LA-9699-MS
Los Alamos National Laboratory is operated by the University of California for the United States Department of Energy under contract W-7405-ENG-36.
     A Regional Recreation Demand
                     and Benefits Model
                         Los Alamos National Laboratory
                         Los Alamos.New Mexico 87545

-------
                        An Affirmative Action/Equal Opportunity Employer
 This  work  was  supported  by  the  US  Environmental Protection  Agency,
 Freshwater Division, Environmental Research Laboratory.
                     Edited by Lidia G. Morales, S Division
This  report  has  been  reviewed  by  the  Corvallis  Environmental  Research
Laboratory,  US Environmental Protection Agency, and approved for  publica-
tion.  Mention of trade  names  or  commercial products  does  not  constitute
endorsement or recommendation for use.
                                       DISCLAIMER

  This report was prepared as an account of work sponsored by an agency of the United States Government.
  Neither the United States Government nor any agency thereof, nor any of their employees, makes any
  warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness,
  or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would
  not infringe privately owned rights. Reference herein to any specific commercial product, process, or
  service by trade name, trademark, manufacturer, or otherwise, does not necessarily constitute or imply its
  endorsement, recommendation, or favoring by the United States Government  or any agency thereof. The
  views and opinions of authors expressed herein do not necessarily state or reflect those of the United
  States Government or any agency thereof.

-------
                                                EPA ERL-Corvallis Library
                                                    00004341

                                                   LA-9699-MS


                                                   UC-11
                                                   Issued: March 1983
               A Regional Recreation Demand
                     and Benefits Model
                        Ronald J. Sutherland
L
Los Alamos National Laboratory
Los Alamos, New Mexico 87545

-------
                                    CONTENTS

ABSTRACT	   1

EXECUTIVE SUMMARY 	   2

CHAPTER I.   INTRODUCTION 	   5

CHAPTER II.  RECREATION BENEFITS AND DISPLACED FACILITIES  	  11
             1.   Introduction	11
             2.   An Overview of Benefits  in the Recreation  Literature.  ...  13
                  A.   Knetsch	13
                  B.   Mi shan	16
                  C.   Freeman	18
                  D.   Support for the Conventional Measure  of Benefits  ...  20
             3.   The   Theoretical   Underpinning   for   the  Conventional
                  Measure of Recreation Benefits	22
             4.   Consumer Surplus with Multiple-Price Changes,  	  26
             5.   Conclusions	30

CHAPTER III. ESTIMATING RECREATION TRIPS WITH A  GRAVITY MODEL 	  31
             1.   Gravity Model Overview	35
             2.   Gravity Model Input Variables  	  39
                  A.   Fraction Factors (F-0 	  39
                  B.   Trip Production Model (P.)	43
                  C.   Attractions Model (A.) -1	51
             3.   Calibrating the Gravity  Model  	  54

CHAPTER IV.  Estimating an Outdoor Recreation Demand  Curve	61
             1.   Estimating  a  Travel-Cost  Demand  Curve  and  Consumer
                  Surplus:  An Overview 	  61
             2.   Travel-Cost   Demand  and  Valuation   Estimates:    Some
                  Illustrations 	  67
                  A.   Aggregating Recreation Activities	67
                  B.   Substitute Sites 	  68
                  C.   Some Empirical Estimates  for Swimming	70

CHAPTER V.   Survey  Estimates  of the  Willingness  to Pay to Recreate  and
             the Value of Travel Time	74
             1.   Introduction	74
             2.   Direct Willingness to Pay Estimates 	  74
             3.   The Value of Recreation  Travel Time	75

-------
                                 CONTENTS (cont)

CHAPTER VI.   THE SENSITIVITY OF TRAVEL-COST ESTIMATES OF RECREATION DEMAND
             AND VALUATION TO VARIOUS COMPUTATION AND SPECIFICATION ISSUES. .  80
             1.    The Three Issues	81
             2.    The  Sensitivity  of  Travel-Cost  Estimates  to  Various
                  Assumptions	83
                  A.    Functional Form of the First-Stage Demand Curve. ...  84
                  B.    Size of Origin Zone	91
             3.    Conclusions and Implications	96

CHAPTER VII. EMPIRICAL ESTIMATES  OF  RECREATION  BENEFITS OF IMPROVED WATER
             QUALITY IN THE PACIFIC NORTHWEST	101
             1.    Determinants of Recreation Value and Use	101
             2.    Demand and Valuation Estimates for Selected Lakes	106
             3.    Benefits of Improving Water Quality in Streams	109
             4.    Conclusions	116

APPENDIX A.   DATA TABLES
             Table A.I.  Population Centroids, Population, and Counties .   . .119
             Table A.2.  Recreation   Centroids   by   Name,   County,   and
                         Centroid Number	123
             Table A.3.  Recreation Activity Days Produced by Centroid. .   . .128
             Table A.4.  Recreation   Facility  Variables,   Existing  and
                         Potential,   from  Improved   Water  Quality   by
                         Recreation Centroid	132
             Table A.5.  Annual  (1979) Recreation Value by Activity and by
                         County  for Washington,  Idaho, and Oregon	137

APPENDIX B.   HOUSEHOLD SURVEY QUESTIONNAIRE	140
             Table B.I.  Frequency Distribution of Recreation  Trips Using
                         1980 Household Survey Data	143

ACKNOWLEDGEMENTS  	 144


REFERENCES	,145
VI

-------
                                     TABLES

Number                                                                      Page

   1    Regression Estimates of Gamma Specification of the Decay Curve.  ...  42

   2    Regression Estimates  of Exogenous and  Endogenous  Attractions  (in
        Natural Logs) 	  53

   3    Trip Interchange Matrix 	  55

   4    Demand   and   Valuation   Estimates   for   Swimming   in   Selected
        Washington Centroids	72

   5    Direct Willingness to Pay Estimates per Recreation Day	  76

   6    Direct Estimates of the Value of Recreation Travel Time 	  78

   7    Annual  Valuation  Estimates  for   Boating  in Selected  Washington
        Centroids Using a Semilog and Double-Log Functional Form	86

   8    Demand  and  Valuation  Estimates Using  a Semilog  and Double-Form
        and Endogenous Quantity Demanded	88

   9    Travel-Cost  Valuation  Estimates  Using  a  Semilog  and Double-Log
        Form and a $0.25 Price  Increment	90

  10    Semilog  and  Valuation  Estimates Using  10-Mile  and 20-Mile Origin
        Zones	92

  11    Estimates  of  Quantity  Demanded   by  Centroid  Using  Semilog  and
        Double-Log  Forms  and  Various  Definitions  of  Origin Zones  (in
        Thousands of Visitor Days)	94

  12    Double-Log  Valuation Estimates Using 10-Mile  and  20-Mile Origin
        Zones	96

  13    Annual  Recreation  Demand  and  Value of  Selected  Lakes  in  the
        Pacific Northwest (1979 Dollars)	107

  14    Annual  Recreation  Benefits  of  Improved Water  Quality in Streams
        by Activity  and by County for Washington, Oregon, and  Idaho	113
                                                                              VI 1

-------
                                     FIGURES



Number                                                                      Page



   1    Recreation Demand and Benefits:  The Knetsch Analysis 	  14



   2    Consumers' Surplus:  The Case of Perfect Substitutes	18



   3    Demand for Two Recreation Sites 	  19



   4    Consumer's Surplus Using Ordinary and Compensated Demand Curves ...  23



   5    Measuring Benefits with Multiple-Price Changes	28



   6    Decay Curves for Camping, Fishing, Boating, and Swimming	42



   7    Estimating Consumers' Surplus Using Bode's Rule 	  66



   8    Price-Quantity Observations for a Recreation-Site Demand Curve. ...  83



   9    The Effect of Substitute Sites on Demand and Value	103
 vm

-------
                 A REGIONAL RECREATION DEMAND AND BENEFITS MODEL

                                       by
                              Ronald J.  Sutherland

                                   ABSTRACT
     This report  describes  a  regional  recreation demand and benefits model  that
is used  to  estimate  recreation  demand  and  value (consumers'   surplus)  of  four
activities  at  each  of  195  sites  in Washington,  Oregon,  Idaho,  and  western
Montana.   The  recreation activities considered are camping, fishing,  swimming,
and  boating.   The  model  is  a  generalization  of  the  single-site  travel-cost
method of estimating  a  recreation demand curve to virtually an unlimited number
of sites.  The major components of the analysis include  the theory of recreation
benefits, a travel-cost recreation demand curve, and a gravity  model  of regional
recreation  travel  flows.   Existing  recreation  benefits  are estimated  for  each
site in the region and for each activity.  Recreation benefits  of improved water
quality in degraded rivers and streams in the Pacific Northwest are estimated on
a  county basis  for Washington,  Oregon,  and  Idaho.   Although  water quality is
emphasized, the  model  has the capability of estimating demand and value for new
or improved recreation sites at lakes, streams, or reservoirs.
     This  research documented in  this  report was  started  in  June  1978  and
completed in September 1982.

-------
                               EXECUTIVE SUMMARY








     A regional  recreation  demand and  benefits model  is described  and  used to



estimate recreation demand and  value  (consumers'  surplus) of four activities at



each  of  195  sites  in  the  Pacific  Northwest.    The  recreation  activities



considered  are  camping,  fishing,  swimming,  and boating.   The essence  of  the



model is that  it  generalizes the single-site travel-cost method of estimating a



recreation  demand  curve to  virtually an unlimited number of  sites.   The major



components  of  the  analysis  include  the  theory  of   recreation  benefits,  a



travel-cost recreation  demand curve,  and a gravity model of regional recreation



travel flows.   Recreation benefits  of improved water quality in degraded rivers



and  streams  in  the  Pacific  Northwest are  estimated  on  a  county  basis  for



Washington, Oregon,  and Idaho.    The  model  is also  illustrated  by  estimates of



existing  recreation benefits  of selected  lakes where  water quality  is good.



Potential and existing recreation benefits are high for sites located near large



urban areas and relatively low for rural sites.  The model provides quantitative



estimates  of  these benefits.   Although water quality  is emphasized,  the model



has the capability of estimating demand and value for new or improved recreation



sites at lakes, streams, or reservoirs.



     Recreation benefits are defined  as willingness to pay, or alternatively as



consumers'  surplus,  and measured as  the area  under the recreation site demand



curve.  An  improvement  in  water quality at one site implies an outward shift in



the  demand  curve  for that  site and a  redistribution of demand from substitute



sites.  The issue of the proper measurement of benefits at an improved site when

-------
there  are  displaced  facilities  is  analyzed  with  the  conventional  utility



maximization  model  for  consumer  behavior.   The  analysis  shows  that benefits



measured under  a  single  demand curve are net benefits and automatically account



for any displaced facilities.



     Two major  limitations  of  the travel-cost  method of  estimating recreation



demand are  its  failure to consider substitute sites and the expense of applying



it  on a  site-by-site basis.   A  gravity  model  is  used  here to  overcome each



limitation.  This model distributes recreation trips to every site in the region



on  the  basis  of relative travel costs and relative attractiveness of each site.



The  output of the gravity model  is  a trip interchange matrix that  is the main



input for travel-cost  demand curves for each site in the region.



     The conventional  gravity model is a distribution model, which means that it



only  estimates  the distribution  of trips between  productions  and attractions,



which are  assumed exogeneous.   Because the model does not estimate total demand



at  each  destination,  its applicability is limited for most recreation purposes.



The  gravity  model  is  extended  here  by  estimating  it   iteratively with  an



attractions model.  As a result,  the desirable  properties  of the gravity model



that  determine  the distribution  of trips also  influence  total  demand at each



s i te.



     After  a  demand  curve and consumers'  surplus  are  estimated  for each of 195



sites in  the  region,  a simulation analysis is used to determine the sensitivity



of  the  results to  three computational  and  specification  choices  that must be



made in the analysis.   A semilog specification of a recreation site demand curve



is  shown  to  be  preferable to  a double-log  specification.   Recreation trip



origins  may be defined as a system of concentric  zones,  or  as  each  population



centroid.   Demand  and  valuation   results  are  shown  to  be  sensitive  to  the



definition  of  an origin,   although the  best  definition   is  not  determined.

-------
Quantity demanded at several sites was estimated using travel-cost demand curves



and  compared  to independent  estimates of  quantity demanded.   Errors  in these



quantity  estimates  are  particularly  large  when  a double-log  specification is



used, and the errors also depend on the definition  of the origin zone.



     The  regional  model   is  used to  estimate  recreation demand  and consumers'



surplus  for the four  activities at each of  195 sites in the region.  Demand and



valuation  are  again   estimated assuming  that each  officially  degraded river



becomes  "fishable and swimmable," which is the goal of the 1977 Clean Water Act.



Recreation benefits of improved water quality are  estimated quantitatively on a



county basis and for each of the four  activities.

-------
                                    CHAPTER I



                                  INTRODUCTION








     The Clean  Water Act  of  1977  (U.S.  Congress  1977)  reaffirms  the  national



goal of  eliminating  the  discharge of pollutants into  navigable  waters  by 1985.



This  Act  defines  an  interim  1983  goal   of protecting  fish,  shellfish,  and



wildlife   and   providing  for  recreation.    These   goals—expensive,   perhaps



impossible to attain in an absolute sense—are becoming less feasible because of



the  increasing  political  importance  of competing goals.  The desire to expand



energy  supplies and  to  reduce  inflation  may conflict  with  regulations  that



attempt  to  achieve  a high level  of water quality.   Furthermore,  the benefits to



be  gained  by  achieving the Federal goals may not be sufficient in some  cases to



justify  their costs.



     The  Environmental  Protection Agency   has  begun  to  incorporate  economic



factors  into  its evaluation  of  water  (and air) quality  improvement  programs.



Although  the  Agency  has  not  completed  its  approach  to  defining  economic



efficiency and to performing marginal analyses, there is a clear movement toward



including  costs and benefits  in  the decisionmaking  process.  However,  a major



difficulty  in attempting to  use  quantitative cost-benefit estimates is that the



Agency has  no well-developed  and tested procedures  for  making these estimates.



Specifically, the  marginal  costs of making incremental  improvements  in water



quality  in  streams  and lakes are  difficult to estimate.   Similarly, the Agency



does  not  have  well-developed  and  tested  procedures   for  obtaining  dollar




estimates of the benefits  of improvements in water quality.

-------
     Although several  uses of  water may  be  enhanced by  quality  improvements,

recreation benefits  appear to  be the most extensive.1   Therefore,  this effort

will  focus  on  the  development of  a model to  estimate  recreation  benefits of

improved  water  quality on  a  regional  basis.   The  model  should  possess  the

conventionally  desirable properties  of  reliability  and  theoretical soundness,

but it is also  important that the model be operational.   Specifically, the model

should be able  to estimate dollar benefits with a consistent methodology over a

large  number of  sites,  quickly and with reasonable  cost.   One function of the

EPA,  both  at  their  headquarters   in  Washington,  D.C., and  at  the   regional

offices,  is  to select  from  a  large number  of  potential  sites  water-quality

improvement  projects  that are  to   be  funded.   Single-site  analyses  are   time

consuming  and  expensive and  therefore of  limited value.  The model presented

here  combines  the  gravity  model   with  a travel-cost  analysis  of recreation

behavior  to  estimate  benefits at  any  site  in  the Pacific  Northwest,  which

corresponds  to  EPA  Region  X,  excluding Alaska.

      Although the EPA is  the  intended user of this work,  other Federal  agencies

may find  the model  appropriate for  their  recreation planning  needs.  The Water

Resources  Council   (1979),  through  its  procedures   for  evaluating  costs  and

benefits,  defines   the  evaluation   procedures  for  water-oriented  construction

projects  that  Federal  agencies  are legally  obliged  to  follow.   The  Water

Resources  Council emphasizes  three  points:   (1)  recreation benefits should be

defined  as  consumers'  surplus; (2)  demand should  be measured with  the travel-

cost  method  or  direct willingness-to-pay  approach;  and  if possible,  (3)   a

regional  estimator model should be  employed.  At  present, fewer  than a handful
  According  to Freeman  (1979a),  recreation benefits  are  more than half  of  the
 total  potential water  quality  benefits  and  more  than three times  larger than  the
 next most  significant  benefit.

-------
of models  meeting these  criteria have  been  constructed and  none has received



widespread  acceptance.   The model presented here  uses  the  travel-cost approach



on  a  regional  basis  and  measures  benefits  in   terms  of   consumers'  surplus.



Because  the model  meets  the  criteria  of the Water  Resources Council,  it is



appropriate  for  use  by  those  Federal  agencies  concerned  with  water-based



recreation.



     The construction of new reservoirs and the upgrading of existing reservoirs



may  encourage  additional  recreation  use,  particularly  if  the  appropriate



facilities  are  provided.   The  model   is  designed  to  estimate  the  change in



recreation  demand and value  resulting from an  increment in  recreation  oppor-



tunities.   The  water-based recreation  activities  analyzed here include camping,



fishing,   boating,   and   swimming.    Because   these   activities   are  treated



separately,  in  effect four models are constructed.  The uniqueness of the model



is that  demand  and benefits can  be  estimated  for any site in the region, which



in  this  study  consists  of  Washington,  Oregon, Idaho,  and  western  Montana.



Demand and value  are estimated separately for 195 recreation  centroids and for



each of  four activities.  Because origin and destination centroids can be added



or deleted,  the model is  capable  of analyzing  demand and value for any site in



the Pacific  Northwest region.



     Chapter II provides the conceptual basis for  estimating value and benefits.



Recreation  benefits  are  defined as  net  willingness  to pay  and  measured as



consumers'  surplus.   An  improvement in water quality produces  an  outward shift



in the recreation-site  demand  curve.   The increase  in  benefits  is measured as



the area  between  the new  and initial  demand  curves  and above the market price,



which  is typically zero.



     A critical  step in estimating recreation benefits  for  a  specific site is



estimating the recreation  demand  curve for that site.  Chapter  IV  is a  review of

-------
the  travel-cost  method for developing these  estimates.   The  travel-cost method



has  been used  extensively with a  good measure  of theoretical  and empirical



support.  However,  there are  several limitations of this approach, for example,



the  time  bias,  but the most  serious  problem for agencies requiring analysis of



several  sites  is the  expense and level  of  effort  required  to analyze a single



site.   Vis it-rate  data by origin are  required for each  site, and the data from



one  site  usually cannot  be applied to other  sites.   These data are obtained from



either  household surveys  or  site attendance  estimates,  and  in  either case are



not  readily available.  When identifying projects  to be  funded, the  Agency must



select  from a  large number of  candidates.   The  time and  survey  expense  required



to  estimate a  travel-cost  demand  curve  limits  its  applicability  when  it is



necessary to select a few sites from among a large number of alternatives.  In



this study, the  travel-cost  demand curve  approach is  generalized to include  a



large  number of sites within a region and can  be  applied with  minimum  time and



expense.  The development  and use of regional  estimator models is  recommended by



the  Water Resources Council   (1979)  and  is  also  recommended  by Dwyer,  Kelly, and



Bowes  (1977).   In addition to  economizing  on information, such  a  model  can more



accurately  reflect the influence of  substitute sites.



      The input  data  required  in a  travel-cost demand  analysis include travel



(mileage) costs and visit rates for each  population center  that  sends  visitors



to  the site being  analyzed.  Obtaining  the  visit-rate data  is the main  time and



financial constraint  to  applying the travel-cost approach over a large number of



sites.   A regional household recreation survey  was undertaken  in 1980  covering



each of the three  Northwestern states.  The survey results are  used  to  estimate



the   number of  recreation  trips by  activity  emanating from  each  population



centroid in the  region.  A  gravity model  is  used to  allocate  recreation trips



from each origin in the  region  and from  external zones to each recreation

-------
destination.  The model  includes  155 population centroids (origins) and has 195



recreation  centroids  (recreation  destinations).   The purpose  of  Chapter III is



to develop a regional recreation gravity model.  The inputs of the gravity model



are also  developed  and  these include a trip production model, an attractiveness



model,  trip-length  frequency distributions  and  a  travel  distance  or impedance



matrix.   The  output of  the gravity model is a trip interchange matrix that, for



each  destination  in the region, is the  number of  trips from each origin in the



region.   When those trips  are divided by their  corresponding population,  visit



rates  are obtained,  and they  are  the  critical input  in a  travel-cost demand



curve.  By combining household recreation survey results with a gravity model,  a



model  is  constructed that has the capability of producing travel-cost demand and



valuation estimates for  any site in the region.



      The  main components of the recreation model include the conceptual  measure



of  benefits and  value  (Chapter II), the  gravity model  (Chapter III),  and the



travel-cost  demand  curve  (Chapter  IV).   Chapter  V  is  an examination  of  some



computation  and  specification   issues   involved  in  calculating a  travel-cost



demand curve.   The functional   form of the  demand curve and the  size  of the



origin zone  are analyzed as possible determinants of travel-cost estimates.  In



Chapter VI, the operation of the model is discussed and some applications of the



model  are presented  for both  lakes and streams.  The  first  application of the



model  is an  estimate  of  recreation  benefits at  five  selected  lakes  in the



Northwest.   The lakes  are  selected as  representative  of both  urban and rural



lakes.   Other things being  equal,  benefits  are  estimated to  be significantly



larger in urban than in rural  lakes.  Recreation benefits which would accrue if



the   degraded  rivers  and  streams   in  the  Northwest  were   made  fishable  and



swimmable are estimated on a county basis.   The  model  is also used to  estimate



demand and  benefits  resulting  from improving  water-based recreation areas and

-------
from  constructing new  facilities.   Agencies that may have an  interest  in  this



work  include  the  Soil  Conservation Service, Water  and  Power Resources  Service



(formerly  the  Bureau of Reclamation), Army  Corps  of Engineers, and others that



need  to  estimate recreation benefits  resulting  from water-related projects.  In



a  study in  progress (Sutherland  1982d),  the model  is  being  used  to estimate



recreation demand and  value of the Flathead Lake  and existing river system in



western  Montana.
  10

-------
                                   CHAPTER II



                  RECREATION BENEFITS AND DISPLACED FACILITIES







1.    Introduction



     The proper measure of the monetary value of a recreation site has long been



of  interest  to  academic researchers  and to recreation  planners in  state  and



federal agencies.  The  economic  concept of net willingness  to  pay (or consumer



surplus)  is  now  widely  accepted  as  the  appropriate  measure  of  benefits.



However,  a  complexity  arises when  the  net willingness  to pay  for a  new  or



improved  site  comes  at the expense of an existing substitute site.  If measured



benefits  of the  new  site contain a  large component  of benefits which have been



redistributed  from other  sites,  then  these  estimated benefits  overstate  true



social benefits.



     The  issue of  how to treat benefits  which  are redistributed from displaced



facilities can be  resolved with  basic  economic principles.   The resolution  has



practical importance to recreation researchers and planners.   If benefits can be



measured  correctly by  estimating  net willingness to pay for the new or improved



site  and  excluding benefits foregone,  then  estimating  recreation site benefits



is feasible.   If,  however,  foregone benefits must explicitly be subtracted from



the benefits of a new site, then all relevant substitutes must be identified and



their  demand  curves  estimated.   Such  a  task  is  empirically  difficult.   The



importance of  being  able to value a recreation site, or more appropriately,  the



recreation use  of a site,  requires  that we have  a  concise  definition of these



benefits.





                                                                              11

-------
     Some recreation literature is reviewed in Section 2, where it is shown that



some  researchers   are   not  sure  how  to  treat  displaced  benefits.   Other



researchers  have  constructed  elaborate  econometric models  which  explicitly



subtract  benefits  redistributed from substitute  sites.   The  most commonly held



view is that benefits can be measured correctly by estimating willingness to pay



at  the  new or improved site and  ignoring  shifts  in the demand for substitutes.



A main objective in reviewing these studies is to show the absence of the neces-



sary  justification  for  this   position.   Indeed,  researchers  who  argue  that



benefits  from  displaced facilities can be ignored often derive their support by



quoting each other.



     One   objective  of  this  chapter  is  to  determine   the  proper  measure  of



benefits  of a  new  or improved recreation site when demand for this site comes at



the expense  of existing  sites.   This  chapter  will  serve as  the theoretical



foundation for the benefit  measure used in the regional  recreation demand model.



In   Section  3,  benefits  are   demonstrated  to  be  measured  correctly as  net



willingness  to pay for a new or  improved site and any displaced  benefits can be



ignored.   The  main  objective  here is  to  provide theoretical  support for this



view.   The appropriate measure of benefits can be  derived from  basic  economics



principles,  and  it depends on  the  assumption of whether the prices  of other



goods,  such as substitute sites,  remain constant.   In Section 4,  the development



of  a new recreation  site  is  assumed  to affect  the  price  of  other goods.  The



proper  measure of  recreation benefits now must include the change in benefits in



those  markets  where prices have changed.   This  case is clearly the exception,



because the price  of a  recreation site is either  zero or fixed, and  is  therefore



insensitive to changes  in other prices.
12

-------
2.    An Overview of Benefits in the Recreation Literature



     A brief  overview of  benefits  measurement in the  recreation  literature is



presented  focusing  on two  questions:   (1) What  is  the  appropriate  measure of



benefits  of a  new or  improved  site  when demand  for  that  site  comes  at  the



expense  of substitute sites?   and  (2) What  is the  explicit theoretical just-



ification  for the commonly accepted definition of benefits?



     All  recreation  benefit analyses  contain some definition  of  benefits,  but



the  issue  of  measuring benefits when there  exist  close substitutes  has  only



recently been considered.   For instance, in the exchange by Stevens (1966, 1967)



and  Burt   (1967)  on  the  fishing  benefits  of water  pollution  control,   no



consideration  was  given   to  demand  shifts  for  fishing  at  substitute  sites.



Reiling,  Gibbs,  and  Stoevener   (1973,  p.  3)  reveal  a  clear preference  for



avoiding this issue by explicity assuming that expanded use of one  site does  not



come at  the expense of substitute sites.   Some of the more recent  literature by



Knetsch  (1977), Mishan (1976), Freeman (1979a) and Cesario and Knetsch (1976) is



reviewed,  which explicitly considers  measuring benefits  at one  site  when there



exist  substitute  sites.    The  focus  is  on  how  benefits  are  measured  and



particularly on how this benefit measure is justified.








A.   Knetsch



     Knetsch  (1977) is  concerned with the evaluation  of benefits  at a proposed



site  when  there  is  an  identical  displaced  facility  requiring  a  greater



travel-cost.   To  review Knetsch's  position,  the demand curve  for  an existing



site  is  depicted  in  Figure  1.   Quantity demanded is  1,000  recreation days  and



consumers'  surplus  is  $2,500.   Assume  that a  second  and  identical  site is



constructed that  reduces  travel  costs by  $1  for  each population centroid.  The



demand curve  for  the  proposed site appears on the  right hand side of Figure 1.






                                                                              13

-------
                            FIGURE 1



            RECREATION DEMAND AND BENEFITS:  THE KNETSCH ANALYSIS
           Existing Site
       Proposed Site
-co-
o>
o
                  1000
       Quantity

  ( Recreation Days)
              1000  1500
     Quantity

(Recreation Days)

-------
The  area  above $1  and under this demand  curve  is equal to  the  area under the



demand curve  for  the existing site.   According to Knetsch, the demand curve for



the  new site  slopes downward and to  the  right from a price less than $1 to P =



0.  The demand curve but becomes horizontal at $1 because at a fee of $1 or more



all  recreationists  return  to the initial  site.   The  increase in total benefits



is  $1,250,  which  is  the  area  under the  new (kinked) demand  curve.   Knetsch



concludes  that the  demand  curve  for  the new  facility must  reflect existing



facilities, but  the  loss  in value  of  the existing facility can  and should be



ignored in calculating the net gain of the new facility.



     Unfortunately,  in  Knetsch1s  analysis  measured benefits at the  new  site do



not  include  a redistribution  of willingness to  pay  from  a substitute  site



because no  redistribution  occurs.   The willingness to pay for the first site is



$2,500 before the new site  is  constructed and,  at a price of $1 or more at the



new  site,  it  is  $2,500 after the  new site is constructed.  As the price of the



new  site  rises above  $1,  the willingness  to  pay  for  the  existing site remains



unchanged.  Knetsch1s  analysis  is based on a special case where the demand curve



for  the  substitute  site  doesn't  shift.   Because  there  is   no decrease  in



willingness to pay  for the substitute site, his analysis provides no support for



the  position  that the  decrease  in willingness  to  pay for substitute sites can be



neglected when measuring net benefits of a new or  improved site.



     However,  a  particularly  important  econometric  implication  of Knetsch1s



analysis  is  the  need to include some measure  of  substitutes when  estimating the



site demand  function.   If the specification  of  the  proposed site excludes the



existing  site,  the  continuous  demand curve  from Q = 1,500 to  P  = $6 would be



estimated.  Benefits would be overestimated by $2,500.  By correctly  specifying



the  demand for the proposed site, the  kinked demand  curve would presumably be
                                                                               15

-------
estimated.    Benefits  of  the  proposed  site  would be  correctly estimated  at

$1,250.



B.   Mis nan

     Mishan  (1976),  in  his  authoritative  treatise  on  cost-benefit analysis,

addresses the  issue of measuring consumers' surplus when  increased purchases of

one good are at the  expense of other goods.  Mishan states that if a  new good is

introduced  or  the price of a good  falls,  consumers'  surplus should  be measured

by  neglecting changes  in  consumers'  surplus  of alternative goods.   He says:


             ...  I append a note to this chapter  containing a simple
             example   in   order   to  reassure   the  reader  that  in
             measuring  the  consumers'  surplus  of a new  good,  or a
             good  the price of which has  changed, he should neglect
             the  induced  shifts  of demand  of related goods,  (p.  32)


The  reduction  in the demand for  the  substitute good shifts the  demand schedule

to  the left  producing a  decrease  in  consumers'  surplus.   According to Mishan,

this   loss  "...   is  not to  be  regarded   as  a loss of  consumers'  surplus...";

instead,   "This   reduction in   area   is   simply  the  consequence of consumers

bettering  themselves by switching from good y to the new lower  priced good x."

(p.  34).

      When   Mishan  considers  relatively   close  substitutes  he  uses a   demand

schedule  for each good, and asserts that  the  area  under  the demand schedule for

the  substitute  good can  be  ignored.   He  defends  his position  by  example and

illustration,  but  changes  the  case so  that  the  two  goods  are  perfect  substi-

tutes.   Because  what  was  two goods  is now only one  good,  an aggregate  demand

schedule  replaces  two  separate  schedules.  Specifically,  Mishan considers the

demand for transportation across a certain water body where a ferry service  is

being  replaced  by  a  bridge.   With  the   ferry service  the price   is  P0  and


16

-------
consumers' surplus  is  area  P0ab in Figure 2.   After  the bridge is constructed,



the  ferry  is  discontinued  and  the  price  of  transportation   falls  to  Px.



According to Mishan, the appropriate measure of the benefits of constructing the



bridge is the  area under the demand curve and between the new and initial  price



(PiPobc).



     When  measuring  the increment  to  benefits,  it  is possible  to think  of



consumers' surplus  foregone  as  being subtracted from  the  gross  increase.   With



the  ferry  service,  consumers'  surplus  was  area  P0ab.   After  the ferry  is



discontinued,  consumers'  surplus  resulting  from  the  bridge   is  P^c.    The



increment in  consumers'  surplus is total surplus after the ferry service (P^c)



minus  consumers'   surplus  foregone  from the  bridge  (P0ab);  this  increment  is



PaP0bc.   The  reason  for subtracting  consumers'  surplus  foregone is that  the



ferry  service  is  discontinued,  and the  bridge  demand  schedule  assumes that the



ferry  is  not in operation.



     Assuming  that the  ferry could operate  if the price  were P0,  the  demand



schedule  for the  bridge  is  deb as before,  but  it  becomes  perfectly  elastic  at



price  P0.   The amount of consumers' surplus is the same as above, but it is the



area above price P^ and below the bridge demand schedule.  No consumers'  surplus



is subtracted  because no consumers' surplus is foregone.



     Mishan1 s  position  is that  benefits  of  a new site  (in this  case a  bridge)



can  be measured by neglecting  shifts in the  demand  for substitutes.   However,



his  justification  is  an  illustration  that,  in  principle,   is   identical  to



Knetsch's.   By considering  the  case of  perfect  substitutes  and a single demand



curve,  Mishan  provides  no  support  for  the  position  that   the markets  for



substitutes  can be ignored  when measuring  benefits at  a  new  or improved site.



In  Mishan's  case,  like  Knetsch's,  there  is  no  redistribution  in   consumers'



surplus.





                                                                              17

-------
                                    FIGURE 2

              CONSUMERS'  SURPLUS:   THE CASE OF PERFECT SUBSTITUTES
                                 Quantity
C.    Freeman

     The  analysis  of Mishan  and  Knetsch are  special  cases and  not useful  in

analyzing the general case  where  the site demand curve  shifts  to the  right and

the demand  for  substitutes  shifts  to the left.  Freeman  (1979b)  has explicitly

addressed the issue  of  measuring  recreation  benefits when the  demand  curve for

substitute  sites shifts,  so  his  analysis is  reviewed.   In Figure  3,  the  initial

demand curves for  site  A and B are  denoted  as  DAI  and DR2.   An  improvement  in

water  quality at  site A shifts the  demand curve outward to  D.3 and the  demand

for  the  substitute  shifts  inward to  DBI.   Benefits  of the  improvement  are

measured  as the area between the new and initial demand curves for site A and

above  the  market  price  (area  BDGE).   According to  Freeman,  no  consideration

should be given to the  decrease in willingness to pay  for the  substitute site,

area RSVU.  He states:
            In utilizing this measure  of  benefits,  there  is  no  need
            to take into account  changes  in  recreation  use at other
            sites or  savings  in travel cost (Knetsch 1977).  These
            are  captured  by  the  BECD  [BDGE in  Figure 3]  benefit
            measure,   (p.  199)

18

-------
                  FIGURE 3



        DEMAND FOR TWO RECREATION SITES
Site A
Site B
                         b  T
    DAI    DA2   DA3
      DBI   DB2  DB3
  Qi

-------
Freeman's conclusion  is  that benefits can be measured by demand curve shifts at

the improved  site  and demand shifts at  substitute sites do not explicitly enter

into the  benefit calculation.   Freeman  provides  no justification for neglecting

the decrease  in  consumers' surplus at  substitute  sites, except for his reference

to  Knetsch.   My review  of  Knetsch's position revealed it  to be a special  case

where there is no decrease in willingness to pay  for  the substitute site.



