International Journal of Disaster Risk Reduction 18 (2016) 107114

Contents lists available at ScienceDirect

International Journal of Disaster Risk Reduction

journal homepage: www.elsevier.com/locate/ijdrr

An empirical assessment of farmers' risk attitudes in ood-prone areas

of Pakistan
Shahab E. Saqib a,n, Mokbul Morshed Ahmad a, Sanaullah Panezai b, Irfan Ahmad Rana a
a
b

Department of Regional and Rural Development Planning, Asian Institute of Technology, Thailand
Department of Disaster Management and Development Studies, University of Balochistan, Quetta, Pakistan

art ic l e i nf o

a b s t r a c t

Article history:
Received 19 April 2016
Received in revised form
20 June 2016
Accepted 20 June 2016
Available online 21 June 2016

Farmers are confronted with several sources of climatic risks. As such, their risk attitudes play an important role in farm management decisions. Few studies have attempted to explore farmers' risk attitudes in ood-prone areas. This study examines the effects of socio-economic factors on risk attitudes of
farmers in a ood-prone area of Pakistan. The data were collected from 168 subsistence farmers through
a standardized questionnaire. The farmers were selected through multi-stage sampling techniques. For
farmers' risk attitude measurement, Equally Likely Certainty Equivalent (ELCE) method and a cubic utility
function were employed. Risk perceptions of farmers were measured by the risk matrix technique. A
Logit model was employed to investigate the effects of socio-economic factors on farmers' risk attitudes.
The ndings of the study reveal that the majority of farmers were risk averse in nature. The results for
the logit model show that education, experience, farmers' group, landholding size, off-farm income, and
risk perceptions of oods signicantly affect the risk attitude of farmers. The study provides useful insights into the most important factors affecting the risk attitude of farmers. The results have implications
for policy makers in providing farmers with accurate risk mitigating and management tools, such as
agricultural credit and crop insurance, to cope with climatic risks.
& 2016 Elsevier Ltd. All rights reserved.

Keywords:
Risk attitude
Risk perception
Socio-economic factors
Floods
Pakistan

1. Introduction
Since 2010, the agricultural sector in Pakistan has faced three
massive oods that had devastating impacts on the entire economy, particularly in the agriculture sector. The monsoon oods of
2010, 2011, and 2014 caused huge damage to agricultural crops,
sheries, forestry, livestock, and primary infrastructure, such as
water channels, tube wells, houses, people, seed stocks, animal
shelters, fertilizers and agricultural equipment/machinery. The
oods struck just before the harvesting period of the main crops:
rice, cotton, sugarcane, maize and vegetables. The total production
loss of paddy, sugar cane, and cotton was assessed at 13.3 million
metric tons. Over two million hectares of standing crops were
damaged, and over 1.2 million livestock, excluding poultry, died in
the 2010 ood [59]. In 2011, another massive ood struck Sindh
and Balochistan provinces, which severely affected these areas.
The people suffered from a loss of livelihood, especially relating to
agricultural activities. Approximately 80% of the Sindh's rural population is dependent upon agricultural activities for their
n

Corresponding author.
E-mail addresses: shahabmomand@gmail.com (S.E. Saqib),
morshed@ait.asia (M.M. Ahmad), sanaullah.panezai@gmail.com (S. Panezai),
irfanrana90@hotmail.com (I.A. Rana).
http://dx.doi.org/10.1016/j.ijdrr.2016.06.007
2212-4209/& 2016 Elsevier Ltd. All rights reserved.

livelihoods; livestock, crops, sheries and forestry [39]. The ood

in 2011, destroyed standing crops of sugar cane, cotton, sorghum,
rice, vegetables and pulses; livestock also suffered heavy losses.
For instance, approximately 115,500 livestock were killed, and
though around 5 million livestock survived, they were also indirectly affected through disease and displacement. The estimated
total loss was US$ 1,840.3 million, of which 89% was direct damage
and 11% indirect losses. The highest damage (approximately US$
1.84 billion) occurred in the agriculture sector, particularly to
sheries and livestock. The total damage caused by the 2011 oods
has been estimated at US$ 3.7 billion, and the total cost of recovery
and reconstruction estimated at US$ 2.7 billion [43]. In the recent
oods of September 2014, 367 persons died, and over 2.5 million
people were affected by heavy rains and oods. Moreover, 129,880
houses were damaged, and more than 1 million acres of cultivated
land and 250,000 farmers were affected. The cost of recovery and
resilience building were estimated at US$ 439.7 million and US$
56.2 million respectively [40]. These statistics illustrate the fact
that agriculture was the most affected sector due to oods in
Pakistan.
The agriculture sector is highly dependent on variations in
climatic conditions, thus making it a risky enterprise. Climate
variability is the main source of risk for agriculture and food systems [13]. The rising severity and frequency of extreme weather

108

S.E. Saqib et al. / International Journal of Disaster Risk Reduction 18 (2016) 107114

