trustworthy y.

     For recreational  fishing the problem is in some waya a  little simpler.
Here we use  the catch rate experienced by the individual for  1980, the  year
the  trips were  taken.  There is  of course  a  complex and uncertain chain of
relationships between improvements in  ambient quality and  growth in the
density of  fish stocks.    There is further uncertainty in the  connection
between  fish  stocks and catch rates.    These are largely,  although  not
completely,  problems of biology and  are not addressed here,  but nonetheless
remain as  imperfectly understood links in the  system.

     Restricting our comments entirely to the behavioral realm does  not
eliminate these uncertainties and potential sources of modelling error.   In what
sense is the  catch rate in the  year the trip,  were taken a good measure of
quality? Fishermen may value higher  catch rates  but their  demand (behavior)
fortripa  this  year  may be baaed on catch  rates experienced in previous
years.  When  the quality of  the good  is uncertain to the consumer,  there may
be one eat of  quality indicators  that stimulate demand and another which
affect the  benefits derived from consumption.   Further, there is no guarantee
that catch rate is the only (or moat  important) variable which  determines the
enjoyment of trips to  catch  fish.   For example,  catching one five-pound
striped baaa may be  batter than  catching two two-pound stripers.

     In addition to  the severe  difficulties in inferring the  relationships
between ambient quality,  there are two other  significant sources of  error in
computing aggregate  benefits.   First,  there is the problem of sampling and
non-sampling error  associated with the measurement  of the number of  trips
per participant and the number of participant  in each  activity, M well as
measurements  of exogenous variables  such  as  costs  per trip.    The boating
survey is a  good example of non-sampling error for trips  This survey waa a
mail  survey, so in a sense the respondents are  volunteers. The  return rate
was 70 percent.   We have no way of  knowing  whether those who  competed
their questionnaire,  were representative  of the boating population as a whole
or if there is a built-in sample selection  bias.


                                     104

-------
     We have also  used only  segments  of the total population in our  analysis
of  benefits.   The  boaters were limited to those  who  trailer their boats,  the
fishermen to those who fish for  striped bass and  the beach _ users to _those
who use public-accees  western shore beaches.   In  the  boating  and  fishing
analysis we have  excluded non-Maryland households.    In the  contingent
valuation and beach uae analysis,  only  20 percent  of Virginia's  population  wae
included  and about 80  percent of Maryland's  households.  In every  instance,  a
major  portion of users is  excluded so any estimates derived will be lower
bounda.

     Another  source of error in  aggregating benefits across activities is
aggregation  bias.    This  comes  in  two forms:     simple doublecounting  and
conceptual aggregation bias.    Doublecounting occurs because a 0ubatantial
number  of boaters also fish, and  many  fishermen have boats.   The conceptual
aggregation biaa occurs because  of the  jointness  of choice  among  sites for  a
given activity  and among activities.  For example, the choice  of  visiting Sandy
Point  versus  Point  Lookout may depend  in part on water quality.  Enhancing
water  quality at both  sites may only increase attendance at one sites  making
the addition of  benefits across sites incorrect. A discussion of  this  problem is
offered in Chapter 3, but both forms  of  aggregation bias are treated in detail
in Chapter 5 of the conceptual volume  of this report.

     Finally, we must remember that we  have  only  three activities: boating,
fishing, and swimming.   There  are many other  recreational  and commercial  uses
of  the Bay whose value is enhanced by cleaner water.    For example,  our
analysis of  fishing rovers only  striped bass;  fishing  for species besides
striped bass  (e.g.  crabbing)  is widespread and  not  covered by our  analysis.
And our  analysis of the effect of  changes in water  quality  covers only
trailered boats, not boats at marinas.   Many other,  especially more  casual,
activities are omitted.   We have limited our analysis  to boating, fishing,  and
swimming because we could obtain data of adequate quality only for these
activities

Estimates

     With these  difficulties  firmly in mind, we are prepared to hazard  some
judgments on the magnitude of the aggregate benefits  of improving the Bay's
water quality.   Cur approach  is to present  low,  middle and high  benefits  for
the beach use  (Chapter 4),  boating (Chapter  5),  and  fishing (Chapter 6)  and
qualitatively compare those benefits with the total benefits  derived  from
Chapter 3.  Comparing the ranges of  these  independent sources of benefits
will help us  to form a judgment, but nothing  more, of the magnitude of
aggregate benefits.

     Chapters  4  through 6 give benefit estimates for activities conditioned on
the  computational method and the proportionate  change in ambient quality  and
catch  rate.  We  adopt  the  convention  of analyzing  a  20 percent reduction in
nitrogen  and phosphorus for boating and beach  use  and a  20  percent increase
in  the catch rata for striped bass  fishing.    These changes should be
interpreted loosely  as considerable  improvements  in the quality of the  Bay
without attaching much  significance to  the  absolute change in ambient
readings which would be  implied.   In particular, one should not interpret the
estimated effect of nitrogen and phosphorus  as an  "all else equal" effect.  The
change in nitrogen  and phosphorus is a proxy for  changes in  moat ambient
                                    105

-------
determinantts  of  water quality so that  the  implicit  assumption  is that a range
of ambient  factors may be  improving.   Further, to counteract the problem of
aggregating  across sites  for a given  activity,  we  select  as a pessimistic
estimate the  lowest estimate of  the benefits of improving  the quality by  20
percent at the one moat important  site.

     Table 7.1  summarizes  some of  the  estimates of aggregate benefita  for our
groups of boaters,   sportfishermen and beach users,   translated into 1987
dollars.   The variation  from pessimistic  to optimistic  is  provided  by  two
sources:   variation  induced by the  method of calculating benefits  (i.e., using
actual trips versus predicted  trips)  and variation caused  by  choosing one site
rather than  the sum  over  all sites. Recall  that because each  site's benefits
are calculated assuming other sites' quality remains unchanged aggregating
these measures  over  sites  will  produce  an upwardly  biased  aggregate benefit.
The pessimistic estimates for  beach use and for boating are the lower  of  the
two estimates  of  the benefits  for a  20 percent  improvement  in water  quality
from  Sand y Point for beach uae  and  Anne  Arundel  County for boating.  One
site  was  chosen as  a lower bound because wtth only one site all (upward)
aggregation  bias is eliminated.    The  average  estimates  for beach use and
boating are  the lower of  the two calculation methods for sums across all sites.
The optimistic estimates are the  higher  of the two calculation methods  for the
sums  across  all sites.  For  striped bass fishing,  the pessimistic  estimate is  the
lower of the  two methods  of  calculation.    The  sites have  already been
aggregated  for the fishing  case,  and  as  we  show in Chapter 5  of the
accompanying volume, the nature of the aggregation bias in  this case is  not
obvious.  The  optimistic estimate is the  higher of the two  calculation arithmetic
methods and  the average is the mean  of the pessimistic and  optimistic.
                                  Table  7.1
         Aggregate Benefits for Three  Water-related Activities  from a
           "20%"  Improvement in the Chesapeake Bay's Water Quality
                               in 1967 dollars
Benefit Estimate
Activity
Public Western Shore Beach Usea
Boating with Trailered Boat"
Striped Bass Sportf ishingc
Pessimistic

16, 853
664
664
"Averaqe"
, . .
1 (? Thousand) "
34,658
4,717
1,366
(Mimistic

44,960
8,129
2,071
• From Table 4.6. Pessimistic estimate is the  Method B value for Sandy Point,
 the average  is the sum of Method B values over all ten sites,  and the
 optimistic is the sum of  Method A values over all sites.
bFrom Table  5.13.    All  per  boater  estimates expanded to 60,000 boaters
 trailering boats.   Pessimistic estimate  is the low  value  (Method A)  for Anne
 Arundel  County,  the  average  estimate in the sum of low values  (Method A)
 across all counties and the optimistic  value is the sum of high values (Method
 B)  across  all  counties.
GFrom Table 6.4.  Pessimistic value is the value using Method B, the  "average"
 value is the average of the pessimistic and  optimistic value,  the  optimistic
 value is the value using Method A.
                                     106

-------
     The   aggregate   meaaurea  of willingness  to  pay for water quality
improvements are revealing  for several  reasons.   First,  regardless of which
benefit measure we use  (pessimistic, average, or  optimistic), the returns  to
beach use are  the greatest.   This ia primarily because a  larger proportion of
the population  engages  in come beach-going during the year than boating  Or
fishing.  Additionally,  this group may be more  sensitive  to  changes  in water
quality than the boating-fishing group.

     A second  interesting implication of the results, although not obvious from
looking at Table 7.1,  is the importance of regional variation in water quality.
If we were able to clean up the water around Anne Arundel  County  only,  we
would still go  a long way  towards  satisfying   some of the  human  needs for
using the Bay.   While we realize that confining a  water qualtty improvement
program to a particular locality may not be technically or ecological y feasible,
any clean-up strategies which result in  significant improvements in this region
of the  Bay will yield substantial  benefits.

     A comparison  of the  behaviorally based measures of benefits  presented  in
Table 7.1  with benefit estimates  derived from contingent  valuation (ace Table
7.2)  is interesting  even  though the valuation questions driving the  two
analyses  are different.  All  of  the estimates in Table 7,1  are partial estimates
in that  they account  for only one activity and involve  only fubaeta  of  the
population.  Table 7.2 presents  contingent valuation produced benefit estimates
associated with  a  broader but less precise  hypothetical  improvement:
improving water quality to an "acceptable"  level.    The  subset  of the
population includes those in the BWSMSA who found water  quality unacceptable
for swimming or related  uses.
                                  Table 7.2

            Aggregate" Benefits  from Water  Quality Improvements-
                             Contingent  Valuation
                                in 1984 dollars

                                Willingness to Pav  for  Improved  Water Quality"
Group                            Pessimistic0      Average0       Optimistic0
User
Non-User

47,254
18,446

67,582
23.556

87,870
28.733
  Total                            65,700            91,137           116,603

 •Population is the Washington,  D.  C.  and  Baltimore SMSA's
 Willingness to    acceept tax increase to raise  Chesapeake Bay Water quality  from
 a level unacceptable for  swimming end/or other related activities to a level
 acceptable fo
 The average will
 See Table 3.8.
 acceptable  for swimming.
GTheaveragewillingnesstopay plus or minus one standard error in est imate.
                                      107

-------
     The numbers in Tables 701 and 7.2 give likely magnitudes for the annual
benefits of dirproving water quality in the  Bay.   The numbers  suggest a range
of from $10  million to over $100 million,   We know  that there  are  numerous
random elements  in all estimates.   Further,  we know that several activities and
populations have been omitted.    But based on these numbers,  it seems
plausible to estimate that the annual returns to cleaning up the Chesapeake
are at least  of this order of magnitude.   We have only the evidence  presented
herein to make this judgment.

     In conclusion, we recapitulate the premise.   Society  haa undertaken an
investment  program.    The  nature  of the program is the  cleanup  of the
Chesapeake  Bay.    The costs  of  the program include  such  things  as sewage
treatment plants,  funding of government  programs  to regulate and monitor
agricultural effluents,   installation of  industrial waste disposal  systems,
restrictions on housing  development, etc.    The annual returns on  the
investment  program are measured by what  people are willing  to pay for the
improved services  of  the Bay.  TMa  is the dividend yielded  by  the  public's
investment program.

     For several  reasons,  we think that the long-run  benefits  are higher  than
the figures Tables  7.1  and 7.2 indicate.   First, as people learn  that the  Bay
haa become  clamor, they will adjust their preferences toward  Bay  recreation.
This  is especially true of people who do  not currently use the Bay and are
largely  excluded from the analyaia.   Second,  the population  and income of the
area  have grown since 1984,  and both are  likely  to grow more, increasing the
demand for  and value ofimprovements in  water  quality.   Finally, we have
ignored the value  (both uae and existence value)  which households outaide the
BWSMSA may  have for the Bay.  The  Chesapeake  Bay  is a nationally prominent
resource.  Its  improved health is of value  to many who will never use it.
                                    10s

-------
                                REFERENCES
Bell,  F. W. 1968.    "The Pope  and the Price of Fish.  "   American Economic
     Review 58142-147.

Binkley,  Clark S.  and W.  Michael Hanemann.    1978. The  Recreation Benefits of
     Water  Quality  Improvements;   An  Analysis of Day  Trips  in  an Urban
     Setting. Washington, B.C.  E. P.A.-600/5-78-010.

Bishop,  R.C. and T.A.  Heberlein.   "Simulated Markets,  Hypothetical  Markets and
     Travel  Cost Analysis:   Alternative Method, of Estimating Outdoor
     Recreation Demand."   Working paper  no.  187.

Bishop,  R. C. and  K.  Sample.    1886.   "Estimating  the  Value of Variations in
     Anglers'  Success Rate.:   An Application of the  Multiple-Site Travel Cost
     Method."  Marine Resource Economics 2:1:55-74

Blackburn, A.J.  1870.   "An Alternative Approach  to  Aggregation and  Estimation
     in  the Non-Linear Model."  Ch. 9  in The Damand for Travel;  Theory  and
     Measurement. Edited by  R E.  Quandt.  B.C.  Heath,  Lexington.

Bockstael,  Nancy,  W.  Michael  Hanemann and Catherine Kling.  1887.  "Modeling
     Recreational  Demand in a Multiple Site Framework."   Water  Resources
     Research. Forthcoming.

Bockstael,  Nancy, W.  Michael Banemann and  Ivar  Strand.  18SS.  Measuring  the
     Benefits  of  Water Quality Improvments   Using  Itocreaticn Demand Models.
     Report   presented   to    the    B.P.A.    under   Cooperative Agreement
     CR-81 143-01-1, Volume II.

Bockstael,  N.  E., and K.  E. McCormell. 1981.   "Theory  and Estimation  of  the
     Household  Production Function for Wildlife Recreation."     Journal of
     Environmental Economics and Management 8:188-214.

Bockstael,  N. E.,  and  K.  E.  McConnell.    1983.   "Welfare Measurement  in  the
     Household  Production Framework." American Economic Review  73:806-814.

Bockstael, N. B.,   and J.  J.  Opaluch.   1983.    "Discrete Modelling of  Supply
     Response under Uncertainty:    The Case of  the  Fishery."    Journal of
     Environmental Economics and Management  10:125-137.

Bockstael,  N.  B., I.  B. Strand,  and W. M.  Hanemann.    1987.    "Time and  The
     Recreational Demand Model." American  Journal of Agricultural   Economics
     69213-302

Bockstael,  N.  E.  and  I. E.  Strand.  1987.   "The  Effect of Common  Sources of
     Regression Error  on Benefit Estimates." Land  Economics 63:11-20.
                                   109

-------
Bocks tael, N. E.,  I.  E.  Strand,  and A. Graefe.   Economic Analysis of  Artificial
     Reefs:   A Pilot Study  of Selected Valuation Methodologies.   Technical
     Report  #6.   Artificial  Reef Development  Center,  Sportfishing  Institute.
     1986.

Brookshire,  D.S.,   L.S.  Eubanks and A.  Randall.  .  1983.    "Estimating Option
     Prices and Existing Values  for Wildlife Resources.  "    Land Economics
     58:1-15.

Brown, Gard  nor  and  Robert  Mendelsohn.    1984.    "The  Hedonic Travel Cost
     Method." Review of Economics  and  Statistics 66:427-33.

Brown, W. G., C.  Sorhus, B.  Chow-Yang, S.  Richards.  1883.  "Using Individual
     Observations to Estimate  Recreation Demand Functions."  " American Journal
     of Agricultural Economics  65:154-157.

Brown, W. G.,  C. N. Sorhus,  and D.  A.  Meyer.  1882.   "Estimation of  Oregon
     Sport-Caught  Salmon and Steelhead Values Via the Travel Cost  Method."
     Mimeo, ASSA Meetings,   25 ppa.

Brown, W. G., A. Singh  and  E. N. Castle.    "Net Economic  Value  of the Oregon
     Salmon-Steelhead Sport Fishery."     Journal  of  Wildlife   Management
     29(2):266-279.

Burt,  Oscar  and Durward Brewer.   1872.   "Estimation of Net Social Benefits
     From Outdoor Recreation."  Econometrica 39:813.

Carson, R.  T.  and  R C. Mitchell   188sO   "The Value of Clean Water:    The
     Public'. Willingness to Pay for Beatable,  Fishable,   and  Swimmable Quality
     Water."    Discussion paper No.  QE85-08 revised,  (Washington,  D. C.,
     Resources for the  Future).

Castle, Emery N.  and W.  G.  Brown.  1884.    "The Economic Value of  a
     Recreational  Resource:   A Case Study  of  the  Oregon  Salmon-Steelhead
     Sport Fishery.  Paper presented before the Committee  on  the  Economics
     of Water Resource* Development of the Western Agricultural Economics
     Research  Council,  San  Francisco,  California.

Caulkins,  Peter,  Richard  Bishop and Nicolass Bouwes.  1982.  "The Travel Coat
     Model  for   Lake  Recreation  A  Comparison of  Two  Methods   for
     Incorporating  Site Quality and Substitution Effects." Mimeo, ABA annual
     meetings,

Caulkins,  Peter P.,  Richard  C. Bishop  and Nicolass W.  Bouwes,  Sr. 1888. "The
     Travel  Cost  Model  for  Lake  Recreation: A  Comparison of  Two Methods  for
     Incorporating  Site Quality and Substitution Effects." Amer.  J.  Air. Econ.
     68(2):291-297.

Caulkins,  Peter,  Richard  Bishop and  Nicolass Bouwes.     1985.    "Omitted
     Cross-Price Variable  Biases  in the Linear Travel Coat Model:  Correcting
     Common Misperceptions.   " Land Economics  61:182-85.


                                   110

-------
&aario,  Frank J.     1969.    "Operations  Research in  Outdoor Recreation.  "
     Journal of  Leisure Research 1:33-52.

Cesario,  Frank J.   1973.   "A Generalized  Trip  Distribution Model." Journal  of
     Regional Science  13:233-48.

Cesario,  Frank J.   1974.    "More  on the Generalized Trip Distribution Model,"
     Journal of  Regional Science 14:389-97

Cesario,  Frank J.  and Jack Knetach.   1976.    "A Recreation Site  Demand and
     Benefit Estimation Model. "  Regional Studies 10:97-104.

Christensen, L. R.,  Jorgenson, D.W.  and Lawrence Lau.   1975.  "Transcendental
     Logarithmic Utility Functions."   American Economic Review 65:367-83.

Cicchetti,  Charles, Anthony Fisher and V. Kerry Smith. 1976.  "An  Econometric
     Evaluation of  a Generalized  Consumer  Surplus Measure: The Mineral King
     Controversy."  Econometrics 44;1268-76.

Cicchetti,  C.J., J.J. Seneca and Paul   Davidson.  1869.    The Demand and Supply
     of  Outdoor  Recreation:  An Econometric Analysis.  Rutgers. 1869.

Clawson,  MartOn.   1859.   Methods for  Measuring the Demand  for and Value of
     Outdoor Recreation.  Resources for the Future, Inc.,   Reprint no. 10.

Clawson,  Marion and Jaok  Knetach.  1866.   Economics  of Outdoor Recreation.
     Resources for the Future, Washington D.C.

Cox,  D.R.  1861.   "Tests of Separate Families of Hypotheses." Proceedings of
     the Fourth Berkeley  Symposium on Mathematical Statistics and Probability.
     Vol  1.  University of California Press.

