Cost of Risk Exposure, Farm Disinvestment and Adaptation to Climate Uncertainties:

The Case of Arable Farms in the EU

Habtamu Yesigat Ayenew*1, Johannes Sauer1, Getachew Abate-Kassa1

1Chair of Production and Resource Economics, Technical University Munich, Germany

Corresponding author: habtamu.ayenew@tum.de

Selected Paper prepared for presentation at the 2016 Agricultural & Applied Economics

Association Annual Meeting, Boston, Massachusetts, July 31-August 2

Copyright 2016 by [Habtamu Yesigat Ayenew, Johannes Sauer, Getachew Abate-Kassa]. All

rights reserved. Readers may make verbatim copies of this document for non-commercial

purposes by any means, provided that this copyright notice appears on all such copies.

Cost of Risk Exposure, Farm Disinvestment and Adaptation to Climate Uncertainties:

The Case of Arable Farms in the EU

Habtamu Yesigat Ayenew*1, Johannes Sauer1, Getachew Abate-Kassa1

1Chair of Production and Resource Economics, Technical University Munich, Germany

Abstract

This paper investigates the implication of the cost of risk exposure on farm diversification, farm

insurance premium payment and investment behavior in agriculture using an extensive Farm

Accountancy Data Network (FADN) panel data of arable farms and weather data from 1989

to 2009 from France and Germany. For this purpose, we develop a two stage empirical model.

First, we estimate the full profit moments distribution using the major inputs of production with

quadratic function. Following this, we estimate the major responses of farmers for exposure to

risk (farm diversification, investment decisions and purchase of insurance) via three-stage least

squares (3SLS) estimation procedure. Our empirical analysis confirms that risk exposure

measured with variance and skewness of farm profit can significantly influence the level of farm

diversification, insurance premium payment and farm investment and disinvestment in arable

farms in both countries. As these strategies seem not to be completely substitutable, this

evidence can be used to support the discussion of improving the availability of market based

instruments to strengthen the adaptive capacity of farms in the developing world. We do find

an evidence that farms can see disinvestment as a response to extreme shock and risk exposure

at least in a short run. Improving the adaptive capacity of farms might not only secure them

pervasive impacts of risk exposure, but also can influence their investment and disinvestment

behavior.

Key words: climate change, cost of risk, disinvestment, EU

1. Introduction

Agricultural risk management has captured a particular attention in research and policy

discussions and dialogues in agriculture and the food sector (IPCC, 2007, Goodwin, 2008,

OECD, 2009, IPCC, 2014). Climate uncertainty is a major concern in this regard, and has a

direct implication on the returns from agriculture and it often has disastrous consequences

(Dercon, 2004, Hardaker et al., 2004, Orea and Wall, 2012, IPCC, 2014). The type and extent

of the risky event, the cost and welfare impacts of risk and adequacy of the capacity of countries

to respond for risk can essentially shape the performance of the agricultural sector (Hardaker et

al., 2004, OECD, 2009, Chavas and Shi, 2015).

Farm production plan in the presence of climatic risk is typically composed of alternative

outcomes with different realization probabilities (Day, 1965, Atwood et al., 2003, Tack et al.,

2012). As these potential outcomes and associated probabilities determine the return from

agriculture, cost of climate risk appears as a key area of research and development concern

(IPCC, 2007, OECD, 2009, Orea and Wall, 2012, Kim et al., 2014). For instance in the EU,

despite a range of price protection and weather insurance packages, climate uncertainties

including extreme weather explain a significant share of the cost of risk in agriculture (Bielza

et al., 2007, OECD, 2009, Ciscar et al., 2011).

The level of risk that farmers experience and the sensitivity of the farmer towards risk shape

farmers’ risk behavior and adaptation responses. As a response to the potential impacts of

climate uncertainties, farmers adopt a wide range of risk mitigation strategies (Hardaker et al.,

2004, Diaz-Caneja et al., 2008, Baumgärtner and Quaas, 2010). These incudes, farm or non-

farm diversification (Di Falco and Chavas, 2009), conservation tillage and agro-environmental

schemes (Ding et al., 2009, Baumgärtner and Quaas, 2010), and the purchase of insurance

policies (Enjolras and Sentis, 2011, Enjolras and Kast, 2012, Finger and Lehmann, 2012).

Farmers experiencing production uncertainty with climate change and other risk sources might

consider disinvestment or gradual farm exit as an option. Ihli et al. (2014) for instance highlights

the role of irreversibility of investment, famers’ learning through repeated investment and its

implication on waiting on investment time and disinvestment. Of particular relevance in this

paper is the cost of risk exposure and its implication on their adaptation and mitigation

instrument choices and its impact on farm disinvestment decision. This paper contributes to

policy and empirical literature in a number of ways. First, this article is the first in its kind, by

analyzing multiple risk mitigation and adaptation strategies and disinvestment as

interdependent decisions in the farm. Second, we use a unique and extensive dataset to explain

dynamic farm and non-farm adjustments including farm disinvestment. This paper is organized

as follows. Next to the introduction, we illustrate the conceptual framework of the paper.

Section 3 presents the data and empirical approach used in this paper. In section 4, we present

and discuss the results. Section 5 concludes.