have substantially damaged agriculture [30]. Farmers are routinely

exposed to various natural disasters, erratic rainfall and pests. For
example, farmers are confronted with heavy rains, oods, pests
and diseases [29,54,55], droughts [54], and market price uctuations [29]. According to Musser and Patrick [38], there are ve
important sources of risk factors in agriculture: production, nancial, marketing, legal, environmental, and human resources.
First, production risks associated with variations in crop yields and
livestock from several sources, such as uncertain weather conditions, incidence of disease and pests. Second, nancial risks: i.e. a
farmer's ability to pay their bills to continue farming and avoid
bankruptcy. Third, marketing risks, which involve uctuations in
prices of agriculturally produced commodities. Fourth, the legal
and environmental risks associated with agriculture. Fifth, limited
human resources, such as the unavailability of family members for
labor and farm management. As the outcomes of these risks can
negatively affect production levels causing considerable production losses, it is therefore crucial for farmers to perceive and
manage production risks accordingly [19].
Farmers' attitudes toward agricultural risks are very important
for planning risk management strategies. Dadzie and Acquah [15]
revealed that farmers' attitudes toward risk are the foremost step
in understanding the behavior and coping strategies they adopt to
mitigate the effects of environmental risks. Farmers' risk attitudes
are critical in the adoption of modern agricultural technologies,
such as production and investment decisions, in agriculture [33].
Han and Zhao [26] argue that special consideration needs to be
given to the risk attitudes of farmers because risk-averse farmers
are less likely to adopt new practices due to uncertainty about the
costs and returns of their strategies. Many studies have reported
that farmers, particularly poor farmers, are at high risk to natural
disasters [4,9,18,29,53]. However, this risk factor is of an adverse
nature, which negatively affects poor farmers' attitudes; they are
therefore reluctant in adopting new technologies in agriculture
[17,20]. Showing a different perspective, Yu et al. [61] reported
that crop and variety selection were the most common methods
adopted by farmers in Northeast China to cope with the effects of
climate change, as opposed to disaster prevention and infrastructure improvement. Hence, due to uncertainty and the frequent occurrence natural disasters, farmers are in a continuous
search for risk coping strategies.
Risk management is a continuous process for farmers. Decisions in these uncertain situations are based on their perception
about the environment, information, their attitudes, and preferences [33]. Ullah et al. [57] found that in risk-prone areas
farmers addressed production risk proactively by using their precautionary savings, agricultural credit, and diversication as risk
management tools at the farm level in Pakistan. Likewise, farmers
adopt diversication beyond the farm, such as diversication in
crops, scheduling of farming practices, migration, and a variety of
other diversication methods such as irrigation and water conservation techniques to cope with climatic risks [7]. Similarly, to
cope with droughts, farmers practiced income diversication, assets depletion, expenditure adjustment, water shortage coping
techniques and migration [6]. However, risk management in
agriculture is not only important for avoiding risk, but also has
ramications concerning the optimum combination of risks and
returns that can result in a wide range of outcomes [27].
Farmers' attitudes toward risk depend on several factors, ranging from cultural background to individual psyche [25]. Farm
household characteristics such as experience and education also
affect risk attitudes and risk perceptions [29,54] stated that educated farmers perceive crop disease as less risky, which resulted in
a negative relationship with risk aversion, whereas experience was
found to be positively related. Likewise, other studies reveal that
the risk attitudes of farmers differ [28,36], with income

[14,15,29,55] and with age [15,29,32,51]. Similarly, farm size

[32,36], land ownership status [36,54], off-farm employment [33],
farm size [29], and farmers' risk perceptions [55] greatly affect the
risk attitudes of farmers.
Climatic risks in the agriculture sector have long been studied,
which has had a substantial inuence on farmers' production
decisions. The literature has not only addressed the risk coping
strategies adopted by farmers, but also the government policies
that are being initiated, particularly risk reduction policies. Risk is
clearly the main characteristic of any agricultural decision. However, there is a gap in our knowledge about the attitudes of
farmers toward risk and where the problems lie, particularly
in situations where risk attitudes are closely associated with the
complex individual characteristics of farmers. Therefore, this study
design is based on two objectives: to determine the risk attitudes
and assess the effects of socio-economic factors of farmers in the
study area.
The paper is divided into six sections. Section 2 is the theoretical framework; Section 3 is about the materials and methods;
Section 4 shows the results of the descriptive analysis and regression model; Section 5 describes the discussion; Section 6 is the
conclusion of the study.

2. Theoretical framework
Different approaches have been adopted by researchers to
measure the attitudes of farmers [15]. Two basic approaches, direct and indirect are used for measuring risk attitude. The direct
method, as suggested by von Neumann and Morgenstern, has
complications that result from the fact that the subjects have
different levels of tolerance or intolerance for gambling and that
the concepts of probability are by no means intuitively obvious,
and moreover, it is a time consuming method [37]. Risk attitude
can be measured through eliciting Certainty Equivalents (CEs) and
the experimental method as gambling with real payoffs [9]. In
interviews for farmers' elicitation of preferences, Anderson et al.
[3] have discussed several techniques. These include the von
Neumann-Morgenstern (N-M) model, Equally Likely Certainty
Equivalent (ELCE) method, a modied version of the N-M model,
and the Equally Likely but risky outcome method. Based on the
above discussion, we have adopted the interview method of the
direct approach with the ELCE, using a Purely Hypothetical Risky
model (explained in Section 3.3). The farmers are categorized into
three groups. First is risk-preferring: those willing to take risks or
the expected outcome is preferred over certain. Second is riskneutral: those who are indifferent to certain and uncertain outcomes, but has the same expected income. Third is risk-averse;
where farmers give preference to certain income over income that
is uncertain. It is assumed that the selection of expected or sure
outcomes is based on utility. Farmers opt for that choice which
gives them more utility. Farmers maximize utility. Utility, in our
case, is a function of wealth, but we use it as a function of income
[27,42].

U = u( w )

(1)

The individual wants to maximize utility with respect to income.

U( w )0

(2)

The rst differential is positive and indicates that more is

preferred over less (also called convex utility function). Likewise,
risk aversion is a state of utility function that shows a decrease in
marginal utility as the payoff increases (also called concave utility
function). Risk neutral has a linear utility function [27].

S.E. Saqib et al. / International Journal of Disaster Risk Reduction 18 (2016) 107114

The expected utility theory is dened by Von Neumann and

Morgenstern [58]. According to this theory there are reasons behind the individual choices involving risks. The decision makers
compare the expected utility in risky and uncertain prospects.
Levy [34] and Gill [21] argued that individuals are reluctant to
accept choices with uncertain payoffs, but rather, are willing to
accept another choice with a low and sure payoff. Farmers will try
to maximize utility within the constraints:

U = u( y , c )

(3)

where y is farm income and c is consumption. The TUF will show

the nature of individual behavior on the basis of convexity or
concavity of the utility function. This will further lead to risk
aversion, which is the central behavioral concept in the expected
utility theory [38]. Risk aversion attitude measures a decisionmakers' unwillingness to accept outcomes with uncertain payoffs.
Instead, they prefer outcomes that are a certain, although with the
probability of lower expected payoffs. A decision-maker's utility
function will shape their risk preferences [27]. A decision-maker's
utility function will have a positive slope, which means that a
greater payoff is always preferred to a lesser one. The nature of risk
attitude is further explained by Arrow [5] and [44], which is
mentioned in Section 3.3.

