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

International Journal of Disaster Risk Reduction 18 (2016) 107–114

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 flood-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 flood-prone areas. This study examines the effects of socio-economic factors on risk attitudes of farmers in a flood-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 findings 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 floods significantly 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 floods that had devastating impacts on the entire economy, particularly in the agriculture sector. The monsoon floods of 2010, 2011, and 2014 caused huge damage to agricultural crops, fisheries, forestry, livestock, and primary infrastructure, such as water channels, tube wells, houses, people, seed stocks, animal shelters, fertilizers and agricultural equipment/machinery. The floods 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 flood [59]. In 2011, another massive flood 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: [email protected] (S.E. Saqib), [email protected] (M.M. Ahmad), [email protected] (S. Panezai), [email protected] (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, fisheries and forestry [39]. The flood 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 fisheries and livestock. The total damage caused by the 2011 floods 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 floods of September 2014, 367 persons died, and over 2.5 million people were affected by heavy rains and floods. 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 floods 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) 107–114

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, floods, pests and diseases [29,54,55], droughts [54], and market price fluctuations [29]. According to Musser and Patrick [38], there are five important sources of risk factors in agriculture: production, financial, 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, financial risks: i.e. a farmer's ability to pay their bills to continue farming and avoid bankruptcy. Third, marketing risks, which involve fluctuations 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 diversification as risk management tools at the farm level in Pakistan. Likewise, farmers adopt diversification beyond the farm, such as diversification in crops, scheduling of farming practices, migration, and a variety of other diversification methods such as irrigation and water conservation techniques to cope with climatic risks [7]. Similarly, to cope with droughts, farmers practiced income diversification, 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 ramifications 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 influence 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 modified 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 first 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) 107–114

The expected utility theory is defined 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 floods, 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 floods 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 floods in 2010. These farmers mostly lived on the bank of the river and faced floods consistently, every year during the monsoon rains. However, the severity of floods 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 identified 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% confidence level with a 77% margin of error:

n=

N (1+Ne2)

(4)

n¼Sample size. N¼ Total number of farming households in an area. e¼Precision 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 120,000–200,000) 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) = α1 + α 2w+α3w2 + α 4w 3

(5)

where αs are the parameters and w represents the wealth of the farmers and their attitudes toward risk, which are dependent on several factors. However, a significant 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 first and second derivatives of the function are:

U′ = α 2+2α3w +3α 4w 2

(6)

U′′=2α3+6α 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 first derivative is with respect to income. According to Arrow [5] and Pratt [44], the risk aversion coefficient indicates the nature of risk attitude. In the language of

110

S.E. Saqib et al. / International Journal of Disaster Risk Reduction 18 (2016) 107–114

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)¼ 0 and U (200,000) ¼1

1

U (120,000) ¼0.5u (0)þ 0.5u (200,000) ¼0.5 U (60,000) ¼0.5u (0) þ0.5u (120,000) ¼0.25 U (30,000) ¼0.5u (0) þ0.5u (60,000) ¼ 0.125 U (20,000) ¼0.5u (0) þ0.5u (30,000) ¼ 0.0625 U (140,000) ¼0.5u (200,000) þ (0.5u (140,000) ¼ 0.75 U (170,000) ¼0.5u (200,000) þ (0.5u (170,000) ¼0.875 U (180,000) ¼ 0.5u (200,000) þ (0.5u (180,000) ¼0.937

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 floods Risk perception is measured by a five 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) 107–114

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 coefficient 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 ¼1; otherwise, we assigned a value ¼ 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 βi + Ɛi ⎣ (1−Pi ) ⎦

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=Xiβi + Ɛ 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). βi=Vector K × 1 of estimated coefficient. Ɛi=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 floods 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 five 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

1¼ Aversion, 0¼ 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 ¼good health, 0¼ 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 field from river river (1 ¼within 500 m from the bank, 0¼ otherwise) Risk perception Risk perception of floods of floods (1 ¼high risk, 0¼ low risk) Farmers' groups Farmers' groups (1¼ lower subsistence farmer, 0¼ 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 fit measure, shows a value of 0.773. Seven of 12 variables show significance at a 90% or higher confidence level. Four variables are found significant at 1%, while one variable is significant at 5% and two at 10%. Therefore, the high Pseudo R2 measured the goodness of fit, combined with the seven significant variables at 10%, 5% and 1%, which indicate that the model has sufficient explanatory power. The logistic regression results for risk aversion in Table 2 show that age was not found to be significant (p-value ¼ 0.760). Education level has a positive co-efficient (0.221) and is highly significant (p-value ¼0.001). 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 significant (p-value ¼0.006) and has a positive coefficient (0.093). The findings for experienced farmers imply that experienced farmers are more risk averse than inexperienced or less experienced farmers. Similarly, the findings of family size show a positive coefficient (0.2720) and is significant (p-value ¼0.075), showing that as 2 According to the State Bank of Pakistan, PKR 1¼ 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) 107–114

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

Coefficients

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 floods 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

Significance p-value

Variables

Coefficients

Standard errors

Significance 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 floods 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 coefficient ( 0.00012) and is highly significant (p-value ¼0.000). Likewise, landholding size was found to be significant (p-value ¼ 0.010) with a negative coefficient (  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 floods. The results in Table 2 show a positive coefficient (4.997) and are significant (p-value ¼0.000). This means that as farmers' perception of floods 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 significant (p-value ¼ 0.031), with a positive coefficient (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 five 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 significant (p-value ¼0.006). Likewise, the positive coefficient (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, significance was found at a 90% confidence level. Monthly off-farm income of households had a negative impact, with a coefficient value of  0.000067, and showed significance (p-value ¼0.000) on the risk aversion of the farmers. The same findings are calculated for landholding size, with negative coefficient of  0.088, and are significant (p-value ¼0.010). Risk perception had a positive influence on the risk averse nature of farmers. It has a positive coefficient (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 significant (p-value ¼0.000). Similar results are obtained for farmers' groups.