2. Conceptual framework

Agricultural yield and income often are uncertain and this uncertainty largely arise from climate

variabilities (IPCC, 2007, OECD, 2009, IPCC, 2014). A number of previous empirical studies

and projections highlight that climate variabilities can significantly determine agricultural

production (Tubiello et al., 2000, Ciais et al., 2005, Weitzman, 2009). Agriculture is in a

continuous adjustment in order to cope with the changing environment and reduce the cost of

risk exposure (Hardaker et al., 2004, Di Falco and Chavas, 2006, Baumgärtner and Quaas, 2010,

Zuo et al., 2014). Farmers might adopt risk mitigation and adaptation strategies including farm

level diversification and biodiversity (Di Falco and Chavas, 2006, Bezabih and Sarr, 2012),

irrigation and temporary water trading (Koundouri et al., 2006, Groom et al., 2008, Zuo et al.,

2014), conservation agriculture and agro-environmental schemes (Baumgärtner and Quaas,

2010, Kassie et al., 2014), and purchase insurance policies (Enjolras and Sentis, 2011, Enjolras

and Kast, 2012, Finger and Lehmann, 2012). Alem et al. (2010) underscore the correlation of

past rainfall variability and fertilizer purchase decisions in Ethiopia. Jalan and Ravallion (2001)

using the portfolio approach indicate non-productive and precautionary liquid wealth holding

of farmers to mitigate idiosyncratic shock in China. All these papers highlight that investments

(like diversification, insurance, irrigation etc.) and savings (precautionary liquid wealth) are

crucial farm and non-farm strategies to mitigate pervasive effects of risk. In this paper, we

would like to focus on analyzing farm responses to risk exposure using an extensive panel

dataset from France and Germany.

Following empirical representation of farmers’ investment behavior by Jalan and Ravallion

(2001), we consider a farmer with a decision on how to allocate his initial wealth (𝑊) between

annual input expenditure (𝐶) and capital investment (𝐼). The utility from farm investment and

production in a multi-period model can be represented as:

𝑈[𝑊 − 𝐶 − 𝐼] + 𝐸𝜀𝑈[𝑓(𝐶, 𝐼, 휀)] (1)

where 𝑈[𝑊 − 𝐶 − 𝐼] represents the utility of investment, and 𝑓(𝐶, 𝐼, 휀) is the farm production

activity as a function of input expenditures, capital investment and depends on the realization

of a random variable (휀). Differentiating equation (1) gives us:

𝑈′[𝑊 − 𝐶 − 𝐼] = 𝐸𝜀𝑈′[𝑓(𝐶, 𝐼, 휀)] (2)

As we can see from equation (2), the wealth of the farm and the distribution of the random

variable (휀) determine the choice of (𝐶) and (𝐼) in order to maximize the expected utility of

profit. Farmers make a decision in investment and annual expenditure depending on their utility

of the expected profit from farm activities and the utility of the profit distribution. The

randomness of the distribution that often arise from climate variabilities, disease and pest

outbreak, etc. is an important element for the integration of risk component in the estimation.

There is a continuous attempt to extend a simple farm profit maximization model with the

inclusion of risk components (Antle, 1983, Antle, 1987, Chavas and Di Falco, 2012). For a

simple representation of a farm decision model that relates the expected utility model of profit

and risk components, we start with a random farm, with resource constraints and working in a

risky environment.

𝐸[𝑈(𝜋)] = 𝑈[𝐸(𝜋) − 𝑅] (3)

U is farmers’ risk preferences based on the von Neumann - Morgenstern utility function with

farm profit (𝜋). Risk premium (R) corresponds to the amount of money that a risk averse farmer

is willing to pay to avoid risk (Pratt, 1964, Arrow, 1965). The elements of the profit and risk

function consist of random variables including yield with climate uncertainties and technology,

and return from randomness in the price and cost of inputs (Chavas, 2004, Hardaker et al.,

2004). In empirical studies, this randomness in the expected profit is the basis to calculate the

farm risk premium (Antle, 1983, Koundouri et al., 2006, Groom et al., 2008).

Following Antle (1983), the Arrow-Pratt risk premium can be approximated as:

𝑅 ≈ ∑ − [1

𝑖!]𝑟

𝑖=2 (𝑈′′

𝑈′ ) 𝑀𝑖 (4)

In this approximation, risk premium is represented as a linear function of the first r moments of

the profit distribution, where 𝑀𝑖 = (𝜋 − 𝐸(𝜋))𝑖 is the ith central moment of profit (𝜋) and i =

2, …, r (Antle, 1983, Chavas, 2004, Chavas and Di Falco, 2012). Building on the existing

literature on risk mitigation strategies, we develop a conceptual and empirical model that

comprises of alternative farm and non-farm responses to exposure to risk. Farmers make

decisions on alternative risk mitigation strategies in order to maximize the expected profit given

the resource constraints and exogenous shock. A farmer with exogenous shock may diversify

his (her) production activities (𝐶ℎ𝑡), buy an insurance policy (𝐼ℎ𝑡), or otherwise reduce the

capital investment. Nonetheless, these adaptation and mitigation responses of farmers to

dynamic environment are often cited as ‘slow’ (Carey and Zilberman, 2002, Richards and

Green, 2003). Part of this inertia is explained by the nature of agriculture itself and the required

time of adjustment. This is especially relevant for adoption of risk mitigation mechanisms

where the farmer might able to respond through learning from past shock experiences (Alem et

al., 2010, Zuo et al., 2014). Hence, we include the lagged estimates of risk exposure (variance

and skewness) in the estimation. This also solves the endogeneity problem that we have to deal

with at this stage had we otherwise decide to use the risk exposure estimates.

This functional relationship can be represented as:

𝐶ℎ𝑡 = 𝑓1(𝑆ℎ𝑡 , 𝐿ℎ𝑡 , 𝐸ℎ𝑡 , 𝑅ℎ(𝑡−1), ) + 𝑒1𝑖𝑡 (5)

𝐼ℎ𝑡 = 𝑓2(𝑆ℎ𝑡 , 𝐿ℎ𝑡 , 𝐸ℎ𝑡, 𝑅ℎ(𝑡−1), ) + 𝑒2𝑖𝑡 (6)

𝐷ℎ𝑡 = 𝑓3(𝑆ℎ𝑡 , 𝐿ℎ𝑡 , 𝐸ℎ𝑡, 𝑅ℎ(𝑡−1), ) + 𝑒3𝑖𝑡 (7)

where (𝑆ℎ𝑡) represents the demographic characteristics of the farm manager, (𝐿ℎ𝑡) captures land

characteristics, (𝐸ℎ𝑡) represents environmental variables and (𝑅ℎ(𝑡−1)) represents the lagged

risk exposure measured with second and third profit moments.

