Journal of Agricultural Economics, Vol. 63, No.

1, 2012, 158–174
doi: 10.1111/j.1477-9552.2011.00326.x

Determinants of Health Costs due to

Farmers’ Exposure to Pesticides: An
Empirical Analysis
Wasantha Athukorala, Clevo Wilson and Tim Robinson1
(Original submitted September 2010, revision received July 2011, accepted
August 2011)

Abstract
Pesticide spraying by farmers has an adverse impact on their health. However,
in studies to date examining farmers’ exposure to pesticides, the costs of ill
health and their determinants have been based on information provided by farm-
ers themselves. Some doubt has therefore been cast on the reliability of these
estimates. In this study, we address this by conducting surveys among two groups
of farmers who use pesticides on a regular basis. The first group is made up of
farmers who perceive that their ill health is due to exposure to pesticides and
have obtained at least some form of treatment (described in this article as the
‘general farmer group’). The second group is composed of farmers whose ill
health has been diagnosed by doctors and who have been treated in hospital for
exposure to pesticides (described here as the ‘hospitalised farmer group’). Cost
comparisons are made between the two groups of farmers. Regression analysis of
the determinants of health costs show that the most important determinants of
medical costs for both samples are the defensive expenditure, the quantity of pes-
ticides used per acre per month, frequency of pesticide use and number of pesti-
cides used per hour per day. The results have important policy implications.

Keywords: Agricultural policy; non-market valuation; policy; production economics.

JEL classifications: Q10, Q18, Q19.

1. Introduction
Sri Lanka has experienced widespread use of pesticides since the end of the 1960s.
The quantity of pesticides used has shown a rapid increase in succeeding years,

1
Wasantha Athukorala is a researcher in the School of Economics and Finance, Queensland
University of Technology, Queensland, Australia. E-mail: wasantha.athukorala@qut.edu.au
for correspondence. Clevo Wilson is an Associate Professor in the School of Economics and
Finance, Queensland University of Technology, Queensland, Australia. Tim Robinson is a
Professor in the School of Economics and Finance, Queensland University of Technology,
Queensland, Australia. The authors acknowledge the useful comments received on previous
versions of the paper from the Editor and two anonymous referees.

 2012 The Agricultural Economics Society. Published by Blackwell Publishing,

9600 Garsington Road, Oxford OX4 2DQ, UK, and 350 Main Street, Malden, MA 02148, USA.
 Determinants of Health Costs due to Farmers’ Exposure to Pesticides 159

especially during the last three decades. The morbidity and mortality arising from
exposure to pesticides have grown in line with this increase.2 Research indicates
there are at least three reasons for the high level of pesticide use in developing
countries, including Sri Lanka. First, the expectation of farmers of future yields
from using pesticides is high. For example, 92% of the general farmers group3 inter-
viewed believed their crop loss would be more than 50% if they used a lower quan-
tity of pesticides than their current levels. Second, some farmers substitute
pesticides for labour and capital input requirements. For example, increased inten-
sity of agricultural production based on green revolution and high yielding varieties
tends to create a shortage of family labour and increases farmers’ dependency on
outside labour input or, more commonly, on chemical inputs. Manual weeding has
almost disappeared in irrigated areas in Sri Lanka (Van Der Hoek et al., 1998).
There is also a greater need for more herbicides when sowing paddy directly than
when transplanting – the latter procedure being more labour intensive. In addition,
it is cheaper for farmers to clear land using herbicides than to plough the land.
Third, sales promotion activities and credit facilities promote the use of pesticides
among farmers. As long as farmers perceive pesticides as indispensable to sustaining
high yields and for crop cultivation, they will continue to use them despite the
health hazards.
All previous studies have examined the incidence of pesticide poisonings of gen-
eral farmers based on the information provided by the farmers themselves (see, for
example, Wilson, 2002). Due to these estimates being based on self-assessment of
pesticide-induced illnesses, more accurate estimates have not been possible. This
study augments these self-reported data with data reported by a hospitalised sam-
ple.4 The research questions to be addressed in this study are as follows. (i) What
are the costs of ill health arising from exposure to pesticides? (ii) What is farmers’
willingness to pay (WTP) to avoid direct exposure to pesticides? (iii) What are the
relative contributions of different components of total costs (namely loss in earn-
ings, medical costs (MC),5 averting expenditure and disutility)? (4) What are the
variables that determine the direct health costs?6 The study uses data from two
field surveys covering 246 general farmers and 217 hospitalised farmers diagnosed
as suffering from exposure to pesticides. The results of the two studies are com-
pared.
The remainder of the article is set out as follows. Section 2 reviews the litera-
ture while section 3 presents a theoretical model of WTP to avoid ill health due

2
Van Der Hoek et al. (1998) estimate that approximately 7.5% of agricultural workers in Sri
Lanka are affected each year. The corresponding figures for Costa Rica and Nicaragua are
4.5% and 6.3%, respectively (Wesseling et al., 1993; Garming and Waibel, 2009).
3
General farmers are those who perceive that their ill health is due to exposure to pesticides
and have obtained at least some form of treatment.
4
Hospitalised farmer group is defined as a group of those whose ill health has been diag-
nosed by doctors and who have been treated in hospital for exposure to pesticides.
5
Medical costs include any costs related to taking medicine due to exposure to pesticides.
For example, it includes direct doctor fees, hospitalisation fees, cost of medicine or tests and
costs relating to hospital or doctor visits.
6
Health costs due to exposure to pesticides include medical costs, averting expenditures and
loss of wage income due to ill health.

 2012 The Agricultural Economics Society.

160 Wasantha Athukorala, Clevo Wilson and Tim Robinson

to direct exposure to pesticides. Section 4 discusses the survey and data collec-
tion methods. Section 5 presents the results of the analyses of the two datasets.
We first examine the costs of ill health between the two samples and estimate
the loss in social welfare due to pesticide related health costs. In this section, we
also report the results of the econometric analyses which examine factors
influencing ill health and medical expenditures. The final section summarises and
concludes.

