EPA-600/5-77-013
September 1977
Socioeconomic Environmental Studies Series
OXIDANT AIR AND WORK
PERFORMANCE OF CITRUS
HARVEST LABOR
Health Effects Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina 27711
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific .and Technical Assessment Reports (STAR)
7. Interagency Energy-Environment Research and Development
8. Special Reports
9. Miscellaneous Reports
This report has been-assigned to the SOCIOECONOMIC ENVIRONMENTAL
STUDIES series. This series includes research on environmental management,
economic analysis, ecological impacts, comprehensive- planning and fore-
casting, and analysis methodologies. Included are tools for determining varying
impacts of afternative policies, analyses of environmental planning techniques
at the regional, state, and local levels; and approaches to measuring environ-
mental quality perceptions, as well as analysis of ecological and economic im-
pacts of environmental protection measures. Such topics as urban form, industrial
mix, growth policies, control, and organizational structure are discussed in terms
of optimal environmental performance. These interdisciplinary studies and sys-
tems analyses are presented in forms varying from quantitative relational analyses
to management and policv-oriented reports.
mis document is available to the public through the National Technical Informa-
tion Service, Springfield. Virginia 22161.
-------
EPA-600/5-77-013
September 1977
OXIDANT AIR POLLUTION
AND
WORK PERFORMANCE OF CITRUS HARVEST LABOR
by
Thomas D. Crocker and Robert L. Horst, Jr.
Department of Economics, University of Wyoming
Laramie, Wyoming 82071
Contract No. 68-02-2204
Project Officer
Donald G. Gillette
Criteria and Special Studies Office
Health Effects Research Laboratory
Research Triangle Park, N.C. 27711
U.S. ENVIRONMENTAL PROTECTION AGENCY
OFFICE OF RESEARCH AND DEVELOPMENT
HEALTH EFFECTS RESEARCH LABORATORY
RESEARCH TRIANGLE PARK, N.C. 27711
EPA - RTP LIBRARY
-------
DISCLAIMER
This report has been reviewed by the Health Effects Research
Laboratory, U.S. Environmental Protection Agency, and approved for
publication. Approval does not signify that the contents necessarily
reflect the views and policies of the U.S. Environmental Protection
Agency, nor does mention of trade names or commercial products
constitute endorsement or recommendation for use.
-------
FOREWORD
The many benefits of our modern, developing, industrial society are
accompanied by certain hazards. Careful assessment of the relative risk
of existing and new man-made environmental hazards is necessary for the
establishment of sound regulatory policy. These regulations serve to
enhance the quality of our environment in order to promote the public
health and welfare and the productive capacity of our Nation's population.
The Health Effects Research Laboratory, Research Triangle Park,
conducts a coordinated environmental health research program in toxicology,
epidemiology, and clinical studies using human volunteer subjects. These
studies address problems in air pollution, non-ionizing radiation,
environmental carcinogenesis and the toxicology of pesticides as well as
other chemical pollutants. The Laboratory develops and revises air quality
criteria documents on pollutants for which national ambient air quality
standards exist or are proposed, provides the data for registration of new
pesticides or proposed suspension of those already in use, conducts research
on hazardous and toxic materials, and is preparing the health basis for
non-ionizing radiation standards. Direct support to the regulatory function
of the Agency is provided in the form of expert testimony and preparation of
affidavits as well as expert advice to the Administrator to assure the
adequacy of health care and surveillance of persons having suffered imminent
and substantial endangerment of their health.
The economic impact on individuals from exposure to high oxidant
concentrations may be reflected in many forms. This study attempts to
measure in economic terms one of these forms - the effect on worker
productivity. The results of this study indicated that the average income
citrus workers in Southern California was reduced by approximately two
percent when working in areas where oxidant concentrations were high.
Considerable differences in performance levels of workers were noted when
exposed to similar environmental conditions. This report represents the
first attempt to document the economic cost of reduced productivity, a
very important and frequently neglected social cost of air pollution.
John H. Knelson, M.D.
Director,
Health Effects Research Laboratory
iii
-------
PREFACE
This project was initiated in the summer of 1975 while the authors were
at the University of California, Riverside. Dr. Donald Gillette of the
Health Effects Research Laboratory of the U.S. Environmental Protection Agency
originally suggested the research. Professor Lester Lave of Carnegie-Mellon
University, Professor Jon Nelson of The Pennsylvania State University,
Professor Wallace Gates of Princeton University, the Resource Economics Group
at the University of New Mexico, and Professors Ralph d'Arge, Robert Rowe,
and Todd Sandier of the University of Wyoming have all provided helpful
comments. Personnel of the Statewide Air Pollution Research Center of the
University of California, particularly Dr. C. Ray Thompson, have under rather
trying circumstances, greatly expedited administrative details of the project.
Computational assistance has been provided by the University of Wyoming
Computer Center.
iv
-------
ABSTRACT
This project assesses the effect of photochemical oxidants on the work
performance of twelve individual citrus pickers in the South Coast Air Basin
of southern California. A model of the picker's decision problem is constructed
in which oxidants influence the individual's picking earnings and leisure-time
via a short-term and reversible morbidity effect. Circumstances are specified
under which this effect can be interpreted as the additional earnings the
individual would have to receive in the presence of oxidants in order to make
him indifferent to the presence of oxidants. This Hicksian compensating
surplus is estimated separately for each of twelve individuals. In terms of
absolute dollar magnitudes, compensating surpluses appear to range from less
than twenty dollars to nearly two hundred dollars over an entire calendar
year, given the piece-work wage rate scales and the levels of air pollution
prevailing in the South Coast Air Basin during 1973 and 1974. As a percentage
of what individual earnings would have been in the absence of air pollution,
the dollar magnitudes range from three-tenths of one percent to nine percent.
The average is about two percent. All estimates of the compensating surplus
are conditional upon the individual not adjusting the hours he picks in
response to air pollution.
Estimates give fairly strong support to the hypothesis that air pollution
impact, measured in terms of the compensating surplus, tends to increase with
increasing numbers of hours worked.
No tendency was found for the individual to substitute leisure-time for
work-effort as ambient oxidant levels increased. However, the procedures
employed to estimate this relationship could have biased the results.
-------
CONTENTS
Foreword
Preface iv
Abstract v
Figures vii
Tables viii
1. Introduction I
Footnotes to Chapter 1 4
2. The Empirical Setting 5
The Setting 5
The Mechanics of Picking Citrus Fruit 9
The Wage System 12
A Reviov oC the Salient Features 15
Footnolci- to Chapter 2 38
3. The Data l',.i ;- 20
Description 20
Possible Sources of Measurement Error 23
Footnotes to Chapter 3 26
4. A Model of the Harvest Operation 27
Footnotes to Chapter 4 38
5. Air Pollution and Earnings: Empirical Results 39
Estimates of the Inverse Supply Function 40
Increases in Income Required to Compensate
Pickers for Earnings Losses Due to Air Pollution 58
Does Air Pollution Vary with
Picker's Physical Condition? 61
Footnotes to Chapter 5 73
6. Air Pollution and Absenteeism: Empirical Results 74
Absenteeism 74
Simultaneous Adjustments of Work Effort and Leisure. ... 79
Footnotes to Chapter 6 80
7. Conclusions and Feasible Extensions .SI
Conclusions 81
Feasible Extensions of this Research 85
Footnotes to Chapter 7 91
Bibliography 92
vi
-------
FIGURES
Number Pace
4.1 Relationship Between Grove Conditions
and Fruit Yields 29
4.2 The Individual Picker's Choice ol
Earnings Opportunities 35
vii
-------
TABLES
Number Page
2.1 Annual Harvesting and Marketing
Cycles for Southern California citrus 7
2.2 Rates of Pay in Cents per Box for Lemon
Picking, by Tree Classes, Yields,
and Fruit Size, Ventura County 1964 14
3.1 Individual Picker Performance
and Grove Condition Data 22
3.2 Temperature and Pollution Stations 22
5.1 Glossary of Variable Names 43
5.2 Simple Correlation Coefficients Between
0Z, 0ZH, and TM for Various Lemon Pickers 44
5.3 Simple Correlation Coefficients Between
0Z, 0ZI1, and TM for Various Orange Pickers 44
5.4A Earnings Estimates by
Two-Stage-Least-Squares for Lemon Pickers 50
5.4B Earnings Estimates by
Ordinary-Least-Squares for Orange Pickers . . 52
5.5A Earnings Estimates by Two-Stage-Least-
Squaras for Lemon Pickers When LTM Deleted 54
5.5B Earnings Estimates by Ordinary-Least-
Squares for Orange Pickers When LTM Deleted 56
5.6 Required Picker Income Compensation 59
5.7A Air Pollution Coefficients for H Partitionings 63
5.7B Air Pollen Coefficients for II
Partitionings When LTM Deleted 65
5.8 Earnings Estimates by Ordinnry-Least-Squares
for }\ Partitionings of Irvine Orange Pickers 68
5.9 Required Picker Income Compensations
Using Results of H Partitionings 70
6.1 Absentee; ism Estimates by Ordinary-
Least-Squnres for Upland Lemon Tickers 76
viii
-------
Chapter 1
INTRODUCTION
Upon having acquired some familiarity with the epidemiological
literature reviews attempting to document the covariation of health effects
and air pollution, one is struck by the frequent inability of these reviews
to discover a substantial number of consistent findings for the effects
thought to be caused by any one pollutant. Various reasons are typically
advanced for this lack of consistency: inadequate characterization of
the pollutants; the use of noncomparable, and sometimes questionable,
estimating techniques; failure to account for other environmental influences
and self-induced health stresses; failure to distinguish between pollution
levels at work and at home; lack of attention to the difference between
indoor and outdoor pollution; and other factors- Nowhere in this refrain
is it pointed out that epidemiology lacks an analytical framework in which
the objects of study, human beings, are viewed as being capable of choice.
In particular, the health effects of air pollution are usually treated as
being absolute, even though all epidemiological findings are statistical
inferences drawn from a sample of individuals with minds of their own.
Basically, a set of inputs, including air pollution is posited to exist
and these inputs are considered to be combined, on grounds of some a priori
investigator knowledge about exogenously determined physical and biological
associations, to produce an output, an observed health effect. The epidemio-
logical literature generally fails to recognize that to the extent health
effects are subject to fixed economic and non-economic constraints, these
effects have to be measured on norms endogenous to the individual human
being. Attempts to explain the etiology of observed health effects must
recognize that these individuals use different input mixes and magnitudes
because: (1) they face different sets of relative prices for various
combinations of preventive and ameliorative health care; (2) they luxve
different biological endowments, measured and nonmeasured; and (3) they
succeed to varying degrees, in the presence of uncertainty, in maximizing
utility. Most epidemiological effort accounts for only the second of these
considerations, even though remedial measures to combat pollution-induced
-1-
-------
health effects may differ, depending on whether the etloJopy of the health
effects depends on economic or biological factors. One purpose of this
study is to provide an example, albeit an incomplete example, of how
microeconomic analysis permits the introduction of the first consideration
into an empirical study fundamentally epidemiological in emphasis. No
serious attempt is made, however, to show how the analytical framework of
the study might be generalized to encompass a broad variety of epidemiological
problems. Nevertheless, although no effort will be made to do so here, it will
be fairly apparent that the analytical framework is easily generalized to
account for the third consideration. The typical epidemiological study of
the health effects of air pollution might capture the health effects that
lie in people's stars; it fails to capture the health effects that lie in
people themselves.
A second and perhaps less controversial motivation for this study is
to provide and apply an analytical framework for assessing the economic
effects of environmental pollution upon the performance of inputs,
particularly labor inputs, in production processes. Nearly all studies of
the economic effects of environmental pollution view changes in relative
3
market prices as being demand-induced. Constant relative unit supply
prices as between and among outputs are assumed. However, any change in
process productivity necessarily alters the price the producer must receive
in order to be willing to supply a given quantity of an output good. These
productivity changes therefore also constitute a source of change in output
market prices. Failure to consider the impact of pollution upon supply
means that an important facet of the total economic effect of environmental
pollution is being neglected. Although the present study is limited to
estimating the effect of air pollution upon worker performance, it does
provide an example of ;i necessary step in any attempt to ascertain the
ultimate economic Impact, upon the market price of the outputs the workers
cooperate in producing.
Those studies oi die: economic effects of environmental pollution upon
inputs that have been performed are known as materials damnge studies. They
have two distinguishing common characteristics. First, they focus entirely
upon specific inputs without devoting attention to the manner in which the
inputs are involved in a production process. This study appears to be the
first dealing with a particular input that explicitly accounts for the producer's
decision problem.
-2-
-------
A second distinguishing char ac tor .1 si ir. of the materials damage studies
is their fixation on nonhuman inputs. Except for a few rather rough efforts
employing highly aggregated data, the economics of the effects of environ-
mental pollution upon worker performance has been left undone. Perhaps the
reason is that data drawn from the performances of individual workers doing
jobs requiring substantial physical exertion in an occasionally highly
polluted environment have been lacking. Data of this sort were available
for this study.
Finally, there currently is virtually no evidence bearing upon the
economic effects of photochemical smog. In spite of this, quite stringent
ambient air quality standards have been adopted for the various chemical
precursors of smog ay well as for the mixture that results from atmospheric
transformation procf.MsoK. Although it cannot be expected that the effects
established in these pages constitute a large portion of the total negative
economic effects of photochemical smog, the results provide defensible
evidence that these effects do in fact exist.
In succeeding chapters, the effect of photochemical smog upon the work
performance of citrus workers is investigated. The next chapter describes
the analytically relevant features of the market setting for the empirical
efforts reported iu the fifth and the sixth chapters. A third chapter is
both a summary of the data base available for the study and a commentary
on the deficiencies of this data base with respect to the analytical model
presented in the fourth chapter. A final chapter summarizes the study,
points out its limitations, and suggests how more information might be gleaned
from the same data base.
-3-
-------
Footnotes: Chapter 1
1. For an extension of this list, see Commission on Natural Resources
(1975) 58-169.
2. A recent paper by Smith (1975) tends to support the argument of
this paragraph. In applying the Ramsey tests for specification error to
some thirty-six epidemiological studies on air pollution and mortality,
Smith found that not a single one of the studies met the Ramsey tests for
the absence of this error!
3. The studies used by Waddell (1974) are almost entirely of this sort,
4. Waddell (1974) lists several such studies.
5. The only really careful study available appears to be Nelson (1975),
-4-
-------
Chapter 2
THE EMPIRICAL SETTING
The Setting. This study deals with the men and women whose primary
occupation is the harvesting of citrus fruit in the South Coast Air Basin
of southern California. The object of the study is to ascertain whether
their work performance is influenced by the presence of photochemical smog.
The occupation of citrus harvesting has the ease of entry and exit, the
geographical and numerical scope, and the absence of idiosyncratic (i.e.,
heterogeneous,, highly differentiated, task-specific skills enabling the
current occupant to possess a degree of monopolistic advantage) character-
istics that Doeringer and Piore (1971) term the secondary labor market.
Harvesting operations in citrus groves are highly labor-intensive activities
for which there at present exist no economic substitutes for hand-labor.
A substantial number of workers choose to be employed on a year-round basis
and are thus exposed to varying air pollution levels over the year. Since
the citrus harvest occurs in both high and low smog months, many individuals
work during periods of relatively low and relatively high smog levels.
Except for backyard citrus trees, citrus is a crop which even in the
smallest commercial groves has harvest labor requirements well in excess of
any labor supply the family of the owner is likely to be able to provide.
Rosedale and Mamer (1974, p. 11), in a study of harvest operations in
Ventura County, the center of California's lemon industry, indicate that
from 1966 through 1972 eighty-five to ninety percent of harvest costs were
direct labor costs. In an earlier study of the Ventura County lemon industry
Smith, et. al. (1965, p. 4) state that "...all labor and material costs
for lemon production on the tree averaged eighty-five cents per field box."
Forty-five cents of this sum was picking cost.
There appears to have been very little change in citrus harvest labor
productivity over the years. Our data indicate that the representative
worker picks about 1900 pounds of lemons and 3000 pounds of oranges por
eight-hour workday. Fellows (1929, p. 71) indicates that in 1929 those
rates corresponded to 1750 pounds of lemons and 3000 pounds of oranges.
Although mechanical harvesting aids and systems do exist, the U.S. Census
-5-
-------
of Agriculture (1973, pp. 6, 29) shows ih-it, in the 1968-69 season, of 764
reporting California lemon growers who harvested nearly 700 million pounds
of lemons, only six growers, harvesting a total of less than five million
pounds of lemons, used harvesting machines. Similarly, of 1969 California
orange growers reporting slightly more than two billion pounds of oranges
harvested, only sixteen growers, whose aggregate harvest was 11.7 million
pounds of oranges, employed machines for picking. For both technological
and economic reasons, it would appear that picking labor is an integral and
necessary part of the California citrus industry.
Harvest operations in the California citrus industry are typically
organized around large packing-houses that are either privately owned and
usually specialize solely in harvesting, packing, and marketing, or are
grower cooperatives (e.g., Sunkist Growers) who participate in all facets of
citrus production. Many growers turn their harvest operations almost
entirely over to the packing-houses, permitting picking policies and the
sequence of picks across different owners' groves to be established by the
packing-house management. This management is said to have a general idea
at any particular time of the sequence in which groves are to be picked,
but the initially selected sequence is subject to alteration according to
weather conditions, the rate at which fruit in particular groves is ripening
and growing, and other factors. Orange picking activities are said to be
somewhat less subject to plan alterations of this sort than are lemons.
This perhaps is duo lo the fact that at least some lemons are normally
picked every week of Che year, while the picking times for orange varieties
are more limited in l:hp. choice of harvest dates. The marketing of oranges
appears to be similarly concentrated in time. Table 2.1 below gives the
relevant picking and marketing calendar time intervals for southern
California lemons and oranges. The geographical area to which the table
refers roughly corresponds to the climates prevailing over the South Coast
Air Basin.
-6-
-------
Table 2.1
Annual Harvesting and Marketing Cycles for Southern California Citrus
Lemons
Oranges: Early,
Midseason, Navels Oranges: Valencias
Full bloom dates March 5-Dec. 30
Begin harvest
Most active
harvest
Begin marketing
Most active
marketing
End marketing
Aug. 1
March 5-March 30
Nov. 20
Jan. 15-July 15 Dec. 15-May 15
Aug. 1
March-July 15
July 31
Nov. 25
Dec. 20-May 20
June 15
March 5-March 30
March 10
May 10-Oct. 25
March 15
May 15-Nov. 1
Dec. 20
Source: Statistical Report Service (1975, pp. 40-44).
Within the range ul: prices that have prevailed since World War II,
the consumer demand for fresh citrus fruit as well as processed citrus fruit
2
is thought to be relatively price inelastic. However, given that citrus
fruit is often stored for as long as six months with only moderate spoilage,
an individual grower is unlikely to exercise meaningful influence upon
market price via his harvesting and marketing decisions. A further
implication is that the size of the crop and factors such as weather, rather
than market price, will be the primary influences upon the quantity and
temporal distribution of harvest labor requirements, assuming, of course,
that the price and availability of labor does not exhibit substantial
seasonal fluctuations. Interviews with packinghouse managers have confirmed
that market prices expected during the next one to six months in a particular
harvest year have little or no influence upon the choice of harvest dates
although, in exceptional circumstances, expected prices may determine
2
whether the season's fruit in a particular grove will be harvested at all.