D.   Support  for the  Conventional Measure of Benefits

     The  validity  of Freeman's position is not an issue at this point.   Rather,

the contention here  is  that the  recreation  benefits  literature (as  exemplified

by  Knetsch,   Mishan, and   Freeman)   does  not  contain persuasive   theoretical

justification for  the position that  displaced benefits  should be ignored  when

calculating  net benefits of a  new  or  improved site.   Although  Mishan,  Knetsch,

and Freeman  reach  the same conclusion, they offer no  evidence that shifts in the

 demand for substitutes  should  be  ignored when calculating benefits  of  a new or

 improved site.  Yet  their position  seems to  be  the  accepted view of recreation

 researchers.   For  instance,  Cesario  and  Knetsch (1976,  p.  101)  state:


             That  is, the   value  measurement for   a  new  site  is
             measured independently of any diminutive effects on the
             use of  existing sites.   Any losses  in consumer  surplus
             at  existing sites  are  irrelevant  to  the calculation
             (even  though it  may be  informative for planning purposes
             to  calculate the  magnitude  of these quantities).   Such
             losses merely reflect changed demand  characteristics and
             losses  in  the  value  of  some fixed  assets,  and should
             have  no  bearing  on  the benefit calculation   for the
             proposed site which would be judged  on  its merits  alone
             (McKean  1958;   Mishan  1971;  Knetsch 1974).


 Cesario and Knetsch provide  no rigorous  justification for  this  position, relying

 instead on references such  as  Mishan  and Knetsch.    The above  review of Knetsch
 20

-------
and Mishan argues  that  they do not support the view that any losses of consumer

surplus at existing sites are irrelevant.

     Although  the  sample  of recreation  benefits  literature reviewed  here  is

small, the work is probably the most  important  in this area.  On  the  basis  of

this  review, two general  conclusions are suggested.   First, the prevailing view

is that benefits of  a new or  improved  site  can  be measured as the area between

the  new  and initial  demand  curve  and above the market  price and,  furthermore,

that  demand  shifts for  substitute sites need  not be  considered.   Second,  the

theoretical  support  for  this  position  has  not  been  made  explicit   in  this

1iterature.

      In an analysis of the potential benefits of a new ski  site at Mineral King,

Cicchetti,  Fisher,  and  Smith  (1976)  challenge  the  commonly  held view  that

benefits  can be measured by considering only  the  impacted  site.1  Cicchetti  et

al.  specify  a  simultaneous demand equation model in which the price of each ski

site  is  an  argument  in  each demand  curve.    They  assert  that  specifying  a

multisite  model  allows  them to estimate the effects of a change in the price at

one  site  on demand and consumers'   surplus at the substitute sites.  In an edited

version of the  Mineral King  study, Krutilla and Fisher (1975, p. 198) state that

the  new  Mineral King site would result  in  a reduction in demand for substitute

sites  and these  effects  are  captured  by  measuring  the   change  in consumers'

surplus  over multiple  sites.   Bishop and Cicchetti  (1973)  further explain the

benefit measure used  in the  Mineral  King paper:
 1Burt  and Brewer  (1971)  used  a  multiequation model  very similar  to  that of
Cicchetti et al. (1976).
                                                                              21

-------
            In a  recent paper  Cicchetti,  Fisher, and  Smith  (1972)
            simultaneously  estimate  the  demand for  various  skiing
            sites in  California.   The  location of other  sites  and
            therefore their  relative prices  are  taken  into account
            explicitly   by   using   a  generalized   least  squares
            regression  approach.    The  benefits   of  new   sites  at
            various  locations  can  be determined  by  simultaneously
            estimating  the  change  in  consumer  surplus  for  the
            alternative sites,   (p. Ill)


The estimate of consumers' surplus in the Mineral  King study explicitly reflects

the reduction in willingness to pay for substitutes.   This position is in marked

contrast  to  that taken  in  the  studies  discussed  above and implies the  need to

define the theoretical underpinnings of the prevailing view.



3.   The  Theoretical  Underpinning  for the  Conventional   Measure  of Recreation
     Benefits

     The  above  review  of the  definition  of  recreation benefits  suggests some

ambiguity  on  the issue and the absence of agreement  on theoretical support for

any  particular   definition.   Benefits  are  now  demonstrated  to   be  properly

measured  by  considering only the demand curve  for the  affected site.  Further-

more,  this  demonstration follows from an application of economic principles.

     The   following   analysis  may  assume  an  environmental   improvement  at  a

recreation site  (hence  a  demand curve shift), a decrease in the price of  a  site,

or  the  introduction  of  a  new site.   On  grounds of expositional  convenience,

consider  the net benefits  of  introducing a new  recreation  site.    Panel  B in

Figure  4  depicts an  ordinary demand  curve (ODC)  and  a Hicks compensated demand

curve  (HCDC), where  OD  depicts  the  quantity of the  new  good demanded at  price

?i.2    Net benefits  of  the  new  good  can be  measured as  the area  under the
  2A  decrease  in  the price  of a good  increases the  quantity demanded  because
people  substitute this  good for other  goods  and because the  lower  price effec-
tively  increases  real  income.   The ordinary  demand curve  reflects both  this
substitution  and  income effect.  The compensated demand  curve reflects  only the
substitution  effect  and presumes that real  income is  unchanged.

22

-------
                          FIGURE 4



CONSUMERS'  SURPLUS USING ORDINARY AND COMPENSATED DEMAND CURVES
                                 Panel A
  a>
  u
             Quantity of X (Recreation Days at New Site)
                                  Panel  B
   Ordinary

.Demand Curve
                         ±
                              i
                              1   Compensated

                              I  Demand Curve
                         C    D

             Quantity of X( Recreation Days at New Site)
                                                                   23

-------
compensated  demand curve  and  above  the market  price,  area PxPoF.   This area



reflects the change in consumers' surplus caused  by introducing the new good and



is  defined  as the  willingness to  pay for  the  new good  over  the above  actual



payment.   On grounds  of  empirical   necessity  and the  work by  Willig (1976),



consumers'  surplus,  as  measured  under the  ordinary  demand curve, is generally



considered  an  acceptable approximation to the area under  the compensated  demand



curve.  According to Knetsch,  Freeman, Mishan, and Cesario and  Knetsch, benefits



of  a new site are  measured  as the area under the  compensated  demand curve, or



approximately  as  the  area under the  ordinary demand curve.  At issue  is whether



this  area correctly measures  the benefits  of  a  new site  and what consideration



if  any  should be given to  benefit from displaced  facilities.



      This question is answered by  deriving  a  demand  curve  for a  new  recreation



site  using  indifference  curves   and price  lines.    Assume  that  a utility-



maximizing  consumer allocates  all his income between  good X (the  new  recreation



opportunity) and  a composite of all  other goods,  which is  termed Hicksian  money.



Before the  recreation opportunity was provided,  the consumer purchased only the



composite good and did  not consume  good  X.   As depicted  in Panel A of  Figure 4,3



this initial  allocation is  defined by point A,  which is  the  point of tangency



between  price  line   P0  and  indifference  curve  I0.    After the   recreation



opportunity  is  provided,  Pt  becomes the  price  of recreating  relative   to the



price index of the composite  good  and the  consumer maximizes  utility by  moving



 from point A to point G.  The change in  welfare,  as measured by the compensating



variation,   is AB  after  it has been converted to dollar terms  by  multiplying  by



the  price   index  of  the  composite  good.   A well-known  proposition  in welfare



economics,   and critical point here,  is  that this measure  of consumer  surplus  in
  3The diagram in Figure 4 was presented by Currie,  Murphy,  and Schmitz (1971).






 24

-------
Panel A corresponds to consumer surplus as measured under the compensated demand



curve in Panel B.   We can now focus on the welfare gain AB in Panel A.



     Recreation use  at  the  new site (good X)  comes  partially at the expense of



substitutes, which  in this  case is the composite  good.   As  a result of the new



recreation opportunity,  use  at the site becomes OD (Panel A) and demand for the



substitute decreases by AJ.   Hence the improvement in welfare, which is measured



by the  movement  from indifference curve I0 to  11,  clearly reflects a reduction



in  demand for  the  substitute  composite  good.  The  demand  curve  and consumer



surplus  in  Panel   B do  not  imply  how  foregone  benefits  should  be  treated.



However,  the  derivation of  this  demand curve and the  corresponding measure of



consumer  surplus  (AB in Panel  A) show clearly that measured consumer surplus is



a net increment to benefits.



     The  above  analysis supports  the  conventional  measurement  of  benefits,



subject,  however,  to a  stringent assumption.   As seen in Figure 4, Panel A, the



composite  good  is  an aggregation or weighted  sum  of all  other goods,  where the



weights are  the  prices  of these goods.  As stated originally by Hicks (1939, p.



33), if the  relative prices of a  group  of commodities are given and unchanged,



these  commodities  can   be  lumped together  and  treated  as  a  composite  good.



Hicks'   theorem  of  group  commodities  is  being  used to  justify  defining  the



decrease  in  relative  price   of  good  X.   Specifically,  the  above  conclusion



assumes  that  lowering  the   price  of  a  new  good  does  not affect  the relative
                                                                              25

-------
prices  of other  goods.4   If  the  introduction of  a new site  affects  relative

prices  of  other  goods,  the composite good theorem  is not applicable.  At issue,

then,  is  determining the proper measure of consumer surplus under conditions of

multiple-price changes.



4.    Consumer Surplus with Multiple-Price Changes

      The  view  that recreation benefits can be measured by considering only the

market for  the  single  affected   site  is  correct  if  we assume  an  ordinary

Marshallian  partial  equilibrium  demand curve.5  In  the  Marshallian demand curve,

prices of all  other goods are fixed,  and therefore Hicks' theorem  of composite

goods is  applicable.6   In this section  we consider  the measure of recreation

benefits  when a  new or  improved  site affects prices  in more than one market.

      Although  the recreation  literature gives  little attention  to this issue, it

 has been treated  at  length  in the  welfare  theory  literature by Harberger (1971)

 and Mohring (1971) among  others.  Borrowing an illustration from Mohring, assume

 two goods, margarine and  butter, whose demand  functions can be  written as
  4According to  the  Cicchetti  et  a]_.  analysis,  for  each individual,  relative
 prices  of  existing  sites  are  invariant to  the construction  of  a new  site.
 However, the  price of the new site  relative  to that of existing  sites  differs
 according to  the  origin of the  individual.   The latter point does  not  nullify
 the use of the  composite  good theorem,  which  seems  appropriate  in  the Cicchetti
 et al.  study  and  in the Burt and  Brewer  study.   In these studies,  there  was  a
 decrease in the price  of  the  new  site that produced  a  shift in the  demand  for
 substitute  sites,  but  relative  prices  of   these  substitute   sites   remains
 constant.   Burt  and Brewer  and  Cicchetti   et  al.  used a  quadratic  benefit
 estimation equation  that is a generalized approach  for integrating  a system of
 equations, in this  case, when prices change at one  or more  sites.   Because only
 one price  changes  (the new  site),  the simpler  technique  of integrating that
 demand equation would have been appropriate.

   5The Marshallian  demand  curve  is a sufficient  but  not  necessary condition to
 consider only  the  affected site.   If all other  prices change  proportionately,
 the composite good theorem still  holds.

  6Freeman (1979b,  p. 35)  and Mishan (1976,  p.  32)  recognize the necessity of
 invoking  Hicks'  composite  good  theorem  when  analyzing  benefits  in a  single
 market.
 26

-------
                              Qm = fi

                              % = ^V V Y)  '


where the price  of margarine (P ) and the  price  of butter (P.) enter into each

demand equation along with income (Y).  Initial equilibrium in the margarine and

butter markets  is defined  in  Figure 5  by points  A  and  C,  respectively.   The

margarine  market  is  analogous  to  our  proposed site,  except  that  Mohring's

initial change is a  reduction in price.   The price of margarine falls from P'  to

P"  which produces  a decrease in the demand  for  butter from D.  to  D'    If the

price of butter remains constant at P'  consumer surplus is measured as the area

under the margarine  demand schedule between the new and initial price.

     Mohring emphasizes that even though the butter demand schedule shifts, this

fact  need  not be  considered when measuring the  increase  in  benefits  resulting
                                                        f
from  lower  priced margarine.   This  point corresponds to  our  conclusion  in the

previous section that benefits are correctly measured under the demand curve for

the affected recreation site, and displaced facilities can be ignored.

     Suppose  the  illustration  is  changed  so  that  a  decrease  in the  price  of

margarine decreases  the demand for butter as before, but now the price of butter

falls  from  P!  to PV.  This  price  decrease  in butter increases the net willing-

ness  to  pay for butter and  the  increment in consumer surplus  must  be added  to

the increase in consumer surplus for margarine to obtain the appropriate measure

of  welfare  change.   This  point is  well   recognized  in  the  welfare  economics

literature  and  can be generalized to state  that  the  change in consumer surplus

resulting from  price changes in several markets  is  the sum of the increment of

consumer surplus in  each market  (Harberger 1971).

     Mohring  emphasizes  the  ambiguity  of  measuring  the  change  in consumer

surplus  in  the  butter  market,  and he notes  that  three  measures  have been

proposed.   Using  the  initial  butter demand  curve, consumer  surplus is P"P'CD,

                                                                              27

-------
                                    FIGURE  5



                MEASURING BENEFITS  WITH MULTIPLE-PRICE CHANGES
                Demand for Margarine
Demand for Butter
but this measure  is  PVP/FE  if the new demand schedule is  used.   If we move  from
                      b  b


the initial to  new  demand schedule,  the change in consumers'  surplus is  P/CEPo-



Because a rationale can  be presented  for each of these definitions, the measure-



ment  of  consumer surplus is,  in general,  sensitive  to the  definition  chosen.



This point is recognized in  the welfare theory literature  and has led Silberberg



(1972) to  conclude  that the appropriate change in utility  or welfare cannot be



defined  unambiguously.    Hotel!ing  (1938)  noted this  indeterminancy  in  the



measure of benefits and also the condition under which consumer surplus could be



measured unambiguously.   This condition is known as the integrability condition,



and means  that  the  demand curves have  identical  cross  partial  derivatives  with



respect  to prices.   The  integrability condition  for  the  margarine  and butter



demand curves is





                                  8Q,   9Q
                                   x      x
                                  9Pm   9Pb
28

-------
which  says  that the  change in  the  quantity of butter  (Q. )  demanded resulting



from  a change  in  the price of  margarine equals the change  in  the quantity of



margarine  (Q )  demanded  resulting  from  a  change  in the  price of  butter   As



noted  by  Mohring  (p.  356), this condition  holds  if  the demand curves are Hicks



income-compensated or if the income elasticity of demand for both goods is zero.



Burt  and  Brewer (1971)  and Cicchetti et al. (1976) recognized this requirement,



and therefore specified their demand equations as linear and symmetrical.



     These conclusions can be restated in terms of our main concern of valuing a



recreation  site that has  a substitute  site.   Let the  price of  recreating be



defined as the  entrance fee or as travel costs.   If a decrease in the price of a



recreation  site (or  the  construction  of  a new  site)  affects the  demand  for



substitutes  or  complements, but  leaves their prices  (entrance fees  or  travel



costs)  unchanged,  benefits are  estimated properly as  the area  under the site



demand  curve and  between the initial and  new  price.   No explicit consideration



should  be given to the decrease in  willingness  to pay for the substitute site.



Alternatively,  if  the  decrease  in  the  price  of  a  site causes  a  change  in



relative  prices  of  other goods,  such  as  a   substitute  site,  the  increment



(decrement)  in  consumer  surplus  in the substitute site resulting from the price



change  must  be  added (subtracted) to that  of the first site to obtain the total



change  in consumer surplus.



      A  peculiar feature  of outdoor recreation is  that  the price of recreating,



as  measured  by entrance  fees,   is  generally  zero.   Where  entrance  fees  are



charged,  for example campgrounds,  these prices  are  insensitive  to  the  intro-



duction  or  improvement  in substitute  sites.  Where travel costs  are  used as a



proxy  for price, the travel cost  to substitute sites  is  invariant to a demand



shift  at the site  being analyzed.   Therefore,  for most all  practical applica-



tions  in  recreation,  including  the travel-cost  approach that  is  used here,






                                                                              29

-------
benefits of a new or improved site can be measured correctly as consumer surplus



at the new or improved site.








5.   Conclusions



     This chapter addresses the issue of the proper measure of benefits at a new



recreation  site when  demand  for  that  site comes  partially  at  the  expense of



substitute  sites.   The literature reviewed indicates that some researchers have



avoided  the  issue;  others have  explicitly subtracted  benefits  foregone.   The



prevalent view  is that benefits can be measured by considering only the new site



demand  curve.   The  limitation with this view  is  the absence of  any  theoretical



justification.   As  shown  here,  benefits are  measured  correctly by  considering



only  the demand curve for  the new site, but this demand  curve must be correctly



specified  to  consider existing  sites.   Use  of  the  conventional microeconomic



model   of   consumer  behavior  shows  that  recreation   benefits,  measured  as



willingness to pay  for  the  new  site,  automatically net out benefits foregone



from  substitute sites.   In  the special case where the introduction of a new site



causes  prices  of other  sites or goods to  change,  the  increment in  benefits is



the net sum of consumers'  surplus in  these  affected markets.
 30

-------
                                   CHAPTER III



                ESTIMATING RECREATION TRIPS WITH A GRAVITY MODEL








     In Chapter II it was established that recreation benefits can be defined as



the area under  the  recreation  demand curve above  the  market  price.   Chapter IV



contains a  discussion of  the  travel-cost approach  to estimating a  recreation



demand  curve.   This chapter presents  the methodology used to  obtain  the  input



data of the travel-cost  demand  curves.   A  1980 regional  household  recreation



survey  is used  to estimate the number of recreation trips by activity from each



origin  in  the  region.   An attractiveness model  is used  to  obtain  preliminary



estimates of  the attractions  of  each  site in the  region.  The distribution of



trips between each origin and destination is estimated by using a gravity model.



The gravity model and  attractiveness  model  are  then integrated, and  quantity



demanded at each  site is estimated with  the  revised attractiveness  model.  The



output  of the  gravity model  is the number of visitor days received by each site



in the  region  by activity and emanating  from each origin in the region.  These



outputs are the basic input required to calculate a travel-cost demand schedule



for each recreation site in the region.



     This  analysis  of  recreation  behavior  differs  from  existing   studies  by



virtue  of  magnitude,  with  195  recreation  centroids defined  over three  and



one-half states.  This  scale is  considerably  larger  than  those in the regional



models  of  Burt and  Brewer (1971),  Cichetti,  Fisher,  and  Smith (1976), Cesario



and Knetsch (1976) and Knetsch, Brown, and Hansen (1976).  The primary advantage



of this size  model  is  that  any  site within  the  region  can be analyzed.  Also,






                                                                              31

-------
the influence  of  all  potential substitute  sites  is  most likely to be reflected



in a  larger  model.   The ability to analyze a large number of sites results from



the  use of  household  surveys to  estimate recreation  trips  by  origin  and a



gravity model to estimate the  distribution  of these trips.



     Most  recreation  analyses  focus  on one activity or  treat recreation as a



composite homogeneous good, for example, Stevens' (1966) estimate of the fishing



benefits  resulting from  improved  water quality.   In  contrast,  this analysis



considers four activities:  camping, fishing, boating, and swimming.   A focus on



one  activity may  be inadequate when several activities  respond to water-quality



improvement.   These four  activities  are not homogeneous;  they differ in their



response to  site  characteristics such  as water quality, average travel distance



and  length  of stay,  and value  per  activity  day.    Furthermore,  the relative



composition  of these  activities  varies widely  across  recreation  sites.   For



these  reasons, the  above  four  activities are analyzed  separately.



      A fundamental  difference between  this study and  other  regional travel-cost



studies is  the  method  of obtaining   input data.   In  the  regional  models of



Cesario (1973, 1974,  1975),  Cesario  and  Knetsch  (1976),  Cheung  (1972),   and



Knetsch,  Brown,  and  Hansen  (1976), origin-destination  data were obtained  from



site attendance records  or on-site  surveys.   In this study, origin-destination



allocations   are  estimated  from a  gravity model  that  uses  origin  data  from



household  recreation  surveys.   The costs and benefits of  this  approach relative



to that of using  site-specific attendance data merit  brief comment.



      The initial  cost  of a  regional  household recreation  survey  and regional



model  is  of course substantial, but  once  the  survey is taken  and  model  con-



structed,  the marginal cost of  analyzing additional  sites  is  less  than  that of



most single-site  analyses.   The cost  of  using existing  attendance  records  is



low,  but  in the   Northwest,  these   data  are   deficient  in both quantity   and






32

-------
quality.    Several  agencies,  such as  the  Corps of  Engineers,  Water  and Power



Resources Service, U.S.  Forest  Service,  and state parks  departments  have total



attendance data, but  not by origin.   Agencies may define attendance in terms of



visits, visitor days,  recreation  days or activity days,  and  the definitions of



these terms tend  to  vary between agencies.   The on-site survey approach is less



expensive when  the number of sites is small, but more expensive when the number



of  sites  is large.   The number  of  sites  at  which  the  costs  of  the household



survey and site survey approach are equal cannot be defined a priori.



     The  household survey approach coupled  with the model presented here offers



significant advantages  over the on-site  survey approach.  The present model can



estimate  demand and   consumers' surplus  for  a proposed  site  at  any location in



the region.  The on-site  survey approach obviously cannot obtain attendance data



for  a proposed site;  so the  demand function  for  the proposed   site  must be



estimated  by assuming that the site  is similar to an existing site.  The demand



for a site depends on site characteristics,  distance to population centers, size



of  the  population centers,  and alternative sites available  to each population



origin.  A model based on these variables can be used to estimate input data for



the  demand curve  of  a proposed  site;  but  the model  would, at  best,  produce



reliable  estimates of  total quantity demanded.   However, the  distribution of



these  trips by origin  would be  estimated  with large  errors unless substitute



recreation  opportunities  were  accurately  modeled for  each origin.   Existing



regional  models do not  have this capability,  and consequently  are  limited in



terms of  estimating  demand curves for proposed sites.  The model presented here



can estimate total quantity  demanded  for a proposed site and the distribution of



this  demand  by  origin.    In  addition  to  being  able  to  estimate  demand and



benefits  for any  site in the region, the estimates should be more  reliable than



those based on  "similar  sites" and site-attendance data.






                                                                              33

-------
     The travel-cost  approach is  not applicable when most  users  come from one



origin  because  travel  distances  and,  hence  travel  costs,  will  not  possess



significant statistical variability.   The average distance traveled for fishing,



swimming, and  boating  is  about 40 miles, and a  large number of recreation sites



are  located  near urban areas.   If the site survey  defines  origin as county or



city, the data will  be inadequate for a large number of sites.  The methodology



used  here  permits dividing  urban  areas into  several  population centroids.   In



this  way,  the  travel  cost  is   measured  accurately  for  a  large  number  of



recreators, and travel costs will  vary over these users.



     As  a  brief  overview,  the model   consists  of four integrated  components:  a



trip  production  model, an  attractiveness  model,  a  trip distribution (gravity)



model,  and  a demand and valuation model.   The  trip  production model  is used to



estimate the  number  of  recreation  days  by  activity  that emanate  from each



population  centroid  in  Washington,   Oregon,  Idaho,  and western  Montana.  The



attractiveness model  is  used to estimate the  attractiveness,  or total quantity



demanded,  of each recreation centroid in the  region.   Recreation days produced



and  attracted enter  a  gravity  model  where  they  affect  the  distribution of



recreational  travel.   A  gravity model estimates a trip  interchange  matrix that,



for  each recreation centroid, is  the  number of  activity days  received from each



origin.   These outputs are  used to estimate a travel-cost demand  curve for each



recreation   destination   and  for  each  of  the  four  activities   considered.



Recreation  value is measured  as  the  area under the  demand  curve and above the



market  price, which  in  this  study  is presumed to  be  zero.  An  improvement in



water  quality coupled with  an  increase  in  facilities produces an outward shift



in  the  demand curve,  and the  area between  the  initial  and  new  curve  represents



the  benefits  of  improved water  quality.
 34

-------
1.    Gravity Model Overview



     The  gravity  model  as applied  to  travel  behavior  is a  trip distribution



model that  is used to  estimate  trip interchanges between  all  pairs  of origins



and destinations.   Normally,  the number of trips produced  and received by each



zone  are exogenous  variables.    The  endogenous  variable  is   the  allocation  of



these productions.  The  basic premise of the model  is  that the number of trips



produced  by  origin  i  and attracted to destination j is directly proportional  to



(1)  the  total  number  of  trips  produced  in  i, (2)  attracted  to  j,  and (3)



inversely proportional  to a  function of spatial separation  between  the zones.



     The  gravity  model   is   ideally  suited  to  estimate   the distribution  of



recreation  travel.   However,  the  most  stringent limitation  of the  model,  for



purposes  of  recreation  analysis,   is  the  requirement  that attractions  are



exogenous.   According  to  this  assumption,  the  quantity  of recreation  use



demanded  at  each  site is known, and  the gravity model solves  for the allocation



of this  demand by origin.  Previous versions of this study, including Sutherland



(1982c),  are subject to this  limitation.  The gravity model is  developed in this



chapter  first, along traditional  lines, and using exogenous attractions.  In the



latter  part of this  chapter, the  gravity  model is  extended to simultaneously



estimate  attractions.   This   extension is  shown  to result   in  a  substantial



improvement,  both  theoretically and empirically,   in  the regional  recreation



demand model.



     The  gravity  model  has  a  long  history  of   successful applications  in



economics  and  in  transportation analysis,  but  has  also  been used  to analyze



recreation  travel.   The primary use  of the  gravity  model  in  economics has been



to analyze  regional  trade flows.  Anderson (1979, p. 106) conjectures that this



model  is the most  successful empirical trade  device to  evolve  in  the last 25



years.   Regional  economics books,  such as  Isard's  (1960),  typically contain  a






                                                                              35

-------
discussion of this model.  However, the most frequent application of the gravity



model is  to  estimate both interurban and  intraurban travel  flows.   The gravity



model appears to  have had a long and successful history as a tool for analyzing



travel  flows.   The  prominent  position   of  this  model  is  confirmed  by  the



attention  given  it  in  the  transportion  engineering  texts,  such  as  those by



Hutchinson (1974), Dickey (1975), and Stopher and Meyburg (1975).



     The  gravity  model   owes   its  theoretical  foundation  to  Newton's  Law of



Gravitational  Force,  which  stated loosely,  is that  the  gravitational  force



between  two  bodies is directly proportional to  the product of their masses and



inversely  proportional  to the square of the distance between them.  A  frequent



criticism  of the  model as applied  in  economics  is  that the theoretical founda-



tions  are  in  physics  and  not  in the  principles  of social  behavior.   This



criticism  has been answered by some recent  work that establishes a theoretical



foundation  for the  gravity  model.   For  example,   Anderson  (1979)  provides  a



theoretical  explanation  of the model as applied to  commodities.  Niedercorn and



Bechdolt (1969)  derive  a gravity model  from  consumer  theory  by using a loga-



rithmic  and  power utility function.



      Despite theoretical support and  extensive  empirical  success in predicting



urban travel, the gravity model  has been  used infrequently in analyzing  recre-



ation travel and with limited  success.   Ellis  and  Van Doren (1966) found the



gravity  model predictions of  camping  in Michigan to  be  less reliable than those



from a  systems  theory  model.  Freund and  Wilson   (1974)  obtained some  rather



large discrepancies  between gravity model predictions  of recreation behavior  in



Texas and observed behavior.



      Several  specifications  of the model  have been  put  forth; the  specification



used  here   is  one  which is  used widely in  transportation  analysis  and was



developed by the  Bureau  of Public  Roads (1965).  The equation  is






36

-------
                                          A. F.
(III.l)                         T,, = P, =-f-^
and the constraints are


(III. 2)                        IT.. = P.  and
                               j
(III.3)                        IT..  = A.
                               ,-  iJ     J
where  i  refers  to origin  and j  to  destination.   The  symbols  in  (III.l)  are

defined as


     T-- = number of activity days produced at i  and attracted to j,

     P.  = number of activity days produced at i,

     A.  = number of activity days attracted to the jth recreation centroid,  and
      J
     F-- = a calibration term for interchange ij, which reflects  the  effect of
       J     distance.

     Equation (III.2) states  that the  estimated trip  interchange matrix (T. .)

must imply  that the total number of  trips from origin i (IT..)  is equal  to the
                                                           j 1J
exogenous number of  trips produced.   In the calibration procedure used here and

elsewhere,  this  constraint  is  satisfied  automatically.    According  to  Eq.

(III.3), the  estimated  trip distribution matrix, which  estimates  the  number of

trips  terminating   at  each site,  must  also be  consistent  with  exogenously

estimated attractions.

     The gravity model,  as  generally  used, is a  distribution  model;  it takes a

given number of recreation activity days emanating from population centroids and

distributes  these  days  according to  the  relative  attractiveness  and  spatial

impedance between centroids.  In the special case where site-attendance data and

trip-production data are available, the gravity model  is well suited to estimate

                                                                              37

-------
allocation of these  trips.   If site-attendance data  are  unavailable,  they must



be estimated  by a demand  model.   Ideally, a  demand  model  would include travel



costs to  all  substitute sites and the relative attractiveness of all substitute



sites.   Such  a demand  model  would  be  quite  similar  to  the gravity model.   In



this  study,  the gravity  model  is  extended to include  endogenous  attractions;



hence, it becomes a demand and distribution model.



     As noted by Ewing  (1980), Eq. (III.l)  has two important properties.  Adding



destinations  to the  system  or   increasing  the  attractiveness  of  the  existing



destinations  will  increase the number of  trips to  that destination, but at the



expense  of alternative destinations.  That  is,  the  total  number  of  trips  is



exogenous.   Second,   the  model  allocates   trips  by  considering the substitut-



ability  between  recreation  centroids,  a  property  particularly  important for



recreation  analysis.   The  proportion of trips emanating from  i  with destination



j  is a  function  of  the  attractiveness  and  spatial  impedance of destination j



relative  to  that  of  alternative   recreation  centroids  in  the  system.    As



reflected in  the  denominator  of  Eq.   (III.l),  all  sites  in  the  region are



considered  as  potential  substitutes being  analyzed.   This  property,  plus the



definition  of substitutes  in terms of both travel  distance and attractiveness,



make the gravity  model appealing for a  regional  recreation  analysis.  Because



the  quantity  of recreation demanded  at each  site depends on  the same  variables



that are  in  the  gravity model,  it is   important  to incorporate  this inter-



dependence  in the  overall  model.



      When applied to transportation  problems, the  dependent variable  is trips;



however,  the variable  of  interest in recreation studies  is recreation days  or



activity days.    In   this  study, the terms  will  be  used  synonymously,  and  a



distinction  will  be made  only  for  trips of more  than  one  day.   Origins and



destinations  are  often defined  as  zones or centroids.   The  term population






38

-------
centroid  is  used to define  the  origin zone, and recreation  centroid  to define



recreation zone.  In each  case,  a centroid  is  a  point but is used to represent



origins and destinations of the neighboring area.



     The  rationale for  using a gravity model is  that  the estimated trip inter-



change  matrix  (T-.)  serves  as  an  input  in  estimating  a large  number  of



travel-cost demand schedules.  Each column vector in T. .  estimates the number of



recreation activity  occasions produced  at origin i with  a  specific  recreation



destination.   In  this  study  j =  1,2, ..., 195 so 195 demand curves can be esti-



mated for  each  of the  four  activities considered.   Because  destinations can be



added to  the analysis,  the  model potentially  can  estimate  a demand  curve and



recreation  value  for   each  activity  and  for  any  site  in  the  region.   The



construction of the gravity model input data is explained in Section 2.








2.   Gravity Model Input Variables



     The three gravity model  input variables are developed in this section.   The



spatial  impedance  variable   (F..)   is  discussed first,  followed  by  the  trip



production model  (P.) and the attractiveness model (A.).








A.   Fraction Factors (F..)



     A necessary  input to construct the F. . terms is the impedance matrix (I..),



which  contains  the  minimum  driving  distance  from  each  population  centroid



(internal  and  external  to  the region) to  each  recreation centroid in the three



and  one-half state  region.    This  matrix was  estimated  by  first  defining the



population  and  recreation  centroids  of  each  county,  and  where  appropriate,



multirecreation  or multipopulation  centroids were used per county.  There are a



total of  129 counties  in  Washington, Oregon,  Idaho,  and  western Montana, but



there are 141  internal  population  centroids and 195  recreation centroids.  In






                                                                              39

-------
this model  the remainder  of the  United States and  Canada is  divided  into 14


external  zones,  so  there  are  a  total   of  155 population  (origin) centroids.


Table  A.I in Appendix A  lists  the population centroids  by  name and county and


gives the corresponding population.  Table A.2 lists the  recreation  centroids by


name and  the corresponding county.  After each centroid  was defined and  located


on  a  highway map,  a network was constructed to show the  distance between  inter-


sections  along major  roads.   Possible  routes from  each  population centroid to


each  recreation  centroid  were thereby identified.  A  computer program was  used


to  solve  for the minimum driving distance between each population and recreation


centroid.    The  resulting  travel  distances constitute   a  155-by-195  impedance


matrix.   Each column vector in this matrix  denotes  the  minimum one-way  mileage


from  each population centroid to  a specific recreation centroid.  The impedance


matrix is an  input  in  the gravity model, and the column vectors  in the  matrix


will  also be used  as  inputs in the travel-cost demand curves.


      The  F.  . variable in  (III.l)  reflects  the  influence  of travel  distance (or


time)  on the propensity to travel.  This variable is estimated  as the dependent


variable  in  a trip-length,  relative frequency distribution, which is also termed


a decay curve.


      Our  1980  regional  household survey  included  a  question  on  the  one-way


travel distance  in miles  for  each recreation  trip.   Because  the  sample  size


exceeds  3,000  and  several  persons  in  each  household may  have taken  numerous


trips, only  a subsample  of  the   sample  results  is  used  to  estimate the decay


curves.   We  sampled  every fifteenth  questionnaire  and  recorded  the  number  of


activity  days by type  and  the corresponding  one-way miles traveled.