3. Data and empirical approach

This paper investigates the cost of risk exposure and its implication on disinvestment in

agriculture using an extensive Farm Accountancy Data Network (FADN) panel data and

weather data from 1989 to 2009 from France and Germany. In total, we used 45526 and 30503

farm observations in France and Germany, respectively. All the values related to the prices of

farm products are deflated towards the base year 1989 by the national price index of the two

countries. In the same analogy, input costs and capital items are deflated with the input price

index of the two countries.

Table 1: Summary statistics

Variables France Germany

Mean Std. dev Mean Std. dev

Land (in hectares) 112.75 77.04 197.74 427.22

Seed cost (in Euros) 7351.81 7505.30 10384.98 24344.89

Labor hours 3574.24 2522.49 7441.696 17150.04

Fertilizer cost (in Euros) 12042.62 9224.24 16014.11 34495.68

Crop protection (in Euros) 11706.12 9208.08 15196.12 31548.42

Energy (in Euros) 4929.07 5736.00 15415.72 37263.88

Asset (in Euros) 230537.5 168207.4 705247.8 1015290

Gross income (in Euros) 65900.22 53184.76 113756.6 252652.1

Insurance (in Euros) 3743.39 2470.82 6156.25 12428.24

Investment (in Euros) 17447.98 36418.1 34469.69 98552.22

Subsidy 26217.96 22938.2 48384.26 110861.4

Crop count (count index) 5.85 2.14 6.02826 2.225557

Age of the manager 52.84 10.63 48.02 17.24

Sunshine hour 1844.29 264.49 1622.81 146.79

Mean annual temperature 11.64 1.19 9.27 0.76

Mean annual precipitation 804.92 164.89 816.68 165.40

This paper investigates the relationship between risk exposure in the farm and the adoption of

risk mitigation strategies by controlling for demographic characteristics, environmental factors,

economic size and production structure etc. Specifically, we explore the impact of risk exposure

in the farm on farm diversification, insurance policy purchase decisions and disinvestment. The

analysis has two major stages; (i) estimation of the second and third order profit moments using

profit moment approach and (ii) the adoption model by including the lagged variance and

skewness estimates of farms from the first stage and a range of control variables.

Starting with a flexible representation of a farm profit function with vector of input variables

(𝑋ℎ𝑡), demographic characteristics (𝑆ℎ𝑡), land characteristics (𝐿ℎ𝑡) and environmental variables

(𝐸ℎ𝑡) and respective coefficients 𝛼, 𝛽, 𝛾 and 𝛿:

𝜋ℎ𝑡 = 𝑔(𝑋ℎ𝑡, 𝑆ℎ𝑡 , 𝐿ℎ𝑡, 𝐸ℎ𝑡; 𝛼, 𝛽, 𝛾, 𝛿) + 𝑢𝑖𝑡 (8)

Building on the moment based approach (Antle, 1983, Antle, 1987, Koundouri et al., 2006,

Groom et al., 2008), we estimate the second moment (variance) and third moment (skewness)

of the profit function with quadratic specification. we use the residual of equation 8 (𝑢𝑖𝑡), which

is assumed i.i.d (independent and identically distributed) with a zero mean and variance (δ2) to

estimate the second (variance) and third (skewness) order moments of profit function. Using

the approach illustrated in equation (4), we can use the first, second and third moments of the

profit function to estimate risk premium. In this particular context, however, we didn’t

particularly specify the functional relationship and estimate the Arrow-Pratt risk aversion

coefficient. We rather restrict ourselves to use these empirical moments as measures of cost of

risk exposure. This is particularly true under the assumption of general risk aversion and

downside risk aversion, where higher second moment (variance) and lower third moment

(skewness) increase the cost of risk exposure (Groom et al., 2008, Weitzman, 2009, Kim et al.,

2014). Apart from the production inputs and climatic variables, we include the level of farm

diversification, capital investment and asset of the farm as determinants of the profit function.

We estimated equations from 5 to 7 using the system of three-stage least squares technique

introduced by Zellner and Theil (1962). 3SLS allows efficiency improvement in the GLS

estimation as the model permits non-zero covariance between the error terms across these

equations (𝑒1𝑖𝑡, 𝑒2𝑖𝑡 , and 𝑒3𝑖𝑡). Furthermore, endogenous variables can be included as

explanatory variables in the estimation. To generate consistent estimates that accounts for

correlation of error terms across equations, 3SLS uses an instrumental variable approach

(Zellner and Theil, 1962, Davidson and MacKinnon, 1993, Greene, 2012). A farmer

experiencing exogenous shock can make simultaneous decisions to engage in diversified

farming, insurance policy purchase or may decide to disinvest. This simultaneous farm decision

requires simultaneous estimation especially when the decisions, and hence the error terms of

each estimation can be correlated to each other.

There are some econometric challenges to estimate equations 5 to 8. First, we need to determine

the functional form of the full profit moments estimations in equation 8. Based on the Akaike’s

Information Criterion (AIC) and Information Criterion (BIC), the quadratic functional form is

selected for the mean, variance and skewness functions (alternatives quadratic, logarithmic).

The fixed effect estimation of the quadratic specification for the profit moments function works

well compared to the random effects estimation. Second, the level of farm diversification and

investment in the farm could be endogenous. For this, we employ fixed effects instrumental

variable method for the estimation of the profit moments functions. We use the lagged values

of farm diversification and investment for this estimation. Third, it is vital to control for

endogeneity in the 3SLS estimations from equations 5 to 7. Greene (2012) suggested the use of

instrumental variables towards this effect. We use cultivated land, the level of diversification,

insurance policy purchase and investment in the previous growing season as instruments in the

3SLS. Fourth, one should control for heteroscedasticity problem in the 3SLS as it can lead to

inconsistent estimates. Following the approach suggested by Wooldridge (2010), we estimate

these systems of equations using Generalized Method of Moments (GMM) approach with a

weight matrix that uses explanatory variables as instruments. The Breusch-Pagan test of

independence in the Seemingly Unrelated Regression estimation also verifies the

interdependence between the major responses of arable farms for exposure to risk.