Cronin, F. J.  1882.   "Valuing Non-Market  Goods Through Contingent Markets.  "
     Rep. PNL-42S5.  Pac. Northw.  Lab., Richland, Wash.

Cropper,  M.  L.,  L.B. Deck, K.B.  McConnell and T.T. Phipps.  1886. "Should the
     Rosen  Model be Used to Value  Environmental  Amenities?"     Papar
     presented at the conference on  the  Economics of Chesapeake Bay
     Management 11,  Annapolis, MD.

Cummings,R G. ,  D.  S.   Brookshire  and  W. D. Schulze.    1984.  valuing
     Environmental  Goods: A State of the  Arts Assessmentof the Contingent
     Valuation Method.   Draft report to  EPA. Universities of New Mexico and
     Wyoming.    "

Daubert,  J.  T. and R A.  Young.  1981.   "Recreational  Demands  for Maintaining
     Instream Flows:  A  Contingent Vaulation  Approach." Amer. J. Apr.  Econ.
     63(4):666-676.

Beaton, A.S.  1974,   "The Analysis  of Consumer Demand  in the United Kingdom,
     1800-1970."  Econometrics 42:341-67.
                                    Ill

-------
Beaton,  Henry.  1978.   "Specification and Testing in Applied Demand Analysis.  "
     Economic Journal 88:524-36.

Desvousges,  W.  H. , V.  K. Smith and  M.  P.  McGivney.   1983.   A Comparison of
     Alternative  Approaches  for Estimation of  Recreation and RelatedBenefits
     of Water  Quality   Improvements.     Washington,  D.  C.:     Environmental
     Protection Agency,  Office of Policy Analysis.

Deuel, D. G.   1978.  "National Marine Recreational Fishery Statistics Program.  "
     Paper  presented  at Canadian Sport Fisheries Conference,  New Brunswick,
     Canada.

Domencich,  Thomas  and  Daniel McFadden.    1975.    Urban Travel Demand;  A
     Behavioral Analysis. North-Holland,  Amsterdam.

Dwyer,  J.   F.  and M.  D.  Bowes.    1978.    "Concepts of Value for  Marine
     Recreational Fishing."   American Journal of  Agricultural  Economics 60(5);
     1008-10120

Feenberg,  Daniel  and Edwin  Mills.  1980.  Measuring the  Benefits  of Water
     Pollution Abatement. New York,  Academic Press.

Fisher,  Ann and William H.  Desvousges.  1881.  "Alternative Contingent Market
     Approaches to Valuation  of Recreation  and Related  Benefits of Improved
     Water  Quality."   Paper presented at  the  Eastern Economics Association
     Meetings,  Washington B.C.

Fisher, Am and  Robert Raucher. 1982.    "Comparison of Alternative Methods of
     Evaluating the Intrinsic  Benefits  of  Improved  Water Qualiy."    Paper
     presented at  ASSA meetings, New York City.

Fisher,  F. and  K.  Shell.  1871.  "Taste  and Quality Change in the  Pure Theory
     ofthe True Cost-of-Living  Index." in Price  Indexes and  Quality Change.
     Edited  by Zvi Griliches.   Harvard University Press, Massachusetts.

Freeman,   A. Myrick III.  1979. The  Benefits  of  Environmental  Improvement:
     Theory and Practice. Johns Hopkins Press for  Resources for  the Future,
     Inc., Baltimore. 1979.

Freeman, A. Myrick  III.  1882.  Air and  Water Pollution Control:  A Benefit-Coat
     Assessment. John Wiley and Sons.

Freund,  R. J.  and  R R Wilson.   1974.    "AnExample  of a Gravity Model  to
     Estimate Recreation Travel." Journal of  Leisure  Research  6:241-56.

German, W.M.   1976.    "Tricks with Utility Function."    in Essays in  Econmic
     Analysis: Proceddings of the 1975  AUTE Conference. Sheffeld. Edited  by
     M.J. Artis and A.R.  Nobay.  Cambridge University Press,  Cambridge.
                                    112

-------
Greenley,  Douglas A., Richard  G.  Walsh  and Robert A.  Young.   1981.   "Option
     Value:   Empirical  Evidence from a Case Study of Recreation and Water
     Quality."   Ouartelv J.  Econ.  96(41:657-673.

Gum, R.L. and  W.E. Martin.   1975.   "Problems and Solutions  in  Estimating  the
     Demand for and Value of  Rural Outdoor Recreation. "  American Journal of
     Agricultural Economics 57:558-66.

Hanemann, W. Michael.    1980.    "Quality  Changes, Consumer's  Surplus,  and
     Hedonic Price Indices."   California Experiment Station working paper
     number 116.

Hanemann,  W. Michael.    1981.   "Quality and  Demand Analysis."^.  3 in  New
     Directions  in Econometric  Modeling  and Forecasting  in U.S.  Agriculture.
     Edited by  Gordon  Rausser, Elsevier/North-Holland,  New York.

Hanemann,  W. Michael.     1882.    "Applied Welfare Analysis  with  Qualitative
     Response Models."    California  Agricultural  Experiment Station working
     paper no. 241.

Hanemann,  W. Michael.    1984a.   "Multiple Site Demand:  ARew'ewofExisting
     Models  and New Development",  Chapter  9  in Measuring  the Benefits of
     Water  Quality  Improvements  Using  Recreation  Demand  Mod«=3g-    Draft
     Report presented to E.P.A. under  Cooperative Agreement CR-80043-01-0,
     VOL  II.

Hanemann,  W. Michael.     1984b.    "Discrete/Continuous Models of  Consumer
     Demand." Econometrics 52:541-61.

Hanemann,  W. M.   1978.   "A Methodological and Empirical Study of Recreation
     Benefits  from  Water Quality  Improvement."  Ph.  D.  dissertation,
     Department  of Economics, Harvard University.

Hanemann,  W. M.  1984.    "Welfare Evaluations with Market Data." American
     Journal of  Agricultural Economics. 66:332-41.

Hanemann,  W. M.    1984.    "Welfare Evaluation in Contingent Valuation
     Experiment with  Discrete Responses." American Journal of Agricultural
     Economics.  VOL  66(3) :332-341.

Harmon,  R.  J.  and Associates.   1883.  Recreational Resources at  Risk.  Taak 2
     Working Paper.  Washington,  DC.

Harrison,  D. and D. Rubinfeld. 1878.   "Hedonic Housing  Prices  and the Demand
     for  Clean Air."    Journal  of  Environmental  Economics and Management
     5:81-102.

Hausman,   J.A.    1886.     "Exact Consumer's  Surplus  and Deadweight Loss."
     American  Economic Review  71:662-76.
                                   113

-------
Heintz,  H.T. Jr., A.  Hershaft  and G.C.  Horak.  1976.  National Damages  of Air
     and Water Pollution.  A Report to the EPA.  Enviro Control,  Inc.  Rockville,
     MD.

Hiett, R  L.  and J. W.  Worrall.  1977.   "Marine  Recreational Fisheries Ability to
     Estimate Catch and to  Recall Catch and Effort over  Time."    Human
     Sciences  Research,  Inc., Report HSR-RR-77/18,  Washington, D.C.

Horowitz, Joel.    1983.    "Statistical Comparison  ofNon-nested Probabilistic
     Discrete Choice Models.  "   Transportation Science 17:319-50.

Hotelling, H.    1948.   Letter  to the National  Park Service in "Economics of
     Outdoor  Recreation" - The Prewitt Report.

Huppert,  D.D.    1983.    NMFS Guideline,  on Economic Valuation of Marine
     Recreational _ Fishing.           NOAA      Technical      Memorandum
     NOAA-TM-NMFS-SWFC-32, Department of Commerce,  Washington, D.C.

Johnston, J.    1972.    Econometric  Methods.   Second Edition McGraw-Hill, New
     York.

Just, Richard, Darrell Hueth  and Andrew Schmits.    1982.   Applied  Welfare
     Economics and Public  Policy.  Englewood Cliffs, N.J. Prentice-Hall inc.

Just,  R E.,  and  D.  L. Hueth.   1979.    "Multimarket Welfare  Measurement."
     American Economic  Review 69947-64.

Kealy, Mary Jo, and Richard  C. Bishop.    1983.    "Theoretical  and  Empirical
     Specification  Issues  in  Travel Cost Demand  Studies. "Unpublished.
Kendall,  M.  and Alan Stuart.   1977.   The Advanced  Theory of shaHgHrw. 4th
     Edition,  (Vol.  1), New York:  MacMillan.

Kling,  Catherine.   1986.   *A Simulation  Approach to Comparing Multiple Site
     Recreation Demand Model.  Using Chesapeake Bay  Survey Data." Paper
     presented at the Conference  on  the Economics of the Chesapeake Bay
     Management II, Annapolis, MD.

Kling, Catherine,  Nancy Bockstael  and Ivar Strand.  1985. "An  Evaluation of
     Multiple Site Model Development."  Paper presented at the AEA meet ings,
     New York,  N.Y.

Kling, Catherine.  1886.   Meaaurin* the Recreational Benefits of Environmental
     AmeniUa* Uain* Multiple Site Model*: An Evaluation of Techniauaa.  Ph.D.
     Dissertation,  University of Maryland.

Krutilla,  J.  V.    1967.    "Conservation Revisited."   American Economic  Review.
     57(  Sept. ):777-86

Lewbel, Arthur.    1975.    "A Unified Approach to Incorporating Demographic or
     Other Effects into Demand Systems."  Review of Economic Studies 42.


                                    114

-------
Loomis, J.   1986.   "An  Examination of  the Variation in Empirical Estimates of
     Site Specific  Marginal Values for Recreational Fishermen." Working  paper,
     Division of  Environmental  Studies, University of California, Davis.

Maddala,   G.  S.     1983.    Limited  Dependent and  Qualitative Variables in
     Econometrics.  Cambridge University Press New York.

Maddala,  G. S.   1977. Econometric McGraw-Hill,  Inc., N.  Y. 516 pp.

Maler,  Karl-Goran.    1974.   Environmental Economics;  A Theoretical Inquiry,
     Baltimore,  Johns Hopkins  University Press  for Resources for the Future.

Mathews,  S. B. and G.  S. Brown.   1970.   "Economic Evaluation of  the 1967
     Sport  Salmon Fisheries  of Washington*"    Washington Department of
     Fisheries,  Technical Report #2.

McConnell, K. E. and N. E. Bockstael. 1984.  "Aggregation  in  Recreation
     Economics:  Issues of Estimation  and Benefit  Measurement."  Northeastern
     Journal of Agricultural and  Resource Economics  13:181-186.

McConnell, K. E. and  I.  E.  Strand.    1981.   'Measuring the Cost of Time in
     Recreation  Demand Analysis." American Journal  of Agricultural Economics
     63{  Feb):153-156.

McConnelltK. B.  1986.     Identification of preference,  in Hedonic Models.
     Report presented  to Environmental  Protection Agency under Cooperative
     Agreement CR-811043-01-0,  Vol I.

McConnell,  K.  E. 1879.    "Values of  Marine Recreational Fishing:  Measurement
     and Impact  of Measurement." American Journal of Agricultural Economics
     61(5):921-923.

McConnellrK. E.  and T.  P. Smith. 1979.  "Marine Recreational Fishing in Rhode
     Island,"  University of Rhode Island Marine  Memorandum #62.

McFadden, Daniel.     1874.    "Conditional Logit  Analysis of Qualitative Choice
     Behavior."  in Frontiers in Econometrics.  Edited  by  P.  Zarembka.

McFadden, D.     1978.     "Modelling   the Choice  of Residential Location. "
     In  Karlqjist,   Lundquist,  Snickars,  and Weibull  (eds) Spatial  Interaction
     Theory and  Planning    Models.    New York   North-Holland Publishing
     company.

Mendelsohn, Robert.   1984.   "Estimating  the  Structural Equations of Implicit
     Markets and Household Production Functions."  Review  of Economics  and
     Statistics  66:673-77.

Meyer, P.  A.  1982.   Net Economic Values for Salmon and Steelhead from  the
     Columbia River  Sustem.   NOM  Technical Memorandum  NMFS F/NWR-3, U.
     S. Department  of Commerce,  Washington, D.C.


                                   115

-------
Miller, Jon.    1982.    "Valuation Methods for Fish  and Wildlife Associated
     Recreation: An Empirical  Comparison of  the Travel  Coat  and Contingent
     Valuation  Method-."   Paper presented at  the ASSA  meetings,  New York,
     N.Y.

Mitchell,  R C.  and  Richard T.  Carson.   1984.    "National  Fresh Water Quality
     Benefits:    Findings from a New  National Contingent Valuation  Survey."
     Presented at the annual AERE meeting,  Cornell University.

Morey, Edward. 1981.  "The Demand for Site-Specific Recreational  Activities:  A
     Charactaristics Approach."    Journal of Environmental Economics  and
     Management  8:345-71.

Morey, Edward.  1984a.  "Characteristics,  Consumer  Surplus and  New Activities:
     A Proposed Ski Area."  Paper presented  at the ASSA  meetings, Dallas,
     TX.

Morey, Edward.  1984b.   "The Choice  of Ski Areas: Estimation of a Generalized
     CBS Preference Ordering with Characteristics."  Review ofEconomics and
     Statistics  66:584-90.

Muellbauer,  J.    "Household  Composition,  Engel  Curvea and Welfare Comparisons
     Between  Households A  Duality  Approach." European  Economic  Review
     5(1974):103-22.

JVorton, V. J., T. P. Smith, and I.  E.  Strand.   1889. Stripers:  The  Economic
     Value ofthe  Atlantic  Coast Commercial  and Recreational Striped  Bass
     Fisheries. Maryland Sea  Grant Publication #UM-SG-TS-83-12.

Pacific  Fisheries  Management Council.  1886. Review of  1985  Ocean  Salmon
     Fisheries.  Portland, Oregon.

Parks,  RS.  1969,  "Systems of Demand Equations:  An Enpirical  Comparison of
     Alternative  Functional  Forma."  Econometrics  37:629-65

Pesaran, M.H.    and   A.S.  Deaton.     1978.     "Testing Non-nested Non-linear
     Regression  Models."  Econometricca 46:677-94.

Pindyck, Robert and Daniel Rubinfeld.  1981. Econometric Models and Economic
     Forecasts.  New York, McGraw-Hill  Book Company.

Pow,   Robert.  1971.   "Additive Utility  Functions and Linear  Engel Curves.  "
     Review of Bconoaic Studies  38401-14.
Pollak,  Robert  and Terence  Wales.     19790    "Estimation  of the  Linear
     Expenditure System."   Econometrics 37:611-29.

Pollak,  Robert and Terence Wales.   1881.   "Demographic Variables in Demand
     Analysis."  Econometrics 4*1533-51.
                                   116

-------
Reiling, S. D., K.C.  Gibbs and H.H. Stoevner.  1973.  Economic Benefits from  an
     Improvement  in Water Quality.  Washington,D.C.,  U.S.  EPA.

Ribaudo,  Marc 0.  and Donald  J.  Epp.   1983.   "The  Importance of  Sample
     Discrimination in Using the Travel Cost  Method to Estimate the Benefits  of
     Improved Water Quality."  Paper  presented at the annual AAEA meeting,
     Purdue,  Indiana.

Rosen,  Sherwin.     1974.     "Hedonic  Prices and  Implicit Markets: Product
     Differentiation in Pure Competition."    Journal  ofPolitical Economy
     82:34-55.

Rowe, Robert,  Edward Morey, Arthur Ross and W.  Douglass  Shaw.    1985.
     Valuing Marine  Recreation Fishing  on  the  Pacific  Coast.   Energy and
     Resurce Consultants,  Report  prepared  for National Marine Fisheries
     Service, NOAA.

Russell, C.  S.,  and W. A.  Vaughan.   1982.   "The  National Recreational  Fishing
     Benefits of  Water Pollution Control." Journal  of Environmental  Economics
     and Management 9:328-354.

Samples, K.  C.,  and  R. C.  Bishop.    (Undated).   "Estimating Social  Value,  of
     Sport  Caught Fish  A Suggested Approach."  Draft,  12  pps.

Samuelson,  Paul A.   1947.    "Some Implications  of Linearity,"     Review  of
     Economic Studies 15:88-90.

Saxonhouse,   G.R.     1977.     "Regressions  from  Samples Having Different
     Character!-tics." Review of Economics and Statistics 59294-7.

Schulse, W.  D., D.  S. Brookshire, B. G. Walther, M.  K. McFarland, M. A.  Thayer,
     R. L.  Whitworth, S.  Ben-David,  W.  Maim, and J. Molenar.  19S3.   "The
     Economic Benefits of Preserving Visibility in the National  Parklands  of  the
     southwest."   Natural  Resources Journal  23:149-173.

Sellar,  Christine,  John  Stoll and Jean-Paul Chavas.    196s.    "Validation of
     Empirical Measures  of Welfare  Change:  A  Comparison  of  Nonmarket
     Techniques.  "  Land Economics 61:156-75.

Sellar, C.,  J. P.  Chavas  and J.  R. Stoll.  1966.    "Specification  of the Logit
     Model:    The Case of  Valuation of  Nonmarket Goods."    Journal  'of
     Environmental Economics  and Manaoment Vol.  13:382-390.

Smith, V.  Kerry and William H.  Deavousge.     1985.   "The Generalized Travel
     Cost Model and Water Quality Benefits:  A Reconsideration."   Southern
     Economic Journal 62371-81.

Smith, V. K., W.  H. Desvousges and  Ann Fisher. 1986.  "A  Comparison of Direct
     and Indirect Methods for Estimating  Environmental Benefits."    Amer.  J.
     Aar. Econ.   68 (2) :280-290.
                                   117

-------
Smith,  V.  Kerry,  William  Ho  Desvousges and M.P. McGivney.  1983.  "Estimating
     Water Quality  Benefits:   An Econometric Analysis. "   Southern  Economic
     -50:422-37"

Smith, V. K., W. H.  Desvousges and M.  P.  McGivney.   1983.   "The Opportunity
     Cost of  Travel Time  in Recreation Demand. Models."    Land Economics.
     59:3:259-78.

Smith, V. Kerry and  Yoshiaki  Kaoru. 1985.   "The Hedonic Travel Cast Model: A
     View from  the  Trenches."  Mimeo.

Stevens, J.  B.   1966.    "Recreation Benefits  from Water  Pollution Control. "
     Water Resources Research  2:167-182.

Stevens, Joe B.   1866.   "Angler Success as  a  Quality Determinant of  Sport
     Fishery  Recreational  Values.  "    Transactions of  the American Fisheries
     Society.  95:357-362.

Stoevener, H.  H.   1863.   "An Economic  Evaluation of Water Pollution Control
     Alternatives:    A  Progress Report."     Water  Resources  and  Economic
     Development of  the West,  No.  12,  Committee on the  Economics of Water
     Resources Development of the Western Agricultural Economic.  Research
     Council,  Salt Lake City,  47-59.

Stoevener,  H. H. 1864.   "Water uee Relationships as Affected  by Water Quality
     on the Yaquina  Bay."  New Horizons for Resources Research: Issues  and
     Methodology. Western Resources Conference,  University of Colorado Press,
     Boulder,  87-86.

Stone,   J.R.M.     1854.    The Measurement of  Consumers'  Expenditure and
     Behaviour in the United Kingdom.  1920-1938,  Vol  L Cambridge University
     Press,  Cambridge.