      The  widespread use of gravity models  has  resulted  in serious  study of the


shape of  decay  curves and ways   to estimate them.  One  approach  is to  use the


power function F.. =  Po0,-,-  where  F..  is the proportion  of  trips  from  i  to  j
                 ' J       ' J           J



40

-------
and  D..  is   the  corresponding  distance.   Another option  is  the  exponential



function F. .  =  p0e  ^1  ij.   Either of these functions may be adequate, but quite



often decay  curves  are  humped and  highly  skewed  to  the  right.   For instance,



people  are more  likely to travel 40  to  50  miles to camp than to travel 5 to 10



miles, particularly if they live near city center in a large city.



     The preferred  decay curve  model of most researchers  is  a gamma distribu-



tion, which is a combination of the exponential and power functions:
(III.4)                  F... = P0D?J.e"p2Dij  .
The  px  coefficient may  be positive  and  thereby allow for a  peak  in the decay



curve.   This  specification is  used to estimate a decay curve  for  each of the



four activities being considered.  The results are presented in Table 1.



     The  coefficients  for the  exponent  are  not negative as  expected, nor are



they  statistically  significant.    The  R2  values  indicate  that each of  the



equations  has  rather   low explanatory  power.   These  apparently  discouraging



results  are  easy  to explain.   The raw data do not depict the above relationship



for  three of the  four  activities.   Consequently, one of the  reasons for using



the  gamma distribution  is not  applicable  to  those  data.    Also,   respondents



tended  to round off  their distance  traveled on  long  trips  to  the  nearest 50



miles.   For example, respondents indicated a  total of 522 recreation  days at 300



miles and no  recreation days  at 310  or  290 miles.  The tendency for long trips



to  consist of  "spikes"  (and  zeros)  means  that  the  regression  estimate is too



high  in  the  tails.   The consequence  of  using the gamma estimates in Table 1 in



the gravity model would be to allocate far too many people on long trips.



     The  data error caused by respondents' rounding distances to the nearest 50



miles implies that  any specification estimated with ordinary least squares would



not yield a  good fit.   Consequently, the  decay  curves are estimated here using






                                                                              41

-------
                                   TABLE 1

       REGRESSION  ESTIMATES OF A GAMMA SPECIFICATION OF THE DECAY CURVE

Activity
Swimming
Camping
Fishing
Boating
Intercept
6.73
(8.22)
5.76
(7.46)
6.77
(7.85)
5.98
(6.78)
Power
-1.30
(-2.58)
-0.051
(-1.05)
-1.18
(-2.20)
-1.09
(-2.03)
Exponent
0.03
(0.99)
0.01
(0.28)
0.02
(0.49)
0.03
(0.71)
R2
0.34
0.10
0.34
0.25

Note:   The numbers in parentheses are t  values.  R2 is the  coefficient of deter-
mination.
                                  FIGURE 6

               ESTIMATED DECAY  CURVES USING EXPONENTIAL SMOOTHING
                                SWIMMING	
                                CAMPING
                                FISHING
                                BOATING	
             0
              0
50   100  150  200  250 300  350  400  450 500
                    MILES
42

-------
exponential  smoothing.   In this  procedure,  the estimated  proportion of people



traveling any distance is equal to the sum of the proportion of people traveling



the previous x  distances  divided by x.   After experimenting with x = 5, 10, and



15,  it  was  decided  to  smooth  over  the previous  10  distance groups,  where



distance  is  also measured  in  10-mile increments.   The estimated  decay curves



using exponential smoothing are depicted in Figure 6.  The trip-length frequency



distributions  in Figure  6 show  that  recreationists who  swim,  fish,  and  boat



strongly prefer to travel short distances.   In contrast, the camping decay curve



is peaked, with most camping trips occurring between 50 and 100 miles.



     The  main  use of these  decay curves is to  transform  impedance values  into



the  F..  matrices.   By  substituting  each  impedance value  into  the  four  decay



curves,  an F..  matrix  is constructed  for each activity.   This matrix  is  one



input  in  the  gravity  model,  Eq.  (III.l).  The estimates in an F.. matrix can be



interpreted as the probability that a recreator residing in origin  i will travel



the distance from i to destination j.








B.   Trip Production Model (P..)



     A  household  recreation  survey was  conducted in the  fall  of 1980 to obtain



data  to  estimate  recreation   trips  produced  by  origin  by type  of  trip.   A



telephone  survey was  undertaken  by the Survey  Research  Center  at  Oregon State



University, specifically  for  use in this model.  Appendix  B contains a copy of



the questionnaire.  A statistical methodology used to construct trip-production



estimates  was  developed  by  Carter (1981)  as part of her  dissertation.   In the



methodology she  developed,  the sampling unit is the county, not the individual



or  household,   as  is  typical  in  most  studies.   Forty counties  (out  of  119



counties  in Washington,  Oregon, and Idaho) were sampled,  with  the average size



of 75  households  per  county.   A  recreation  trip production model,  based on the






                                                                              43

-------
40  sampled  counties,  was  then  used  to  extrapolate to  the  remaining counties.



Because  trip  productions  are estimated  to be  negative for few  counties,  the



overall  reliability of  the estimates cannot be  confirmed.   An  alternative trip



production model is developed here, estimated with the Oregon State survey data.



     In  most  recreation  participation  analyses,  the  sampling  unit  is  the



individual, and a specific activity is being considered.  Because a high propor-



tion of  individuals generally do not participate in the  specific activity, there



is  a  corresponding  large  number  of  zeroes.    The  assumption of  normality is



therefore  likely to  be  violated.   Most  resarchers   have  employed  a two-step



procedure.   First,  a dichotomous  dependent variable  is used to  denote whether



the  person  participated and, for those persons  who  participated, the  number of



days   participating  is  the  dependent  variable  in  the  second  model.   The



independent  variables  in  these  models are  demographic, such  as  age, sex, and



income,  and some measure of  the supply of recreation opportunities.



     A  common  and  serious  problem  shared  by  these  models  is  their very low



overall  explanatory power.   For instance,  Davidson,  Adams, and Seneca  (1966)



obtained R2  values  of 0.28,  0.11,  and 0.11  for  the probability of participating



in   swimming,  fishing,   and  boating,  respectively.    Hay and  McConnell   (1979)



obtain R2 of 0.02 and  0.03  for  participating  in nonconsumptive  recreation  such



as  wildlife  photography.    Cicchetti  (1973) reports  the goodness  of  fit for



several  recreation  participation  equations  (p.  69, 73,  75),  and each is below



0.18,  and several  are  less  than  0.10.   In previous  versions  of this model,  I



used the conventional  two-step  procedure and  obtained  unsatisfactory  results.



      In   addition  to  the  statistical  difficulty  of   a  large  number  of  zero



responses,  there may be conceptual  difficulties  with focusing on  individuals and



single activities.  Where  the family  unit recreates together, individuals do not



act independently.   Also,  household members may  participate in  several  activ-






44

-------
ities during one  recreation  trip;  hence activities, like household members, may



not be separate and independent.



     An  implicit   assumption  in  previous  participation   analyses   is  that



participation per  capita  varies  across  regions.   However,  this  assumption has



apparently  been  untested in  the  literature.    One  estimator  of  per  capita



participation  is  the  sample  mean  number  of  recreation  days  per  person.



Considering  the  low explanatory  power of most  recreation  participation models,



this  estimator  may  be  quite  reasonable.   As  a minimum,  one  should  test



statistically whether  mean participation varies  across  geographical  boundaries



within  the  sample  region  before  estimating  a  regression  model.    If  the



hypothesis  of  equal means cannot  be rejected,  the  regression approach will  be



futile and the sample mean becomes the appropriate and certainly most convenient



estimator.



     The recreation participation model developed here differs from those in the



literature  first  by using the household  as the sample unit and  by  focusing  on



recreation  trips  as a composite  variable and then explaining the composition  of



a trip  by  activity.   The conceptual rationale  for  focusing  on  the household  is



that  recreation decisions may often be joint decisions  where  the entire family



participates.   The  probable  interdependent  decisionmaking  within  the  family



suggests that  the  household  is  a  more  appropriate unit  for analysis  than the



individual.  The  statistical  benefit of focusing on households is the increased



probability  that  at  least one member of the household  participates  in recrea-




tion.



     By considering a composite  of  recreation  activities,  the  probability that



someone in the household participates is again  increased.  A zero  response  is



obtained only  when  no one  in the  household participates  in  any of  the four
                                                                              45

-------
activities.   The number  of  zero responses in most  studies  and the conventional



two-step estimation procedure is no longer necessary.



     More than one recreation activity is often undertaken during one recreation



trip.  For  example,  during  a weekend camping trip, some family members may fish



and  boat  while others  swim,  and  some  family members  may  enjoy each activity.



The  demand  curve  for recreating by a single  activity may be different from one



for  the same activity where other activities also occur.  The interdependence of



recreation  activities will  be  considered  by analyzing  recreation  trips  as  a



composite and then explaining the activity composition of these trips.



     Consider  first  the  possibility  that  the  most  appropriate  recreation



participation  estimator  is  the  sample  mean  of  trips  per  household.   The



hypothesis  that populations  have the same participation rate can be  tested by a



one-way  analysis  of  variance.   The formal  statistical  hypothesis  is  that the



mean number  of  trips per  household is  constant  across subregions  within the



total  region.   The first test  is whether the mean number of trips per household



is  constant across the three states.  The  sample  means equal 5.5, 5.5, and 8.6



for  Oregon, Idaho, and Washington, respectively, for summer trips and 1.6, 1.3,



and  2.9 trips  per household for  winter trips.   The  observed F statistics are



20.23  (summer) and  17.55  (winter), which reject the  hypothesis of  equal means



across the  three stages.



     The  second  test is  whether  the  mean  number  of  trips  per  household is



constant  across counties for  the 40 counties  sampled.  The observed  F statistics



are  3.67 (summer) and  2.90  (winter),  which are larger than  expected at the 95



percent  level if  the means  were constant.   The third  hypothesis is that means



are  equal  across  counties  where counties are grouped  by state.  Reporting the



summer  F values first  and  the winter values second,  the F statistics are 2.42



and  2.90  for Oregon, 5.11 and  1.85 for  Idaho, and 1.93 and 1.18  for  Washington.






46

-------
Each of these  F  values suggests rejecting the  hypothesis  of equal means at the

95 percent  level.   However,  some of the F values are close to their theoretical

value,  which is not true of the above two tests.1

     The  implication  of  these  tests  is  that recreation  participation  (in

camping, fishing,  boating,  and swimming) differs across the three states in the

region  and  between  counties   within  each  state.    The  main   source  of  this

variation comes  from  Washington residents who recreate more on the average than

Oregon and Idaho residents.

     Because  mean  trips  per  household  are  apparently   not  constant  across

counties  in  the region,  the nonrandom  variation  in household  trips  should be

explained.  The number of trips per household (summer plus winter) is postulated

to  be  a  linear  function of demographic and recreation  supply  variables.   The

only  demographic  variables  included   in  the  model  are  household  size  and

household  income,  because these  are the  only  demographic data  for which data

were collected.

     The  relevant  supply  measure  of  recreation   opportunities  includes  the

necessary  recreation  facilities and  the distance of these  facilities  from the

population  centroid.    The  recreation   facilities  used  here are:   number  of

camping  units,  boat  ramps,   linear  designated  beach  feet,  and  river-plus-

shoreline miles  for camping, boating, swimming, and fishing, respectively.  The

recreation  supply  variables,  defined as recreation accessibility, are estimated

as  a function  of the availability  of  facilities,  and the willingness to travel

the necessary  distance to these facilities.   Let F..  denote the probability of
                                                     J
 xAn examination  of the raw data  indicated  that six households reported taking
more than 150 trips during either the summer or winter.  These observations were
treated as  outliers and the analysis of  variance  tests were rerun.  Even after
omitting these  six  observations, all the  above  hypotheses  were rejected at the
95 percent  level.


                                                                              47

-------
driving the distances from population centroid i to the jth recreation centroid.


The recreation accessibility  of each population centroid (RA.) for one activity


is  estimated  by  summing  recreation  facilities  (Fac.)  over  all  recreation
                                                        J

centroids in the region weighted by the probability of driving the corresponding


distances.  That is,




                                    195

(III. 5)                       RA.. -  I  F^. FaCj

                                     J




where  i  = 1,  2,  ..., 155 and where RA. measures the accessibility of recreation


opportunities  to  the  ith  population  centroid.   Equation  (III. 5)  must  be


estimated  separately  for  each of  the  four activities  because  the  friction


factors  (F. .)  and  facilities  are  unique to  each  activity.   Using Eq.  (III. 5),


recreation accessibility was estimated for each activity and for each population


centroid  in the region.


     As   a  measure  of  the  supply  of  recreation  opportunities,  recreation


accessibility   has  some   commendable   properties.    First,   every  recreation


destination  in  the  region  is  considered  in  this  measure.    Second,  these


opportunities  are  summed,   but weighted  by  the  probability  of driving  the


necessary distance.   Limitations of  this measure  are  the  data requirements to


estimate  it and that congestion  is ignored.


     From the above-defined variables, the trip production model is expressed as
 (HI. 6)           TJ = f  (HS.j, Y-, RAc, RAf) RAb, RAg , Dl5 D2)
where  the variables are defined as:
            T. =  number of trips produced by  household  i,


           HS. =  number of people  in  household  i,


     RA   , .    =  recreation accessibility   for  camping,   fishing,  boating,  and
         '  '  '     swimming,
48

-------
            D! = dummy variable = 1 if Oregon, and 0 otherwise, and

            D2 = dummy variable = 1 if Idaho, and 0 otherwise.


The state  dummy variables  are included because the analysis  of  variance tests

revealed recreation participation rates vary across states.

     The number of  households  surveyed exceeded 3,000, which  yielded more data

than  is  necessary  for  regression  analysis.  Those  respondents  who  failed  to

answer a question,  particularly on family income, were deleted as were one-half

of  the  remaining  responses.   Using a  sample  size of  1545  households,  a trip-

production model is estimated to be:


(III.7)  T. = 5.71 + 0.983 HS. + 0.879 Y. + 0.0001 RA  - 0.014 RA  + 0.0008 RA
          1         (4.35)   1  (3.84)  "*  (0.14)    S (-2.55)   C  (0.14)

             + 0.346 RA  - 1.695 Di - 3.345 D2
              (2.01)     (-1-25)    (-2.83)


where t  values  are  in parentheses and R2 = 0.053.  The encouraging results from

Eq. (III.7)  are that  household size and  income  have  positive coefficients that

are  highly  significant.   Unfortunately,  only  one  recreation  accessibility

variable (boating) is significant and of proper sign.

     The main  purpose of  Eq.   (III. 7)  is  to estimate  the number  of trips per

household for each population  centroid in the region.   Because the model will  be

used for estimating  purposes,  it should not contain insignificant coefficients.

After eliminating the insignificant variables, the model becomes


(III.8)     T. = 5.005 + 0.993  HS. + 0.876 Y. - 4.084 Da - 3.053 D2
             1           (4.399)   1  (3.846) "" (-4.709)   (-3.865)


where R2 = 0.049,  and where household size and income remain highly significant.

The negative coefficients for  the dummy variables are consistent with the
                                                                              49

-------
analysis of  variance  result  that  participation  rates  differ across  the  three

states.

     The recreation accessibility variables do not appear in Eq.  (III.8) because

they are not  significant.   Recreation facility variables are subject to serious

measurement errors, which  at  least partially explains their estimated insignif-

icance.   An  implication  of the insignificance of the accessibility variables is

that  increasing  recreation facilities  will  not  cause  people  to increase their

participation, although they may redistribute their demand for recreation sites.

     Equation  (III.8)  is  used to  estimate the  expected number  of recreation

trips produced by  household  for each  county  in the  three-and-one-half  state

region  and western Montana.   Census  data for 1980 on  household size by county

and  1979  Department of  Commerce  county  income  data were  substituted into Eq.

(III.8) to  estimate trips  per household by county.   The number  of households by

county—obtained  from  the  1980 census—was multiplied by trips  per household to

estimate total trips per county.2

     The Oregon  State University  survey data were also  used  to allocate total

recreation  days  by county to the four  activities:   camping,  fishing, boating,

and  swimming.   Treating  each  state separately,  frequency  distributions  were

constructed  showing the proportion  of  days  of  participation in  each activity

(see  Table B.I in Appendix B).  These proportions were then multiplied by total
 2Total  county trip  data  were transformed into total  recreation days by  first
multiplying  trips  by  the  average  length  of  stay.    For  Oregon,  Idaho, and
Washington,  the  sample survey estimates are:  2.439,  2.194,  and 2.453 days per
trip,  respectively.   The average size of  a  recreation party is  estimated  to be
the  mean household size,  which  is  2.60,  2.85, and  2.61 for Oregon,  Idaho, and
Washington according  to the 1980 census.   Total  recreation  days per  county are
estimated as the product of total trips,  average  length of stay,  and number of
persons  per  trip.   For  the  three-state  region,  households  average about 8.6
trips  per  year  and,  considering household size and  length  of stay,  about 55.4
recreation activity days per year.


50

-------
recreation  days  by  county to  estimate  number  of  days  by  activity  for each


county.


     Estimates  of   activity  days  were  also  constructed for  ten  counties  in


western  Montana.    Regional   mean  sample  data  were  used  to  produce  these


estimates.   The estimates of  recreation  trips  produced by  activity  and  by


population centroid appear in Appendix A, Table A.3.





C.   Attractions Model (A.)
                         J

     The  gravity  model  also  requires  an estimate of  the attractions (quantity


demanded)  of each  recreation  centroid.   Attractions  are postulated to  be  an


exponential  function  of   recreation  facilities and the  accessibility  of  the


recreation  centroid,  which measures  the likely demand on that centroid.  Demand


for  recreation  sites  tend to  vary  inversely with  the  distance  to population


centers.   The  responsiveness  of  attractions  to  changes in  facilities  should


therefore  be  positively  related to the  nearness of  these facilities to popula-


tion  centers.   Furthermore,   attractions   should  respond  to  increments  in


facilities at a diminishing rate, because demand cannot increase indefinitely in


proportion  to facilities.  The  attractiveness model  is specified in exponential


form  to allow  for  the  diminishing  returns  effect  and  the  interaction between


facilities and accessibility.


     Accessibility  of  recreation  centroids,  called population accessibility, is


a  function of the  number  of trips produced  by each  population centroid and the


likelihood  that these trips  will  terminate  at  that  recreation  centroid.   The


accessibility of each recreation centroid is estimated by  summing trips produced


(P.)  by  all  population  centroids weighted  by  the  probability of  driving the


distance  to  the  recreation centroid.  That is, population accessibility for the


jth centroid is




                                                                              51

-------
                                        155
(HI.9)                           PA.. =  I  F..P.
where the  F-- values  are  obtained  from  the decay  curves.   Estimates from Eq.

(III.9) were constructed for each recreation centroid in the region and for each

of  the  four activities being analyzed.   Population  accessibility estimates are

one input in the recreation attractiveness model.

     The attractiveness model  also  assumes that demand  at a site is  a positive

function of the site  characteristic.   The facility  variables  used are camping

units,  river  and  shoreline miles, boat ramps,  and linear designated  beach feet

for  camping,  fishing,  boating,  and swimming, respectively.  U.S. Forest Service

data  on visitor  days  and facilities  by ranger  district  were used with the

accessibility  data  obtained  from  Eq.  (III.9)  to estimate  the attractiveness

model.   As  seen in  the first  four rows  in Table 2,  the accessibility coeffi-

cients  are  significant in  only two  of  the four equations.  This insignificance

is  due  partially  to poor  quality data  because  similar estimates based on older

survey  data showed this variable to be significant.  The  positive accessibility

coefficients  indicate that  use for each  activity is greatest  for those sites

located near  large  production centroids.   The facility  variables  are overall

significant and have positive signs as expected.   As the equations are in multi-

plicative  form,  a positive accessibility  exponent implies that the responsive-

ness  of use to facilities is positively related to the accessibility  of a site.

That is, for  a  given  increment  in facilities, use will  be greatest for those

sites   that  are  most  accessible.   Facility   and  accessiblity  data  for  each

recreation  centroid  were  substituted  into  Eq.   (III.10)-(III.13)  to estimate

relative attractiveness of each centroid  in the region.   The sum of attractions

to  all  sites  estimated by  the attractions model  will  not  likely equal  total
52

-------
                                     TABLE 2

 REGRESSION ESTIMATES OF EXOGENOUS AND ENDOGENOUS ATTRACTIONS (in natural logs)

Equation
Number
(III. 10)
(III. 11)
(III. 12)
(III. 13)
(III. 10' )
(III. 11' )
(III. 12' )
(III. 13')
Activity
Swimming
Camping
Fishing
Boating
Swimming
Camping
Fishing
Boating
Intercept
1.060
(0.943)
-0.396
(0.372)
-5.637
(-2.408)
1.242
(0.698)
-4.052
(-2.309
-2.763
(-2.315)
12.020
(2.419)
-9.716
(3.730)
Recreation
Facil ity
0.194
(2.902)
0.631
(5.460)
0.533
(15.862)
0.586
(3.363)
0.163
(2.585)
0.509
(4.781)
0.545
(16.199)
0.621
(4.287)
Recreation
Access.
-0.216
(-0.721)
0.466
(2.123)
0.354
(1.956)
0.691
(1.394)
0.576
(2.487)
0.591
(3.955)
0.248
(2.636)
1.408
(4.210)
Coef. of Det.
Sample Size
R2 -
n =
R2 =
n =
R2 =
n =
R = 0
n =
R2 =
n =
R2 =
n =
R2 =
n =
R2 =
n =
0.18
42
0.41
49
0.78
74
.25
36
0.29
42
0.52
49
0.79
74
0.47
36

     Note:   the  numbers in  parentheses  are t  values.   The  dependent variables
are  activity days  for swimming,  camping,  fishing,  and  boating,  respectively.
The  first  independent  variable is the facility variable, which is linear desig-
nated beach feet (BF.), camp sites (CS.), acceptable river miles (RM-), and boat
ramps  (BR.)-   The  second  independent' variable  is  accessiblity  for swimming,
camping, fishing, and boating, respectively.


trips  produced   in  the  region.   An  accounting  identity  and condition  of the

gravity model  is that  total  trips  produced  equals  total  trips  received.   The

attractions  model   therefore   estimates  relative  attractiveness,  and  these

attractions are scaled to sum to total trips produced.

     The three inputs  in the gravity model, P., A.,  and F-- have been estimated

with a  trip  production model,  a trip  attractions  model,  Eq.  (III.10)-(III.13),

and  by  transforming  the impedance matrix with  the decay curves.   The output of


                                                                              53

-------
the gravity  model  is a  trip  interchange matrix (T. .)  that  gives  the number of



trips  emanating  from population  centroid  i  with  recreation  centroid  j as the



destination.



     The  statistical  estimates of the  attractiveness  model  are unimpressive in



terms  of overall  explanatory  power and  in  the  failure of  the  accessibility



variable  to  be positive and significant.   Recreation  data are typically of low



quality  and  the  data used in  the  attractiveness  model  are  no  exception.  In



addition,  there  may  be  a specification  problem with  the attractiveness model.



The  gravity  model  has the desirable property  of distributing  trips according to



the  attractiveness  of a recreation  site  relative to all  substitute sites in the



region,  and  according to effect of  distance to the site  (R^-,-). relative to all



sites  in the region.   The gravity model  includes the  effect of substitute  sites



in  terms of  relative travel distance (or travel time)  and relative attractions.



Incorporating  this  property  into  the  attractiveness  model  would  be highly



desirable.   Because  one  input for  this  extension  results from calibrating the



gravity  model, a  discussion  of this  calibration  procedure  is  provided first.








3.    Calibrating  the  Gravity Model



      A trip  interchange (T. •)  matrix is  illustrated in Table 3.   A row depicts



the  number of  trips  received by each destination centroid emanating from a  given



origin.   Similarly,  the columns depict  the number  of  trips emanating  from each



population  centroid with a given  destination.  Because the  region  is defined to



be  closed,  the  total number  of  trips  produced must  equal  the total  number of



trips received,  which in turn  equals the total sum of trips  in the trip inter-



change matrix.




      Unfortunately,  the  best estimates  of T..  are not  obtained simply by substi-



tuting the input data  into the gravity model  [Eq.  (III.l)]  and  solving.   First,





54

-------
                                     TABLE 3


                             TRIP INTERCHANGE MATRIX

j

i
1
2

Trip Interchange Matrix (T..)
"i J

1 2 ...
TH T12
j j
21 22



m
Tlm
T
'2m
Trip
Productions
pi
ITlj = Pl
VT D
II 2j P2
                  Tnl      Tn2              '        '       Tnm        ZTnj    Pm
                                                                     J
Attractions      i        i                                              J

    A-           = Ax    = A2                                        =IITii
     j                                                                 i j    J
the estimated  trip-length  (miles  one way) frequency distribution  obtained  from


using  the  estimated  T..  values  and  the impedance  matrix  typically would  not


correspond with  the  assumed known  distributions,  that  is,  the  decay  curves.


Second, the estimated number of trips received at each  recreation centroid would


not correspond with  the  attractiveness  input data,  which means  that the sum of


the column vectors in Table 3 would not equal A,.
                                               J

     The gravity model is therefore calibrated with  an  iterative technique where


a  new  trip  interchange  matrix  (T-.)  is   estimated  by  each   iteration.   The


elements  of  the  new T. .  matrix  are  used to  estimate a trip-length  frequency


distribution and  are  summed vertically  to estimate  A..   These  estimates  are
                                                        J

compared with the assumed known decay curves and A.  values,  and  if a significant
                                                  J
                                                                              55

-------
discrepancy  exists,   the   iterative  process  continues.   The  gravity  model  is


calibrated to  produce a T..  matrix  that yields a  decay  curve corresponding to


the exogenous decay curve and estimated attractions that correspond to  exogenous


attractions.   When  the estimated  and observed  A.  values and decay curves are
                                                 J

satisfactorily  close, as  judged  by  some  predefined criteria,  the iterations


conclude.


     To  define  this  calibration  technique  more  precisely,  recall  that the


conventional gravity  model  includes  the constraint, Eq.  (III. 2), which in  terms


of Table 3 is




                          M

(III. 14)                  IT,. = A.   , for each j.
                           n  J    J




Each   iteration  of   the  gravity  model  necessarily  satisfies  the  production


constraint,  Eq.  (III. 2), because the  ratio component of Eq. (III.l) sums to one.


However,  Eq. (III. 14)  is  not generally  satisfied  by the  first or  even second


iteration.   The calibration  technique brings  the  estimated  and observed trip


length  distributions together  and  also  satisfies  Eq.  (III. 14).   In  each


iteration,  attractions are multiplied by the coefficient bc,  which reflects the


discrepancy  between  A.  and IT.,  estimated  in  the previous   iteration.   This
                       J       i   \J

adjustment coefficient is  obtained from
 (III. 15)                        bc = b0'1 ZT  c-1

                                         i 1"J



 where c  designates  the number  of the  iteration.   The attractions  for  each


 iteration  after the  first  iteration  are  estimated by multiplying the  previous
 56

-------
attractions  by the  adjustment  coefficient  obtained  from  Eq.  (III.11).   This



procedure results in Eq. (III.10) being approximately satisfied.



     According to  this  conventional  calibration technique,  the number of trips



received by  each  recreation centroid is exogenous, and the gravity model solves



for the distribution of recreation travel.  The number of trips received by each



recreation  centroid is  estimated by  the  attractiveness  model,  Eq.  (III.10)-



(III. 13), on  the  basis  of facilities at the site and accessibility of the site.



The  attractiveness  model,  as defined thus far, does  not consider the effect of



substitute sites as does the gravity model.



     A procedure  similar  to Eq.  (III.15) is used to adjust the friction factors



F...  The  travel  distance factors used in the cth iteration (F..c) are equal to



the product of the factors used  in the previous iteration (F-.   ) and the ratio



of observed to calibrated trips  which occur from i to j.  That  is,





(HI.16)                       F..C = F0:1^
                                 ij     ij  GM





where the  numerator is  the  percent of  trips  implied  by the decay curves and GM



is the percent of trips for  the  same distance that is predicted from the gravity



model.  The  gravity model is calibrated  using  an  iterative  approach as defined



by  Eq.  (III.15)  and (III.16).   Three iterations  are  generally required for the



trip   interchange  matrix   (T..)  to   approximately   satisfy  the  attractions



constraint,  Eq.  (III.3),  and to produce  a  decay  curve that closely corresponds



with the observed decay curve.



     The  empirical  estimates   of  the  attractiveness  model  in  Table  2  are



disappointing, particularly  because two of the accessibility coefficients failed



to  be  significantly  positive  as  expected.   Recall  that   accessibility  is



estimated  as the  sum  of trips  produced  weighted by  the  F. .  values,  which are



probabilities  of  driving  various distances.   The  F. .  values are estimated from




                                                                              57

-------
decay curves, which in turn are estimated with regionwide trip-length data.  The


decay  curves  are  probably  an  accurate  representation  of  recreation  travel


overall, but  they  are not  necessarily  accurate for any  individual  site.   If a


recreation  site  is  close to  a large  urban  area,  most  trips will  have short


travel  distances,  and the tail of  the  decay  curve  will  terminate  close  to the


origin.  Alternatively,  if all origins to a  site  are  several  miles away, the


appropriate  decay  curve  must  reflect  a  large  area under  these corresponding


distances.


     The  attractiveness   model  estimated  above  presumed  that  a  decay curve


estimated  with regionwide data would be applicable to  each site.   A preferred


alternative  is to  estimate  a  decay  curve  for  each  site  which  reflects the


influence  of substitute sites.


     The gravity model produces a T.. matrix (Table  3), but  it also estimates an


F..  matrix via the iterative procedure.  An F-. matrix is a gravity model input


variable  and  it is based on  a single  regional decay  curve.   The algorithm for


computing  T. . is  iterative,  and it continues to  adjust the F..  values until


estimated  attractions balance  with  A-  and the decay curve  implicit in the T..
                                      J                                         J

matrix balances  with the regional  decay  curve.   The  iterative  calibration


process  [Eq.  (III.16)] results in a  new F-. matrix  in each  iteration.  Implicit


in   this  matrix  is   a  decay  curve  that  is  unique to  each  site.   The final


iteration  produces  an F.. matrix where each column  vector implicitly contains a


decay  curve unique to the corresponding destination.   As  these F.. values are


computed  by the gravity  model, they  reflect  the  influence of the  independent


variables  in the gravity  model.


     The  gravity  model was  estimated  using the  input variables defined  above,


including  the  attractiveness variables  predicted  from Eq.   (III. 10)-(III-13)  in


Table  2.   From this  version of the  gravity model,  the estimated  F..  values  were



58

-------
obtained.    These  values  were  then  used  to  reestimate the  recreation access-

ibility measure and then to reestimate the attractiveness model.

     Empirical  estimates  of   the  second  version of  the  attractiveness  model

appear  as   Eq.  (III.10')-(III.13')  in  Table  2.   The  explanatory power  of the

model,  as  measured by  R2,  shows  an improvement  in each  of  the four equations

over the previous estimates.   Each of the accessibility coefficients is positive

and  is  significant  at the 1 percent level.  Overall, on empirical grounds, this

two-stage  procedure  for  estimating  the  attractiveness  model  results   in  a

dramatic improvement  in  the model.3  On theoretical grounds,  the model is also

improved because the  same varibles that determine  the distribution of recreation

travel  also  influence   total   demand  at  each  site.   In  addition  to  being  a

distribution  model,  the  gravity  model,  along  with the  attractiveness model,

becomes a trip demand model.

     The gravity model as estimated  in this study  produces  two  outputs necessary

to  estimate demand  and  benefits for recreation sites.   First,  quantity demanded

is  estimated  for each centroid and for each of the four activities.  By changing

the  level   of  facilities  at  a centroid,  the attractiveness  of  the  centroid

changes  [Eq.  (III.10')-(III.13')] and,  through the gravity  model,  so does the

total number  of trips received.  For each recreation centroid,  the gravity model

also  estimates  the number  of trips received from each  origin.  These data are

transformed into visit rates and are a critical output in estimating travel-cost

demand  curves.   Estimating a  gravity model  requires  constructing an  impedance
 3Three of the four equations  in Table 2  use data from only 49  ranger  districts,
whereas the  fishing  equation  is based on 74 observations.   Destinations  on  the
original  highway  network  conformed  to   only  49 ranger  districts.   When this
network was  expanded  to  include all  ranger districts,  and  a larger  impedance
matrix  was  constructed,  the  new  attractiveness equation  (except for  fishing)
failed to show a statistical improvement.   For this  reason, only  the new fishing
equation is  used.


                                                                               59

-------
matrix, which reflects  the  minimum travel distance from each origin (population



centroid)  to each  destination  (recreation  centroid)  in  the  region.   These



minimum  travel   distances,  when  multiplied  by  travel   cost per  mile,  yield



travel-cost  estimates  that  are necessary to estimate  recreation demand curves.
60

-------
                                   CHAPTER IV



                  ESTIMATING AN OUTDOOR RECREATION DEMAND CURVE








     In Chapter II recreation benefits are defined as net willingness to pay,  or



alternatively as consumers'  surplus, and measured as the area under a recreation



demand curve and  above  the  market price.   A detailed explanation of the travel-



cost method of estimating a recreation demand curve is presented in Section 1  in



this chapter.  Recreation demand curves and net willingness to pay are estimated



for each of  195 recreation  centroids and  for  each  of the four activities being



studied.   A  gravity model  of  recreation  travel  was developed  in  Chapter III.



The purpose of this model is to estimate recreation trips by origin to each site



in the region,  and thereby  to provide an input in estimating travel-cost demand



curves.  A  sample of these  demand  estimates  is presented in  Section  2 of this



chapter.   Section 2 also includes a discussion of the significance of substitute



sites as well as disaggregating recreation into four specific activities.








1.   Estimating a  Travel-Cost Demand  Curve  and  Consumer  Surplus:   An Overview



     Willingness  to pay  for  a  recreation  site  can  be estimated  directly  or



indirectly.   In the  direct  approach an interviewer confronts the recreationist,



and  using   an   appropriate   survey  instrument,  asks  the  recreationists  their



willingness  to  pay.   There  are some numerous and impressive case studies of the



direct approach,  but  for purposes  here,  it  has  two serious  limitations.   An



expensive and time-consuming survey must be undertaken  for  each site analyzed.