3. Materials and methods

3.1. Study area
The aim of this paper is to study farmers' nature and behavior
in disaster-prone areas. For this purpose, the study was conducted
in Khyber Pakhtunkhwa province of Pakistan. This is the northernmost province of Pakistan. Khyber Pakhtunkhwa was purposively selected for two reasons. First, the province is vulnerable to
natural disasters such as oods, droughts and storms [45]. Second,
the majority of the people live in rural areas and agriculture is
their main source of income [55]. Mardan District was purposively
selected among 25 districts of the province due to its vulnerability
to oods and heavy rains. Moreover, it is the second largest district
in the province and the 19th largest district of Pakistan [47]. The
total area of the district is 1632 square km, and 80% of the population are dependent on the agricultural sector [46].
3.2. Sampling procedure
The data were collected by multi-stage sampling. First, the
Khyber Pakhtunkhwa province was purposively selected due to its
high vulnerability to natural disasters, as mentioned in the previous section. Second, the Mardan District was selected, as mentioned in Section 3.1. Third, the rural population, composed mainly
of farmers, was purposively selected as the target population.
Fourth, vulnerable farmers, as per the criteria set by the Provincial
Disaster Management Authority [45], were purposively selected. It
is pertinent to note that these farmers were hit by severe oods in
2010. These farmers mostly lived on the bank of the river and
faced oods consistently, every year during the monsoon rains.
However, the severity of oods differs from year to year. Fifth,
subsistence farmers, who make up about 97% of the farming
community and possess landholdings of up to 12.5 acres, were
therefore purposively selected [1]. Last, the data were collected
through random sampling from the lists prepared with Kisan1
councilors. A total of 970 households were identied by the PDMA
1
A Kisan Councilor is the farmers' elected representative as per the K.P.K local
government act of 2013.

109

as vulnerable farmers in the study area that are eight villages

(Fig. 1). Applying the Yamane [60] formula, a sample size of 168
households was determined to be at a 95% condence level with a
77% margin of error:

n=

N
(1+Ne2)

(4)

nSample size.
N Total number of farming households in an area.
ePrecision value, set at 77% (0.07).
3.3. Risk attitude
The Equally Likely Certainty Equivalent Method (ELCEM) is used
to calculate the risk attitude of farmers. Several studies have
adopted this model [27,29,41,50,52]. Certainty equivalence for
several risky outcomes was then compared with associated utility
values [53]. For example, farmers were asked to mention a monetary value between two risky outcomes that would make them
indifferent: the annual income of a sample farmer is PKR 200,000,
with an associated probability of 0.5, and in case of loss, 0 income
with the same probability of 0.5; the farmer is asked to choose the
income in this range. For example, say the farmer was indifferent in
PKR 120,000, which was an assured outcome. The farmer then had
to choose in the range between PKR 0 and 120,000, and was found
indifferent at PKR 60,000. Likewise, in the next step, a he is asked to
choose in the range between PKR 0 and 60,000 and was found
indifferent at PKR 30,000. The experiment was repeated and the
next amount was PKR 20,000 to which the farmer was indifferent.
Likewise, the farmer was asked to choose between the higher
ranges (PKR 120,000200,000) and were indifferent at PKR 140,000.
Similarly, between PRK 140,000 and 200,000, the farmer was indifferent at PKR 170,000. Similarly, the experiment was repeated,
and several CE points were derived with their associated probabilities. This procedure was repeated for every farmer and the
values were incorporated in cubic utility function (Eq. (5)).
Utility values for certainty equivalence were put in the cubic
utility function that divides the farmers into three categories: risk
seeker, averse or neutral. The utility function is:

ui( w) = 1 + 2w+3w2 + 4w 3

(5)

where s are the parameters and w represents the wealth of the

farmers and their attitudes toward risk, which are dependent on
several factors. However, a signicant theoretical argument has
been shown that there is a link between risk attitude and wealth.
Arrow [5] and Pratt [44] stated that for an individual, absolute risk
aversion should be a decreasing function of wealth. Instead of
wealth, we have used annual income as a substitute for the
household in the cubic utility function [42,55].
After estimation of the model, the rst and second derivatives
of the function are:

U = 2+23w +3 4w 2

(6)

U=23+6 4w

(7)

Then, by using the derivatives, the absolute risk aversion is

calculated by the formula:

ra(w) =

U( W)
U( W)

(8)

where U( w) is 40, and the rst derivative is with respect to

income.
According to Arrow [5] and Pratt [44], the risk aversion coefcient indicates the nature of risk attitude. In the language of

110

S.E. Saqib et al. / International Journal of Disaster Risk Reduction 18 (2016) 107114

Fig. 1. Map of the study area.

mathematics:
ra( w) < 0 implies risk aversion.
ra( w)=0 implies indifference.
ra(w)>0 implies risk seeker.
Example of elicitation of certainty equivalents and computation
of utility values.
Step Elicited CE
Scale

Utility calculation
U (0) 0 and U (200,000) 1

U (120,000) 0.5u (0) 0.5u

(200,000) 0.5
U (60,000) 0.5u (0) 0.5u
(120,000) 0.25
U (30,000) 0.5u (0) 0.5u
(60,000) 0.125
U (20,000) 0.5u (0) 0.5u
(30,000) 0.0625
U (140,000) 0.5u (200,000)
(0.5u (140,000) 0.75
U (170,000) 0.5u (200,000)
(0.5u (170,000) 0.875
U (180,000) 0.5u (200,000)
(0.5u (180,000) 0.937

2
3
4
5
6
7

(120,000; 1.0)(0,
200,000; 0.5, 0.5)
(60,000; 1.0)(0,
120,000; 0.5, 0.5)
(30,000; 1.0)(0, 60,000;
0.5, 0.5)
(20,000; 1.0)(0, 30,000;
0.5, 0.5)
(140,000; 1.0)(200,000,
140,000; 0.5, 0.5)
(170,000; 1.0)(200,000,
170,000; 0.5, 0.5)
(180,000; 1.0)(200,000,
180,000; 0.5, 0.5)

Authors' calculations.

Fig. 2. Risk matrix.

3.4. Risk perception of oods

Risk perception is measured by a ve point Likert scale. This
scale ranges from 1 to 5, where 1 is very low, 2 is low, 3 is medium,
4 is high and 5 is very high. To calculate risk perception, the data
were collected for two dimensions: frequency and severity, and
were entered into the risk matrix (Fig. 2) [41,48].