Strand,   I.  E. and N.  E.  Bockstael. 1985.   "The Effect of Common  Sources of
     Regression Error  on Benefit  Estimation."    Working paper, University of
     Maryland.

Sutherland, R.J. 1881.   "A  Regional Approach to  Estimating Recreation Benefits
     of Improved Water  Quality. "   Journal of Environmental  Economics  and
     Management  9:229-47.

Thayer, Mark.    1981.     "Contingent   Valuation  Techniques for  Assessing
     Environmental  Impacts: Further Evidence."   Journal  of Environmental
     Economic* and Management  8:27-44.

Theil,  Henry.   1976.  Theory and Measurement of Consumer Demand. Vols  1  and
     2. Amsterdam.

Tobin,   James.    1985.    "Estimation  of Relationships for  Limited Dependent
     Variables."  Econometrics 26:24-36.
                                   118

-------
u. S. EPA.     1983.    Water  Quality  Standards  Handbook.   office ofWater
     Regulation,  and  Standards.  Washington,  DC.

U. S. Department of  Commerce.  1984.   "Marine  Recreational Fishery  Statistics
     Survey,  Atlantic and Gulf  Coasts, 1979  (Revised)  -  1980, Current Fishery
     Statistics 8322." Washington,  DC.

u, S. Department of  the Interior. "1980 National Survey of  Fishing, Hunting,
     and Wildlife-Associated Recreation."    U.  S.  Fish  and  Wildlife Service.
     Washington, DC. Undated.

Vaug ban,  W. J.  and C. S. Russell.     1982.     "Valuing a  Fishing Day: An
     Application of a Systematic Varying Parameter Model."   Land  Economics
     58:450-62.

Vaughan, W.  J.  and C.  S.  Russell.   1982. Fresh Water Recretional Fishing; The
     National Benefits of Water Pollution Control.  Washington D. C.:RFF, Inc.

Vaughan, W. J., C. S.  Russell, L.  P.  Gianessi and  L.  A. Nielsen. "Measuring and
     Predicting Water Quality  in  Recreation Related Terms."   J."  Env.  Mornt.
     15:363-380.

Violette,  Daniel.    188s.    "A  Model  to  Estimate  the  Economic  Impacts on
     Recreational Fishing  in the  Adirondacks from Current   Levels of
     Acidification."    Paper presented  at the AERE Workshop on Recreation
     Demand Modeling,  Boulder,  CO.

Wales, Terence  J.   1971.    "A  Generalised  Linear Expenditure  Model  of  the
     Demand forNon-Durable  Goods in Canada." Canadian Journal  of Eonomics
     4:471-84.

Wales, Terence  and A. D. Woodland.    1883.   "Estimation of Consumer Demand
     Systems   with  Binding   Non-Negativity     Constraints."      Journal of
     Econometrics  21:263-85.

Walsh, Richard  G.    Undated.    '*An Extension  of  the  Contingent  Valuation
     Approach  to Estimate the  Effect oflnstream  Flow on  Recreational
     Demand."   Unpublished.

Walsh, R G.,  D. A.  Greenley, R A. Young, J.  RMcKean,  anc&. A.  Pinto.  1978.
     "Option Values and Recreational Benefits  of  Improved Water Quality:  A
     Case Study  of  the  South Platte River  Basin,  Colorado."     Rep.
     EPA-600/5-78-100.     U. S.  Environmental Protection Agency, Research
     Triangle Park,  N. C.

Walsh, R G., L. D.  Sanders,  and J. B.  Loomis.   1985.  Wild and Scenic  River
     Economics: Recreation  Use  and Preservation Values.  American Wilderness
     Alliance,  Englewood, Colorado.
                                   119

-------
Wegge,  Thomas,  W. Michael  Hanemann and Ivar Strand.   1985.   "An Economic
     Assessment  of Marine Recreational Fishing in Southern California."
     Report  subritbed to National Marine Fisheries Service Southwest  Regional
     Office.

Wennergren, E.  B. and D. B.  Nielson.     1970. '    "Probability Estimates of
     Recreation Demands." Journal of Leisure Research 2:112-22.

Williams,  J.  B.,  T. P. Smith,  M.  J. Speir,  and S.  Early.   1983.   "1980 Maryland
     Saltwater Sport  Fishing Survey.  "    Maryland Department of Natural
     Resources Tidewater Administration Report No. TA-CRD-83-1.   Annapolis,
     MD.

Willig,  R.D.   1976.   "Consumer's  Surplus  without Apology."  American Economic
     Review  66:589-97.

Willig,  R.  D.     1978.     "Increment Consumer's Surplus and  Hedonic  Price
     Adjustments."  Journal  o/Economic Theory 17:227-253.

Wilman,  Elizabeth A.    1884.    External Costa of Coastal Beach Pollution.
     Resources for the  Future,  Washington,  DC.

Wilman,  Elizabeth.   1985.   "Valuing Quality  Changes  in  Recreation  Resources.™
     Paper presented at  the AERE Workshop on Recreation  Demand Modeling,
     Boulder, CO.

Zellner, A .   and  Soo-Bin Park.    1879.    "Minimum Expected Lose  (MELO)
     Estimators for Functions  of Parameters and  Structural Coeffiecients of
     Econometric Models." Joimial of the  American  Statistical Association.
     74:365:185-193.
                                   120

-------
                                REFERENCES
Bell,  F. W. 1968.    "The Pope  and the Price of Fish.  "   American Economic
     Review 58142-147.

Binkley,  Clark S.  and W.  Michael Hanemann.    1978. The  Recreation Benefits of
     Water  Quality  Improvements;   An  Analysis of Day  Trips  in  an Urban
     Setting. Washington, B.C.  E. P.A.-600/5-78-010.

Bishop,  R.C. and T.A.  Heberlein.   "Simulated Markets,  Hypothetical  Markets and
     Travel  Cost Analysis:   Alternative Method, of Estimating Outdoor
     Recreation Demand."   Working paper  no.  187.

Bishop,  R. C. and  K.  Sample.    1886.   "Estimating  the  Value of Variations in
     Anglers'  Success Rate.:   An Application of the  Multiple-Site Travel Cost
     Method."  Marine Resource Economics 2:1:55-74

Blackburn, A.J.  1870.   "An Alternative Approach  to  Aggregation and  Estimation
     in  the Non-Linear Model."  Ch. 9  in The Damand for Travel;  Theory  and
     Measurement. Edited by  R E.  Quandt.  B.C.  Heath,  Lexington.

Bockstael,  Nancy,  W.  Michael  Hanemann and Catherine Kling.  1887.  "Modeling
     Recreational  Demand in a Multiple Site Framework."   Water  Resources
     Research. Forthcoming.

Bockstael,  Nancy, W.  Michael Banemann and  Ivar  Strand.  18SS.  Measuring  the
     Benefits  of  Water Quality Improvments   Using  Itocreaticn Demand Models.
     Report   presented   to    the    B.P.A.    under   Cooperative Agreement
     CR-81 143-01-1, Volume II.

Bockstael,  N.  E., and K.  E. McCormell. 1981.   "Theory  and Estimation  of  the
     Household  Production Function for Wildlife Recreation."     Journal of
     Environmental Economics and Management 8:188-214.

Bockstael,  N. E.,  and  K.  E.  McConnell.    1983.   "Welfare Measurement  in  the
     Household  Production Framework." American Economic Review  73:806-814.

Bockstael, N. B.,   and J.  J.  Opaluch.   1983.    "Discrete Modelling of  Supply
     Response under Uncertainty:    The Case of  the  Fishery."    Journal of
     Environmental Economics and Management  10:125-137.

Bockstael,  N.  B., I.  B. Strand,  and W. M.  Hanemann.    1987.    "Time and  The
     Recreational Demand Model." American  Journal of Agricultural   Economics
     69213-302

Bockstael,  N.  E.  and  I. E.  Strand.  1987.   "The  Effect of Common  Sources of
     Regression Error  on Benefit Estimates." Land  Economics 63:11-20.
                                   109

-------
Bocks tael, N. E.,  I.  E.  Strand,  and A. Graefe.   Economic Analysis of  Artificial
     Reefs:   A Pilot Study  of Selected Valuation Methodologies.   Technical
     Report  #6.   Artificial  Reef Development  Center,  Sportfishing  Institute.
     1986.

Brookshire,  D.S.,   L.S.  Eubanks and A.  Randall.  .  1983.    "Estimating Option
     Prices and Existing Values  for Wildlife Resources.  "    Land Economics
     58:1-15.

Brown, Gard  nor  and  Robert  Mendelsohn.    1984.    "The  Hedonic Travel Cost
     Method." Review of Economics  and  Statistics 66:427-33.

Brown, W. G., C.  Sorhus, B.  Chow-Yang, S.  Richards.  1883.  "Using Individual
     Observations to Estimate  Recreation Demand Functions."  " American Journal
     of Agricultural Economics  65:154-157.

Brown, W. G.,  C. N. Sorhus,  and D.  A.  Meyer.  1882.   "Estimation of  Oregon
     Sport-Caught  Salmon and Steelhead Values Via the Travel Cost  Method."
     Mimeo, ASSA Meetings,   25 ppa.

Brown, W. G., A. Singh  and  E. N. Castle.    "Net Economic  Value  of the Oregon
     Salmon-Steelhead Sport Fishery."     Journal  of  Wildlife   Management
     29(2):266-279.

Burt,  Oscar  and Durward Brewer.   1872.   "Estimation of Net Social Benefits
     From Outdoor Recreation."  Econometrica 39:813.

Carson, R.  T.  and  R C. Mitchell   188sO   "The Value of Clean Water:    The
     Public'. Willingness to Pay for Beatable,  Fishable,   and  Swimmable Quality
     Water."    Discussion paper No.  QE85-08 revised,  (Washington,  D. C.,
     Resources for the  Future).

Castle, Emery N.  and W.  G.  Brown.  1884.    "The Economic Value of  a
     Recreational  Resource:   A Case Study  of  the  Oregon  Salmon-Steelhead
     Sport Fishery.  Paper presented before the Committee  on  the  Economics
     of Water Resource* Development of the Western Agricultural Economics
     Research  Council,  San  Francisco,  California.

Caulkins,  Peter,  Richard  Bishop and Nicolass Bouwes.  1982.  "The Travel Coat
     Model  for   Lake  Recreation  A  Comparison of  Two  Methods   for
     Incorporating  Site Quality and Substitution Effects." Mimeo, ABA annual
     meetings,

Caulkins,  Peter P.,  Richard  C. Bishop  and Nicolass W.  Bouwes,  Sr. 1888. "The
     Travel  Cost  Model  for  Lake  Recreation: A  Comparison of  Two Methods  for
     Incorporating  Site Quality and Substitution Effects." Amer.  J.  Air. Econ.
     68(2):291-297.

Caulkins,  Peter,  Richard  Bishop and  Nicolass Bouwes.     1985.    "Omitted
     Cross-Price Variable  Biases  in the Linear Travel Coat Model:  Correcting
     Common Misperceptions.   " Land Economics  61:182-85.


                                   110

-------
&aario,  Frank J.     1969.    "Operations  Research in  Outdoor Recreation.  "
     Journal of  Leisure Research 1:33-52.

Cesario,  Frank J.   1973.   "A Generalized  Trip  Distribution Model." Journal  of
     Regional Science  13:233-48.

Cesario,  Frank J.   1974.    "More  on the Generalized Trip Distribution Model,"
     Journal of  Regional Science 14:389-97

Cesario,  Frank J.  and Jack Knetach.   1976.    "A Recreation Site  Demand and
     Benefit Estimation Model. "  Regional Studies 10:97-104.

Christensen, L. R.,  Jorgenson, D.W.  and Lawrence Lau.   1975.  "Transcendental
     Logarithmic Utility Functions."   American Economic Review 65:367-83.

Cicchetti,  Charles, Anthony Fisher and V. Kerry Smith. 1976.  "An  Econometric
     Evaluation of  a Generalized  Consumer  Surplus Measure: The Mineral King
     Controversy."  Econometrics 44;1268-76.

Cicchetti,  C.J., J.J. Seneca and Paul   Davidson.  1869.    The Demand and Supply
     of  Outdoor  Recreation:  An Econometric Analysis.  Rutgers. 1869.

Clawson,  MartOn.   1859.   Methods for  Measuring the Demand  for and Value of
     Outdoor Recreation.  Resources for the Future, Inc.,   Reprint no. 10.

Clawson,  Marion and Jaok  Knetach.  1866.   Economics  of Outdoor Recreation.
     Resources for the Future, Washington D.C.

Cox,  D.R.  1861.   "Tests of Separate Families of Hypotheses." Proceedings of
     the Fourth Berkeley  Symposium on Mathematical Statistics and Probability.
     Vol  1.  University of California Press.

Cronin, F. J.  1882.   "Valuing Non-Market  Goods Through Contingent Markets.  "
     Rep. PNL-42S5.  Pac. Northw.  Lab., Richland, Wash.

Cropper,  M.  L.,  L.B. Deck, K.B.  McConnell and T.T. Phipps.  1886. "Should the
     Rosen  Model be Used to Value  Environmental  Amenities?"     Papar
     presented at the conference on  the  Economics of Chesapeake Bay
     Management 11,  Annapolis, MD.

Cummings,R G. ,  D.  S.   Brookshire  and  W. D. Schulze.    1984.  valuing
     Environmental  Goods: A State of the  Arts Assessmentof the Contingent
     Valuation Method.   Draft report to  EPA. Universities of New Mexico and
     Wyoming.    "

Daubert,  J.  T. and R A.  Young.  1981.   "Recreational  Demands  for Maintaining
     Instream Flows:  A  Contingent Vaulation  Approach." Amer. J. Apr.  Econ.
     63(4):666-676.

Beaton, A.S.  1974,   "The Analysis  of Consumer Demand  in the United Kingdom,
     1800-1970."  Econometrics 42:341-67.
                                    Ill

-------
Beaton,  Henry.  1978.   "Specification and Testing in Applied Demand Analysis.  "
     Economic Journal 88:524-36.

Desvousges,  W.  H. , V.  K. Smith and  M.  P.  McGivney.   1983.   A Comparison of
     Alternative  Approaches  for Estimation of  Recreation and RelatedBenefits
     of Water  Quality   Improvements.     Washington,  D.  C.:     Environmental
     Protection Agency,  Office of Policy Analysis.

Deuel, D. G.   1978.  "National Marine Recreational Fishery Statistics Program.  "
     Paper  presented  at Canadian Sport Fisheries Conference,  New Brunswick,
     Canada.

Domencich,  Thomas  and  Daniel McFadden.    1975.    Urban Travel Demand;  A
     Behavioral Analysis. North-Holland,  Amsterdam.

Dwyer,  J.   F.  and M.  D.  Bowes.    1978.    "Concepts of Value for  Marine
     Recreational Fishing."   American Journal of  Agricultural  Economics 60(5);
     1008-10120

Feenberg,  Daniel  and Edwin  Mills.  1980.  Measuring the  Benefits  of Water
     Pollution Abatement. New York,  Academic Press.

Fisher,  Ann and William H.  Desvousges.  1881.  "Alternative Contingent Market
     Approaches to Valuation  of Recreation  and Related  Benefits of Improved
     Water  Quality."   Paper presented at  the  Eastern Economics Association
     Meetings,  Washington B.C.

Fisher, Am and  Robert Raucher. 1982.    "Comparison of Alternative Methods of
     Evaluating the Intrinsic  Benefits  of  Improved  Water Qualiy."    Paper
     presented at  ASSA meetings, New York City.

Fisher,  F. and  K.  Shell.  1871.  "Taste  and Quality Change in the  Pure Theory
     ofthe True Cost-of-Living  Index." in Price  Indexes and  Quality Change.
     Edited  by Zvi Griliches.   Harvard University Press, Massachusetts.

Freeman,   A. Myrick III.  1979. The  Benefits  of  Environmental  Improvement:
     Theory and Practice. Johns Hopkins Press for  Resources for  the Future,
     Inc., Baltimore. 1979.

Freeman, A. Myrick  III.  1882.  Air and  Water Pollution Control:  A Benefit-Coat
     Assessment. John Wiley and Sons.

Freund,  R. J.  and  R R Wilson.   1974.    "AnExample  of a Gravity Model  to
     Estimate Recreation Travel." Journal of  Leisure  Research  6:241-56.

German, W.M.   1976.    "Tricks with Utility Function."    in Essays in  Econmic
     Analysis: Proceddings of the 1975  AUTE Conference. Sheffeld. Edited  by
     M.J. Artis and A.R.  Nobay.  Cambridge University Press,  Cambridge.
                                    112

-------
Greenley,  Douglas A., Richard  G.  Walsh  and Robert A.  Young.   1981.   "Option
     Value:   Empirical  Evidence from a Case Study of Recreation and Water
     Quality."   Ouartelv J.  Econ.  96(41:657-673.

Gum, R.L. and  W.E. Martin.   1975.   "Problems and Solutions  in  Estimating  the
     Demand for and Value of  Rural Outdoor Recreation. "  American Journal of
     Agricultural Economics 57:558-66.

Hanemann, W. Michael.    1980.    "Quality  Changes, Consumer's  Surplus,  and
     Hedonic Price Indices."   California Experiment Station working paper
     number 116.

Hanemann,  W. Michael.    1981.   "Quality and  Demand Analysis."^.  3 in  New
     Directions  in Econometric  Modeling  and Forecasting  in U.S.  Agriculture.
     Edited by  Gordon  Rausser, Elsevier/North-Holland,  New York.

Hanemann,  W. Michael.     1882.    "Applied Welfare Analysis  with  Qualitative
     Response Models."    California  Agricultural  Experiment Station working
     paper no. 241.

Hanemann,  W. Michael.    1984a.   "Multiple Site Demand:  ARew'ewofExisting
     Models  and New Development",  Chapter  9  in Measuring  the Benefits of
     Water  Quality  Improvements  Using  Recreation  Demand  Mod«=3g-    Draft
     Report presented to E.P.A. under  Cooperative Agreement CR-80043-01-0,
     VOL  II.

Hanemann,  W. Michael.     1984b.    "Discrete/Continuous Models of  Consumer
     Demand." Econometrics 52:541-61.

Hanemann,  W. M.   1978.   "A Methodological and Empirical Study of Recreation
     Benefits  from  Water Quality  Improvement."  Ph.  D.  dissertation,
     Department  of Economics, Harvard University.

Hanemann,  W. M.  1984.    "Welfare Evaluations with Market Data." American
     Journal of  Agricultural Economics. 66:332-41.

Hanemann,  W. M.    1984.    "Welfare Evaluation in Contingent Valuation
     Experiment with  Discrete Responses." American Journal of Agricultural
     Economics.  VOL  66(3) :332-341.

Harmon,  R.  J.  and Associates.   1883.  Recreational Resources at  Risk.  Taak 2
     Working Paper.  Washington,  DC.

Harrison,  D. and D. Rubinfeld. 1878.   "Hedonic Housing  Prices  and the Demand
     for  Clean Air."    Journal  of  Environmental  Economics and Management
     5:81-102.

Hausman,   J.A.    1886.     "Exact Consumer's  Surplus  and Deadweight Loss."
     American  Economic Review  71:662-76.
                                   113

-------
Heintz,  H.T. Jr., A.  Hershaft  and G.C.  Horak.  1976.  National Damages  of Air
     and Water Pollution.  A Report to the EPA.  Enviro Control,  Inc.  Rockville,
     MD.