Also,  it is  particularly difficult  to estimate potential  benefits  of  a site






                                                                              61

-------
which  doesn't  exist  or  to  estimate  increased  benefits  from   the  potential
improvement  of  a site.   The  critical  need to assess  potential  benefits over a
large  number of sites precludes  the use of direct  estimates  of willingness to
Pay.
     In the  travel-cost method (TCM), willingness to pay is estimated  indirectly
on  the basis of observed travel  patterns,  and not, as  in the direct approach,
from what  people  say  they would do  in response to hypothetical situations.  For
this reason, most analysts have preferred the travel-cost  approach to  the direct
approach.   Although the  TCM has  numerous  limitations,  some  of  which  will be
dealt with here, it will serve as the basis for estimating recreation  demand and
value.   The  rationale  for using the TCM is first its credibility, which results
from  its widespread use  and official  sanction  by  the  Water  Resources Council
(1979).  The objective of this study is to develop,  test,  and apply a  model  that
can  estimate recreation  demand and  value at  any  site  in  a large  region.  There
are  no viable alternatives  to the TCM in terms of models  that are theoretically
sound  and  operational on a regional  basis.
      In  this study, travel-cost  demand curves  are  estimated for each of  four
activities  (fishing,  swimming,  camping,  and boating)  and  for  a large number of
sites,  which are termed recreation  centroids.  A travel-cost demand schedule is
now  developed,  but the notation  is  simplified  by assuming one activity and one
recreation centroid.  Let T.  be the  annual number of visitor days  emanating  from
the  ith  population centroid and  recreating at  the  site being analyzed, and let
N..  be the population of  the ith population centroid.   Using  C.  for  the travel
cost per person per visitor-day from the ith zone, the equation

(IV.1)                            T1/N1 - f(C.)
62

-------
relates visit  rates to  travel  costs.   Equation  (IV.1)  is  the  general  form of

what Clawson (1959)  termed the demand curve  for  the recreation experience, and

it  is  often referred  to as  a  per capita demand  curve or visit-rate schedule.

The regression estimate of this equation is used to generate a site demand curve

by  first  multiplying the  equation by the  population  of  the  ith  zone  (N.) to

obtain
                                  n. - f(C.)N.   ,
then summing all origins to obtain
 (IV.2)                          ZTn. = If(Ci)N1   ,
which  yields  an  estimate  of total  visitor-days  as  a  function  of total travel

costs.

     The  essence of  the  TCM is  that a site demand  curve  is  inferred from the

empirical  relationship of  visit  rates by  origin  to  corresponding travel costs

[Eq. (IV.1)].  Although travel costs  are a  transaction cost, not a market price,

they are  treated as an implicit market price.  The response of total  recreation

days  to  hypothetical  prices  is obtained  by  assuming that recreationists would

respond  to prices  (entrance  fees) just  as they  respond  to  the  same change  in

travel  costs.   To estimate total  visitor-days as  a function of increased travel

costs  or  market  prices, AP  is inserted  in  Eq. (IV.2)  to obtain


(IV.3)                       IT =  Zf(C. + AP)Ni   .
                             i     i


The  prices for  a site demand curve  may  be selected  somewhat arbitrarily,1 but
  *The  issue of  the sensitivity  of consumer  surplus  estimates to  the size  of
price  increment  is  considered  in  Chapter  V.

                                                                               63

-------
should begin at zero and cover the full range of the demand curve.  The quantity



of visitor  days  demanded at each  price  is obtained from  Eq.  (IV. 3)  by  letting



each  price   equal  AP  and  solving  for   the  corresponding  quantity  (IT.).   A



recreation  site  demand curve  can  then be  estimated from  these price-quantity



observations.  The  site  demand  curve  is  usually estimated  as  a  regression



equation obtained from the price-quantity points.  The final step is to estimate



consumers'  surplus,  which  is  typically  the integral  of the  estimated demand



equation.



     The  focus of  this study is on  total  quantity demanded at a zero price and



on  consumers'  surplus, but  not  on the site demand  curve  per se.  Furthermore,



using  regression  analysis to  estimate a  site demand curve  raises  the issue of



the proper functional form.   Also, a regression estimate may be highly sensitive



to the  choice  of hypothetical  prices substituted in Eq. (IV.3).  Because a site



demand  curve is  unnecessary  and regression  analysis  introduces some potential



problems, an alternative procedure is developed.



     The  following chapter will  demonstrate  that a semilog  form of the visit-



rate demand  schedule is reasonably good and  superior  to that of the double-log



form.  Using this  form, Eq.  (IV.1) becomes





(IV.4)                    In (Tj/N^ = a + $C. + e.  .





Taking antilogs of the regression estimate of Eq. (IV.4), multiplying by  N., and



summing yields







(IV.5)                       ZT. =ea + P(Ci+ AP)








which corresponds  to Eq.  (IV.3).   The price increments  used here are $1  from $0



to $4, $2 from $4 to $12, $4 from $12 to $76, or until a successive price incre-
64

-------
merit  increases  consumers'  surplus  by  less  than  one  percent.   Initially,

one-dollar price increments were used from $0 to $76, but experimentation showed

that  most of  the consumers'  surplus occurs at  relatively  low  prices.   Also,

extensive computer time  is required to perform the large number of calculations

required  for 780  (195 x 4), first-stage  demand  curves.   For these two reasons,

larger price increments were used as higher prices.

     The  hypothetical prices and the quantities generated from Eq. (IV. 5) can be

used  to  estimate recreation  demand  and value.2   In   lieu  of  estimating  the

site-demand  curve,  consumers'  surplus is estimated  directly  by applying Bode's

Rule  to  the  price-quantity data.  Bode's Rule is an algorithm for integrating a

fourth degree  polynomial  that  fits five  points equally spaced on the horizontal

axis.   Suppose that  we  are given  five   such points  x.,  where  i  = 0,  ...,  4.

Bode's rule approximates
                                   X4
                                   J  f(x) dx
 by  fitting  a  fourth  degree polynomial  through  the  five  points  (x.. f(x.)).

 Bode's Rule  is3
               X4          Ok
              S  f(x) dx = ^  (7f0 + 32fj +  12f2 + 32f3 + 7f4)
                                 Of6 f-  h7
                       where E =          , x0 < £ < x.
 2The  approach  here follows Clawson's original  two-step  method of estimating  a
visit-rate schedule and using  it to generate a site-demand  schedule, except  that
the  integral  of the site-demand schedule  is estimated without  actually  estimat-
ing  that schedule.  An  alternative and simpler  approach would be to integrate
the  first-state curve directly.

 3Bode's  Rule  is  given in  Davis and  Robinowitz  (1967, p.  30)  and  in Abramowitz
and  Stegun (1964,  p. 886).


                                                                               65

-------
but the remainder E  is  set equal  to zero.   The h term is  the  interval,  which in


our case is the price increment used in  Eq.  (IV.5).


     The use  of Bode's Rule  is illustrated  by  Figure 7.  The first series  of


five equally spaced  points is  the  prices from $0  to  $4 in  increments  of  $1.   The


corresponding  quantities   are  obtained  from  Eq.   (IV.5).    A  fourth  degree


polynomial  is connected to these five points,  and Bode's Rule  is used to measure


the  area  under  this segment  of  the  demand  curve.   The next series  of  five


equally  spaced points  includes the  price-quantity  observations  where  prices


ranged  from $4  to   $12  in $2 increments.   Bode's  Rule  is   again  applied  to


estimate the consumer surplus  corresponding to this  segment of the  demand curve.


The  process continues  until   the  last  application  of the algorithm increases


consumer surplus by  less than  one  percent of the  total.




                                   FIGURE 7



                ESTIMATING CONSUMERS'SURPLUS USING BODE'S RULE
      O

      Q
      UJ
      O
      LU

      O
      <


      O
                                    6        8


                                      PRICE    $
10
12
66

-------
2.    Travel-Cost Demand and Valuation Estimates:   Some Illustrations



     This study departs from the recreation literature not only in the number of



sites considered,  but in  the  number of separate  recreation  activities.   Also,



the gravity model permits each site in the region to be considered as a possible



substitute  for  every other  site  in  the region.   This section  first presents a



brief  discussion on  aggregating  recreation  activities  and then  discusses  the



issues of  substitute  sites.   The  model is illustrated by presenting some demand



and value estimates of swimming.








A.    Aggregating Recreation Activities



     In  most  recreation  analyses,  recreation is construed  as  a single homogen-



eous good.  Such an assumption may be appropriate when estimating the demand for



a  national  park, but  it is inappropriate when analyzing the  demand for water-



based  recreation.   In  this  study,  recreation  is  disaggregated  into  four



activities:   swimming,  fishing, boating,  and camping.   The  first three activ-



ities are  water dependent and camping is water related.   In the Northwest, most



camping  occurs  near  water,  and  camping  is therefore  a  potential  benefit of



improving water  quality  or of constructing water recreation areas.



     The  optimal  degree  of disaggregation  is  a matter  of  judgment  because



increased reliability and  realism must be weighed against costs, complexity, and



lack  of  data.   It  is important  to  consider the  above  activities  individually



because  they  often occur  separately;  they  have  different trip-length frequency



distributions  and  they  respond   to  different  water-quality  parameters.   For



instance, an  increase in water temperature may be  lethal  to cold-water fish, but



may enhance water quality  for swimming.



     None  of   the  four  activities is  homogeneous,  implying  that even further



disaggregation  could be  useful.  For  instance, kayaking  is  a specialized type of






                                                                              67

-------
boating requiring  rapidly flowing water,  and this activity  is  quite different



from  speed  boating  or  sail  boating.    Similarly,   fishermen  may   have  quite



different  preferences  for  salmon,   trout,  and  catfish.   Data  requirements



preclude  disaggregating  activities   beyond  the  four  being  considered,  and



therefore a qualification is required.   If a change in water quality affects one



type of activity  that  is not representative of  the  general  category, the model



will not produce reliable results.   For instance, if an exogenous change affects



kayaking,  where willingness  to pay exceeds  that for  boating  in  general,  the



model will produce valuation estimates  with a downward bias.








B.   Substitute Sites



     According  to  the economic  theory  of consumer behavior,  the  quantity of a



good  demanded  depends on the price  of  the good, the budget  constraint and the



price  of  substitute  goods.   The conventional travel-cost  analysis excludes the



price of substitutes and this omission is one of the more serious limitations of



the  analysis.   At least  three  issues are  associated with  the  availability of



substitute  sites:   (1)   the correct  measure  of the  increment  in  consumers'



surplus  given  that  consumers'  surplus  may be  redistributed  from  a  substitute



site;  (2)  the  statistical  bias  in the travel-cost demand curve;  and (3) esti-



mating  the response of  use  to  a quality or facility change  given the attract-



iveness of alternative sites.  The first issue is discussed in Chapter II, where



it  is argued  on  the  basis  of  conventional theory  that  benefits  are measured



correctly  by not subtracting benefits foregone from substitute sites.



     The  omission  of the price  of substitutes  may  introduce a  statistical bias



into  the  price coefficient estimate.    For those recreationists  located rela-



tively  near  a  site,  there  are  likely  to be few substitute  sites, hence their



demand  schedule  may be  relatively inelastic.     As   we consider  travel zones





68

-------
farther from the  recreation  site,  the number of  substitute  sites increases and



the price elasticity correspondingly diminishes.   The relationship between visit



rates  and  travel  costs  may  therefore  produce  a  biased   estimate  of  price



elasticity.



     This bias is one reason for the recent interest in regional models.   Dwyer,



Kelly, and  Bowes (1977),  in their review of the  recreation  demand literature,



conclude  .that  regional  models  are  an  improvement over  single-site  analyses.



Regional   simultaneous  equation models  have been constructed  by  Burt  and Brewer



(1971) and by Krutilla and Fisher (1975) where six sites were considered in each



study.   Neither   of  these models  is  easily  transferable  to other  regions  or



activities, nor  is  it clear that all relevant substitute sites were considered.



     The  substitute  sites  considered in most demand analyses are those near the



site  being  analyzed,  but  this  consideration  is  insufficient.    For  example,



suppose that people travel up  to 100 miles to camp and that we are interested in



the  benefits of  a new campground.  The market area for the new campground would



be  within a circle with  100-mile  radius with the proposed  site  in the center.



Any  existing  site within  this area is a  potential  substitute  for the proposed



site.  However,  the  area encompassing substitute  sites  that  must be considered



is  significantly larger than  the market  area for  the  proposed site.   Visitors



who  would travel  up to 100  miles  to the new or  improved  site  would travel 100



miles  in  any direction  to an  identical  site.    Consequently,  a  site 200 miles



from  the  proposed could be a  substitute  for  that site because it would attract



visitors  who reside half-way between the two sites.  If visitors would travel up



to  x  miles to   recreate  at  a  new  or improved  site,  the  area  of potential



substitute  sites is  a circle with  radius 2x miles.   The  number  of potential



substitute  sites is  therefore much larger than  is commonly recognized.   Price
                                                                              69

-------
elasticity  estimates   in  this  study,  as  in  previous  studies,  may  be  biased



because the influence  of substitute sites is not adequately considered.



     Recreation analyses are  as  much  concerned with estimating  the quantity of



recreation demanded as with  estimating  a price  elasticity.   The  increment in



quantity demand at an  improved site depends upon  the  attractiveness of substi-



tute sites  and the relative  travel distances  to these sites.   A  virtue  of the



gravity  model  [Eq.  (III.l)]   is  that recreation  trips are  distributed  simul-



taneously to all  sites on the  basis of attractions to each site and the relative



effect of  spatial  impedance.   The gravity model permits each site in the region



to be a substitute for every other site and the model also considers substitutes



in terms of travel distances and attractions.








C.   Some Empirical Estimates  for Swimming



     The data necessary to estimate a travel-cost demand schedule include visits



or  visit  rates  by origin  and  the corresponding travel  cost.   Visit  data by



origin are  estimated  for each site by the  gravity model,  which is discussed in



the  previous  chapter.   The travel-cost  data  include  the travel  distance  from



each population centroid to each recreation centroid and the round trip cost per



person  per  vehicle  mile.    One-way  travel  distances  are   obtained  from  the



impedance  matrix, which  is  explained  in Chapter III.   According to  the  U.S.



Federal Highway Administration, the total cost per mile for an intermediate size



car  in 1981 is 23.8 cents per mile.  However, the variable cost  is 6.6 cents per



mile  for gas  and oil  plus 5.6  cents for  maintenance,  accessories,  parts, and



tires,  for a  total  of 12.2  cents  per mile.   This  estimate  of 12.2  cents was



doubled to adjust for  round-trip costs and then divided by the average number of
70

-------
persons per vehicle  (3.47)  to obtain 7.753  cents.4   The average length of stay

is one  day for  swimming,  boating,  and  fishing,  but  two days  for  camping;  so

7.753 cents  is divided by  two  to obtain mileage costs  for  camping activities.

     A  recreation  experience demand  function is estimated  in  semilog form for

each  of four  activities  and  for each  of  195 centroids  in the  origin (these

centroids  are  defined in  Appendix  A).   Total  quantity demanded,  consumers'

surplus  and  surplus  per  trip  were also  estimated  for  these activities  and

centroids.   A sample  of  the  demand  and valuation  estimates  for  swimming  is

presented  in   Table  4.  The  first  three columns  in this  table  identify  the

recreation  centoid  by number,  county,  and  name.   Linear  designated beach feet

and  recreation accessibility  (RA..)  are  inputs in the attractiveness model, and

A. are estimated attractions from this model.  The gravity model estimates total
 J
quantity demanded with Eq.  (III.3),  and  these estimates are presented in column

7 of Table 4.   As indicated by column 8,  most recreation centroids receive trips

from  over  100 origin zones.  An  inspection of the trip  interchange matrix (T. .)

indicated  that the  large  majority of trips  emanate  from relatively few origin

zones.   Columns  9-12  are  the first-stage demand statistics  and overall show a

high  level  of significance.  Consumers'   surplus  is  estimated with Bode's Rule,

and  surplus  per trip is  simply total  surplus  divided  by  quantity  demanded

(column 7).

     The demand and  valuation estimate presented  in Table 4 reflect one  activity

(swimming)  out of  four being considered, and 20  recreation  sites out of 195 in

the  region.    However, the  demand and valuation  estimates based on this sample

are  representative   of the  other activities  and  of  the entire   region.  The
 4The  number  of  persons  per vehicle  is  estimated as  the sample  mean  of the
household  regional   recreation  survey.   This number  is  larger  than  the mean
household size in each of the four states  in  the region.


                                                                               71

-------
(V)
                                                                           TABLE 4



                                         DEMAND AND VALUATION  ESTIMATES  FOR SWIMMING  IN SELECTED WASHINGTON  CENTROIDS
Recreation
Centroid
Number
(1)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Notes:
form. RA .

County
(2)
Adams
Asotin
Benton
Chelan
Chelan
Clallum
Cl all urn
Clall urn
Clark
Columbia
Cowlitz
Cowlitz
Douglas
Oerrt
Franklin
Garfield
Grant
Grant
Grant

Recreation Centroid
(3)
Northeast Corner
Fields Spring St. Park
Crow Butte State Park
Lake Wenatchee
Lake Chelan State Park
Bogachiel State Park
Neah Bay State Park
Dungeness State Park
Battleground State Park
Lewis and Clark St. Park
Merwin Reservoir
Seaquest State Park
Chief Joseph
Twin Lakes
Lyons Ferry State Park
Pataha Creek
Potholes State Park
Sun Lakes State Park
Steamboat State Park
Gray's Harbor Bay City
Quantity
measures
demanded was estimated from
Linear
Beach
Feet
(4)
1
2000
1850
200
870
1200
1100
1100
1085
1
1
1
100
400
1000
1
1000
2930
1000
1

RA .
(5)SJ
413
116
211
641
145
129
79
722
912
219
522
881
111
201
220
167
238
210
219
576

A,
(6)J
119,815
199,415
277,589
366,156
198,291
194,775
145,165
517,662
590,879
83,203
137,210
185,382
119,167
210,193
257,529
71,319
269,245
298,494
256,888
145,131
the gravity model. NOZ
the swimming accessibility of a recreation
centroid
Quantity
Demanded
(7)
109,926
82,308
215,045
391,439
111,302
49,393
24,431
534,887
806,747
80,600
109,345
229,694
65,854
99,762
220,907
44,194
224,317
194,961
153,903
118,689
is the number
divided by 1,
Experience Demand
Curve Statistics C

NOZ In a
(8) (9)
126 7.67
134 8.34
135 8.85
125 9.06
116 8.55
107 8. 18
94 8. 14
124 9.34
135 9.29
136 7.29
121 7.83
127 8.12
108 8.09
122 8.64
139 8.64
131 7.34
133 8.89
130 8.86
127 8.73
120 7.90

B
(10)
-0.225
-0.225
-0.243
-0.239
-0.242
-0.230
-0.238
-0.234
-0.228
-0.220
-0.231
-0.229
-0.242
-0.244
-0.230
-0.229
-0.244
-0.237
-0.238
-0.233
origin zones. The
000. These

t R2
(11) (12)
-32.3 0.877
-31.5 0.872
-31.6 0.873
-39.2 0.914
-37.6 0.907
-31.8 0.876
-35.3 0.899
-35.7 0.898
-38.8 0.912
-29.0 0.852
-36.2 0.900
-36.7 0.903
-36.3 0.901
-33.3 0.885
-31.7 0.873
-31.8 0.874
-33.7 0.887
-35.2 0.895
-34.3 0.890
-32.4 0.878
demand curve is
lonsumers'
Surplus
$
(13)
436,053
324,576
674,364
1,661,722
645,566
260,539
131,485
2,515,514
5,435,151
196,873
797,108
1,345,905
454,320
595,718
631,644
151,382
687,471
777,465
606,162
613,219
specified
estimates were obtained from Eq.
Surplus
per Trip
$
(14)
3.96
3.94
3.13
4.24
5.80
5.27
5.38
4.70
6.73
2.44
7.28
5.85
6.90
5.97
2.85
3.42
3.06
3.98
3.93
5.16
in semi log
(III. 13).

-------
consumers'  surplus estimate of $4.31 per swimming day is  comparable in magnitude



to the other activities and of the other recreation centroids.
                                                                              73

-------
                                    CHAPTER V

                   SURVEY ESTIMATES OF THE WILLINGNESS TO PAY
                    TO RECREATE AND THE VALUE OF TRAVEL TIME
1.    Introduction

     The willingness to pay  to  recreate can be estimated  directly using sample

surveys or indirectly by  estimating  a recreation  demand curve and measuring the

area  under  this  curve.   The  indirect  demand  curve approach  is used  in  this

study.  Taking  a  regional  household  recreation  survey in  the  summer  of  1980

afforded  the  opportunity  to include  a  question  on willingness to  pay.   The

objective of obtaining  direct  estimates of consumer surplus  is to compare them

to the direct estimates  when all other factors are equal.

     Recreation travel  patterns are influenced by  travel cost, which is measured

at least partially by vehicle operating expenses.   An opportunity cost of travel

is foregone time,  and  recreationists may consider travel  time  as an additional

travel  cost,  or as  a  benefit.    Empirical  evidence  from  the current household

survey on the value of recreation travel time is also presented in this chapter.

The  objective  in presenting  these results is to provide  empirical  evidence on

the travel time bias in recreation studies.



2.   Direct Willingness-to-Pay Estimates

     Obtaining credible estimates of willingness to pay using sample survey  is  a

challenging endeavor.   Even  at  best, the estimates  may not inspire much confi-

dence.   Obtaining  these   estimates  sometimes  involves  lengthy  and  expensive
74

-------
personal interviews using a "bidding game" approach.  The approach taken here is

much  less  ambitious.   The resources  available  permitted  one  question  to  be

included on the questionnaire.

     Most  recreation benefit  studies  use  either  the  indirect  or  the  direct

approach,  but  not both.   The  observed  differences  between results  from  these

approaches owe partially  to  the different approaches and to  other differences,

such as  sites  analyzed,  activities included, and  date  of  study.   The objective

here  is  to compare direct willingness-to-pay estimates with  indirect estimates

of the same activities, region, and time period.

     Although  the complete  household  survey  is  included  in  Appendix B,  the

direct willingness-to-pay question is provided here.  The question is


            What  is  the  maximum daily use fee  you would  be willing
            to pay for this recreation  facility  rather than  forego
            using  it?


We  explicitly  asked for  the  daily fee to obtain  a consumers'  surplus estimate

per visitor-day.

     The frequency distribution of responses is presented in Table 5.  Virtually

all the  respondents  indicated  a willingness to pay between $2 and $10 per trip

to recreate.   The  mean value per trip is $5.62, and the mode and median are each

$5.  The  indirect estimates  of willingness to pay obtained in this study are in

the $3 to $6 range with an average of about $4.20.



3.   The Value of  Recreation Travel Time

     The essence  of  the travel-cost approach to  estimating a  recreation demand

curve  is  that the cost of traveling  is an  empirical  proxy  for  price and the

relationship  between travel costs and  visit  rates is used  to impute  a  site-

demand curve.  If travel  costs are estimated as vehicle operating costs and the


                                                                              75

-------
                                     TABLE  5

             DIRECT WILLINGNESS-TO-PAY  ESTIMATES  PER RECREATION DAY

Monetary1
Value
($)
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15+
Frequency
1
7
27
58
38
0
33
26
21
8
39
3
3
2
0
8
Relative
Frequency
0.0027
0.0192
0.0741
0.1594
0.1044
0.2472
0.0906
0.0714
0.0577
0.0219
0.1071
0.0081
0.0081
0.0054
0.0000
0.0220
                   Total               364               1.00

Other statistics  are:   Mean =  $5.619,  Median  =  $5.00, Mode =  $5.00,  Standard
Deviation = $2.93


 1The monetary values are less  than or equal  to these numbers.   For example, all
estimates above $1 and less than or equal  to  $2 are recorded as  $2.


correct  measure  of travel  costs  includes some value of travel  time,  then the

estimated  demand curve  is a biased  representation  of  the true  demand  curve.

     The  above point  is  well  recognized in  the  recreation  literature.   The

concensus  in  this literature is that travel  time is a positive  cost of  travel

and  therefore must  be  included  in empirical  estimates  of travel-cost  demand

curves.   In one of the more widely quoted studies, Cesario (1976) concludes that

the  recreation value of  travel  time  is approximately  1/3  the  average national

wage rate.    This  study   has received  the official   endorsement of  the Water
76

-------
Resources  Council  in  that the  Council  recommends  use  of  this  value  in the

travel-cost analyses.

     In  the  infinitely  flexible and continuously  adjustable  world of neoclass-

ical economics,  the  cost  of  travel time  may be  defined in  terms  of foregone

earnings.  At  the conceptual  level,  the  cost  of traveling  is  its opportunity

cost, which  is  what  one gives up  in order to travel.   The neoclassical view is

questionable  on empirical  grounds, because  generally people  do not  have the

flexibility  to  trade work  time for travel  time.   However,  a more fundamental

objection  to  using  the  wage rate  as  an opportunity  cost  can  be  raised  on

conceptual grounds.  The  real cost of  foregone  work time is  not wages, but the

utility  of  income  minus the  disutility  of work.   Gross wages are not likely to

be a good proxy to the net benefits of  employment.

     In  addition to the conceptual objections to valuing travel time in terms of

foregone  earnings,  there  is room  for  skepticism about the  reliability  of

Cesario's  empirical  estimate.   Cesario's estimate was derived from a  literature

review  of several studies  of how  commuters  value their  journey-to-work travel

time.   Recreation is  a  leisure  time,  discretionary  activity,  which  is  quite

different  from the daily required  journey-to-work trip.   The recreationist has

the  option of choosing a destination  so as to  have a  positive  value of travel

time.   The  commuter  is  generally rigidly  constrained to arrive at a destination

not of  his own choosing and to do so during peak traffic hours.

     The  above  reservations  about the  accepted  view on the travel time bias led

to  the   inclusion  of  two questions  on the household  recreation survey.    These

questions are:


            Ql.  Some  people  feel  that  time spent traveling  to a
                 recreation  site  is  an  inconvenience  while others
                 enjoy  it.  How about you?
                                                                               77

-------
                      1.    Enjoyed travel  time
                      2.    Prefer to shorten travel time
                      3.    Refused
                      4.    Don't know,  no  answer

            Q2.   About  how much  would you  be  willing  to pay  to
                 shorten the total travel  time for this last trip by
                 one half?


The total  sample size is  2,249,  of which 107 respondents  refused  to  answer or

did not  know.   Of  the  remaining respondents, 1,865 enjoyed their  travel  time,

whereas  only 276  would  prefer  to  shorten  their travel  time.  These  results

suggest that travel  time  is a net benefit, not a cost.  Vehicle operating costs

probably overstate recreation travel costs, not understate them.

     The objective  in  question  one is to  determine whether travel time is a net

cost  or  a  benefit.   The  objective  of   the  second  question is  to obtain  a

quantitative estimate  of  what is presumed to  be  a cost.   The frequency distri-

bution of survey results is presented in Table 6.


                                    TABLE  6

             DIRECT ESTIMATES OF THE VALUE OF RECREATION TRAVEL TIME

Boundaries
0.0 -
2.0 -
4.0 -
6.0 -
8.0 -
10.0 -
12.0 -
14.0 -
16.0 -
18.0 -
2.0
4.0
6.0
8.0
10.0
12.0
14.0
16.0
18.0
20.0
Frequency
1178
23
46
4
2
33
0
12
0
29
Relative
Frequency
0.888
0.017
0.035
0.003
0.002
0.025
0.000
0.009
0.000
0.022
                   Total               1327               1.00

Other  statistics  are:    Mean  =  $1.069,  Median  =  0.0,  Mode  =  0.0,  Standard
Deviation = 3.65
78

-------
     The most impressive result in Table 6 is that 88 percent of the respondents



are not  willing  to pay  anything to  shorten their  travel  time by  50 percent.



Fewer than  3% of  the  total   respondents  are willing  to pay  more than  $5  to



shorten  their travel  time by 50 percent.  The results in Table 6 cast doubt that



recreationists,  at  least in the Northwest, perceive their travel time as a cost.



The  question that  now  arises  is  whether  recreationists   deliberately  incur



vehicle  operating costs  in order to spend more  time traveling.   Unfortunately,



the survey evidence is insufficient to answer this question.   The main result is



that the  cost of  recreation travel time, at least in the Pacific Northwest, and



for the four activities considered, is not positive.   On this basis, travel cost



will be  measured  as  vehicle  operating costs.  Of course, these results should



not be applied to value travel time by commuters in  large urban areas.
                                                                               79

-------
                                   CHAPTER VI

        THE SENSITIVITY OF TRAVEL-COST ESTIMATES OF RECREATION DEMAND AND
           VALUATION TO VARIOUS COMPUTATIONAL AND SPECIFICATION ISSUES
     The travel-cost  demand curves developed  and estimated  in Chapter  IV  are

based  on  a semilog  form of the  first-stage demand  curve,  origins  defined  as

recreation centroids, and total  quantity demanded estimated from a gravity model

with  endogenous  attractions.    This   chapter  uses   a  Monte  Carlo  simulation

analysis to test the robustness  and correctness of some of the input assumptions

in the model.   Specifically, the focus of this chapter is on three specification

and computational choices  required by the TCM which  may  influence estimates of

the demand curve and consumers'  surplus.  The three issues investigated here are

(1) the  functional  form of the  first-stage  demand curve; (2)  the  width  of the

concentric  zones;  and  (3)  the  estimate  of  total  quantity  demanded.   The

objective  is  to determine  the  sensitivity of travel-cost  demand  and valuation

estimates  to  various assumptions  concerning  these four  points.   The method of

analysis is to  apply the TCM to  several  sites  under various assumptions and to

contrast the results.

     Applying  the  TCM   and estimating  consumers'  surplus  requires  that  some

assumption  be  made  on  each of  these  points.   Choices are  often made inadvert-

ently;  at  least there  is  little  analysis of the  sensitivity  of the results to

variations  in  the computational  procedure.   The  first  section of this chapter

contains a brief discussion of the possible significance  of  the  three points.

Section 2  contains  the  empirical  estimates of travel-cost  demand  and valuation
80

-------
estimates under these various assumptions.   The conclusions and implications are

discussed in Section 3.



1.   The Three Issues

     Many empirical  demand  curves  in the economics  literature  are specified in

double-log  form,  perhaps because  the coefficients may  be  interpreted as elas-

ticities.1   In the  recreation  literature,  the  semi log specification  is  most

prevalent,  although  linear  functions have  been used.2  An issue considered here

is  the  relative   merit  of  the  semi log  and double-log  specification  of  the

first-stage demand  curve and the sensitivity of  the  valuation  estimates to the

choice of these two  functional forms.

     In  the TCM,  visit  rates from various  origins  are regressed against corre-

sponding travel costs.  Since the pioneering work of Clawson and Knetsch  (1966),

origins  have  been defined by a  series of concentric rings around the  recreation

site.   For  instance, if  recreationists travel a  maximum of 200 miles and rings

are  defined every 20 miles, then  there  are 10 origin zones and 10 observations

for  the experience-demand  schedule.   Similarly,  if a ring  is  defined every 10

miles,  there will  be 20 travel  zones and  20  observations for  estimating the

visit-rate  schedule.  Alternatively, each  population  centroid  may be construed

as  a separate origin, and  the number of observations  is therefore determined by

the  number  of such  centroids.   A  second  issue  is the sensitivity of  the demand

and  valuation  estimates  to  the definition of  the  origin  zone.
  xln  their  literature  surveys  on  the demand  for  money,  Laidler  (1977)  and
 Goldfeld (1973) present  empirical estimates  in favor  of  a  log-log  specification.

  2Linear  demand  curves  have  been  used   by Burt   and  Brewer  (1971)  and  by
 Cicchetti,  Fisher,  and  Smith  (1976) because this  specification is required  by
 some properties of their models.


                                                                               81

-------
     There  are  at  least  two ways  to  estimate  total  quantity  of  recreation



demanded  at  a  zero  price.   An  estimate  can  be  generated   endogenously  by



substituting a zero price increment in the experience demand curve.  Cesario and



Knetsch  (1976,  p.   100)  apply  this  method  and  it  is   generally  used  when



site-attendance  data  are unavailable.   In  most  travel-cost  analyses,  quantity



demanded  is estimated  exogeneously  by  site-attendance  data.    Clawson  (1959)



estimated quantity demanded  in  this manner and Knetsch (1974, p. 83) and others



have  followed  his  lead.   Estimates  of consumers'  surplus, and  particularly



consumers'  surplus  per  trip,  may  be  sensitive  to  the  choice  between  these



quantity-demanded estimates.



     If  the demand  curve is  constrained  to  intersect  the observed  quantity



demanded,  then  the   magnitude  of  the  hypothetical   price  increments  in  the



first-stage  demand  curve  may affect  consumers'   surplus.   Figure 8  depicts  a



hypothetical demand  curve generated  from  price-quantity  observations  using $1



price  increments in  the  first-stage demand curve.   In Panel A,  quantity  a is



estimated from the first-stage demand curve by letting AP = 0, and quantity c is



assumed  to  be the  correct estimate.  Consumers'  surplus  estimated  as  the  area



under  cbd will  be less  than the surplus estimated as the area under abd.   Panel



B depicts an  estimate of consumers' surplus when price increments of $0.25 are



used from 0 to $1,  and $1 price  increments are  used thereafter.  If the demand



curve  is  constrained to  include  the  correct  quantity   demanded,  consumers'



surplus  in  Panel B  (Oced) exceeds  that  in  Panel  A (Ocbd) by an amount equal to



ceb.   If  quantity  demanded  is  estimated  incorrectly,   the  magnitude of  the



hypothetical price increment could affect the results.



     The  discussion  to   this  point  offers  some  a  priori  possibilities  that



specification and computational  choices in estimating a travel-cost demand curve



may affect the results.   Empirical evidence on the  sensitivity of the results to





82

-------
                                    FIGURE 8
                   PRICE-QUANTITY OBSERVATIONS FOR A RECREATION SITE DEMAND CURVE
                             PANEL A
                           DEMAND CURVE
                         SI PRICE INCREMENT
                  PANELS
                DEMAND CURVE
              $0.25 PRICE INCREMENT
           O
           Q
           UJ
           Q
           D
           >
           <
           O
O
D
UJ
O
<
UJ
O
<
o
                        PRICE S
                                                      PRICE S
these choices is presented in Section  3  by  estimating several  travel-cost demand

curves under alternative conditions  using the model  described below.