2. Literature Review
Antle and Pingali (1994) examined the effects of on-farm pesticide use on farmers’
health and the effects of farmer ill health on farm productivity, finding that pesti-
cide related health impairments caused significant reductions in labour productivity.
Using the same sample, Pingali et al. (1994) studied the impact of prolonged pesti-
cide use on farmers’ health. They quantified the magnitude of chronic health effects
and health costs directly related to exposure to pesticides. When the estimated
health costs were incorporated in their benefit-cost calculations, the net present
value of pesticide use was found to be negative. Antle and Capalbo (1994) intro-
duced a conceptual framework to explain the health-productivity tradeoffs of pesti-
cide use in developing countries. Crissman et al. (1994) identified a number of
health consequences including acute poisoning, chronic dermatitis, and chronic cen-
tral nervous system damage directly linked to the use of pesticides. According to
them, ill health due to exposure to pesticides resulted in loss of work days, private
health costs, a reduction in work productivity and impairment in decision-making
abilities.
Kishi et al. (1995) examined correlations between exposure to pesticides and
symptoms of pesticide toxicity among Indonesian farmers. According to the
authors, 21% of spray operations resulted in three or more neurobehavioural, respi-
ratory and intestinal symptoms. The number of spray operations per week, the use
of more toxic pesticides and skin and clothes becoming wet with the spray solution
were significantly and independently associated with the number of signs and symp-
toms. Lichtenberg and Zimmerman (1999) analysed the adverse health experiences,
environmental attitudes, and pesticide usage behaviour of 2,700 corn and soybean
growers in Maryland. According to this study, farmers who thought that they had
had adverse health experiences from being exposed to pesticides were likely to have
heightened concerns about the use of pesticides and were more likely to take greater
precautions in dealing with pesticides.
Wilson (2005) used field survey data from Sri Lanka to estimate farmers’
expenditure on defensive behaviour (DE) and to determine factors influencing
DE. His findings show that farmers’ expenditures on DE were inversely related
to incidence of ill health among farmers using pesticides. Pimentel (2005) investi-
gated the costs resulting from using pesticides in agriculture. He estimated the
value of different cost components which included pesticide impacts on public
health, livestock production losses and crop product losses. This study shows
that the public health cost due to the application of pesticides in the USA is
around US$1.1 billion per year. In addition, the study shows that the cost of
crop losses due to the eradication of beneficial insects (e.g. honey and wild bees)
due to the use of pesticides is US$1.4 billion per year. Garming and Waibel
(2009) used a contingent valuation (CV) approach to assess the health effects of

 2012 The Agricultural Economics Society.

 Determinants of Health Costs due to Farmers’ Exposure to Pesticides 161

exposure to pesticides among Nicaraguan vegetable farmers. They found that

farmers were willing to spend an additional amount of about 28% of their cur-
rent expenditures on pesticides to avoid health risks. This implies that farmers in
this sample were willing to pay more to purchase pesticides that are less harmful
to human health.
Typically, the data in these studies have been collected from farmers based on
their perceptions of various symptoms arising from exposure to pesticides. None of
the studies examine farmers who have been diagnosed and treated by doctors for
symptoms resulting from exposure to pesticides. This study uses data collected by
interviewing farmers who have been identified by doctors as suffering from pesti-
cide-induced illnesses and evaluates their costs of ill health. These data are then
compared with a control group of general farmers’ self-reports on pesticide use and
effects. Our objective is to analyse the costs of ill health and to divide the WTP
to avoid pesticide exposure related ill health into different components for the
two samples. An analysis of the costs of ill health and WTP by comparing the two
samples has not been attempted before. In addition, we use the data from the two
surveys to compare the factors that influence the resulting medical costs.

3. Theoretical Approach
We use a model developed by Harrington and Portney (1987) to derive the WTP to
avoid an increase in pollution. Based on the model, CV WTP estimates are greater
than the direct cost of illness and aversion behaviour estimates, because the bids
also capture the disutility7 arising from ill health due to exposure to pesticides. In
the Harrington and Portney (1987) model an individual derives utility from a com-
posite good – his health condition and leisure. We assume that an individual’s well
being increases with aggregate consumption (C) and leisure (L), but is negatively
affected by sick days (S):
U ¼ UðC; L; S; Xu Þ: ð1Þ

Utility is increasing in C (¶U ⁄ ¶C > 0) and L (¶U ⁄ ¶L > 0), while it is decreasing
in S(¶U ⁄ ¶S < 0). Xu is a vector of individual characteristics capturing preferences
for income, leisure and health. We assume that the health of a person is measured
by the number of sick days (S). The number of sick days depends on the level of
exposure to pesticides (P), averting activities (A) and medical treatment (M). While
the level of exposure increases the number of sick days, individuals can spend on
defensive (averting) behaviour – such as wearing protective clothing and use of
masks – to lessen their chances of being harmed by exposure to pesticides. Since
the health outcome is a function of these three variables, the relationship between
exposure to pesticides and health outcomes can be summarised in a dose-response
function as follows:
S ¼ SðP; A; M; Xs Þ ð2Þ

where ¶S ⁄ ¶P > 0, ¶S ⁄ ¶A < 0 and ¶S ⁄ ¶M < 0.

Xs is a vector of individual characteristics such as age and occupation, which also
has some influence on health. It is hypothesised that the number of sick days is

7
Examples of disutility include stress, pain and suffering.

 2012 The Agricultural Economics Society.

162 Wasantha Athukorala, Clevo Wilson and Tim Robinson

negatively related to averting expenditures and medical expenditures, whereas the

number of sick days is positively related to levels of exposure to pesticides. We
assume that an individual allocates his total time (T) between work (W) and leisure
(L) and spends income on aggregate consumption, medical care and averting activi-
ties. The individual chooses the levels of C, L, A and M to maximise utility, subject
to the following budget constraint:
I þ wT ¼ C þ Pm M þ Pa A þ wL þ wSðP; A; M; XS Þ ð3Þ
where I is the non-wage income, Pm and Pa are prices of medical care and averting
activities, respectively. The wage rate is denoted by w. The price of a unit of the
aggregate consumption good is normalised to one. This budget constraint assumes
that individuals allocate their time between work and leisure. The left hand side of
Equation (3) shows the sum of wage income and non-wage income, while the right
hand side of the equation shows the total possible expenditures. In the budget con-
straint, the time allocated to work and medical care is expressed as a function of
the number of sick days. The number of sick days can reduce working hours, while
it can increase medical expenses. This information is used to derive a measure of
the benefits of reductions in the deleterious health effects from farmers’ exposure to
pesticides during handling and spraying. A farmer’s decision problem can then be
expressed as:
€
Max UPðC; L; A; MÞ ¼ U½C; L; SðP; A; M; XS Þ; Xu 
þ k½I þ wT  C  Pm M  Pa A  wL  wS ð4Þ
where k is the Lagrangian multiplier which can be interpreted as the marginal utility
of income. The simultaneous solutions to the first order conditions of (4) show the
demand functions for the composite commodity, leisure, averting activities and
medical treatment. Except for the composite commodity, the other three demand
functions can be expressed as follows:

L ¼ L ðPm ; Pa ; w; I; PÞ ð5Þ

A ¼ A ðPm ; Pa ; w; I; PÞ ð6Þ

M ¼ M ðPm ; Pa ; w; I; PÞ ð7Þ
This model can be used to derive an observable measure of an individual’s (farm-
er’s) WTP to avoid exposure to pesticides. This is expressed as the largest amount
of money that can be taken away from that individual (farmer) without reducing
his or her utility (Freeman, 2003). Here, exposure to pesticides affects utility only
through health.8 WTP is the cost of achieving the optimal level of health made pos-
sible by avoiding exposure to pesticides. Accordingly, we can express the WTP to
avoid ill health from exposure to pesticides as follows:9

8
Since the benefits of using pesticides in terms of greater potential income are exogenous to
this model, this issue is not considered in the analysis.
9
This can be proved using the expenditure function of the associated dual problem. Under
expenditure minimisation, this component is similar to ¶E ⁄ ¶S where E is the expenditure
function.

 2012 The Agricultural Economics Society.

 Determinants of Health Costs due to Farmers’ Exposure to Pesticides 163

dS @A @M @U=@S dS
WTP ¼ w þ Pa þ Pm  ð8Þ
dP @P @P k dP
Note: Proof can be submitted on request.
The first three terms of the right-hand side expresses loss in earnings due to ill-
ness resulting from exposure to pesticides, the cost of averting activities and expen-
ditures on medical treatment. The last term expresses the monetary equivalent of
the disutility of illness (US ⁄ k). Using the CV approach, an individual asked directly
for his or her WTP to avoid harm from exposure to pesticides may consider all
expenditures shown in equation (8) in revealing their WTP bid. Depending on the
adequacy of the defensive expenditures, the first two terms and the last term can be
large or small. If defensive expenditures undertaken are small (inadequate) then the
first two terms and the last term are likely to be larger, ceteris paribus, and vice
versa.
It is important to determine the four components of WTP for policy decision-
making purposes. Consequently, in addition to using a CV WTP question we also
use the cost of illness and the aversion behaviour to derive the values for the respec-
tive components shown in the RHS of equation (8).
It is also seen as important to identify factors influencing ill health and health
expenditures due to exposure to pesticides. For this purpose, a reduced form equa-
tion is estimated to explain direct health costs due to exposure to pesticide as a
function of its determinants. Accordingly we use OLS and WLS methods for a sam-
ple of hospitalised farmers who have been treated for exposure to pesticides and a
Tobit model for the sample consisting of general farmers. Separate models are used
because, for the hospitalised sample, the health cost is always positive, while for the
general farmers’ sample the MC can be zero. Our explanatory variables include
averting expenditure per month (Rs.), average quantity of pesticides used per month
(litres), frequency of pesticide use per month, spraying hours per day, number of
crop varieties grown, number of pesticides used. Socio-economic variables include
age, farm income, education in terms of number of years, and whether or not farm-
ers have read instructions (dummy).

4. Sample and Data Collection

The main objective of this survey was to collect data to show the relationship
between farmers’ exposure to pesticides and the resulting health costs. Two surveys
were conducted covering 217 farmers who were hospitalised due to exposure to pes-
ticides while spraying and 246 general farmers. The approach adopted by Yamane
(1967), Gomez (1984), Bartlett et al. (2001), Lukas (2007) and Dattalo (2008) was
utilised to collect the field survey data.10
In June 2008, a pilot survey of the general farmers was conducted to assess the
impact of exposure to pesticides on the health of farmers in randomly selected areas

10
The sampling technique for the general farmers allowed us to obtain a representative sam-
ple from the selected agricultural areas. However, we are aware of the possible selection bias
that can arise with the hospitalised sample. Most of the hospitals that were visited provided
us with the addresses of the hospitalised farmers. However, a few hospitals could not find
their records or refused to provide them.

 2012 The Agricultural Economics Society.

164 Wasantha Athukorala, Clevo Wilson and Tim Robinson

of the Anuradhapura district in Sri Lanka. The main survey commenced in July
2008 and was completed in December 2008. A questionnaire was completed through
face-to-face interviews with the head or other working member of farm households
in the Galnawa divisional secretariat area in the Anuradhapura district.11 The sur-
veys were conducted covering six villages. Of the 257 interviews conducted, only
data from 246 interviews are used in the analysis.
A slightly different approach was adopted for the hospitalised farmer’s survey.
The farmers studied were those who had been treated by doctors for symptoms aris-
ing from exposure to pesticides while spraying on farm crops. These farmers, as for
the general farmers, cultivate their own land and include both men and women. We
conducted this survey covering farmers in the Anuradhapura, Kurunagala and Am-
para districts. We first visited 36 hospitals and dispensaries to obtain the medical
records of the affected farmers.12 In general, doctors maintain records under three
headings: accidental poisoning, spray poisoning and observations of doctors who
suspect a patient needs treatment due to pesticide related illnesses.
The survey of hospitalised farmers was conducted from September, 2007 to Febru-
ary, 2008. The following procedures were adopted: once the interviewers visited a
hospital in the selected area, they requested doctors to provide addresses of those
patients who have been diagnosed as suffering from ill health due to exposure to pes-
ticides. We then sought the permission of patients (farmers) to conduct the inter-
views. Only farmers who provided their consent to take part in the survey were
interviewed. We collected data from 221 farmers (representing a response rate of
62%). Four responses were incomplete or too inconsistent to be used in the analysis.
The questionnaire used in the survey was validated in a pilot survey and through
focus group discussions. The final questionnaire was adjusted following the pilot
survey and focus group discussions. The gathering of data was conducted by a
trained group of researchers under the close supervision of the research team. The
interviews took place in the interviewee’s home where participants were informed
about the purpose of the study and verbal consent to take part in the study was
requested. A field supervisor reviewed the quality of the data gathered and entered
it into a database for analysis. Confirmation was provided that the survey questions
were clearly understood by respondents and that we were obtaining appropriate
information regarding exposure to pesticides and health effects.