4. Result and discussions

In this section, we present and discuss the findings of this study. We use farm level and climatic

data from France and Germany to estimate the functions illustrated from equation 5 to 8. As

mentioned in the above section, the mean, variance and skewness functions were estimated

using the fixed effect instrumental variable panel data model with quadratic specification. This

flexible functional form helps us to capture the varied contributions of the production inputs,

their square terms and the interaction effects to the mean, variance and skewness of the profit

function. In addition to the variable inputs of production in a quadratic specification, we control

for climatic variables, economic structure and the level of diversification of farms.

4.1. Cost of risk exposure

We did several tests to compare the results of the alternative estimation methods. The Du-

Hauseman test in the first profit function (test statistic=3723.04 and p=0.00 for France and test

statistic=14842.73 and p=0.00 for Germany) verified the problem of endogeneity in the pooled

data analysis when we neglect the farm fixed effect in the estimation. The lagged values of farm

diversification and investment are significantly correlated with the level of diversification and

investment and they can influence profit moments only through their effect on diversification

and investment respectively. Both the individual p- values and the F- test of the overall model

verifies that these variables are strong instruments, and can be used in the specification. The

estimation results of the mean, variance and skewness functions are presented in Table 2. Most

of the production inputs and explanatory variables significantly influence the profit moments

functions in the expected direction. Except for fertilizer in Germany and fertilizer and energy

in France, the coefficients of production inputs in the mean profit function are positive.

An interesting finding worth noting at this point is the effect of farm diversification on the first,

second and third order profit functions. Specialized farms are likely to gain higher profits

compared to diversified arable farms in both countries. Nonetheless, we do find a varied result

on the role of farm diversification for the variance and skewness functions in France and

Germany. Despite higher variance, diversification contributes towards positively skewed profit

in France. On the other hand, diversification is associated with lower variability and negative

skewness in Germany. These all indicate that the economic role of diversification in arable

crops production in the two countries depends on the relative implication of diversification on

productivity and risk mitigation.

Table 2: Estimation of the first, second and third moment function

Variables France Germany

Mean Variance Skewness Mean Variance Skewness

Seed .072***(.007) .034***(.011) .159***(.069) .0147***(.011) .008 (.009) .079**(.034)

Seed^2 .002***(3.1e-04) -.002***(5.1e-04) -.001 (.003) .009***(3.9e-04) -.012***(3.6e-04) -.029***(.001)

Labor .211***(.008) -.239***(.013) -.592***(.082) .436***(.013) .137***(.008) .247***(.033)

Labor^2 .016***(7.2e-04) .029***(.002) -.040***(.009) -.002***(5.6e-04) .018***(.001) .038***(.004)

Fertilizer -.010 (.006) .003 (.932) .201 (5.702) -.076***(.006) .055***(.004) .122***(.014)

Fertilizer^2 -.004***(6.9e-04) -.001 (.001) -.013*(.007) .014***(5.2e-04) .011***(4.6e-04) .013***(.002)

Crop protection .043***(.005) 6.4e-04 (.037) -.028 (.229) .008***(.003) -.013***(.002) -.061***(.006)

Crop protection^2 -.004***(2.9e-04) -7.0e-04 (5.7e-04) .011***(.003) .001**(4.8e-04) .002***(2.8e-04) .008***(.001)

Energy .490***(.021) .136***(.033) 1.812***(.205) -.004***(6.9e-04) .105***(.010) .708***(.042)

Energy^2 -.011***(7.4e-04) .036***(.002) .008 (.015) -.003***(.001) .034***(7.8e-04) .067***(.003)

Seed*Labor -.009***(8.0e-04) .017***(.002) .159***(.012) .002*(.001) -.012***(.002) -.082***(.007)

Seed*Fertilizer -.005***(.001) .005**(.002) .153***(.013) -.003***(6.7e-04) -.009***(.001) -.011***(.003)

Seed*Crop protection .016***(.001) -.008***(.002) -.001 (.013) -.026***(7.6e-04) .006***(.001) .049***(.004)

Seed*Energy -.009***(9.3e-04) -.052***(.004) -.016 (.022) -.009***(7.1e-04) .034***(.001) .013***(.005)

Labor*Fertilizer .021***(.002) .030***(.004) .077***(.025) -.002**(8.5e-04) .004***(.001) .028***(.004)

Labor*Crop prot. -.016***(.001) 1.1e-04 (.002) -.049***(.014) .003**(.001) -.009***(.001) -.048***(.004)

Labor*Energy -.013***(.002) -.066***(.006) -.215***(.035) .016***(.001) -.027***(.002) .196***(.008)

Fertilizer*Crop prot. .007***(.002) -.001 (.003) .037*(.020) -.006***(.001) -.003***(9.9e-04) -.027***(.004)

Fertilizer*Energy .027***(.002) .078***(.007) -.002 (.039) -.016***(6.6e-04) .027***(.001) -.052***(.004)

Crop prot.*Energy .004***(9.4e-04) -.025***(.003) .014 (.018) .023***(9.2e-04) -.073***(.001) -.184***(.006)

Asset 4.5e-05***(1.3e-06) .101***(.007) .039 (.042) 3.1e-06***(3.8e-07) -.001 (.002) .053***(.010)

Investment -2.1e-06 (2.3e-06) -.005**(.002) .048***(.013) 4.6e-05***(1.3e-06) .006***(.001) .020***(.003)

Crop count -.347035***(.034) .011**(.005) .109***(.032) -1.291***(.036) -.129***(.003) -.352***(.013)

Sunshine hour 9.3e-06 (1.1e-05) 5.2e-05***(1.6e-05) 3.1e-04***(9.7e-05) 3.4e-05*(2e-05) 1.4e-06 (1.1e-05) 6.1e-05 (4.3e-05)

Mean annual temprature -.007 (.005) -.046***(.007) -.352***(.042) .009 (.006) 4.7e-04 (.003) .016 (.013)

Mean annual precipt. -1.6e-05 (1.3e-05) 5.4e-05***(1.9e-05) 4.7e-04***(1.2e-04) -6.6e-06 (2.2e-05) 2.2e-05*(1.2e-05) 1.1e-04**(4.7e-05)

Constant 0.021 (0.058) .399***(.099) 3.24***(.607) .014 (.067) -.042 (.035) -.293**(.145)

Model summary F = 552.29

Prob > F =0.000

F=44.54

Prob > F =0.000

F=14.56

Prob > F =0.000

F= 657.68

Prob > F = 0.000

F=143.63

Prob>F= 0.000

F=85.83

Prob>F= 0.000

Notes: The year dummies are not reported for space limitations. The numbers within the parenthesis are standard errors.