In effect, therefore, over a fairly wide range of piece-work wage rates for
harvest labor during n particular growing season, there will be near-zero
covariation between this wage rate and the number of harvest labor man-hours
expended in a particular grove. The man-hours expended will primarily be a
function of the amount of fruit in the grove.
-7-
-------
The .Individual picker Is part of. ,\ i-ukinj1, crew l.h.il , .iccoril.i nj-, lu I In-
stage in the picking cycle, may be as small as ten people and as high as
forty people. A typical size appears to be about thirty people. The crew
is supervised by a foreman who is paid in proportion to the amount of fruit
his crew picks. This foreman is responsible for getting the pickers and
equipment to the grove to be picked and for maintaining his crew at the
desired size. Once in the grove, the foreman tries to assure that the
fruit is picked in accordance with the specified conditions. He also
maintains a record of the amount of fruit picked by each picker.
Over each crew foreman is a salaried field superintendent. This
superintendent answers to the general manager of the packing-house or
growers' cooperative and is responsible for the over-all operation and
coordination of harvesting and grove maintenance activities within and
among groves and growers. In the great preponderance of situations where
the piece-work wage rate for pickers varies with the relative difficulty
of picking conditions, it is he, prior to the entrance of pickers into
the grove, who estimates the relative difficulty of the picking opportunity
and thereby establishes the piece-work rate to apply to the particular
grove. In the words of one packing-house manager, the base rate that is
adjusted according to the degree of picking difficulty is established in
accordance with "prevailing market conditions."
Frequently, the responsibility for securing a suitable labor supply
for harvesting purposes is transferred from the packing-house or cooperative
to a labor contractor, a specialist in the recruitment and supervision of
citrus workers. Crew foremen are then employees of the contractor rather
than the packing-house or cooperative, although the actual performance of
crews will continue to be monitored by a field superintendent. Pickers
are paid for their production performance by the contractor and it is he
who sets the piece-work wage rate for each grove. The contractor, in effect,
assumes the functions and associated risks of picker recruitment, supervision,
payment, and provision of wh?u ever picker lj Ce support lac Mil Jo? arr-
standard. In returr , t'»r • .si* rac I.or i« guaranteed a certain rate of
compensation. The l.-ibor rontractors i.nvnlved in the present .study all
appear to have had Long e>.j.t:rinnce in their business and to have rather
-8-
-------
l;jr;;o opernlions. They have m.ide the 1'luw of their lousiness regular .mil
rouiinr and depend upon established customers and a large number of workers
who have been previously employed by them and to whom they can offer fairly
regular employment.
The Mechanics of Picking Citrus Fruit. Actual picking procedures for
lemons and for oranges have many common features but they also differ in
several important respects. The differences tend to be due mainly to the
fact that during any one season the fruit in a lemon grove does not mature
for picking purposes at the same time but instead is distributed over much
of the year. This means that any single grove might be picked as many as
four times in a given season. During the first three picks, only the fruit
that is "up to size" is picked. In order to ascertain whether a given
piece of fruit is of picking size, the picker must manipulate a measuring
device. In addition, as he must also do when picking oranges, he must
avoid damaging the fruit quality by leaving long stems, cutting the point
at which the stem is attached to the fruit, or pulling the fruit off the
tree. Only during the terminal or "strip" pick does the picker take all
lemons from the grove. In contrast, all orange picking activities are
"strip" picks. Further intensifying the difficulties of picking lemons
relative to oranges is the fact that lemon trees have thorns and are
generally bushier than orange trees. In fact, lemon pickers typically
wear rather heavy clothing and shoulder-length, rather awkward looking
gloves in order to protect their persons from the thorns. Many individuals
specialize in lemon-picking and will pick oranges only when there are no
lemons available; whereas relatively few people who primarily pick oranges
A
will pick lemons in the absence of oranges .
Apart from the degree of difficulty of the picking operation, the
actual mechanics of picking of the two types of fruit appear to be identical.
Citrus groves in southern California are universally planted in long,
straight rows -so as to facilitate irrigation, maintenance, and harvesting
activities. Upon the arrival of pickers in the grove, the grove is divided
according to "drive" rows down which a collection device (e.g., a truck)
periodically makes an appearance. Individual pickers are then assigned
row sets of three trees on both sides of the drive row. Each picker is
-9-
-------
usually initially assigned the same six rows with which to initiate his
picking activities at each new grove. The ease of the pick for the first
row set thus varies randomly from grove to grove for each picker. Only
this first row set is assigned. After the initial assignment, the pickers
leapfrog, although unless there are only a small number of rows remaining,
once a picker starts a row it is his to complete. In cases where the number
of remaining rows is inadequate for a one-to-one correspondence between
pickers and rows, everyone picks what remains. These procedures appear to
vary not at all among groves.
Citrus pickers are paid on a piece-rate basis; that is, each picker is
paid a unit price for the quantity of fruit he picks rather than the number
of time units he expends. The relative ease of picking therefore helps to
explain the quantity of fruit he picks and the amount of money he earns for
any given time interval during which he picks. Seamount and Opitz (1974)
disaggregate the picking activity into three facets: net picking time;
time moving within and between trees while picking; and time moving to and
from field containers and dumping fruit into these containers. The
proportion of time passed in each of these three facets will vary according
to whether the picker is engaged in skirt or ladder picking. Ladders are
used with trees the fruit of which cannot be reached with both feet on the
ground. Ladder picking is thought to slow the picker's rate of pick by
forty to sixty percent relative to skirt picking.
Net picking time is the picking act of searching, reaching, clipping,
and placing the fruit in the bag the picker has hanging diagonally across
his shoulders and carries on his hip. For ladder strip picks, it accounts
for sixty percent or more of the picker's time and more than five minutes
per standard 3115 cubic inch field box- Seamount and Opitz (1974, p. 165)
list the following nonpicker factors as probably influencing net picking
time: fruit density; distance of the fruit from the picker; fruit size;
fruit stem characteristics; tree leafiness; picker orientation; platform
stability; freedom of various picker body members; portion of the tree
being picked; and tret« height, diameter, nnd surface characteristics.
Frequency of movement within and between trees is though to be related
8
to fruit density, fruit clustering, and ease of reaching the fruit-
The trees in most citrus groves are planted sufficiently close together so
that movement between trees is thought to have little or no influence on
rate of pick. The speed with which others in the picker's crew pick could
-10-
-------
have some influence on the individual picker's rate of pick since it can
influence the distance he lias to move among row sets. However, movement
from one row set to another by a single picker is sufficiently infrequent
to make it implausible that the factor has other than trivial importance
to the rate of pick. For strip picking with ladders and for an extremely
small sample of pickers' time,moving within and between trees an average of
9
about one and one-half minutes per 3115 cubic inches in box of fruit is required.
Transporting the picked fruit from the row sets to field receptacles
placed in the drive rows and dumping the fruit into these receptacles
consumes an average of about forty seconds per 3115 field box for most
pickers. Little other than the roughness of the terrain is thought to
influence this facet of picking activity. Movement from one grove to
another is thought by Seamount and Opitz (1974, p. 169) to be a greater
influence.
In most respects, the citrus picking endeavor is ideally suited to
application of the piece-work wage rate. Output is readily defined,
measured, and monitored, the results of each picker's efforts are separable
from those of other pickers, and the difficulty of the worker differentiating
his task from the tasks of other pickers (thus making it hard for him to
argue that his task is in some sense "more difficult") all serve to make it
easy to assign the entire responsibility for perfunctory work performance
solely to the individual picker himself.
Grove factors are, of course, likely to be the major influence upon
differences in the individual picker's rate-of-pick from one time period to
another. However, it should be noted that the responses of individual
pickers to these factors can differ greatly from one picker to another.
Thus, one must be extremely cautious, in trying to generalize from the
responses of a few pickers to the entire picker population. This caution is
well supported by some of the findings of Smith, £t. al. (1965), with respect
to lemon pickers. While studying an "example" crew, they noted that the
fastest worker picked an average of 3.375 field boxes per hour while the
slowest picked only 1.750 field boxes per hour. The crew mc.in was 2.570
boxes per hour with a standard deviation of +0.389 boxes per hour. In
a separate sample of 2500 pickers only 24 percent of the total variance in
rate of pick could be accounted for by grove factors, while 64 percent
12
was accounted for by variations in pickers. They also note that variations
in rate-of-pick appear to be much greater among U.S. citizens than among
13
the Mexican nationals working in identical groves-
-11-
-------
The Wage System. As earlier note.I, citrus pickers are paid by Lhe
quantity of fruit they pick, rather than by the number of hours they work.
Pickers having two or three months of experience who are unable to earn the
14
minimum wage regularly are simply terminated. Three different classes
of means of determining the piece-work wage rate for a particular grove on
a given day appear to prevail. Two of the three are, in effect, sequential
spot contract systems in which the picker and his employer are continually
renegotiating on terms that must be satisfactory to picker and to grower.
The three classes may be distinguished according to the extent to which and
the manner in which the factors that contribute to the difficulty of picking
are taken into account.
The most sophisticated means of determining piece-work wage rates per
box of fruit picked is employed for lemons. This means is simply a component
of. a labor management system designed to reduce rates of picker turnover and
absenteeism and thereby lower grower screening and recruiting costs for
pickers as well as reducing the likelihood of having to reallocate inputs
because of the unexpected absence of a picker. Unless the grower has available
a perfect substitute at equivalent cost, each picker who quits or each day a
picker is absent means that the grower must, at a cost, attempt to adjust
either by juggling the distribution of tasks among the remaining workers
or by initially hiring more workers than the picking process requires in
order to ensure duplication of the services of absent or terminated pickers.
The motivation is to do away with the historically casual nature of the
supply of pickers to lemon growers. In order to enhance the likelihood of
assuring themselves a reasonably stable labor supply of more-or-less known
quality, the lemon growers have tacitly shifted part of the risk of the
picker's uncertain income stream and living conditions to themselves; that is,
by providing health, disability, unemployment, and life insurance, retirement
plans, explicitly stat.ed promotional tracks, paid vacations, and other
accoutrements of the modern industrial blue or white collar worker, the
growers have to some degree transferred many risks that historically have
accrued to the picker to the income streams of the growers and their
15
creditors•
One major means California lemon growers have adopted to unburden
the picker of variability in his income stream is to adjust piece-work wage
rates in accordance with the degree of difficulty in picking conditions.
-12-
-------
The more difficult: Che picking chance, i ho highlit: Liu; piece-work waj;.- r;iU-.
For lemon pickers, the wage per box of Iruit picked associated with each
combination of three key grove variables that supposedly influence the rate
of pick are published and are applicable for several weeks or perhaps an
entire season. Table 2.2 presents a pay schedule used in Ventura County
during the middle 1960's. The pay schedules relevant to lemon picking in the
current study are identical in structure, although the piece-work wage rates
have been altered over time.
As Table 2.2 indicates, the supposedly influential variables are the
number of fruit on the tree that meet the specified conditions (e.g., color),
size of fruit, and tree height. The values of these variables are recorded
at the time of picking for each grove in which the picker's crew works.
Since all fruit meeting prespeeified conditions is to be picked, a
picker's earnings in any particular grove are then the number of boxes of fruit he
picks multiplied by ..It1 prr box wage r.ite as determined by the fruit density,
fruit size and true. l'ci;;,ht In the grove. It should be: noted, however, that
these three grove variables do not always completely determine the per box
wage rate, for they do not capture all grove attributes thought to contribute
to the relative (.difficulty of picking. For example, as mentioned elsewhere, the
slope of the ground in the grove and the bushiness of the tree are also
influential. 'In Droves where variables in addition to the three variables
mentioned above are thought to be relevant, the foreman of the picker crew
apparently announces the adjustment before the picking performance.
Moreover, since all fruit meeting prespecified conditions is to be picked,
pickers have little, if any, incentive on a particular day to urge each
other to slow the rate of pick, given that all pickers are at least
earning the minimum wage. To do so would reduce the earnings of the better
pickers without enhancing the earnings or reducing the required work effort
of the slower pickers. Of course, the schedule of the per box wage rates
with respect to a particular grove variable might be adjusted over time if
it became particularly noticeable that certain pickers were receiving
earnings greatly in excess of what might normally be expected. This
adjustment might redound to the disadvantage of those pickers whose
performance was not so responsive to variations in the variable in quest: ion.
It is then conceivable that the latter pickers might urge the. former
-13-
-------
Table 2.2
Rates of Pay in Cents per Box for Lemon Picking,
by Tree Classes, Yields, and Fruit Size, Ventura County, 1964.
Class I*
Class IT
Class III
Class IV
Fruit size (number per box)
bxs/tree**
0-1/4
£ 1/4-1/2
i
1/2-3/4
3/4-1
1-1 1/2
1 1/2-2
2-3
3+
Under
240
47
41
36
33
31
29
240-
300
56
46
40
37
33
31
Over
300
70
53
45
41
36
33
Under
240
57
48
42
38
35
32
30
240-
300
64
54
46
42
38
35
32
Over
300
75
59
51
45
41
37
35
Under
240
cents
66
55
47
42
38
34
31
28
240-
300
72
60
51
46
41
37
34
32
Over
300
79
66
56
50
44
41
38
36
Under
240
78
67
58
52
47
43
38
35
240-
300
86
73
63
56
51
46
42
39
Over
300
95
81
70
6.2
54
50
46
43
*Tree classification:
Class I - Picked without a ladder.
Class II - Ladder-picked trees less than 9 1/2 feet tall.
Class III - Ladder-picked trees 9 1/2 to 12 feet tall.
Class IV - Ladder-picked trees over 12 feet tall.
**Field box capacity: 2,926 cubic inches.
Source: Snith, et. al. (1965, p. 6).
-------
pickers to reduce their per to nuances. N.-verthcltiss, since there arc: .'several
thousand pickers employed in any one crop seasoni it does not seem far-
fetched to view the picker as a wage-taker; that is, he acts as jLf his picking
performance does not influence the per box wage rate he will receive and,
furthermore, other pickers act as if he does not influence the per box wage
rate they receive.
The second class of means of determining the piece-work wage rate is
considerably less formal. It is what is in effect a sequential spot
contracting system found in orange harvest efforts where the grove variables
likely to influence picking performance differ from one grove to another.
Even for those crews who, when picking lemons, work under a published fee
schedule that matches wage rates to combinations of picking conditions, the
per box wage rate applicable to a particular orange grove is only determined
shortly before the entrance of the crew into the grove. Upon the discovery
that the prior determination of the wage rate does not accurately reflect
picking conditions, tlii.'s v/age rate may be adjusted. However, at least for
the crews for whom wc->. collected dita, the wa^e-rate was never reduced after
entrance to the grove.. it was only increased and then only infrequently.
Finally, for sets; of groves that are extremely uniform in quality and
for which pickers will therefore be picking for extended periods of time
under more-or-less uniform conditions, piece-work wage rates are established
only in accordance with the labor supply and demand conditions prevailing
at the beginning of the season or picking period. This, of course, raises
the possibility that faster workers may be urged by their slower fellows
to reduce their picking rates so as to reduce the possibility of management
demands to raise average performance levels. Management is undoubtedly
aware of these group pressures but, to judge from the pay system they have
adopted, it apparent 1 v I-.-cJ.s that the cost i.-f the loss in jdcker productivity
is outweighed by :;hc: rost reductions due to not having to keep detailed
picker performance and grove attribute records when groves do have uniform
attributes. In any case, tor the data we possess, it is only in the
Irvine area where this could constitute an analytical problem.
A Review of the Salient Features. Since the purpose of this study is to
estimate the response of the citrus picker's work performance to variations
in air pollution, an analytical model of the picker's decision problem is
-15-
-------
required in order to gencraLe testable hvpotheses. Most important, the
model must be a reasonable representation of reality. From the discussion
of the preceeding pages, the following salient features of the market for
citrus picking labor can be culled. It is desirable to account for these
features in any model of the picker's decision problem.
1) At least on the supply side, the market for citrus pickers embodies
the major feature of a competitive labor market, i.e., the individual
picker is a wage-taker.
2) The picker is paid entirely on a piece-work basis.
3) Entry into the market is easy. Exit appears to be even easier.
4) The market has substantial geographical and numerical scope.
5) Citrus picking, at least for any single variety, is a homogeneous
activity for which individual pickers cannot differentiate their
particular tasks from those of other pickers.
6) The citrus harvest is a highly labor-intensive activity. Except
for ladders, cutting shears, and bags into which to deposit picked
fruit, complementary capital inputs exercise little, if any, influence
on the individual picker's output. Moreover, there are no good
econdmic or even technical substitutes for the individual picker.
7) Market price of citrus fruit is not a primary influence on the
quantity and temporal distribution of harvest labor requirements
within a single harvest season.
8) The picker's output is readily defined, measured, and monitored.
9) Picking procedures are standardized from one grove to another.
10) While picking a particular grove, picking procedures do not require
the picker to take involuntary leisure.
11) Each picker's efforts are separable from those of other pickers.
12) A learning curve of two or three months duration exists for
picking citrus fruit.
13) Substantial differences are known to exist among pickers in the
responses of their picking rates to certain grove attributes.
14) The citrus picker's immediate supervisor, the picking crew foreman,
is typically paid on the basis of the quantity of fruit his crew
picks per unit time.
15) A salaried field superintendent from a growers' cooperative or a
packing-house oversees the crew foreman.
-16-
-------
16) For given labor market conditions, piece-work wage rates vary witli
the degree of picking difficulty in a particular grove.
17) The piece-work wage rate is set by the crew foreman or field
superintendent before initiation of picking activity in a
particular grove. However, this wage rate may later be modified
if initial expectations about picking conditions are not fulfilled,
18) During a particular harvest season, the individual grower is a
price-taker for both his fruit crop and his use of harvest labor.
-17-
-------
Footnotes: Chapter 2
1. In a study of family and hired labor on U.S. farms, Sellers
(1966, p. 35) states, in effect, that all commercial citrus growers
employ hired labor.
2. See Bell (1965, p. 4).
3. Interview of the first author with Mr. Robert Lamberson and
Mr. Edward Ruiz of Upland Lemon Growers, March 11, 1976.
4. Interview of the first author on March 12, 1976, with
Mr. Xavier Piedra, Manager of the San Gabriel Valley Labor Association.
5. The description in this paragraph is a synthesis of conversations
of the first author with Mssrs. Lamberson, Ruiz, and Piedra, as well as
Mr. Mack Garcia of the River Growers Association in East Highlands,
California.
6. Seamount and Opitz (1974, p. 165).
7. Ibid.
8. Ibid., p. 167.
9. Ibid.
10. Ibid.
11. Smith, et_,_ al_._ (1965, p. 20).
12. Ibid., p. 36.
13. Ibid., p. 23.
14. Smith, et. al. (1965, pp. 46-51) state that this occurs.
Interviews of one of the authors with labor camp managers confirmed the
Smith, et. al. statement.
15. See Manpower Administration (1969) and Rosedale and Mamer (1974)
for detailed descriptions of the features of the system. The description
offered in the latter source which, among other things, refers to special
leaves, birthday greetings and cake, counseling, and legal aid, Christmas
greetings, adult educations and entertainment, is reminiscent of newspaper
accounts of the Japanese firm or perhaps an academic environment.
Mr. Jack Lloyd of the Coastal Growers Association in Vonturn County
is widely credited with developing the system. Insights into the
motivations for developing the system are available in Smith, et. al.