5. Results and Discussion

Table 1 provides the details of descriptive statistics of the variables used in this
study. The average number of work hours lost due to ill health from exposure to
pesticides is approximately 9 hours per person per month for the hospitalised farm-

11
Anuradhapura district was selected because it is an agricultural district with high rates of
poisonings in Sri Lanka. A secretariat division and six villages were selected randomly. The
selection of villages for the sample was conducted using the list of active farmers in the area
provided by the village agricultural extension officer. Every third name on the list was
selected for the interview.
12
Accessing this type of record can be effected with the permission of the registrar of the
respective hospital. As we didn’t need the name of the patient, only addresses were recorded
and used to locate farmers for interview.

 2012 The Agricultural Economics Society.

 Determinants of Health Costs due to Farmers’ Exposure to Pesticides 165

Table 1
Descriptive statistics of the variables used in the analysis

Variables Mean Max Min SD

Hospitalised farmers
Farm income (Rs.)* 25,991 100,000 6,100 13,842
Land size (acres) 2.56 9.00 0.25 1.48
Number of commercial crops grown** 4.00 7.00 1.00 1.25
Education (years) 6.00 12.00 0.00 3.25
Age (years) 37.00 59.00 20.00 11.40
Quantity of pesticides used per month (litres) 4.75 12.00 0.50 2.44
Number of pesticide brands used 5.00 9.00 1.00 1.54
Frequency of pesticides used (per month) 10.00 19.00 6.00 2.78
Direct exposure to pesticides (per day ⁄ hours)*** 4.20 8.00 0.30 1.69
Number of reported symptoms 3.00 7.00 1.00 1.44
Number of work hours lost per month 8.45 44.00 1.00 6.47
General farmers
Farm income (Rs.)* 28,113 180,000 6,000 21,962
Land size (acres) 1.40 5.25 0.50 1.02
Number of commercial crops grown** 3.00 6.00 1.00 1.40
Education (years) 9.00 13.00 0.00 3.47
Age (years) 36.00 58.00 21.00 10.18
Quantity of pesticides used per month (litres) 2.15 8.00 0.20 1.35
Number of pesticide brands used 3.00 6.00 1.00 1.18
Frequency of pesticides used (per month) 3.00 10.00 1.00 2.01
Direct exposure to pesticides (per day ⁄ hours)*** 2.30 7.30 0.30 1.45
Number of reported symptoms 2.00 6.00 0.00 1.40
Number of work hours lost per month 3.00 13.00 0.00 2.89

Notes: i. Sample sizes for the hospitalised and general farmers are 217 and 246, respectively.
*Farm income includes only direct income received by selling farm products.
**Only the crops cultivated for sale in the market are included in this category.
***This includes spraying hours as well as time spent handling and mixing pesticides.
ii. The daily wage rate in Sri Lanka varies between Rs. 400 and Rs. 500 for different areas.
The average daily wage is approximately Rs. 450 (USD 4.5 using the exchange rate of Rs.
100 equals USD 1). Accordingly, the hourly wage rate used to calculate the loss in earnings
was Rs. 56.25.

ers. On the other hand general farmers lost only 3 hours on average. Reflecting this
disparity, the average farm income is higher for the general farmers, even though
land size is smaller. Hospitalised farmers were shown to use more pesticides and use
them more frequently.
The average number of self-reported illnesses on a typical pesticide spraying
day13 is three and two for the hospitalised and general farmers, respectively. The
average number of schooling years for the general farmers was nine, while it was
six for the hospitalised farmers. As expected, the quantity of pesticides used per
month (in litres), number of pesticides brands used, frequency of use (pesticides

13
This is while spraying ⁄ mixing pesticides or at most within 12 hours after spraying or han-
dling pesticides.

 2012 The Agricultural Economics Society.

166 Wasantha Athukorala, Clevo Wilson and Tim Robinson

Table 2
Medical costs incurred under different scenarios14 (Rs. ⁄ per month)

Scenarios Hospitalised farmers Percentage General farmers Percentage

A, B and C 324 (0.72) 45.62 287 (0.64) 1.63

A and B 250 (0.56) 13.82 262 (0.58) 1.22
A and C 180 (0.40) 23.96 165 (0.37) 1.22
A 157 (0.35) 16.58 138 (0.31) 0.81
B and C – – 154 (0.34) 1.63
B – – 120 (0.27) 1.22
C – – 53 (0.11) 56.50
Average 228 (0.51) 100.00 168 (0.37) 64.23
NSDs 50 (0.11) – 29 (0.06) –
Zero Cost 0.00 0.00 0.00 35.77
N 217 – 246 –

Notes: i. Monthly medical costs are reported for a typical cultivation season for farmers who
have different experiences due to exposure to pesticides. The medical costs for hospitalised
farmers or those who take treatment for being sick soon after spraying (within 12 hours) or
while spraying ⁄ mixing pesticides were calculated. For the general farmers’ sample, only 158
farmers reported any form of costs. NSD denotes average medical costs incurred due to
exposure to pesticides on non-spraying or mixing days. Zero costs refer to farmers who did
not fall into any of the above mentioned categories during the specified cultivation seasons.
ii. Costs as a ratio of daily wages are shown in parentheses.

used per month) and direct exposure to pesticides (per day ⁄ hours) were higher for
hospitalised farmers than for the general farmers. The health symptoms most com-
monly reported by all respondents (both samples) included headaches (68%), eye
irritation and tearing (49%), pain in muscles, joints, or bones (31%), a rash or
cramps (30%), and difficulty in breathing (16%).
Sample t-tests (mean comparisons) show that there is no significant difference in
the two samples for variables such as age, education, number of pesticide brands
used and the number of commercial crops grown on their farms. However, farm
income, direct exposure time and quantity of pesticides used per month show signif-
icant differences between the two samples (at 5%). Land size, frequency of pesti-
cides used per month and the number of work hours lost also show significant
differences between the two samples (significant at 1%).
Using these data, we first investigated changes in MC under different scenarios.
This analysis considered the variation of costs to farmers who have different experi-
ences of illnesses due to direct exposure to pesticides. The average cost was calcu-
lated for farmers classified as serious (A: hospitalisation), moderate (B: a doctor is
consulted, but no hospitalisation is required) and mild cases (C: no visits to the doc-
tor, yet medication is taken). It is noted that a farmer may have experienced one,
two or all three of the above. Table 2 shows MC incurred by the two groups of

14
These costs were calculated using average monthly health costs as well as by taking into
account the number of times a farmer has suffered from ill health. For example, MC of gen-
eral farmers who have experienced B and C (Rs.154.16) are greater than the cost for farmers
who have only been hospitalised (Rs. 138.33). The reason for this is that the frequency of
hospital visits is lower than B and C.