2.   Sensitivity of Travel-Cost  Estimates to Various Assumptions

     Because  the  195  recreation centroids  and four  activities  included in the

regional model  are  more than sufficient for this analysis, we quite arbitrarily

consider the  demand for boating at  20 Washington recreation centroids,  numbered

17.0  to 26.0  (column  1  in  the  accompanying tables).   These centroids  include

those  in  King County,  which  contains Seattle  and  is heavily populated,  as  well

as sparsely populated  counties  east  of the  Cascade Mountains.  By  including  both

urban  and  rural  counties  in the  sample,   the  travel-cost  estimates  reflect  a

diversity  of  realistic  conditions.   The  rationale  for  sampling a  relatively

large  number  of  centroids  (20)  is  that   certain  adverse  consequences may  be

observed  only occasionally, and a large sample  increases  the  likelihood of  such

a result.   Also,  results  based on a  single  site may  reflect  a  special case, and

be inconsistent with results obtained over  a wide range  of experience.
                                                                                83

-------
     The  sensitivity  of  travel-cost  estimates  to  each  of the  computational



issues  being  considered  depends  upon the  assumption made  on  the  other three



issues.    The  interdependence  of  these  issues precludes  analyzing  them indi-



vidually.   We  consider  first the  functional   form  of  the first-stage  demand



curve,  focusing  on  the semi log form  and the double-log form.   Results  will be



presented  by  generating  quantity  estimates endogenously  and by  assuming that



quantity  demanded  is  exogenous.   Travel-cost  estimates will then be presented



using various size origin zones.   We will show that the results  are sensitive to



the definition of origin zone and  this sensitivity in turn depends on the choice



of quantity demanded and on the functional  form.








A.   Functional  Form of the First-Stage Demand Curve



     The  issue of proper  form of  a recreation  demand  curve has been studied by



Zeimer  et  al.  (1980)  and by Smith (1975).   The studies are similar in that only



one site was considered and a statistical analysis, namely a Box-Cox transforma-



tion, was  used to  statistically  estimate the most  appropriate  functional form.



Smith rejected the linear form because it provided a poorer fit of the data than



the double-log and semilog form.   However,  Smith also concluded that even though



the  latter two forms  fit the data  and  provided  reasonable  results,  each form



must be considered inappropriate.   Zeimer et a_[. used the Box-Cox transformation



procedure  and concluded  that  a semilog form is appropriate and a  linear  form is



inappropriate,  and   further   that   consumers'   surplus  estimates  are  highly



sensitive to the choice of functional form.



     In  considering  the  various  functional  forms,  double-log  and   semilog



(logarithm of the  dependent variable) are candidates,  but the  linear form need



not be  considered.  Ziemer et al. and Smith provide evidence against the linear



form, and  scatter plots   of several   sites  indicate  a distinct  curvilinear





84

-------
relationship.   The  evidence against  the appropriateness of  the  linear form is



persuasive, and in this study, we consider the double-log and semilog functional



form.



     The objective  of analyzing  these two  forms  is first to  determine if the



results are sensitive  to  the choice  of  functional  form and if so, to determine



that of the two  forms seems most appropriate.  Four criteria are suggested that



may be useful  in identifying the most appropriate form.   First, the coefficients



of  determination  are  a  relevant  but  not  decisive indicator,  particularly  if



estimated  over several  sites.   Second, estimates of consumers' surplus per trip



should be  somewhat  stable across sites  and  should  be  similar to those reported



elsewhere  in  the  literature.   Third,  the  first-stage  demand  curve  should



estimate closely the known quantity demanded  at a zero price when AP = 0 is used



in  Eq. (IV.5).  Finally,  goodness of  fit and  consumers'  surplus estimates should



be  insensitive  to other  computational  decisions, particularly  if  the decisions



are made arbitrarily.  These properties are not espoused as rigorous statistical



criteria  that will  necessarily  determine  the  unambiguous  superiority  of one



functional  form.   Because  previous  studies  have not been  able to resolve this



issue on statistical or theoretical grounds,  it is appropriate  to employ a Monte



Carlo  analysis,  where  a  demand curve for  several  sites  is estimated with each



functional form and the results  are compared.



     First-stage  demand  curves  for  boating  [Eq.  (IV.4)] are  estimated for 20



centroids  using  both double-log and  semilog  forms,  where the  logarithm  is taken



of  the dependent  variable.   These estimates  are based on  population centroids as



origin  zones,  a  $1  price  increment  in  Eq.  (IV.5),  and   quantity   demanded



estimated  exogeneously.   The  results  are presented  in Table 7.  The coefficients



of  determination,  columns  2  and  5,  indicate that  each form  fits  the data



reasonably well,  but the semilog model  has more explanatory power in  19  of  the






                                                                              85

-------
                                     TABLE 7
ANNUAL VALUATION ESTIMATES  FOR  BOATING
SEMI LOG AND DOUBLE-LOG FUNCTIONAL FORM
            IN SELECTED WASHINGTON CENTROIDS USING A
                     Semilog Results*
                            Double-Log Results
Recreation
Centroid
Number
(1)
17.0
17.1
17.2
18.0
19.0
19.1
20.0
21.0
22.0
22.1
22.2
23.0
23.1
23.2
24.0
24.1
24.2
24.3
25.0
26.0
R2
(2)
0.851
0.863
0.873
0.885
0.740
0.829
0.727
0.888
0.766
0.751
0.673
0.884
0.884
0.874
0.841
0.810
0.805
0.812
0.874
0.785
Consumers '
Surplus
(in $1000)
(3)
2,163
5,239
5,596
283
141
536
548
1,008
15
27
34
228
231
1,253
19
31
39
67
205
52
Surplus
per Day
$
(4)
6.22
5.48
5.12
4.50
4.28
6.15
5.20
5.83
2.57
2.56
2.32
5.44
5.08
5.78
2.67
2.55
2.72
2.87
4.89
2.42
R2
(5)
0.645
0.605
0.666
0.724
0.660
0.686
0.582
0.702
0.744
0.729
0.699
0.695
0.691
0.636
0.781
0.740
0.747
0.747
0.732
0.724
Consumers'
Surplus
(in $1000)
(6)
2,757
5,577
11,564
334
142
620
526
886
18
58
69
220
331
1,975
35
136
60
1,606
168
109
Surplus
per Day
$
(7)
7.94
5.83
10.58
5.31
4.78
7.11
5.00
5.13
3.18
5.57
4.79
5.24
7.26
9.12
4.97
11.05
4.11
68.95
4.00
5.17
column mean   0.820
886
4.24
0.696
1,360
9.23
 * In the semilog form the logarithm is taken of the dependent variable.


20  cases.    The  semilog  surplus-per-day  estimates  are more  stable  than  the

corresponding  double-log estimates.   Dwyer,  Kelley,  and  Bowes  (1977)  review

several  empirical  studies  of  recreation behavior,  but  only a  few  of  these

studies  deal  specifically with  boating.   If we presume that other water-based

activities  have a  value  comparable to  boating or  that boating  is  typical  of

outdoor  recreation  in general, we  may  conjecture  on the basis  of Dwyer et aJL

that value-per-day estimates below $1 or above $10 are outside the range of many
86

-------
existing  studies.   The  double-log estimate  of surplus  per  day  of  $68.95 for



centroid 24.3 is clearly untenable, and the double-log surplus-per-day estimates



of $10.85 and $11.05 appear suspiciously high.



     A  few  of the  surplus-per-day estimates, such as those  for centroids 18.0



and 20.0, are  insensitive  to the  choice of  functional  form,  but some estimates



are highly  sensitive to this  choice.   This  result indicates  the inadequacy of



analyzing  the  issue of  functional form  by  considering  only  one  site.   The



results  in  Table 7 do  not  establish  that  either form is  correct or incorrect,



but the consistently lower explanatory power of the double-log form and the wide



variation in surplus-per-day estimates cast some doubt about the appropriateness



of this form.



     The  sensitivity of the above  results to the choice of quantity demanded is



observed by reestimating the above  equations where quantity demanded is obtained



from the visit-rate schedule using  a zero price increment.  Table 8 compares the



results  of  demand  and valuation estimates  obtained  with  a   semilog   and  a



double-log form where quantity demanded is estimated endogeneously.  The assumed



known  quantity  demanded is  in column 2 and  the semilog and double-log quantity



estimates are  in  columns 3 and  6  respectively.   Several  of the double-log form



estimates of  total  quantity demanded contain very  large  errors.   For instance,



the double-log  form produces  an  estimate of 111 million  boating days  at Lake



Washington  (centroid 17.2),  which errs by approximately  110  million days.  The



quantity  estimates  from a  semilog form are  much  closer  aproximations to total



use, but the discrepancies are notable.



     Comparing  the  consumers'   surplus  and  surplus-per-day  estimates  of the



semilog and double-log  form (Table  6) indicates dramatic differences in results.



Total   surplus estimates with a double-log form average about four times those of



a  semilog form,  but the surplus-per-day estimates  are  considerably smaller for






                                                                              87

-------
oo
oo
                                                         TABLE 8



           DEMAND AND VALUATION ESTIMATES USING A SEMI LOG AND DOUBLE-LOG FORM AND ENDOGENOUS QUANTITY DEMANDED


Recreation
Centroid
Number
(1)
17.0
17.1
17.2
18.0
19.0
19.1
20.0
21.0
22.0
22.1
22.2
23.0
23.1
23.2
24.0
24.1
24.2
24.3
25.0
26.0
Mean
Exogeneous
Quantity
Demanded
(in 1000)
(2)
347
956
1,092
63
33
87
105
113
6
10
15
42
45
217
7
12
14
23
42
21
166
Semi log Results

Quantity
Demanded
(in 1000)
(3)
534
1,270
1,354
70
36
134
135
244
4
8
9
57
59
303
5
9
10
19
49
17
216

Consumers'
Surplus
(in $1000)
(4)
2,221
4,337
5,677
285
142
551
557
1,030
14
26
32
233
235
1,280
18
30
38
66
207
50
902

Surplus
per Day
$
(5)
4.16
4.20
4.19
4.10
3.96
4.11
4.12
4.22
3.52
3.33
3.73
4.06
4.02
4.22
3.60
3.49
3.67
3.40
4.25
2.96
3.87
Double-Log Results

Quantity
Demanded
(in 1000)
(6)
3,751
17,657
111,807
572
122
475
882
800
15
130
102
389
1,279
10,549
52
1,223
84
7,081
91
1,015
8,234

Consumers'
Surplus
(in $1000)
(7)
3,817
10,774
45,793
492
169
741
767
1,081
21
95
97
328
715
7,003
50
512
81
3,988
181
418
3,856

Surplus
per Day
$
(8)
1.02
0.61
0.41
0.86
1.38
1.56
0.87
1.35
1.37
0.74
0.95
0.84
0.56
0.42
0.94
0.42
0.97
0.52
1.99
0.41
0.91

-------
the double-log  form.   Without a benchmark for  comparison,  we cannot be certain



which estimates  are most  accurate.   Because the  double-log form  yields  gross



errors in the  quantity estimates,  it is possible that similar errors character-



ize the surplus estimates.   Comparing total surplus semilog estimates in Table 7



with  those  in Table 8 indicates  a very close  correspondence.   The result that



total  surplus  estimates  are  insensitive  to  the  choice  of quantity  demanded



(given  a semilog  form) is  significant.   In  contrast,  the total  surplus  and



surplus-per-day  estimates  using a  double-log  form are  highly  sensitive  to  the



choice of quantity estimate.




     The  sensitivity  of the  valuation results to the  size  of  the hypothetical



price  increment  is  analyzed by using a price increment of $0.25 from zero to $1



in the first-stage demand curve and a $1 price  increment thereafter.  The choice



of  price increment does  not affect the (experience)  demand  statistics,  but it



may affect the area under the site demand curve.  Table 9 depicts double-log and



semilog  estimates  of  total  consumers'  surplus and  surplus  per visitor-day  for



each  of  the 20 centroids considered, using a $0.25 price increment  up to $1.   We



again  observe  significant  discrepancies  between  the  double-log  and  semilog



results.  The  double-log surplus-per-day estimate of  $133.61 for centroid 24.3



is  beyond  any  tenable limit,  and  several  other   double-log  results  appear



unreasonably  high.   In contrast, the  surplus-per-day  estimates using a semilog



form  are between $2 and $6, which is in the area of other studies.



      Tables 7,  8,  and 9 present a  comparison of semilog and double-log results



under  alternative  computational  assumptions.   A  comparison   of  the  average



results  across  the  three tables provides  one  measure of the appropriateness of



these  two forms.   Using a semilog form, consumers' surplus averaged $886, $902,



and $897 thousand per  site and $4.24, $3.87, and $4.25 per visitor  day in Tables



7, 8, and 9 respectively.  Using a double-log form, consumers'  surplus estimates






                                                                              89

-------
                                     TABLE 9

TRAVEL-COST VALUATION ESTIMATES USING  A SEMILOG AND DOUBLE-LOG FORM AND A $0.25
PRICE INCREMENT

Semi log Results
Recreation
Centroid
Number
(1)
17.0
17.1
17.2
18.0
19.0
19.1
20.0
21.0
22.0
22.1
22.2
23.0
23.1
23.2
24.0
24.1
24.2
24.3
25.0
26.0
Mean
Consumers'
Surplus
(in $1000)
(2)
2,205
5,307
5,651
285
142
547
554
1,024
14
26
32
232
234
1,272
19
31
38
66
206
50
897
Surplus
per Day
$
(3)
6.34
5.55
5.17
4.52
4.30
6.27
5.27
5.93
2.50
2.51
2.23
5.53
5.14
5.88
2.60
2.49
2.66
2.83
4.93
2.38
4.25
Double-Log Results
Consumers'
Surplus
(in $1000)
(4)
3,503
8,284
24,974
437
159
710
627
1,023
20
82
89
284
536
3,919
45
303
74
3,111
177
245
2,430
Surplus
per Day
$
(5)
10.08
8.66
22.85
6.95
4.81
8.14
5.96
5.92
3.56
7.83
6.12
6.76
11.76
18.09
6.34
24.66
5.12
133.61
4.23
11.57
15.66

Note:
demanded.
These results are

based on an exogenous

estimate of

total quantity

per  site  are $1,360, $902, and  $2,430  thousand per site and  $9.23,  $0.91,  and

$15.66 per  visitor-day.   Double-log  results are highly  sensitive  to  the choice

of  hypothetical  price  increment  in  Eq.  (IV.5)  and to  the choice  of quantity

demanded at  a  zero  price.   Double-log results also show wide differences across

sites,  even when  computational   assumptions  are  identical.    In  contrast,  the

semilog results  are relatively  stable  across sites and much  less sensitive to

the choice   of price  increment  in  Eq. (IV.5)  and to  the choice  of quantity


90

-------
demanded.    These  results  do  not  support  the use  of  the  double-log  form and

suggest that a semi log form is to be preferred.



B.   Size of Origin Zone

     When the travel-cost  method was presented by Clawson (1959) and by Clawson

and  Knetsch  (1966),  origins  were  aggregated into zones defined  by  a series of

concentric circles.   There  does  not appear to have been any serious analysis of

the  appropriate  size  of  these  origin  zones, nor  of  the  sensitivity  of the

results to  various  size zones.8  The above results use each population centroid

in  the region  as  a  potential  origin  zone.   Evidence on  the  sensitivity of

travel-cost  demand  and valuation   to   the  definition  of  the  origin zone is

obtained  by comparing  the  above  results to  those  obtained  using  10-mile and

20-mile  origin  zones.   Consider   two   systems  of concentric  circles,  one at

10-mile and one at 20-mile intervals from the  recreation centroid.  Origin zones

are  now defined as  the area  between  each ring and visit  rates  are defined as

total  trips  from each  zone  per  1,000  population of the zone.   The  travel  cost

from  each  zone  is  the weighted  average  travel cost of all  centroids within the

zone where the weights are the number of trips per centroid.

     Travel-cost  demand and valuation  estimates  using 10- and  20-mile   origin

zones  are  presented in  Table 10.   Comparing the  results  using  a 10-mile  zone

with  those  of a 20-mile origin shows similar estimates  for several sites, but

quite  dissimilar estimates for others.    The estimates in Table 10 are comparable

to the  semi log results in Table  7 because they are based on  a  semi log specifica-

tion,  a $1 price   increment  in  Eq.  (IV.5),  and quantity   demanded  estimated
 8Brown  and  Nawas  (1973)  have  argued that  observations  should  be  based  on
 individuals,  rather  than aggregations  of people.   As they use  site-attendance
 data,  visit  rates  reflect  the  frequency  of  participation  of  and  not  the
 aggregate participation rate of  the population.


                                                                               91

-------
                                     TABLE 10

      SEMILOG VALUATION ESTIMATES USING 10-MILE AND 20-MILE ORIGIN ZONES*
                   10 Mile Origin Zones
                                        20 Mile Origin Zones
Recreation
Centroid
Number
(1)
17.0
17.1
17.2
18.0
19.0
19.1
20.0
21.0
22.0
22.1
22.2
23.0
23.1
23.2
24.0
24.1
24.2
24.2
25.0
26.0
R2
(2)
0.675
0.764
0.768
0.935
0.631
0.547
0.619
0.824
0.258
0.620
0.170
0.882
0.895
0.815
0.916
0.903
0.835
0.830
0.840
0.699
Consumers'
Surplus'
(in $1000)
(3)
$1,394
3,463
3,775
111
82
444
394
767
117
60
361
118
120
890
9
53
29
139
98
102
Surplus
Per Trip
(4)
$4.01
3.62
3.45
1.77
2.47
5.10
3.75
4.44
20.52
5.70
24.87
2.82
2.62
4.11
1.36
4.31
2.02
5.97
2.35
4.80
R2
(5)
0.787
0.749
0.868
0.977
0.668
0.470
0.751
0.827
0.216
0.716
0.474
0.948
0.928
0.913
0.919
0.903
0.877
0.806
0.915
0.698
Consumers'
Surplus
(in $1000)
(6)
$1,415
3,606
2,830
91
81
587
337
882
217
44
313
114
109
868
7
44
23
106
75
80
Surplus
Per Trip
(7)
$4.07
3.77
2.59
1.44
2.45
6.73
3.21
5.11
38.12
4.21
21.58
2.73
2.39
4.01
0.97
3.56
1.61
4.56
1.79
3.76
   Mean
0.720
$626
$5.50
0.771
$591
$5.93
 *These estimates are  based  on a $1 price increment in Eq.
demanded estimated exogeneously.
                                              (IV.5), and quantity
exogenously.    Comparing  the   results   on   these  two  tables  indicates  that

aggregating  population  centroids  into  concentric  zones  increases  consumers'

surplus  by an  average of  over $1 per  trip.   Furthermore,  consumers'  surplus

estimates  on Table  7  appear uncorrelated with those  on  Table 10.   Estimates of

total  surplus   for  centroids  17.1  and  17.2  are  over  $1  million  lower  when

population  centroids  are  aggregated  into  zones.    However,  the  aggregation

process increases the  surplus  estimates  per trip for centroids 22.0 and 22.2 by
92

-------
over  300  percent.   The  surplus-per-trip  estimates for  these  two  recreation



centroids exceed $20, and the coefficients of determination are relatively  lower



for these two centroids.   The results for these two centroids may be regarded as



outliers  and  therefore  dismissed,  but  it  is   significant  that  aggregating



population  centroids into  zones produced  outliers  whereas  use  of  population



centroids as origins did not.




     The  conclusion  that travel-cost  valuation estimates  are  sensitive to the



definition  of  the origin zone  raises  the question  of  which  definition is most



appropriate.  The average of the coefficients of determination favor the use of



population  centroids as  origin  zones;  but the  differences  in  R2  values between



models  do  not  provide  sufficient  evidence  to  resolve this  issue.    The two



extreme  estimates  (centroid 22.0  and  22.2)  obtained from the 10-  and 20-mile



origin  zone equations  raise a question  about aggregating,  but are  also not



compelling  evidence against  it.   A third potential indicator of the proper model



is  the  ability of  the  statistical  estimate of the  first-stage demand  curve to



estimate  known quantity demanded at a zero price.



     Table  11 depicts  the  assumed known quantities  and endogenous estimates of



this  variable using  10-  and 20-mile origin zones and using recreation  centroids



as  origin zones.   The  main   result is that aggregating population centroids into



either  10-  or  20-mile  zones substantially  improves  the ability of the model to



predict  total  use at  a  zero price.   Although  aggregating populations  improves



the  predictive  ability of  the  model in this sense,  the quantity estimates for



several centroids still contain substantial errors.



     The  result that aggregating population centroids into  concentric zones does



not  improve the  R2  values, but  does  improve  the estimates of  total   quantity



demanded,  is easily explained.   Visit  rates  diminish with distance  from the



site, but the number  of population centroids  increases with  distance from the






                                                                              93

-------
                                     TABLE 11

ESTIMATES OF  QUANTITY DEMANDED BY  CENTROID USING SEMILOG  AND  DOUBLE-LOG FORMS
AND VARIOUS DEFINITIONS OF ORIGIN  ZONES (IN THOUSANDS OF VISITOR-DAYS)
                             Semi log Results
                                        Double-Log Results

Recreation
Centroid
Number
(1)
17.0
17.1
17.2
18.0
19.0
19.1
20.0
21.0
22.0
22.1
22.2
23.0
23.1
23.2
24.0
24.1
24.2
24.3
25.0
26.0

Exogeneous
Quantity
Demanded
(2)
347
956
1,092
63
33
87
105
113
6
10
15
42
45
217
7
12
14
23
42
21


Recreation
Centroids
(3)
534
1,270
1,354
70
36
134
135
244
4
8
9
57
59
303
5
9
10
19
49
17
Ten -
Mile
Origin
Zone
(4)
352
912
1,009
47
32
97
114
148
18
18
43
37
39
206
7
26
16
61
30
32
Twenty -
Mile
Origin
Zone
(5)
313
829
833
34
27
87
98
145
24
14
40
31
33
185
4
21
13
43
24
26


Recreation
Centroids
(6)
3,751
17,657
111,807
572
122
475
882
800
15
130
102
389
1,279
10,549
52
1,223
84
7,081
91
1,015
Ten-
Mi le
Origin
Zone
(7)
492
1,357
1,656
56
45
134
261
243
34
39
101
48
55
345
6
20
15
68
37
106
Twenty
Mile
Origin
Zone
(8)
540
1,381
1,333
44
59
155
240
337
51
48
113
51
50
325
4
18
12
43
34
116
   Mean
166
216
171
141
8,234
256
248
     Note:   The  quantity estimates  in  columns  3  through 8  are  obtained  by
 letting AP = 0 in the appropriate least squares estimate of Eq.  (IV.4).


 site.  When population  centroids are used as origins, there is a large number

 of  observations  of  low  visit rates  that   are close  to  the  regression line.

 The very  few  origin zones that  have  high  visit  rates and account for most of

 the total visits have relatively little  influence on the regression line.  The

 visit  rates  of  the  close  origin  zones  are  often estimated with  large

 residuals.  Aggregation results  in a  large number of good-fitting observations
94

-------
being  combined  into a few observations  and,  hence,  reduces  their influence on
R2.
     Aggregation  decreases   the   total  number  of  observations  and  thereby
increases  the  relative weight of  the  close  origins  in  determining the  regres-
sion  line.   The error in estimating  these visit rates thereby  decreases,  and
hence,  so does  the error in estimating total  visits.   The "solution" to  the
visit  estimation problem is  not  increased  aggregation;  because  aggregating
from  a 10-mile  origin zone  to  a  20-mile  origin zone actually  decreases  the
reliability  of  predicting  total   visits  (see  Table  11,  columns 4  and  5).
 Indeed,   total  visits  could  be   predicted   exactly  if  populations  were   of
constant  size across  origins.9
      Simulation  estimates  of each of  these cases  were again  made using  a
 double-log  form.   The  results  using  a  $1 price  increment  and  exogenous
 quantity  demanded are presented  in Table  12.  The coefficients  of  determina-
 tion  in columns 2  and 5  are lower for  a  double-log model  than  for  a  semilog
 model  when  population centroids   are  aggregated into  10-  or 20-miles zones.
 Furthermore, most of  the surplus-per-day  estimates  are higher than  one  could
 reasonably expect.    Overall,  the aggregation process provides  no credibility
 to the   double-log  form.   This   result  also   follows  when  we   consider  the
 double-log  estimates  of total  use  at  a zero  price.   As  seen   in  Table  11,
 aggregating  population centroids  into  10-  or 20- mile origin  zones improved
 the predictability  of the model  in terms of total  use.   However, the double-
 log  model   predicts   total   use  with  a  larger  error  than  a semilog model,
 regardless of the choice  of  origin zone.
 9The estimated  residuals  in predicting  visit rates  necessarily sum  to  zero,
that is,  I(V.  -  V.)  = I(T./N.  -  T./N.) - 0.   If the population of each origin is
identical,  NI(T. -  T.) = 0,  visits (T.) are also predicted exactly.

                                                                              95

-------
                                     TABLE 12




    DOUBLE-LOG VALUATION  ESTIMATES  USING  10-MILE AND 20-MILE ORIGIN ZONES
                    10-Mile  Origin Zones
                            20-Mile Origin Zones
Recreation
Centroid
Number
(1)
17.0
17.1
17.2
18.0
19.0
19.1
20.0
21.0
22.0
22.1
22.2
23.0
23.1
23.2
24.0
24.1
24.2
24.3
25.0
26.0
R2
(2)
0.556
0.599
0.608
0.803
0.493
0.536
0.535
0.723
0.366
0.491
0.175
0.775
0.844
0.736
0.797
0.813
0.782
0.711
0.764
0.461
Consumers'
Surplus
(in $1000)
(3)
8,810
34,694
36,984
405
500
2,312
8,386
7,171
1,050
906
4,326
902
926
9,984
15
180
62
238
505
3,242
Surplus
per Day
$
(4)
25.36
36.29
33.84
6.44
15.10
26.54
79.72
41.52
184.24
87.75
298.29
21.52
20.31
46.10
2.04
14.69
4.26
10.23
12.05
153.28
R2
(5)
0.596
0.588
0.673
0.804
0.454
0.434
0.555
0.668
0.283
0.469
0.399
0.878
0.903
0.855
0.776
0.792
0.822
0.670
0.772
0.459
Consumers'
Surpl us
(in $1000)
(6)
18,546
38,203
19,616
770"
1,631
5,044
7,732
15,291
2,114
1,658
5,345
1,445
1,150
9,324
15
227
60
230
561
4,083
Surpl us
per Day
$
(7)
53.39
39.96
17.95
12.25
49.38
57.90
73.51
88.54
371.10
158.72
368.54
34.46
25.25
43.06
2.15
18.49
4.15
9.92
13.39
193.05
               0.627
6,080
55.93
0.643
6,653
.81.76
3.    Conclusions and Implications



     This chapter presents travel-cost demand and value estimates for boating in



20  recreation  centroids  in  Washington.   The  objective of  the analysis  is  to



determine the  sensitivity of the  results to  three  specification  and  computa-



tional  assumptions,  and  thus  to  determine which assumptions  are  most  appro-



priate.



     A  Monte  Carlo analysis  is  used  to  examine questions  that have  not been



resolved theoretically  or empirically.   We  can find  plausible results  for at



least one centroid  in  each of the tables;  but by  observing results for several




96

-------
sites,  the deficiencies in various assumptions becomes apparent.  The recreation



literature has  given only  cursory attention to the  issues  considered  here and



many existing studies have been based on a single site.



     The preference  of  most recreation analysts for  a semilog specification of



the first-stage demand function over a double log is confirmed by these results.



In  terms  of goodness  of fit,  stability of  results  across sites,  accuracy of



predicting quantity  demanded at  a zero  price,  and a  priori  reasonableness of



results, this specification is clearly superior to the double log.



     Some  recreation analysts,  such  as  Common  (1973),  have used  a double-log



specification with satisfactory  results.   However,  Common and others have tried



alternative  specifications  for only  one site.   For  some centroids, consumers'



surplus estimates  are  insensitive  to the  specification,  but this  result is  a



special  case that  may  be  observed  in  a sample of  one site.   A  particularly



serious problem with the  double-log specification  is that on  occasion  it can



produce totally unrealistic results.   The source of this problem is unclear and



cannot  be determined  from  the  regression  estimates  of the  experience  demand



schedule.    The  cause  of  these  occasional  drastic  results may,  to  a  lesser



extent, affect  the apparently tenable results, hence these estimates should also



be considered suspect.



     This analysis of boating at 20 recreation centroids  reflects a  small sample



of the  195  centroids and four recreation  activities  considered in  the regional



model.   Recreation experience demand curves were estimated for each  centroid and



for  each  activity  using   both  a  double-log and  semilog  specification.   The



results are  simliar  to  those reported  here, with  some estimates of consumers'



surplus varying in  sensitivity  to the  choice  of  functional  form.   About  five



percent of the  results using a double-log specification  are  unreasonable.
                                                                               97

-------
     The price-quantity observations depend upon the price increment used in Eq.



(IV.5)  and  hence  the area  under  these  price-quantity observations,  which is



consumers'  surplus, could  also  be affected by the  size of the price increment.



As seen by comparing Table 7 with Table 9, when a double-log form is used, total



surplus and surplus per  trip are sensitive to the size of this price increment.



With a semilog specification, the results using a $0.25 price increment up to $1



and a  $1  increment thereafter are virtually identical to those obtained using a



$1 increment.   The robustness  of the semilog also suggests  its  superiority to



the double-log form.



     Quantity  demanded  at a  zero price  is  usually estimated  exogenously  from



site data, but it can also be  estimated  by  setting AP =  0  in  Eq.  (IV.5).   The



first estimate is based on observed (visit-rate) data and the second estimate is



based  on  visit rates  estimated from a regression  equation.   The two estimates



are not identical, but one would hope that differences would be small and have a



mean of zero.   When origin zones are defined as population centroids, we observe



wide  differences   between  exogenous  quantity  estimates and  quantity estimates



obtained from  Eq.  (IV. 5).   One implication of this  result is that if empirical



estimates of  Eq.  (IV.5)  were used to predict visits at a similar proposed site,



a substantial  error would be expected.  Second, visit-rate schedules would yield



inaccurate estimates  of  the effect of  initiating an  entrance fee on total  use.



As  seen in  Panel  B  of  Figure  1,  imposing a  fee  may  lead  to  an  increase in



predicted visits.



     When  a  semilog model  is used, discrepancies  in  quantity  estimates do not



produce  discrepancies  in  total  consumers'  surplus  estimates (compare Table 7,



column  3,  with Table  8,  column 4).  This result implies that it  may be feasible



to estimate surplus at a proposed  site,  even  when  use cannot be estimated with



reliability.    Errors in estimating quantity demanded have a negligible effect on





98

-------
total surplus because  surplus  is estimated with Bode's Rule and not as the area



under a  regression equation.   A regression estimate  of  a site-demand equation



would be affected  by  the choice of quantity demanded.   However, by using Bode's



Rule  to  measure  area under  several  points,  the  choice  of  one price-quantity



point  affects  the area only  in  the  neighborhood  of that  point.    A  second



compelling reason for using Bode's Rule is to avoid the issue of the appropriate



functional  form of the site-demand curve.



     The most disconcerting  result of this chapter  is  that valuation estimates



are  sensitive  to  the  definition  of the  origin  zone.   When  each  population



centroid is  construed  as a separate  zone, the explanatory power of the model is



higher on  the  average  than  when  centroids  are aggregated  into 10-  or 20-mile



zones.   Furthermore,  aggregating  centroids  results  in  a  substantial  loss  of



degrees  of  freedom,  which with  other things being equal,  is undesirable, and in



this  case  causes  the results to  become unstable.   However,  aggregating popula-



tion  centroids  into origin  zones  improves  the  accuracy  by  which  total  use is



predicted at a  zero price.



      Most  travel-cost  studies  have  been  based  on  an  aggregation of population



centroids  into  concentric zones.  The choice of a 10-mile versus 20-mile system



of  concentric  circles  affects  the  results,  but  there is  a  greater  disparity



between  using zones and  using population centroids as origins.   A consequence of



using  each centroid as  an origin  is that a  large  proportion of the  centroids



account  for a small proportion  of  the trips.   In rough  numbers,  about 95 percent



of the centroids account  for only  10  to 15 percent of the  trips.  The experience



demand   curve   is  therefore  influenced  disproportionately  by  centroids  that



account  for  very  few  trips.    There  is  some  justification  for  using  each



population  centroid  as  an   origin   zone  and  for  aggregating  centroids   into



concentric  zones.   The  best  choice  is  unclear.   Because travel-cost  valuation






                                                                              99

-------
estimates  are  sensitive  to  the  definition of  the  origin  zone,  this is  an



important topic for future work.
100

-------
                                   CHAPTER VII

             EMPIRICAL ESTIMATES OF RECREATION BENEFITS OF IMPROVED
                     WATER QUALITY IN THE PACIFIC NORTHWEST
     This chapter  presents  empirical  estimates  of recreation  benefits to  be

gained through improving water  quality of degraded rivers and preserving water

quality in selected  lakes  in the Pacific Northwest.  The  first  section of  this

chapter presents a brief  overview of the main determinants of recreation demand

and  valuation.   The  objective  is  to  provide an  intuitive  explanation of  the

model  and  of the  subsequent empirical estimates.  The  second section  presents

estimates of existing  recreation  benefits of eight selected  lakes.   The sample

permits  a  contrast in  benefits  between  urban  and rural  locations.  Section 3

estimates recreation  benefits  on  a county  basis of improving water  quality in

all the degraded rivers in the Pacific Northwest.



1.   Determinants of Recreation Value and Use

     Travel-cost  analyses  have  documented  that  recreation behavior  can  be

explained quite  well  by  four  independent  variables:   population  size,  travel

cost  to  the  site, site  characteristics,  and   the  availability  of  substitute

sites.   The  population  centers  that  send  recreators  to  a specific  site  are

obviously a critical  determinant of potential demand.   The actual number of lake

users  is  influenced  more by the  potential  number of users  than  perhaps by any

other  variable.  However,  populations  of equal   size do  not  necessarily produce

the same  number of   recreation trips.   Demographic   characteristics   such as
                                                                             101

-------
household  size  and  income  influence  participation  rates.   In  the  Northwest,



these variables influence the  number  of trips per  household  [Eq.  (III.8)], but



population size is the main  determinant of days spent recreating.