Hiett, R  L.  and J. W.  Worrall.  1977.   "Marine  Recreational Fisheries Ability to
     Estimate Catch and to  Recall Catch and Effort over  Time."    Human
     Sciences  Research,  Inc., Report HSR-RR-77/18,  Washington, D.C.

Horowitz, Joel.    1983.    "Statistical Comparison  ofNon-nested Probabilistic
     Discrete Choice Models.  "   Transportation Science 17:319-50.

Hotelling, H.    1948.   Letter  to the National  Park Service in "Economics of
     Outdoor  Recreation" - The Prewitt Report.

Huppert,  D.D.    1983.    NMFS Guideline,  on Economic Valuation of Marine
     Recreational _ Fishing.           NOAA      Technical      Memorandum
     NOAA-TM-NMFS-SWFC-32, Department of Commerce,  Washington, D.C.

Johnston, J.    1972.    Econometric  Methods.   Second Edition McGraw-Hill, New
     York.

Just, Richard, Darrell Hueth  and Andrew Schmits.    1982.   Applied  Welfare
     Economics and Public  Policy.  Englewood Cliffs, N.J. Prentice-Hall inc.

Just,  R E.,  and  D.  L. Hueth.   1979.    "Multimarket Welfare  Measurement."
     American Economic  Review 69947-64.

Kealy, Mary Jo, and Richard  C. Bishop.    1983.    "Theoretical  and  Empirical
     Specification  Issues  in  Travel Cost Demand  Studies. "Unpublished.
Kendall,  M.  and Alan Stuart.   1977.   The Advanced  Theory of shaHgHrw. 4th
     Edition,  (Vol.  1), New York:  MacMillan.

Kling,  Catherine.   1986.   *A Simulation  Approach to Comparing Multiple Site
     Recreation Demand Model.  Using Chesapeake Bay  Survey Data." Paper
     presented at the Conference  on  the Economics of the Chesapeake Bay
     Management II, Annapolis, MD.

Kling, Catherine,  Nancy Bockstael  and Ivar Strand.  1985. "An  Evaluation of
     Multiple Site Model Development."  Paper presented at the AEA meet ings,
     New York,  N.Y.

Kling, Catherine.  1886.   Meaaurin* the Recreational Benefits of Environmental
     AmeniUa* Uain* Multiple Site Model*: An Evaluation of Techniauaa.  Ph.D.
     Dissertation,  University of Maryland.

Krutilla,  J.  V.    1967.    "Conservation Revisited."   American Economic  Review.
     57(  Sept. ):777-86

Lewbel, Arthur.    1975.    "A Unified Approach to Incorporating Demographic or
     Other Effects into Demand Systems."  Review of Economic Studies 42.


                                    114

-------
Loomis, J.   1986.   "An  Examination of  the Variation in Empirical Estimates of
     Site Specific  Marginal Values for Recreational Fishermen." Working  paper,
     Division of  Environmental  Studies, University of California, Davis.

Maddala,   G.  S.     1983.    Limited  Dependent and  Qualitative Variables in
     Econometrics.  Cambridge University Press New York.

Maddala,  G. S.   1977. Econometric McGraw-Hill,  Inc., N.  Y. 516 pp.

Maler,  Karl-Goran.    1974.   Environmental Economics;  A Theoretical Inquiry,
     Baltimore,  Johns Hopkins  University Press  for Resources for the Future.

Mathews,  S. B. and G.  S. Brown.   1970.   "Economic Evaluation of  the 1967
     Sport  Salmon Fisheries  of Washington*"    Washington Department of
     Fisheries,  Technical Report #2.

McConnell, K. E. and N. E. Bockstael. 1984.  "Aggregation  in  Recreation
     Economics:  Issues of Estimation  and Benefit  Measurement."  Northeastern
     Journal of Agricultural and  Resource Economics  13:181-186.

McConnell, K. E. and  I.  E.  Strand.    1981.   'Measuring the Cost of Time in
     Recreation  Demand Analysis." American Journal  of Agricultural Economics
     63{  Feb):153-156.

McConnelltK. B.  1986.     Identification of preference,  in Hedonic Models.
     Report presented  to Environmental  Protection Agency under Cooperative
     Agreement CR-811043-01-0,  Vol I.

McConnell,  K.  E. 1879.    "Values of  Marine Recreational Fishing:  Measurement
     and Impact  of Measurement." American Journal of Agricultural Economics
     61(5):921-923.

McConnellrK. E.  and T.  P. Smith. 1979.  "Marine Recreational Fishing in Rhode
     Island,"  University of Rhode Island Marine  Memorandum #62.

McFadden, Daniel.     1874.    "Conditional Logit  Analysis of Qualitative Choice
     Behavior."  in Frontiers in Econometrics.  Edited  by  P.  Zarembka.

McFadden, D.     1978.     "Modelling   the Choice  of Residential Location. "
     In  Karlqjist,   Lundquist,  Snickars,  and Weibull  (eds) Spatial  Interaction
     Theory and  Planning    Models.    New York   North-Holland Publishing
     company.

Mendelsohn, Robert.   1984.   "Estimating  the  Structural Equations of Implicit
     Markets and Household Production Functions."  Review  of Economics  and
     Statistics  66:673-77.

Meyer, P.  A.  1982.   Net Economic Values for Salmon and Steelhead from  the
     Columbia River  Sustem.   NOM  Technical Memorandum  NMFS F/NWR-3, U.
     S. Department  of Commerce,  Washington, D.C.


                                   115

-------
Miller, Jon.    1982.    "Valuation Methods for Fish  and Wildlife Associated
     Recreation: An Empirical  Comparison of  the Travel  Coat  and Contingent
     Valuation  Method-."   Paper presented at  the ASSA  meetings,  New York,
     N.Y.

Mitchell,  R C.  and  Richard T.  Carson.   1984.    "National  Fresh Water Quality
     Benefits:    Findings from a New  National Contingent Valuation  Survey."
     Presented at the annual AERE meeting,  Cornell University.

Morey, Edward. 1981.  "The Demand for Site-Specific Recreational  Activities:  A
     Charactaristics Approach."    Journal of Environmental Economics  and
     Management  8:345-71.

Morey, Edward.  1984a.  "Characteristics,  Consumer  Surplus and  New Activities:
     A Proposed Ski Area."  Paper presented  at the ASSA  meetings, Dallas,
     TX.

Morey, Edward.  1984b.   "The Choice  of Ski Areas: Estimation of a Generalized
     CBS Preference Ordering with Characteristics."  Review ofEconomics and
     Statistics  66:584-90.

Muellbauer,  J.    "Household  Composition,  Engel  Curvea and Welfare Comparisons
     Between  Households A  Duality  Approach." European  Economic  Review
     5(1974):103-22.

JVorton, V. J., T. P. Smith, and I.  E.  Strand.   1889. Stripers:  The  Economic
     Value ofthe  Atlantic  Coast Commercial  and Recreational Striped  Bass
     Fisheries. Maryland Sea  Grant Publication #UM-SG-TS-83-12.

Pacific  Fisheries  Management Council.  1886. Review of  1985  Ocean  Salmon
     Fisheries.  Portland, Oregon.

Parks,  RS.  1969,  "Systems of Demand Equations:  An Enpirical  Comparison of
     Alternative  Functional  Forma."  Econometrics  37:629-65

Pesaran, M.H.    and   A.S.  Deaton.     1978.     "Testing Non-nested Non-linear
     Regression  Models."  Econometricca 46:677-94.

Pindyck, Robert and Daniel Rubinfeld.  1981. Econometric Models and Economic
     Forecasts.  New York, McGraw-Hill  Book Company.

Pow,   Robert.  1971.   "Additive Utility  Functions and Linear  Engel Curves.  "
     Review of Bconoaic Studies  38401-14.
Pollak,  Robert  and Terence  Wales.     19790    "Estimation  of the  Linear
     Expenditure System."   Econometrics 37:611-29.

Pollak,  Robert and Terence Wales.   1881.   "Demographic Variables in Demand
     Analysis."  Econometrics 4*1533-51.
                                   116

-------
Reiling, S. D., K.C.  Gibbs and H.H. Stoevner.  1973.  Economic Benefits from  an
     Improvement  in Water Quality.  Washington,D.C.,  U.S.  EPA.

Ribaudo,  Marc 0.  and Donald  J.  Epp.   1983.   "The  Importance of  Sample
     Discrimination in Using the Travel Cost  Method to Estimate the Benefits  of
     Improved Water Quality."  Paper  presented at the annual AAEA meeting,
     Purdue,  Indiana.

Rosen,  Sherwin.     1974.     "Hedonic  Prices and  Implicit Markets: Product
     Differentiation in Pure Competition."    Journal  ofPolitical Economy
     82:34-55.

Rowe, Robert,  Edward Morey, Arthur Ross and W.  Douglass  Shaw.    1985.
     Valuing Marine  Recreation Fishing  on  the  Pacific  Coast.   Energy and
     Resurce Consultants,  Report  prepared  for National Marine Fisheries
     Service, NOAA.

Russell, C.  S.,  and W. A.  Vaughan.   1982.   "The  National Recreational  Fishing
     Benefits of  Water Pollution Control." Journal  of Environmental  Economics
     and Management 9:328-354.

Samples, K.  C.,  and  R. C.  Bishop.    (Undated).   "Estimating Social  Value,  of
     Sport  Caught Fish  A Suggested Approach."  Draft,  12  pps.

Samuelson,  Paul A.   1947.    "Some Implications  of Linearity,"     Review  of
     Economic Studies 15:88-90.

Saxonhouse,   G.R.     1977.     "Regressions  from  Samples Having Different
     Character!-tics." Review of Economics and Statistics 59294-7.

Schulse, W.  D., D.  S. Brookshire, B. G. Walther, M.  K. McFarland, M. A.  Thayer,
     R. L.  Whitworth, S.  Ben-David,  W.  Maim, and J. Molenar.  19S3.   "The
     Economic Benefits of Preserving Visibility in the National  Parklands  of  the
     southwest."   Natural  Resources Journal  23:149-173.

Sellar,  Christine,  John  Stoll and Jean-Paul Chavas.    196s.    "Validation of
     Empirical Measures  of Welfare  Change:  A  Comparison  of  Nonmarket
     Techniques.  "  Land Economics 61:156-75.

Sellar, C.,  J. P.  Chavas  and J.  R. Stoll.  1966.    "Specification  of the Logit
     Model:    The Case of  Valuation of  Nonmarket Goods."    Journal  'of
     Environmental Economics  and Manaoment Vol.  13:382-390.

Smith, V.  Kerry and William H.  Deavousge.     1985.   "The Generalized Travel
     Cost Model and Water Quality Benefits:  A Reconsideration."   Southern
     Economic Journal 62371-81.

Smith, V. K., W.  H. Desvousges and  Ann Fisher. 1986.  "A  Comparison of Direct
     and Indirect Methods for Estimating  Environmental Benefits."    Amer.  J.
     Aar. Econ.   68 (2) :280-290.
                                   117

-------
Smith,  V.  Kerry,  William  Ho  Desvousges and M.P. McGivney.  1983.  "Estimating
     Water Quality  Benefits:   An Econometric Analysis. "   Southern  Economic
     -50:422-37"

Smith, V. K., W. H.  Desvousges and M.  P.  McGivney.   1983.   "The Opportunity
     Cost of  Travel Time  in Recreation Demand. Models."    Land Economics.
     59:3:259-78.

Smith, V. Kerry and  Yoshiaki  Kaoru. 1985.   "The Hedonic Travel Cast Model: A
     View from  the  Trenches."  Mimeo.

Stevens, J.  B.   1966.    "Recreation Benefits  from Water  Pollution Control. "
     Water Resources Research  2:167-182.

Stevens, Joe B.   1866.   "Angler Success as  a  Quality Determinant of  Sport
     Fishery  Recreational  Values.  "    Transactions of  the American Fisheries
     Society.  95:357-362.

Stoevener, H.  H.   1863.   "An Economic  Evaluation of Water Pollution Control
     Alternatives:    A  Progress Report."     Water  Resources  and  Economic
     Development of  the West,  No.  12,  Committee on the  Economics of Water
     Resources Development of the Western Agricultural Economic.  Research
     Council,  Salt Lake City,  47-59.

Stoevener,  H. H. 1864.   "Water uee Relationships as Affected  by Water Quality
     on the Yaquina  Bay."  New Horizons for Resources Research: Issues  and
     Methodology. Western Resources Conference,  University of Colorado Press,
     Boulder,  87-86.

Stone,   J.R.M.     1854.    The Measurement of  Consumers'  Expenditure and
     Behaviour in the United Kingdom.  1920-1938,  Vol  L Cambridge University
     Press,  Cambridge.

Strand,   I.  E. and N.  E.  Bockstael. 1985.   "The Effect of Common  Sources of
     Regression Error  on Benefit  Estimation."    Working paper, University of
     Maryland.

Sutherland, R.J. 1881.   "A  Regional Approach to  Estimating Recreation Benefits
     of Improved Water  Quality. "   Journal of Environmental  Economics  and
     Management  9:229-47.

Thayer, Mark.    1981.     "Contingent   Valuation  Techniques for  Assessing
     Environmental  Impacts: Further Evidence."   Journal  of Environmental
     Economic* and Management  8:27-44.

Theil,  Henry.   1976.  Theory and Measurement of Consumer Demand. Vols  1  and
     2. Amsterdam.

Tobin,   James.    1985.    "Estimation  of Relationships for  Limited Dependent
     Variables."  Econometrics 26:24-36.
                                   118

-------
u. S. EPA.     1983.    Water  Quality  Standards  Handbook.   office ofWater
     Regulation,  and  Standards.  Washington,  DC.

U. S. Department of  Commerce.  1984.   "Marine  Recreational Fishery  Statistics
     Survey,  Atlantic and Gulf  Coasts, 1979  (Revised)  -  1980, Current Fishery
     Statistics 8322." Washington,  DC.

u, S. Department of  the Interior. "1980 National Survey of  Fishing, Hunting,
     and Wildlife-Associated Recreation."    U.  S.  Fish  and  Wildlife Service.
     Washington, DC. Undated.

Vaug ban,  W. J.  and C. S. Russell.     1982.     "Valuing a  Fishing Day: An
     Application of a Systematic Varying Parameter Model."   Land  Economics
     58:450-62.

Vaughan, W.  J.  and C.  S.  Russell.   1982. Fresh Water Recretional Fishing; The
     National Benefits of Water Pollution Control.  Washington D. C.:RFF, Inc.

Vaughan, W. J., C. S.  Russell, L.  P.  Gianessi and  L.  A. Nielsen. "Measuring and
     Predicting Water Quality  in  Recreation Related Terms."   J."  Env.  Mornt.
     15:363-380.

Violette,  Daniel.    188s.    "A  Model  to  Estimate  the  Economic  Impacts on
     Recreational Fishing  in the  Adirondacks from Current   Levels of
     Acidification."    Paper presented  at the AERE Workshop on Recreation
     Demand Modeling,  Boulder,  CO.

Wales, Terence  J.   1971.    "A  Generalised  Linear Expenditure  Model  of  the
     Demand forNon-Durable  Goods in Canada." Canadian Journal  of Eonomics
     4:471-84.

Wales, Terence  and A. D. Woodland.    1883.   "Estimation of Consumer Demand
     Systems   with  Binding   Non-Negativity     Constraints."      Journal of
     Econometrics  21:263-85.

Walsh, Richard  G.    Undated.    '*An Extension  of  the  Contingent  Valuation
     Approach  to Estimate the  Effect oflnstream  Flow on  Recreational
     Demand."   Unpublished.

Walsh, R G.,  D. A.  Greenley, R A. Young, J.  RMcKean,  anc&. A.  Pinto.  1978.
     "Option Values and Recreational Benefits  of  Improved Water Quality:  A
     Case Study  of  the  South Platte River  Basin,  Colorado."     Rep.
     EPA-600/5-78-100.     U. S.  Environmental Protection Agency, Research
     Triangle Park,  N. C.

Walsh, R G., L. D.  Sanders,  and J. B.  Loomis.   1985.  Wild and Scenic  River
     Economics: Recreation  Use  and Preservation Values.  American Wilderness
     Alliance,  Englewood, Colorado.
                                   119

-------
Wegge,  Thomas,  W. Michael  Hanemann and Ivar Strand.   1985.   "An Economic
     Assessment  of Marine Recreational Fishing in Southern California."
     Report  subritbed to National Marine Fisheries Service Southwest  Regional
     Office.

Wennergren, E.  B. and D. B.  Nielson.     1970. '    "Probability Estimates of
     Recreation Demands." Journal of Leisure Research 2:112-22.

Williams,  J.  B.,  T. P. Smith,  M.  J. Speir,  and S.  Early.   1983.   "1980 Maryland
     Saltwater Sport  Fishing Survey.  "    Maryland Department of Natural
     Resources Tidewater Administration Report No. TA-CRD-83-1.   Annapolis,
     MD.

Willig,  R.D.   1976.   "Consumer's  Surplus  without Apology."  American Economic
     Review  66:589-97.

Willig,  R.  D.     1978.     "Increment Consumer's Surplus and  Hedonic  Price
     Adjustments."  Journal  o/Economic Theory 17:227-253.

Wilman,  Elizabeth A.    1884.    External Costa of Coastal Beach Pollution.
     Resources for the  Future,  Washington,  DC.

Wilman,  Elizabeth.   1985.   "Valuing Quality  Changes  in  Recreation  Resources.™
     Paper presented at  the AERE Workshop on Recreation  Demand Modeling,
     Boulder, CO.

Zellner, A .   and  Soo-Bin Park.    1879.    "Minimum Expected Lose  (MELO)
     Estimators for Functions  of Parameters and  Structural Coeffiecients of
     Econometric Models." Joimial of the  American  Statistical Association.
     74:365:185-193.
                                   120

-------
                                Appendix A

            The Random  Digit  Dialing Survey Telephone Procedures


    For  the Random  Digit Dialing Survey, three instruments (copies included in
the Appendix)  were  developed:    a two-page screening form,  an 1 1-page,
28-item questionnaire, and a Record of Calls sheet.

1.   Screening Form

    The  screening form was intended to determine the  eligibility of  the
location served by a randomly generated telephone number.  The  number was
printed on a label affixed to the  top of  the screening  form.   Pay  phones  and
phones  used only by businesses  were  not considered eligible, since people
answering  ouch  telephones would be eligible  at  their residence  phone.. In
addition,  if  the household served by the phone waa  not located  within the
counties/cities  making up the selected SMSA's  then that phone  (residence) was
not eligible.   Once an eligible phone  (residence)  waa identified,  a  member  of
the household who waa 18 years of age or  older waa  required.    If all
residents  were  under  18,  the screening  waa completed with  •member of the
household who  waa 14  or older.

    The  screening  form waa composed of  five  sections:   an Identification
sectdoB consisting of an area  code, telephone number, and  five-digit  case
identification number,   all  printed on the  aforementioned label; »brief
introduction  to be  read  by the  interviewer which explained the study;  a
screening  section which waa used to  eliminate pay phones, businesses without
living  accommodations,  and residences not  located  in  certain  specific SMSA's;. a
screening status section to record the  screening eligibility of the location;  and
a questionnaire status  section to record whether or not a questionnaire  wafl
completed with an  rfigiHlft person.