 2012 The Agricultural Economics Society.

 Determinants of Health Costs due to Farmers’ Exposure to Pesticides 167

farmers. As can be seen, there is considerable variation in the costs incurred for the
three categories. It is interesting to note that the reported average cost of hospita-
lised farmers is higher than that of the general farmers for most of the three catego-
ries and on non-spraying days (NSDs).
Table 2 shows that on a typical pesticide spraying day or soon afterward, approx-
imately 4.8% of the general farmers interviewed had been admitted to hospital and
incurred costs while 8% had taken treatment from a doctor. This figure is consis-
tent with other studies conducted (see, for example, Murray et al., 2002; Garming
and Waibel, 2009). Approximately 56% of the general farmers took home-made
self-treatment and incurred other private costs. Approximately 71% of this group
said that they have suffered from some form of acute ill health and incurred costs
from pesticide spraying days during the last 3 years. However, 34% of the inter-
viewed farmers in this group said that they did not suffer any form of ill health
and did not incur any form of expenditure due to exposure to pesticides during the
previous cultivation season.
Of the hospitalised farmers, the majority (45%) incurred all forms of costs (i.e.
A, B and C) and 84% said that they had been hospitalised plus taken treatment at
home without consulting a doctor. This is because most farmers in this group
resorted to home-made treatment which did not incur significant costs.
In addition to incurring medical expenditures, loss in earnings from being unable
to work is a large cost to farmers (see, for example, Huang et al. 2000; Garming
and Waibel, 2009). The opportunity cost of lost labour hours due to ill health from
direct exposure to pesticides and averting expenditures is shown in Table 3.
The results indicate that during a typical cultivation season, farmers are losing
between half and one day value of labour each month due to ill health from

Table 3
Loss in earnings due to exposure to pesticides and averting expenditures (Rs. ⁄ per month)

Loss in earnings* Averting expenditures**

Hospitalised General Hospitalised General

Scenario farmers farmers farmers farmers

A, B and C 731 (1.62) 590 (1.31) 23 (0.05) 34 (0.07)

A and B 519 (1.15) 379 (0.84) 44 (0.09) 54 (0.12)
A and C 316 (0.71) 412 (0.91) 43 (0.09) 41 (0.09)
A 332 (0.73) 346 (0.77) 65 (0.14) 56 (0.12)
B and C – 253 (0.56) – 50 (0.11)
B – 241 (0.53) – 53 (0.12)
C – 177 (0.39) – 68 (0.15)
Average*** 475 (1.05) 343 (0.76) 44 (0.09) 51 (0.11)
N 217 246 217 246

Notes: *The daily wage rate in Sri Lanka varies between Rs. 400 and Rs. 500 for different
areas. However, Rs. 450 was used as the average daily wage rate. Accordingly, the hourly
wage rate is assumed as Rs. 56.25 when calculating the loss in earnings.
**Mitigating expenditure mainly includes costs incurred wearing protective clothing, masks,
gloves and shoes.
***Average for general farmers: calculated using the 158 farmers who had at least one expe-
rience of the above conditions.
Costs as a ratio of daily wages are shown in parentheses.

 2012 The Agricultural Economics Society.

168 Wasantha Athukorala, Clevo Wilson and Tim Robinson

Table 4
Estimated total costs to farmers due to exposure to pesticides (Rs. ⁄ per month)

Hospitalised As a percentage General As a percentage

Scenario farmers of farm income farmers of farm income

A, B and C 1136 (2.52) 4.37 970 (2.15) 3.45

A and B 864 (1.92) 3.32 725 (1.61) 2.58
A and C 591 (1.31) 2.27 668 (1.48) 2.00
A 602 (1.34) 2.31 561 (1.25) 1.99
B and C – – 482 (1.07) 1.71
B – – 432 (0.96) 1.54
C – – 307 (0.68) 1.10
Average 798 (1.77) 3.07 592 (1.32) 2.10

Notes: i. Total costs include mainly medical, labour and averting costs. Medical costs include
both costs incurred on spraying days and non-spraying days. The sample average income is
used to show the total costs as a percentage of farmers’ average income. Average farm
incomes of hospitalised and general farmers are Rs. 25,991 and Rs. 28,113, respectively (see
Table 1).
ii. Costs as a ratio of daily wages are shown in parentheses.

exposure to pesticides. We specifically asked whether they were aware of the value
of labour hours lost due to exposure to pesticides in a typical cultivation season.
More than 90% of farmers said they have not thought about it. This indicates the
opportunity cost of lost labour hours is low – a problem common to markets not
fully monetised.
Although in the study areas pesticide spraying is undertaken every 2 to 6 days,
the extent of the precautions taken is low for both farmer groups (Table 3). A simi-
lar conclusion is drawn by Wilson (2005) in a study analysing the averting behav-
iour of pesticides of farmers in Sri Lanka. Our data also show that the average
monthly averting expenditure of farmers who said that they incurred no costs from
spraying pesticides is more than double (Rs. 108) the average of farmers who
reported expenditures due to pesticide related ill health.
Average total cost estimates for both samples are shown in Table 4. The average
estimated total cost incurred by the hospitalised sample is equal to approximately
3% of an average farmer’s monthly income during a typical cultivation season. This
figure is equivalent to a loss of approximately 2% of their monthly farm income for
the general farmer’s sample. The costs to farmers who have incurred all three types
of medical expenditures are relatively higher for both samples. As a percentage of
farm income it is approximately 4% and 3% for the two groups, respectively.
In this article, we have not estimated the magnitude of the benefits of using pesti-
cides. However, the majority of farmers (87%) surveyed were of the view that their
crop loss will be more than 75% if they do not use pesticides on their crops.
It is important to note that more than 60% of total health costs incurred were
due to loss in earnings and represented a hidden cost to farmers. Labour is a scare
resource during the peak cultivating and harvesting periods in Sri Lanka, but less
so during other periods when spraying usually takes place. Farmers are therefore
less likely to place a high value on their opportunity cost of lost hours in non-peak
cultivating and harvesting periods. This would explain why most farmers do not
take into account this cost when deciding to use pesticides.