(1965, pp. 14-19). A study of the variability of the degree of risk-shifting
-18-
-------
from pickers to growers with respect to such factors as the productivity
and dependability of the picker, the market for lemons, societally
provided benefits, labor supply, and other factors would be most
interesting. At the abstract level, a framework for approaching these
questions is to be found in Alchian and Demsetz (1972) and Crocker (1973)
A much more thorough development is presented in Azariadis (1975) .
16. Rosedale and Mamer (1974, p. 19) state that in 1973, 3335
pickers were employed by the Coastal Growers Association of Ventura
County alone.
-19-
-------
Chapter 3
THE DATA BASE
Description. Two classes of data make up the empirical basis of
this study: (1) observations on indicators of picker work performance
such as boxes of fruit picked and hours worked; and (2) observations on
the conditions under which the picker worked such as the piece-work
wage rate and grove and environmental conditions. These two data classes
are available on no less than a day-to-day basis for each individual
•picker studied. Air pollution and temperature data are usually available
on an hour-by-hour basis. Since no systematic effort was made to collect
data on individual picker characteristics such as age and state of health,
no comparisons across individuals of the reasons for variations in work
performance are possible.
Except for the air pollution and temperature observations, all data
were acquired from records maintained by citrus packing-houses and
labor camps in southern California. These packing-houses and labor camps
were selected from a list supplied by Sunkist Growers Cooperative. Every
packing-house and labor camp on the list was sent a copy of the original
research proposal along with a letter explaining the type of data in which
we were interested. The various packing-houses and labor camps were then
contacted by telephone in order to ascertain their willingness to cooperate
in the study and the nature of the data they possessed. The following
criteria were developed for the collection of data from the packing-houses
and labor camps during the summer of 1975. It should be recognized that
the application of these criteria resulted in a nonrandom sampling of the
citrus picker population.
1) The study is a p.mel study in which the objects of interest .ire the.
daily work [ KM: f.o i •mnncos of individual citrus pickers. Data files
containing detailed information on the day-to-day work performances
and conditions of individual pickers are therefore to be sought.
-20-
-------
This perspective of Ihe ceatr:il objective of the study avoided the
necessity of collecting possibly sensitive data on individual picker
socioeconomic attributes such as age, state of health, etc. Each
individual worker selected for study can then be treated as a
separate and distinct study.
2) Most of the individual pickers for whom work performance and
conditions data are acquired must have worked at times and locations
where ambient concentrations of photochemical smog were substantially
above background levels. Given that the central objective of the
study is to ascertain the covariation of picker work performance and
photochemical smog the rationale for this criterion is obvious.
3) Pickers are to be selected having near continuous records of
employment as citrus harvesters during 1973 and 1974. The years
1973 and 1974 were selected because citrus growing conditions,
according to packing-house managers, exhibited substantial differences
between the years. Moreover, the most detailed ambient smog data
was available for these years. Pickers with long employment
histories during the two-year period were desired in order to
maximize available degrees of freedom for hypothesis testing.
4) Pickers are to be selected having at least one year of experience
in citrus harvesting. It is hoped that the application of this
criterion negated any of the learning effects to which Smith,
et. al. (1965, pp. 41-46), refer.
5) Pickers having relatively high, moderate, and low records of average
daily earnings are to be selected. Although there was no intent in
the study to maka detailed explanatory comparisons of work
performance among pickers, it was thought desirable that a set of
pickers having a fair distribution of apparent potential producti-
vities be selected. The reason was an intuition that the influence
of air pollution upon picker performance might vary witli the
potential productivity of the picker.
In Table 3.1 .! ,-i provided a. listing of all picker perlormance and
grove condition data obtained for 237 individuals, with 103 individuals from
Upland (U) and Riverside (H) , 60 individuals l"rom VonLura (V), 32 iiullviilu.il;-:
from Irvine (I), and 42 individuals from San Bernard ino-Uc'dl .inds (S) .
Temperature and air pollution data consist of records of a number of
monitoring stations throughout southern California. These records are
maintained on computer tape by the Statewide Air Pollution Research Canter
of the University of r.-il 1 F«rrH.i , Riverside. Table 3.2 provides those
tcmperaCu re: UIK! air pol I u --.ii monitoring slatiuns by mmc1 tli.iL were used
-21-
-------
A.
Table 3.1
Individual Picker Performance and Grove Condition Data
Data Organized by Crew to which Individual Picker Belongs.
Data Description
Unit of Measure
Location
Calendar date
Grove location
Camp departure
Camp return
Picking initiated
Picking terminated
Wage rate
Fruit type
Fruit size
Tree class
Average boxes picked
per tree by crew
Total trees picked by crew
Tree age
Day
1/4 Section
Military time
Military time
Military time
Military time .,
Cents per 3115 in. box
Lemons, navels, valencias
Fruit per 3115 in. box
Height in feet
3115 in. boxes
Trees
Years
B. Data Organized by Individual Picker.
U,V,S,I,R
U.V.S.I.R
U, S, R
U,S,R
U,S,I,R
U,S,I,R
U,V,S,I,R
U,V,S,I,R
U,V,I
U,V
U,V,S,I,R
U,V,I
S
Data Description
Unit of Measure
Table 3.2
Temperature and Pollution Stations
Location
Work time
Boxes picked
Refused to work
Sick, did not work
No reason, did not work
Nonpicking work activity
Weekly gross income
Weekly net income
Lives in labor-camp
Hours _
3115 in. boxes
(0,1)
(0,1)
(0,1)
Hours
Cents
Cents
(0,1)
U,V,S,I,R
U,V,S,I,R
U
U
U
U,V,S,I,R
U,V,S,I,R
U,V,S,I,R
U,V,S,I,R
Grove
Locations
Temperature Station Name
Pollution
Station Name
Upland
Ventura
San Bernardino-
Redlands
Irvine
Riverside
Upland
Santa Paula (1973)
Summit Fire Lookout (1974)
San Bernardino
El Toro Air Station
UC, Riverside
Upland Civic Center
Santa Paula
San Bernardino
El Toro
Norco
-22-
-------
for each of lite general grove locations. Ml temperatures used in this
study are maximum hourly arithmetic average dry-bulb temperatures in
F° on each work-day of interest. Air pollution measures are hour-by-hour
arithmetic averages of ambient concentrations of ozone or oxidants in
parts per million by volume.
Possible Sources of Measurement Error. Known as well as suspected
measurement errors lurk throughout the data set used for this study.
Some are perhaps sufficiently severe to introduce serious possibilities
of bias into empirical estimates of relationships developed from the
analytical framework of the next chapter.
Given the objective of this study, by far the most unkind source
of measurement error is the air pollution data. The following quote, in
a December 18, 1975, memorandum entitled Errors in Ozone/Oxidant Monitoring
Systems from Mr. Roger Strelow, Assistant Administrator for Air and Waste
Management, USEPA, to all USEPA regional administrators, succinctly states
the most dismaying facet of the problem with the air pollution data,
"Based upon results to date, we suspect that the existing data
could possibly contain some positive and some negative errors...
Therefore, I do not believe we should attempt to make any
modifications to the existing data; we simply do not know what
adjustments to make, or even if the data is generally too high or
too low." (p. 3)
Earlier in the memorandum, Mr. Strelow notes that certain combinations of
instrumentation, calibration procedure, and operator performance appear
to result in a variable negative bias.
The above does not exhaust the sources of error in the air pollution
data. With the exception of the air pollution and temperature monitoring
stations relevant to the fruit harvesting sites in Irvine and Ventura,
all monitoring stations are generally located five to eight miles from
the groves. In both Irvine and Ventura, the monitoring stations are
central to and only a short distance from all picking sites. However,
in Upland, San Bernardino-Rcdlands, and Riverside the stations are in
downtown areas and niro typically at somewhat lower elevations thnn in
the groves. The Locations uf these stations relative to the groves made
it impossible, by h.r ian^ulutinji among stations, to arrive at a weighted
-23-
-------
menu of harvest site air pollution concentrations and temperatures.
Instead, the temperature and the concentration at the monitoring
stations closest to the harvest site have in all cases been used as a
measure of the temperature and air pollution at the harvest site. We
have absolutely no basis for judging discrepancies in measures realized
at the stations and the actual measures at the harvest site. If a guess
is required we would assert, on no basis other than casual observation,
that readings at the Upland, San Bernardino-Redlands, and Riverside
stations were slightly higher for some hours on some days more frequently
than they were slightly lower than the actual state of affairs in the
groves. This assertion is made on the basis of the downtown locations
and lower elevations of the monitoring stations; it is not an assertion
we are anxious to defend.
Relative to the measurement errors in the environmental conditions data,
sources of this error in the grove conditions and work performance data seem
innocuous and limited indeed. Perhaps the most serious is the
rounding-off of the number of hours a picker has worked to the nearest
half-hour. In circumstances where the work-day has been rather short,
this could lead to some bias in estimates, although it seems likely that
there is no systematic bias with respect to the sign of the error.
It is possible that error exists in the size-of-fruit variable,
when observations on this variable are available. Typically, the daily
value for this variable is determined by having the foreman of the picking
crew select five boxes of fruit harvested that day from the grove being
picked. The total number of fruit in the boxes divided by five then
represents the "size-of-fruit" recorded for determining the piece-work
rate of pay. Although an effort is apparently made to select individual
boxes from a number of locations within a particular grove, a sample
of five boxes from the daily population of several hundred boxes a crew
is likely to pick is at bust a "small" sample; that is, it will probably
be biased. We possess no information, however, permitting us to evaluate
the direction or the magnitude of this possible bias.
-24-
-------
Other than the instances referred Lo in this section we are unaware
of any other possible sources of measurement error in the data we have
used.
-25-
-------
Footnotes: Chapter 3
1. Worker performance data were obtained from the San Gabriel
Valley Labor Association of Cucumonga, the Lemoneira Ranch of Santa Paula,
the River Growers Association of East Highlands, and Irvine Valencia
Growers of Irvine. Grove condition data were provided from Upland
Lemon Growers of Upland, Lemoneira Ranch of Santa Paula, Western Fruit
Growers Packing Company of Mentone, Irvine Valencia Growers of Irvine,
and Corona College Heights Citrus Company of Riverside.
-26-
-------
Chapter 4
A MODEL OF THE HARVEST OPERATION
Our fundamental purpose is to explain the influence, if any, that
photochemical smog has upon the work performance of the individual citrus
worker. It is obvious that any attempt to establish empirical values for
this influence requires that the expressions to- be estimated be explicitly
derived from an analytical statement of the picker's decision problem. It
is perhaps not so obvious that a complete analytical statement of the
picker's work performance requires some attention to the grower's decision
problem. The reason is that the picker's work performance is influenced
by certain of the choices the grower makes. In turn, these grower choices
are plausibly influenced in part by the grower's past observations on picker
work performance. Thus, at least initially, one must recognize the inter-
dependent nature of the two sets of parties' decision problems. Only then
can one legitimately consider making a set of assumptions that will form
the basis of the analytical model to be estimated. Sound judgment of the
value of what is ultimately retained relative to what has been cast aside
requires knowledge of the scope of this initial problem framework.
As noted in Chapter 2, with or without the intermediation of a labor
contractor or a grower cooperative, the picker-grower relationship can be
described as a sequence of spot contracts. The individual citrus picker
is an independent contractor who daily sells his labor services in response
to various combinations of piece-work wage offers, expected picking and
environmental conditions, and prospective hours of work. The product the
picker is selling is the number of boxes of fruit he picks within a given
time interval. His realized daily earnings are determined by his wage per
box of fruit picked, the relative ease of picking the fruit, and the number
of hours he is able to work. The relative ease of picking the fruit may
plausibly influence his innate productivity as well as the number of hours
he chooses to work. In either case, his realized earnings will be affected.
Just as pickers can trade-off reduced effort and gains in income, so can
growers substitute between fruit output and those grove conditions that enhance
-27-
-------
the ease of picking. Of course, fruit ouLput and certain grove conditions
(e.g., fruit density) that enhance picking are highly complementary. However,
other grower investment activities increase grower costs without contributing
anything to and perhaps even detracting from the amount of fruit grown. For
example, growers can reduce the height of their trees or clear the ground of
large stones so as to aid picking ease. In the extreme, a grower might
remove any tree from his grove once it reaches a height requiring a ladder
to pick its fruit. Younger and shorter trees yield less fruit, however. If
the grower reduces his fruit output in order to make life easier for the
pickers, he often reduces his gross revenues; but if he increases his fruit
output in order to increase his gross revenues, he sometimes makes life harder
for pickers. Making life harder for pickers requires, if they are to be
willing to accept the harder life, that the grower increase his costs by
increasing the piece-work wage rate for picking.
Figures la and lb below present the non-pecuniary essence of the
grower's long-term problem. In Figure la, the B-isoquants represent output
levels embodied in the citrus fruit production function. The citrus fruit
that will be hanging on the trees is influenced by the non-harvest labor
and capital, L and K applied to the grove, and the composite grove conditions,
G. G is an output as well as an input; that is, it is a product the grower
sells to the picker in exchange for reduced piece-work wage rates as well
as being a determinant of fruit yields. Now, viewing G as an output rather
than an input, the line XX' shows, for a given stock of fully employed L and
K, the relation between composite grove conditions developed solely to ease
picking and non-harvest labor and capital devoted to improving fruit yields.
Alternatively, the X'-intercept can be taken as the origin. The labor and
capital devoted to producing grove conditions rather than fruit yields
increases as one moves to the left from the X'-intercept. Thus, when the
X'-intercept is taken as the origin, the XX' line simply indicates the
ratio of G to the amount of I,, K committed to the production of G.
Figure lb is derived, as indicated by the dotted lines, from the
intersections oC the Lsoqiuiutrf in Figure la wLlh XX'. Over llu- 015..
interval on the B-axis of Figure lb, an improvement in grove conditions
-28-
-------
Figure 4.1
Relationship Between Grove Conditions and Fruit Yields
G
X
X' L,K
(a)
(b)
-29-
-------
occurs jointly with an increase in fruit yields. For example, the careful
pruning of trees will both increase hanging fruit and make the harvesting
of fruit easier. The OB9 interval of Figure Ib corresponds to a movement
from X' to the tangency of B_ with XX' in Figure la. If, from the grower's
perspective, both fruit output and grove conditions command positive prices
(the first by increasing gross revenues and the second by reducing the
harvest costs) , the grower will never intentionally select a combination
of fruit output and grove conditions in the increasing portion of the
production possibility frontier in Figure Ib. He will instead select a
portion in the declining portion of this boundary because it is only in this
portion that the grower, in order to increase his revenues .by producing more
fruit, must at the same time increase his harvest costs by making picking
more difficult.
The implications of Figure 4.1 are unnecessarily complex for this
study. Little, if any, violence is done to the nature of the grower's
decision problem if the harvest problem is treated as being entirely
separable from decisions about investing in grove conditions and fruit yields.
From the perspective of the current harvest, all prior investments are
predetermined. Moreover, except for extreme circumstances where one decides
to harvest the fruit by bulldozing the trees, current harvest decisions have
little or no effect upon future fruit yields or grove conditions. Assuming
all growers to be net revenue maximizers, the representative grower's
harvest (i.e., short-run) decision problem can therefore be represented as:
(1) Max: n =» pB - bK - vL - C,
where:
(2) B = B(C), a concave function, and B >_ 0 >
and
(3) C = C(w,K,L,G,E) , C.., C > 0; C J 0.
J\ JL — W <
A subscript indicates a derivative taken with respect to the subscript, and
tr is the grower's net revenue from the harvest.
p is the constant daily selling price of a box of fruit.
B is the number of boxes of fruit actually picked by a picking crew.
b is the unit rental price of composite capital.
-30-
-------
K. is the number of units of composite capital the grower employs.
v is the hourly wage rate of nonpicking labor.
L is the man-hours of nonpicking labor.
w is the wage the individual picker receives for each box of fruit he
picks.
C is the grower's expected total wage bill for pickers. It is thus
the piece-work wage rate multiplied by the number of boxes of fruit
the grower expects to have picked. Consistent with the static state-
ment of the nature of the grower's harvest decision problem, the
speed with which picking occurs and thus the number of workers he
hires are presumed to be matters of indifference.
G is an invariant composite variable representing existing grove
conditions that influence the ease of picking. In a longer-run
setting, it would be a function of nonpicking labor, capital, and
environment, and their respective prices. It includes the quantity
of fruit hanging on the grower's trees.
E is a composite variable representing environmental conditions such
as air temperature and photochemical smog concentrations that may
influence the ease of picking. It is exogenous to the grower.
Expression (2) states that the number of boxes of mature fruit the
grower will have picked is a function of the total picker wage bill the
grower expects to have to pay. As (3) indicates, this total picker wage
bill depends upon the piece-work wage rate, grove and environmental conditions,
and the quantity of nonpicking labor and capital provided that it is
complementary to picking labor. It appears in practice, however, that the
provision of non-harvest labor and capital differs only trivally from one
grove to another. We therefore disregard it in subsequent discussion.
Upon substituting (2) and then (3) into (1), and partially differentiating
the result with respect to w, K, and L, one obtains the usual first-order
conditions. These conditions determine the net revenue maximizing values
w,* K,* and LA for the grower in terms of p, b, and v as well as the
parameters of B(.) and .€(.)• One of the conditions:
(4) IT = pB C - C =0,
w v c w w '
or
(4a) p - B'1 = CB
is a standard result. This expression states that short-run grower net
revenue maximization requires the marginal cost of fruit picking, C,,, to
b
be set equal to the selling price of a box of picked fruit.
-31-
-------
The value of w the grower chooses constitutes part of the picker's
decision problem. The daily decision problem the picker faces may be
stated as:
(5) Max: U(It,H) Uj > 0, UR < 0
subject to:
(6) I - w(G)-B (E,G,H+ - Z) + M = 0
(7) H(E,G,I(._1) + Z = H+
where U(.) is concave, all partial derivatives are continuous, and where:
I is the picker's daily consumption expenditures and savings.
11 For notational simplicity, it is assumed the picker works in
only one grove a day.
H is the daily number of hours the picker harvests fruit in a
particular grove.
H is the length of the picking crew's work-day in a particular
grove. The individual picker is unable to influence the length
of this work-day.
I , is the picker's earnings in the previous pay period.
Z is the leisure time the picker voluntarily takes when he
otherwise could have been working.
M is all nonplcking income accruing to the picker.
All other variables are defined as they were for the grower.
This formulation of the picker's short-run decision problem states that
he obtains utility, U, from income (or the physical goods and services that
income can buy) and that he receives disutility from work. Utility for each
day directly depends only on the level of earnings and the hours of work
during that day, although the hours of work may be influenced by earnings in
the previous pay period. The incentive effects, if any, of income and social
security taxes and minimum wages are disregarded.
The first constraint, (6), implies that the picker's daily consumption
expenditures and savings are exactly equal to his daily earnings from harvesting
citrus plus whatever outsldo. income he is able to obtain. Outside income, M,
is fixed for the day in question. The second constraint implies that the
daily number of hours the worker is able to pick cannot exceed the number of
hours that the crew to which he belongs picks. Time during which his crew
-32-
-------
picks but the worker does not pick is used by the worker to pursue leisure
activities from which he obtains positive utility.