2. Questionnaire

    The Random  Digit Dialing Survey  Questionnaire was  intended to determine
the following:

    Uae or  intent  to  use the Chesapeake Bay for recreation during  1984;
    Reasons for  nonuse;

    Activities  that the respondent (and his/her  family)  participated in
    while  visiting beaches;

    Reasons the respondent  or other  members of  his/her family do  not go
    in the water during visits  to the western shore beaches;

    Changes in swimming  participation  in the Chesapeake brought about
    by change,  in  the water quality;

    The respondent's  perception of the  water quality in the Chesapeake;

    The value respondents  place  on the Bay and how  they visualize that
    improvement should  be made and financed.
                                     121

-------
    AS in the User  Intercept Survey, a  series of demographic questions
which will enable analysts to establish profiles of beach users  and nonuser
were  included in the  questionnaire.

    The Random Digit  Dialing Survey Questionnaire waa also  divided  into
sections. The first, Identification,  had space provided for recording the case
ID number from the  screening forms the telephone interviewer's  initials,  and
the date the interview was  completed.   The  second section,  as in the  User
Intercept Survey, was a  lengthier  introduction to  be read by the  telephone
interviewer, which went into greater detail  regarding the purpose of the
survey and contained statement informing the respondent  ofthe voluntary
nature of  his/her participation   in  the study  and  assurances of the
confidentiality of  the data collected.   The third section, Recreational Uae of
the Chesapeake Bay, sought specific responses which would:     (1) enable
analysts to determine if  and how the beaches  were used and  (2)  what the
overall perception of  the water quality waa.  Thie  waa followed by a fourth
and  final  section consisting of some  18 demographic questions.


Data Collection Methods

    Two field interviewers  were  trained in  Baltimore for the data collection of
the Ueer  Intercept Survey on May  25,  1884.  A Field Interviewer Manual was
developed  (which is available upon  request)  and included  quemtion-by-question
specification,   probing   techniques,   confidentiality   procedures,   refusal
conversion  strategies,  and  other measure*necessary  to  assure the  collection
of standardized, quality data during  the  course of  the  field survey process.
Alao  covered in the manual were:   background information,  assignment
information,  sampling  procedures and administrative  procedures.

    The final day of field work on the  ueer survey was August 16,  1984.  The
confirmation portion of the user survey was  completed  on  September 1.   The
following  represents  the response rates forthe  field  work:
                                  Table A.I

                    Response Rates for Beach User Survey
  Sample
Individual
Successfully
  screened
 Eligible
Individuals
Eligible  Individuals
     Interviewed
   468

   100%)
    463

 (98.79%)
    414

    100%)
         408

       (88.55%)
                                     122

-------
Of the 468  individuals screened, 60 were not  administered questionnaires for
the  fclawing reasons:

          Ineligible  because of residence	39
          Refused screening	7
          Language barrier-screening	6
          Other	2
          Refused questionnaire	3
          Language -barrier - questionnaire	3

    Regarding the confirmation portion of the  ueer  survey,  340 of  the  people
interviewed  gave telephone  numbers or come other piece  of  information
through  which contact  could be made to conduct  a confirmation/intention
interview.    Approximately  240  (71 percent) of these individuals were
successfully  contacted during the time period allowed.

    Training of telephone  interviewers  for the Random Digit Dialing  Survey
started on  July 23.  A total  of 11 telephone interviewer, were  hired with three
of these  spending the majority of their time making  confirmation/intention
calls  to participants  in  the User Intercept  Survey.

    As in the User Intercept Surveyt each interviewer  received  a copy of  a
Telephone Interviewer Manual  specifically developed for this phase of the
project,  as well as copies of the Random Digit  Dialing instruments.    The
Telephone Interviewer Manual  (available upon request)  included question-by-
question specification, probing techniques,  confidentiality  porcedures, refusal
conversion • trategiea, and other  measures necessary to assure the  collection
of standardized, quality  data during the course  of the  telephone  •nvey
process.

    Approximately 192 telephone  interview,  were completed with western shore
beach users.  The remainder consisted of  approximately 804  nonusers and 48
intended users.  The  following two  tables represent questionnaire completions
per  strata and final totals for screening  and questionnaire  status codes.
                                     123

-------
                                 Table A.2

                   Questionnaire Completions Per  Strata
Stratum
Number
1
2
3
4
5
6
Totals
Cases
Avail .
1,230
1,100
408
1,014
820
1.560
6,132
Cases
Assigned
1,060
1,000
408
1,014
820
1.560
5,962
Quest.
Complete
155
225
70
96
111
293
1,010
Quest.
Partial
10
7
0
4
6
7
34
Total
Quest.
165
232
70
100
111
300
1,044
Quest .
Needed
138
220
" 77
112
158
295
1,000
Diff.
+ 27
+ 12
- 7
- 12
+ 19
+ 5
-44
                                 Table A.3

Final Telephone Result Totals for  Screening and  Questionnaire Status Codes


                          Screening Status Cod»»

  Eligible Identified/Screener  Completed	   1,108
  Nota  Working Telephone Number	   2,866
                                                                      13
                                                                     843
                                                                     897
                                                                      11
                                                                      10
                                                                     203
Pay Telephone 	
Business Telephone	
No Answer After Repeated  Calls	
Telephone Not Located in Bait./Wash. SMSA  	
No Eligible  Respondent Available After Repeated Calls  .  .  .
Refused to Answer Screening Questions  	
Language  Barrier 0	
Other  	

                      Questionnaire Status  Codes
                                                                       3
  Questionnaire Completed  	   1,010
  Questionnaire Partially  Completed  	      34
  Language Barrier  .  ...'	       0
  Questionnaire Refused  	      63
  Other	       1
                                    124

-------
         Appendix B




Telephone  Survey instrument
               125

-------
                             CHESAPEAKE BAT BEACH USE SURVEY
                               TELEPHONE SCREENING FORM

A.    IDENTIFICATION

     TsuPllGNs  NUMBER    (  ) .=	-	        CASE  ID NUMBER
B.    INTRODUCTION
     I   an calling fro*  the Research  Triangle Institute  near Raleigh,  North  Carol in* to
conduct • telephone survey • bout the  Chesapeake Soy for the  University of  Maryland. To
find out if I'vt contacted the  proper  type  of place, I need to ask • few  sieple questions.
First,
c.    SCREENING
     1.   Is chi* telephone niaabcr (READ  THE 7 DIGIT NUMBER ABOVE) in • rea cede
         (READ 3 DIGIT AREA CODE ABOVE)?
                                      Me JT1 (RANG UP AND DIAL NUMBER AGAIN)                (j;
                                   TOS  m

     2.   It this • pay  phone?
                                      *« m                                                (2i
                                   Yes   |1"| (CO TO *)
     3. • .    Is  tblt telephone located in • private residence or a business?
                     Private  residence   JTJ  (GO TO 4a)                                    (3)
                     Business           iTl
         b.    Aro thereby living aeconodatians* t this  place  of business?
                                 No     Qj  (GO TO 6)                                     (4)

                                 Yes    GO
         6.    Do the people  living there US. this phone for their calls?

                                 No     CD  (00 TO 6)                                     (s)
                                 Tes     Qj

         d.    Whoa can I apeak  to one of the people who lives • t this business
              location and  uses this telephone for personal calls?
              ANSWER :
              (CALL  BACK IF NECESSARY TO COMPLETE SCREENING)

     4.e.    Are you • neaber of the household  serviced by this telephone?
                                  TesQFI1  (m  TO  So)

         b.    Unen will* •caber of the household bea> liable  to  talk tone?
              ANSHP;
                                     __   _________ _
              (CALL BACK IT MECISSAKI TO COMPLETE SCREENING)

     S. a.    Is this residence located in Maryland,  Virginia,  Vathin|ton,  DC
              or son* other place?

                              Maryland  ITI  (GO TO b)                                     <6>

                              Virginia  ITi  (GO TO b)

                         Washington, DC  ITI  (GO TO 7)

                      Sosw other place  PH  (CO TO 6)

         b.    In whet county is this residence located?
              ANSWER : _ (RECORD ANSWER AND CODS BELOW)
                   MARYLAND                                                                l?>
                     (Anne Aruadel,  Baltinore  including city, Carroll,
                     Charles, Harford, Howard,  Nontgonery,
                        Prince Georges  ...................................  J3  (GOTO 7)

                   VIRGINIA
                     (Arlington, Fairfax. Loudon, • od  Prince William.
                     Also include Alexandria  city, Fairfax city and        __
                     Falli Church city ..................................   ITI  (CO TO 7)

-------
     6.   I'm sorry but we   are  not interested in talking  to people (at Pay  telephones/at
         business  telephones/in th« • rea where YOU live).  Think You for • svermg my
         question* • nd could I hove your  OOOO  in case my  supervisor want* to check my
         work?
         NAME  	(COMPLETE SCREENING STATUS IN PART D ONLY)

     7.   •    This telephone is located • tsresidence in • • rea where we* re  interested
             in talking to people .    Are  you under  18 years of • ge or over 18 years  of
             • ge?

                            under 18    JT]  (GO T0 b)

                            over 18     FT1  (GO TO c)


         b.   Whoa will  I  be • bLo to talk to  SOMOQC  over 18 years of • ge who lives in
              this household?     ANSUER:	
              (CALLBACK  IF NECESSARY TO  VERIFY ANSWERS AND  COMPLETE  QUESTIONNAIRE.  IF
             ALL  RESIDENTS ARE  UNDER 18 PROCEED WITH ANT RESIDENT 14 OR OLDER.)

         c.   How swny telephones with different nuabers, not  • xtensions,  service  this
             household?                 ,    ,    |                                           (8-9)
         d.    Hay I have your oaaM in case sqr  supervisor  wants to check eiy work?
              NAME           _____                   (PROCEED TO QUESTIONNAIRE BUT RETURN
              TO AND CODE  SCREENING STATUS AND  QUESTIONNAIRE STATUS AFTER QUESTIONNAIRE
              ADMINISTRATION.)
D.   y<^fprfi»c STATUS  COPES

     Eligible Identified/Screener Completed  ........................ l"oT!

     Not a Working Telephone Nuaber ................................ ("oTi

     Pay Telephone ................................................. I 03 i

     felines* Telephone  ............................................ CoTI                 (20-11.

     No Answer After Repeated Cello  ................................ foTl

     Telephone Not  Located In l*lt. /Wash. S. M.S. A .................. FSTi

     lo IligibU  toistadiat  Anlliblt Adit Itptattd  CtUi  ......... llTi

     Refused To  Answer Sereenint Questions  , ........................ 1 97 i

     Ut|U|i brtiit .............................................. I II i

     Other [[[ m   I


E.   QUESTIONNAIRE STATUS CODES

     Questionnaire Coexisted  (No Mil  followup) .................... L2U

     Questionnaire Completed  (Mail  followup)  ....................... I °2 '

     Questionnaire Partially  Completed ............................. l"03j
                                                                    -^— ^                 (2 w» " ~
     Lan|uage Sorrier  .............................................. I 04 i

     Questionnaire Refused  ......................................... I *1 i

     Other .................................................. - ...... I 061
                                                                    COLS. U'73 'blank

-------
                 CHESAPEAKE BAY  TELEPHONE  SURVEY QUESTIONNAIRE
                    Conducted by Research Triangle Institute
                         for  the University of Maryland
A.   IDENTIFICATION
          CASE  ID
     Interviewer  Initials
     Date Interview Completed I    I   I     |   |
                              Month    Date
B.    INTRODUCTION

     As I  said Earlier,  researchers «t the University of Maryland «re current-
     ly studying citizens'  use of the Chesapeake Bay.  I will  • sk you some
     questions regarding your recreational  use of the Chesapeake Bay,  parti-
     cularly tt  the beaches.   I flso have tofsk  some questions which will
     enable the  researchers to fstablish profiles of typical users fnd non"
     users  of the Bay.  There  is no direct  benefit from taking part in this
     study tnd  you have the right to refuse  to taswer  fay or til of the  ques~
     tions  or discontinue your  participation at any time.   The information
     that you provide will be combiaed with that  provided by other people who
     participate in the survey  to •ssure complete confidentiality• nd your
     same will not be  released or revealed  to 0nyone other than  authorized
     project staff.  The results of  this survey may be helpful in  Effectively
     Allocating mosey to cleaning up  the Bay.

-------
RECREATIONAL USE OF  THE  CHESAPEAKE  BAY

1.    Have you  orfnymembers of your  household used  the Chesapeake Bay
     for recreation in  1984?

          No	FT! (GO TO  Q.  2.)

          Yes  ..... JT|  (GO  TO Q.  4.)

2.    Do  you or any members of your household intend to  use the Chesapeake
     Bay for recreation during the  rest of 1984?

          No	Qj (GO  TO Q.3. )

          Yes	jTj (GO TO Q.4)

3.    What tre  the reasons  you tnd members  of your household have  not used
     and do not intend to use the  Chesapeake Bay for recreation  during
     1984?  (CODE ALL RESPONSES  GIVEN INTO-THE CATEGORIES BELOW)
                                                           CODE  IF  GIVEN

     •\   Not interested in water related  recreation  	   I 2  I

     b.   Unable for health reasons	   I 2  I

     c.   Coiti too auch	   I 2  I

     d.   Takes  too much time to get  there  (too far to           	
          travel)	   I 2  I

     •\   Unacceptable water quality  	   I 2  |

     f.   Too atny jellyfiih	   fTj

     g.   Too crowded	   ( 2  I

     h.   Have not had • chance (too  busy)	   I 2  I

     i.   Other	   l~2~l

      (GO TO Q.  8.)

-------
     What* ctivities did or will you  (and/or members of your household)
     participate in while using  the Chesaoeake?  (READ EACH OF THE FOLLOW-
     ING AND INDICATE  PARTICIPATION FOR EACH ACTIVITY.)
                                                              NO    YES

     a.    Fishing	0 .  • • TJ]   |T"i

     b.    Swinaning	 .  . . TT]   ITl

     c.    Boating	  Qj   [Tj

     d.    Hunting	•  • «.LLJ   IT"!

     ».    Beach Activities	•...-...• LJj   I 2 I

     f.    Sightseeing	o= ...CD   QQ

     g.    Other	  CD
5.    This  next  question 01SO pertains  to YOU and •«•!»*" Of
     household.   During 1984 did  0ny of you  or will 0ny of you...   (READ
     THE  FOLLOWING.   )
                                                              NO    YES
     »    Visit beaches, on the Eastern Shore  of the
          Chesapeake, for^xample shores  close to              	   	
          Cambridge, Salisbury or Chestertown?	I 1 I   I 2 j

     b.    Visit beaches  on  the  ocean, such »s                  	   	
          Ocean  City?	« •.. « | 1 I   I 2 \
     c.    Go iwianing from •boat  in the Chesapeake?  ....    I 1 I   I 2. I

     d.    Go swiming  in public or private  stunning pools?  .    I 1 I   I 2 I

     e.    Visit beaches on  the Western  Shore of the
          Chesapeake, for •xample beaches near Baltimore,       	   	
          Annapolis,  Prince Frederick or Lexington  Park? .  .  .  (JJ   I 2 1

     (IF  YES TO PART  e.,  ASK f.   IF  NO  TO PART t, GO TO  QUESTION 8.)

     f.    During visits  to Western Shore beaches did
          nr will  anyone tttend but not go                     ,	.   ,	
          in the water fflt     reMOn?	m   Qj
                for any reason?

(IF YES TO  PART f, GOTO Q. 6.  IF  NO TO PART f, GO TO  Q.  7.

-------
What* re  the reasons you or others  do not go in  the  water during
visits to the Western Shore beaches?

Do  (READ EACH PART AND CODE NO OR  YES.)
                                                          NO    YES
a.    You or they believe the water is                      	   	
b.

c.
Can
You or they believe there t-re too many
jellyfish 	 , 	
You or they have some other reason 	
you tell me which Western Shore beaches you (and
have visited in 1984 or plan to visit during the rest

. . i_u 23
. . Cu Qu
your family)
of this year?
(CHECK A/0 OR YES FOR EACH BEACH LISTED. )


a.
b.
c.
d.
e.
f.
8"
h.
i.
j.
k.
1.
m.
n.
0.
P-
q-
r.


BEACH
Sandy Point St. Park 	
Fort Siallvood 	
Bay Bridge Beach 	
Herrington Harbor 	 *
Kurtz Pleasure Beach 	
Caop Merrick 	
Breezy Point Beach 	
Chesapeake Beach 	
North Beach 	
Rod and Reel Dock 	
Point lookout St. Park 	
Elm'f Beach 	
Horgantown Beach 	
Miaii Beach (Baltimore) 	
RockyPointPark 	
Conrad' sRuthVilla 	
PorterNewPark 	
Other (SPECIFY)

VISITED
NO YEs
..mm
..mm
. .QJ m
. . LU UJ
..mm
. -Lll LLJ
. . m LLJ
. -LU LLJ
. tix: m
• LU LLJ
. CD m
. . m m
. m m
..mm
..mm
..mm
..mm

QJ CD

-------
       Have you (or members of  your  family who live with your)  »ver changed
       your swimming participation  in  the  Chesapeake because of  changes in
       the  Bay's water quality?

             No	|T"l  (GO  TO Q.  9.)  .                             (4f

             Yes  (stopped) .  .  f2~|

             Yes  (started).  .  FT]

       a.    In what year  did you  (or  members of your  family)  last change
             your  swimming  habits in the Chesapeake because  of changes in
             the Bay's water quality?
                      Year.
                                                                               (4
        We  would like  to  find out how people currently perceive the water
        quality  in the  Chesapeake Bay.

        • \   Lto you consider  the water  quality in the  Chesapeake to  be
            w  cceptable or unacceptable  for  swioBini • rid/or other water
            tctivities.

                 Acceptable ....... jTj                                  (4

                 Unacceptable  .....  ^ m
        b.    Do  you believe the water quality varies ft different  beaches
             • ions the Western Shore of the Chesapeake?

                 No ........... [Tj (GO TO CHECKPOINT I.)            (5

                 Yes  .  .   . '. ...... JTl

        c.    (IF YES, SAY:)  In  general, which statement best describes your
             beliefs?

                 The water quality is better North of Annapolis .  . !JJ         , c
                                                                               < «r

                 The water quality  is better  South  of Annapolis .  .  I 2 |
                          INTERVIEWER CHECKPOINT I

REFER TO  QUESTION 9.A. .

WAS THE WATER QUALITY IN THE CHESAPEAKE RATED AS UNACCEPTABLE?

        No	Qj  (GO TO  Q.  11. )

        Yes	[Tj  (GO TO CHECKPOINT TABLE.)

-------
                               CHECKPOINT TABLE

  ENTER THE LAST DIGIT OF THE  CASE  ID NUMBER HERE.
                                                           CIRCLE AND
  IF THE LAST DIGIT IN                                 USE THIS AMOUNT IN
  THE CASE  ID NUMBER  IS                                    QUESTION  10

          1	$5.00
          2	$10.00
          3	$15.00
          4	$20.00
          5	$25.00
          6	$30.00
          7 ....             . .                  . .          .  $35.00   "
          8 ....:::::..:::::::..::::. $40.00
          9	'	$45.00
          0	0  .,  .  .  ..$50.00


     10.   You  indicated that  in  your opinion  the water quality in the  Chesa-
          peake is unacceptable for (visaing.   Would  you be willing to pay
           (AMOUNT FROM CP TABLE)  in txtra state or federal  taxes per year,  if
          the water quality were improved  so  that  you found it  • cceptable to
          •via  in the  Chesapeake?
               No. .  .  .".
               Kes .  .  .