     The number of  users  of a lake is influenced significantly by the distances



to  the  population  origins.    Recreators  typically  are  adverse  to  travel  and



therefore, other  things being  equal,  the greater the  required  travel distance,



the  fewer  will  be the users of a  lake.   The  increase in travel costs, particu-



larly gasoline,  in the last  several  years  could increase the  demand for lakes



closer to  population  centers  at the expense of more distant sites.   In addition



to  the  aversion  to  travel,  distance  also  tends  to  diminish  use because  the



greater  the  distance from  origins  to  the  recreation  site,  the  greater  the



probability of preferred substitutes closer to the population  origins.



     The  use  and  value of a recreation  site  depends  on  the existence of compe-



titive  or  substitute recreation  sites.   If  a  site  has  one  or  more  close



substitutes,  the  value-per-unit day will be  less  than if the site has no close



substitutes.   A site preferred over  its competitors will have a high use even



though  a  low value per use day.   A site that is generally less  preferred than



its  competitors will  have low use and  low  value per day.  Furthermore,  substi-



tute  sites may  be  located  in the  same proximity,  but  this  is certainly not



necessary.   Figure  9 depicts  two  population centers  that,  for  illustrative



purposes,  are assumed to be  located  on the same straight road.   In  this illu-



stration,  there  are  four  recreation sites  that   are  assumed to  be identical



except for location.  Most,  if not all, recreators from population center 1 will



visit site  A.   Site B is not  a close substitute for site A  because  of greater



travel costs.   Recreators from population center 2 will  recreate at  sites C and



D, which are close substitutes because of their identical travel distances, even
102

-------
                                    FIGURE 9

               THE EFFECT OF SUBSTITUTE SITES ON DEMAND AND VALUE
Distance
(in miles)       <- 5 ^ <-     25     -»• •*-  10  -» <-      30      -* <-      30      ->

Center       Site A Pop. 1         Site B   Site C           Pop. 2       Site D
(Population,
Recreation)
Trips from 100
Origin
Trips to 95 10 48
Destination
Value per $5 $0.50 $1
Day
100

47

$1


though  they are  60  miles  apart.   In  contrast,  sites B  and  C are  not close

substitutes even though they are only  10 miles apart.

     In this illustration, site A receives the greatest use and has the greatest

value-per-user day.   Demand is relatively  price inelastic because the site has

no  close  substitute.   Sites C and D receive significant use but have low values

per  user  day;  the demand for  each  of  these sites  is very price elastic because

each site is a close  substitute for the  other.   Site B  is closer to a population

center  than is either site C  or  site  D, however B  is  dominated by other sites

preferred  by both population  centers.  Thus, site  B  receives  only minimal use

and  has a  low value per  day.   By implication  it  may  not be cost-effective  to

improve recreation  opportunities  at site B, because the presence of a preferred

substitute  discourages use at B.

     This  example illustrates  the  importance  of  considering  substitute  sites

when  selecting lakes for  restoration.  To  emphasize   this  point,  consider the

conditions  in  Figure 9,  and assume that water  quality is  uniformly poor at all

sites.    Now,  which   lakes would  be  cost-effective  to restore?   The  highest

                                                                              103

-------
priority in  items  of  recreation benefits is likely to be site A.  This site has



the greatest potential demand because it is located near a population center and



has no  close  substitutes.   Sites C or D  are  also a high priority, but it would



probably be  cost-effective  to  restore only one of these sites.   If one of these



two lakes were  restored,  use and value per day  would be high at that site.   If



both were  restored,   use  would be divided  between the  lakes and  value  per  day



would be low.  If site A and either C or D were restored, restoring lake B would



probably  not  be  cost-effective  because  of  the  availability  of  preferred



substitutes.



     The fourth  major determinant of recreation  demand  and value  is  the site



characteristics  including  lake  size,  aesthetics,  recreation   facilities,  and



water quality.   These characteristics,  in combination, determine the ability of



a  site  to  attract recreators from  various  origins.   Defining,  weighing,  and



measuring  these characteristics  has proven  a  major  challenge  to  researchers



analyzing  recreation   demand  and  value.   Recreation  facility  data serve as  a



proxy for  these characteristics  because  these  data  are available  on  a county



basis across the entire region.



     If a  lake  is  to  be used  for public recreation, it must have public access.



Also, there must be facilities appropriate to the various recreation activities.



For  instance,  a  swimming   beach  is  important  for   swimming,  boat ramps  are



necessary  for  boating,  and camping  facilities  are  required for  camping.   No



particular   facilities  are  required  for  fishing,  but the  appropriate  site



characteristic is probably the anticipated catch.



     Water  quality is one  characteristic  of a  site  and  like  other aesthetic



qualities,  it  is subjective and difficult to define and to measure.  Estimating



the  response of recreation  use and  value  to changes  in water  quality is also



difficult  because  this response  varies  widely across  sites  and depends on the





104

-------
initial  value  of the  other  determinants of  recreation  demand.   To illustrate,



assume  that  water  quality  of each  site in  Figure  9 is  identical and  can be



described  as  moderate   to   good.    Under  these  assumptions,  water  quality



improvement would encourage additional demand at site C or site B, but not both,



because these  sites are  close  substitutes.   Improving water  quality  at  site A



will not encourage additional use because site A is already heavily utilized and



cannot attract recreationists from substitute sites.



     Water-quality  improvement  efforts  may  be  directed  toward  maintaining



existing  good  water  as  well  as  improving  the quality of degraded  water.



Maintaining existing value  and  use may  be the  goal  of preventive water-quality



programs if  in  the  absence of such action,  water would  become degraded and use



would decline.   Referring again  to Figure 9, water-quality protection would not



appear  to  be justified  at  site  B,  because   of  negligible demand.   This  result



depends critically on two assumptions:   (1) that water quality is uniformly good



at  other sites;  and  (2) there  exists a  site  that  is  preferred to site B.



However, if  water quality were uniformly poor,  improving  site B would encourage



a  significant  increase  in use from both population centers 1  and 2.  Preventive



actions may  not  be  justified at  C  or D because if either site became unusable,



demand  would merely  shift  to  the  other site,  which  is a  close substitute.



However, if  both sites were threatened, maintaining quality at both sites or at



least  at one  site  would induce  large benefits.   The  site  with  the greatest



potential preservation value is clearly  site  A.  The uniqueness of  site A is its



high  existing  value and use, which are  determined by its nearness to a popula-



tion  center  and  the absence of close substitutes.  This  point can be general-



ized.   The  greater  the current use of a site and  the fewer its substitutes, the



stronger is  the  justification for preserving water  quality at that site.   Even



with  the stringent  assumption that water quality  is uniformly good to moderate,






                                                                              105

-------
the  basic  conclusion  is that  those  sites  that  offer the  greatest potential



increase in  use and  value  are not  necessarily the  sites  where preserving the



existing level of water quality is cost-effective.








2.   Demand and Valuation Estimates for Selected Lakes



     A  sample of eight  lakes  was  selected to  indicate the recreation value of



preserving good  water quality  at  urban versus  rural  locations.   The demand and



valuation estimates of these lakes are explained in terms of the determinants of



demand,  as  described  in the  previous  section.   Although  three of  the lakes



considered here  are  affected by the EPA's Clean Lakes Program (Liberty, Medical



and Fernridge Reservoir), demand and benefits are estimated under the assumption



that existing water quality does not discourage use.



     A  summary  of  the demand and valuation  estimates  is  presented in Table 13.



Column  3 contains  data  on  facilities  for  the  corresponding  activity.   This



variable is  a proxy for the characteristics of a  recreation site and should be



positively correlated with demand and value of the site.  The accessibility of a



recreation  site is  measured  as  the weighted  sum  of  recreation  activity days



emanating from  each  population center, as defined by  Eq.  (III.5).  The weights



are  the probability  that a person will  travel  the  distance from the respective



origin  to  the corresponding recreation  site.   Accessibility  reflects the joint



influence  of population  size  and  distance.   The closer  a site  to population



centers  and  the  greater the  number  of  recreation  trips  produced  by these



centers,  the  larger  the accessibility  number.   The  level  of  facilities  and



accessibility of a site jointly determine total visitor-days by  activity  for the



site.




     The first  two  lakes shown in Table 13, Lake Washington and Lake Sammamish,



are  large  urban  lakes  with numerous recreation  facilities.    These  lakes are





106

-------
                                                TABLE  13

      ANNUAL  RECREATION  DEMAND  AND  VALUE OF  SELECTED  LAKES  IN THE  PACIFIC NORTHWEST  (1979 DOLLARS)
Recreation
Centroid
Number
(1)
17.1
17.2
10.0
24.0
48.1
32.0
32.2
103.2
Lake and County
(2)
Lake Sammamish
King County
Lake Washington
King County
Twin Lakes
Ferry County
Perrygin Lake
Okanogan County
Priest Lake
King County
Medical Lake
Spokane County
Liberty Lake
Spokane County
Fernridge Reservoir
Lane County
Activity
(3)
Swimming
Campi ng
Fishing
Boati ng
Total
Swimming
Camping
Fishing
Boati ng
Total
Swimming
Camping
Fishing
Boating
Total
Swimming
Camping
Fishing
Boating
Total
Swimming
Camping
Fishing
Boating
Total
Swimming
Camping
Fishing
Boating
Total
Swimming
Camping
Fishing
Boating
Total
Swimming
Camping
Fishing
Boating
Total
Facility*
(4)
3,000
150
220
12

7,000
1
100
39

400
117
525
4

275
120
800
4

6,750
1,265
310
26

150
4
61
5

100
25
50
1

6,608
200
75
33

Access.**
(5)
2,755,403
2,633,395
2,003,932
1,596,838

2,883,916
2,511,231
2,199,562
1,656,586

201,214
507,609
117,229
223,566

90,521
333,922
45,169
139,078

218,585
550,561
143,338
240,518

989,653
1,357,021
650,469
705,431

1,232,264
1,637,891
851,309
880,733

309,616
777,142
335,729
254,786

Annual
Visitor
Days
(in 1000)
(6)
4,069
2,123
1,745
4,424
12,361
4,837
153
1,791
9,616
16,397
100
170
95
41
406
46
93
47
18
204
157
640
112
137
1,046
747
134
509
563
1,953
844
441
639
303
2,218
383
446
311
193
1,333
Recreation
Value
(in $1000)
(7)
18,446
8,038
5,085
30,968
62,557
21,880
603
4,589
66,773
93,845
596
905
535
342
2,378
393
729
402
194
1,718
818
2,743
485
1,005
5,051
3,114
551
1,806
3,594
9,065
3,970
1,829
2,387
2,114
10,300
2,236
2,147
1,117
1,741
7,241
Value
per
Trip
S
(8)
4.54
3.79
2.91
7.00
5.06
4.52
3.94
2.56
6.94
5.72
5.97
5.33
5.64
8.33
5.86
8.58
7.87
8.54
10.76
8.42
5.22
4.29
4.31
7.35
4.83
4.17
4.10
3.55
6.38
4.64
4.70
4.15
3.79
6.97
4.64
5.83
4.81
3.59
9.01
5.43
 *The facilities for the  activities  are:   linear beach feet,  camping  units,  acceptable river and shore-
line miles,  and number of boat ramps.

**Access.  means accessibility  and  is  a positive function of  the  nearness to population centers  and  the
size of these centers.   See Eq. (III.13),  p.  52.


                                                                                                    107

-------
located  in  the  Seattle  Standard  Metropolitan  Statistical   Area   (SMSA),  the

largest  concentration  of  people in  the  Northwest.   The  annual  value  of the

water-based  recreation  activity  on these  lakes  is estimated to  be  $93.8 and

$62.6 million, respectively,  making them the most  valuable  recreation lakes in

the region.1  These  high  annual  values reflect the combination of short travel

distance, large population centers,  and numerous recreation facilities, particu-

larly for swimming and boating.

     Of  the  other  six lakes  in  Table  13,  three are urban and three are rural.

Twin  Lakes,  Perrygin Lake, and  Priest  Lake  are each more than  50  miles from a

major population center.   In  contrast, Medical  Lake and Liberty Lake are within

20  miles  of  Spokane,  Washington  (SMSA population is 304,058),  and Fernridge

Reservoir  is  about  12 miles  from  Eugene-Springfield,  Oregon  (SMSA population,

271,130).  Fernridge Reservoir is used more extensively than the other lakes and

has a  corresponding  higher value.   The use and value of the other five  lakes is

similar.

     The significance of an urban versus rural  location is easily appreciated by

comparing  swimming estimates  of  two urban lakes, Medical and Liberty, with those

of  the  two rural  lakes,  Perrygin and Twin Lakes.   Medical and  Liberty have only

150  and 100  linear  beach feet,  respectively,  whereas Perrygin and Twin Lakes

have  275  and  400  linear swimming   beach  feet.   The  urban  lakes are  very

accessible   (see  column  5,   Table  13),  and  the  two rural   lakes  relatively

inaccesible.   Thus,   even  though   the  two  urban  lakes   offer fewer   swimming

opportunities (measured by linear beach feet), they receive more use and have  a

corresponding higher value than  the rural  lakes.
 1Crater Lake, in Oregon, attracts visitors nationwide and even from abroad, but
not   for   recreation  reasons,  because  fishing,  boating,  and  swimming  are
prohibited.   Crater  Lake certainly has a high  value,  but it is not included  in
this  study.


108

-------
     Although benefit estimates  are  presented here for  only  eight lakes in the



Northwest,  some  general  conclusions  are  suggested.    First,  those  lakes  that



offer extensive  recreation  opportunities and are  located  near large population



centers will  receive extensive  recreation  use,  and  this use will  have a high



total value.   The  demand  and valuation estimates for  Lake  Washington  and Lake



Sammamish reflect their proximity to large population centers and their abundant



recreation facilities.  The  estimate of annual recreation value of $7.2 million



for  Fernridge  Reservoir is  significantly  less than the estimates  of  the above



two  lakes.   However,  this  reservoir is  smaller,  located  further  from  a popula-



tion center, and the population center has fewer people.  The counterpart to the



principle  stated above  is  that  lakes  without  recreation facilities  that  are



located a significant distance from major population centers will  not be heavily



used and  will  have  corresponding low recreation values.  The two lakes farthest



from population  centers,  Twin  Lakes and Perrygin Lake,  have  the lowest total



recreation  value.   Two of  the lakes (Newman and  Liberty)  are not particularly



attractive  in  terms of  their  recreation  facilities,  but  are   located  near



Spokane,  Washington,  which  is a large  urban  center.   The  other three lakes are



located  in  rural areas  but offer  appealing  site characteristics  that attract



recreators from  several miles.








3.   Benefits of Improving Water Quality in Streams



     This section presents estimates of  recreation benefits that would accrue if



the  degraded  rivers  and streams in the  Pacific Northwest were made fishable and



swimmable.



     In  the  Northwest,  camping, fishing, swimming,  and boating generally occur



where water  quality  is  high and appropriate facilities are available.   In those



areas where  water  is degraded, recreation facilities have not been provided and






                                                                             109

-------
recreation  does  not  occur.    Water  quality  and  recreation  facilities  are

complements and  for purposes  of  this study  are  assumed to  be  perfect comple-

ments.   Improved water quality will  not stimulate recreational use unless there

is a corresponding  improvement of related facilities such  as swimming beaches,

boat ramps, or campsites.   Recreation facilities are not viewed as a true causal

variable,   but as  a statistical  proxy  for  a  large number  of  nonquantifiable

variables   that  in   combination  determine the  attractiveness  of a  recreation

site.2  Fishing is the exception as no facilities  are required for fishing.  The

quality variable  that  is assumed  comparable  to  recreation  facilities  is  the

number  of  fishable  river  miles  and  lake  shoreline  miles.   The  exogenous

variables  that  drive the  model are:   linear swimming beach feet, number of boat

ramps,  camping  units,  and fishable  river and  shoreline miles.   An  increase in

any of  these  variables  will  increase demand and consumers'  surplus,  where it is

implicit   that   water  quality   is   "acceptable"   for   recreational   purposes.

Similarly, an improvement in water quality must be accompanied by an increase in

one  or more  of the  above  variables if  use  and benefits  are to be  affected.

     The  Region  X  (Seattle)  office of the U.S. Environmental  Protection Agency

(EPA)  has  published a series of  water-quality  assessment  reports covering each

major river basin in the Pacific Northwest.   Water quality was assessed for each

major  stream  and for  various  reaches on these streams  using both recreational

and biological  criteria.   Water quality is indicated for  recreation  in general
 2A report by the Institute of Transportation and Traffic Engineering (1971, ch.
7) includes an effort to construct a recreation attractiveness index for camping
by  defining  28 characteristics  of campgrounds and  applying  factor analysis to
select the most important factors.  The factors that were selected accounted for
41.5%  of the  variance  in  the  observed data.   It is  not  feasible to  apply a
similar  approach  in  this  study because much  of  the  required data do not exist,
and the  data that  are  available  are  of poor quality.   The  large scale effort
that  would  be  required  to  complete  the  analysis  is  not  justified  by the
improvement in the results.


110

-------
and  is  not estimated  for each  of  the four activities  analyzed in this paper.



Water quality was not measured on a continuous scale, rather it was judged to be



acceptable,  objectionable,   or  not  acceptable.    An acceptable  stream  is  one



meeting the 1983  Federal  water-quality goals of fishable and swimmable streams.



Although standard water quality parameters, for example, turbidity, were used in



assessing  water  quality,  professional  judgment  was also  a  factor.   The objec-



tionable and not acceptable river stretches were noted on U.S.  Geological Survey



base maps  for  each  of the three states.  A planimeter was then used to tabulate



degraded and acceptable  river miles on a county basis.   The resulting estimates



of  acceptable  and  degraded  river  miles  serve  as   the  basic  water-quality



inventory  data for this study.



     Given the assumption that water quality and facilities are perfect comple-



ments,  it  is  feasible to  estimate  water-quality  benefits  by  estimating  the



benefits of  increasing facilities  on degraded rivers.   The number and type of



facilities  that  could   be   constructed  if  water  quality were   improved  was



estimated  by  state  recreation planners.   The assistance  of  these planners was



sought  because  of their  first-hand knowledge of  the  recreational potential of



the  various  areas  in  their  respective  states.   The  recreation  planners were



shown  a U.S.  Geological  Survey  base map (scale  1:500,000)  of  their state with



the  degraded  rivers  marked  and asked  the  following question  for each degraded



river segment.  "If water quality were  improved, would  this area be conducive to



any  of  the  four  activities  being considered  here?"   When  the  answers were



affirmative,  the  next question  asked  was,  "How many  facilities  by type could



reasonably be  constructed along the degraded river?"   Although this method of



estimating the  potential increment  in  facilities  certainly  lacks  scientific



rigor,  all recreation  facilities  data were  obtained  from the  state  recreation
                                                                              111

-------
officials, whose responsibility  is  to  recommend the development of state recre-



ation areas.



     Based on  these  interviews  with the  state recreation  planners,  estimates



were  made of  the  potential  increment  in  recreation  facilities (boat  ramps,



swimming beach feet,  and campsites)  that could be constructed at each recreation



centroid  if   the  degraded  water  were  improved.    Estimates  of  observed  and



potential  facilities  by  activity  and  recreation  centroid  are  presented  in



Appendix  A,  Table A. 5.   The  incremental   variable  that  enters  the  model  for



fishing  is degraded  river  miles  by centroid.   This  variable was  estimated with



EPA  data and did not require the  assistance of the  recreation  planners.   The



estimated increment in facilities should be interpreted as the maximum potential



change  and  not as  an  estimate  of what  would  occur  if  water quality  were



improved.  Recreation benefit estimates therefore represent an upper bound  that



can  be  attained  only by  cooperation  with those  responsible for planning  and



developing recreation sites.



      Recreation demand and  value was estimated for  each  of  four  activities and



for  each of  the 195  selected recreation centroids in the Northwest on the basis



of  existing  water  quality  and level of  facilities  (see  Table A.4).   Demand and



value  were  again estimated  assuming that  all degraded water  was  made fishable



and  swimmable  and the  assumed  facilities  were constructed.   The increments in



benefits  for  each  activity are  presented on a county basis in Table 14.  The 16



recreation  centroids  in  western Montana  are  not  included  in  this  analysis.



      The  main  result  from  Table 14 is  that  substantial  incremental  benefits,



(for  example,  over one million  dollars) are  concentrated  in a few counties and



that  most counties show much lower benefits.  As expected, the counties with the



largest  potential benefits are those accessible to the largest populations.  The



Washington counties  with  the largest populations in order are King, Pierce, and





112

-------
                                    TABLE 14

ANNUAL RECREATION BENEFITS OF  IMPROVED  WATER QUALITY IN STREAMS  BY ACTIVITY AND
BY COUNTY FOR WASHINGTON,  OREGON,  AND IDAHO




County
Adams
Asotin
Benton
Chell an
Cl all urn
Clark
Columbia
Cowl i tz
Douglas
Ferry
Franklin
Garfield
Grant
Grays Harbor
Island
Jefferson
King
Kitsap
Kittitas
Klickitat
Lewi s
Lincoln
Mason
Okanogan
Pacific
Pend Oreille
Pierce
San Juan
Skagit
Skamania
Snohomish
Spokane
Stevens
Thurston
Wahkiakum
Walla Walla
Whatcom
Whitman
Yakima
Total Incre-
mental
Benefits


Swimming
$
0
0
20,471
0
0
0
0
0
65,794

0
0
0
26,692
0
0
0
0
0
0
0
0
0
0
0
0
0
0
467,354
0
0
637,896
0
251,302
0
0
0
-0
0

1,469,509



Camping
$
0
222,016
0
0
0
0
0
0
414,745
138,208
0
0
194,394
0
0
0
2,358,906
0
204,163
0
0
0
0
127,453
0
0
1,347,167
0
0
0
0
1,051,884
0
0
0
0
0
0
453,419

6,512,335

Washington

Fishing
$
0
2,598
14,071
0
0
0
0
0
7,207
8,101
0
0
1,830
0
0
0
14,373
0
0
0
0
0
0
0
0
0
165,450
0
0
0
0
0
36,335
26,445
0
7,960
0
0
11,993

296,363



Boating
$
0
30,327
0
0
0
0
0
25,015
24,616
0
0
0
0
0
0
0
693,112
0
0
0
154,300
0
0
136,550
0
0
1,050,138
0
0
0
0
870,667
0
5,360,468
0
0
0
0
1,013,475

9,360,668


Total
Recreation
Benefits
$
0
254,941
34,542
0
0
0
0
27,015
512,362
146,309
0
0
196,224
26,692
0
0
3,066,391
0
204,163
0
154,300
0
0
264,003
0
0
2,562,755
0
467,354
0
0
2,560,447
36,335
5,638,215
0
7,960
0
0
1,478,887

17,638,895


113

-------
TABLE 14 (continued)





County
Ada
Adams
Bannock
Bear Lake
Benewah
Bingham
Blaine
Boise
Bonner
Bonnevil le
Boundary
Butte
Camas
Canyon
Caribou
Cassia
Clark
Clearwater
Custer
Elmore
Franklin
Fremont
Gem
Gooding
Idaho
Jefferson
Jerome
Kootenai
Latah
Lemhi
Lewi s
Lincoln
Madison
Minidoka
Nez Perce
Oneida
Owyhee
Payette
Power
Shoshone
Teton
Twin Falls
Valley
Washington
Total Incre-
mental
Benefits



Swimming
$
0
0
131,356
0
0
0
3,808
0
0
7,705
0
34,910
0
29,048
21,400
139,965
22,589
12,091
2,988
0
2,662
20,775
0
129,722
0
22,572
0
0
0
618
0
52,829
18,864
0
0
64,168
0
0
0
185,954
40,826
86,172
0
5,647

$1,036,669




Camping
$
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
73,216
17,311
0
0
0
0
0
0
0
0
0
0
1,300
0
0
0
0
0
0
0
0
0
67,601
0
0
0
0

$159,428

Idaho


Fishing
$
0
0
0
0
0
37,707
88,530
0
0
15,237
0
34,481
0
12,798
0
19,925
2,390
0
106,986
0
4,184
5,506
0
0
0
164,082
0
0
0
0
0
40,659
57,764
0
0
21,117
10,828
0
0
0
13,092
43,814
0
149

$679,249




Boating
$
0
652
0
925
0
5,993
0
0
0
0
0
0
0
37,532
0
8,706
0
17,922
5,267
0
2,435
697
0
4,583
0
0
0
0
0
653
0
0
0
0
0
2,956
0
0
0
45,743
2,107
3,740
11,954
0

$151,865


Total
Recreation
Benefits
$
0
652
131,356
925
0
43,700
92,338
0
0
22,942
0
69,391
0
79,378
21,400
168,596
24,979
103,229
132,552
- 0
9,281
26,978
0
134,305
0
186,654
0
0
0
2,571
0
93,488
76,628
0
0
88,241
10,828
0
0
299,298
56,025
133,726
11,954
5,796

$2,027,211

114

-------
TABLE 14 (continued)




County
Baker
Benton
Clackamas
Clatsop
Columbia
Coos
Crook
Curry
Deschutes
Douglas
Gilliam
Grant
Harney
Hood River
Jackson
Jefferson
Josephine
Klamath
Lake
Lane
Lincoln
Linn
Malheur
Marion
Morrow
Multnomah
Polk
Sherman
Tillamook
Umati 1 1 a
Union
Wai Iowa
Wasco
Washington
Wheeler
Yamhill
Total Incre-
mental
Benefits
Total Regional
Incre-
mental
Benefits


Swimming
$
0
1,597,504
0
0
0
0
2,858
0
157,796
65,229
0
0
0
0
0
0
0
0
0
0
0
875
0
0
0
0
112,649
0
0
0
0
0
0
56,586
0
1,353,222

$3,346,719


$5,852,897




Oregon



Camping Fishing Boating
$
0
0
0
0
0
0
179,842
0
83,710
485,883
0
0
6,877
0
0
0
0
1,099
0
711,274
62,287
0
0
0
0
0
0
0
0
0
0
0
0
62,669
0
766,698

$2,360,339


$9,032,122


$
0
42,056
0
0
0
21,716
0
0
11,016
110,447
0
0
1,734
47,798
0
0
0
0
0
87,815
64,894
0
0
0
0
0
30,448
0
0
0
0
0
0
4,515
0
26,336

$448,775


$1,424,387


$
0
0
97,130
0
0
0
12,601
0
4,506
40,153
0
0
0
0
0
0
0
0
0
0
0
10,720
0
0
0
0
192,895
0
0
0
8,325
0
0
358,683
0
181,793

$906,806


$10,419,339



Total
Recreation
Benefits
$
0
1,639,560
97,130
0
0
21,716
195,301
0
257,028
701,712
0
0
8,611
47,798
0
0
0
1,099
0
799,089
127,181
11,595
0
0
0
0
335,992
0
0
0
8,325
0
0
482,453
0
2,328,049

$7,062,639


$26,728,745


                                                                              115

-------
Spokane, and these  counties  show corresponding large  recreation  benefits.   The



most populated counties  in  Idaho and Oregon are Ada and Multnomah.  Each county



has no water-quality  benefits,  but neither county has  a  water-quality problem.



     Fifty-eight of the total 119 counties indicate zero total potential recrea-



tion  benefits  and  several  more  show no  benefits  for certain activities.   Of



these  58  counties,  33 have  no  officially degraded water and  therefore have no



potential   benefits.   An  additional  16  counties  that  have  a  water-quality



problem  were  judged to  be  not  conducive  to recreation even  if water quality



were  improved.   These  counties  are  typically rural  where agriculture  is  the



economic base.



     Zero  or  low benefits  also occur for  those  counties  that are significant



distances  from population  centers.    Even  if water  quality were  improved  and



facilities  added,  demand would  not  increase significantly  in  those areas that



are  inaccessible.   Preservation values are significantly  larger  than potential



incremental benefits  for each  activity and for each state.   This result owes to



the abundance  of  existing accessible recreation opportunities.  The attractions



model,  Eq. (III-61)-(III-9')  (Table 2),  provides  empirical evidence  that  the



response  of attractions  to facilities  diminishes  as  the  level  of facilities



increases.  The results  in Table 14 cannot be extrapolated to other regions that



may have  different  population  densities,  existing recreation opportunities, and



water-quality problems.








4.   Conclusions



     A model has been presented that can be used to estimate recreation benefits



for four water-based activities within a three and one-half  state  region.  Bene-



fits can  be estimated for any  single  site or for several sites  simultaneously.



Benefits also  can  be estimated for preserving existing water quality as well as





116

-------
improving  degraded  water.   The main  conclusion  is  that,  with respect  to the



three Northwestern states,  the  largest potential  recreation benefits exist near



the  population  centers.    In  contrast,  improving  water  quality  in  sparsely



populated agricultural areas will  probably not stimulate a substantial  increase



in recreation demand.



     The benefit estimates  in  Table 14 may appear  discouraging in terms of the



economic  viability  of  meeting  the  national   goal  of "fishable and swimmable"



water.   Indeed,  improving water  quality  in some agricultural  areas  may  not be



cost-effective.  However, potential recreation benefits at several  sites exceeds



$1  million per  year.   Also,  certain  nonrecreation  benefits  such  as  property



values,  aesthetic values,  option demand,  and  perhaps  drinking  water and health



benefits  are  likely   to display  the  same  geographic   pattern   as  recreation



benefits.  That  is,  these potential benefits  may also correlate with population



densities.  A  more  comprehensive analysis of  benefits, focusing particularly on



those  listed  above  could  conclude  that  total  water-quality  benefits  are



substantially  larger  than  those  presented  in  Table  14.    For  example,  in  a



valuation  study  of  the Flathead Lake  and  River  system in western Montana using



this  model,  recreation  values  are   estimated  to  be  $6.3  million  per  year



(Sutherland  1982d).   However,  in  the  same  study,  nonuser values  (option,



existence, and bequest)  are estimated to be $97.3 million per  year  for the  same




region.
                                                                              117

-------
                                  APPENDIX  A



                                  DATA  TABLES
118

-------
                    TABLE A.I



POPULATION CENTROIDS,  POPULATION,  AND COUNTIES

Population
Centroid
Number
1.0
2.0
3.0
3.1
4.0
5.0
6.0
7.0
8.0
9.0
10.0
11.0
12.0
13.0
14.0
15.0
16.0
17.0
17.1
17.2
17.3
17.4
18.0
19.0
20.0
21.0
22.0
23.0
24.0
25.0
26.0
27.0
28.0
29.0
30.0
31.0
31.1
32.0
33.0
34.0
35.0
36.0
37.0
38.0
39.0



County
Adams
Asotin
Benton
Benton
Chellan
Clallum
Clark
Columbia
Cowl itz
Douglas
Ferry
Frank! i n
Garfield
Grant
Grays Harbor
Island
Jefferson
King
King
King
King
King
Kitsap
Kittitas
Klickitat
Lewi s
Lincoln
Mason
Okanogan
Pacific
Pend Oreille
Pierce
San Juan
Skagit
Skamania
Snohomish
Snohomish
Spokane
Stevens
Thurston
Wahkiakum
Walla Walla
Whatcom
Whitman
Yakima
Washington


Population Centroid
Othello
Clarkston
Kennewick
Richland
Wenatchee
Port Angeles
Vancouver
Dayton
Kelso
Watervi 1 le
Republ ic
Pasco
Pomeroy
Moses Lake
Aberdeen
Oak Harbor
Port Townsend
Seattle
Auburn
Kent
Renton
Bellevue
Port Orchard
Elensburg
Goldendale
Chahal is
Davenport
Shelton
Omak
Raymond
Newport
Tacoma
Friday Harbor
Mt. Vernon
Stevenson
Everett
Edmonds
Spokane
Col vi lie
Olympia
Cathlamet
Walla Walla
Bel 1 ingham
Pul Iman
Yakima



Population
13,322
16,822
42,383
66,291
44,980
51,224
192,060
4,098
79,489
22,156
5,748
34,613
2,483
48,040
66,356
44,016
15,903
998,909
50,568
37,925
38,397
139,061
145,990
24,866
15,879
55,450
9,597
30,896
30,654
17,234
8,561
482,692
7,793
63,184
7,914
221,739
114,214
341,058
29,008
124,249
3,824
47,267
105,198
40,321
170,767







(continued)
119

-------
TABLE A.I (continued)

Population
Centroid
Number
40.0
41.0
42.0
43.0
44.0
45.0
46.0
47.0
48.0
49.0
50.0
51.0
52.0
53.0
54.0
55.0
56.0
57.0
58.0
59.0
60.0
61.0
62.0
63.0
64.0
65.0
66.0
67.0
68.0
69.0
70.0
71.0
72.0
73.0
74.0
75.0
76.0
77.0
78.0
79.0
80.0
81.0
82.0
83.0

County
Ada
Adams
Bannock
Bear Lake
Benewah
Bingham
Blaine
Boise
Bonner
Bonnevil le
Boundary
Butte
Camas
Canyon
Caribou
Cassia
Clark
Clearwater
Custer
Elmore
Frankl in
Fremont
Gem
Gooding
Idaho
Jefferson
Jerome
Kootenai
Latah
Lemhi
Lewi s
Lincoln
Madison
Minidoka
Nez Perce
Oneida
Owyhee
Payette
Power
Shoshone
Teton
Twin Falls
Valley
Washington
Idaho
Population Centroid
Boise
Council
Pocatel lo
Montpel ier
St. Maries
Blackfoot
Ketchum
Horseshoe Bend
Sandpoint
Idaho Fall
Bonners Ferry
Arco
Fairf ield
Namp
Soda Springs
Burley
Dubois
Orof ino
Chalis
Mountain Home
Preston
St. Anthony
Emmet
Gooding
Grangevil le
Rigby
Jerome
Couer d'Alene
Moscow
Salmon
Kami ah
Shoshone
Rexburg
Rupert
Lewiston
Mai ad City
Homedale
Payette
American Falls
Kel logg
Driggs
Twin Falls
McCall
Weiser