 2012 The Agricultural Economics Society.

 Determinants of Health Costs due to Farmers’ Exposure to Pesticides 169

Table 5
Farmers’ CV WTP to avoid exposure to pesticides under its four components (Rs. ⁄ per
month)

Hospitalised Costs as a General Cost as a

Scenario farmers percent of WTP farmers* percent of WTP

Medical expenditure 279 29.35 198 31.80

Loss in earnings 475 49.94 343 55.08
Averting expenditure 44 4.64 51 8.28
Disutility 152 16.05 30 4.82
WTP 951 1.00 622 1.00

Notes: WTP, willingness to pay; CV, contingent valuation.

*Only 196 farmers who have incurred any form of expenditure due to exposure to pesticides
are used to calculate WTP for general farmers. This includes 32 farmers who reported some
costs, but stated that their CV WTP was zero.

Dividing the total costs into MC, loss in earnings and averting expenditures
allows a comparison with farmers’ WTP to avoid exposure to pesticides. The esti-
mated costs under total medical expenditures, loss in earnings, averting expenditure
and disutility are shown in Table 5. The value of disutility is obtained from the
residual of equation 8. As discussed in the methodology section, WTP to avoid any
cost resulting from pesticide induced illnesses has four components namely, the
monetary value of direct medical expenditures, loss in earnings, averting expenditure
and the monetary value of disutility. The monetary value of disutility is the differ-
ence between the average WTP and the sum of the monetary value of direct medical
expenditures plus loss in earnings and averting expenditure.15 In the hospitalised
sample, the average WTP is approximately Rs. 950 per month for a typical cultiva-
tion season – equal to two days wage of an average farmer. The average WTP of
those general farmers who have had at least one or more incidence of pesticide
induced illness is approximately Rs. 620. As expected, the cost share of the averting
expenditure is relatively higher for general farmers (8.2%) than the hospitalised
sample (4.6%).
As can also be expected the value of the disutility arising from ill health is more
than five times larger for the hospitalised farmers than for the general farmers.
Equally, the relative WTP share of the disutility is 16% for hospitalised farmers,
compared to approximately 5% for the general farmers.
According to the Sri Lanka Department of Census and Statistics, the total num-
ber of smallholding farmers16 in 2002 was approximately 1,780,000. Assuming that
all these farmers are using pesticides and 5–20% suffer from pesticide exposure
related illnesses, the annual costs of medical expenditures, loss in earnings and

15
When we asked the CV question to obtain the WTP bid to avoid any cost resulting from
pesticide induced illnesses, we asked the farmers to consider the monetary value of all direct
and indirect costs (i.e. medical expenditures, loss in earnings, averting expenditure and the
cost of the disutility).
16
Farmers reporting a holdings size of 40 perches (0.10 hectares) or less were excluded. Fur-
thermore, we did not include plantation workers who mainly cultivate tea, rubber and coco-
nut. Only farmers who cultivate cash crops for commercial purposes are used in the analysis.

 2012 The Agricultural Economics Society.

170 Wasantha Athukorala, Clevo Wilson and Tim Robinson

Table 6
Cost scenarios for general farmers per year (Rs. million)

Type of costs Scenario (a) Scenario (b)

Medical expenditure 140 563

Loss in earning 244 976
Averting expenditure 36 146
Total cost 421 1,687

Notes: The estimated average cost of medical expenditures, loss in earnings, averting expen-
ditures as well as their total are shown in the table. It is assumed that (a) 5% (89,000), and
(b) 20% (356,000) of total farmers growing commercial crops (1,780,000) are affected by pes-
ticide use.

averting expenditures for the country can be estimated. Table 6 shows the estimated
annual costs of using pesticides when 5% (Scenario (a)) and 20% (Scenario (b)) of
Sri Lankan farmers are assumed to be affected.17
It is clear that the loss in social welfare is significant in each year. Of this, the loss
in earnings is the most significant component. However, low expenditures on defen-
sive behaviour, low levels of education and poverty help explain why these losses
continue to occur. Evidence of pesticide-related ill health, both direct and indirect,
is also shown by Maumbe and Swinton (2003) and Wilson (2005). According to the
Maumbe and Swinton (2003) study, cotton growers’ loss in income in Zimbabwe
was equivalent to 45–83% of the annual household pesticide expenditures in the
study areas. The time spent recuperating from illnesses attributed to exposure to
pesticides averaged between 2 and 4 days during the 1998 ⁄ 1999 growing season.
According to Wilson (2005) the cost of defensive expenditure is approximately 40
million rupees a year if 100,000 farmers use pesticides in the country. However, if a
figure of 300,000 farmers is used the cost rises to more than 121 million rupees a
year. Aggregate health costs due to pesticide exposure are, therefore, large for devel-
oping countries such as Sri Lanka. Moreover, long-term costs arising from exposure
to pesticides are also likely to be high (Wilson, 2005). The case for reducing the cur-
rent high levels of exposure to pesticides is therefore very strong. To do so an
understanding of what factors influence direct health costs is needed so that appro-
priate policy instruments for remedial action can be put in place.
A number of factors influence MC. They include averting expenditure (AE), aver-
age quantity of pesticides varieties used per month (APQ), number of pesticides
used (NPV), number of hours spraying pesticides per day (HPS), frequency of pesti-
cide use per month (FP), number of crops varieties grown (NCV), income (IN),
level of education (EDU), age (AGE) and whether or not farmers have read instruc-
tions on pesticide containers (dummy-D). It is hypothesised that the higher the
APQ, NPV, HPS, FP, NCV and IN, the higher would be the level of MC. How-
ever, AE can reduce MC. That is, there could be a negative relationship between
MC and EDU given more educated people are more likely to take appropriate
precautions when using pesticides.