Since the picker is unable to influence the length of the crew's work-day,
the above decision problem may be written as:
(8) L = U (It,H) - A [lt - w(G)-B [E,G,H(.)] + M] = 0
and the necessary conditions for an interior utility maximum are:
(9) L = U - X = 0
t t
(io) LH - UH - AWBH = o
(ii) L - i. - W(.)-B PE.G.HC.)! + M = o
AD *"*
Expressions (9) and (10) above represent, respectively, the marginal
utility of earnings and the marginal disutility of work presuming that the
opportunity to acquire earnings by working exists. Taken together, (9) and
(10) imply
which is the value of work to the picker and the rate at which in equilibrium
he is willing to substitute leisure for earnings. From (4) and (12),
simultaneous individual grower net revenue and individual picker utility
maximization thus requires that:
(13) Cg - wBH;
that is, the rate at which the grower's expected total wage bill changes in
response to changes in boxes picked must be equal to the value of work to
the picker.
From the individual picker's perspective, the left-hand side of (1.1)
is predetermined. Although this picker may have some trivial influence
upon C_, the thousands of citrus pickers available to growers in southern
15
California make it worthwhile for the individual picker to behave as if he
had no influence whatsoever. Each day the picker is considering whether or
not to work therefore, the picking opportunities available to him are
composed of a set of discrete points, one point for each grower, where the
coordinates of a point indicate the total earnings a picker can expect to
be paid by a grower in exchange for picking fruit over a work-day of given
length.
-33-
-------
Temporarily assume that all growers are identical in terms of their
grove attributes, except that their groves differ in size and therefore
require a greater expenditure of hours from a given number of men in order
to be picked. This implies that the piece-work wage rate will be constant
across groves and that the individual picker's earnings opportunities will
differ only according to the number of hours it will take his crew to pick
each grove. One can therefore construct an indifference function for the
individual picker showing the change in leisure necessary to compensate
him for a marginal change in perceived earnings opportunities from picking
while maintaining a constant level of utility.
At the beginning of any given day, the individual picker faces the
situation depicted in Figure 4.2(a). Each point in the figure represents a
picking or nonpicking earnings opportunity, one point to an opportunity.
Only the points on the Indifference function represent picking opportunities.
All others represent other types of jobs such as fruit loading, truck driving,
pruning, box repair, etc. In the situation depicted, U*, which lies on
the picking opportunity indifference function, is on the highest indifference
function passing through any of these points, and this point will be the
earnings opportunity the individual will choose for that day. If earnings
opportunities other than picking always lie below the picking opportunity
indifference function, the individual will choose to pick each and every
day, given that picking opportunities are available. However, if on some
days, nonpicking earnings opportunities become available that lie above the
picking opportunity indifference function, the individual will take the
alternative job rather than picking fruit. As for the picking opportunities,
they will change from day to day as the sizes of the groves ripe for picking
change. Over time therefore, one will observe the individual picking at
various points on his picking opportunity indifference function.
The above reasoning is not altered by the fact that grove attributes
are dissimilar across groves. This is because growers adjust piece-work
wage rates so that for any particular expenditure of his hours over the
picking day, the individual picker is led to expect his earnings will be
(nearly) equal from one grove to another. This means that as crew hours
-34-
-------
Figure 4.2
The Individual Picker's Choice of Earnings Opportunities
Daily $
per grove
M
24
.U*
Daily $
per grove
R
Q
Leisure
--Work
U*
Work
(a)
(b)
Leisure
24
-35-
-------
increase and if the picker works as long as his crew, he expects to be sliding
along the same indifference function as he moves from grove to grove. Thus,
for example, all groves having the attributes associated with U* in Figure 4.2 (a)
are expected by the picker to provide the same level of earnings for the
expenditure of H hours of his time.
After subtracting nonpicking income, the S curve in Figure 4.2(b) is
the mirror image of the picking opportunity indifference function in Figure
4.2(a). Since the indifference function has a negative slope throughout, the
slope of the S function is the negative of the individual picker's marginal
rate of substitution between earnings and leisure. It thus provides a constant
real income supply function which, because of the convexity of the picking
opportunity indifference function, has a positive slope. Since the S function
is a compensated supply function, it has no backward bending portion as do
ordinary labor supply functions. Once the individual has actually selected
a grove in which to pick, the S function also represents the number of hours the
picker is willing to supply the grower at different levels of earnings.
The above commentary has almost entirely concentrated upon the individual
picker's decision problem at the start of each work day. However, once he
has made his choice of a grove in which to pick, he may discover that his
initial perceptions were mistaken. For example, assuming that his first-stage
decision of whether or not to pick is not influenced by his expectations
about levels of air pollution, he may find, once he has started picking,
that his earnings are distressingly low because air pollution levels are
2
reducing his physical picking prowess- Similarly, he may find that the
piece-work wage rate being paid is imperfectly adjusted to grove attributes
so that his earnings for a given time expenditure are less than he had
been led to expect. These disappointments are reflected in the shift of
the supply function in Figure 4.2(b) from S to S'. If in spite of his
disappointments the picker continues to work as long as his crew, the
crass-hatched area RQUU* represents the additional income; required to return
the individual to his former indifference function. It is thus a measure of
the compensating surplus and Js representative of the social loss caused by
air pollution that attaches to this picker. However, since the picker is,
-36-
-------
by assumption, constrained to work the same number of hours as his crew, the
area overstates the compensation required if he were allowed to adjust his
hours downward. Upward adjustments of hours are infeasible because the
picker is institutionally constrained from working longer hours than does
his crew.
Without further information, economic theory does not permit prediction
of the combination of hours and earnings the picker will choose for his
adjustment. Nevertheless, assuming that leisure is not an inferior good
for the picker, the substitution and income effects of earnings changes
possess the same sign in our compensated supply function: we should observe
nonincreasing hours of work as the earnings of a picker are reduced.
-37-
-------
Footnotes: Chapter 4
1. Pickers whose earnings after a two or three month training period
frequently fail to meet the minimum wage are no longer permitted to work.
In the empirical work of the next chapter only pickers who have continued
to pick long after the training period are analyzed.
2. The assumption that the worker's choice among groves on any
particular day is independent of air pollution levels is fairly innocuous,
given the more-or-less constant distribution of expected air pollution
concentrations over the locale in which the picker is likely to have
picking opportunities. For example, air pollution magnitudes and magnitude
durations are unlikely to differ perceptibly in those areas of Upland
in which citrus is grown. Expected air pollution levels might influence
the picker-'s decision whether or not to pick at all; however, our empirical
efforts do not attempt to deal with this issue.
-38-
-------
Chapter 5
AIR POLLUTION AND EARNINGS: EMPIRICAL RESULTS
Standard least-squares estimating techniques are readily applied to the
second stage (the stage in which the individual has decided to do at least
some picking during a day) of the representation in the last chapter of the
picker's decision problem. Empirical implementation of this representation
requires information on earnings, the piece-work wage rate, boxes of fruit
picked, picker and crew hours worked, and grove and environmental attributes.
The objective is to estimate, with respect to changes in air pollution, the
changes in the individual picker's earnings when he is picking fruit and the
changes in the leisure lie voluntarily takes when he could have been picking
fruit. The maintained hypothesis is that picker earnings vary negatively,
and voluntarily taken hours of leisure vary positively, with respect to increases
in air pollution. This chapter deals with the impact of air pollution upon
earnings. The next chapter presents results for absenteeism.
Throughout the empirical investigation described in this and•the
succeeding chapter, the overriding criterion has been to arrive at estimated
expressions yielding unbiased, or at least consistent, coefficients for a
particular explanatory variable, the air pollution variable. Consistent
estimates of these coefficients allow us to make inferences about the
compensation in terms of earnings the picker requires to make him indifferent
to an increase in the level of.air pollution. They can also enable us to infer
the change in the picker's equilibrium hours worked with respect to changes
in air pollution.
One common source of bias in estimated coefficients is "data-grubbing."
In the words of Selvin and Stuart (1966, p. 21):
"... any preliminary search of data for a model, even when the
alternatives are ptredosignated, affects the probability levels of nil
subsequent testt? based on that model on the same data, and in no very
simple way, and also affects the characteristics of subsequent estimation
procedures. The only valid course is to use different data for testing
the model dredged from the first set of data."
In order to assure that the data used to test the hypotheses generated from
the analytical framework of the previous chapter are unsullied by any prior
-39-
-------
efforls to puL together an empirical model from a combination of analytical
and empirical investigations, the picking histories of four experienced
pickers who worked more-or-less continuously picking lemons over an entire
year were used for preliminary estimation. All "data-grubbing" for the
empirical results presented in this report was limited entirely to these four
pickers; that is, the estimates for all pickers other than the four issue from
virgin data.
Estimates of the Inverse Supply Function. The analytical framework
surrounding Figure 4.2 makes it convenient to estimate the inverse supply
function, the function in which earnings are determined by hours worked. The
coefficient attached to the air pollution variable in an empirical counterpart:
I - f (H, air pollution, other factors that shift S in Figure 4.2(b))
to the S function of Figure 4,2(b) will, for the sample observations on the
Dicker's work performance, then measure the shift of S averaged over all the
hours the individual picker worked.
For estimation purposes, several assumptions, in addition to those
embodied in the analytical framework, were imposed upon the stochastic form
of the above earnings expression. First, in the absence of unique directions
about functional form from the picker's decision model, all estimated
expressions have been specified in double-logarithmic form. Although no
formal comparisons were made, exploratory manipulations with the four prelim-
inary test pickers made it appear that the double-log form fit each picker's
data equally as well as arithmetic or semi-log forms. The double-log form
was ultimately selected because of its greater flexibility. In particular,
it permits the marginal effect of air pollution upon individual picker earnings
to be constant, decreasing, or increasing; it makes the coefficients of the
explanatory variables easily interpretable as constant elasticities; it
restricts the dependent variables to positive values; it reduces the influence
of extreme data values; and, finally, it may reduce heteroskedasticity.
Second, the picker was always assumed to work the same number of hours as
his crew. Thus only those observations in which the picker's work-day was the
same length as that of his crow were used for estimation purposes. This
implies that the picker does not view his hours-worked as a decision variable,
but rather accepts H as exogenously determined by the crew foreman through the
-40-
-------
foreman's choice, of H . This is equivalent to assuming that Z = 0 in (12).
The assumption was adopted for- two reasons: (1) to provide an estimate of
the response of earnings to air pollution independent of states in which
nonzero combinations of earnings and leisure could be chosen; and most
importantly, (2) to reduce possibilities of introducing bias into the
estimated air pollution coefficients because of failure to include a variable
relevant to explanation of the variations in the picker's earnings. As will
be shown in a later chapter dealing with absenteeism, our data does not permit
us to account for very much of the variation in the hours the picker chooses
to work. Since hours-worked is an integral element of the earnings expression,
an inability to explain very much of the variation in hours-worked
substantially increases the risk that a relevant variable, nonorthogonal to the
air pollution variable, is being neglected. The assumption that hours-worked
is exogenous rather than endogenous to the picker's problem neatly avoids this.
Third, the variable representing hours-worked could, because of the common
practice in the packinghouse crew records of rounding to the nearest half hour,
contain a relatively high degree of measurement error when the picker worked
for only a short time in a given grove. In order to correct for this possible
source of error, we assumed for the earnings expression that the picker always
worked at least two hours when he worked at all. The adoption of this
assumption required that all observations in which the picker worked less than
two hours be excised from the data used for estimation purposes.
Finally, an examination of the residual pattern of some ordinary-least-
squares regressions for the four preliminary test lemon pickers revealed a
definite drift of the residuals across time, even though each of these four
individuals were known to have been picking lemons for years. An obvious
means of ameliorating this is to introduce a calendar date variable into the
regression specification. This additional variable might capture the work
performance effects of selective picking as opposed to clean picking of lemon
groves. Workers engaged in lemon picking are required to use rings and pick
by color during most of the. multiple harvests in a lemon prove durinp a
calendar year. Inclusion of a variable representing calcMidar dale may capture
the effect of prior picking that has occurred in a grove or it may register
factors not explicitly recorded that do influence picking ease. Since all
-41-
-------
orange picking is clean picking, calendar date did not seem relevant.
Inspection of the residuals for ordinary-least-squares estimates of an
orange picker's earnings expression seemed to confirm this irrelevancy.
Table 5.1 below presents the acronyms of the variables used to estimate
the earnings expression. The initial ordinary-least-squares estimates of the
earnings expression for the four preliminary test pickers revealed several
additional statistical problems. As has been stressed in previous pages, our
primary concern was to obtain consistent estimates of the coefficient for
the air pollution variable. However, if it is not statistically confirmed
that the air pollution variable is significantly different from zero at the
usual test levels, then any claims as to the effect of air pollution upon the
performance of citrus pickers is unfounded. As is well known, collinearity
inflates the standard errors of the set of explanatory variables. In turn,
this implies a reduction in the t-statistics and an unnecessarily conservative
test of significance. Different equation specifications for the four
preliminary test pickers verified that TM and DE were relevant explanatory
variables and consequently must be included in the empirical specification.
Additionally, both were collinear with the air pollution variables and thus
rendered difficult an interpretation of the levels of statistical significance
of these variables.
Tables 5.2 and 5.3 provide detail on the extent of collinearity between
the air pollution variables and temperature. Each picker is identified by the
general locale in which he picked fruit as well as by a number immediately
following the locale. The parenthetic numbers indicate the years to which
data for the picker refer. Thus Upland 1 (1973) refers to the same individual
as Upland 1 (1974), but the year from which the data is drawn is different.
Those to whom we refer to as pickers are thus on occasion the same individual
distinguished by year and/or crop. A distinction was made between crop years
for the picking activities of the same individual because the time pattern of
the fruit harvest is said by growers and packinghouses to have differed fairly
substantially between 1973 and 1974.
-42-
-------
Table 5.1
Glossary of Variable Names
B - The number of 3115 cubic inch field boxes picked by the picker during
the work-day in a particular grove.
BT - The average number of 3115 cubic inch field boxes picked per tree during
the work-day by the picker's crew in a specific grove.
DE - The calendar date of the work day: Jan. 1=1; Dec. 31 = 365.
I - The picker's daily gross earnings in dollars from picking activities for
each grove worked.
I - The picker's gross earnings in dollars from picking activities in the
t- previous pay period.
H - The number of hours the picker spent in picking activities in each
grove.
FR - The number of fruit from the grove required to fill a 3115 cubic inch
field box.
0Z - The arithmetic average 24-hour ambient concentration on the work clay of
0~ in parts per million by volume as measured by the CHEMILUM method
ac the monitoring station closest to the grove site.
0ZH - The arithmetic average of the hourly ambient concentrations of 0 occurring
during the time interval the worker was actually engaged in citrus picking.
This variable is also measured by the CHEMILUM method at the monitoring
station closest to the grove site.
TM - The maximum hourly arithmetic average dry-bulb temperature in F on the
work-day at the monitoring station closest to the grove site.
TR - An index indicating the height of the trees picked by the worker's crew
during the work day.
1 = tree can be picked without a ladder.
2 = ladder picked trees up to 9 1/2 feet tall.
3 = ladder picked trees 9 1/2 to 12 feet tall.
4 = ladder picked trees in excess of 12 feet tall.
w - The rate-of-pny (in dollars x 10) the picker receives for each 3115 cubic
inch field box of citrus he picks.
-43-
-------
Table 5.2
Simple Correlation Coefficients Between 0Z,
0ZH, and TM for Various Lemon Pickers.
Worker — '
Upland 1 (1973)
Upland 1 (1974)
Upland 2 (1973)
Upland 2 (1974)
Upland 3 (1973)
Upland 4 (1973)
Upland 22 (1974)
Santa Paula 10 (1973)
Santa Paula 10 (1974)
Santa Paula 11 (1973)
Santa Paula 11 (1974)
0Z.0ZH
.735
.643
.737
.658
.711
.752
.558
N.A.
N.A.
N.A.
N.A.
0Z.TM
.744
.753
.749
.747
.749
.835
.753
.478
.486
.472
.453
0ZH . TM
.588
.555
.590
.557
.593
.653
.554
N.A.
N.A.
N.A.
N.A.
N.A. - Not Available,
Table 5.3
Simple Correlation Coefficients Between 0Z, 0ZH,
and TM for Various Orange Pickers
Worker — '
Upland 2 (1973)
Upland 4 (1973)
San Bernardino 5 (1973)
San Bernardino 7 (1973)
Irvine 38 (1974)
Irvine 39 (1974)
Irvine 40 (1974)
0Z.0ZH
N.A.
N.A.
.824
.747
N.A.
N.A.
N.A.
0Z.TM
.688
.703
.763
.751
.223
.205
.223
0ZH.TM
N.A.
N.A.
.654
.585
N.A.
N.A.
N.A.
N.A. - Not Available
-44-
-------
For both lemon and orange pickers, the collinearity between air pollulion
and temperature is somewhat less for 0ZH Chan for 0Z. Whenever available
data allow it, the former measure of air pollution is used in the estimated
expressions. Some collinearity between the temperature and air pollution
variables was, of course, expected. The standard way to resolve this
complication is to obtain an unbiased estimate of the temperature coefficient
from an extended sample, or from a sample where the correlation between the
collinear variables is less severe, and then to subtract the temperature term
from the dependent variable, thus forming a new regression specification.
Because air pollution and temperature in the Ventura region do not appear to
be highly correlated, it would seem that data collected from that region would
be ideal for this purpose. However, after considerable consternation, we
decided not to follow this procedure. Our reasons were two. First, it was
felt that the transformation of the dependent variable would alter the
interpretation of the estimated coefficients. It is not clear to us, in
terms of the analytical framework of Chapter 4, what the use is of a measure
of Che compensating surplus for air pollution where the picker has already
received the compensation required to make him indifferent between existing
temperatures and some hard-to-identify temperature he regards as ideal for
his citrus fruit picking activities. Second, and perhaps more important, the
temperature coefficient used to transform the earnings variable for one
individual would, of necessity, be the coefficient estimated for another
individual or set of individuals. It is generally acknowledged that responses
to identical perturbations in meterological and environmental variables can
differ greatly across individuals. The possibility of introducing bias into
the other estimated coefficients, particularly the air pollution coefficients,
therefore seemed, in our judgment, to be excessive. We thus instead chose to
present regression results where the temperature variable is both included and
excluded, leaving it to the reader to judge for himself where the "true" level
of significance lies.
With respect to Lho calendar date variable, we attempted to mollify the
collinearity problem by viewing the system as recursive. Specifically, we
hypothesized the following pair of expressions:
-45-
-------
(15a) Lw = f(LBT, LFR, LTR, LDE),
I" " "7
(15b) LI = gJLw, LH, L (air pollution), LTM, L (grove attributes)]
where L denotes the natural logarithm of the original arithmetic value of
the variable. (15a) is' interpreted as the. piece-work wage rate faced at the
beginning of each day by the crew of the individual picker, and, given that
the individual has chosen to pick during that day, (15b) is the picker's actual
earnings expression. The expression of empirical interest, (15b), can be
estimated by the two-stage-least-squares method. Assuming the usual classical
conditions for the general linear model hold, consistent estimates of coefficient
for the explanatory variables will be obtained.