Population
172,843
3,347
65,448
6,946
8,295
36,473
9,825
2,998
24,155
65,971
7,302
3,351
809
83,601
8,689
19,476
798
10,383
3,392
21,502
8,892
10,806
11,967
11,845
14,724
15,316
14,804
59,914
28,667
7,444
4,084
3,439
19,502
19,693
33,232
3,233
8,239
15,827
6,879
19,234
2,907
52,869
5,633
8,815
                                                                     (continued)
120

-------
TABLE A.I (continued)

Population
Centroid
Number
84.0
85.0
86.0
86.1
87.0
88.0
89.0
89.1
90.0
91.0
91.1
92.0
93.0
94.0
95.0
96.0
97.0
98.0
98.1
99.0
100.0
101.0
102.0
103.0
104.0
104.1
105.0
106.0
106.1
107.0
107.1
108.0
109.0
110.0
111.0
112.0
113.0
114.0
115.0
116.0
117.0
118.0
119.0

County
Baker
Benton
Clackamas
Clackamas
Clatsop
Columbia
Coos
Coos
Crook
Curry
Curry
Deschutes
Douglas
Gil 1 iam
Grant
Harney
Hood River
Jackson
Jackson
Jefferson
Josephine
Klamath
Lake
Lane
Lincoln
Lincoln
Linn
Malheur
Malheur
Marion
Marion
Morrow
Multnomah
Polk
Sherman
Tillamook
Umati 1 la
Union
Wai Iowa
Wasco
Washington
Wheeler
Yamhill
Oregon
Population Centroid
Baker
Corval 1 is
Lake Oswego
Oregon City
Astoria
St. Helens
Coqui 1 le
Coos Bay
Prinevi 1 le
Gold Beach
Brookings
Bend
Roseburg
Condon
Canyon
Burns
Hood River
Medford
Ashland
Madras
Grants Pass
Klamath Falls
Lakeview
Eugene
Newport
Lincoln City
Albany
Vale
Ontario
Salem
Woodburn
Heppner
Portland
Dallas
Moro
Til lamook
Pendleton
La Grande
Enterprise
The Dalles
Hillsboro
Fossil
McMinnvi 1 le

Population
16,127
68,078
193,085
44,120
32,467
35,709
15,453
48,477
13,097
13,186
3,749
62,117
93,100
2,061
8,216
8,306
15,810
115,279
16,156
11,556
52,937
59,048
7,523
271,130
15,185
20,129
87,743
18,727
8,164
181,964
22,490
7,525
559,058
45,201
2,177
21,170
58,816
23,935
7,269
21,711
245,684
1,511
55,230

                                                                     (continued)
                                                                             121

-------
TABLE A.I (continued)

Population
Centroid
Number
120.0
121.0
122.0
123.0
124.0
125.0
126.0
127.0
128.0



County
Cascade
Flathead
Gal latin
Flathead
Lake
Lewis and Clark
Lincoln
Missoula
Silver Bow
Western Montana


Population Centroid
Great Falls
Kali spell
Bozeman
Whitefish
Poison
Helena
Libby
Missoula
Butte



Population
89,367
41,462
67,414
10,000
19,098
49,992
17,731
79,091
95,067
                                          External Zones
129.0
130.0
131.0
132.0
133.0
134.0
135.0
136.0
137.0
138.0
139.0
140.0
141.0
143.0
Eastern Montana
British Columbia
British Columbia
Alberta
Wyomi ng
Utah
Nevada
Cal i form' a
Alaska
Eastern Canada
North Central
Northeast
Southeast
South Central
Bil 1 ings
Vancouver
Cranbrook
Calgary
—
—
—
—
—
—
—
—
—
— - —
159,117
2,206,608
200,000
1,838,037
470,816
1,461,037
799,184
23,668,562
330,000
18,687,959
50,571,000
61,880,000
41,487,000
31,440,000

     Notes:   The  population  estimates  for  western  Montana  counties  include
neighboring  counties,  for
Missoula  includes  Mineral
and  Powell;  and Jefferson
estimates  for all  United
from the 1980 census, U.S.
of Population and Housing
which  no  population
 and Granite;  Butte
 includes  Beaverhead
States  counties and
 Department of Commerce,
(by state), 1981.
centroid is  used.   For instance,
includes Silver  Bow,  Deer Lodge,
and Ravalli counties.   Population
states  are preliminary estimates
    Bureau of Census, 1980 Census
122

-------
                        TABLE A. 2




RECREATION CENTROIDS BY NAME, COUNTY,  AND CENTROID NUMBER

Recreation
Centre id
Number
1.0
2.0
3.0
4.0
4.1
5.0
5.1
5.2
6.0
7.0
8.0
8.1
9.0
10.0
11.0
12.0
13.0
13.1
13.2
14.0
14.1
14.2
15.0
15.1
16.0
16.1
16.2
17.0
17.1
17.2
18.0
19.0
19.1
20.0
21.0
22.0
22.1
22.2
23.0
23.1
23.2
24.0
24.1


County
Adams
Asotin
Benton
Chelan
Chelan
Clal Turn
Clallum
Cl all urn
Clark
Col umbia
Cowlitz
Cowl i tz
Douglas
Ferry
Frankl in
Garfield
Grant
Grant
Grant
Grays Harbor
Grays Harbor
Grays Harbor
Island
Island
Jefferson
Jefferson
Jefferson
King
King
King
Kitsap
Kittitas
Kittitas
Klickitat
Lewi s
Lincoln
Lincoln
Lincol n
Mason
Mason
Mason
Okanogan
Okanogan


Recreation Centroid
Northwest corner
Field Springs State Park
Crow Butte State Park
Lake Wenatchee State Park
Lake Chelan State Park
Bogachiel State Park
Neah Bay State Park
Dungeness State Park
Battleground State Park
Lewis and Clark State Park
Merwin Reservoir
Seaquest State Park
Chief Joseph
Twin Lakes
Lyons Ferry State Park
Pataha Creek
Potholes State Park
Sun Lakes State Park
Steamboat State Park
Bay City
Ocean City State Park
Lake Quinalt
Camano Island State Park
Deception Pass State Park
Kalaloch
Olympic National Park
Dosewallips State Park
Snoqualm
Lake Sammamish
Lake Washington
Horshoe Lake
Wawapum State Park
Lake Kachess
Horsethief Lake State Park
Ike Kinswa State Park
Grand Coulee Dam
Fort Spokane
Sprague Lake
Lake Cushman
Bel fair
Dash Point State Park
Pearrygin Lake State Park
Conconolly State Park





(continued)
123

-------
TABLE A. 2 (continued)
Recreation
Centroid
Number
24.2
24.3
25.0
26.0
26.1
27.0
27.1
27.2
28.0
29.0
29.1
30.0
30.0
31.0
31.1
32.0
32.1
32.2
32.3
32.4
33.0
33.1
33.2
34.0
35.0
36.0
37.0
37.1
37.2
37.3
38.0
39.0
40.0
41.0
42.0
43.0
44.0
45.0
46.0
46.1
47.0
48.0
48.1
49.0
50.0
County
Okanogan
Okanogan
Pacific
Pend Oreille
Pend Oreille
Pierce
Pierce
Pierce
San Juan
Skagit
Skagit
Skamania
Skamania
Snohomish
Snohomish
Spokane
Spokane
Spokane
Spokane
Spokane
Stevens
Stevens
Stevens
Thurston
Wahkiakum
Walla Walla
Whatcom
Whatcom
Whatcom
Whatcom
Whitman
Yakima
Ada
Adams
Bannock
Bear Lake
Benewah
Bingham
Blaine
Blaine
Boise
Bonner
Bonner
Bonnevil le
Boundary
Recreation Centroid
Alta Lake State Park
Osoyoos Lake State Park
Fort Canby
Skookum Lakes
Crawfield
Alder Lake







Mount Ranier National Park
Tolomerie State Park
Morgan State Park
Bayview State Park
Rockport State Park
Spirit Lake
Beacon Rock State Park
Wenberg State Park
Skyomish Park
Four Lakes
Newman Lake
Liberty Lake
Lake Williams
Long Lake
Wains Lake
Loon Lakes
Kettle Falls Recreation
Miller State Park
Cathlamet
Columbia State Park
Birch Bay State Park
Mount Baker
Colonial Bay
Ross Lake
Ross Lake
Rimrock Lake
Lucky Peak Reservoir
Oxbow Dam
Lava Hot Springs















Area












Bear Lake Recreation Area
Heyburn State Park
Blackfoot River
Sun Valley
Alturas Lake
Lowman
Lake Pend Oreille
Priest Lake
Palisades Reservoir
Copeland










                                                                      (continued)
 124

-------
TABLE A.2 (continued)

Recreation
Centroid
Number
51.0
52.0
53.0
54.0
55.0
55.1
56.0
57.0
58.0
58.1
59.0
59.1
60.0
61.0
62.0
63.0
64.0
64.1
64.2
64.3
65.0
66.0
67.0
68.0
69.0
70.0
71.0
72.0
73.0
74.0
75.0
76.0
76.1
77.0
78.0
79.0
80.0
81.0
81.1
82.0
82.1
83.0
84.0
85.0
86.0


County
Butte
Camas
Canyon
Caribou
Cassia
Cassia
Clark
Clearwater
Custer
Custer
Elmore
Elmore
Frankl in
Fremont
Gem
Gooding
Idaho
Idaho
Idaho
Idaho
Jefferson
Jerome
Kootenai
Latah
Lemhi
Lewi s
Lincol n
Madison
Minidoka
Nez Perce
Oneida
Owyhee
Owyhee
Payette
Power
Shoshone
Teton
Twin Falls
Twin Falls
Valley
Valley
Washington
Baker
Benton
Clackamas


Recreation Centroid
Craters Moon
Magic Reservoir
Lake Lowell
Blackfoot Reservoir
Lake Cleveland
Snake River
Sheridan Reservoir
Dworshak Reservoir
Mackay Reservoir
Stanley Basin Recreation Area
Anderson Ranch
Atlanta
Devil Creek Reservoir
Island Park Reservoir
Black Canyon Dam
Hagerman Valley
Corn Creek
Pittsburg Landing
Selway Falls
Powell Recreation Area
Snake River
Snake River
Fernan Lake
Deary Helmer Area
Yellow J. Lake
Winchester Lake
Richfield Area
Snake River
Snake River
Hells Gate
Daniels Reservoir
Mountain View Reservoir
Bruneau State Park
Payette
American Falls Reservoir
St. Joe River
Victor Area
Cedar Creek Reservoir
Snake River
Dagger Falls
McCall Lake
Brownlee Reservoir
Phillips Reservoir
River Park
Milo McLeur State Park





(continued)
125

-------
TABLE A.2 (continued)
Recreation
Centroid
Number
86.1
87.0
87.1
88.0
89.0
90.0
91.0
91.1
92.0
92.1
93.0
93.1
93.2
93.3
94.0
95.0
96.0
97.0
98.0
98.1
99.0
100.0
101.0
101.1
102.0
103.0
103.1
103.2
104.0
104.1
105.0
106.0
107.0
108.0
109.0
110.0
111.0
112.0
113.0
114.0
115.0
116.0
117.0
118.0
119.0
County
Clackamas
Clatsop
Clatsop
Columbia
Coos
Crook
Curry
Curry
Deschutes
Deschutes
Douglas
Douglas
Douglas
Douglas
Gil liam
Grant
Harney
Hood River
Jackson
Jackson
Jefferson
Josephine
Klamath
Klamath
Lake
Lane
Lane
Lane
Lincoln
Lincoln
Linn
Malheur
Marion
Morrow
Multnomah
Polk
Sherman
Til lamook
Umatil la
Union
Wai Iowa
Wasco
Washington
Wheeler
Yamhill
Recreation Centroid
Mount Hood Area
Ecola State Park
Fort Stevens State Park
Scaponia
Sunset Bay State Park
Prineville Res. State Park
Boardman State Park
Humbug Mountain State Park
Wickiup Reservoir
Tumalo
Winchester Bay
Diamond Lake
Wildlife Safari
Sutherlin
J. S. Burres State Park
Clyde Holiday State Park
Malheur Lake
Bonneville Dam
Howard Prairie
Lost Creek Area
Cove Palisades State Park
Indian Mary C. Park
Klamath Lake
Crater Lake
Goose Lake
Honeymoon State Park
MacKenzie Bridge
Fern Ridge Reservoir
Otter Crest
Devils Lake State Park
Foster Lake
Lake Owyhee State Park
Detroit Lake
Boardman Park
Rooster Rock
Independence
Deschutes River State Park
Tillamook Bay
Weston Area
Hilgard Junction State Park
Wai Iowa Lake
Memaloose State Park
Scoggins Reservoir
She! ton Wayside
Stewert Grenfeld State Park

                                                                      (continued)
126

-------
TABLE A. 2 (continued)

Recreation
Centroid
Number
120.0
120.1
121.0
121.1
121.2
121.3
121.4
125.0
126.0
127.0
127.1
126.1
129.0
130.0
131.0
132.0
County
Lake
Lake
Flathead
Flathead
Flathead
Flathead
Flathead
Lincol n
Missoula
Canada1
Canada1
Missoula
Deer Lodge
Meagher
Cascade
Park
Recreation Centroid
Flathead Lake (1)
Flathead Lake (2)
Flathead River (1)
Flathead River (2)
Hungry Horse Dam
Whitefish Lake
McGregor Lake
Lake Koocanusa
Lake Alva
Calgary Rec.2
Cranbrook Rec.
Missoula Rec.
Butte Rec.
Helena Rec.
Great Falls Rec.
Bozeman Rec.

  1These two recreation centroids  are  in  Canada.

  2The  recreation centroids  defined by  Rec.  reflect a  proxy for the  composite
 recreation sites close to  a  particular population  center.
                                                                              127

-------
                                    TABLE A.3




                 RECREATION ACTIVITY DAYS PRODUCED  BY CENTROID

Recreation
Centroid
Number
1.0
2.0
3.0
3.1
4.0
5.0
6.0
7.0
8.0
9.0
10.0
11.0
12.0
13.0
14.0
15.0
16.0
17.0
17.1
17.2
17.3
17.4
18.0
19.0
20.0
21.0
22.0
23.0
24.0
25.0
26.0
27.0
28.0
29.0
30.0
31.0
31.1
32.0
33.0
34.0
35.0
36.0
37.0
38.0
39.0


County
Adams
Asotin
Benton
Benton
Chellan
Clallem
Clark
Columbia
Cowlitz
Douglas
Ferry
Frank! in
Garfield
Grant
Grays Harbor
Island
Jefferson
King
King
King
King
King
Kitsap
Kittitas
Klickitat
Lewi s
Lincoln
Mason
Okanogan
Pacific
Pend Oreille
Pierce
San Juan
Skagit
Skamania
Snohomish
Snohomish
Spokane
Stevens
Thurston
Wahkiakum
Walla Walla
Whatcom
Whitman
Yakima
Total
Washington Activity
Swimming
987
1,299
3,399
5,316
3,710
4,134
14,721
341
6,460
1,666
391
2,678
290
3,560
5,293
3,247
1,266
95,739
4,352
3,264
4,352
1,088
11,242
1,816
1,204
4,289
820
2,409
2,358
1,413
593
36,682
658
4,698
554
17,027
8,771
26,374
2,079
9,884
289
3,524
8,205
2,595
13,006
327,962
Occasions
Camping
1,128
1,484
3,884
6,076
4,240
4,724
16,824
389
7,383
447
447
3,061
239
4,069
6,050
3,711
1,447
109,415
4,973
3,730
4,973
1,243
12,848
2,075
1,376
4,902
937
2,753
2,695
1,615
678
41,922
752
5,369
633
19,459
10,025
30,142
2,376
11,296
330
4,028
9,377
2,966
14,864
374,812
(in 100)
Fishing
671
882
2,309
3,611
2,521
2,808
10,001
231
4,389
266
266
1,819
142
2,419
3,596
2,206
860
65,042
2,956
2,217
2,956
739
7,637
1,233
818
2,914
557
1,636
1,602
960
493
24,920
447
3,992
376
11,567
9,959
17,917
1,412
6,715
196
2,394
5,574
1,763
8,836
222,801

Boating
661
870
2,276
3,561
2,485
2,769
9,861
228
4,327
262
262
1,794
140
2,385
3,546
2,175
848
64,129
2,915
2,186
2,915
729
7,530
1,216
807
2,873
549
1,613
1,580
947
397
24,571
441
3,147
371
11,412
5,879
17,666
1,393
6,621
193
2,361
5,496
1,738
8,712
219,691

128
                                                                     (continued)

-------
TABLE A.3 (continued)

Recreation
Number
40.0
41.0
42.0
43.0
44.0
45.0
46.0
47.0
48.0
49.0
50.0
51.0
52.0
53.0
54.0
55.0
56.0
57.0
58.0
59.0
60.0
61.0
62.0
63.0
64.0
65.0
66.0
67.0
68.0
69.0
70.0
71.0
72.0
73.0
74.0
75.0
76.0
77.0
78.0
79.0
80.0
81.0
82.0
83.0


County
Ada
Adams
Bannock
Bear Lake
Benewah
Bingham
Elaine
Boise
Bonner
Bonnevi 1 le
Boundary
Butte
Camas
Canyon
Caribou
Cassia
Clark
Clearwater
Custer
Elmore
Franklin
Fremont
Gem
Gooding
Idaho
Jefferson
Jerome
Kootenai
Latah
Lemhi
Lewis
Lincoln
Madison
Minidoka
Nez Perce
Oneida
Owyhee
Payette
Power
Shoshone
Teton
Twin Falls
Valley
Washington
Total
Idaho Activity
Swimming
4,143
73
1,412
141
181
714
232
64
500
1,434
152
69
2
1,771
185
423
19
225
72
417
169
206
247
239
313
287
311
1,313
571
162
91
73
330
391
793
66
147
336
144
436
58
1,194
130
186
20,422
Occasions
Camping
11,287
199
3,846
383
493
1,944
632
174
1,363
3,907
413
188
6
4,824
504
1,151
52
614
19
1,136
460
561
673
652
852
782
846
3,576
1,555
440
247
199
900
1,066
2,160
180
400
915
393
1,187
157
3,253
353
508
55,627
(in 100)
Fishing
9,972
176
3,398
339
436
1,717
558
154
1,204
3,452
365
166
5
4,262
445
1,017
46
543
173
1,003
406
496
594
576
753
691
747
3,159
1,374
489
218
176
795
942
1,908
159
354
809
347
1,048
139
2,874
312
449
49,146

Boating
3,172
56
1,081
108
139
546
178
49
383
1,098
116
53
2
1,356
142
323
15
173
55
319
129
158
189
183
239
220
238
1,006
437
124
69
56
253
300
607
51
112
257
110
333
44
914
99
143
15,634











(continued)
129

-------
TABLE A.3 (continued)

Recreation
Centroid
Number
84.0
85.0
86.0
86.1
87.0
88.0
89.0
89.1
90.0
91.0
91.1
92.0
93.0
94.0
95.0
96.0
97.0
98.0
98.1
99.0
100.0
101.0
102.0
103.0
104.0
104.1
105.0
106.0
106.1
107.0
107.1
108.0
109.0
110.0
111.0
112.0
113.0
114.0
115.0
116.0
117.0
118.0
119.0


County
Baker
Benton
Clackamas
Clackamas
Clatsop
Col umbia
Coos
Coos
Crook
Curry
Curry
Deschutes
Douglas
Gil 1 iam
Grant
Harney
Hood River
Jackson
Jackson
Jefferson
Josephine
Klamath
Lake
Lane
Lincoln
Lincoln
Linn
Malheur
Malheur
Marion
Marion
Morrow
Multnomah
Polk
Sherman
Til lamook
Umatil la
Union
Wai Iowa
Wasco
Washington
Wheeler
Yamhill
Total
Oregon Activity
Swimming
466
1,779
181
1,450
983
1,097
482
1,447
395
398
112
1,918
2,741
60
243
253
498
3,371
460
323
1,651
1,724
222
8,229
618
466
2,605
494
243
5,428
671
259
19,768
1,283
68
668
1,703
681
216
696
8,175
47
1,653
82,228
Occasions
Camping
948
3,618
12,571
2,949
2,000
2,232
981
2,943
804
810
229
3,991
5,575
122
494
515
1,012
6,857
935
547
3,359
3,507
451
16,738
1,257
948
5,299
1,005
495
11,040
1,365
527
40,207
2,609
139
1,359
3,465
1,384
440
1,416
16,628
96
3,361
167,248
(in 100)
Fishing
554
2,115
7,351
1,724
1,169
1,305
574
1,721
470
474
134
2,281
3,260
71
289
301
592
4,010
547
384
1,964
2,050
264
9,787
735
554
3,098
588
289
6,455
798
308
23,510
1,526
81
794
2,026
809
257
828
9,723
56
1,966
97,792

Boating
294
1,122
3,900
915
620
692
304
913
249
251
71
1,210
1,730
38
153
160
314
2,128
290
204
1,042
1,088
140
5,193
390
294
1,644
312
154
3,425
423
163
12,475
809
43
422
1,075
430
136
439
5,159
30
1,043
51,887

                                                                     (continued)
130

-------
TABLE A.3 (continued)

Recreation

Number
120.0
121.0
122.0
123.0
124.0
125.0
126/0
127.0
128.0

Western

County
Cascade
Flathead
Gal latin
Jefferson
Lake
Lewis and Clark
Lincol n
Missoula
Silver Bow
Total
Montana

Swimming
3,591
2,317
1,729
261
761
2,020
731
3,702
3,876
17,181
Activity Occasions

Camping
4,948
3,192
2,382
359
1,047
2,780
1,008
5,101
5,341
23,656
(in 100)

Fishing
3,798
2,450
1,829
276
804
2,130
774
3,916
4,100
18,160


Boating
2,439
1,574
1,175
177
516
1,370
497
2,515
2,633
11,663
                                        External Zones
129.0
130.0
131.0
132.0
133.0
134.0
135.0
136.0
137.0
138.0
139.0
140.0
141.0
142.0
Eastern Montana
Vancouver, B.C.
Cranbrook, B.C.
Calgary
Wyoming
Utah
Nevada
Cal i form' a
Alaska
Eastern Canada
North Central
Northeast
Southeast
South Central
4,400
2,481
275
1,300
120
2,106
220
15,603
64
1,456
1,298
595
323
1,132
5,200
1,684
187
900
198
3,465
479
24,736
105
971
2,137
980
531
1,864
4,500
1,172
130
650
68
1,183
164
8,436
36
652
729
335
181
637
2,700
3,432
381
1,906
96
1,699
236
12,156
52
1,906
1,045
479
260
914

    Note:   Missoula  county includes Mineral  and Granite  counties.   Silver  Bow
 county  includes:  Deer  Lodge,  Powell,  Beaverhead, and  Ravalli  counties.
                                                                              131

-------
                                    TABLE A.4

RECREATION  FACILITY VARIABLES,  EXISTING  AND  POTENTIAL,  FROM  IMPROVED  WATER
QUALITY BY RECREATION CENTROID

Recreation
Centroid
Number
1.0
2.0
3.0
4.0
4.1
5.0
5.1
5.2
6.0
7.0
8.0
8.1
9.0
10.0
11.0
12.0
13.0
13.1
13.2
14.0
14.1
14.2
15.0
15.1
16.0
16.1
16.2
17.0
17.1
17.2
18.0
19.0
19.1
20.0
21.0
22.0
22.1
22.2
23.0
23.1
23.2
Campsites
County Exist.
Adams
Asotin
Benton
Chelan
Chelan
ClaTlum
Cl all urn
Clallum
Clark
Columbia
Cowl itz
Cowl itz
Douglas
Ferry
Franklin
Garfield
Grant
Grant
Grant
Grays Harbor
Grays Harbor
Grays Harbor
Island
Island
Jefferson
Jefferson
Jefferson
King
King
King
Kitsap
Kittitas
Kittitas
Klickitat
Lewis
Lincoln
Lincoln
Lincoln
Mason
Mason
Mason
0
0
108
340
359
92
125
75
147
40
75
138
33
117
67
0
263
296
173
191
177
100
348
254
125
125
150
138
150
0
198
25
415
104
350
80
80
67
140
250
156
Pot.
0
35
108
340
359
102
125
75
147
40
75
138
130
167
67
0
292
425
173
191
177
113
348
254
155
125
150
179
241
0
198
66
415
104
350
80
80
67
140
250
156
Linear
Beach Feet
Exist.
0
2000
1850
200
870
1200
1100
1100
1085
0
0
0
100
400
1000
0
1000
2930
1000
0
450
270
0
600
2360
2000
3350
1925
3000
7000
1400
7000
4500
1325
1995
1300
1300
1400
240
400
240
Pot.
0
2231
2850
200
870
1200
1100
1100
1085
0
0
0
294
400
1000
0
1140
3124
1000
0
723
358
0
600
2658
2000
3350
2198
3601
7000
1400
7000
4500
1325
2383
1300
1300
1400
240
400
240
Boat Ramps
Exist.
3
7
17
4
4
4
5
4
14
3
14
4
3
4
9
0
15
2
5
8
10
8
10
14
8
2
6
11
12
39
16
4
9
12
14
2
1
1
5
5
5
Pot.
3
10
17
4
4
4
5
4
14
4
14
4
4
4
9
0
16
1
5
8
12
9
10
14
8
2
6
13
16
39
16
4
9
12
16
2
1
1
5
5
5
River Miles
Exist.
381
148
300
362
362
200
56
250
291
241
250
140
314
525
246
210
308
58
276
150
235
244
0
0
122
281
450
483
220
100
41
142
489
622
514
198
180
220
120
120
110
Pot.
381
198
503
370
370
200
87
250
291
365
250
140
452
745
246
215
350
100
276
150
300
263
0
0
180
281
450
542
350
100
41
200
489
706
598
198
180
220
120
120
110

                                                                     (continued)
132

-------
TABLE A.4 (continued)

Recreation
P a nt* vn "i r4
L»tr 1 1 L, I U I U
Number
24.0
24.1
24.2
24.3
25.0
26.0 "
26.1
27.0
27.1
27.2
28.0
29.0
29.1
30.0
31.0
31.1
32.0
32.1
32.2
32.3
32.4
33.0
33.1
33.2
34.0
35.0
36.0
37.0
37.1
37.2
37.3
38.0
39.0
40.0
41.0
42.0
43.0
44.0
45.0
46.0
46.1
47.0
48.0
48.1
Campsites

County
Okanogan
Okanogan
Okanogan
Okanogan
Pacific
Pend Oreil le
Pend Oreille
Pierce
Pierce
Pierce
San Juan
Skagit
Skagit
Skamania
Snohomish
Snohomish
Spokane
Spokane
Spokane
Spokane
Spokane
Stevens
Stevens
Stevens
Thurston
Wahkiakum
Walla Walla
Whatcom
Whatcom
Whatcom
Whatcom
Whitman
Yakima
Ada
Adams
Bannock
Bear Lake
Benewah
Bingham
Blaine
Blaine
Boise
Bonner
Bonner

Exist.
120
120
300
168
300
610
100
40
186
40
675
254
240
152
137
140
4
4
25
4
117
70
70
78
248
0
189
179
150
101
125
99
592
290
198
663
297
167
276
275
413
434
675
1265

Pot.
120
170
350
179
300
616
100
40
200
92
675
254
248
152
137
140
4
26
25
4
139
70
70
78
248
0
189
179
150
101
125
104
811
290
198
663
297
167
276
275
413
434
675
1365
Linear
Beach Feet

Exist.
275
1100
500
867
0
1450
400
900
900
1800
1030
300
0
500
1600
1715
150
40
100
0
40
300
300
300
849
0
700
800
1000
750
1150
800
5200
1960
400
0
300
1100
650
33
32
250
3830
6750

Pot.
275
1588
592
1028
0
1491
400
900
992
2145
1030
300
69
500
1674
1715
150
40
100
0
1040
300
300
300
1349
0
700
800
1000
750
1150
832
6640
1960
450
300
350
1100
750
85
32
250
3830
6800
Boat Ramps

Exist.
4
4
4
4
12
5
2
6
5
6
7
10
9
3
9
7
5
1
1
3
1
4
4
3
4
0
8
7
6
18
11
10
2
4
5
1
3
15
8
13
8
3
16
26

Pot.
4
7
4
5
12
5
2
6
6
8
7
10
9
3
9
7
5
2
1
3
2
4
4
3
12
0
8
7
6
18
11
10
11
4
6
1
4
15
11
14
8
3
16
28
River Miles

Exist.
800
241
80
265
281
241
40
100
178
45
7
260
285
275
334
340
61
94
50
75
119
219
250
100
174
114
263
40
413
180
60
639
728
79
345
266
241
159
130
0
407
504
152
310

Pot.
800
347
100
300
290
250
40
100
198
120
7
260
300
275
360
340
61
125
50
75
150
319
400
100
235
114
360
40
413
180
60
646
1041
79
373
290
283
159
240
104
407
509
152
325

(continued)









133

-------
TABLE A.4 (continued)

Recreation
Centroid
Number
49.0
50.0
51.0
52.0
53.0
54.0
55.0
55.1
56.0
57.0
58.0
58.1
59.0
59.1
60.0
61.0
62.0
63.0
64.0
64.1
64.2
64.3
65.0
66.0
67.0
68.0
69.0
70.0
71.0
72.0
73.0
74.0
75.0
76.0
76.1
77.0
78.0
79.0
80.0
81.0
81.1
82.0
82.1
83.0
Campsites
County
Bonnevil le
Boundary
Butte
Camas
Canyon
Caribou
Cassia
Cassia
Clark
Clearwater
Custer
Custer
Elmore
Elmore
Frankl in
Fremont
Gem
Gooding
Idaho
Idaho
Idaho
Idaho
Jefferson
Jerome
Kootenai
Latah
Lehmi
Lewis
Lincoln
Madison
Minidoka
Nez Perce
Oneida
Owyhee
Owyhee
Payette
Power
Shoshone
Teton
Twi n Fal 1 s
Twin Falls
Val 1 ey
Valley
Washington
Exist.
472
221
83
93
40
131
400
0
41
323
112
1013
182
180
340
458
47
324
15
200
200
80
22
95
1123
65
474
34
6
28
102
302
36
128
85
0
112
237
78
16
147
623
267
128
Pot.
472
221
83
93
40
131
400
0
41
423
162
1013
182
180
365
458
47
324
40
200
200
80
22
95
1123
65
524
34
6
28
102
302
36
128
85
0
112
237
78
16
147
623
267
128
Linear
Beach Feet
Exist.
100
525
0
50
150
0
0
0
0
155
431
2444
2340
260
100
0
725
0
250
500
25
100
100
0
5500
50
200
0
0
50
1465
1200
0
1100
0
0
0
0
0
0
0
9000
3350
150
Pot.
200
525
50
50
300
50
25
25
50
305
831
2444
2340
260
150
100
725
200
300
500
25
100
200
0
5700
50
400
0
0
100
1465
1200
50
1150
0
0
0
300
100
25
25
9000
3650
150
Boat Ramps
Exist.
60
8
0
1
4
10
1
7
0
9
1
4
7
1
5
8
5
3
1
0
0
0
0
1
82
1
5
3
0
0
4
10
2
4
6
0
10
0
1
4
5
12
5
0
Pot.
62
8
1
1
8
11
2
8
1
12
5
4
7
1
7
10
5
4
1
0
0
0
1
1
85
1
9
3
1
1
4
10
3
5
6
0
11
2
2
5
6
12
9
1
River Miles
Exist.
177
457
66
408
65
533
316
140
153
534
76
0
288
311
114
238
76
111
470
550
555
545
16
43
236
292
1003
150
31
30
31
243
58
174
149
78
104
682
51
107
83
374
320
179
Pot.
275
457
170
437
90
594
389
200
200
577
425
50
300
311
149
353
76
116
552
550
555
545
71
43
262
292
1053
150
91
58
31
243
90
574
174
78
124
716
82
167
150
374
375
219

                                                                     (continued)
134

-------
TABLE A.4 (continued)

Recreation
r*Q nt Y*n "1 H
VsC 1 1 1* [ \J IU
Number
84.0
85.0
86.0
86.1
87.0
87.1
88.0
89.0
90.0
91.0
91.1
92.0
92.1
93.0
93.1
93.2
93.3
94.0
95.0
96.0
97.0
98.0
98.1
99.0
100.0
101.0
101.1
102.0
103.0
103.1
103.2
104.0
104.1
105.0
106.0
107.0
108.0
109.0
110.0
111.0
112.0
112.1
113.0
114.0
Campsites

County
Baker
Benton
Clackamas
Clackamas
Clatsop
Clatsop
Columbia
Coos
Crook
Curry
Curry
Deschutes
Deschutes
Douglas
Douglas
Douglas
Douglas
Gi 1 1 iam
Grant
Harney
Hood River
Jackson
Jackson
Jefferson
Josephine
Klamath
Klamath
Lake
Lane
Lane
Lane
Lincoln
Lincoln
Linn
Malheur
Marion
Morrow
Multnomah
Polk
Sherman
Til lamook
Til lamook
Umatil la
Union

Exist.
488
37
755
495
520
520
71
1525
221
1236
281
750
776
500
500
674
500
20
273
321
427
650
693
1750
1087
630
633
324
658
1200
200
800
538
861
273
1628
156
184
16
132
1100
1088
273
183

Pot.
488
37
830
495
520
520
71
1625
421
1236
281
750
916
584
585
874
700
20
273
371
447
650
693
1790
1087
730
633
324
788
1330
348
920
568
886
273
1628
156
184
16
132
1100
1088
273
183
Linear
Beach Feet

Exist.
8300
0
900
1000
11415
11415
0
2000
300
7750
3000
40
35
1300
1300
1450
1300
0
0
0
1280
700
700
4900
0
7080
7083
0
5000
5000
6608
150
50
4300
4725
7510
0
10218
1350
0
18000
18960
1300
10

Pot.
8300
940
1300
1000
11415
11415
0
2075
450
7750
3000
40
435
2100
2100
2210
2100
0
0
0
1280
700
700
5200
0
7800
7083
0
5400
5400
6938
150
50
5530
4725
7510
0
10218
2140
0
18000
18960
1300
10
Boat Ramps

Exist.
14
2
3
3
9
9
1
23
3
13
1
6
6
13
13
12
13
3
12
8
7
20
19
31
30
22
22
11
33
33
33
7
20
26
7
14
2
21
11
4
18
18
14
3

Pot.
14
2
4
3
9
9
1
25
5
13
1
6
7
16
15
15
15
3
12
8
7
20
19
32
30
25
22
11
36
36
35
7
20
29
7
14
2
23
14
4
18
18
14
4
River Miles

Exist.
827
155
375
295
165
169
308
131
552
200
206
169
85
150
670
25
70
382
1077
804
116
375
400
409
451
340
264
355
170
673
75
127
50
624
1897
460
460
130
172
282
175
174
752
524

Pot.
827
249
425
295
165
169
308
507
642
200
206
169
140
175
700
156
200
382
1120
1238
194
375
400
432
451
340
336
389
22
700
225
222
100
747
1935
493
493
147
251
282
175
177
842
535

(continued)









135

-------
TABLE A.4 (continued)

Linear
Recreation
Centroid
Number
115.0
116.0
117.0
118.0
119.0


County
Wai Iowa
Wasco
Washington
Wheeler
Yamhill
Camp si


Exist.
522
590
67
80
52
tes


Pot.
522
590
77
80
127
Beach


Exist.
400
2000
1200
0
0
Feet


Pot.
400
2000
1720
0
500
Boat


Exist.
4
8
2
0
0
Ramps


Pot.
4
8
4
0
2
River


Exist.
998
805
204
375
198
Miles


Pot.
998
805
256
375
290
            County
Campsites
Beach Feet    Boat Ramps    River Miles
120.0
120.1
121.0
121.1
121.2
121.3
121.4
125.0
126.0
127.0
128.0
126.1
129.0
130.0
131.0
132.0
Lake
Lake
Flathead
Flathead
Flathead
Flathead
Flathead
Lincoln
Missoula
Canada1
Canada1
Missoula
Bear Lodge
Meagher
Cascade
Park
48
96
1
1
175
10
20
30
40
500
500
480
876
482
159
750
1100
1100
1
1
100
600
300
900
300
350
3300
350
4200
4500
900
900
7
8
4
3
7
4
4
3
3
4
4
13
19
9
4
14
55
55
96
58
72
16
120
50
50
100
100
70
50
64
40
120

     Notes:   Exist,  means currently  existing.   Pot.  means  potential;  that is,
the  potential  facilities  (or  river  miles)  that  could  be constructed  if all
degraded  rivers were  improved  so as to be  acceptable for recreation purposes.