17
If we assume that approximately 2–8% of farmers are hospitalised (Garming and Waibel,
2009) each year, cost scenarios for hospitalised farmers per year would be 22 million and 90
million.

 2012 The Agricultural Economics Society.

 Determinants of Health Costs due to Farmers’ Exposure to Pesticides 171

Table 7
Results of WLS, OLS and Tobit models

Hospitalised farmers General farmers

Variables MC (WLS) MC (OLS)a MC (Tobit)b

Constant 113.04 (27.75)* 114.67 (45.37)**

AE )0.16 (0.08)** )0.22 (0.09)** )0.14 (0.06)**
APQ 8.80 (1.51)* 12.12 (2.70)* 15.45 (3.89)*
NPV 9.71 (2.59)* 6.76 (3.62)*** 1.05 (3.15)
HPS 13.02 (3.52)* 14.34 (4.87)* 21.08 (4.05)*
FP 6.65 (1.31)* 4.22 (1.71)** 12.37 (3.22)*
NCV 10.92 (3.16)* 12.05 (4.01)* )4.16 (3.25)
IN 0.01 (0.00)* 0.02 (0.00)* 0.01 (0.00)**
EDU )4.13 (1.30)* )5.22 (1.70)* )9.45 (1.80)*
AGE )1.22 (0.36)* )1.45 (0.48)* )1.88 (0.47)*
D 5.30 (8.52) 4.55 (11.38) )2.73 (9.31)
Adjusted R2 0.824 0.849
N 217 217 246

Notes: WLS, weighted least squares; MC, medical costs.

i. Standard errors are shown in brackets.
*Denotes the significant variables under 1% level of significance while ** and *** show the
significant variables under 5% and 10% levels, respectively.
a
OLS estimators are with robust standard errors.
b
Marginal effects on the latent dependent variable are reported for the Tobit model. These
coefficients indicate how a one unit change in an independent variable alters the latent
dependent variable.

Tests performed showed some degree of heteroscedesticity. There are a number

of ways of overcoming this problem including using weighted least squares (WLS)
and taking robust standard errors. Both these methods were used for the hospita-
lised sample when estimating the health cost function. A Tobit analysis is used for
the general farmers sample given, as discussed above, it is the more theoretically
appropriate method. The results of the WLS, OLS and Tobit analyses for the two
samples are shown in Table 7.
It is clear that the AE variable is significant in all models and shows the expected
sign. The results therefore indicate that when farmers take more precautionary
action, their MC due to exposure to pesticides is relatively low. This means that the
greater the precautions taken, the lower the level of health expenditures ⁄ ill health.
As expected, APQ, HPS and FP are significant and all models have the correct
signs. However, NPV and NCV are significant only for the hospitalised sample,
while EDU, AGE and INC are significant for both groups. This indicates that the
higher the level of EDU, the more careful they are when handling pesticides. Fur-
thermore, it can be assumed that educated farmers are more likely to follow instruc-
tions when they are handling pesticides.18 The sign of the age variable is negative,
implying that older farmers’ costs of pesticide related illnesses are less than for
young farmers. A study conducted by Sivayoganathan et al. (1995) to examine the

18
Correlation matrix among variables revealed a significant correlation between education
and averting expenditure.

 2012 The Agricultural Economics Society.

172 Wasantha Athukorala, Clevo Wilson and Tim Robinson

use of protective gear by pesticide applicators and its relationship with ill health in
Sri Lanka came up with similar results. A possible reason could be that farmers’
tolerance to exposure to pesticides could increase with age (see, for example,
Jeyaratnam et al., 1997).
The IN variable is significant and has the expected sign with a low coefficient
value. The ‘read instruction’ (dummy) variable is not significant for both models.
One possible reason is that farmers are unable to understand instructions on the
pesticide bottles and, therefore, do not adhere to them. Other reasons why farmers
do not often adhere to instructions and warnings are examined by Sivayoganathan
et al. (1995) and Wilson (2005).
For all the models the signs reported for these variables are not surprising for Sri
Lankan subsistence farmers. Given the low level of agricultural extension services
and education, farmers tend to spray excessive amounts of pesticides in line with
expectations of higher future yields. This has resulted in significant additional costs
being incurred from exposure to pesticides induced ill health. The results show that
a significant portion of costs could be reduced using appropriate educational
campaigns targeting pesticide users in the country.

6. Conclusions and Policy Implications

The major findings from this study can be summarised as follows. First, on average
a farmer incurs a cost between 2% and 3% of their monthly income due to expo-
sure to pesticides during a typical cultivation season. Second, the monthly average
loss in earnings in a typical cultivation season for a farmer in the hospitalised sam-
ple is Rs. 475, compared to Rs. 345 for the general farmers’ sample. These levels
are approximately equal to a farmer’s day’s wage during the survey, which is in the
range of Rs. 450–500 depending on the district. Third, the mitigation expenditures
are very low among Sri Lankan farmers. For example, the average mitigating
expenditure was only 1.12% and 1.73% of their monthly pesticide expenditure for
the hospitalised and general farmers’ sample, respectively. Net gains could be
derived from higher expenditure on mitigation.
Fourth, it was found that loss in earnings is the dominant element of a farmer’s
WTP to avoid exposure to pesticides accounting for approximately half of the total.
The estimated percentage contribution to WTP for each of medical expenditures,
loss in earnings, mitigating expenditures and disutility are 29%, 50%, 5% and
16%, respectively for the hospitalised farmers and 32%, 55%, 8% and 5%, respec-
tively for the general farmers. The value of the disutility is shown to be low given
most farmers place a low valuation on these costs. Finally, it was found that avert-
ing expenditure, the quantity of pesticides used per acre per month, the frequency
of pesticide use and the number of hours spent spraying pesticides per day are the
most important determinants of MC for both samples.

19
The discussion with village officers, agricultural extension workers and some of the doctors
revealed that the weak regulation, low pesticide hazard awareness among users, inadequate
use of personal protective equipment, lack of proper care during application and the use of
highly toxic pesticides are some of the important reasons for the resulting adverse health
effects due to exposure to pesticides.