S.E. Saqib et al. / International Journal of Disaster Risk Reduction 18 (2016) 107114

3.5. Regression model

Table 1
Descriptive analysis of variables.

In this study, we have investigated the factors involved at the

farm level that affect the attitudes of farmers. Our dependent
variable was a binary variable based on the categorization of the
risk aversion coefcient value; in our case only two types of
farmers were found either risk averse or risk seeker. As shown in
Section 4.1, 56% of the farmers were risk averse and the remaining
were risk seekers. If the farmer was risk averse, we assigned a
numerical value 1; otherwise, we assigned a value 0. The dependent variable was a binary in nature, therefore, both the
Classical Linear Regression Model (CLRM) and the Linear Probability Model (LPM) could not be applied. A Logistic regression was
employed because it has several advantages compared to linear
models [22,35]. The CLRM model cannot be applied, and the LPM
model has several problems, such as non-normality of disturbance
terms (ui), the possibility of the Yi value being beyond zero, and
the heteroscedasticity of ui having a lower R2 value [23]. Keeping
in mind the stated problems, the logistic regression model was
considered the most suitable model and was therefore applied.

P
i
Logit:log
= Xi i + i
(1Pi )

111

Variables

Dependent variable
Y
Risk aversion
Independent variables
X1 Age
X2 Education
X3 Experience
X4 Health status
X5
X6
X7
X8
X9
X10

X11

(9)

This can be expressed as:

y=Xii + i
where:
P
y=log (1 i P ) .
i
Pi=The predicted probability of that particular condition occurs.
Xi=Vector of 1 K of independent variables (factors).
i=Vector K 1 of estimated coefcient.
i=Error term.

4. Results
4.1. Descriptive analysis of study variables
The dependent variable in our study is the risk attitude of
farmers. The results from the cubic utility function show that all
the farmers are either risk averse or risk seekers. However, no
farmer was found risk indifferent/ neutral. For this purpose, only
one dummy is used: 1 for the risk-averse farmer, and 0 for
otherwise. Results show that the majority of farmers were risk
averse in nature. The binary variable used for risk aversion had a
mean value of 0.56 (Table 1). In the case of risk attitude, most
farmers in the survey were lower subsistence farmers and had less
landholdings. Due to low assets and income they were risk averse
in nature, while the farmers with larger landholdings were risk
seekers. Likewise, for the independent variables, the mean and
standard deviations were calculated. These variables are categorized into three groups: socio-economic factors, risk perception of
oods and farmers' category (lower subsistence farmers and above
lower subsistence). Among the socio-economic factors, age had an
average value of 48.6 years. Farmers' education level was measured by completed years of school attendance at the time of data
collection. The mean years of schooling was 5.6 for all selected
farmers, which was very low. Similarly, farmers' experience was
also measured in years, with a mean value of 23.9. For the health
status of the farmers, we use the ve point Likert scale, ranging
from very poor (1) to very good (5). Later, 1 and 2 were categorized
into poor health, and 3, 4 and 5 were categorized into good health
[12,24]. In the model, a dummy is included for this variable: 1 for
good health and 0 for otherwise. In addition, family size is the
number of family members living within the same boundary and
sharing a kitchen, income and expenditures. The average family

X12

Description and forms of

expression

Mean

SD

1 Aversion, 0 otherwise

0.56

0.49

46.80
5.60
23.90
0.53

13.80
5.50
14.60
0.50

Age of farmers in years

Education as years of schooling
Farming experience in years
Health status (1 good health,
0 poor health)
Family size
Total number of family members
Monthly off-farm Average off-farm monthly inincome
come in PKR
Land holding size Land holding size in acres
Owned land
Proportion of owned land of total
proportion
land holding in acres (ratio)
Field labor
Ratio of family members working
as labor to total family members
Distance from
Distance of eld from river
river
(1 within 500 m from the bank,
0 otherwise)
Risk perception
Risk perception of oods
of oods
(1 high risk, 0 low risk)
Farmers' groups
Farmers' groups (1 lower subsistence farmer, 0 otherwise)

9.10
3.30
15,694.7 13,494.0
4.40
0.41

4.20
0.42

0.31

0.42

0.60

0.49

0.59

0.49

65.5

0.45

Source: Field survey, 2015.

size was nine members per household. Family income is measured

in Pakistani Rupees (PKR).2 The mean off-farm income per
household per month was PKR 13,494.0. The mean landholding
size was 4.4 acres, while the mean for land ownership proportion
was 0.41. Likewise, the mean value of proportion of farm labor was
0.31. Distance from the river was a dummy, with a mean value of
0.60. The mean value of risk perception was 0.59. We used a
dummy for farmers' groups, where 1 was assigned to subsistence
farmers and 0 to others. The farmers were divided into two
groups: lower subsistence and others. Lower subsistence farmers
had landholdings of less than 5 acres.
4.2. Results of the logistic model
The results of the logistic regression analysis are given in Table 2. The regression model was estimated by using STATA-12. The
Pseudo R2 as a goodness of t measure, shows a value of 0.773.
Seven of 12 variables show signicance at a 90% or higher condence level. Four variables are found signicant at 1%, while one
variable is signicant at 5% and two at 10%. Therefore, the high
Pseudo R2 measured the goodness of t, combined with the seven
signicant variables at 10%, 5% and 1%, which indicate that the
model has sufcient explanatory power.
The logistic regression results for risk aversion in Table 2 show
that age was not found to be signicant (p-value 0.760). Education level has a positive co-efcient (0.221) and is highly signicant (p-value 0.001). This shows that educated farmers are
more risk averse than the uneducated or less educated farmers.
Educated framers can perceive disasters more wisely, and it was
found that they were more risk averse in nature. Likewise, the
experience level of farmers was found to be statistically signicant
(p-value 0.006) and has a positive coefcient (0.093). The ndings for experienced farmers imply that experienced farmers are
more risk averse than inexperienced or less experienced farmers.
Similarly, the ndings of family size show a positive coefcient
(0.2720) and is signicant (p-value 0.075), showing that as
2
According to the State Bank of Pakistan, PKR 1 0.00982 US$, dated 30 June
2015. URL: http://www.sbp.org.pk/.