The adoption for lemon pickers of the system represented by (15a) and
'(15b) introduced an additional collinearity problem. In particular, a linear
combination of the grove attributes included as arguments in the earnings
expression is highly collinear with the estimated rate-of-pay variable also
appearing on the right-hand side of (15b). Theoretically, the inclusion of
all these variables is required; but a specification of this sort reduces the
rank of the data matrix below that required for satisfaction of the order
condition for identification. Consequently, it was necessary to delete one
of the grove attribute variables from (15b). -The correlations of the various
grove attribute variables with the air pollution variables served as our
principal guide in determining the best variable to delete from the second
structural equation. It can be shown that exclusion of a potentially relevant
but orthogonal variable will not bias the air pollution coefficient, although
it will increase the standard error. For the four preliminary test pickers,
a review of the simple correlation coefficients revealed that the tree height
variable, TR, was relatively uncorrelated (approximately -.06) with the air
pollution variable for three of the pickers. For one picker the simple
correlation between the two variables was high; however, it was also positive,
even though the citrus industry universally expects, for given piece-work wage
rates, that tree height and earnings per grove picked will vary inversely.
These two facts for this single picker (the high simple correlation and its
positive sign), imply that the bias imparted to the air pollution coefficient
of this picker would be negative. All these considerations for the four
-46-
-------
preliminary test pickers led us to delete TR for all pickers in all two-stage-
least-squares regression specifications.
Apart from the variables explicitly introduced through the theoretical
model, conversations with various labor camp managers produced suggestions
about potentially relevant factors. It appears to be a widely held notion that
the performance of the typical citrus harvest worker will markedly decrease on
Fridays and Mondays. Expectations of a wild weekend and supposed fulfillment
of these expectations are offered as a rationale for this occurrence. For the
four workers initially surveyed, however, estimated coefficients for dummy (0,1)
variables for Friday and Monday did not yield estimates significantly different
from zero at the usual levels. Consequently, these variables were not included
2
in our final specification.
A second common observation is that many pickers set an earnings goal and
will not Vork as productively once this goal is achieved. Since pickers
receive weekly paychecks, one way of ascertaining the validity of this
hypothesis is to include a measure of the worker's total earnings in previous
weeks. Inclusion of such a variable in the two-stage-least-squares formulation
for the four preliminary test workers did not yield statistically significant
estimates. Hence, this variable was also not included in the estimated
expressions for other workers.
Finally, labor camp managers believe that multiple groves worked in
a day seriously impairs the productivity of the worker. It is thought that
moving three or four times a day causes the worker to go through three or
four "warm up" periods, thus slowing down his picking output. Again, inclusion
of a single explanatory variable representing number of groves picked did not
yield a coefficient significantly different from zero.
It should be mentioned that although none of the three variables alluded
to above proved to be statistically significant, the expected signs were in
fact obtained.
We have referred .several times to "the air pollution variables" in Ulu:
preceding discussion without explicitly stating in each circumstance what
measure we arc postulating. Ideally, one would like the pollution monitoring
stations to be located in each and every grove, with hourly readings having a
one-to-one correspondence with hourly picking performance. Unfortunately, the
-47-
-------
picking data cannot be disaggregated to this extent. Moreover, we also must
rely on the hourly readings from the closest recording station.
In some instances, we were unable to obtain the start and stop times
for the picking crews for the various groves, and in these cases we employed
the arithmetic average 24-hour ambient concentration of air pollution in
parts per million by volume as measured by the CHEMILUM method. For the
majority of workers for whom we were able to ascertain the actual time
period of the day during which they picked fruit, we used the arithmetic
mean of the hourly concentrations during this time span as our measure of
air pollution.
Several alternative characterizations of the air pollution measure can
be hypothesized. These might include higher moments (e.g., the variance)
or perhaps some distributed lag structure. We have not attempted a
distributed lag specification, but inclusion of the variance of the air
pollution measure has not proven to be significant in various trial runs
for the four preliminary test pickers.
The reader is by now no doubt aware that we are aware of the likely
existence of measurement error in the air pollution variable. Since it does
not appear possible to identify any systematic deviations in the values of
this variable, it would seem that an instrumental variable would be our best
recourse. The most likely candidate for an instrument, among those variables
available to us, would be temperature, TM. However, this variable is already
included in the regression specification. The next best alternative was
posited to be TM lagged one period. A sample run for two of our trial
pickers using the maximum temperature of the previous day as a proxy for the
actual air pollution during the period of the next day in which the worker
picked did not produce any interesting results. Given that the correlation
between this lagged temperature and air pollution was only about 0.70, little
gain from this reformulation could be expected. In the results to be
presented below, the air pollution variable itself is utilized.
This completes the description of the basic model specifications used for
estimation. The empirical specifications finally settled upon for the four
preliminary test pickers were carefully checked for conformity with the
-48-
-------
c.la:TiieaJ linear model. An effort was, in fact, made to rmnJoy a program
package for the formal Ramsey (1969) tests for specification error. Since
time and available resources did not permit the correction of package
programming errors, resort was had to less formal means. Heteroscedastic
disturbances were searched for by evaluating scatter diagrams of residuals
versus values of the dependent variable. No heteroscedasticity appears to
be present in the final estimated expressions for the four preliminary test
pickers. All variables for which data was available and which might plausibly
be nonorthogonal to the air pollution variables were included at one time or
another in specifications for the preliminary test workers. All those not
found wanting in terms of statistical significance were included in all
subsequent specifications. Autocorrelation was evaluated by means of the
Durbin-Watson statistic. In the multiplicative form of the earnings
expression finally selected, checks were made to assure that the disturbances
were at least approximately log-normally distributed.
This completes the description of how we arrived at the basic specifi-
cations used for estimation of the earnings expression. Estimates for these
expressions appear in Tables 5.4 and 5.5. Note that with one statistically
insignificant exception for lemon pickers and two statistically insignificant
exceptions for orange pickers, the coefficients for tlie air pollution variables
are consistently of tin.- expected sign. Moreover, even though a col.Linear
temperature variable is included, the coefficients for almost half (seven of
eighteen) the pickers are statistically significant at traditional levels for
non-rejection of the maintained hypothesis that air pollution has a detrimental
influence upon picker earnings. An indication of tlie impact that collinearity
between the air pollution and temperature variables has upon the statistical
significance of the former can be obtained from Table 5.5, where the temperature
variable has been deleted from the empirical specification. In Table 5.5, all
but two air pollution coefficients (both for orange pickers) have the expected
negative sign and these negative coefficients are statistically significant
for thirteen of the eighteen pickers. A further comparison of Table 5.5 with
Table 5.4 makes it appear that the impact of the deletion of the temperature
variable upon the sl.ii lsl.u:.il signif i enure of the1 air poll.uti.mi e:oe f I ic I rnl s
in Table 5.5 was greulcj- Tor those pickers having nonsignificant air pollution
-49-
-------
Table S.4A
Earnings Estimates by Two-Stage-Least-Squares for
Lemon Pickers. Dependent Variable * LI..
i
en
o
i
Picker:
Variable
Constant
Lw
LH
LFR
LBT
LOZ
LOZH
LTH
-2
R
S.E.
D-H
F
Sample
Size
Sample
Period
Upland la
(1973)
1.326
(1.680)
-0.207
(0.472)
1.084
(0.041)
0.073
(0.297)
-0.018
(0.108)
-0.015
(0.022)
-0.173
(0.175)
0.824
0.246
1.88
131.46
168
March 17-
Dec. 21
Upland 1
(1974)
1.270
(1.210)
-0.780
(0.423)
1.080
(0.039)
0.268
(0.219)
-0.204
(0.114)
-0.023
(0.024)
0.028
(0.148)
0.856
0.209
2.09
153.18
162
April 1-
Nov. 2
Upland 2a
(1973)
3.180
(1.740)
-0.475
(0.479)
1.010
(0.046)
-0.092
(0.342)
0.113
0.108
-0.038**
(0.021)
-0.164
(0.186)
0.824
0.287
1.79
143.24
189
March 19-
Oec. 21
Upland 2
(1974)
4.713
(1.306)
-0.509
(0.404)
1.028
(0.043)
-0.129
(0.221)
-0.067
(0.111)
-0.017
(0.026)
-0.414**
(0.158)
0.822
0.206
1.86
120.52
156
April 1-
Nov. 2
Upland 3
(1973)
0.831
(1.502)
-0.604
(0.493)
1.021
(0.043)
0.255
(0.359)
-0.070
(0.359)
-0.029*
(0.018)
* -0.085
(0.178)
0.849
0.217
1.92
127.87
136
March 17-
Dec. 20
Upland 4b
(1973)
-2.220
(1.570)
1.610
(0.409)
0.960
(0.047)
0.107
(0.329)
0.339
(0.098)
-0.012
(0.018)
-0.178
(0.172)
0.888
0.176
1.70
124.86
101
March 16-
Dec. 21
Upland 22 Santa Paula 10a
(1974) (1973)
0.214
(1.700)
0.830
(0.594)
1.074
(0.049)
-0.199
(0.349)
0.243
(0.162)
-0.047*
(0.031)
-0.037
(0.190)
0.822
0.257
1.82
104.74
143
April 17-
Nov. 2
0.931
(2.588)
-0.270
(0.604)
0.986
(0.103)
0.388
(0.354)
-0.314
(0.102)
-0.242**
(0.128)
-0.266
(0.455)
0.640
0.304
1.68
15.11
58
May 14-
July 11
Santa Paula 10 Santa Paula lla
(1974) (1973)
0.842
(0.523)
-0.122
(0.163)
1.070
(0.023)
0.068
(0.129)
-0.119
(0.040)
-0.034***
(0.016)
-0.042
(0.073)
0.877
0.169
1.94
381.42
293
Jan. 3-
Nov. 20
3.800
(1.490)
-0.473
(0.889)
1.150
(0.055)
-0.002
(0.289)
-0.155
(0.145)
-0.029
0.074
-0.351
(0.295)
0.927
0.174
1.72
99.92
54
May 13-
July 11
Santa Paula 11
(1974)
1.650
(0.495)
-0.320
(0.122)
1.060
(0.021)
0.120
(0.103)
70.106
(0.026)
0.002
(0.019)
-0.142***
(0.070)
0.912
0.155
1.57
456.11
264
March 3-
Dec. 4
-------
Table 5.4A
(continued)
The results reported for this picker have been derived from various empirical testing
procedures and regression specifications.
This picker is a woman.
Levels of significance are explicitly shown for only the air pollution and temperature
variables.
*Coefficient is significantly different from zero at the 0.10 level of the one-tailed
t-test.
**Coefficient is significantly different from zero at the 0.05 level of the one-tailed
t-test.
***Coefficient is significantly different from zero at the 0.025 level of the one-tailed
u, t-test.
i—'
i
(The numbers in parentheses are the standard errors of the estimated coefficients).
-------
Table 5.4B
Earnings Estimates by Ordinary-Least-Squares for
Orange Pickers.^ Dependent Variable = LI
Picker:
Variable
Constant
Lw
LH
LTR
LET
L0Z
L0ZH
LTM
-2
R
S.E.
D-W
F
Sample
Size
Sample
Period
Upland 2
(1973)
2.210
(2.062)
0.290
(0.288)
1.371
(0.109)
0.065
(0.075)
-0.567
(0.043)
0.755
0.181
1.276
44.094
57
June 18-
Sept. 9
Upland 4
(1973)
2.094
(1.325)
0.323
(0.181)
1.114
(0.686)
-0.009
(0.047)
-0.468**
(0.276)
0.864
0.110
2.043
77.782
54
June 20-
Sept. 8
San Bern. 5
(1973)
0.898
(1.285)
0.119
(0.178)
0.884
(0.077)
-0.162
(0.112)
-0.043
(0.051)
-0.061***
(0.030)
0.028
(0.283)
0.487
0.410
1.800
24.724
163
Feb. 29-
Dec. 31
San Bern. 7
(1973)
-0.786
(1.363)
1.189
(0.252)
1.029
(0.090)
0.082
(0.167)
0.150
(0.095)
-0.054
(0.053)
-0.226
(0.312)
0.814
0.397
1.502
127.562
152
Mar. 4-
Oct. 18
Irvine 38
(1974)
3.583
(1.190)
a
1.001
(0.067)
-0.081*
(0.059)
-0.568*
(0.270)
0.692
0.308
1.328
82.392
114
Apr. 4-
Aug. 28
Irvine 39
(1974)
1.279
(1.282)
a
1.146
(0.062)
-0.027
(0.062)
-0.096
(0.293)
0.759
0.332
1.557
116.543
115
Apr. 4-
Aug. 28
Irvine 40
(1974)
2.387
(1.175)
a
1.172
(0.066)
0.065
(0.057)
-0.366*
(0.265)
0.756
0.303
1.111
113.519
114
Apr. 4-
Aug. 28
-------
Table 5.4B
(continued)
Jlate-of-pay in 1974 for Irvine is a constant 32 cents per box since grove conditions
are very nearly uniform.
For orange pickers, there is no a priori reason to think that grove attributes contributing
to picking ease vary systematically by calendar date. Thus LDE is not a relevant explanatory
variable and two-stage-least-squares estimating procedures were not necessary to avoid the
collinearity problem between LDE and the environmental variables.
*Coefficient is significantly different from zero at the 0.10 level of the one-tailed t-test.
**Coefficient is significantly different from zero at the 0.05 level of the one-tailed t-test.
***Coefficient is significantly different from zero at the 0.025 level of the one-tailed t-test,
(The numbers in parentheses refer to the standard errors of the estimated coefficients).
-------
Table 5.5A
Earnings Estimates by Two-Stage-Least-Squares for tenon
Pickers When LTH Is Deleted. Dependent Variable = LIt-
i
en
Picker:
Variable
Constant
Lw
LH
LFR
LBT
LOZ
LOZH
-2
R
S.E.
D-W
f
Sample
Size
Sample
Period
Upland 1
(1974)
1.410
(1.020)
-0.772
(0.411)
1.080
(0.039)
0.260
(0.214)
-0.201
(0.111)
-0.021
(0.020)
0.856
0.208
2.090
185.190
162
Apr. 1-
Nov. 2
Upland la
(1973)
1.177
(1.515)
-0.159
(0.458)
1.088
(0.041)
0.041
(0.291)
-0.002
(0.103)
-0.026**
(0.018)
0.826
0.244
1.89
159.69
168
Mar. 17-
Dec. 21
Upland 2 Upland 2a
(1974) (1973)
2.731
(1.109)
-0.640
(0.404)
1.035
(0.044)
-0.035
(0.221)
-0.109
(0.111)
-0.058***
(0.021)
0.814
0.210
1.81
136.84
156
Apr. 1-
Nov. 2
2.620
(1.610)
-0.448
(0.472)
1.020
(0.045)
-0.127
(0.335)
0.124
(0.005)
-0.046**
(0.019)
0.824
0.286
1.470
171.860
189
Mar. 19-
Dec. 21
Upland 3
(1973)
0.662
(1.424)
-0.517
(0.469)
1.247
(0.041)
0.188
(0.329)
-0.048
(0.105)
-0.032**
(0.017)
0.857
0.211
1.901
162.63
136
Mar. 17-
Dec. 20
Upland .4
(1973)
-2.226
(1.479)
1.606
(0.409)
0.963
(0.047)
0.107
(0.329)
0.342
(0.012)
-0.031**
(0.013)
0.888
0.301
1.704
137.842
101
Mar. 16-
Dec. 21
Upland 22 Santa Paula 10* Santa Paula 10 Santa Paula 11 Santa Paula lla
(1974) (1973) (1974) (1974) (1973)
0.029
(1.394)
0.816
(0.586)
1.074
(0.049)
-0.189
(0.342)
0.239
(0.160)
-0.050**
(0.026)
0.822
0.256
1.827
126.644
143
Apr. 17-
Nov. 2
0.862
(2.741)
-0.291
(0.610)
0.987
(0.103)
0.426
(0.352)
-0.325
(0.102)
-0.256***
(0.110)
0.619
0.304
1.67
14.14
58
Hay 14-
July 11
0.736
(0.446)
-0.112
(0.162)
1.070
(0.023)
0.053
(0.129)
-0.114
(0.040)
-0.038***
(0.015)
0.887
0.169
1.930
327.020
293
Jan. 3-
Nov. 20
0.046
(0.217)
-0.428
(0.160)
1.124
(0.020)
0.130
(0.110)
-0.158
(0.028)
-0.088***
(0.034)
0.894
0.163
1.42
484.79
264
Mar. 3-
Oec. 4
0.029
(1.090)
-0.850
(0.881)
1.140
(0.057)
0.095
(0.295)
T0.200
(0.149)
-0.072
(0.069)
0.916
0.186
1.470
104.390
54
May 13-
Ouly 11
-------
Table 5.5A
(continued)
The results reported for this picker have been derived from various empirical testing
procedures and regression specifications.
Coefficient is significantly different from zero at the 0.10 level of the one-tailed
t-test.
^Coefficient is significantly different from zero at the 0.005 level of the one-tailed
t-test.
***Coefficient is significantly different from zero at the 0.025 level of the one-tailed
<-" t-test.
Ln
I
(The numbers in parentheses are the standard errors of the estimated coefficients).
-------
Table 5.5B
Earnings Estimates by Ordinary-Least-Squares for Orange Pickers
When LTM is Deleted.b Dependent Variable = LI .
I
-------
Table 5.5B
(continued)
Rate-of-pay in 1974 for Irvine is a constant 32 cents per box since grove conditions
are very nearly uniform.
For orange pickers, there is no a. priori reason to think that grove attributes contributing
to picking ease vary systematically by calendar date. Thus LDE is not a relevant explanatory variable
and Cwo-stage-least squares estimating procedures were not necessary to avoid the collinearity
problem between LDE and the environmental variables.
*Coefficier.t is significantly different from zero at the 0.10 level of the one-tailed t-test.
**Coefficient is significantly different from zero at the 0.05 level of the one-tailed t-test.
***Ccefficient is significantly different from zero at the 0.025 level of the one-tailed t-test.
Oi
-•4
1 (The numbers in parentheses refer to the standard errors of the estimated coefficients).
-------
coefficients but relatively significant temperature coefficients in Table 5.A.
Of course, since we have no reason to suppose that temperature is an irrelevant
explanatory variable and since temperature is obviously nonorthogonal to air
pollution, only the air pollution coefficients in Table 5.A should be viewed as
unbiased estimates. Nevertheless, since the air pollution coefficients of
Table 5.5 generally do not exhibit major change from those in 5.A, one can
tenatively and somewhat hesitantly conclude that, over the observed ranges of
variation of the two variables, air pollution has greater relevance to picker
3
earnings than does temperature.
One must temper the generalizations of the above paragraph with the
observation that the estimated equations for three orange pickers (Upland 2
(1973), Irvine 38 (1974), and Irvine AO (197A)) have Durbin-Watson statistics
probably indicative of negative autocorrelation of disturbances. It therefore
seems likely that the estimated expressions for these individuals, as presented
in Tables 5.A and 5.5, are misspecified. More will be said shortly about
the sources of this specification error.
Finally, when reviewing Tables 5.4 and 5.5, the careful reader will have
noted that the three grove attributes (BT, FR, TR) used by the packinghouses
to determine the piece-work wage rate are frequently statistically non-
significant and often seemingly of the wrong sign. However, because the
piece-work wage rate is adjusted in accordance with changes in the values of
these grove attribute variables, one cannot interpret their coefficients in
the customary manner. Instead of representing the response of the picker's
earnings to changes in the grove attributes variables, the coefficients
represent the deviation in the individual picker's adjustment to the change
from the adjustment to the grove attribute reflected in the piece-work wage
rate. If this rate were always adjusted perfectly for the individual picker,
the coefficients attached to the grove attribute variables would each be
zero; that is, any change in one or more of the three grove attributes would
have no effect whatsoever upon tiie picker's earnings.