 2012 The Agricultural Economics Society.

 Determinants of Health Costs due to Farmers’ Exposure to Pesticides 173

No previous study has conducted an analysis of this nature which compares

the results of two samples of farmers using pesticides. This study shows clearly
that pesticide use in Sri Lanka incurs a considerable annual welfare loss to farm-
ers. And unlike many previous studies of acute pesticide poisoning, this study
more clearly identifies the specific issues relating to causes and outcomes. The
findings therefore provide a more persuasive case for the implementation of poli-
cies19 by relevant authorities which encourage mitigating behaviour among farmers
and thereby reduce the costs associated with ill health as shown in Table 6. The
total savings are shown to be approximately Rs. 443 million per annum under
the minimum possible outcome scenario. In order to minimise the cost of gov-
ernment mitigation programs, a targeted awareness programs could be imple-
mented through agricultural extension services. This study also indicates that
given the extent of over use of pesticides their subsidisation could be reduced or
possibly removed altogether in the long term. Government extension services are
clearly best positioned to bringing about these changes and effect a significant
improvement in farmers’ welfare.

References
Antle, J. M. and Capalbo, S. M. ‘Pesticides, productivity, and farmer health: Implications
for regulatory policy and agricultural’, American Journal of Agricultural Economics, Vol.
76, (1994) pp. 598–605.
Antle, J. M. and Pingali, P. L. ‘Pesticides, productivity, and farmer health: Implications for
regulatory policy and agricultural research’, American Journal of Agricultural Economics,
Vol. 76, (1994) pp. 598–602.
Bartlett, J. E., Kotrlik, J. W. and Higgins, C. C. ‘Organizational research: Determining
appropriate sample size in survey research’, Information Technology, Learning, and Perfor-
mance Journal, Vol. 19, (2001) pp. 43–50.
Crissman, C. C., Cole, D. C. and Carpio, F. ‘Pesticide use and farm worker health in Ecu-
adorian potato production’, American Journal of Agricultural Economics, Vol. 76, (1994)
pp. 593–597.
Dattalo, P. Determining Sample Size: Balancing Power, Precision and Practicality (New York:
Oxford University Press, 2008).
Freeman, A. M. The Measurement of Environment and Resource Values: Theory and Methods
(Washington DC: Resources for the Future, 2003).
Garming, H. and Waibel, H. ‘Pesticides and farmer health in Nicaragua: a willingness-to-pay
approach to evaluation’, The European Journal of Health Economics, Vol. 10, (2009) pp.
125–133.
Gomez, K. A. Statistical Procedures for Agricultural Research (Oxford: John Wiley & Sons,
Inc., 1984).
Harrington, W. and Portney, P. R. ‘Valuing the benefits of health and safety regulation’,
Journal of Urban Economics, Vol. 22, (1987) pp. 101–112.
Huang, J., Qiao, F., Zhang, L. and Rozelle, S. ‘Farm Pesticide, Rice Production, and Human
Health’ (EEPSEA Research Reports, International Development Research Centre: Ottawa,
2000).
Jeyaratnam, J., Alwis, J., Seneviratne, R. S. and Copplestone, J. F. ‘Survey of pesticide poi-
soning in Sri Lanka’, Bulletin of the World Health Organization, Vol. 60, (1997) pp. 615–
619.
Kishi, M., Hirschhorn, N., Djajadisastra, M., Satterlee, L. N., Strowman, S. and Dilts, R.
‘Relationship of pesticide spraying to signs and symptoms in Indonesian farmers’, Scandi-
navian Journal of Work, Environment and Health, Vol. 21, (1995) pp. 124–133.

 2012 The Agricultural Economics Society.

174 Wasantha Athukorala, Clevo Wilson and Tim Robinson

Lichtenberg, E. and Zimmerman, R. ‘Adverse health experiences, environmental attitudes,

and pesticide usage behaviour of farm operators’, Risk Analysis, Vol. 19, (1999) pp. 283–
294.
Lukas, B. A. Marketing Research (New South Wales: McGraw-Hill Higher Education, 2007).
Maumbe, M. B. and Swinton, S. M. ‘Hidden health costs of pesticide use in Zimbabwe’s
smallholder cotton growers’, Social Science and Medicine, Vol. 57, (2003) pp. 1559–1571.
Murray, D., Wesseling, C., Keifer, M., Corriols, M. and Henao, S. ‘Surveillance of pesticide-
related illness in the developing world: Putting data to work’, International Journal of
Occupational and Environmental Health, Vol. 8, (2002) pp. 243–248.
Pimentel, D. ‘Environmental and economic costs of the application of pesticides primarily in
the United States’, Environment, Development and Sustainability, Vol. 7, (2005) pp. 229–
252.
Pingali, P. L., Marquez, C. B. and Palis, F. G. ‘Pesticide and Philippine rice farmer health: A
medical and economic analysis’, American Journal of Agricultural Economics, Vol. 76,
(1994) pp. 587–592.
Sivayoganathan, C., Gnanachandran, S., Lewis, J. and Fernando, M. ‘Protective measure use
and symptoms among agropesticide applicators in Sri Lanka’, Social Science and Medicine,
Vol. 40, (1995) pp. 431–436.
Van Der Hoek, W., Konradsen, F., Athukorala, K. and Wanigadewa, T. ‘Pesticide poison-
ing: A major health problem in Sri Lanka’, Social Science and Medicine, Vol. 46, (1998)
pp. 495–504.
Wesseling, C., Castillo, L. and Elinder, C. G. ‘Pesticide poisonings in Costa Rica’, Scandina-
vian Journal of Work, Environment & Health, Vol. 19, (1993) pp. 227–235.
Wilson, C. ‘Private costs and the relation between pesticide exposure and ill health: Evidence
from Sri Lanka’, Environmental Economics and Policy Studies, Vol. 5, (2002) pp. 213–227.
Wilson, C. ‘Exposure to pesticides, ill-health and averting behaviour: Costs and determining
the relationships’, International Journal of Social Economics, Vol. 32, (2005) pp. 1020–1034.
Yamane, T. Statistics: An Introductory Analysis, 2nd edn (New York: Harper and Row,
1967).

 2012 The Agricultural Economics Society.