112

S.E. Saqib et al. / International Journal of Disaster Risk Reduction 18 (2016) 107114

Table 2
Factors affecting risk attitude (logistic model).
Variables

Coefcients

Age
 0.012
Education
0.221
Experience
0.093
Health status
0.333
Family size
0.272
Monthly off-farm income  1.2  10
Landholding size
 0.358
Owned land proportion
1.375
Field labor
 2.662
Distance from river
1.158
Risk perception
Risk perception of oods 4.997
Famer category
Dummy 1: Lower sub1.795
sistence farmers
Log-Likelihood Value  26.370
LR Test Chi2(12) 179.560
Prob-Chi 2 0.000
Pseudo R2 0.773
Total number of observations 168

4

Table 3
Marginal effects (logistic model).

Standard
errors

Signicance
p-value

Variables

Coefcients

Standard
errors

Signicance
p-value

0.040
0.081
0.042
1.249
0.152
0.000
0.139
1.132
1.877
1.266

0.760
0.006***
0.028**
0.789
0.075*
0.000***
0.010***
0.224
0.156
0.360

 0.003
0.055
0.022
0.082
0.067
 2.8  10
 0.088
0.340
 0.658
0.281

0.010
0.020
0.010
0.307
0.071
0.000
0.035
0.277
0.465
0.291

0.760
0.006***
0.028**
0.788
0.075*
0.000***
0.010***
0.220
0.157
0.336

1.228

0.000***

0.836

0.087

0.000***

0.830

0.031**

Age
Education
Experience
Health status
Family size
Monthly off-farm income
Landholding size
Owned land proportion
Field labor
Distance from river
Risk perception
Risk perception of oods
Famer category
Dummy 1: Lower subsistence farmers

0.421

0.171

0.014**

Source: Field survey, 2015.

*

P r0.10.
Pr 0.05.
***
Pr 0.01
**

5

Source: Field survey, 2015.

*

Pr 0.10.
P r0.05.
***
Pr 0.01
**

If the group of farmers is changed from upper subsistence to lower

subsistence, the probability of risk aversion changes by 42.1%.

5. Discussion
family size increases, farmers are more likely to be risk averse in
nature.
Unlike the previous results of education, experience and family
size which were positively associated with risk aversion, the offfarm income has a negative coefcient ( 0.00012) and is highly
signicant (p-value 0.000). Likewise, landholding size was found
to be signicant (p-value 0.010) with a negative coefcient
(  0.358). This implies that as landholding size increases, farmers
are less likely to be risk averse in nature.
For risk perception, we use risk perception of oods. The results
in Table 2 show a positive coefcient (4.997) and are signicant
(p-value 0.000). This means that as farmers' perception of oods
rises from 0 to 1, their probability to be risk averse also increases.
Likewise, the dummy variable that is included for lower subsistence farmers is signicant (p-value 0.031), with a positive
coefcient (1.795). The results for farmers' groups imply that lower
subsistence farmers are more risk averse in nature than other
farmers with landholdings of over ve acres.
4.2.1. Marginal effects
The marginal effects are shown below in Table 3. However, the
change in probability of being risk averse is 5.5%, with one
instance of change in education being highly signicant
(p-value 0.006). Likewise, the positive coefcient (0.022) of experience means that the probability of risk aversion would be 2.2%
higher with each unit increase in experience. Similar results are
obtained for family size; however, signicance was found at a 90%
condence level. Monthly off-farm income of households had a
negative impact, with a coefcient value of  0.000067, and
showed signicance (p-value 0.000) on the risk aversion of
the farmers. The same ndings are calculated for landholding
size, with negative coefcient of  0.088, and are signicant
(p-value 0.010).
Risk perception had a positive inuence on the risk averse
nature of farmers. It has a positive coefcient (0.836), which
means that if the risk perception of farmers increases from lower
to higher, it increases risk aversion to 83.6%, and is signicant
(p-value 0.000). Similar results are obtained for farmers' groups.

The ndings of this study reveal that more than half of the
farmers were risk averse in nature, and their perceptions about
oods were found to be high. In natural disasters, In terms of
economic loss, ooding is the most destructive natural disaster [2].
In the study area, farmers were the most affected in terms of damages to crops, livestock, irrigation systems, water contamination
and other agricultural operations. Further, the impacts of oods on
agricultural systems aggravated the problems in terms of losses in
farm yields and food security. The same results were obtained by
Deen [16] and Khan et al. [31]. Due to these huge losses and damages to the agriculture sector in the foods of 2010, 2011 and
2014, farmers had very high risk perception of oods and heavy
rains compared to other natural disasters. This high risk perception of farmers led to the high risk attitude averse nature of
farmers (Section 4.1). Our results for risk aversion are consistent
with the ndings of Iqbal et al. [29], Ullah et al. [55], Bond and
Wonder [11] and Kitonyoh [33]. They reported that the majority of
farmers in their studies were risk averse in nature. Among the
socio-economic factors, education was highly signicant in affecting the risk aversion of farmers. Educated farmers may have
better knowledge on sources of risk, and also the possible strategies they can adopt at the farm level to secure themselves from
such risks. Our ndings for the relationship of education with risk
aversion are in agreement with Lucas and Pabuayon [36]. They
found that most of the educated farmers in the Philippines were
risk averse in nature compared to illiterate farmers. Likewise, Kitonyoh [33,55] also reported the same results for education and
risk attitudes of farmers. However, Dadzie and Acquah [15] and
Binici et al. [8] have reported an inverse relationship. They stated
that as their education increased, farmers were less risk averse in
nature. In regard to experience, our ndings reveal that highly
experienced farmers are more likely to be risk averse than less
experienced farmers. Experienced farmers have indigenous
knowledge of the environment, weather, natural hazards and the
possible pests and diseases, which makes them more careful and
less likely to take risks. Our ndings support the ndings of Lucas
and Pabuayon [36]. Their results revealed that highly experienced