4.2. Risk mitigation and disinvestment in the farm

As we can learn from the descriptive statistics of the arable farms in the FADN dataset, we

observe the increasing trend for participating farms in the insurance market and the increasing

insurance premium paid by arable farms through years in both countries. Nonetheless, the level

of farm diversification didn’t show any specific trend. Though this looks surprising merely from

risk mitigation perspective, such a development can be explained at least from two perspectives.

First, farm diversification has a multitude of advantages in addition to its implication for risk

mitigation. This sometimes might even goes beyond the farm context including the role for

nutrition diversity. Second, farm diversity is one of the rural development packages promoted

by Common Agricultural Policy (CAP) in the EU.

Using the covariance analysis, we do find interdependence for most of the relationships between

risk exposure captured by the variability and skewness of farm profit in the preceding years

with diversification, farm investment and insurance in the season in the two countries. We

further explore the issue using estimation with the three-stage regression technique analyzed

through the General Method of Moments (GMM) approach, and the major results are discussed

hereafter.

We now turn in to discussing the results of the implications of risk exposure on risk mitigation

instruments and investment characteristics using the output from the 3SLS estimation. The

general test statistics confirm a model that controls the farm fixed effect is more robust than the

pooled cross section estimation. We also conduct the analysis using the Seemingly Unrelated

Regression. The Breusch-Pagan test of independence in the SUR estimation also verifies the

interdependence between the major responses of arable farms for exposure to risk.

Table 3: GMM estimation of three-stage Least Squares

Variables

France

Coeff. (Std. err.)

Germany

Coeff. (Std. err.)

Diversification Insurance Investment Diversification Insurance Investment

Variance (t-1) .027*** (.001) -18.67***(.685) -22.94***(8.14) .015** (.006) -20.77***(4.81) -622.02*** (65.35)

Skewness (t-1) 6.7e-04***(2.6e-04) -1.206*** (.141) -.224 (1.673) -.004* (.002) 3.61**(1.51) 95.00*** (20.43)

Subsidy (t-1) -.9.5e-05***(3.1e-06) .019***(.002) 1.2e-05 (.019) 2.5e-05**(1.0e-05) 0.083*** (.009) -.801*** (.055)

Income (t-1) 7.5e-06***(3.6e-07) .042***(1.9e-04) .106***(.002) 1.2e-05***(2.4e-07) .018***(1.7e-04) .149*** (.002)

Asset (t-1) 9.1e-06***(2.2e-07) .003***(1.2e-04) .021***(.001) 5.9e-06***(7.7e-08) .004*** (5.5e-05) 8.1e-04 (7.4e-04)

Sunshine hour 2.3e-05***(2.7e-06) 6.5e-04 (.001) -.004 (.017) 6.1e-05*** (6.5e-06) .012***(.004) -.081 (.059)

Mean temp. 9.9e-04 (6.3e-04) 1.652***(.339) -9.63**(4.04) -.003** (.001) .142 (.843) -1.147 (11.435)

Mean Preciptatio 2.9e-05***(3.5e-06) .002 (.002) .029 (.022) 4.8e-05***(6.5e-06) .019*** (.004) .090* (.054)

Sunshine_hr (t-1) 1.7e-05***(2.6e-06) .007*** (.001) .016 (.017) 4.4e-05*** (5.9e-06) -.039*** (.009) .135** (.055)

Mean temp (t-1) 0.001* (6.3e-04) -1.21***(.34) 3.37 (4.05) -0.005*** (.001) -5.288***(.832) -49.72*** (11.26)

Mean precip (t-1) 1.9e-05***(3.6e-06) .013***(.002) .0101(.026) 1.5e-05** (6.6e-06) .007 (.005) .229*** (.063)

Age 3.7e-04***(3.8e-05) -.121***(.020) 4.79***(.242) 8.5e-05* (4.5e-05) .057* (.032) 3.195*** (.431)

Economic size -3.5e-07***(5.1e-09) -6.9e-05***(2.7e-06) 1.9e-04***(3.3e-05) -2.7e-08***(2.1e-09) 5.6e-06***(1.5e-06) 7.7e-05***(1.9e-05)

Altitude zone -.003**(.001) 5.25***(.715) 2.404 (8.515) .009***(.001) 4.611*** (.795) .558 (10.77)

Constant -.041*** (.005) -21.51***(2.70) -214.83***(32.19) 3.559*** (.309) -3519.96***(62.73) -12103***(234.17)

Model N=36894 GMM estimation

Instruments= land (t-1), diversification (t-1), insurance (t-1) and

investment (t-1)

N=21392 GMM estimation

Instruments= land (t-1), diversification (t-1), insurance (t-1) and

investment (t-1)

Risk exposure and farm diversification

Our empirical estimation confirms the impact of risk exposure captured with variance of profit

functions for farm diversification decisions in arable farms in both countries. Arable farms are

likely to diversify their farm production activities as a response to higher profit variance in the

preceding years. We do also find an evidence that farmers that experience downside risk with

respect to their farm profit would likely to diversify their farm activities in Germany. These all

confirm that risk exposure of arable farms in the preceding years enhance the likelihood of

diversifying farm activities in both countries. Nonetheless, the analysis also confirm that farms

that experience positive skewness in France are likely to continue towards diversification in

their farm activities. The mixed results of the implication of skewness of farm profit in the

preceding years on farm diversification might indicate the incomplete risk protection function

of farm diversification. This is an important finding as farm diversification might have different

implications in different contexts. Despite the positive role of farm diversification for risk

mitigation as a buffer against environmental fluctuations and income variabilities (Di Falco and

Chavas, 2006, Di Falco and Chavas, 2009, Lin, 2011), farm diversification might not give

complete protection against climate extremes which often cause lower farm returns (Bradshaw

et al., 2004, Cafiero et al., 2007). Farms in the two countries partly rely on farm level

adjustments for risk mitigation even in the existence of market based risk mitigation instruments

including hell insurance. This finding reveals the incomplete protection of either of the risk

mitigation schemes, and we question the existing belief on the complete substitutability of farm

level and market based risk mitigation instruments.

