Hi-Stat
Discussion Paper Series
No.226
Weather Risk, Wages in Kind, and the Off-Farm
Labor Supply of Agricultural Households
in a Developing Country
Takahiro Ito
Takashi Kurosaki
November 2007
Hitotsubashi University Research Unit
for Statistical Analysis in Social Sciences
A 21st-Century COE Program
Institute of Economic Research
Hitotsubashi University
Kunitachi, Tokyo, 186-8603 Japan
http://hi-stat.ier.hit-u.ac.jp/
Weather Risk, Wages in Kind, and the Off-Farm Labor Supply
of Agricultural Households in a Developing Country
∗
Takahiro Ito
†
and Takashi Kurosaki
‡
7 November 2007
Abstract
This paper investigates the effects of weather risk on the off-farm labor supply of
agricultural households in a developing country. Faced with the uninsurable risk of
output and food price fluctuations, poor farmers in developing countries may diversify
labor allocation across activities in order to smooth income in real terms. A key feature
of this paper is that it distinguishes different types of off-farm labor markets: agriculture
and non-agriculture on the one hand, and, wages paid in cash and wages paid in kind on
the other. We develop a theoretical model of household optimization, which predicts that
when farmers are faced with more production risk in their farm pro duction, they find it
more attractive to engage in non-agricultural work as a means of risk diversification, but
the agricultural wage sector becomes more attractive when food security is an important
issue for the farmers and agricultural wages are paid in kind. To test this prediction, we
estimate a multivariate two-limit tobit model of labor allocation using household data
from rural areas of Bihar and Uttar Pradesh, India. The regression results show that the
share of the off-farm labor supply increases with weather risk, the increase is much larger
in the case of non-agricultural work than in the case of agricultural wage work, and the
increase is much larger in the case of agricultural wages paid in kind than in the cash
wage case. Simulation results based on the regression estimates show that the sectoral
difference is substantial, implying that empirical and theoretical studies on farmers’ labor
supply response to risk should distinguish between the types of off-farm work involved.
JEL classification codes: Q12, O15, J22.
Keywords: covariate risk, non-farm employment, self-employment, food security, India.
∗
This is a thoroughly revised version of the COE Discussion Paper no.161, titled “Weather Risk and
the Off-Farm Labor Supply of Agricultural Households in India,” April 2006. The authors are grateful to
Nobuhiko Fuwa, Stefan Klonner, Daiji Kawaguchi, and the participants of the 2006 IAAE Conference for
their useful comments on earlier versions of this paper. All remaining errors are ours.
†
Graduate School of Economics, Hitotsubashi University. E-mail: ed044001@srv.cc.hit-u.ac.jp
‡
Corresp onding author. The Institute of Economic Research, Hitotsubashi University, 2-1 Naka, Kunitachi,
Tokyo 186-8603 Japan. Phone: 81-42-580-8363; Fax.: 81-42-580-8333. E-mail: kurosaki@ier.hit-u.ac.jp.
1
1 Introduction
This paper investigates the effects of weather risk on the off-farm labor supply of agricul-
tural households in a developing country. In low-income developing countries like India,
markets for agricultural inputs and outputs are well-developed, while the development of
credit and insurance markets has been lagging behind (Townsend, 1994; Kochar, 1997a;
1997b). This means that people in general, and particularly poor farmers, have few means
to hedge against the vagaries of production and price shocks that may put their livelihood
at risk (Fafchamps, 2003; Dercon, 2005). It has long been argued that poor farmers in
developing countries attempt to minimize their exposure to risk by growing their own neces-
sities (Fafchamps, 1992; Kurosaki and Fafchamps, 2002), diversifying their activities (Walker
and Ryan, 1990; Kurosaki, 1995), and through other income smoothing measures. If risk
avoidance inhibits gains from specialization and prevents farmers from achieving the output
potential they would be capable of, the provision of efficient insurance mechanisms becomes
highly important in poverty reduction policies.
As an example of such inefficiency due to risk avoidance, we focus on the labor supply of
farmers in developing countries. In the development literature, the relationship between risk
and labor market participation has been analyzed by several authors. For example, Kochar
(1999) and Cameron and Worswick (2003) examined the role of labor market participation as
an ex post risk-coping mechanism for households hit by idiosyncratic shocks, such as injury
or plot-level crop failure. The two studies showed that additional wage income was criti-
cally important for shock-hit households in India (Kochar) and in Indonesia (Cameron and
Worswick) to maintain consumption levels. Rose (2001) focused on the role of labor market
participation both as an ex ante and an ex post response to covariate shocks. She showed
that households facing a greater risk in terms of the reliability of rainfall were more likely to
participate in the labor market (ex ante response). Moreover, unexpectedly bad weather and
low rainfall also increased labor market participation (ex post response). Finally, Townsend
(1994) showed that Indian villagers found it more difficult to insure against covariate risk
than against idiosyncratic risk.