S.E. Saqib et al. / International Journal of Disaster Risk Reduction 18 (2016) 107114

farmers were more risk averse than less experienced farmers. The
results of landholding size indicate that larger landholders are
more likely to be risk seekers than smaller landholders. The
farmers with more land had greater diversication in crops and
varieties, as they have more land available for cultivation. In addition, farmers with larger landholdings can spread their elds so
as to mitigate the effects of oods. However, in the oods of 2010,
the entire study area was ooded. This resulted in the upper
subsistence farmers being more risk seeking than lower subsistence farmers. Lower subsistence farmers had limited landholdings. These farmers had a lack of opportunity in diversication
of crops and preparing elds at different locations. However, our
results are consistent with those of Sewando et al. [49], who stated
that large landholders were more risk seeking than small landholders. However, Iqbal et al. [29], Ullah et al. [55] and Dadzie and
Acquah [15] found no signicant relationship of landholding size
with risk aversion of farmers. Off-farm income is purposively used
as an independent variable in this study on the basis of a reconnaissance survey, where most of the farmers had secondary
income generating activities. The farmers were sending their adult
male children to Saudi Arabia and the United Arab Emirates, who
sent back remittances to support their families. The ndings for
off-farm income show that as off-farm income increases, farmers
were more likely to be risk-seekers in nature. Our results for the
negative coefcient of off-farm income are consistent with the
ndings of Dadzie and Acquah [15], Iqbal et al. [29] and Ullah et al.
[55]. They reported that poor farmers were more risk-averse than
wealthy farmers. The risk perceptions of farmers in the study area
are also found to affect the risk aversion of farmers. The farmers
with high risk perception were more risk averse in nature than
those with lower perception. Risk perception is very important
indicator in the disasters literature. It demonstrates individual and
community responses to natural disasters [10] and a positive
correlation is found between public response and adaptation/
management to natural hazards. This means that when risk perception of farmers is high, they will be more risk averse and will
adopt risk mitigating activities. For example, farmers had a high
risk perception of oods so they adopted agricultural credit [47,56]
and off-farm diversication [56] as agricultural ood-risk management tools. Likewise, farmers may use diversication in income, precautionary savings, diversication in crops and several
other farm risk management tools in post and pre disaster situations. Large farmers have more land and greater diversication of
income and crops. Therefore, the dummy for the farmer category
reveals that small subsistence farmers are more risk averse in
nature than large subsistence farmers. Hence, farmers' socio-economic factors and other disaster-related factors play key roles in
determining their risk attitude. After the 2005 earthquake and
major oods in 2010, Pakistan still has poor disaster management,
mitigation, preparedness and institutionalized coping strategies. It
is important that disaster risk reduction and preparedness should
be a national priority. Moreover, it is imperative that government
support programs such as crop insurance and agricultural nancing, which should be extended to disaster-prone areas. Through
these initiatives and programs, the interests of the farming community, the largest portion of the population, can be secure.

6. Conclusion
Based on the results and signicant ndings of this study, it is
clear that risk and uncertainty are the main causes of low yields
and crop production in the study area. The majority of farmers
were risk averse and had a high perception of oods. Farmers' risk
attitudes were signicantly inuenced by education, experience,
family size and income. Moreover, risk perception of oods and

113

farmers' category also played a role in their attitudes. This implies

that these factors are very important for consideration under a
policy framework. There is a positive correlation between risk
perception and preparedness for disasters. Therefore, for governmental agencies, it is recommended to design policies and programs to secure farmers from such natural disasters. Flood risk
perception and farmers' risk attitude could play an important role
in employing ood management strategies. In addition, these socio-economic factors are also important for farmers' risk management strategies. The ndings of the study can also be used in
future studies. There is a need to explore the role of information
sources, particularly formal sources such as print media, electronic
media and extension services, in farmers' risk attitudes and risk
perception. Moreover, this research could be extended to exploring the role of informal sources of information, such as face-toface information sharing, and input and output dealers in farmers'
risk attitudes. The study further suggests that as the farmers were
more risk averse in nature, other studies should explore the risk
management activities that farmers are practicing in the study
area. Research could also include investigating farmers' willingness to adopt and pay for crop loan insurance, which has been
introduced by the government but not yet extensively practiced by
farmers.

Acknowledgments
We would like to thank the two anonymous reviewers and the
editor of this journal for providing helpful comments on the earlier draft of the manuscript that have substantially improved its
quality. We would also like to show our gratitude to Miss. Tran Thi
Nhu Ngoc, Agriculture University Hanoi, Vietnam for sharing her
pearls of wisdom with us during the course of this research.

References
[1] Agriculture Census Organization, Agriculture Census: 2010 Report, Islamabad:
Statistics Division, Government of Pakistan, 2010. Available at: http://www.
pbs.gov.pk/sites/default/les/aco/publications/agricultural_census2010/
WRITE-UP%20AGRI.%20CENSUS%202010.pdf, (accessed 15.10.15).
[2] A.M.S. Ali, September 2004 ood event in southwestern Bangladesh: a study
of its nature, causes, and human perception and adjustments to a new hazard,
Nat. Hazards 40 (1) (2007) 89111.
[3] J.R. Anderson, J.L. Dillon, J.B. Hardaker, Agricultural Decision Analysis, Iowa
State University Press, Iowa, U.S.A., 1977.
[4] J.M. Antle, Econometric estimation of producers' risk attitudes, Am. J. Agric.
Econ. 69 (3) (1987) 509522.
[5] K. Arrow, Essays in The Theory of Risk Bearing, North-Holland, Amsterdam,
The Netherlands, 1970.
[6] M. Ashraf, J.K. Routray, Perception and understanding of drought and coping
strategies of farming households in north-west Balochistan, Int. J. Disaster Risk
Reduct. 5 (2013) 4960, http://dx.doi.org/10.1016/j.ijdrr.2013.05.002.
[7] T. Below, A. Artner, R. Siebert, S. Sieber, Micro-Level Practices to Adapt to
Climate Change for African Small-Scale Farmers: A Review of Selected Literature, International Food Policy Research Institute, Washington DC, 2010,
IFPRI Discussion Paper 00953.
[8] T. Binici, A. Koc, A. Bayaner, The Risk Attitudes of Farmers and the Socioeconomic Factors Affecting Them: A Case Study for Lower Seyhan Plain
Farmers in Adana Province, Turkey, 2001. Available at: http://www.tepge.gov.
tr/Dosyalar/Yayinlar/b6d9769e27f549e1a8884cb0333cc39b.pdf, (accessed
4.12.15).
[9] H.P. Binswanger, Attitudes toward risk: experimental measurement in rural
India, Am. J. Agric. Econ. 62 (3) (1980) 395407, http://dx.doi.org/10.2307/
1240194.
[10] S. Birkholz, M. Muro, P. Jeffrey, H. Smith, Rethinking the relationship between
ood risk perception and ood management, Sci. Total Environ. 478 (2014)
1220, http://dx.doi.org/10.1016/j.scitotenv.2014.01.061.
[11] G.E. Bond, B. Wonder, Risk attitudes amongst Australian farmers, Aust. J. Agric.
Econ. 24 (1) (1980) 1634, http://dx.doi.org/10.1111/j.1467-8489.1980.tb00367.x.
[12] J. Bond, H.O. Dickinson, F. Matthews, C. Jagger, C. Brayne, M. CFAS, Self-rated
health status as a predictor of death, functional and cognitive impairment: a
longitudinal cohort study, Eur. J. Ageing 3 (4) (2006) 193206.
[13] V. Choudhry, T. Baedeker, T. Johnson, Making the Risky Business of Agriculture