In the empirical estimation, we do find an evidence that climatic variables and their lags are

essential elements of the farm diversification decision. Farm managers can learn from past

climatic experiences and seem to consider current climate records in the crop choice and

diversification decisions. These result confirms the role of learning and experience in the

adoption of risk mitigation strategies in the farm. There are similar findings on the impact of

climatic variables on the level of farm diversification both in the developed and developing

world (Di Falco et al., 2010, Bezabih and Sarr, 2012, Finger and Sauer, 2014). Finger and Sauer

(2014) for instance evidenced the role of the standard deviation of major climatic variables on

the farm diversification in the UK. On the other hand, Di Falco et al. (2010) and Bezabih and

Sarr (2012) found a significant contribution of the lagged climatic events on the level of farm

diversification in Ethiopia. The empirical estimation reveal that the age of the farm manager,

economic size of the farm and altitude zone of the farm significantly determine the level of farm

diversification of arable farms in both countries. In addition, agricultural subsidies to the farm,

income and asset of the farm in the preceding year significantly determine the level of farm

diversification in both countries.

Risk exposure and insurance

The estimation result confirms the effect of risk exposure of farms in the preceding years and

their propensity to get insured against climate extremes. Nonetheless, we do find mixed effects

on the implication of the skewness of the profit function in the two countries. While farms with

positively skewed profit in the previous seasons pay less insurance premium in France, the story

is completely the opposite for Germany. It is worth noting that the FADN dataset didn’t

exclusively differentiate the insurance premium paid for hell insurance, insurance for

machinery and buildings etc. In addition, except for the choice of the crops that farmers

cultivate, most of the other crucial elements of insurance premium rate determination are

associated to climatic variables in the region1 (European Commission, 2006). Considering these

facts, one has to be cautious when interpreting the results in this estimation.

We do find consistently significant effects of climatic variables and their lags on the level of

insurance premium paid by the farms. As discussed in the previous section, diversified farms

payoff especially when there exist varied inter-crop effects with climate variabilities (Lin,

2011). Nonetheless, farm diversification or other farm based risk management strategies might

not give complete protection at times of environmental extremes (Bradshaw et al., 2004, Cafiero

et al., 2007). In such a scenario, it is rather essential for farms to get protection against extreme

events from the farm insurance market. The agricultural subsidy, asset holding and the

agricultural income in the preceding year positively contributes to the insurance premium paid

by arable farms in the two countries. Decoupled payment and subsidies might often have an

effect on the propensity to buy insurance policies (Chakir and Hardelin, 2010, Finger and

Lehmann, 2012). In the same line, the age of the farm manager, economic size and altitude zone

of the farm significantly influence the insurance premium paid by arable farms in both

countries.

Risk exposure and disinvestment

Our estimation using arable farms in both countries confirm the impact of risk exposure

captured by profit variability on farm investment. Farms that experience higher farm profit

1 These variables include the frequency of risks in time and on area, the type of risk (hail, drought) and the number

of risks covered, the sensitiveness of crops, the number of farms insured, technicalities like deductibles

EUROPEAN COMMISSION 2006. Agricultural Insurance Schemes. Brussels: Institute for the Protection and

Security of the Citizen, Agriculture and Fisheries Unit.

variability in the preceding production year are likely to devote lower investment2 for farm

production activities. In addition to this, we do find an evidence that farms with positive

skewness in their profit are likely to invest more in the farm. This result confirms the research

hypothesis that risk exposure is likely to determine the propensity to invest in the farm at least

in a short run. In addition, Jalan and Ravallion (2001) indicated that wealth can be held

unproductive in the presence of risk as a buffer against low income levels.

We do find an evidence that climatic variables (the mean precipitation, temperature and

sunshine hours) and their lags significantly determine the level of farm investment in the two

countries. Alem et al. (2010) found similar result on the impact of climate variabilities on

fertilizer purchase decisions in Ethiopia. The past climatic trends and current climate readings

are essential elements of farm investment (disinvestment) decision in arable farms in the two

countries. Agricultural subsidy contribute negatively to the level of farm investment in

Germany while has no significant impact in France. This might partly be associated with the

significant proportion of transfers to decoupling, and other rural development supports in the

CAP that are not necessarily associated with production. Age of the farm manager and

economic size of the farm significantly determine the farm investment (disinvestment) in both

countries.

5. Summary and conclusions

This paper empirically investigates implication of cost of risk exposure on risk mitigation and

adaptation mechanisms and investment behavior of farms. For this purpose, we use an extensive

Farm Accountancy Data Network (FADN) panel data of arable farms and weather data from

1989 to 2009 from France and Germany. We employ a fixed effect panel data estimation for

the profit moments estimation. Three SLS regression using GMM approach is used for

analyzing the impact of cost of risk exposure on risk mitigation and adaptation strategies and

investment (disinvestment) in arable farms. Our empirical analysis confirms that risk exposure

measured with variance and skewness of farm profit can significantly influence the level of

farm diversification, insurance premium payment and farm investment and disinvestment in

arable farms in both countries. In the same way, we do find an evidence that climate variabilities

in the production season and preceding year have a significant implication on the farm level

and market based risk responses in the two countries.

2 Includes total expenditure on purchases, major repairs and own production of fixed assets and growth of young

plantations.