 1These  recreation  centroids  are  near  Calgary  and  Cranbrook,  respectively.
136

-------
                                    TABLE A.5

ANNUAL (1980)  RECREATION  VALUE  BY ACTIVITY AND BY COUNTY FOR WASHINGTON,  IDAHO,
AND OREGON

Zone
Number
1
2
3
4
6
9
10
11
13
14
15
16
17
20

23
25

29
31
32
34

35
36
39
42
46
47
49
52
53
55
57

59
64
67
68
69
70
74
75


County
Name Swimming
Adams $418,732
Asotin 317,286
Benton 694,835
Chelan 2,215,635
Clallum 2,792,041
Clark 5,219,251
Columbia 189,053
Cowlitz 2,057,886
Douglas 520,114
Ferry 572,054
Franklin 606,553
Garfield 143,368
Grant 2,010,921
ura^S 2,601,004
Harbor '
Island 5,103,889
Jefferson
4,309,672
King 49,635,416
Kitsap 2,808,435
Kittitas 6,747,085
Klickitat
1,334,830
Lewis 4.582,449
Lincoln 3,458,897
Mason 14,629,231
Okanogan 4,482,324
Pacific 359,415
J!end,-. 1,237,846
Oreille ' '
Pierce 18,545,658
San Juan 1,404,485
Skagit 4,982,385
Skamania 4,175,050
Snohomish
16,375,406
Spokane 13,261,349
Stevens 3,995,138
Thurston 7,317,709
Wahkiakum 968,438
Walla 10? 2Qg
Walla '
Whatcom 3,660,551
Whitman 1,506,345
Yakima 1,784,770
State $198,134,715
Total
Recreation Val
Camping
$124,046
267,561
744,426
4,477,158
2,105,443
4,166,888
436,056
5,256,482
859,498
1,043,430
582,571
59,728
3,740,331
5,089,433

9,601,347

3,615,155
16,920,861
2,637,427
6,415,688

1,305,933
5,351,179
2,903,928
13,323,855
5,776,672
2,190,611
2,958,325
9,979,171
3,119,350
6,578,156
4,068,523

9,365,677
6,895,249
3,900,172
6,468,939
231,069
1,368,908
4,476,638
1,542,352
3,972,679
$163,910,915

ue, Washington (in dollars)
Fishing
$825,649
310,729
496,187
1,539,314
1,457,025
2,644,397
489,887
3,308,244
484,039
542,865
500,445
409,595
1,312,042
2,211,118

2,400,678

1,996,718
14,558,351
1,087,164
2,571,265

1,020,827
2,044,047
2,054,917
6,965,910
2,846,955
744,379
925,770
1,904,041
532,618
2,953,124
2,883,744

5,086,977
8,856,214
2,976,602
2,872,549
1,598,016
765,048
2,014,450
1,111,546
838,726
$94,762,172

Boating
$674,596
204,062
564,788
1,609,433
1,637,269
6,081,738
242,627
5,319,961
221,079
322,250
439,460
81,671
1,122,036
4,254,056

13,171,954

2,964,559
115,215,125
3,859,918
5,671,063

1,020,862
6,991,453
906,117
18,191,191
1,566,587
1,141,501
639,492
22,029,498
1,127,575
7,846,645
2,242,283

19,330,802
10,521,165
2,991,068
11,579,364
1,596,378
763,027
3,589,988
1,724,077
1,605,125
$281,060,843

Total
$2,043,023
1,099,638
2,500,236
9,841,540
7,991,778
18,112,274
1,357,623
15,942,573
2,084,730
2,480,599
2,129,029
696,362
8,185,330
14,155,611

30,277,868

12,885,104
195,219,753
10,392,944
21,415,101

4,682,452
18,969,128
9,323,859
53,110,187
14,672,538
4,425,906
5,761,433
57,458,368
6,184,028
22,360,310
13,369,600

50,878,862
39,533,977
13,862,980
28,238,561
4,393,901
4,004,192
13,741,627
7,884,320
8,201,300
$737,868,645












(conti nued)
137

-------
TABLE A.5 (continued)


Zone
Number
76
77
78
79
80
81
82
84
85
87
88
89
90
91
92
93
95
96
97
99
101
102
103
104
105
109
110
111
112
113
114
115
116
117
118
119
120
122
123
124
125
126
128
130



Recreation Value, Idaho
County 	
Name Swimming
Ada $643,736
Adams
Bannock
Bear Lake
Benewah 1
Bingham
Blaine
Boise
Bonner 2
Bonnevil le
Boundary
Butte
Camas
Canyon
Caribou
Cassia
Clark
Clearwater
Custer
Elmore
Franklin
Fremont
Gem
Gooding
Idaho
Jefferson
Jerome
Kootenai 4
Latah
Lemhi
Lewi s
Lincoln
Madison
Minidoka
Nez Perce
Oneida
Owyhee
Payette
Power
Shoshone
Teton
Twin Falls
Valley
Washington
State $14
Total
40,860
223,012
95,777
,041,242
221,638
55,173
73,166
,211,064
102,547
376,725
42,735
71,696
386,587
26,197
224,741
27,653
167,255
103,848
317,244
102,925
20,087
334,305
101,542
277,789
300,407
112,146
,660,328
117,772
8,218
89,253
51,082
251,059
422,061
546,957
78,552
104,255
108,783
148,075
129,753
39,476
138,367
168,998
129,365
,849,336

Camping
$1,348,072
200,200
1,354,811
544,829
1,664,328
727,532
537,698
488,164
6,151,289
664,612
983,203
254,790
275,700
472,809
238,547
1,392,720
136,727
709,087
606,000
893,402
621,482
347,379
384,406
1,057,230
824,850
298,537
610,502
9,306,606
326,918
107,438
291,485
88,233
324,523
701,064
1,288,813
321,227
558,574
52,761
715,241
1,123,893
225,307
806,319
686,855
389,560
$41,447,118

Fishing
$1,279,777
150,935
1,153,465
603,709
753,164
1,025,614
337,267
321,526
1,274,625
688,426
430,045
493,668
461,566
1,106,917
440,539
2,424,434
378,111
304,967
388,627
1,045,508
663,081
308,287
817,416
951,653
685,628
1,326,688
979,517
1,929,019
230,153
74,139
325,064
494,576
1,359,088
1,174,084
537,271
959,115
844,756
758,055
1,599,766
409,826
517,499
1,410,276
399,642
455,427
$34,386,352

(in dollars)
Boating
$92,765
12,723
11,566
8,307
1,533,043
46,766
22,141
13,290
2,691,692
77,149
317,660
6,725
5,968
83,869
11,235
87,126
4,247
143,045
9,302
45,242
15,229
8,602
70,295
36,584
53,443
21,989
19,771
18,634,145
40,118
1,835
99,814
8,907
20,510
51,953
424,070
14,031
43,942
23,963
90,236
102,220
4,708
61,442
44,877
13,925
$25,225,203


Total
$3,354,350
404,718
2,742,854
1,252,622
5,001,777
2,021,550
952,279
896,146
12,328,670
1,532,734
2,107,633
797,918
814,930
2,050,182
716,518
4,230,132
546,738
1,324,354
1,107,777
2,301,396
1,402,617
684,355
1,606,422
2,147,009
1,952,710
1,987,621
1,721,936
34,530,098
714,961
191,630
805,616
640,798
1,955,180
2,349,162
2,797,111
1,372,925
1,551,527
943,562
2,553,329
1,765,692
786,990
2,416,404
1,300,372
988,277
$115,908,009

138
                                                                     (continued)

-------
TABLE A.5 (continued)

Zone
Number
131
132
133

135
137
138
139
140
142
144
148
149
150
151

152
154
155
156
158
159
162
164
165
166
167
168

169
170
171

173
174
175
176
177

178
179
State
Region


Name Swimming
Baker $141,457
Benton 827,292
Clackamas
4,772,207
Clatsop 3,353,411
Columbia 912,365
Coos 175,528
Crook 110,386
Curry 172,808
Deschutes 480,344
Douglas 1,846,194
Gilliam 164,645
Grant 35,176
Harney 9,036
Hood
". 3,107,820
River ' '
Jackson 598,764
Jefferson 521,123
Josephine 95,934
Klamath 611,989
Lake 82,309
Lane 3,471,636
Lincoln 1,411,179
Linn 1,851,516
Malheur 166,984
Marion 1,493,100
Morrow 130,070
Multnomah
8,050,763
Polk 3,316,113
Sherman 350,726
Til lamook
4,017,140
Umatilla 365,126
Union 101,271
Wallowa 50,938
Wasco 1,634,881
Washington
3,145,591
Wheeler 53,209
Yamhill 2,176,145
Total $51,438,145
Total $265,008,995
Recreation

Camping
$561,255
1,252,484

10,898,672
5,539,193
1,957,309
993,108
536,669
539,296
3,634,461
5,250,715
333,351
503,063
194,990

5,045,580
2,021,986
3,194,512
1,148,556
1,782,220
909,666
7,317,266
6,308,622
4,341,859
355,532
5,265,461
808,591

4,697,102
906,166
1,342,707

8,290,208
828,568
540,853
384,261
3,517,436

2,003,162
341,281
2,255,131
$95,891,547
$300,057,296
Value, Oregon

Fishing
$200,539
1,667,586

3,131,003
1,553,691
1,685,883
197,613
216,667
211,894
804,946
1,662,190
451,784
205,665
81,409

1,541,053
731,808
438,121
345,897
568,538
352,077
2,249,034
1,683,258
1,218,805
473,956
1,001,303
427,914

2,784,334
1,890,496
802,698

1,809,872
373,757
293,411
114,082
1,166,457

1,726,816
221,735
1,602,844
$36,093,970
$165,669,098
(in dollars)

Boating
$63,442
5,599,918

1,663,320
1,401,502
570,635
79,706
43,073
14,001
324,188
805,572
175,736
78,217
8,944

1,836,999
223,571
552,449
140,261
213,087
114,146
2,723,584
1,607,926
1,492,862
26,483
817,479
110,449

7,293,007
2,241,450
444,574

2,105,897
279,471
74,768
21,527
857,965

1,160,209
28,802
588,034
$30,592,428
$336,035,293


Total
$966,693
4,307,280

20,465,202
11,847,797
5,126,192
1,445,955
906,795
937,999
5,243,939
9,564,671
1,125,516
822,121
294,379

11,641,452
3,576,129
4,706,205
1,740,648
3,175,834
1,458,198
14,761,480
11,110,985
8,905,042
1,134,066
8,677,343
1,481,024

22,825,206
8,374,225
2,940,705

16,223,117
1,846,922
1,010,303
570,808
7,176,739

8,036,778
645,027
6,622,254
$214,016,090
$1,066,770,682
                                                                              139

-------
                                   APPENDIX B

                         HOUSEHOLD SURVEY QUESTIONNAIRE



     This appendix contains the questionnaire used by the Survey Research Center

at  Oregon  State.   The  telephone  survey  included  3,000  households  and  was

conducted  in  the Fall of  1980.   Columns  1-4 on the code sheets  are  household

identification numbers;  columns  5-8 are  card numbers;  and column 9 is  a state

verification number.   The responses to question one were coded in columns 10-11.


                        OREGON OUTDOOR RECREATION SURVEY
1.
2.
2a.
10-11 Number
99 DK, NA
12-13 Number
99 DK, NA
14-15 Number
99 DK, NA
During the past 12 months, how many persons,
including yourself, have lived in your household?
How many of these people are 18 years or older?
And, how many are under 18 years of age? (INT:
RESPONSE TO Q. 2 AND 2a MUST EQUAL TOTAL IN Q. 1)
3.   I'd  like  to complete  picture  of your household.  Some  of  these questions
     concern  each person,  while others  are  about your  household as  a  group.
     Thinking  about  everyone  who lived  in  your  household  during the past  12
     months,  I  would like to  list  each  person from the oldest  to the youngest
     just to  make sure  we are talking about everyone.  (INT:   STARTING WITH THE
     OLDEST,  GET ALL INFORMATION AND ENTER ON FIRST LINE.   CONTINUE WITH EACH
     FAMILY MEMBER DOWN TO THE YOUNGEST.)

                                                   Sex (Circle)       Age
Person 1
Person 2
Person 3
Person 4
Person 5
140
Relationship to "R" First Name
16
19
22
25
28

Male
1
1
1
1
1
Female
2
2
2
2
2
                                                                     17-18
                                                                     20-21
                                                                     23-24

                                                                     26-27

                                                                     29-30

-------
Person 6
Person 7
Person 8
Person 9
Person 10
Person 11
Person 12

31
34
37
40
43
46


1
1
1
1
1
1
1
2
2
2
2
2
2
2
32-33
35-36
38-39
41-42
44-45
47-48


Now I'd like to ask you some questions about your household's outdoor recreation
activities for the past 12 months.

4.    Thinking back to the first of June 1980 to the present,  how many trips,  all
     together, did you or any member of your household take for these four kinds
     of outdoor recreation:   swimming in a lake or  river, boating,  fishing,  or
     camping?
       49-51
   Number of trips
                                   99  DK, NA
     (INT:   IF "NONE," WRITE 0 AND SKIP TO Q.  7)

The next series of questions refers only to the last trip you or someone in your
household took.
5.   $
52-56
     99 DK  (SKIP TO Q.  6)
/day     First, how much was the daily use fee,  if
         any, for the recreation facilities used?
         (INT:   IF NONE, WRITE 0 AND SKIP TO Q.  6)
5a.  $
57-61
     99 DK
/day     What is the maximum daily use fee you would
         be willing to pay for this recreation
         facility rather than forego using it?
6.   $
52-56
     999 DK
         About how much money did you spend
         travelling to and from your home to the
         recreation area on this last trip?  This
         includes meals, gas, oil, car rental or air
         fare, and so forth.   (Just your best
         estimate please.)
6a.  1  Enjoyed travel time

     2  Prefer to shorten
 66  9  DK
                      Some people feel time spent travelling to a
                      recreation site is an inconvenience, while
                      others enjoy it.  How about you?  Did you
                      enjoy the time spent travelling on this
                      trip, or would you rather have shortened the
                      travel time?
                                                                              141

-------
6b.   $
  67-70
About how much money would you be willing to
pay to shorten the total travel time for
this last trip by half?
(ASK OF EVERYONE)

7.        71-73
     Number of trips
     99  DK
                        Now,  thinking back to the first of September
                        of last year to the first of June 1980,  how
                        many  trips,  all together, did you or any
                        member of your houehold take for recreation
                        purposes?  (INT:   IF NONE,  WRITE 0 AND SKIP
                        TO Q.  8)
Finally, for statistical purposes  only,  we have a few last questions about your
household.
8.
     Town or City
     999  Refused
                        First,  in or near which town or city is  your
                        home located?
9.
70-76
     County

     99  Refused; DK
And, in which
your home located?
10.  01  Less than $10,000
     02  $10,000 to $14,999
     03  $14,000 to $19,999
     04  $20,000 to $24,999
     05  $25,000 to $34,999
     06  $35,000 to $40,000
     07  over $40,000
     99  Refused; DK
                        Would you please tell  me if the total  gross
                        income for your household in 1979 was  ...
                        (READ LIST)
11.  Is there anything else you would like to say about outdoor recreation?
                       (THANK YOU FOR YOUR COOPERATION)
142

-------
                                   TABLE  B.I

   FREQUENCY DISTRIBUTION OF  RECREATION TRIPS  USING  1980  HOUSEHOLD  SURVEY  DATA


                                    Oregon
Days
1
2
>2
Total
Proportion
Number of
Trips
273
130
100
403
Number of
Days
273
260
694
1227
Swimming
414
133
283
830
0.206
Boating
143
98
283
524
0.130
Fishing
182
320
484
986
0.245
Camping
10
485
1194
1689
0.419
                                      Idaho
Days
1
2
>2
Total
Proportion
Number of
Trips
262
89
144
495
Number
Days
262
178
646
1086
of
Swimming
218
48
338
604
0.415
Boating
111
44
305
460
0.111
Fishing
576
247
630
1453
0.349
Camping
4
350
1290
1644
0.395
Washington
Days
1
2
>2
Total
Proportion
Number of
Trips
479
113
177
769
Number
Days
479
226
1181
1886
of
Swimming
748
250
1278
2476
0.315
Boating
470
211
982
1663
0.211
Fishing
398
337
952
1687
0.214
Camping
12
502
1528
2042
0.260

     Note:   The above estimates  are  based on a subsample of 313 households  (123
from Washington,  100 from  Oregon,  and  90  from  Idaho),  but  a total of  1767
recreation  trips.
                                                                             143

-------
                               ACKNOWLEDGEMENTS

     The   research   for   this   report  was  conducted  at  the   Environmental
 Protection Agency,  Corvallis Environmental Research Laboratory.   I  am  grateful
 to   my   EPA   colleagues   John   Jaksch  and  Neils  Christiansen  for   several
 discussions  during  the   developmental  stage of  the  model.   This  work  also
 benefited  from discussions with  several   recreation  planners in  the  Pacific
 Northwest  at Federal and  state agencies.   I  am  grateful to  Richard  Walsh,  John
 Loomis, Russell Gum, Louise Arthur, Richard Adams, and John Duffield for their
 review  of the manuscript.  I  also  thank Jack  Gakstatter,  the  EPA  Project
 Officer for his assistance during the final two  years  of the study.
144

-------
                                  REFERENCES



Abramowitz, Milton  and  Irene Stegun, Handbook  of  Mathematical  Functions (Dover



     Publication Inc., New York, 1965).



Anderson,  James  E. ,  "A  Theoretical  Foundation  for  the Gravity  Equation,"  The



     American Economic Review 69 (1), 106-116 (March 1979).



Bishop, John  and  Charles Cicchetti,  "Some Institutional and Conceptual Thoughts



     on the Measurement of Indirect and Intangible Benefits and Costs," in Henry



     M  Perkin and  Eugene M.  Seskin  (eds.),   Cost  Benefit Analysis  and Water



     Pollution Policy (The Urban Institute, Washington, D.C., 1973).



Brown,  William  B.   and  Farid W.  Newas,  "Impact  of Aggregation  on Estimation of



     Outdoor  Recreation  Demand  Functions,"  American  Journal  p_f_  Agricultural



     Economics 55, 246-249 (1973).



Bureau  of  Public Roads,  Calibrating and Testing a Gravity  Model  for Any Sized



     Urban Area (U.S. Department of Commerce, Washington, D.C., 1965).



Burt,  0. R.,  "Comments  on 'Recreation Benefits  from Water Pollution Control' by



     Joe B.  Stevens," Water  Resources  Research   6  (4),  905-907  (August 1967).



Burt,  0.   R.  and D.  Brewer,  "Evaluation of  Net  Social Benefits  from  Outdoor



     Recreation," Econometrica 39, 813-827 (September 1971).



Carter,  Nancy,  "Predicting  Unit  Variate Values in  a Finite Population," Ph.D.




     Thesis, Oregon State University, 1981.



Cesario, Frank  J. ,  "A Generalized Trip Distribution  Model," Journal  of Regional



     Science 13, 233-248  (1973).



Cesario, Frank J.,  "More  on  the Generalized Trip Distribution Model,"  Journal  of



     Regional Science 14, 389-397 (1973).
                                                                              145

-------
Cesario, Frank J. ,  "A  New Method for Analyzing  Outdoor Recreation Trips Data,"




     Journal of Leisure Research 7,  200-215 (1975).



Cesario, Frank J., "Value of Time in Recreation Benefit Studies," Land Economics




     52, 32-41 (1976).



Cesario, F.  J.  and J.   L.  Knetsch,  "A Recreation Site  Demand  and Benefit Esti-




     mation Model," Regional Studies 10 (1), 97-104 (1976).




Cheung, H.  K., "A Day-Use Park Visitation Model," Journal £f Leisure Research 4,




     139-156 (1972).



Cicchetti,   Charles  J. ,  Forecasting  Recreation  j_n the  United  States (Lexington



     Books, 1973).



Cicchetti,  C. J. ,  A.  C. Fisher and V. K.  Smith, "An Econometric Evaluation of a




     Generalized  Consumer  Surplus  Measure:    The  Mineral  King  Controversy,"




     Econometrica 44, 1259-1276 (November 1976).



Cicchetti,  Charles J.,  Joseph J. Seneca and Paul Davidson, The Demand and Supply



     of  Outdoor  Recreation:  An  Econometric  Analysis  (Rutgers  University,  New



     Brunswick, N.J., 1969).




Clawson, Marion,  Methods  of Measuring the Demand for  Values  of Outdoor Recrea-




     tion,   Reprint No.  10  (Resources for  the  Future, Inc.,  Washington,  D.C.,



     1959).




Clawson, Marion  and Jack  Knetsch,  Economics of Outdoor  Recreation  (The Johns



     Hopkins University Press, Baltimore, 1966).




Common,  M.   S. ,  "A Note on  the  Use  of the Clawson Method  for  the Evaluation of



     Recreation Site Benefits," Regional Studies 7, 401-406 (1973).




Currie,  John  A.  Murphy  and Andrew Schmitz, "The Concept of Economic Surplus and




     Its Use  in  Economic Analysis," The Economic Journal  81,  741-799 (December




     1971).
 146

-------
Davidson,  Paul  F. ,  Gerand  Adams  and Joseph Seneca," The  Social  Value of Water



     Recreational  Facilities  Resulting  from  an  Improvement  in  Water Quality:



     The  Delaware  Estuary,"  in  Water Research, Allan V.  Kneese  and Stephen C.



     Smith  (eds. )  (The  Johns Hopkins  University  Press,  Baltimore,  1966),  pp.



     175-211.




Davis,  Phillip  J.   and   Phillip  Rabinowitz,  Numerical   Integration (Blaisdell



     Publishing Co., Waltham,  Massachusetts, 1967).



Dickey,  John W. , Metropolitan Transportation  Planning  (McGraw-Hill,  New York,



     1975).



Dwyer,  John F. , John  R.  Kelly  and  Michael  D.  Bowes,   Improved  Procedures  for



     Valuation £f  the Contribution  o_f  Recreation  to National  Economic Develop-



     ment  (University of  Illinois,  Water Resources Center,  Urbana,  Illinois,



     September 1977).



Ellis,  Jack B.  and  Carlton S. Van  Doren,  "A  Comparative Evaluation of Gravity



     and  System Theory Model  for  Statewide  Recreation Traffic  Flows," Journal o_f



     Regional  Science 6,  57-70 (1966).



Ewing,  Gordon  0. ,  "Progress  and  Problems  in  the  Development  of  Recreation Trip



     Generation  and  Trip Distribution Models," Leisure  Sciences  3_(1),  1-23



      (1980).



Freeman,  A.  Myrick,  III, The Benefits  of Air and Water Pollution Control:   A



      Review  and Synthesis  of Recent  Estimates,  prepared  for  the Council  of




      Environmental Quality.  December 1979.



Freeman,  A.  Myrick,  III, The  Benefits  of  Environmental  Improvement:  Theory  and



      Practice  (Johns Hopkins University Press,  Baltimore,  1979).



Freund,  R.  J.  and  R.  R.  Wilson,  "An  Example of a Gravity  Model to  Estimate



      Recreation  Travel," Journal of Leisure  Research 6,  241-256 (Summer 1974).
                                                                              147

-------
Goldfeld,  Stephen  M. ,  "The Demand  for Money  Revisited," Brookings  Papers on



     Economic Activity 3, 577-646 (1973).



Gordon,  Irene  M.  and Jack L.  Knetsch, "Consumer's  Surplus Measures  and the



     Evaluaion of Resources," Land Economics ^5, 1-10 (February 1979).



Harberger, Arnold C., "Three Basic Postulates for Applied Welfare Economics:  An



     Interpretive Essay," Journal  of Economic Literature IX,  785-797 (September



     1971).



Hay,  Michael  J. ,  and Kenneth  E.  McConnell, "An  Analysis of  Participation in



     Nonconsumptive  Wildlife  Recreation,"   Land  Economics  55,  460-471 (November



     1979).



Hicks, J. R., Value  and Capital, Second  Edition (Oxford University Press, 1939).



Hotelling,  Harold, "The  General Welfare in Relation to Problems of Taxation and



     of  Railway and  Utility Rates," Econometrica 6, 242-269 (1938); reprinted in



     Readings j_n Welfare Economics (Irwin,  Homewood, Illinois, 1969).



Hutchinson, B.  G. ,  Principles  of Urban  Transport  Systems  Planning (Washington,



     D.C.,  Scripta Book Co., 1974).



Institute   of  Transportation  and  Traffic   Engineering,  Transportion  Analysis



     Procedures  for  National  Forest  Planning   (University  of  California,



     Berkeley, 1971).



Isard,  Walter,   Methods  of Regional  Analysis:   An  Introduction to_  Regional



     Science (The MIT Press, Cambridge,  1960).



Knetsch, Jack L. ,  "Outdoor  Recreation Demands and Benefits,"  Land Economics 39,



     387-396 (November 1963).



Knetsch, Jack  L.,  "Displaced  Facilities and Benefit  Calculations,"  Land  Econ-



     omics 53 (1),  123-129  (February 1977).



Knetsch,  Jack  L. ,  Outdoor   Recreation   and  Water   Resources  Planning (American



     Geophysical Union, Washington, D.C., 1974).





148

-------
Knetsch, Jack  L. ,  R.  E.  Brown  and  W.  J. Hansen,  "Estimating Expected Use and



     Value  of  Recreation  Sites," in  Planning  for Tourism,  Development, Quanti-



     tative  Approaches.  C.  Bearing,  W.  Swart  and  T.   Var  (eds. )  (Praeger



     Publishers, New York, 1976).




Krutilla,  J.  V. and A.  C.  Fisher.   The  Economics  of Natural  Environments (The



     Johns Hopkins University Press, Baltimore, 1975).



Laidler,  David W.  E. , The  Demand  for  Money:   Theories  and  Evidence,  Second



     Edition (Dun-Donnelly, New York, 1977).



Ma'ler,  Karl-Goran,  Environmental Economics:   A Theoretical  Inquiry  (The Johns



     Hopkins University Press, Baltimore, 1974).



McAllister,  Donald  M.   and  Frank Klett,  "A  Modified Gravity  Model  of Regional



     Recreation Activity  with  an Application to  Ski  Trips,"  Journal  erf Leisure



     Research 8 (1) 21-34 (1976).



McConnell,  Kenneth   E. ,  "Some  Problems  in  Estimating the Demand for Outdoor



     Recreation,"  American Journal  of  Agricultural  Economics  ^7  (2), 330-334



     (May  1975).



Mishan,  E.  J. ,  Cost-Benefit Analysis,  Second Edition (Praeger,  New York, 1976).



Mohring, Herbert, "Alternative Welfare  Gain and Loss  Measures,"  Western Economic



     Journal 9  (4), 349-368 (December 1971).



Niedercorn,  J.  H.   and  B.  V.   Bechdolt, Jr. ,  "An  Economic  Deviation  of the



     'Gravity  Law'  of  Spatial  Interaction," Journal  of Regional Science 9 (2),




     273-282 (1969).



Reiling,  S.  D. ,  K.  C.  Gibbs  and  H.  H.  Stoevener,  Economic  Benefits from  art



     Improvement in  Water Quality.  Environmental  Protection  Agency,  Washington,



     D.C.(January 1973).
                                                                             149

-------
Rowe, Robert D. ,  Ralph  C.  d'Arge and Davis S.  Brookshire, "An Experiment on the




     Economic  Value  of Visibility,"  Journal  of  Environmental   Economics  and




     Management 7, 1-19 (1980).



Seneca, Joseph  J. ,  Paul  Davidson,  and F.  Gerard  Adams,  "An Analysis of Recrea-




     tion Use of the TVA Lakes," Land Economics 44 (4), 529-534 (November 1968).




Silberberg,  Eugene,  "Duality  and  the   Many  Consumer's  Surpluses,"  American




     Economic Review 62 (5), 942-952 (December 1972).



Smith,  V.   Kerry,   "Travel  Cost  Demand  Models   for  Wilderness  Recreation:   A




     Problem  of Non-Nested  Hypotheses,"  Land  Economics  51 (2),  103-111  (May




     1975).



Smith, V.  Kerry and Charles J.  Cicchetti, "Regression Analysis with Dichotomous



     Dependent  Variables,"  presented at Econometric  Society Meetings,  1972.




Smith,  V.   Kerry  and Vincent G.  Munley,  "The  Relative Performance  of Various



     Estimators  of  Recreation  Participation  Equations,"   Journal  of  Leisure




     Research 10, 165-176  (1978).



Stevens,  Joe  B. ,  "Recreation  Benefits  from  Water  Pollution Control,"  Water



     Resources  Research 2  (2), 167-182 (Second Quarter 1966).




Stevens, Joe  B. ,  "Recreation Benefits from Water Pollution  Control:   A Further



     Note  on  Benefit Evaluation,"  Water  Resources Research  3  (1), 63-64 (First



     Quarter 1967).




Stopher,  Peter  R.  and  Arnim  H.  Meyburg, Urban Transportation,  Model ing  and



     Planning (Lexington Books,  Lexington, Massachusetts, 1975).




Sutherland,  Ronald J. , "Recreation  Benefits and  Displaced Facilities," Journal



     of Leisure Research 14 (3), 248-262  (1982).




Sutherland,  Ronald  J.,   "The  Sensitivity  of  Travel  Cost  Estimates to  the




     Functional  Form  and  Definition   of  Origin  Zones,"  Western  Journal  of



     Agricultural Economics 7 (2), 87-98  (July 1982).






 150

-------
Sutherland, Ronald J.,  "A Regional Approach to Estimating Recreation Benefits of



     Improved Water  Quality,"  Journal  of Economics and Environmental Management



    14 (3), 229-247  (September 1982).




Sutherland, Ronald J.,  "Recreation and Preservation Valuation Estimates for the



     Flathead  River  and  Lake System,"  Flathead  River Basin  Study,  Kali spell,



     Montana, 1982.



U.S. Department  of   Commerce,  Bureau  of  Census,  Current  Population Estimates,



     Series  P-25 (U.S.  Government  Printing  Office,  Washington,  D.C.,  1976).



U.S. Department  of  Commerce,  Bureau  of Census,  1980  Census of  Population  and



     Housing for Washington, Oregon and Idaho, January 1981.



Water  Resources  Council,   "Procedures  for  Evaluation  of  National   Economic



     Development (NED) Benefits and Costs in Water Resources Planning (Level  C)"



     Federal Register (December 14, 1979), pp. 72950-72965.



Watson,  Peter  L. ,  "Choice of  Estimation Procedure  for Models  of Binary Choice:



     Some  Statistical  and Empirical Evidence," Regional  and Urban  Economics 4,




     187-200 (1974).



Williams,  Martin and V.  Kerry Smith,  "Non-Price Determinants  of  Model  Choice



     Decisions:   An Econometric  Analysis,  Regional  and  Urban  Economics  9,




     197-217 (1979).



Willig,  Robert  D. ,  "Consumers'   Surplus  Without  Apology,"  American  Economic




     Review 66  (4),  589-597  (September  1976).



Ziemer,  Rod F., Wesley  N.  Musser and  R.  Carter  Hill, "Recreation Demand Equa-



     tions:   Functional  Form  and Consumer Surplus,"  American Journal  of Agri-




     cultural  Economics 62  (1), 136-141 (1980).
                                                                             151

-------
                                          Printed in the United States of America
                                                    Available from
                                          National Technical Information Service
                                              US Department of Commerce
                                                 5285 Port Royal Road
                                                 Springfield, V A 22161

                                                   Microfiche (A01)
               NTIS
Page Range   Price Code
  001-025
  0264)50
  051 -075
  076 100
  101 125
  126150
A02
A03
A04
A05
A06
A07
                               NTIS
                Page Range   Price Code
               NTIS
Page Range  Price Code
               NTIS
Page Range  Price Code
151-175
176-200
201 225
226250
251-275
276,300

A 08
A09
A 10
All
AI2
AI3

301 325
326350
351 375
376-400
401 425
426-450

A14
A15
A16
AI7
AI8
AI9

45 1 475
476500
501 525
526550
551 575
576600
601 up*
A 20
A2I
A22
A23
A 24
A25
A9Q
'Contact NTIS for a price quote.

-------