Increases in Income Required to Compensate Pickers for Earnings Losses
Due to Air Pollution. For expressions in which tfie temperature variable lias
been included, Table 5.6 below presents the calculated effects of air
pollution upon the daily earnings of those lemon and orange pickers for whom
-58-
-------
Table 5.6
Required Picker Income Compensations'
Picker:
Statistic
n
«t
E0Z
E0ZH
I
0Z
0ZH
b
V in dollars
V in dollars
nV in dollars
/ nV \
= 1100
v I +nV/
t '
Lenons
Upland 2
(1973)
129
2899.30
1291.30
15.34
6.83
-0.038
-0.090
-0.280
-52.92
-1.8%
Lemons
Upland 22
(1974)
143
1645.93
1403.40
11.51
9.87
-0.047
-0.055
-0.105
-15.02
-0.9%
Lemons
Upland 3
(1973)
136
1489.07
895.30
10.95
6.58
-0.029
-0.048
-0.144
-19.58
-1.3%
Lemons
Santa Paula
10 (1973)
58
821.86
262.1
14.17
4.52
-0.242
-0.760
-0.900
-52.20
-6.0%
•Lemons
Santa Paula
10 (1974)
293
4063.85
13.87
3.38
-0.034
-0.140
-0.380
-111.34
-2.7%
Oranges
San Bern. 5
(1973)
163
1886.23
1064.06
11.57
6.528
-0.061
-0.108
-1.138
-185.49
-9.0%
Oranges
Irvine 38
(1974)
114
2063.40
475.95
18.1
4.175
-0.081
-0.351
-0.446
-50.84
- 2.5%
These calculations were made from estimated expressions in which it was assumed the picker's
work-time was institutionally fixed.
-------
statistically significant air pollution coefficients have been obtained. A.
table showing the same calculations when the temperature variable has been
excluded is not presented because the air pollution coefficient in these
expressions is thought to be biased; that is, since temperature is known to
be nonorthogonal to air pollution and since we are unable to show that
temperature is an irrelevant explanatory variable, calculations of required
compensation using statistical results that do not account for temperature
would be highly untrustworthy. The sole purpose of presenting estimates for
expressions in which temperature is deleted has been to provide the reader
a sense of the extent to which collinearity between temperature and air
pollution affects the standard error and thus the statistical significance
of the air pollution coefficient in estimated expressions including both
variables.
Table 5.6 does not Include pickers with statistically insignificant
air pollution coefficients because we are able to reject, for these pickers
only, and only within the context of the particular empirical specification,
the hypothesis that air pollution influenced their earnings. For
those pickers exhibiting statistically significant air pollution coefficients,
the calculated losses represent, in accordance with the analytical construct
presented in Chapter 4, the compensation the picker requires to make him
indifferent between the presence or absence (except for "background" levels)
[ of photochemical oxidants, given that he works as long as his picking crew.
;t!!!|( In Table .5.6, nV represents this total required compensation for the picker
•'||| during the period of observation and the bottom row of figures shows this
required compensation as a percentage of what the picker's earnings would
have been in the absence of air pollution. Thus, for example, in the 293
lemon groves in which Santa Paula 10 picked from January 3, 1974, to
November 20, 1974, he required in compensation 38 cents per grove that he
picked, $111.34 in total, and 2.7% of what his income would have been in the
absence of air pollution.
Table 5.6 actually contains two calculations of picker's required
compensations. Both assume that the response of the picker's earnings to
variations in air pollution is a constant. The first calculation, V, is
111'1
-'''
-60-
-------
arithmetic mean of I
arithmetic mean of 0Z or 0ZH 0Z or 0ZH
where b is the estimated coefficient of the air pollution variable. This
calculation gives the picker's required income compensation for the average
grove he picks. The second calculation, V, is the picker's required income
compensation per grove that he picked during the period of observation. It is;
V =
i=l ith air pollution observation
where the i subscript indexes the groves in which the worker picked and n is
the number of groves. Only the dollar magnitudes and percentages associated
with V are presented in Table 5.6 because V takes greater account of the
peculiarities of each grove in which the picker has worked. Calculations of
required compensations that use V rather than V will obviously give lower
dollar and percentage magnitudes.
Does Air Pollution Impact Vary with the Picker's Physical Condition?
+
The picking of citrus fruit is a physically strenuous activity, giving
reason to speculate that over relatively long work-days the picker will
become fatigued and therefore be more susceptible to the deleterious effects
4
of air pollution. However, the results reported in Tables 5.4 and 5.5
reflect the impact of air pollution on the earnings of various pickers for
a wide range of work-day lengths. By including this entire range of
work-day lengths in the sample used for each estimate in Tables 5.4 and 5.5,
one obtains air pollution coefficients representing weighted averages of
the picker's responses over all work-day lengths. In the absence of further
analysis, it is impossible to disentangle the separate contribution to these
weighted averages of assorted work-day lengths. Furthermore, it could be
that the failure oC the procedures used in Tables 5.4 niul 3.5 to consider
the differential effect of hours worked upon air pollution impact may, for
certain pickers, have incorrectly resulted in statistical rejection of the
hypothesis that air pollution influences picker earnings.
-61-
-------
The results exhibited in Table 5.7 .ire the air pollution coefficients
obtained by running the exact specifications of Tables 5.4 and 5.5 on
partitionings by hours worked of the identical observations of picker
performance used in Tables 5.4 and 5.5. No Irvine workers are included
in Table 5.7. They will be discussed separately. The results included in
Table 5.7A include temperature as an explanatory variable, while those in
Table 5.7B do not. Although the covariance F-test for single coefficients
developed by Tiao and Goldberger (1962) could be used to test for statistically
significant differences in the air pollution coefficients across partitionings,
time did not permit the completion of this task for this report. A glance
at the differences in magnitude among many of the coefficients for single
pickers nevertheless leaves little doubt that statistically significant
differences are fairly common.
The speculation that picke-r responsiveness to air pollution increases
directly with the length of the work-day receives some support from the
'!?" results exhibited in Table 5.7A. For ten of the fifteen pickers in the
';*"!: table, the air pollution coefficient, an elasticity coefficient, increases
p?: in negative magnitude with increases in the length of work-day. Indeed,
v given the near-universal lack of significance (Upland 4 (1973) in lemons
is the sole exception) of the air pollution coefficients in the
partitionings representing relatively short work-days, it is tempting to
assert that air pollution has little, if any, impact unless the work-day
is in excess of about six or seven hours. The results of Table 5.7A are
at least consistent with this interpretation.
The apparent tendency of air pollution impact to increase with increased
work-day length is associated with the most interesting and important
feature of Table 5.7A: with the exceptions of Upland 2 (1973) in oranges
and Upland 4 (1973) in oranges, all pickers for whom statistically non-
significant air pollution coefficients were obtained in Table 5.4 now have
significant air pollution coefficients for the work-day partitioning grcater
than or equal to seven hours. In fact, it is plausible that the failure of
Upland 4 (1973) in oranges to be significant is due to collinearity between
temperature and air pollution. Note that in Table 5.7B, where temperature
-62-
-------
Table 5.7A
Air Pollution Coefficients (and Standard Errors) for
H Partitionings. Dependent Variable = LI .
Picker
Upland 1 (1973)
Upland 1 (1974)
Upland 2 (1973)
Upland 2 (1974)
Upland 2 (1973)
Upland 3 (1973)
Upland 4 (1973)
Upland 4 (1973)
Upland 22 (1974)
Fruit
Lemons
Lemons
Statistic
A
bL02H
s
n
\0ZH
s
n
Lem°nS bL02H
5
n
Lemons
\0ZH
s
n
Oranges, b
s
n
Lemons
Lemons
i
bL0ZH
n
hL0ZH
s
n
Oranges b^
, . n
I
Lemons b M
L0ZH
, s
.' n
2.0
-------
Table 5.7A
(continued)
"
Picker
Santa Paula 10 (1973)
Santa Paula 10 (1974)
Santa Paula 11 (1973)
Santa Paula 11 (1974)
San Bern. 5 (1973)
San Bern. 7 (1973)
Fruit
Lemons
Lemons
Lemons
Lemons
Oranges
Oranges
Statistic
bL0Z
s
n
bL0Z
s
n
bL0Z
s
n
^
>
n
bL0ZH
S
n
bL0ZH
s
n
2.0j7.0
-0.299*
(0.201)
28
-0.020
(0.021)
116
-0.108**
(0.053)
24
-0.055**
(0.030)
90
-0.027
(0.081)
79
-0.062***
0.030
95
^Coefficient is significantly different from zero at the 0.10 level of
the one-tailed t-test.
**Coefficient is significantly different from zero at the 0.05 level of
the one-tailed t-test.
***Coefficient is significantly different from zero at the 0.025 level
of the one-tailed t-test.
-64-
-------
Table 5.7B
Air Pollution Coefficients (and Standard Errors) for li Partitioiiings
When LTM is Deleted. Dependent Variable = LI .
Picker
Upland 1 (1973)
Upland 1 (1974)
Upland 2 (1973)
Upland 2 (1974)
Upland 2 (1973)
Upland 3 (1973)
Upland 4 (1973)
Upland 4 (1973)
Upland 22 (1974)
Fruit
Lemons
Lemons
Lemons
Lemons
Oranges
Lemons
Lemons
Oranges
Lcinous
Statistic
bL0ZH
n
bL0ZH
n
bL0ZH
5
n
^
bL02H
s
n
>
n
bL0ZH
s
n
bL0ZH
n
bL0Z
n
bL0ZH
n
2.CKH<4.0
0.015
0.037
44
0.006
(0.047)
37
-0.046*
(0.035)
67
-0.038
(0.046)
31
-0.038
-0.033
37
-0.047***
(0.020)
33
4.07.0
-0.097***
(0.044)
60
-0.054**
(0.030)
62
-0.052
(0.045)
67
-0.103***
(0.045)
62
0.037
(0.068)
43
0.038
(O.O'il)
52
-0.055*
(0.036)
31
-0.067*
(0.044)
40
-0.031
(0.056)
48
-65-
-------
Table 5.7B
(continued)
Picker
Santa Paula 10 (1973)
Santa Paula 10 (1974)
Santa Paula 11 (1973)
Santa Paula 11 (1974)
San Bern. 5 (1973)
San Bern. 7 (1973)
Fruit
Lemons
Lemons
Lemons
Lemons
Oranges
Oranges
Statistic
;
s
n
b
s
n
b
s
n
T (fi'y
\ .ylff
s
n
bL0ZH
S
n
b ^
s
n
2.07.0
-0.275*
(0.180)
28
-0.022
(0.021)
116
-0.055**
(0.030)
90
-0.108***
(0.054)
24
-0.038
(0.061)
79
-0.069***
(0.030)
95
Coefficient is significantly different from zero at the 0.10 level of the one-tailed t-test.
**Coefficient is significantly different from zero at the 0.05 level of the one-tailed t-test.
***Coefficient is significantly different from zero at the 0.025 level of the one-tailed t-test.
-------
has been deleted from the estimated expression, the air pollution coefficient
is fairly significant for the longer work-day partitioning.
In Table 5.8 are presented the estimates obtained for work-day partitionings
of the three Irvine orange pickers. The partitioning of work-day lengths for
these pickers causes the air pollution coefficients for the shorter work-days
to be statistically significant. This is diametrically opposed to the
observed tendency for most other pickers of air pollution impacts to increase
with increasing work-day lengths. In fact, the overall statistical results
for the shorter work-day length estimates accord rather closely to those
obtained for other pickers. The magnitudes and signs of the coefficients
for each variable are similar to those obtained for other pickers, the
Durbin-Watson statistic, is very close to 2.00, and the F-values for the
entire expression are highly significant. The estimates for these shorter
work-days thus seem quite reliable. For the longer work-days, reliability
must be sought elsewhere. The Durbin-Watson statistics imply negative
autocorrelation and F-values for the entire expression are statistically
nonsignificant. Plots of the residuals against the values of the dependent
variable displayed a classic case of heteroscedasticity. Clearly, the
'statistical estimation procedure employed for this longer work-day length
partitioning must be found wanting. With some consternation, we violated our
data-grubbing ethic, and, now including a calendar date variable, we re-
estimated the same expression for the longer work-days of these three
Irvine pickers. The coefficient for this variable was nonsignificant; it
added extremely little to the total explanation of the variations in the
dependent variable; and neither the coefficients nor the standard errors of
other variables were altered in any more than a minor way. Upon plotting
the residuals of the estimates against time, however, a sine-curve pattern
could be distinctly discerned. The period between each peak of the wave
was consistently about two weeks long. We have no explanation for this
phenomenon, nor do we understand why the apparent statistical quality of
the estimates for the shorter and the longer work days should be so utterly
different.
-67-
-------
Table 5.8
Earnings Estimates by Ordinarv Least Squares for H
Partitioning.? of Irvine Orange Pickers. Dependent Variable
= LI
;
I')1'1
Picker:
Variable
Constant
Lw
1.H
L0Z
LTM
_2
IT
S.E.
D-W
F
Sample
Size.
Sample
Period
Irvine 38
(1974)
2.0_7.0
1.276
(1.503)
a
0.581
(0.263)
0.027
(0.065)
0.109
0.367
0.031
0.280
1.27
2.480
63
Apr. 4-
Aug. 28
Irvine 40
(1974)
2.0_7 . 0
1.700
(1.524)
a
0.898
(0.272)
0.164
(0.065)
-0.176
(0.370)
0.176
0.293
1.43
1.76
66
Apr. 4-
Aug. 28
Rate-of-pay is a constant 32 cents per box since grove conditions are
very nearly uniform.
*Coefficient is significantly different from zero at tlie 0.10 level of
the one-tailed t-test.
**Coefficient is significantly different from zero at the 0.05 level of
the one-tailed t-test.
***Coefficient if; significantly different from zero at the 0.025 level
of the one-tailed t-Lost.
(The numbers in parentheses refer to the standard errors of the
estimated coeff tc Lents).
-68-
-------
In spite of our puzzlement with respect to the longer work-day estimates
Cor the Irvine orange pickers, the fact still remains that a partitioning by
work-day lengths for these pickers did result in statistically significant
air pollution coefficients for the shorter work-day partitioning for each
picker. If the reader is willing to take one instance of a statistically
significant air pollution coefficient for each picker in either Table 5.4A,
5.5A, 5.7A, and/or 5.8 as being acceptable evidence of a deleterious air
pollution impact upon a picker, then the proposition that sixteen of eighteen,
or eighty-nine percent, of the pickers studied appear to have been significantly
and negatively impacted cannot be rejected. Only Upland 2 (1973) and Upland 4
(1973), both in oranges, refuse to yield negative and statistically significant
air pollution coefficients. Remembering, however, that those whom we have
called different pickers are often the same pickers picking a different crop
or in a different year, twelve of twelve, or one hundred percent, of the
pickers studied appear to have had their work performance damaged by air
pollution for at least one of the two crops in both the years studied. It
would place some strain upon one's credulity to- insist that this observed
frequency of deleterious air pollution impacts across individuals is simply
due to chance, particularly when it is recognized that the collinearity
between temperature and air pollution in Tables 5.4A, 5.5A, 5.7A, and 5.8
increases the standard error of the air pollution coefficient and thus
reduces its statistical significance, without, of course, biasing the
coefficient itself.
The calculations in Table 5.9 are performed in a manner identical to
those in Table 5.6. Unless the longest work-day length partitioning for a
picker did not yield a statistically significant air pollution coefficient,
only the air pollution coefficients, the earnings observations, and the
air pollution observations falling within the longest partitioning are used
to calculate V and V. Otherwise the coefficients and observations for a
lesser work-day length partitioning that did have a statistically significant
air pollution coefficient are used. £1.9 however, refers to all work-day
lengths. The percentage in the last column of the table is thrrefore
defined in exactly the same manner as the last row in Table 5.6: it is the
compensation, in terms of a percentage of what his total earnings would be in
-69-
-------
i* — "tC* *-> X
Table 5.9
Required Picker Income Coir.pensations Calculated Using
Results of H Partitionings
o
I
Picker
Upland 1 (1973)
Upland 1 (1974) ,
Upland 2 (1973)a'°
Upland 2 (1974),
Upland 3 (1973)°
Upland 4 (1973)a
Upland 22 (1974)
Partitioning
H>7.0
K>7.0
K>7.G
K-7.J
4.37.0
H>7.0
Santa Paula 10 (1973)Jj H>7.0
Santa Paula 1C (1974) , H>7.0
Santa Paula 11 (1973)
Santa Paula 11 (1974)
San Bern. 5 (1973)b
San Bern. 7 (1973)
Irvine 38 (1974)b
Irvine 39 (1974)
Irvine 40 (1974)
H>7.0
H>7.0
4.07.0
2.0
-------
the absence of air pollution during Che entire period of observation, the picker
requires to make him indifferent to the presence of air pollution.
Upon taking the percentage required compensations For the pickers in
Table 5.6, as well as the same compensations for those pickers in Table 5.9 who
do not appear in Table 5.6, and then calculating an unweighted arithmetic
mean over all eighteen pickers, one obtains a figure of 2.0 percent. Calculating
this same mean for all twelve individuals yields a lesser required compensation
of 1.3%, where each crop and/or year for each individual is weighted by n.
The partitionings in Tables 5.7A, 5.7B, and 5.8 lack an analytical basis.
They were selected to provide similar numbers of degrees of freedom across
partitionings for most pickers. Moreover, as the careful reader will have
already noted from Table 5.1, they do not refer to actual work-day lengths
but rather to work-day lengths in a particular grove. Thus a picker could
conceivably have worked three hours in one grove and six hours in another
on a given day, yet not have his actual work-day length appear as nine hours
in our data. Instead, the three hours would be counted as an observation
in the less than four hours partitioning, while the six hours would appear
in the middle partitioning. This means, then, that the air pollution
coefficients for the-lower and middle ranges in Table 5.7 are not representative
of the interaction between hours worked and air pollution impact since II doc.
not represent actual work-day length. However, this problem is trivial for
the upper partitioning because, with only an extremely few exceptions
involving no more than .an hour, all observations in this upper partitioning
have a one-to-one correspondence between hours worked in a particular grove
and the length of the actual work-day. Of course, since it is likely the
lower and middle partitionings for at least some pickers include hours toward
the end of long actual work-days, the calculation? of required picker
compensations in Table' 5.7A are biased downward, i.e., actual required
compensations are higher. Some idea of the magnitude of this downward bias
is provided by comparing the calculated required compensations for pickers in
Table 5.6 with the calculated required compensations for these same pickers in
Table 5.9. For the pickers appearing in both tables, the percentage required
compensations in Table 5.6 exceed those appearing in Table 5.9 by factors of as
little as one-half (Upland 22 (1973)) and as great as thirty (San Bernardino (1973))
-71-
-------
an
KIT-
S'
»*•
I"
.'I' i
f
Given the downward biases in the absolute magnitudes of the percentage required
compensations of Table 5.9, it. seems reasonable to conclude that Lkc
I£pr_eS£u.t_a.t_iY_e_J-ud.ivid.ua.JL.i.n _thi_s_sj;u dy_wpul.d. hay e__Iiad_ t.o._rt\c.eive_two to
thr.ee. per cent of ..what, his .income would._be .in. the absence of air pollution to
make_him...indif ferent to.the presence. ..of . the...air pollution leve_l_s_tp__wlvic.h
h.e_was._s.ubj.e^:..t.ed. This statement applies only to circumstances in which the
picker chose to work each day as many hours as he was institutionally allowed.