The findings in this paper can have several policy implications. First, farm diversification and

insurance seem to work together in arable farms to mitigate the pervasive impacts of risk

exposure. We verified from the analysis that on-farm diversification, insurance and

disinvestment remain to be important responses to risk exposure in arable farms in France and

Germany. In addition, despite a sharp rise in the uptake of insurance and premium insurance

scheme for risk management, we didn’t see a specific trend on farm level diversification

through years. This findings might reveal the incomplete protection of either of the risk

mitigation schemes, and we question the existing belief on the complete substitutability of farm

level and market based risk mitigation instruments. As these strategies seem not to be

completely substitutable, this evidence can be used to support the discussion of improving the

availability of market based instruments to improve the adaptive capacity of farms in the

developing world. Second, farms might sometimes see disinvestment as a response to extreme

shock and risk exposure at least in a short run. Improving the adaptive capacity of farms might

not only secure them pervasive impacts of risk exposure, but also can influence their investment

and disinvestment behavior. These findings emphasize the practical relevance of better

targeting in policy formulation and strategy development. Third, this empirical work highlights

implication of subsidies on farm resource allocation decisions. Agricultural policies to promote

farm diversification or other risk mitigation instruments should be targeted enough to bring

about the intended outcome.

6. References

ALEM, Y., BEZABIH, M., KASSIE, M. & ZIKHAIL, P. 2010. Does Fertilizer Use Respond to Rainfall Variability? Panel Data Evidence from Ethiopia. Agricultural Economics. Agricultural Economics, 41, 165-175.

ANTLE, J. M. 1983. Testing the Stochastic Structure of Production: A Flexible Moment-Based Approach. Journal of Business and Economic Statistics, 1, 192-201.

ANTLE, J. M. 1987. Econometric Estimation of Producers’ Risk Attitudes. American Journal of Agricultural Economics, 69, 509-522.

ARROW, K. J. 1965. Aspects of the Theory of Risk Bearing, Helsinki, Finland, Yrjo Jahnsson Saatio. ATWOOD, J., SHAIK, S. & WATTS, M. 2003. Are Crop Yields Normally Distributed? A Reexamination.

American Journal of Agricultural Economics, 85, 888–901. BAUMGÄRTNER, S. & QUAAS, M. F. 2010. Managing Increasing Environmental Risks through

Agrobiodiversity and Agrienvironmental Policies. Agricultural Economics, 41, 483-496. BEZABIH, M. & SARR, M. 2012. Risk Preferences and Environmental Uncertainty: Implications for

Crop Diversification Decisions in Ethiopia. Environmental and Resource Economics, 53, 483-505.

BIELZA, M., STROBLMAIR, J. & GALLEGO, J. 2007. Agricultural Risk Management in Europe. Paper prepared for presentation at the 101st EAAE Seminar ‘Management of Climate Risks in Agriculture’ Berlin, Germany, .

BRADSHAW, B., DOLAN, H. & SMIT, B. 2004. Farm-level Adaptation to Climatic Variability and Change: Crop Diversification in the Canadian Prairies. Climatic Change 67, 119-141.

CAFIERO, C., CAPITANIO, F., CIOFFI, A. & COPPOLA, A. 2007. Risk and CrisisManagement in the Reformed European Agricultural Policy. Canadian Journal of Agricultural Economics, 55 419–441.

CAREY, J. M. & ZILBERMAN, D. 2002. A Model of Investment under Uncertainty: Modern Irrigation Technology and Emerging Markets in Water. American Journal of Agricultural Economics 84, 171–183.

CHAKIR, R. & HARDELIN, J. 2010. Crop Insurance and Pesticides in French Agriculture: An Empirical Analysis of Multiple Risks Management. INRA and AgroParisTech Working Papers 2010/04. .

CHAVAS, J. P. 2004. Risk Analysis in Theory and Practice, Academic Press. CHAVAS, J. P. & DI FALCO, S. 2012. On the Role of Risk Versus Economies of Scope in Farm

Diversification with an Application to Ethiopian Farms. Journal of Agricultural Economics, 63, 25-55.

CHAVAS, J. P. & SHI, G. 2015. An Economic Analysis of Risk, Management, and Agricultural Technology. Journal of Agricultural and Resource Economics, 40, 63–79

CIAIS, P., REICHSTEIN, M., VIOVY, N., GRANIER, A., OGEE, J., ALLARD, V., AUBINET, M., BUCHMANN, N., BERNHOFER, C., CARRARA, A., CHEVALLIER, F., DE NOBLET, N., FRIEND, A. D., FRIEDLINGSTEIN, P., GRUNWALD, T., HEINESCH, B., KERONEN, P., KNOHL, A., KRINNER, G., LOUSTAU, D., MANCA, G., MATTEUCCI, G., MIGLIETTA, F., OURCIVAL, J. M., PAPALE, D., PILEGAARD, K., RAMBAL, S., SEUFERT, G., SOUSSANA, J. F., SANZ, M. J., SCHULZE, E. D., VERSALA, T. & VALENTINI, R. 2005. Europe-wide Reduction in Primary Productivity Caused by the Heat and Drought in 2003. Nature 437, 529 - 533.

CISCAR, J. C., IGLESIAS, A., FEYEN, L., SZABÓ, L., VAN REGEMORTER, D., AMELUNG, B., NICHOLLS, R., WATKISS, P., CHRISTENSEN, O. B., DANKERS, R., GARROTE, L., GOODESS, C. M., HUNT, A., MORENO, A., RICHARDS, J. & SORIA, A. 2011. Physical and Economic Consequences of Climate Change in Europe. PNAS (Proceedings of the National Academy of Sciences of the United States of America), 108, 2678–2683.

DAVIDSON, R. & MACKINNON, J. G. 1993. Estimation and Inference in Econometrics, New York, Oxford University Press.

DAY, R. H. 1965. Probability Distributions of Field Crop Yields. Journal of Farm Economics, 47, 713–741.

DERCON, S. 2004. Growth and Shocks: Evidence from Rural Ethiopia. Journal of Development Economics, 74, 309–329.