Taking these findings as our point of departure, we argue that in low-income developing
2
countries, it is important to distinguish different types of off-farm labor markets: agriculture
and non-agriculture on the one hand, and, wages paid in cash and wages paid in kind on
the other. Rose’s (2001) analysis simply considered a single labor market outside the farm,
which, however, raises the following problems. First, the covariance between farming returns
and agricultural wages is likely to be different from the covariance between farming returns
and non-agricultural wages. When an area is hit by bad weather, this may lead to a decline
not only in a farmer’s own farm income but also reduce the demand for agricultural labor
outside the farm, resulting in a high covariance between own-farm returns and wages available
from agricultural work. In contrast, wages outside agriculture are likely to be less correlated
with own-farm returns because they are less likely to be affected by the same kind of shocks.
This line of reasoning suggests that agricultural households would find it more attractive
to engage in non-agricultural work as a means of ex ante risk diversification. Second, the
covariance between wages and food prices also matters in determining the level of real income
(Fafchamps, 1992; Kurosaki and Fafchamps, 2002; Kurosaki, 2006). For farmers for whom
food security is an issue, agricultural work may nevertheless be more attractive than non-
agricultural work if agricultural wages are paid in kind, since the monetary value of wages
paid in paddy (the staple crop) is positively correlated with the paddy price. This paper
shows that both of these considerations do indeed play a role in determining the off-farm
labor supply of farmers in a developing country.
The remainder of the paper is organized as follows. In Section 2, we present a theoretical
model to explain how farmers decide to allocate their labor, incorporating considerations of
food security. We test the predictions of the model using household data from two Indian
states, Bihar and Uttar Pradesh. The dataset is described in Section 3, while the regression
results of a multivariate two-limit tobit model of labor allocation are presented in Section
4. The results robustly show that the share of the off-farm labor supply increases with
weather risk, the increase is much larger in the case of non-agricultural work than in the
case of agricultural wage work, and the increase is much larger in the case of agricultural
wages paid in kind than in the cash wage case. Section 5 shows simulation results based on
the regression estimates in order to examine whether the sectoral difference is economically
significant. Section 6 concludes the paper.
3
2 A Theoretical Model of Labor Allocation
In this section, we present a theoretical model to guide our empirical analysis. Throughout
the section, we assume a unitary decision making process at the household level with respect
to labor allocation (Singh et al., 1986).
1
To stylize the conditions of low-income developing
countries, we assume that there are only two consumption items: “food,” which is the main
output in production and the main item in consumption; and “non-food,” whose price is
normalized at one. The food price is p (= θ
p
p¯), where θ
p
is the multiplicative price risk with
a mean of one.
¯
For simplicity, we fix the total labor supply at L, ignoring the labor-leisure choice. The
welfare of the household is measured by its expected utility, which is defined as E[v(y, p)]
with the following properties:
v
y
> 0, v
p
< 0, v
yy
< 0, v
pp
< 0, v
yp
> 0, v
yyy
> 0. (1)
The first two properties are required for a valid indirect utility function. The third property
guarantees that the household is risk-averse in the Arrow-Pratt sense, and the fourth implies
that, for a given income level, the household’s welfare decreases when the food price variabil-
ity increases. The fourth property is especially appropriate for a (potentially) food-insecure
household in a developing country (Kurosaki, 2006). The last assumption, v
yyy
> 0, corre-
sponds to “risk prudence,” which is required for the welfare cost of consumption fluctuations
to decrease with the level of expected consumption (Kimball, 1990). In effect, these assump-
tions guarantee that the household behaves in a risk-averse and prudent way with respect
to income variability, suffers if food price variability is higher, and gains if the correlation
between the food price and income is higher.
2
¯
There are four different types of activity to which the household can allocate labor L
(indicated by subscript j): own farming (j = a), agricultural wage work paid in cash (j = b),
agricultural wage work paid in kind (j = c), and non-agricultural wage work (j = d). We
1
This assumption is based on our preliminary result from various demographic and health surveys in the
world that bargaining issues are less important in South Asia than in Sub-Saharan Africa. Extending the
analysis of this paper under a non-unitary household modeling framework and empirically testing whether
bargaining among members within a household is important in the current dataset are left for further study.
2
Note that when the food price and nominal income are positively correlated, real income is more stable.
4