-72-
-------
Footnotes: Chapter 5
1. These differences among crop years could have been accounted for by
a dummy variable. However, this would have presumed that the difference was
explained solely by a shift in the intercept of the earnings expression. The
responsiveness of earnings to air pollution would have implicitly been
assumed to have remained the same in the two years. Because of the large.
number of observations we had available on each picker's performance, the course
we adopted appeared less restrictive.
2. In retrospect, a dummy variable approach might not have been the best
way to handle this problem. The labor camp managers may be saying that Fridays
and Mondays are separable and distinct blocks of time; that is, they cannot
be embodied in the time constraint (7) of the picker's decision problem, but
must be treated as additional time constraints. This would imply that separate
earnings expressions should be estimated for each of these two days.
3. Although it co;ild easily be a coincidence, it is worth noting that the
negative magnitudes of the nir pollution coefficients for the three Irvine
workers vary directly with the worker's ages. Irvine 38 is in his early
forties, Irvine 39 is in his late twenties, and Irvine 40 is only eighteen
years old.
A. There exists sound empirical evidence in the economics literature
to support this notion of declining marginal productivity with respect to
increases in the number of hours worked. Feldstein (1967), for example, in
a study using British data finds that the elasticity of output with respect
to hours substantially exceeds that with respect to men.
5. There exists an alternative explanation for the association between
picker responsiveness to air pollution and long work-days: high air pollution
levels and long work-days may themselves be associated. For example, long
work-days may occur primarily in the summer months when high air pollution
levels also occur. Although no attempt was made to calculate a simple
correlation coefficient between crew work-day lengths and air pollution
levels, a scanning of the data files for several workers made it appear that
long work-days are distributed more-or-less rectangularly over the entire
calendar year.
6. One interesting alternative strategy, which involves using first
differences in the dependent variable, is presented in Ashcnfelter and
Heckman (1974).
-73-
-------
Chapter 6
AIR POLLUTION AND ABSENTEEISM: EMPIRICAL RESULTS
Absenteeism. One must remember that the calculated required compensations
in Chapter 5 refer solely to situations in which the picker chose to work as
long as he was permitted: he did not choose to adjust to the presence of
air pollution by taking more leisure. Does he sometimes take more leisure when
air pollution is high? If so, the required compensations presented in
Tables 5.6 and 5.9 will be excessive if extrapolated to circumstances in which
pickers do occasionally voluntarily choose leisure as a mode of adaptation.
As (12) of Chapter 4 implies, for leisure to be chosen, its marginal utility
must exceed the marginal utility of the earnings the picker could have
obtained by continuing to work. Although the picker suffers disutility from
£u the loss of earnings consequent upon his decision not to work, he also acquires
rf| some positive utility by having more leisure available. The compensation he
Jvi* requires to return to his original utility level in the presence of air
'*| pollution is therefore less than if he were to suffer a similar loss in
/* earnings but still work the same number of hours he would without air pollution.
;i-'*I The model presented in Chapter 4 readily captures the additional
I dimension of the picker's voluntary taking of leisure as an adaptation to
V the presence of air pollution. Remembering that H = H - Z, upon rewriting
:.-;jL (14) in the standard form for a labor supply function one obtains
I ffl +
'"'^ (15) Z = H - H(I , air pollution, other factors that shift S in
Figure 4.2(b))
Given that environmental conditions help to determine the hours the picker
chooses to work, a regression specification of Z on crew-hours and the factors
determining the hours the picker works yields an estimate of the covariation
'between the hours the pLcker chooses not to work when he could have worked
and the level of air pollution. Note that Jti (15), 1C air pollution actually
influences the picker's decision, the dependent variable Z, the hours the
picker chooses not to work, and the level of air pollution are expected to
vary directly with one another.
-74-
-------
Because of the myriad of factors whicli may enter into the picker's
decision not to pick and which are unrepresented in (15), we cannot hope
to achieve a particularly high degree of explanatory power. Of course, to
the extent air pollution is a relevant explanatory variable and is uncorrelated
with excluded but relevant explanatory variables, explanatory power is of no
importance. The air pollution coefficient will still be unbiased and that is
all that matters for our purposes.
In order possibly to reduce the extent tc which relevant variables that
are nonorthogonal to the air pollution variable are excluded from attempts
to estimate (15), the estimates appearing in Table 6.1 include observations
during which the picker worked less than two hours. They do not, however,
include observations in which the picker chose not to appear for work at
all. Thus the results should be interpreted as showing the picker's
propensity to work a leaser number of hours than his crew worked on a
particular day, given that the picker chose to work for at least some time
during that day. Regressions run with observations included in which the
picker chose not to work at all gave results in which much less than one
_2
percent of the variation in Z was explained (for two pickers the R 's were
negative) and the F-tests for the entire expression were always less than
unity. Air pollution coefficients were never statistically significant.
The somewhat better robustness of the estimates appearing in Table 6.1
can plausibly be attributed to the greater number of factors (family illness,
vacation plans, etc.) influencing the decision not to appear for work at all
as opposed to the factors influencing the decision to quit picking and undertake
leisure after having already expended picking effort on a particular day.
Available data limited the absenteeism estimates to Upland lemon pickers.
The estimates appearing in Table 6.1 are certainly not encouraging
if one initially suspected that pickers adapt to air pollution by substituting
leisure for picking effort. Any discouragement can, however, be tempered by
at least three factors that might have influenced the character of the estimates,
First, the pickers whose work performances are reported iu this study were
all deliberately selected because of LhcLr long and more-or-loss continuous
work records. In short, these pickers tend on a day-to-day basis to persevere
-75-
-------
Table 6.1
Absenteeism Estimates by Ordinary-Least-Squares for Upland
Lemon Pickers.3 Dependent Variable = Z.
Picker:
Variable
Constant
w
4.
H
FR
BT
0ZH
T\1
A. 4. i.
TR
I, .
t-1
_2
R
S.E.
F
D-W
Sample
Size
Sample
Period
Upland 1
(1973)
-0.409
(1.124)
0.052
(0.047)
0.073
(0.024)
-0.002
(0.003)
-0.008
(0.047)
-0.012
(0.012)
0.003
(0.008)
0.181
(0.121)
0.006
(0.018)
0.042
0.818
2.01
1.95
186
Mar. 17-
Dec. 21
Upland 1
(1974)
-0.183
(0.678)
-0.041
(0.060)
0.054
(0.013)
(0.001)
-0.034
(0.116)
0.003
(0.005)
-0.001
(0.005)
0.125
(0.125)
0.008
(0.010)
0.105
0.428
3.54
2.18
174
Apr. 1-
Nov. 2
Upland 2
(1973)
-0.650
(0.566)
0.114
(0.023)
0.076
(0.013)
-0.003
(0.004)
0.0001
(0.0001)
0.006
(0.006)
0.004
(0.004)
0.070
(0.066)
0.008
(0.008)
0.213
0.470
7.91
2.18
205
Mar. 19-
Dec. 21
Upland 2
(1974)
-0.616
(0.999)
-0.124
(0.105)
0.067
(0.022)
0.003
(0.002)
-0.221
(0.185)
0.004 •
(0.008)
0.003
(0.005)
0.283
(0.197)
-0.018
(0.016)
0.057
0.619
2.25
2.14
168
Apr. 1-
Nov. 2
Upland 3
(1973)
1.149
(0.927)
0.108
(0.043)
0.110
(0.022)
-0.010
(0.003)
-0.016
(0.038)
0.009
(0.011)
0.002
(0.006)
0.147
(0.106)
-0.019
(0.021)
0.218
0.653
6.54
2.13
160
Mar. 17-
Dec. 20
Upland 4b
(1973)
1.854
(1.706)
0.027
(0.081)
0.091
(0.042)
-0.008
(0.005)
-0.199
(0.098)
0.018
(0.019)
0.011
(0.012)
-0.048
(0.172)
-0.063
(0.032)
0.148
0.989
3.82
1.70
131
Mar. 16-
Dec. 21
Upland 22
(1974)
-2.942
(1.215)
0.030
(0.145)
0.224
(0.026)
0.002
(0.003)
-0.427
(0.240)
0.007
(0.009)
0.007
(0.006)
0.403
(0.264)
0.048
(0.025)
0.372
0.737
12.56
1.90
157
Apr. 17-
Nov. 2
-------
Table 6.1
(continued)
These estimates include observations in which the picker worked less than two hours.
This picker is a woman.
(The nurr/s&rs in parentheses refer to the standard errors of the estimated coefficients)
-------
in citrus harvesting. Perhaps pickers who, relative to the population of
pickers, exhibit greater perseverance in their day-to-day picking activities
will exhibit similar tenacity within any single day.
Second, the values of the air pollution and temperature variables used
in Table 6.1 are those employed in all earlier tables; that is, they are the
arithmetic mean ambient pollution concentrations and temperatures over the
crew's work-day rather than the picker's work-day. Thus, on those occasions
where Z is positive, pollution concentrations and temperatures occurring
when the picker was taking leisure are included in the observed values of the
pollution and temperature variables. The use of the latter values would be
justified if and only if the picker's expectations about future pollution
and temperature concentrations during the rest of the crew's work-day were
always realized. This seems unlikely. It is therefore preferable to presume
that the picker formulates his expectations about air pollution and temperatures
for the rest of the work-day on the basis of the pollution and temperature
levels he has already experienced. The values employed for estimation
purposes should therefore have been some combination of pollution and
temperature levels occurring while the picker was actually working. The
failure to do so in the estimates of Table 6.1 means that the air pollution
and temperature variables include measurement error and that their coefficients
are therefore biased. The direction of bias for either coefficient is not
immediately evident, particularly since the errors for these two somewhat
collinear variables probably interact in a complex way.
Third, in contrast to the earnings estimates of Chapter 5, no attempt
was made for the absenteeism estimates to experiment with the empirical
specifications for one or two pickers. That is, no effort was expended to
gain insight into the absenteeism relationship by learning from the empirical
results for "test" pickers. The estimates appearing in Table 6.1 are the first
absentee estimates attempted for each picker.
In summary, the results presented in Table 6.1 make it easy to reject the
maintained hypothesis that air pollution influences picker absenteeism. This
rejection must, however, be highly conditional because of the character of the
pickers for whom estimates were made, because of the inclusion of air pollution
levels occurring after the picker had chosen leisure, and because no experimen-
tation on "test" pickers was performed.
-78-
-------
Siiiiul tanuuuH Adjustments uf Work-Kf1 art and Leisure. The model of the
picker's decision problem presented in the fourth chapter implies that the
picker, when faced with actual or expected air pollution, may adjust his
work-effort, his leisure-time, or both. In the section immediately preceding
this one and in Chapter 5, empirical tests of the model were undertaken, but
no effort was made to estimate expressions in which the picker was permitted
to adjust simultaneously his work-effort and his leisure. Thus no account has
been taken of circumstances in which the picker suffered reduced but still
positive productivity during part of his work-day, and, perhaps in response
to this reduced productivity, then decided to substitute leisure for additional
expenditure of picking effort. Nevertheless, the model of the picker's
decision problem would seem to lend itself readily to simultaneous treatment of
the influence of work-effort and voluntarily taken leisure-time upon earnings,
or the influence of earnings and work-time upon voluntarily taken leisure-time.
All one need do is replace Z on the left-hand side of (15) with H - H,
substitute the H(.) of (15) into the II of (14), and state the resulting
expression in stochastic form. The resulting or "reduced-form" expression
is overidentified, however, meaning that one cannot allocate the estimated
coefficients of the reduced form between (14) and (15) , the original
structural expressions. The air pollution coefficient should nevertheless
provide an estimate of the combined effect upon earnings of work-effort and
voluntarily taken leisure-time.
Following the procedures outlined in the above paragraph, we have
estimated reduced-form expressions by two-stage-least-squares for two
pickers, Upland 2 (1974), and Upland 22 (1974). In both cases, earnings,
I , was the dependent variable, i.e., expression (15) was substituted into
expression (14). Without further analysis, the results cannot be considered
enlightening. The highly statistically significant but positive air pollution
coefficients obtained with the reduced form have thus far defied interpretation.
Time has not permitted a detailed investigation of what may be causing these
positive coefficients.
-79-
-------
Footnotes: Chapter 6
1. Additional remarks on the nature of the bias introduced are
available in Anderson and Crocker (1971, p. 174).
12
-80-
-------
Chapter 7
CONCLUSIONS AND FEASIBLE EXTENSIONS
Conclusions. Whenever one constructs a formal model of a real situation,
one is attempting to express in an internally consistent manner a conviction as
to which of the elements of the situation are trivial and which are essential.
A bounded rationality requires that one pick and choose. It is a common-
place of statistical inference that one's convictions about the essential
elements can, by comparing their implications with observations on real
situations, be rejected but never completely accepted. The best one can
do is fail to reject the convictions. It is worth noting, however, that
even a rejection must rest upon some self-convincing interpretation of one's
observations. The pressures of uncertainty operate symmetrically.
In order to secure the reality of citrus picking in the presence of
air pollution somewhat more closely, a model of the picker's decision at
the beginning of each day of whether or not to pick citrus and his supply
of work-effort once he has decided to pick has been constructed. In the
model, the response of work-effort to air pollution was viewed strictly as
a biological relation: that is, air pollution entered the picker's production
function for citrus fruit but it did not enter his utility function. The
response was, in essence, viewed as a short-term and reversible morbidity
effect. No attempt was made to capture any long-term and irreversible
effects. Certainly neither the model nor its empirical implementation
captures all features that may influence the individual picker's work
performance each and every day. For example, in addition to disliking the
loss in earnings that the presence of air pollution causes, he may also
dislike air pollution in and of itself. Nevertheless, the essential features
of the determinants of the individual's day-to-day earnings from the picking
of citrus fruit, appear, for the most part, to have been captured, both
analytically and empirically.
-81-
-------
The decision agent of the model, tlu.' individual citrus picker, corresponds
exactly to the decision agent in each of the empirical analyses. It should be
recognized that the study does not consist of a single empirical analysis
of a collection of decision agents, but rather of a series of analyses, one
analysis to each of eighteen pickers distinguished by individual, crop, or
year. This series of analyses involved the consistent application of the
identical analytical model to each picker. Although data did not permit
explicit investigations of the reasons for differences among pickers in their
responses to the presence of air pollution, these responses were consistently
of the same order of magnitude.
With the exception o£ four "preliminary test" pickers, the basic model
was empirically applied only once to each picker, thus avoiding the often
difficult-to-interpret effects upon subsequent tests and estimation procedures
of ,the "data-grubbing" or "massaging" techniques so frequently employed in
similar studies. Given the resources and time available to the study, this
procedure was not without its costs, however. For a few pickers (in particular,
all three of the Irvine; orange pickers and the two Upland orange pickers),
the stochastic statement of the model was clearly misspecified. The obvious
way to repair this is to respecify in a correct fashion the expressions to be
estimated for these pickers, and then apply these expressions to the data for
new pickers drawn from the same crews. We possess the requisite data, but
it has not been put in ,3 form susceptible to computer treatment. Moreover,
.;f.
jt if we had investigated in detail the sources of the misspecifications for the
m Irvine and Upland orange pickers, we would have drawn available resources
away from applying the estimated expression for the four preliminary test
pickers to additional pickers in locales and for crops where the expression
appeared to be robust. In our judgment, the application of the expression for
the four preliminary test pickers to more pickers appeared more valuable to
the study than correction of the misspecifications for the Irvine and Upland
orange pickers.
The analytical model developed in Chapter 4 explains both the picker's
decision whether or not to pick each day as well as his decision about how much
effort to put forth once liu lias decided to pick. Only the latter aspect of
-82-
-------
the modal has been subjected to empirical test, and this aspect is considered
to be separable from the daily decision of whether or not to pick at all.
Fortunately, the available data for each picker permitted us to distinguish
days in which the picker worked for as many hours as he was institutionally
allowed from days in which he started working but voluntarily substituted
leisure part way through the work-day. In the model, it was shown that in
circumstances where the picker did not substitute leisure for earnings, the
loss, if any, in earnings due to air pollution could be interpreted as a
measure of the Hicksian compensating surplus, i.e. the social loss caused by
air pollution that attaches to the picker. This measure of social loss is
simply the additional earnings the picker requires in the presence of air
pollution to make him indifferent to its presence. This compensating
surplus was estimated from an inverse constant real income supply function.
Disregarding the specification problems for the Irvine and Upland orange
pickers, this surplus was positive for all twelve individuals and sixteen
of the eighteen pickers studied. (Again, two pickers may be the same
individual distinguished by crop and/or year). For ten of the eighteen
pickers and six of the. twelve individuals, however, the surplus did not
become apparent until account was taken of the possibility that the impact
of air pollution might differ according to the length of time a picker has
worked on a given day. In fact, estimates give fairly strong support to the
hypothesis that air pollution impact, measured in terms of the compensating
surplus, tends to increase with increasing numbers of hours worked. For
pickers, the surplus amounts to an average of about 2.0 percent of the picking
earnings they would have had in the absence of air pollution, while for
individuals the same percentage is 1.3 percent. If one removes from these
calculations the pickers whose estimated expressions appear to be misspccified,
one is left with positive compensating surpluses for thirteen of the thirteen
pickers and nine of. the nine individuals studied. The average compensating
surplus as a percentage of picking earnings in the. absence of air pollution is
then 2.3 percent for the pickers and 1.3 percent for the individuals.
Percentage compensating surpluses differ between pickers and individuals
because the former surplus is an unweighted arithmetic mean, while the latter
-83-
-------
arithmetic mean is established by weight ing each picker by the number of.
observations for him in a crop or year. In terms of absolute dollar magnitudes,
these compensating surpluses appear to range from less than twenty dollars
to nearly two hundred dollars over an entire calendar year, given the piece-
work wage rate scales and the levels of air pollution prevailing in the
South Coast Air Basin during 1973 and 1974.
In addition to estimates of the compensating surplus for a number of
pickers, two attempts were made to estimate the effect of air pollution upon
the picker's substitution of leisure time for picking effort. One set of
estimates presumed that while still picking, the picker did not adjust his
picking effort. No statistically significant impact of air pollution upon the
substitution of leisure time (absenteeism) for picking effort was discovered.
It is possible this lack of significance is due to the fact that all the
pickers tested had stable work histories, that the measure of air pollution
JK employed covered the crew's work-day rather than just the period during which
!•* the picker in question was working, or that the statistical procedures
4
jW employed were unsuite.d to the problem. No attempt has been made to ascertain
«?
jg whether any of these factors have influenced the estimates for absenteeism.
I
,f* A second set of estimates tried to treat simultaneously the picker's
Jp reduction of picking effort during a single work-day and his outright
s
•& substitution of leisure time for picking effort. Few research resources
I were devoted to this treatment, and the results obtained were not susceptible
Ai to immediate interpretation.
A
; ft In summary, the results of the study provide quite strong evidence that
'^i
citrus pickers would have had to be compensated by about two percent of what
their incomes would have been in the absence of air pollution in order to
be indifferent to the air pollution levels prevailing in the South Coast Air
Basin in 1973 and 1974. This presumes that pickers did not reduce the. hours
they spent picking citrus fruit in response to these air pollution levels.
Although much more Lrn.il.i.vc than the preceding estimates of required
compensations, estimates of the responsiveness of absenteeism to air pollution
during the same period were* statistically insignificant. Thus, for the pickers
studied, required income compensations averaging about two percent of picking
earnings are quite accurate, given that air pollution did not influence
absenteeism.