               Don't  know
D.    BACKGROUND INFORMATION

    11.    The next few questions tre  tbout you tnd your household. How many
          of tach of the following types  of people live  in  your household?
          (READ EACH OF THE  FOLLOWING AND ENTER THE NUMBER OF  EACH TYPE.  )


          *.   Adults (age 18 tnd older)	


          b.   Children between the tjges ofl4*)nd 18	
          c.   Children under age of 14
     12.   What  best describes your status in  the  household?
          a.    Grandparent	I  1 j
          b.    Pirent	jTj
          c.    Child. .  .  .  ,	fTi
          d.    Other relative	GQ
          e.     I live flone or with unrelated individuals .  .  .15!

-------
     13.   How many years have you (and your family) lived  in  either  Maryland,
          Virginia,  or  Washington, DC?
               Number  of years  .  . I   I   |
     14.   Do you or any other members  of your household own (READ THE FOLLOW-
          ING) ...   '
• \
b.
c.
d.
'e.
f.
Are
• ar
stuc
»
b.
c.
d.
»
f.
NO
a boat? 	 « i Qj "|
a boat trailer? 	 	 01 1
fishing tackle (rod, reel, 	
tackle box, etc.)? 	 i 01 i
a recreational vehicle (RV)?. . . ; 01 j
• •wining pool? 	 | 01 t
Other recreational items (SPECIFY)
! Q.
ron
YES
Q)2J
DDL
Pol"
COLS.
ron
ron
PbT",
you one of the principal waie tamers in your
ner but not the principal wage earner, or «re yo
lent or retired?
One of the principal wagetarners in the family
A wage earner but not the principal wage earner
Homemaker 	 	 . . .
Retired 	
Student
Other (SPECIFY)





DK RF
|~98^ |TH
- rw i~9Ti
rgTi r?7j
ilFi p97~i
73-80 - » CASE Oil
OD OD
3ES QD
[Hj LIE
household, «wage
u •homemaker, •
	 CTj
	 ijj
	 (Tj
. fTi
	 * •* '
IT I
                                                                                (55-
                                                                                (67-

                                                                                (65-
                                                                                (71-
                                                                                 (1-2
                                                                                 (3-4

                                     		     (5-6

     15.
INTERVIEWER NOTE:   ASK  QUESTIONS  16 THROUGH  19  IF CODE  1 OR 2  IS MARKED  IN
QUESTION 15 OTHERWISE  GO TO QUESTION 20.

    16.    How many hours  do  you usually work per week?

                                                                                 ff-.
               Hours
    17.    How many paid  vacation days  wi 11 you have • Itogether in 1984, in-
          cluding those you've Already  taken?
               Vacation Days .  .  ]_

-------
    18.   For •typical recreational  outing, if you did not go, could you work
         • tsome paying job  instead?

              No	fTi

              Yes	|"Tj

    19.   If  you  could  have worked,  what  hourly wage might you have  been  paid
         specifically for the hours you worked?

         a.   $3.35/hour	"oTi

         b.   $3.36 - $5.00/hour	foTi

         c.   $5.00 - $7.50	r5T>

         d.   $7.50  -  $10.00/hour  .  . .  •  m

         ».   $10.00- $15.00/hour ....  foTj

         f.   $15.00 • $20.00/hour ....  ("581

         g.   $20.00 - $25.00/hour  ....  Qoj

         h.   Over $25.00/hour	I 11 i

         i.   Don't knew	i 98 !

         j.   Refuse	i 97 i

    20.   Are  there *ny (other)  major wage  tamers in your family?

              NO	  Q"^0010 Q-25. )

              Yes	jTj

The next few  questions tre tbout the other  major wage tamer.

    21.   How many  hours does he/she usually work per week?


              Hours  	

    22.   How Mny  paid vacation days will he/she  have   Itogether in  1984?


              Vacation  Days  .   .  I 1

    23.   For the  typical  recreational outing,  if he/she did  not go could
          he/she work-at some paying job initead?

              No	T

              Ye»	m

-------
    24.    If he/she could have worked,  what  hourly rate  would he/she  have  been
          paid specifically for  the  hours  worked?

          • \    $3.35/hour	j 01 .

          b.    $3.36  -  $5.00/hour	roTj

          c.    $5.00 - $7.50	f03"l

          d.    $7.50  -  $10.00/hour  ....  \W\

          e.    $10.00 -  $15.00/hour  ....

          f.    $15.00 -  $20.00/hour  ....

          g.    $20.00 -  $25.00/hour  .... \Wi

          h.    Over $25.00/hour	0m

          i.    Don't know	|~98l

          j.    Refuse	I 97 I

    25.    We need  «n •stimate  of your household's incoae for •!! of.  1984.  I
          will read • series of income categories.   Please stop «e when I  read
          the  category which best describes the total  aaount of income  •!!
          members  of  your household  will receive  during  1984.

                           INTERVIEWER CHECKPOINT II

ENTER  THE LAST DIGIT OF  TIE CASE ID NUMBER  HERE.  	

IF THE LAST  DIGIT                                         START READING THE
      IS                                                  ANSUER CHOICES  AT
l,3,s,7,9	a. less than $5,000
                                                                  AMD ASCEND
2,4,  6,8,0	1. over $100,000
                                                                AND DESCEND
          • \   less  then $5,000 .  .  .  I 01 . |.

          b.   $5,000 to  $10,000  .  . I 02 I  h.

          c.   $10,000 to $20,000 .  .  rbTl  i.

          d.   $20,000  to  $30,000 .   .  \W\\.

          ».   $30,000 to $40,000 .   .  PoTI k. '

           f.   $40,000 to $50,000 .  .  foe"!  1 .

                              Don't  know

                              Refused  ....
$50,000 to $60,000  .  .  101

$60,000 to $70,000  .  . flTl

$70,000 to$80,000  ,  • I 12 !

$80,000  to  $90,000  .  . ! 13 I

$90,000 to $100,000.  .  :"TTj

Over $100,000  ....  ~

 .  .  3E

 .  • iTTi

-------
26.    CODE SEX  BASED  ON  NAME, PREVIOUS  ANSWERS/REMARKS OR  >SKi  > - you
      female  or male?

           Female .....   fTT
                             	,
           Hale	I 2 ;

27.    Which racial  group  do you identify  with?

           White	• Olj

           Black	'_02 .»

           Oriental	• °3J

           Other (SPECIFY)	I"0*1

           Refused	"  I 97 i

           Don't know	1 98J
                      COLS.  28-73 * blank •
                       COLS. 74-80  = CASE02i

-------
28. •.    This  is  the last  question.   We would like to  send  short  ques-
          tionnaires about  the Chesapeake Bay to people through the mail.
          We would include •postagepaid 0 nvelope  to  return  the  completed
          questionnaire,  so it would not cost  anything  to mail it back to
          us.
               Would  you be  willing to receive and  complete such  a
          questionnaire?


          No	JTI (GO TO  C.)

          Yes	(Tj (GO  TO b.)

     b.    What  is your mailing tddress?

          (VERIFY  NAME) 	
                           (P.O.  Box/Street number • nd name)


                          CityStateZip"

          ENTER CASE ID NUMBER
          Thank you  for taking time to • nswer our questions.  Your respon-
          ses will  be very helpful in  determining the status  of twinning
          • nd other •ctivities on the  Chesapeake Bay.
          IF  YES TO  28*. ALSO SAY:  When the questionnaire comes through
          the mail, please complete •nd return it «s quickly »s  possible.

-------
                                 Appendix C

                 The User Survey and  Sampling  Procedures1


    This  section is devoted to  a  description of the samplingprocedures  used
in  a survey  of Chesapeake Beach  Use  conducted in 1984.    Unlike  the data
used in analyses of boating and fishing in  Chapters 5 and 6, the  data used in
this chapter were collected during an earlier budget period of this cooperative
agreement.     Great  care was  taken with the sampling  frame to  improve
confidence in  the results.   Because  the survey itself  dsimportant to the
project,  we describe the content  and  procedures  extensively.   Copies of the
survey instruments can be  found in  Appendix C.

    From May  26, 1984 to August  19,  1984, Research Triangle  Institute  (RTI)
interviewed individual on the  western  shore beaches in Maryland. The  study
population consisted of all residents of the Baltimore  and Washington, D.  C.,
SMSA's, age 14 or older, that used  these beaches for recreation  in 1984.  More
specifically,  the population waa  limited to  recreational users of the following 12
beaches:


                                    	Strata	
                                    Geographic             size

10  sandy Point                        north.                large
2.  Point Lookout                      south                large
3.  Fort  Smallwood                    north                small
4.  Miami                             north                small
5.  Rocky Point                        north                small
6.  Elm's Beach                        south                small
7.  Bay Ridge                         south                large
8.  Kurtz                             north                small
9.  Breezy Point                       south                small
10.  Rod & Reel                         south                small
11.  Morgantown                         south                ml  1
12.  North Beech                        south                Oman
    Four  hundred and  eight individuals were interviewed at  the beach  to
learn of their  recreational patterns  and perception  of  water quality  at  these
beaches. These  individual were randomly selected from sample beaches and
days. The  sampling design  can be described as a  two-stage stratified sample
in which  a probability sample ofbeaches and days was selected,  and a random
systematic sample  of persons was interviewed at each sample site (day-beach
combination) .
1  The discussion of the  sample is composed  ofselected  excerpts from Devore,
 McDonald,  Myers and Williams,  Chesapeake  Bay Beach  Use Survey.  Research
 Triangle Institute, 1984.


                                     139

-------
    A sample  of at least 300 completed interviews was intended so that relative
sampling  errors  would be approximately 11 percent  or less for  estimating
proportions  of .30 or larger.    This assumes  that  a design effect  of
approximately 1.5 might result  because  of  clustering  (a beach-day unit
constitutes a cluster of individuals)  and  unequal, weighting.  A self -weighting
sample was  sought but  difficult to obtain because of  errors  in size measures
(projected number ofusers  for a beach-day unit) .

    The number  of sample primary sampling units (PSU)  waa baaed   on the
requirement  to obtain at  least 300 completed  interview, such  that each PS U
would  involve  approximately one-half day  ofinterviewing for a two-person
team.   To match this design, each PSU was  randomly designated  a.  a morning
or afternoon interview  period, beginning at 1000  and  1300 hours, respectively.
Hence,   an  average of eight or nine  completed interview, waa expected for each
of the  36 PSU's  (see Table 4,1 for  sample  allocation).


Stratification and Sample Selection

    Primary  sampling unit. (PSU's) consisted of  beach-day units (1,204 PSU's
=  66 days x 14 locations).   The 14  locations consisted  of 11 beaches plus three
beach  locations at Sandy  Point -  partitioned  into three segments because of
its relatively large size  and usage.  Beaches  were stratified into north  (Sand y
Point and beaches north)  and couth (the prior listing  of beaches indicates the
stratification).   The beaches wore  further stratified  by size (expected weekly
usage). Large is  defined as  greater  than 7,000 visitors  per week in north
beaches and greater  than 3,000  in  the south beaches. Additionally, days were
stratified into weekdays, Saturdays,  and Sundays  (and holidays) .  The strata,
their population and sample PSU counts are indicated in  Table Cl.
                                  Table  C.  l
              Population Counts  and PSU Allocation by Stratum8
                   Anticipated Number of Beach
                   Visitors May 26 to
                   August 19,  1964  (thousands)
                   Sat  Sun  Weekday   Total
Number of PSU's  (beach-day
units)  Allocated to  Bach
Stratus

Sat  Sun   Weekday  Total
North/large
Nbrth/srBll
south/large
South/ small
Totals
77
6 7
37
20
201
110
6 4
43
24
261
106
94
44
17
259
293
243
124
61
721
3
3a
?2
22
10
5,
42
2
2
13
5
4
22
2
13
13
13
6
6
36
 1  Population  counts are  baaed on cite interviews prior to the  survey,  and PSU
  allocation is proportional  to number of visitors with  the constraint that
  greater or  equal to two per cell are selected.
 20ne fewer resulted in these cells because of random subsampling needed in
  the latter  part of  the survey period.    (A slightly larger number of
  interviews  than  expected  were obtained early in the survey.)
                                     140

-------
    The selection of PSU's within strata involved equal probability y for days
and probability proportional to size  (expected number  ofvisitors)  for beaches.
Within  selected PSU's,   individual were selected on  site with  equal
probabilities.   Approximately  eight or  nine  interviews were  needed from each
PSU. The procedure used to satisfy this objective  waa for the  interviewers to
estimate   the  number  of beach  users just prior to  starting  time  for
interviewing.    By   comparing  this  estimated  count  to a table prepared
specifically for this survey,  the interviewers obtained a sampling interval  and
a random  starting point.    By  using the interval  and starting point in a
pro-designated pattern  for the entire site,   a valid systematic sample of users
was obtained.


Sampling  Weights

    Whenever observation units (in this caae the individual users)  enter the
sample  with  different  probabilities,   weights   must  be  used  with the
observation to obtain unbiased estimates for the study  population.   Because
of the complexities  in  an  intercept-type survey,  selection probabilities  are
often not known. In the present survey, however, PSU selection probabilities
are known,  and final-stage selection  ofindividual  can  be  reasonably estimated
 (even though a systematic sample of visitors at  the site  is taken), their
chance of selection will vary with the  amount of  time spent at the  beach
during the  sample day as well  as number of visits  to this  and other stud y
beaches during the  survey period.  A  sampling weight for the jth individual
is calculated  aa follows:
where

    p    =    * +*j «  the  "selection probability"  (expected number of hits)
     1          >+> for PSU (i) ;

    d    =    the number of daya during  the  survey period  for  the particular
              type of day being sarpled (d = 13 for  Saturdays,  for example);

    n    =    number of PSU's being selected for the stratum (12 strata);

   s,     =    size measure for PSU  (i)  = the expected number of visitors for
              that type  of day (S+ = tSt);

  p, j     *    sampling rate of users within sample PSU  (i);  and
  and    =    factors to adjust for  number of houra user  spent at  the beach
, .
 '
              on the sarple day  and number of trips  to the study beaches
              during the sumar,  respectively.
                                     141

-------
Note that  fjj  and fjj  are baaed  in part on intentions.    In  fact, f» j is
particularly uncertain if the interview is being conducted  early  in  the  season.
To verify the accuracy of these intentions both  to  obtain selection multiplicity
and to estimate related statistics,  intention-based  questions  were verified b y
telephone at the end of the season for interviews  taken early in the  survey.


Screening and Confirmation

    A screening form waa designed to identify as eligible  respondents only
those people living in certain counties of Virginia  or Maryland, or Washington,
Do C.   In addition,  the  screening process  ruled  out as  ineligible any person
who was  under 14 years of age;  however,  if  the  selected  sample  individual was
over 14 but  less  than  18 years of ageand in the  company  of someone he/she
lived with who  was over 18 years of age, the interviewer deferred  to the
older  individual.

    The  User Intercept Survey  Questionnaire was  designed  to  record and
collect  the following:

         Frequency  of visits made  to beaches  on the western shore of the
         Chesapeake

         Activities that the respondent  (and his/her family) participated in
         when visiting beaches

         Activities not participated in and  the reason why they were not

         Coat related to a  typical  trip  to  each beach that the  respondent  had
         visited  since January 1, 1984

         The respondent's perception of the  quality of the beach  and the
         beach facilities 11 each beach wtth which he/she was familiar

         Factors  that  influence a  respondent's decision to visit or not visit a
         beach

         The respondent's  willingness to  continue to viait  the  sample site if
         coots related to the use of the beach  were to rise.

    In  addition,  a series of demographic questions to enable  analysts  to
establish profiles  of beach  users were also included in the questionnaire.

    A Confirmation/Intention  Contact sheet  (used in  conjunction  wtth  a Record
of Calls sheet)  was  developed  not  only  to confirm the number of visits to
sample  beaches reported in the  Intercept  Survey Questionnaire,  but  also to
ascertain the number of  visits the  respondents intended to make  at  any
sample beach  for  the  rest  of 1984.   The  confirmation/intent ion  contacts were
only made with those respondents  who  had provided adequate information  to
contact  them by telephone.
                                     142

-------
Data Collection Procedures

    The  data collection schedule  originally called for a total of 36 visits  to the
12  sample beaches, with an expected yield of 300 or more  completed interviews.
This  schedule  was  later revised  and the subsequent number of  total  visits  wae
reduced.   Table  C.2  shows the original data  collection schedule.    Those
sampling days which were eliminated have been denoted by  an asterisk.   The
resulting   number of  visits  are  summarized as   follows  (AM  indicates
interviewing beginning  in the morning  at 10 o'clock,  and PM  indicates
interviewing beginning  at 1 o'clock in the afternoon):
Beach
1
2
3
4
6
7
8
9
10
11
12

Beech
Name
Sandy Point
Point Lookout
Fort Smallwood
Miami
Elms
Bay Ridge
Kurtz Pleasure Beach
Breezy Point
Rod and Reel Dock
Morgantown
North Beach
TOTALS
AM
Visi ts
5
1
1
0
0
0
0
0
1
1
1
12
PM
Vimits
6
1
0
3
0
3
1
1
0
1
0
18
Total
Visits
11
2
1
3
0
3
1
1
1
2
1
30
      For the User  Intercept Survey,  each field  interviewer waa  given a
      of the names,  addresses,  and ID  numbers for each PSU  (beach)  in the
assignment, along with  area maps with  the beaches  marked.   In addition, each
interviewer wae given a sketch of each  beach.   Other materials included were
the  table, to  determine sampling rates  and listing forma forcounting and
selecting sample individuals.    The interviewers  always worked as a  team,
splitting up only to interview eligible  respondents.

      The  field  interviewers were asked to review their materials and
determine the moat efficient route oftravel to reach each beach.  Upon arrival
at a beach, they had  first to check that they had correctly  identified and
located  the precise boundaries of the area.   Once  this waa verified,  the
interviewers  estimated  the number of sample individual on the  beach  ,
spending no more  than 30 minutes in doing ao.   When  the estimate was
determined,  they  looked  up  that number  in the table of sampling intervals.
Marking  the estimate  and sampling interval  at  the  top  of the listing form, they
next consulted their list of random numbers to determine the  number of  the
first person to  be  interviewed and  marked that in  the  space provided on the
listing form.   They  circled the  number  of  the randomly selected start person,
counted  the proper  interval, and circled the  last  interval  number  as  the next
selected person.   This  activity continued until they had  gone through the
entire list interviewing the selected individuals.


                                    143

-------
                                  Table C.2
          Chesapeake Bay  Beach Use Study Data  Collection Schedule
Date
May 26
27
28
29
30
31

June 1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30

July 1
2
3
4
5
6
7
Beach Beach
Name NLnber
Pt . Lookout
Morgantown
Miami
Miami





Sandy Pt. (NofS)










Sandy Pt. (NofS)
Morgantowm
Sandy Pt . (NofS)
Bay Ridge

Sandy Pt. (NofS)

Rocky Pt.
North Beach
Sandy Pt. (E)
Rocky Pt.





Breezy Pt.

sandy Pt. (SofS)


Sandy Pt. (E)



2
11
4
4





1










1
11
1
7

1

5
12
I
5





9

1


1



Sample
Time
1300
1000
1300
1300





1000










1000
1300
1300
1300

1300

1000
1000
1000
1300





1300

1000


1300



Date
July 8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31

Aug 1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19

Beach Beach Sample
Name Number Time
Kurtz



Bay Ridge

Rocky Pt.