The findings of this study reveal that more than half of the farmers were risk averse in nature, and their perceptions about floods were found to be high. In natural disasters, In terms of economic loss, flooding 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 floods 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 floods 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 findings 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 significant 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 findings 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 findings 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 findings support the findings of Lucas and Pabuayon [36]. Their results revealed that highly experienced

S.E. Saqib et al. / International Journal of Disaster Risk Reduction 18 (2016) 107–114

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 diversification in crops and varieties, as they have more land available for cultivation. In addition, farmers with larger landholdings can spread their fields so as to mitigate the effects of floods. However, in the floods of 2010, the entire study area was flooded. 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 diversification of crops and preparing fields 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 significant 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 findings 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 coefficient of off-farm income are consistent with the findings 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 floods so they adopted agricultural credit [47,56] and off-farm diversification [56] as agricultural flood-risk management tools. Likewise, farmers may use diversification in income, precautionary savings, diversification in crops and several other farm risk management tools in post and pre disaster situations. Large farmers have more land and greater diversification 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 floods 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 financing, 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 significant findings 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 floods. Farmers' risk attitudes were significantly influenced by education, experience, family size and income. Moreover, risk perception of floods 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 flood management strategies. In addition, these socio-economic factors are also important for farmers' risk management strategies. The findings 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/files/aco/publications/agricultural_census2010/ WRITE-UP%20AGRI.%20CENSUS%202010.pdf〉, (accessed 15.10.15). [2] A.M.S. Ali, September 2004 flood event in southwestern Bangladesh: a study of its nature, causes, and human perception and adjustments to a new hazard, Nat. Hazards 40 (1) (2007) 89–111. [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) 509–522. [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) 49–60, 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) 395–407, http://dx.doi.org/10.2307/ 1240194. [10] S. Birkholz, M. Muro, P. Jeffrey, H. Smith, Rethinking the relationship between flood risk perception and flood management, Sci. Total Environ. 478 (2014) 12–20, 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) 16–34, 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) 193–206. [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) 107–114 ‘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) 29–37, http://dx.doi.org/10.5923/j.ijaf.20120202.06. S. Deen, Pakistan 2010 floods. Policy gaps in disaster preparedness and response, Int. J. Disaster Risk Reduct. 12 (2015) 341–349, 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) 159–173, 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) 425–435. 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) 1–9, 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/pdffiles/–item-1266–223-No-fear-19–12-07. 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) 260–268. K. Hamal, J.R. Anderson, A note on decreasing absolute risk aversion among farmers in Nepal, Aust. J. Agric. Econ. 26 (3) (1982) 220–225. 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) 1238–1245, 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. Rutström, Estimating risk attitudes in denmark: a field experiment, Scand. J. Econ. 109 (2) (2007) 341–368, 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) 68–74. H. Ji-kun, Climate change and agriculture: impact and adaptation, J. Integr. Agric. 13 (4) (2014) 657–659. A.N. Khan, S.N. Khan, A. Ali, Analysis of damages caused by flood-2010 in district Peshawar, J. Sc. Tech., Univ. Peshawar 36 (2) (2010) 11–16. 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) 28–39. 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. 07–101. M.P. Lucas, I.M. Pabuayon, Risk perceptions, attitudes, and influential factors of rainfed lowland rice farmers in Ilocos Norte, Philippines, Asian J. Agric. Dev. 8 (2) (2011) 61–77. E. Moscardi, A. de Janvry, Attitudes toward risk among peasants: an econometric approach, Am. J. Agric. Econ. 59 (4) (1977) 710–716. 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. 537–556.

[39] National Disaster Management Authority, Flood Rapid Response Plan, NDMA, Islamabad, 2011 〈http://ndma.gov.pk/Documents/flood_2011/Pakistan_RRP_ Floods2011.PDF〉 (accessed 6.10.15). [40] National Disaster Management Authority, Recovery Needs Assessment and Action Framework 2014–2016, NDMA, Islamabad, 2014 (accessed 3.10.15) 〈http://www.humanitarianresponse.info/operations/pakistan/document/re covery-needs-assessment-and-action-framework-2014–16〉. [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) 39–58, 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) 399–408. [43] Pakistan Economic Survey, 2011  12, Highlights of the Pakistan Economic Survey. Islamabad: Ministry of Finance. Available at: 〈http://www.finance.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) 122–136. [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 influencing farmers' adoption of agricultural credit as a risk management strategy: the case of Pakistan, Int. J. Disaster Risk Reduct. 17 (2016) 67–76, 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) 313–321. [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) 557–571, 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) 138–154, 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 diversification and precautionary savings, Int. J. Disaster Risk Reduct. 12 (2015) 268–277. [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) 151–157, http://dx.doi.org/10.1016/j.ijdrr.2015.05.005. [56] R. Ullah, G.P. Shivakoti, A. Kamran, F. Zulfiqar, Farmers versus nature: managing disaster risks at farm level, Nat. Hazards 82 (3) (2016) 1931–1945, 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 diversification, precautionary savings and agricultural credit, Pak. J. Agric. Sci. 52 (4) (2015) 1139–1147. [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-flood-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) 1599–1608, http://dx.doi.org/10.1016/S2095-3119(14)60805-4.