-84-
-------
Feasible Extensions of this Research. The research possibilities residing
in this data set are by no means exhausted. One might, in fact, map out a
lengthy research program consuming several man-years of effort. No attempt is
made in this section to do so; nevertheless, as embodied in this report, the
study is somewhat incomplete. In this section, work that can quite readily
be done with the existing data set and analytical framework is briefly discussed,
The section and the report conclude with an even briefer discussion of
extensions requiring the acquisition of additional data and/or the development
of an analytical framework capable of encompassing a greater range of phenomena.
Currently in our possession is the same information used for the pickers
and individuals studied in this report for approximately an additional 350
pickers comprising about 220 distinct individuals. As this report is being
written, the information for six of these pickers has been tabulated and
keypunched but has not yet been submitted to the ministrations of the computer.
In spite of the consistency of the results generally obtained across pickers,
the small number of pickers and individuals studied (note, however, that a
large number of observations were available for each picker), make one a bit
hesitant to testify resoundingly that it has been definitively established that
photochemical oxidants are detrimental to the work performance of citrus
pickers. After all, fifteen or so replications of an experiement do constitute
a small sample. Running the experiment for additional pickers would be a
time-consuming task, but, given what has thus far been accomplished, mostly
a mechanical one. In a letter to the project officer dated November 21, 1975,
we stated that an analysis of about sixty pickers would be an adequate sample
size. Clearly, this has not been accomplished.
Perhaps the most troublesome problem in discovering the statistical
significance of the air pollution coefficients has been the collinearity
between the air pollution and temperature variables. This collinearity is
reflected in the considerable reductions in standard errors of the air
pollution coefficients that frequently occur when the temperature variable is
deleted from the expressions to be estimated. One obvious but as yet untried
way possibly to reduce this collinearity is to partition each picker's data
set by temperature intervals so as to lessen the variability of temperature.
-85-
-------
This, of course, does; not guarantee thai i. ue variability ol" air pollution
might not be reduced as much or even more. The partitioning would h.ivc the
additional advantage of providing a partial test of the hypothesis thai nir
pollution impacts vary with temperature levels.
A partitioning of the air pollution variable itself can provide
information on whether the assumption used in this study of a constant
elasticity of air pollution impact with respect to earnings is reasonable.
The partitioning can also provide insight into the form of the relationship
between earnings and air pollution, including whether there is some positive
level of air pollution at which a negative impact of air pollution upon
earnings begins. Some preliminary efforts along these lines for six pickers
show no tendency whatsoever for the air pollution elasticity of earnings to
increase with increa.s i. 1114 air pollution levels. Some imagination and perhaps
a bit o£ wishful think in;-, enables one to discern a slight tendency for the
elas'ticity to decline with increasing air pollution.
J>" Only the arithmetic mean of the concentration air pollution during the
*K crew's work-day or over a twenty-four hour period are used in this report as
A[ a measure of air pollution, although the variance of the air pollution
2f distribution over the crew's work-day was introduced and found to be
statistically insignificant for the four preliminary test pickers. Of course,
the arithmetic mean and variance are not the only characterizations of the air
pollution measure that might be relevant. For example, given the intuition
of many familiar witli the health impacts of air pollution that peak
concentrations account for a disproportionate share of impact, a measure of
skewness could also be relevant. In addition, it is plausible that within
any given work-day the effect of air pollution is cumulative, requiring that
one account for prior air pollution levels during the day via a distributed
lag structure.
The speculations of the preceding paragraph suggest that characterizations
of the air pollution variable in addition to its arithmetic mean might bo.
relevant. No suggestion is made in analytical terms, however, as to why
they might be relevant. An analytical framework for their relevance is
readily provided in terms of the economics of uncertainty. By using the
arithmetic mean of air pollution, we have implicitly and strongly assumed that
the picker makes his picking decisions solely on the basis of the expected
-86-
-------
value of air pollution concentration. In short, certainty equivalence, with
its implication that the concentrations the picker expects are always
realized, has been assumed. The picker's behavior at any time is therefore
viewed as invariant with respect to the probability of error in his forecast
of expected air pollution dosages. It is readily shown that, independently
of actual pollution dosages, the existence of the possibility of discrepancies
between the picker's expected and realized air pollution dosages is costly
in and of itself. This study, as presently constituted, could therefore
be neglecting a fairly important facet of the impact of air pollution upon
citrus picker work performance. The air pollution measures of the previous
paragraph would permit one to test for the existence and magnitude of this
uncertainty effect without necessitating any fundamental changes in the
analytical framework already employed in the study.
All expressions estimated in this study are based upon the assumption
of an additive error term with the usual Gauss-Markov properties. Since the
study is actually a series of analyses of the work histories of individuals,
one analysis to an individual, this is not an altogether innocent assumption.
It implies that the error for a picker during work-day t is independent of
the error at t-1. Except for certain orange pickers, this assumption seems,
in fact, innocent enough since Durbin-Watson statistics consistently hovered
around 2.00. For these orange pickers, however, the Durbin-Watson statistics
were consistent with the presence of rather severe negative autocorrelation.
There exist several possible responses to this presence, none of which have
been applied to the orange pickers in this study. For example, one might
resort to maximum likelihood or two-stage-least-squares estimating procedures.
Or there may be some neglected variable, such as lagged earnings, that has
a systematic influence on the current earnings of orange pickers but has none
upon the current earnings of lemon pickers. It would be a rather simple
matter to designate1. as "preliminary test" workers a couple of the orange
pickers for whom cxpres:. ions have already been estimadnl, cxpcri.iiu'nL will)
their stochastic specifications, and then apply the resulting improved
specification to additional orange pickers.
-87-
-------
The stochastic expressions of the inverse supply function that have been
estimated in this study have not embodied ail the £ prior i. information the
analytical framework is capable of generating. For example, given that air
pollution does not enhance the ability of the picker to harvest fruit, a few
simple manipulations of the compensating surplus model of Chapter A will
generate the conclusion that air pollution must affect picker earnings either
not at all or negatively. This is a result of the inability of the picker to
work for longer hours than does his crew and it implies the coefficient £or
the air pollution variable can be restricted to zero or negative values. It
is possible that this knowledge about the sign of the air pollution coefficient
would, if a restricted estimation technique such as the mixed estimation
method of Theil and Goldberger (1961) is used, would reduce the variance of
the air pollution estimator and thereby increase the efficiency of estimation.
The empirical sections of this study that attempt to estimate the
relationship between air pollution and absenteeism, as well as the simultaneous
relationship between air pollution, earnings, and absenteeism, are incomplete.
As was indicated in the main text, any investigation of the simultaneity
between earnings and absenteeism was set aside as soon as difficulties of
interpretation were faced. If only because a full description of the picker's
response to air pollution requires that account be taken of the possibility that
he adjusts both work effort and leisure time during a single day, more thorough
investigation of this simultaneity would be useful. As for absenteeism, again
as was indicated in the text, a measure of air pollution that measures
cumulative work-day exposures occurring prior to the picker's termination of
work effort would be preferred to the air pollution measure actually employed.
The use of this preferred measure might show some effect of air pollution
upon absenteeism.
The discussion to this point in this last section of the report has dealt
solely with additional research that can be done within the confines of the
existing data set and analytical model. It was pointed out that opportunities
are available to acquire greater confidence in the measured air pollution
Impacts reported here, to pain better understanding of the intcuactious between
air pollution impact and factors such as temperature, and to delve further into
-88-
-------
the reasons why air pollution impacts for certain pickers and certain classes
of relationships were not obtained. If one combines the existing data
set with additional data and if one expands the analytical model, a number of
further issues can be empirically investigated. For example, it would be
most interesting to know why air pollution impacts appear to differ among
individuals. If differences in biological endowments are the source, these
differences are easily built into a slightly expanded version of the
analytical model of Chapter A. It would be a relatively easy matter to acquire
information on such obvious sources (or proxies for sources) of differences as
age, weight and height, and sex. A much more ambitious effort might involve
interviews to acquire data on medical histories, life styles, and occupational
histories. Acquisition of data of this sort would allow a pooled time series-
cross sectional analysis capable of explaining the relationship between the
disturbances of a specific picker population at two or more different times.
The reader familiar with the biomedical literature will have long ago
noted that no mention is made anywhere in the previous pages of biomedical
evidence of the effects of photochemical oxidants upon the human organism.
Although a cursory review of the biomedical literature did not yield any
studies that could be employed as a_ priori information in this study, a more
thorough review might provide such information. Biomedical information might,
in fact, be particularly useful in specifying functions explaining the sources
of differences in air pollution impacts among individuals. This information
would be extremely valuable in a pooled time series-cross sectional study.
A final question perhaps of interest is the feasibility of extending the
approach used here to other industries in order to study the effects of
environmental pollutants upon labor performance. The data available to this
study are certainly unusual to some degree because of their detail on the
day-to-day pay scales, working conditions, and environmental conditions of
individual workers. It is, in fact, difficult to think of another single
industry for which similar detail would be available, in which the supply of
worker effort is strongly separable from the efforts of other workers, and
where complementary capital equipment is of no more than trivial import.
In the absence of these conditions, the analytical model of the worker's supply
-89-
-------
of effort decision would be more complicated. This is not to say, however,
that the model cannot be built, nor that the data is unlikely to be available
with which to implement it empirically. For example, many firms in the
construction industry and in agricultural industries monitor the output of
individual workers and maintain records of the conditions under which they
are working. Collection of the data is normal business practice. This is
the only data required to implement some approximation of the approach taknu
in this study.
-90-
-------
Footnotes: Chapter 7
1. It should be noted, nevertheless, that there exist orderly and
rigorous techniques for inductive model-building that permit one t:o weigh
the gains in improved specification against the losses in robustness of
estimators. For an analysis of the properties of one of these techniques,
see Wallace and Ashar (1972) .
2. For more detail and a less cautious statement about these tendencies,
see Crocker and Horst (1976) .
3. For formal proofs and arguments showing this in pollution contexts,
see Crocker (1971, pp. 21-26), and National Academy of Sciences (1974,
pp. 427-470).
-91-
-------
BIBLIOGRAPHY
Alchian, A.A., and H. Demsetz. 1972. Production, Information Costs, and
Economic Organization. The American Economic Review 62: 7 /7-795.
Anderson, R.J., Jr., and T.D. Crocker. 1971. Air Pollution and Residenl.ia 1
Property Values. Urban Studies 8:171-180.
Ashenfelter, 0., and J. Heckman. 1974. The Estimation of Income, and
Substitution Effects in a Model of Family Labor Supply. Econometrica
42:73-85.
Azariadis, C. 1975. liq.1 licit Contracts; and Underemployment Equilibria.
J. of Political Economy 83:1183-120Z.
Bell, D. Jr. 1967. l^fodcls of Ccmmodity Transfer. Berkeley, Calif. Agric.
Exp. Sta., Gianniui. Foundation Monograph No. 20.
Commission on Natural l-U*M.:urccs, National Academy of Sciences. 197.5. Ait-
Quality and Stationary Source Emission Control. A report prepared for
Committee on Public Works, U.S. Senate, 94th Congress, 1st Session,
Washington D.C. , USGl'U.
Coordinating Committee on Air Quality Studies, National Academy of Sciences.
1974. Air Qua.ll.ty ;•!•! Automobile Emission Control: 4. The Costs and
Benefits of Auton.ob i ' <• Emission Control. USGPO, Washington, D.C. 470 pp.
Crocker, T.I). 1971. Urh;"i Air Pollution Damage Functions: Theory and
Measurement. NAPCA 22-69-52, National Air Pollution Control
Administration, Washington, D.C. J35 pp. plus appendices.
Crocker, T.I). .1073. t'.i>:i'_r.:tcluaJ Choice. Natural Resources Journal 13:
5IU-577.
Crocker, T.U., and R.L. lluryt, Jr. 1976. The Effeels of Oxidants on the
Productivity of Citrus Workers. A paper presented at the annual
meetings of the Western Economic Association. San Francisco. June 25.
Dooringor, P.B., and M.J. I'iorc. .197.1. Internal L.ibor M.-trkcls and Manpnwc-r
Analysis. Lexington. Mass., D.C, Heath and Company.
-92-
-------
Felclstcin, M.S. 1967. Spec if ie.ntion o! flic l.iibor Input in I he- i\£rji"\.-\ic.
Production Function. The Review of Economic Studies 34:375-386.
Fellows, L. 1929. Economic Aspects of the Mexican Rural Population in
California with Special Emphasis on the Need for Mexican Labor in
Agriculture. San Francisco, R and E Associates. 1971.
Fisher, L.H. 1953. The Harvest Labor Market in California. Cambridge,
Harvard University Press.
Freeman, A.M. III. 1974. On Estimating Air Pollution Control Benefits
from Land Value Studies. J. of Environmental Economics and Management
1:74-83.
Green, H.A.J. 1964. Aggregation in Economic Analysis: An Introductory
Survey. Princeton, N.J., Princeton University Press.
Lave, L., and E. Seskin. 1970. Air Pollution and Human Health. Science 169:
723-733.
Manpower Administration. 1969. Personnel Practices of Citrus Growers of
the West. Farm Labor Developments. Washington, D.C., U.S. Dept. of
Labor, 1-7.
Muth, R.F. 1966. Hou;;elioLj Production and Consumer Demand Functions.
Economctrica 34:699-708.
Nelson, J.P. 1975. The Li.i'ects of Mobile-Source Air and Noise Pollution on
Residential Property V'jlues. Report No. DOT-OS-40094, Washington, D.C.,
Office of University Research, USDOT.
Padfield, H., and W.E. Martin. 1965. Farmers, Workers, nnd Machines. Tucson,
The University ol" Arizona Press.
Ramsey, J.B. 1969. Tests for Specification Errors in Classical Linear
Least Squares Regression Analysis. Journal of the Royal Statistical
Society Series B.-J50-371.
Rosedale, D., and J. M.mcr. 1974. Labor Management for Seasonal Farm
Workers: A Case Study. Berkeley, California Agric. Exp. Sta., Info.
Series in Agric. Econ. No. 74-1.
Rosen, S. 1974. llodonic Prices and Implicit Markets: Product Differentiation
in Pure Competition. Journal of Political Economy 82:34-55.
Seivin, 11.C., and A. Stuart. 1966. Data-Dredging Procedures in Survey
Analysis. The American Statistician 20:20-23.
-93-
-------
SmLt.h, II..I., '>t_. n_l . 1965. Lemon Pi.ck.hij' .'irid the Wm.nr,i County 'I'm-
Product ion- Incentive Wage System. Berkeley, Calif. Agric. lixp. Sta.,
Bulletin 809.
Smith, V.K. 1975. Mortality-Air Pollution Relationships: A Comment.
Journal of the American Statistical Association 70:341-343.
Stafford, F.P., and M.S. Cohen. 1974. A Model of: Work Effort and Productive
Consumption. J. of Economic Theory 7:333-347.
Statistical Report Service. 1975. Fruits and Tree Nuts: Blooming,
Harvesting, and Marketing Dates. Agric. Handbook No. 186, Washington,
D.C., U.S. Dept. of Agric.
Thaler, R., and S. Rosen. 1975. The Value of Saving a Life: Evidence
from the Labor Market. N.E. Terleckyj, ed. Household Production
and Consumption. New York, Columbia University Press.
Thiul, II. , and A.S. Go 1 dl-.ergei:. 1961. Un Pure ami Mixrd Si at Lsl u-.-i I
Estimation in KcOHMme.tries. International Economic Review 2:65-78.
Thompson, E.A. 1970. Konpcr.uniary Rewards and the Aggregate Production
Function. The Review of Economics and Statistics 52:395-404.
Tiao, G., and A. Goldberger. 1962. Testing Equality of Individual Regression
Coefficients. Univ. of Wisconsin-Madison, Social Systems Research
Institute, Workshop on the Economic Behavior of Households, Paper 6201.
10 PP.
U.S. Bureau of the Census. 1973. Census of Agriculture, 1969, Volume V,
Special Reports, Part 6, Fruits, Nuts and Berries. Washington, D.C.,
USGPO.
Waddell, T. 1974. The Economic Damages of Air Pollution. USEPA Publication
600/5-74-012. 156 pp.
ar-u, T.D., and V.C. Apinr. 1972. Sequential Methods in Model
Construction. '!h> Review of Economics and Statistics 54:172-178.
-94-
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-600/5-77-013
2.
3. RECIPIENT'S ACCESSI ON> NO.
4. TITLE AND SUBTITLE
OXIDANT AIR POLLUTION AND WORK PERFORMANCE OF CITRUS
HARVEST LABOR
5. REPORT DATE
September 1977
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
Thomas D. Crocker and Robert L. Horst, Jr.
8. PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Department of Economics
University of Wyoming
Laramie, Wyoming 82071
10. PROGRAM ELEMENT NO.
1AA601
11. CONTRACT/GRANT NO.
68-02-2204
12. SPONSORING AGENCY NAME AND ADDRESS
Health Effects Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park. N.C. 27711
13. TYPE OF REPORT AND PERIOD COVERED
RTP,NC
14. SPONSORING AGENCY CODE
EPA-600/11
15. SUPPLEMENTARY NOTES
16. ABSTRACT
This project assesses the effect of photochemical oxidants on the work performance of
twelve individual citrus pickers in the South Coast Air Basin of southern California.
A model of the picker's decision problem is constructed in which oxidants influence
the individual's picking earnings and leisure-time via short-term and reversible
morbidity effect. Circumstances are specified under which this effect can be
interpreted as the additional earnings the individual would have to receive in the
presence of oxidants in order to make him indifferent to the presence of oxidants.
This Hicksian compensating surplus is estimated separately for each of twelve indivi-
duals. In terms of absolute dollar magnitudes, compensating surpluses appear to range
from less than twenty dollars to nearly two hundred dollars over an entire calendar
year, given the piece-work wage rate scales and the levels of air pollution prevailing
in the South Coast Air Basin during 1973 and 1974. As a percentage of what individual
earnings would have been in the absence of air pollution, the dollar magnitudes range
from three-tenths of one percent to nine percent. The average is about two percent.
All estimates of the compensating surplus are conditional upon the individual not
adjusting the hours he picks in response to air pollution. Estimates give fairly
strong support to the hypothesis that air pollution impact, measured in terms of the
compensating surplus, tends to increase with increasing numbers of hours worked.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.IDENTIFIERS/OPEN ENDED TERMS C. COS AT I Field/Group
air pollution
unskilled workers
performance evaluation
photochemical oxidants
05 J
02 B
06 T
18. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19. SECURITY CLASS (ThisReport)
UNCLASSIFIED
21. NO. OF PAGES
im
20. SECURITY CLASS (This page)
UNCLASSIFIED
22. PRICE
EPA Form 2220-1 (9-73)
95
-------
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Research and Development
Technical Information Staff
Cincinnati, Ohio 45268
OFFICIAL BUSINESS
PENALTY FOR PRIVATE USE. S3OO
AN EQUAL OPPORTUNITY EMPLOYER
POSTAGE AND FEES PAID
U S ENVIRONMENTAL PROTECTION AGENCY
EPA-335
Special Fourth-C lass RatP
Book
o
%
\
o
address is incorrect, please change on the above label;
tear off; and return to the above address.
If you do not desire to continue receiving this technical report
series, CHECK HERE d. tear off label, and return it to the
above address.
------- |