Miami



sandy Pt. (SofS)
sandy Pt. (SofS)
Rod & Reel





North Beach
Sandy Pt. (SofS)




Pt. Lookout

Bay Ridge
Ft . Smallwood

Rocky Pt.



Kurtz
Pt . Lookout

Sandy Pt. (E)

sandy Pt. (SofS)


Sandy Pt. (NofS)

8



7

5

4



1
1
10





12
1




2

7
3

5



8
2

1

1


1

1300



1300

1300

1300



1300
1300
1000





1000*
1300




1000*

1300
1000*

1000



1300*
1000

1300*

1000


1000*

*0rijinally scheduled  but  subsequently eliminated.

-------
      Appendix D





User  Survey instrument
            145

-------
I.
ASSIGNMENT INFORMATION:


CASE   ID  #:
                                 SCREENING  FORM
                       Date:QI]
                              Month
                                                 Date
                                                               Tiae:|
     SITE #:J. I 1
     INTERVIEWER  NAME
                                Weather:   s
                                                                 (5-8)

                                                       PC | 2 |  C m  R 1 4
II,  INTRODUCTION:
          Hello,  my name is
                                           Researchers tt  the  Univer-
sity  of Maryland  are currently  studying citizens' use  of the Chesapeake Bay
and they are looking for certain  types of people to aoiwer*questionnaire.  I
would like  to t sk you  4  questions  to determine if  you «re the type of person
we are  looking for.   First. .  .

III. ELIGIBILITY:                                                (20)

     A.    Do you  currently live in Maryland,	   I 1 I  (Go  to B)

                                    Virginia,	   (^2 I   (Go  to B)

                                    Washington, D.C, or	   fTl  (To to C)

                                    Some other place?	   (4 )   (Go  to F)

     B.    What  county do you live  in?  	 (ENTER, COMPARE AND  CODE)
          MARYLAND
           (Anne Arundel, Baltimore  including city,
          Carroll,  Charles,  Bar ford, Howard,                     (21[
          MoatjoMry, Prince George'i)	   | 1 |    (Go to C)
          VIRGINIA
           (Arlington,  Fairfax,  Loudoun,
          Prince William)	   | 2 I   (Go to C)

          SOME  COUNTY NOT LISTED	   fTI    (Go to F)

     c.    How old tre you?                                       ^2)
                   Less than 14 yein of age	   | 1 |   (Go  to F)

                   Over  14, but less than 18 years of age	[ 2 |   (Go  to D)

                   18 years of age or  older	   | 371   (Go  to G)

     D.    Did you  cone  here  . ..with other people including
                               someone you live with  who  is      (13)
                               18 years of tge or older?	j 1 I   (GotoE)

                             . ..other people but none  you
                               live with who  is  18 years
                               of age or older?	   (JJ    'Go  to G'

                                          	   fTl    (Go toG)
                  . ..by your self?	  {_3J

I would like  to  talk to  the person  you live  with who  as 18 years of
age or  older and is here  with you today.   Point that  person out to
me or tell me how and  where to  find  him or her.    (WHEN  YOU  LOCATE
THE  PERSON,  DETERMINE  ELIGIBILITY BY  VERIFYING OR REASKING ANY OR
ALL OF  THE  QUESTIONS  IN PART  III.  BEFORE  YOU BEGIN THE QUESTIONS
READ THE  INTRODUCTION TO THE  NEW PERSON.)

-------
          (CODE 2 or 3  IN Iv BELOW AND SAY:)

          You  (do not live in the  area/belong to  the age group) that  we wish
          to survey  on  this study.  Will you give me your name •nd telephone
          number in  case my supervisor wants to  check on ay  work?

          NAME	  TELEPHONE (      )
           (CODE  I  IN  IV BELOW AND SAY:)

          You  are the type of person we want to interview for  this  study.  The
          questions  will  take 15 to 20 minutes.   Can we start  now? (AFTER
          MEETING OBJECTIONS SAY:) Before I ask  the questions I need to keep
          a record of  the  people I  speak to.   What is your name,  address, and
          telephone number?   what time did you  enter  the  beach?   What  time
          will you leave?
          NAME
TELEPHONE
          ADDRESS
IV.   RECORD OF  ELIGIBILITY  (Screening  Status)
TIME ENTERED BEACH :|  I   |   |   |
                      (^-17)
TIME LEAVING BEACH:!I   1   I   I

                      (18-21)
               Eligible  	   PHi  (INTERVIEW AND/OR COMPLETE  PART V)

               Ineligible  because  of
               Residence	   f02J  (STOP.   DO NOT  INTERVIEW.  READ
                                            STATEMENT FROM III F, RECORD
               A8e                      foal INFORMATION AND TERMINATE CONTACT.

               Refused Screening  ....   I 971

               Language Barrier	   [^5) (RECORD DETAILS
                                              IN NOTES SECTION BELOW)
               Other (Specify)	   [06)

               	                                   (22-23)


v.   RECORD OF INTERVIEW STATUS

               Completed	   foil    (EDIT CASE AND  SHIP TO RTI)

               Partial/Breakoff.  ...   QL    (RECORI) DETAILS IN NOTES
               Refused	   ("97J     SECTION BELOW)

               Language Barrier ....

               0ther:(SpeCi£y)	   —'                             (24-25)
VI.   NOTES :
        COLS.  26-72 * blank
        COLS.  73-80  =  SITE/CASE  1

-------
                         CHESAPEAKE BAY BEACH  USE  SURVEY

                    Conducted by Research Triangle Institute
                         for  the  University of Maryland
     NOTIOE • la/onatioa coataiatd o« tai* fern that vould pacaxt idmtificatioa of «ay iadividiul bit b»«e
           eolltcttd with * fturtat** taat it vill ba bald U ttriet caafidaac* «ad vsad oaly by pcrioai
           «|«iad IB or for purpetM of Uii mi*»j.  All result* will b« (iBBUtixad (or ft«up* of pcopl*
           •ad a* lafonutioa about iadividiuli vill »• nlt«i«d.
A.   IDENTIFICATION
     SITE NO. 1
CASE ID
     Field  Interviewer Name
     Date Interview Completed 	
                               Month
B.    INTRODUCTION

     As I said Earlier,  researchers «t  the University ofMaryUadtre current-
     ly studying citizen*'  use of  the Chesapeake Bay.   I will  • sk you «o«e
     questions reprdinf your  recreational use  of the Chesapeake Bay, parti
     cularly «t  the beaches.   I flso have  to *sk  IOM questions  which will
     tnable the researchers to establish profiles  of  typical users of the Bay.
     There U no direct  benefit froa taking  pert in this study  • nd you have
     the  right  to  refuse to  e-nswer tny or til  of  the questions  or discontinue
     your participation tt tny time.   The  information that you provide will be
     combined with  that  provided by  other  people who participate  in the  survey
     to essure complete confidentiality tnd your uoe  will not be released or
     revealed to  anyone  other than Authorized project staff.    It may be
     necessary to  recontact  you  later to verify the number of times you
     visited beaches during  the  season.   The results  of  this survey  may be
     helpful in • ffectively • llocating money to detains up  the Bay.

cT    BEACH UTILIZATION

10    How  many  of the following types  of  people ere  in  your party today?


          Adults (tge 18 tad older)	

          Teenagers • ge 14 years  to  18	

          Children  under  age 14 	
2.    Are you at  the  beach today with your family?  By  family I mean people who
     • re related  to you by  blood,  marriage,  or adoption, tnd who  live with
     you .

          NO	CD                            (7)
          Yes	fn

-------
3.    (SAMPLE  SITE NAME]  is  situated on the  Western Shore of the  Chesapeake
     Bay.   Since January 1. 1983, have you (and members of  your family who
     live with  you)  (READ THE FOLLOWING)

                                                      NO        YES
     a.    Visited  beaches  on the Eastern
          shore of the Chesapeake, for
          example,  shores close to                    	        	
          Cambridge, Salisbury, or Easton?  . ,  .  .  •  I LJ  •  • •  I 2 I  (  S)

     b.    Visited  beaches on the ocean,                          	
          such tt Ocean City?	[Tj        121  (9)

     c.    gone  swimming from •boat in
          the Chesapeake?	•     m

     d.    gone winning in  public or                   	        	
          private twining pools?	Ill     •   [_2J  (22)
                         THIS SPACE INTENTIONALLY BLANK

-------
4    When  you  visit  a  Westers Short beach such at  this 000, do you  (aadMabcrs  of your tastily who live with  You)  parti
     pat*  ia (READ THE rOLWwTMG)

                                                                           4A.  How sway  adults    41.  How  aaay childre
                                 NO (Nut activity      YES  (Aok  Q.s4A..     participate ia         * teenager* part
          ACTIVITY                    »t qiwatioa)             & 4B.)            (ACTIVITY)             cipate in (ACTIV
                                                                                     *                      b
         iia(  or  wading?
b.    hooch • ctivitioo-
     tuabathing, picnicking,
     ihelling?  	
c.    beach •ctivitios-
     playgrouad?  . .
d.   boating?
     Other AetiTitits?
     (Specify)
. CD
                                                        .  . (T] .
                                 [T]	 . Q] .  (i 7)



                                 Q]	[T]  .  (22) .
                                                        ca
                                  CD	EH .  <*•?)

                                                                                      i   J
                                                                                                       U5-7?)
                                                                                                      CD
                                                                                                       (J0-J2 )
                                                                                                        3s-s6

-------
S    If you (or any mtmbm of  your family who
     describe  why?   (READ THE FOLLOWING)
with you) don't «o  into  the water, which •* the following. »tJ
                                                                             5A. How ewny adults    SB. How uoy  eh
                                                                                 do not jo into         *t«««ugtri
No (co to aext TCS (Ask Q.a 5A. •» the water becauae |o into the
R1ASON reason) *SB.) (XCASOM)T because (WA
+ *
»\
b.
e.
d.
»
f.
Don't kao«
No bathb
at beach
Person's
precludes
Tou or th
water is
Too Deny
Other res
(SKCIFY



,n,i
ous. facilities 	 ^ m 	 ^^
hailrk —
it m [2] (J2)
.«, («-*:>
	 1 1 1
Ui-«4) (6S-66)
mm W« t f e**e> 	 — 	 	
dirty/vollutel . . . ITI 	 121 . (.87) 	
(S
jellrfiah . FH 	 TS1 . ( ;) 	
'OD 	 ..m 	 m.c.) 	 r
	 1 I !
53S)" " " " r^-7:)
COLS, 72-SO •» SITUCASE Ul 1

~n i i i
n 	

-------
6.    How many  years have you  (and members  of your  family who  live with  you)
     been  coming to the beaches on the Western Shore?


          Number of years	I  I  I  (12-12)

7.    Have you (or anyone  in your family who live  with you) changed your
     swimming habits in the Chesapeake because of  changes in  the Bay's water
     quality?

          No	ITI  (GO TO Q.8)

          Yes (stopped)	ITl

          Yes  (started)	f~3~l                           (13)

     A.    In what year  did you  (or members of your family) last change  your
          svinaing  habits   in  the Chesapeake because  of changes in  the  Bay's
          water quality?
                                m
          Year	Ill                        ( 14-15)

The next  few  pueftioni deal with the frequency with which.you (and your
family) visit  beachesfnd the  cost related to-visits aadetofach beach.

HAND RESPONDENT CARD A.
                        THIS SPACE INTENTIONALLY BLANK

-------
I.   look  »t  the  beache* listed oa thia card.  Pleaae give M the letter  that  appear* beside the MM of each  I
     you (*a4 your  fully) riiitid ilact January 1.  1913.  Pita** iaclude  this  vi»it to thi* beach.   (CHECK (J)  i
     VISITED  AMD THEJI ASK QOtSTIOMS SA aadTF
        BEACH VISITED
                            Q.t. Chock if
                                  visited
                                                       Q.IA. How many trip*  did you
                                                             (aad  your faaily) aiake
                                                             to (BEACH)ia 1983?
                                                                                         Q.SB. How Nay tript
                                                                                               (aad your famil
                                                                                               to (BEACH) in
»\   tody Point St.  Park


b.   Tort Seallwood ...


c.   Bay Ridjt Beach


d.   Kcrrisftoa Harbor


».   Kurtx Pltacur* Beach


f.   Caa» derrick ....


«.   Brwry  Point Beach


h.   Chtsapeaka Beach


t.   lorth Beack
                                  m  a
                                                                        .   .   0748)	
                         .....  ma
                                           	rn  (22-231	

                                      •  (mu)	DTI •   (27-28  )	

                                      .  (M)	
                                                                                                  CD. 6
                                                               jlfl  .0/^1


                         ..... GO .  («)  ......... 1X1(42-43   )   ........ m.   . I
                              0
                                                                      •  «M*)


                                      -  («)  .........  . m    , .'(28-ss)


                                  Q] .  (M)  .......... lCf\
                                                                                                  Ill  .
J.   lod awl Reel Dock	~. (cl)	(M-«J| .  |	171   I - (s


k.   Poiat Lookout St. Park .  . . . m . (»»)	I   I   I . $7-88  )	
1.    £!•'• Beach                  . £1 .  (J)	f\\  -  (  24)  .  .

•.    Horiaatova Sooth	[T] .(»)....	I   I   I  . (  7-4  )   .


                                                                       .   02-13  )
                                                                                      COLS.  71-80 »»»SITI/CAi


                                                                                                         .(
                                                                                                  •in-
a.    Hlaai Sooth (Baltiaore)  .  .  .
                                      .  CD
0    Rocky Poi»t Park	m 3	


p.   Coarad* • R«th Villa	[T] . («)  .  . '.	


4.   Porter Mew Park	m   .   (26)   ....   ll-l	(2748  •)....».. IN. (
                                                                      .  (27-18 ) .......... m    | .  (
r.   Other (SKCX7T)
                                                                     32-33  )	

-------
9.    Do YOU  (and you* faaUly) P1"  'o "*• "T trtp* to «*T of *** h«ach«i listed ou thia card  during  the  reMiader of
     sutJBer?
                                             °	      { (GOTO 0.10)                            (38)
                                            Doa't kao«	m  t

                                            Tea	GO

     A.    Which of  these  beaches do You plaa  to visit this sue»erT  Pleaae give aw the letter that appears beside  the
          of  «ach beach (CRECX(V)EACH BEACH(ffMTIOXCO AM)THEN ASK Q.9I FOR £ACH  BEACH.)


                                                                             Q.9B. How  MBJT trips do  you  («nd your
                                                Q.9A. Check  if pica                fMily) iatnd to  take  te  (BEACH)
                      BEACH                            to vi«it                     Skis  i«Mt?



          »\    Sandy  Poiat St.  p a r k	(TI  .  (37)	I  I  I .  (38-3$)


          b.    Fort Saallwood	m  •  up  )	I  I  I .  (
                                                      »

          k.    Point   Lookout   St. Park	fTI  .  (S?)	m      .W-49)


          1.    £!•'• Btaeh                           •  • CD  .  (»>	I  I   I . (72-72)


                                                                                         I COM.  73-SO =  SITE/CASE 03 I



          «.    Heriaatow  Baach                       .  . GD  •  f ')	• •  • '• •  J •  ( *-J )


          n.    HiaaU   Buck  (Baltlaara)	•  • S  .  (4)	I  I   I .(  S-6)


          0.    Rocky  Point Pork	•  • C2  .  (7)	I  I   I • ( 4-9)


          p.    Conrad's    Ruth   Villa	•  • H  • («)          	I   I   I . (JJ-J2)


          q.    Pertat  Htw Park	QQ   (33)	I   I   I  . (,24-1s )


          r.    Other (SPECIFY)

               		JT|.(J

-------
INTERVIEWER NOTE:  ASK QUESTIONS  10A-H SEPARATELY  FOR EACH BEACH CHECKED IN
QUESTIONS 8 AND 9A.

10.   Now think • bout the cost  of  a typical  (family)  trip to  (BEACH)  in 1984.


     A.   What would you say the total  •ntrance  fee per day,  including
         parking,  would  be  for you  (and your family)  »t  fBEACH) ?   (ENTER
         AMOUNT IN COLUMN A BESIDE APPROPRIATE BEACHI.


     B.   What is the typical  coat for you  (and your family) to  travel to
          (BEACH) ? (ENTER  AMOUNT IN COLUMN  B  BESIDE APPROPRIATE BEACH NAME)  .


     c.   How much will  it,  cost you (and  your  family)  per day  in  other
         expenditures such «s  food, hotel or camming fees, tnd entertainment
         • t  (w=)?  (ENTER  AMOUNT IN COLUMN C  BESIDE  APPROPRIATE BEACH
         NAME)  .


     D.   ADD  AMOUNTS  ENTERED  IN COLUMNS A-C  AND ENTER  TOTAL  IN COLUMN D.


     E.   How  much travel tiM does it typically take  you  (and  your family) to
         make •round-trip  fro* your home to  (BEWE)? (ENTER TIME IN COLUMN
         E BESIDE BEACH NAME)  .
     F.
How many miles if it round-trip  from your borne to (BEACH)? (ENTER
NUMBER OF MILES II? COLUMN F BESIDE BEACH NAME) .
     G.   How much tiM do you  (and your family)  typically spend on the beach
         per day when you visit  (BEACH)? PROBE FOR BESTESTIMATE OF THE
         NUMBER OF HOURS AND ENTER NUMBER IN COLUMN  G BESIDE BEACE NAME.
     H*   What is the t verage number of days that you tnd your family  spend on
         • typical trip to (BEACH)?   (ENTER NUMBER IN COLUMN H BESIDEBEACH
         NAME)

-------
TABU 1
QUESTION 10
(Coatiauod) .

B«aCB
k. Point Lookout St.
Park
1. EU' a B«aca
a. Horgaatow Boaca
a. Hiaa* B*aca
(Bolt-se)
• Bocky Poiat Park
p. Coarad' a Butt Villa
a,. Pertar N«v Park
r. Other (SPECI1T)

A
Eatraaea
r««

( \
(2t-2t)
ffl
rn
m
( 2.1 )
ra
(*«-*•)
B
Travti
Coots


*


3-


4 j

(JO-tt]

(

j

1


J-

M.

*.