114

[14]

[15]

[16]

[17]

[18]
[19]
[20]

[21]

[22]
[23]
[24]
[25]
[26]

[27]
[28]

[29]

[30]
[31]
[32]

[33]

[34]
[35]

[36]

[37]
[38]

S.E. Saqib et al. / International Journal of Disaster Risk Reduction 18 (2016) 107114
Climate-smart, The World Bank, Washington, D.C., United States, 2015 http://
blogs.worldbank.org/voices/making-risky-business-agriculture-climatesmart (accessed 7.2.15).
A. Cohen, L. Einav, Estimating Risk Preferences from Deductible Choice, National Bureau of Economic Research, Massachusetts Avenue, Cambridge, MA
02138, 2005 http://www.nber.org/papers/w11461.pdf (accessed 5.1.16).
S.K.N. Dadzie, H. Acquah, Attitude toward risk and coping responses: the case
of food crop farmers at Agona Duakwa in Agona East District of Ghana, Int. J.
Agric. For. 2 (2) (2012) 2937, http://dx.doi.org/10.5923/j.ijaf.20120202.06.
S. Deen, Pakistan 2010 oods. Policy gaps in disaster preparedness and response, Int. J. Disaster Risk Reduct. 12 (2015) 341349, http://dx.doi.org/
10.1016/j.ijdrr.2015.03.007.
S. Dercon, L. Christiaensen, Consumption risk, technology adoption and poverty traps: evidence from Ethiopia, J. Dev. Econ. 96 (2) (2011) 159173, http:
//dx.doi.org/10.1016/j.jdeveco.2010.08.003.
J.L. Dillon, P.L. Scandizzo, Risk attitudes of subsistence farmers in Northeast
Brazil: a sampling approach, Am. J. Agric. Econ. 60 (3) (1978) 425435.
S.A. Drollette, Managing Production Risk in Agriculture, Department of Applied Economics Utah State University, United States, 2009.
A.K.A. Ghadim, D.J. Pannell, M.P. Burton, Risk, uncertainty, and learning in
adoption of a crop innovation, Agric. Econ. 33 (1) (2005) 19, http://dx.doi.
org/10.1111/j.1574-0862.2005.00433.x.
T. Gill, No Fear: Growing Up in a Risk Averse Society, Calouste Gulbenkian
Foundation United Kingdom Branch, United Kingdom, 2007 (accessed 9.3.16)
http://www.gulbenkian.org.uk/pdfles/item-1266223-No-fear-1912-07.
pdf.
W.H. Greene, Econometric Analysis, 7th ed., Prentice-Hall, Englewood Cliffs,
New Jersey, 2008.
D.N. Gujarati, D.C. Porter, S. Gunasekar, Basic Econometrics, 5th ed., Tata
Mcgraw-Hill Education, UP, India, 2013.
M.A.B. Halima, E. Rococo, Wage differences according to health status in
France, Soc. Sci. Med. 120 (2014) 260268.
K. Hamal, J.R. Anderson, A note on decreasing absolute risk aversion among
farmers in Nepal, Aust. J. Agric. Econ. 26 (3) (1982) 220225.
H.-y. HAN, L.-g. ZHAO, Farmers' character and behavior of fertilizer application-evidence from a survey of Xinxiang County, Henan Province, China, Agric.
Sci. China 8 (10) (2009) 12381245, http://dx.doi.org/10.1016/S1671-2927(08)
60334-X.
J.B. Hardaker, R.B. Huirne, J.R. Anderson, G. Lien, Coping with risk in agriculture, CABI Publishing, Wallingford, Oxon, UK, 2004.
G.W. Harrison, M.I. Lau, E.E. Rutstrm, Estimating risk attitudes in denmark: a
eld experiment, Scand. J. Econ. 109 (2) (2007) 341368, http://dx.doi.org/
10.1111/j.1467-9442.2007.00496.x.
M.A. Iqbal, Q. Ping, M. Abid, S.M.M. Kazmi, M. Rizwan, Assessing risk perceptions and attitude among cotton farmers: a case of Punjab province, Pakistan, Int. J. Disaster Risk Reduct. 16 (2016) 6874.
H. Ji-kun, Climate change and agriculture: impact and adaptation, J. Integr.
Agric. 13 (4) (2014) 657659.
A.N. Khan, S.N. Khan, A. Ali, Analysis of damages caused by ood-2010 in
district Peshawar, J. Sc. Tech., Univ. Peshawar 36 (2) (2010) 1116.
M. Kisaka-Lwayo, A. Obi, Risk perceptions and management strategies by
smallholder farmers in KwaZulu-Natal Province, South Africa, Int. J. Agric.
Manag. 1 (3) (2012) 2839.
C.K. Kitonyoh, A Farm Level Analysis of Risk Attitude, Sources and Risk Measurement Strategies among Farmers in Trans Nzoia County, Kenya (MSc thesis), Moi University, Kenya, 2015.
H. Levy, Stochastic Dominance: Investment Decision Making under Uncertainty, Springer, New York, U.S.A., 2006.
T.F. Liao, Interpreting Probability Models: Logit, Probit, and Other Generalized
Linear Models, Sage University Paper Series on Quantitative Applications in
the Social Sciences, Sage, Thousand Oaks, CA 1994, pp. 07101.
M.P. Lucas, I.M. Pabuayon, Risk perceptions, attitudes, and inuential factors of
rainfed lowland rice farmers in Ilocos Norte, Philippines, Asian J. Agric. Dev. 8
(2) (2011) 6177.
E. Moscardi, A. de Janvry, Attitudes toward risk among peasants: an econometric approach, Am. J. Agric. Econ. 59 (4) (1977) 710716.
W.N. Musser, G.F. Patrick, How much does risk really matter to farmers? A
Comprehensive Assessment of the Role of Risk in US Agriculture, Springer,
U.S.A., 2002, pp. 537556.