DI FALCO, S., BEZABIH, M. & YESUF, M. 2010. Seeds For Livelihood: Crop Biodiversity and Food Production in Ethiopia. Ecological Economics, 69 1695-1702.

DI FALCO, S. & CHAVAS, J. P. 2006. Crop Genetic Diversity, Farm Productivity and the Management of Environmental Risk in Rainfed Agriculture. European Review of Agricultural Economics, 33, 289-314.

DI FALCO, S. & CHAVAS, J. P. 2009. On Crop Biodiversity, Risk Exposure, and Food Security in the Highlands Of Ethiopia. American Journal of Agricultural Economics, 91, 599-611.

DIAZ-CANEJA, B. M., CONTE, C. G., DITTMANN, C., PINILLA, G. F. J. & STROBLMAIR, J. 2008. Agricultural Insurance Schemes. : European Commission, Joint Research Centre, Ispra, Italy.

DING, Y., SCHOENGOLD, K. & TADESSE, T. 2009. The Impact of Weather Extremes on Agricultural Production Methods: Does Drought Increase Adoption of Conservation Tillage Practices? . Journal of Agricultural and Resource Economics, 34, 395-411.

ENJOLRAS, G. & KAST, R. 2012. Combining Participating Insurance and Financial Policies. Agricultural Finance Review, 72 156 - 178.

ENJOLRAS, G. & SENTIS, P. 2011. Crop Insurance Policies and Purchases in France. Agricultural Economics, 42, 475-486.

EUROPEAN COMMISSION 2006. Agricultural Insurance Schemes. Brussels: Institute for the Protection and Security of the Citizen, Agriculture and Fisheries Unit.

FINGER, R. & LEHMANN, N. 2012. The Influence of Direct Payments on Farmers’ Hail Insurance Decisions. Agricultural Economics, 43, 343-354.

FINGER, R. & SAUER, J. 2014. Climate Risk Management Strategies in Agriculture-The Case of Flood Risk. Agricultural and Applied Economics Association Annual Meeting. Minneapolis, USA.

GOODWIN, B. K. 2008. Climate variability implications for agricultural crop production and risk management: discussion. American Journal of Agricultural Economics, 90, 1263–1264.

GREENE, W. H. 2012. Econometric Analysis, Upper Saddle River, NJ, Prentice Hall. GROOM, B., KOUNDOURI, P., NAUGES, C. & THOMAS, A. 2008. The Story of the Moment: Risk Averse

Cypriot Farmers Respond to Drought Management. Applied Economics, 40, 315-326. HARDAKER, J. B., HUIRNE, R. B. M., ANDERSON, J. R. & LIEN, G. 2004. Coping with Risk in Agriculture,

Wallingford, UK, CABI Publishing. IHLI, H. J., MAART-NOELCK, S. C. & MUSSHOFF, O. 2014. Does Timing Matter? A Real Options

Experiment to Farmers’ Investment and Disinvestment Behaviours. Australian Journal of Agricultural and Resource Economics 58, 430–452.

IPCC 2007. Impact, Adaptation and Vulnerability‖. IPCC 4th Assessment Report on Climate Change: Working Group II. Intergovernamental Panel on Climate Change.

IPCC 2014. Impacts, Adaptation and Vulnerability. Cambridge, UK: Intergovernmental Panel on Climate Change.

JALAN, J. & RAVALLION, M. 2001. Behavioral Responses to Risk in Rural China. Journal of Development Economics, 66, 23-49.

KASSIE, M., TEKLEWOLD, H., MARENYA, P., JALETA, M. & ERENSTEIN, O. 2014. Production Risks and Food Security under Alternative Technology Choices in Malawi: Application of a Multinomial Endogenous Switching Regression. Journal of Agricultural Economics.

KIM, K., CHAVAS, J. P., BAHRAM, B. & FOLTZ, J. 2014. Rice, Irrigation and Downside Risk: A Quantile Analysis of Risk Exposure and Mitigation on Korean Farms. European Review of Agricultural Economics, 41, 775-815.

KOUNDOURI, P., NAUGES, C. & TZOUVELEKAS, V. 2006. Technology Adoption under Production Uncertainty: Theory and Application to Irrigation Technology. American Journal of Agricultural Economics, 88.

LIN, B. B. 2011. Resilience in Agriculture through Crop Diversification: Adaptive Management for Environmental Change. BioScience, 61.

OECD 2009. Managing Risk in Agriculture: A Holistic Approach (Extracts). OECD Publishing. OREA, L. & WALL, A. 2012. Productivity and Producer Welfare in the Presence of Production Risk.

Journal of Agricultural Economics, 63, 102-118. PRATT, J. W. 1964. Risk Aversion in the Small and in the Large. Econometrica, 32, 122-136. RICHARDS, T. J. & GREEN, G. 2003. Economic hysteresis in variety selection. Journal of Agricultural

and Applied Economics, 35, 1–14. TACK, J., HARRI, A. & COBLE, K. 2012. More than Mean Effects: Modeling the Effect of Climate on the

Higher Order Moments of Crop Yields. American Journal of Agricultural Economics, 94, 1037-1054.

TUBIELLO, F. N., DONATELLI, M., ROSENZWEIG, C. & STÖCKLE, C. O. 2000. Effects of Climate Change and Elevated CO2 on Cropping Systems: Model Predictions at two Italian Locations. European Journal of Agronomy 13.

WEITZMAN, M. L. 2009. On the Modeling and Interpreting the Economics of Catastrophic Climate Change. Review of Economics and Statistics, 91, 1-19.

WOOLDRIDGE, J. M. 2010. Econometric Analysis of Cross Section and Panel Data, Cambridge, MA, MIT Press.

ZELLNER, A. & THEIL, H. 1962. Three Stage Least Squares: Simultaneous Estimate of Simultaneous Equations. Econometrica, 29, 54–78.

ZUO, A., NAUGES, C. & WHEELER, S. A. 2014. Farmers’ Exposure to Risk and Their Temporary Water Trading. European Review of Agricultural Economics, 41, 1-24.