4 j






~~~)






1
(30.33)


( s-


*


1 1
U0-JJ)

Ei

I

Otter
VM



1 1


\
7-11)


(34-31)
| |
(
1 |

|

|

1

|

(3




\
r-ii)

4-1
1
1)
1 1



_r
(J4-U)

(

(j

1
7-11)

1-3
|

u
Total
Co

] 1


(11-17)
J 1


(39-44)
r
<
\

~\ 1

1 J '





|

|

1
(Jt-17)
} 1

(39-44)

1


(12-17)
\ 1

(J9-44)
| |

E
B-T
Tiaw

1
(l»-l>
( '
P3
ui-17,
0«-;» )
Z«-3*)
r
B-T
Hilts



\
(20-21)


«7-W)
COLS.
COU.
G
Tiaw
00
Batch
m
(i4-2$:
\

(Si-Si)
H
0*
I
Tr
n
Q
15-73 sblaak
73-80 = SITE/CASE I

1
(20-23)



(47-U)
COU.
COLS.

| |
(u
(
-23)

(51-52)
5
y
M-72 • aak
73-sO • SITE/CASE

• 1
(20-23)
1 1
(47-50)
COU.
COU.
m
(24-25)
n
T5F
s
39-72 • blaak
73-80 • SITE/CASE




(*0-2 3)



(47-50).

|
(24.



•25)

(S1-S2)
a
n
53:

-------
HAND RESPONDENT CARD B


11.   Please read the water quality characteristics listed  on this  card and
     rank  them • ccording to how  important they 9 re to you (ind your family)
     when  deciding  whether  ornot  to visit  • beach.   Please  rank them on a
     scale of 1  to 5, with  1 being  least  important  and 5  being  most  important.

                                                           RANK

          a.   Presence of  cloudy water	'.  	   fSS)

          b.    Presence of  floating debris or oil	_____   (£S)

          c.    Presence of  odors		    (s7)

          d.    Presence of  jellyfish	•   .    (58)

          ».    Presence of  seaweed  §nd other
               • quaticplants	• 	   (59)


HAND RESPONDENT CARD C


12.   Each  sussttr jellyfish, tlso called sea nettles, • ppearin the  waters of
     the Chesapeake Bay.   lookt-t the  statements on this  card  tnd tell me
     which one  describes  your  (and your faaily'i)  behavior after  jellyfish
     tppear.

          ».    Stop going to  the  beich when                 ^_^
               jellyfish appear	I 1 I
          b.    Still go to the beach, but less often  .  .  .  __G    (60)

          c.    Continue to 80 to  the beach, but             	
               don't fo into the w«ter	l_3 I

          d.    Don't consider  presence  of
               jellyfish atoll	
                                                  COLS. 61-72  =  blank
                                                  COLS. 73-80 = SITE/CASE 12

-------
THIS PAGE  INTENTIONALLY BLANK

-------
13*   A.    HAND RESPONDENT CARD A.  We would like to find out bow beach  users
        perceive the water quality, the quality of  the beach facilities, the
          beach quality, *nd crowding ft the beaches on  the Chesapeake  Bay.
          Please give me the letter that appears beside the name of tach beach
          on  this card that  you  (and your family)* re familiar with.  By
          "familiar,"  I mean that you  know  something tbout  the  beach t ither by
          having been  there  or you have heard  •bout  it through some other
          source.   Please  include this beach  in the  ones that you mention.
          (CHECK COLUMN A FOR EACH BEACH MENTIONED.)


ASK QUESTIONS 13B-ESEPARATELY  FOR EACH BEACH MENTIONED.


     B.    Do you consider the  water quality tt  (BEACE) Acceptable or unaccep-
          table for twinning • rid/or other water •ctivities?    (CHECK APPRO-
          PRIATE CODE IN  COLUMN B BESIDE BEACH NAME) .

     C.    Wftafabout the beach facilities,  such fsspace,  bathhouses,  tables,
          swinging guards, etc.  ft  (BEACH).  Would yourste it  «s • cceptable
          or unacceptable?   (CHECK APPROPRIATE  CODE IN COLIMT C BESIDE  BEACH
          NAME)  .

     D.    Think*bout the beach itself ft (BEACH).   Do you feel  that the
          quality  of the   beach is 0 cceptable  or  unacceptable?    (CHECK
          APPROPRIATE  CODE  IN COLUMH D BESIDE  BEACH  NAME) .

     E.    Do you feel that  the  size of the  crowd  tt (BEACH)   is  0 cceptable or
          unacceptable?   (CHECK  APPROPRIATE  CODE IK COLUMN  E BESIDE BEACH
          NAME)  .

-------
    TABU  2



QUESTIONS  13A-I


Beach
Sandy feint
st. Pttk . .
Fort Saallvoo

Bay Kidie
Beach . . .

. Harrington
Harbor . . .
Kurtz Pleaaui
Beach
. Caap derrick
. Breexy Point
Beach . . .

u Chesapeake
Death . . .
. North Beach
Rod and Keel
"tick ....
;. Point Lookout
St. Park . .
.. Ela'a Beach
t. Morgaatown
Beach
t. Miami Beach
(Baltimore)
i. Rocky Poiat
Park ....
p. Coarad't But
Villa . . .
q. Porter leu
Park ....

r. Other
(SPECIFY)


r— A — 1
Zheck if
••'liar
ffr
55
TT )"
Oj.
UJ)
S»
i
a-
en
CM)"
rri
rn
9-
•ffi.
' '™3
•$,
•a-
•3v
•ffr
•53
(IT)"


• CD.
• (28)
i 	 1 	 : 	 r
Water
:eeptab 1 1
.m..
. CD • •

. m- •

.m-.
.03-
.m-
.23-

. m .
.CD-
. m .
. m .
.23-
-23-
. m .
. m .
.23-
.23.



. [73 •

lualitT
oaccep table
• O3 ci)
. 23 CM

. 23(">

.23(17)
. m (221
. 23 
.23 (w)
• El (S4)

.23 (22)
• ffl (44)
Cti ini
Acceptable I UnacceptabLi
.CD.
. CD-

.23-

.23-
.m-
.23-
.23-

.23-
.03-
.O] (48) • 23 •
.Q3 (s4)
. 03 (««)
. S3 (64)
. 23 <«)
1 COU
. m ( 4)
. m c *)
. 1 2 1 (I*)



. m (»)

CD-
,03.
.03.
,23-
. 71-80 >
.03.
.23-
.23-



.OD.

p ( s )
.CE t:s

. GDCi

. m (20
.23 (2s
.CD (30
. m CM

.23(40
.23 (4s
. 1 2 1 (-50
.CD(sS
. me,
. rn (5^
..mo,
» SITE/ CASI
. .mc^
. . 23 t:
. . 23 di



. . n~' ca


-------
                             INTERVIEWER CHECKPOINT 1

  CHECK THE  RESPONSES TO QUESTION SB  FOR THE NUMBER OF TRIPS THE RESPONDENT
  HAS MADE TO TEE SAMPLE  SITE IN 1984 AND QUESTION 9B FOR THE NUMBER OF'
  TRIPS THE RESPONDENT PLANS TO TAKE TO  THE SAMPLE SITE DURING THE REMINDER
  OF THIS SUMMER.  ADD THE NUMBER ENTERED FOR BOTH QUESTIONS AND RECORD IT
  HERE.  •»
                     TOTAL NUMBER OF 1984 TRIPS =
CHECK COLUMN D IN TABLE 1 FOR THE TOTAL COST OF A TYPICAL 1984 TRIP TO
SAMPLE SITE .  ENTER THE AMOUNT
                                                                     THE
                     TOTAL COST OF TYPICAL 1984 TRIPS
  INSERT THE TOTAL NUMBER OF 1984 TRIPS ANDTHE TYPICAL 1984 COST PER TRIP
  IN THE APPROPRIATE SPACES WHEN QUESTION 14 IS ASKED.

  ENTER THE LAST DIGIT OF THE CASE ID NUMBER »
       If  last digit it
             0
             1
             2
             3.
             4
             5.
             6
             7.
             9..
                                           Ust  $ amount  below in
                                           Question  14
                                                  $50.00
                                                  $45.00
                                                  .$40.00
                                                  .$3s.OO
                                                  $30.00
                                                  $25.00
                                                  $20.00
                                                  $15.00
                                                  $10.00
                                                  $s.OO
14.  According to your responses  to previous questions,  you (and  your  family)
    take •bout   (TOTAL NUMBER OF  1984  TRIPS)  to this site  per year *t an
    approximate  cost  of  (TOTAL  COST  OF TYPICAL  1984 TRIP)  per day. If your
    costs.per day were  to riit  by  (USE INITIAL AMOUNT FROM TABLE ABOVE),
    would you  still visit this site? Keep in mind  that the  costs  of visiting
    other sites on the Chesapeake or participating in other  •ctivities would
    rauia the
                         NO.
                         YES
                                             1
                                             2
(22)
     IF  NO, DECREASE THE DOLLAR
     AMOUNT IN $5.00 Increments
     UNTIL A "YES" ANSUER IS GIVEN.
     WHEN A YES ANSUER IS GIVEN,
     RECORD DOLLAR AMOUNT BELOW.
                         14.
                                          IF YES, INCREASE THE DOLLAR
                                          AMOUNT IN  $5.00 INCREMENTS
                                          UNTIL A "NO" ANSUER IS GIVEN.
                                          WHEN A "NO" ANSUER IS GIVEN,
                                          RECORD DOLLAR AMOUNT OF LAST
                                          "YES"  RESPONSE.

                                           DOLLARS
                                 (22-25)

-------
ENTER THE LAST
If last digit is
1 ....
2. ...
3. ...
4 ....
5
6
7 ....
8 ....
9
0. ...

INTERVIEWER CHECKPOINT 2
DIGIT OF THE CASE ID NUMBER =
Use $ amoun
Questions
	 $ 5
	 $10
	 $ 1
	 $20
. $25
.$30
	 $35
	 $40
. $45
	 $50


t below in
L5 and 16
.00
.00
5.00
00
.00
00
.00
00
.00
.00
15.   Jellyfish «re frequently identified  «s •nuisance  to tvimm. Would you
     (and your fully)  be willing topty  (AMOUNT FROM CP 2) per yeas io • xtra
     state or  federal taxes  if jellyfish could  be t laminated «s •nuisance
     without tny tdverse ecological tffects?
         NO	roil

         Yet	0 CH3                         (26-2?)

         Don't  Know ....  ^  m
                           INTERVIEWER  CHECKPOINT 3

    REFER TO COLUMN B IN TABLE 2 (QUESTIONS 13A-D)

    WAS THE WATER QUALITY OF  THIS SITE RATED AS UNACCEPTABLE?

         No	fTl •»  (GO TO Q. 17)

         Yes	0m"              (28>
16.   You indicated that the water quality «t this  site  is  unacceptable for
     •wining.  Would you be willing to pay (AMOUNT FROM CP 2)  in extra  state
     or federal taxes if the  water  quality were improved so that- you found
     it Acceptable to swim here?
                                                            (29-30)

-------
17.   The next few  questions  • re •bout you  •nd your household.  How many of
     each of the following types of people live in  your household? (READ EACH
     OF THE FOLLOWING  AND ENTER THE NUMBER OF EACH  TYPE).
         a.
          Adults  (age 18 ind older)
         b.   Children between  the ages of 14 and   H
          c.
          Children under fge 14
                                                 m
                                                           (32-32)
                                                                 (33-$4)
                                                                     (ZS-ZS)
18.
HAND CARD D.   Which  of  the  relationships  listed on this card best de-
scribes your  status in your household?
         I.   Grandparent  	

         b. .  Husband	

         c.   Wife  	

         d.   Child 	

         »    Other Relative 	

         f.   I  liveflone or with
              unrelated individuals  . '. .  .
                                          m
                                                                 (37)
                                          m
                                              EJ
19.   How many.years have  You (and your faaily)  lived  in either Maryland,
     Virginia or Washington;  D.C.?

         Number of years	J.J.J.                   (38-39)

20.   Do  you or tny other lenber of your household own  (READ THE FOLLOWING)

                                        NO      YES      DK       RF
• \    tboat?	

b.    •boattrailer? .   .  .

c.    fishing tackle (rod,
     reel,  tackle box,
     etc.)?	

d.    •recreational
     vehicle (RV)? .  .  .  .
                                           m  *  * m *

                                          cm=mzl*
                                                             TT

                                                             TT
                                           m
                                              *  *
                                                  m
e.
f.

• swming pool? 	 I 01 1 . .
Other recreational
iteu (SPECIFY)
roTi M
rsn . .
I 02

FoT
f02
                                      .  C2D -. dS •  • EEI  •  • QD
                                                          9i"i.  . nrn
(40-4:

(42-43



(44-45


(45-4?
                                                       .  r?n .  .
                                                       .  nn .  . irn
                                                                   (so-s:

-------
21.   Are you one of  the   principal wage earners in  your household, «wage
     earner  but not the principal wage • arner,  or «re  you •homemaker,  a
     student orretired?

         a.   One of  the  principal  wife earners                    	
              in the family .„»,,,,,,,»,,	I  1 I

         b.   Awage»arner but  not the  principal                    	
              wage  earner	6	0 IT]

         c.   Homemaker	I  3 I   (S2)

         d.   Retired	0 CD

         e.   Student	CE3

          f.   Other  (SPECIFY) 	
INTERVIEWER  NOTE:   ASK QUESTIONS 22 THROUGH 24 IF CODE  1  OR 2  IS  MARKED IN
QUESTION  21.   OTHERWISE GO TO QUESTION 26.

22.   How many hours do you  usually work per week?


          Hours	I  I   I                                (53-54)

23.   How many paid vacation  days  will  you have W lto@er  in 1984,  including
     those you've  Already  Uken?


          Vacation  Dayi	1  I   I                                (5s-56)

24.   For your typical trip  tothebeach,  if you hadn't gone to the beach,
     could you have  worked •tsoat job instead?


          No	0m
                                                                        (57)

-------
25.  if .You could have worked, what hourly  wage Bifht you have  been paid
     specifically  for  the hours 'you Worked?


          a.   $3.35/hour	I 01 I

          b.    $3.36-$5.00/hour

          c.    $5.00-$7.50/hour 	

          d.    $7.50-$10.00/hour  	
          ».    $10.00-$15.00/hour	% |JKI

          f.    $15.00-$20.00/hour 	 • QJE

          g.    $20.00-$25.00/hour	0 OH

          h.    Over $25.00/hour	0 QD

          i.    Don'tKnow	I 98 J

          j .    Refuse	Q I 97 I

26.  Are  there • ny  (other) major wage»arners in your family with  you today?


          No	• El   (00  TO  Q.31)
                                ^   i                                     (60)
          Ye«	0|   (ASK Qf  .  27-30)                     '   '

The next few  questions • re about  the  other major waget-arner  who is with  you
today .

27.  How  many hours does he/she  usually work per week?


          Hours	IT!                                  (51-?72)

28.  How many paid vacation days will he/she have 0 1 together  in 1984?
          Vacation  Days  .  .  .  .
(83-64)
29.  For the  typical trip to  the beach, if he/she  hadn't gone to  the  beach,
     could he/she have worked • t some paying job  instead?
          1)0 .........                                           
          Yes

-------
30.   If he/ibt could have worked,  what hourly rate would he/she have  been  paid
     specifically for the hours worked?



          • \    $3.35/hour	foTl


          b.    $3.36-$5.00/hour	'  02  I


          c.    $5.00-$7.50/hour	0 "m


          d.    $7.50-$10.00/hour  	  •  m


          ».    $10.00-$15.00/hour	f"bT"l                      (66-$7)


          f.    $15.00-$20.00/hour  	  •  m


          g.    $20.00-$25.00/hour  	  9m


          h.    Over $2 5. 00/hour	CITI


          i.    Don'tKnow	9m


          j.    Refuse	
31.   HAND  CARD E.   Which one  of the categories  on this  card best describes
     your  (family's)  incoae during  1984?   Pleate  give me  the letter that
     • ppears beside  the category?


          •	• ££/          s	 . • . . rrb~i

          b	ron          h	rrn
          c	V  m          i		
          H                  am             i                   rrn   (68-69)
          d	9  m          J	UU

          e	rsn          *	nn

          f	ron          i	rm

                        DK	!~98~1

                        RF	r?n
32.   CODE SEX BY OBSERVATION.
          Female	 I 1 !


          Male	FTI

-------
33.   CODE RACE  BY  OBSERVATION.

          White

          Black	m

          Oriental	FT1

          Other (SPECIFY)
                             .  .  .  . m
                                             I COLS. 72-80  = » SITE/CASE 141
Thank you for participating in our survey of btach users. Your responses  will
be helpful  to  u»0nd  hopefully  to the Sutetnd Federal government* in deter-
•ining the itatui of sviaaiBf and other •ctivities on Chesapeake Bay.

-------
                       TELEPHONE CONFIRMATION QUESTIONS
FOR EACH BEACH LISTED ASK 34. A.  AND B. BEFORE GOING TO  KEX1 BEACH .
34.   A.
     B.
During 1984 how many times have you (or  members  of your  family
live with you) visited  (ENTERED NAME  OF BEACH)?
     RECORD ANSWER IN COLUMN A AND ASK:

During the remainder  of 1984 how many  times will you (or members
your family who live  with YQU) visit (ENTER NAME OF BEACH)?
     RECORD ANSWER  IN COLUMN BAND GOTO NEXT BEACH.

                                                        34.B
                                                        PLANNED
                                                        VISITS
                                                                           who
                      of
                         BEACH
          •\   Sandy Point St. Park ....

          b.   Fort  Smallwood	

          c.   Bay Ridge Beach	

          d.   Her ring ton Harbor	

          »    Kurtz Pleasure  Beach ....

          f.   Canp Merrick	  .

          g.   Breezy  Point  Beach	

          h.   Chesapeake Beach	|	|_

          i.   North Beach	I	j_

          j.   Rod tnd Reel Dock	I   I

          k.   Point  Lookout  St.  Park  .  .  .

          1.   Elm'sBeach		

          m.   Horgantown  Beach	I   I   i

          n*   MiMi  Beach  (Baltimore)  .  .  I   I   I

          o.   Rocky Point Park	I   I   I

          P«   Conrad's  Ruth Villa .  .  .  .  	

          q.   Porter New  Park	I   I   I

          r.   Other (SPECIFY)               I   I   I
 (1-2)

 (3-6)
 (13-14)  •

(17-18) -

 (21-22) »

 (25-28)  •

 (29-30)  •

 (33-34)"  '

 (37-38)"

 (41-42)  •

 (45-46)  •

 (49-50)"

 (53-54) •

 (57-58)  •

 (61-62)-

 (65-66)*

 (69-70)
                                                        OZJ
                                                        en
(3-4)

(7-8)

(11-12)

(25-16)

(19- 20)

(23-24)

(27-28)

(31-32)

(35-36)

(39-40)

(43-44)

(47-48)

(51-52)

(55-56)

(59-60)

(65-64)

(67-68)

(71-72)
                                                   I  COLS.  73-80 SITE/CASE IS I

     AFTER PART  rSAY:  Thank  you  for taking part in the  survey tnd tilkinf to
•e today.  'The  information you-have provided will be very helpful in determin-
ing use of the  Chesapeake.

-------
                       CHESBEEBKE BAY BEACH USE  SURVEY
                           TELEPHONE CONFIRMATION
                           SCRIPT AND INTRODUCTION
     I am calling to speak   with (     INSERT NAHE	 ).  IF NOT THERE,
FIND OUT WHEN HE/SHE WILL BE AVAILABLE. WHEN AVAILABLE  SAY:

     My name  is (INSERT NAME ) »nd I'm calling for  the University of  Maryland.
On (INSERT DAY AND  DATE you were interviewed  »t  (INSERT  THE NSME OF THE  BEACH
WHERE INTERVIEW TOOK PLACE)  by • lady working  for the University on • survey
about  the Chesapeake  Bay.   Do you remember that  interview?  (IF YES PROCEED.
IF NO,  TRY TO REFRESH  THE RESPONDENT'S MEMORY  BY TELLING  HIM OR HER THE TYPE
OF QUESTIONS ASKED.   IF STILL  NO,  TERMINATE THE CALL).

     I'm calling today to confirm some of the information you reported the day
you were interviewed.   I want to  know which beaches you  have visited or plan
to visit  on  the Western Shore of the Chesapeake  Bay.   I will read through*
list of  beaches •nd • sk •bout  each one individually. First,

     GO  TO QUESTION 34.A.

-------