[39] National Disaster Management Authority, Flood Rapid Response Plan, NDMA,
Islamabad, 2011 http://ndma.gov.pk/Documents/ood_2011/Pakistan_RRP_
Floods2011.PDF (accessed 6.10.15).
[40] National Disaster Management Authority, Recovery Needs Assessment and
Action Framework 20142016, NDMA, Islamabad, 2014 (accessed 3.10.15)
http://www.humanitarianresponse.info/operations/pakistan/document/re
covery-needs-assessment-and-action-framework-201416.
[41] V. Ogurtsov, M. Van Asseldonk, R. Huirne, Assessing and modelling catastrophic risk perceptions and attitudes in agriculture: a review, NJAS-Wagening. J. Life Sci. 56 (1) (2008) 3958, http://dx.doi.org/10.1016/S1573-5214
(08)80016-4.
[42] L. Olarinde, V. Manyong, J. Akintola, Attitudes towards risk among maize
farmers in the dry savanna zone of Nigeria: some prospective policies for
improving food production, Afr. J. Agric. Res. 2 (8) (2007) 399408.
[43] Pakistan Economic Survey, 2011  12, Highlights of the Pakistan Economic
Survey. Islamabad: Ministry of Finance. Available at: http://www.nance.gov.
pk/survey/chapter_12/highlights.pdf, (accessed 6.4.16).
[44] J.W. Pratt, Risk aversion in the small and in the large, Econometrica 32 (1)
(1964) 122136.
[45] Provincial Disaster Management Authority, Contengency Plan Khyber Pakhtunkhwa, 2013. Available at: http://pdma.gov.pk/downloads/Monsoon_Con
tingency_Plan_KP_2013.pdf, (accessed 14.10.14).
[46] S.E. Saqib, Access, Adequacy and Utilization of Agricultural Credit to Farmers
in Pakistan: The Case of Mardan District, Khyber Pakhtunkhwa, Asian Institute
of Technology, Thailand, 2015.
[47] S.E. Saqib, M.M. Ahmad, S. Panezai, U. Ali, Factors inuencing farmers' adoption of agricultural credit as a risk management strategy: the case of Pakistan,
Int. J. Disaster Risk Reduct. 17 (2016) 6776, http://dx.doi.org/10.1016/j.
ijdrr.2016.03.008.
[48] E.M. Senkondo, Risk Attitude and Risk Perception in Agroforestry Decisions:
The Case of Babati, Tanzania, Vol. 17, Wageningen University, Netherlands,
2000.
[49] P.T. Sewando, N. Mdoe, K. Mutabazi, Farmers' preferential choice decisions to
alternative cassava value chain strands in Morogoro rural district, Tanzania,
Agric. J. 6 (6) (2011) 313321.
[50] A. Smidts, L. Wageningen, Decision Making Under Risk: A Study of Models and
Measurement Procedures with Special Reference to the Farmer's Marketing
Behaviour, Wageningen University, Netherlands, 1990.
[51] T. Tanaka, C.F. Camerer, Q. Nguyen, Risk and time preferences: linking experimental and household survey data from Vietnam, Am. Econ. Rev. 100 (1)
(2010) 557571, http://dx.doi.org/10.1257/aer.100.1.557.
[52] J. Torkamani, Using a whole-farm modelling approach to assess prospective
technologies under uncertainty, Agric. Syst. 85 (2) (2005) 138154, http://dx.
doi.org/10.1016/j.agsy.2004.07.016.
[53] R. Ullah, Production Risk Management and Its Impacts at the Farm Level: The
Case of Paksitan (Ph.D. thesis), Asian Institute Of Technology, Thailand, 2014.
[54] R. Ullah, D. Jourdain, G.P. Shivakoti, S. Dhakal, Managing catastrophic risks in
agriculture: simultaneous adoption of diversication and precautionary savings, Int. J. Disaster Risk Reduct. 12 (2015) 268277.
[55] R. Ullah, G.P. Shivakoti, G. Ali, Factors effecting farmers' risk attitude and risk
perceptions: the case of Khyber Pakhtunkhwa, Pakistan, Int. J. Disaster Risk
Reduct. 13 (2015) 151157, http://dx.doi.org/10.1016/j.ijdrr.2015.05.005.
[56] R. Ullah, G.P. Shivakoti, A. Kamran, F. Zulqar, Farmers versus nature: managing disaster risks at farm level, Nat. Hazards 82 (3) (2016) 19311945, http:
//dx.doi.org/10.1007/s11069-016-2278-0.
[57] R. Ullah, G.P. Shivakoti, M. Rehman, M.A. Kamran, Catastrophic risks management at farm: the use of diversication, precautionary savings and agricultural credit, Pak. J. Agric. Sci. 52 (4) (2015) 11391147.
[58] J. Von Neumann, O. Morgenstern, Theory of Games and Economic behavior,
Princeton University Press, Princeton, NJ, 1944.
[59] World Food Programme, Pakistan Flood Impact Assessment, WFP (2010),
Rome, Italy, https://www.wfp.org/content/pakistan-ood-impact-assess
ment-september-2010 (accessed 20.10.15).
[60] T. Yamane, Statistics: An Introductory Analysis, 2nd ed., Harper and Row, New
York, 1967.
[61] Q.-Y. Yu, W.-B. Wu, Z.-H. Liu, P.H. Verburg, X. Tian, Y. Peng, H.-J. Tang, Interpretation of climate change and agricultural adaptations by local household
farmers: a case study at Bin County, Northeast China, J. Integr. Agric. 13 (7)
(2014) 15991608, http://dx.doi.org/10.1016/S2095-3119(14)60805-4.