Modern Economy, 2010, 1, 149-155
doi:10.4236/me.2010.13017 Published Online November 2010 (http://www.SciRP.org/journal/me)
Income Shocks and Consumption Smoothing Strategies:
An Empirical Investigation of Maize Farmer’s Behavior in
Kebumen, Central Java, Indonesia
Teguh Dartanto1, Nurkholis2
1PhD Student, GSID, Nagoya University, Japan and Lecturer, Department of Economics,
University of Indonesia, Indonesia
2Lecturer, Department of Economics, University of Indonesia, Indonesia
Received August 28, 2010; revised September 30, 2010; accepted N ovember 5, 2010
Abstract
Farmers in most developing countries usually face vulnerability in consumption due to income shocks
caused by crop loss, price falls, disaster, sickness and death and unexpected expenditure. They will respond
differently to income shocks depending on their asset ownership, labor endowment, access to loan, family
assistance, and family structure. We quantitatively analyze the consumption smoothing strategies of maize
farmers’ response to income shocks in Kebumen, Central Java-Indonesia. The Ordinary Least Square (OLS)
method confirms that selling cattle plays a central role in protecting consumption especially from income
shocks such as price falls and crop loss. Farmers that experienced income shocks related to demographic
shocks such as sickness and death and experienced expenditure shocks related to custom such as birth, mar-
riage and religious event require large effort by combining strategies to smooth their consumption. In this
case, farmers are forced to sell their land even though it is a costly alternative. However, widening access to
the loan market helps them maintain their consumption. Unfortunately, the hypothesis that the government
policies such as cheap rice, cash transfer and health insurance are effective instrument to smooth
consumption is not supported by consistent statistical evidences in all models.
Keywords: Income Shock, Consumption Smoothing Strategy, Rural Economic Development, Government
Assistance
1. Introduction
Chronic poverty is a major obstacle faced by Kebumen,
an agriculture based regency in the southern part of Cen-
tral Java. Reference [1] reported that Kebumen’s poverty
incidence in 2002 was 31.7% which is higher than either
the provincial level (23.1%) or the national level (18.2%).
In 2003, as one of poverty alleviation policies and creat-
ing job opportunities, the government of Kebumen, in
collaboration with a seed company and the central gov-
ernment, supported farmers through sev eral incentives in
maize cultivation. The incentives were a price discount
of hybrid seeds, fertilizers and pesticides subsidies, as-
sistances of post harvest process and a price guarantee1.
Farmers started to cultivate maize and substituted other
crops into maize. However, similar to farmers in most
developing countries, they face vulnerability in con-
sumption due to income shocks caused by crop loss,
price falls, disaster, sickness and death and unexpected
expenditure. Both external shocks such as disaster, cli-
matic risks and economic fluctuations and individual
specific shocks such as sick- ness, death and other cus-
tom shocks make farmer households vulnerable to seri-
ous hardship.
A survey of 220 maize-farm-households was collected
from the rural area in Kebumen during August 2009 . The
survey shows that during the last five years, farmers
faced crop loss due to disaster, climate shocks, pests,
rodents and other calamities was about 59% and price
falls was 73%. Moreover, households experienced demo-
1Suara Merdeka (03/17/2003), http://www.suaramerdeka.com/harian/
0303/17/dar16.htm. Ministry of trade and industry, “MOU kerjasama
Penanaman Jagung di Kebumen,” 2003. http://www.depeag.go.id/
files/publikasi/sigran pers/2003/kebuman2.htm

150 T. DARTANTO ET AL.
graphic shock related to sickness and death was 16.81%.
About 53.6% of households experienced expenditure
shocks related to customs such as such as wedding, cir-
cumcision and birth. In the same period, about 85% of
households experienced at least one shock, and every
household had approximately two income shocks on av-
erage. Maize-Farmers in Kebumen tried to combine
many strategies to smooth their consumption in response
to income shocks.
Similar to the current survey, other studies also showed
how farmers respond differently to income shocks de-
pending on their asset ownership, labor endowment, ac-
cess to loan, family assistance, and family structure.
Reference [2] reported that an increase in labor supply
was the key response to income shocks in rural India.
References [3,4] showed that credit markets played a
central role in protecting consumption from income
shocks. Reference [5] found the sale of assets for
smoothing consumption. Moreover, references [6,7] sur-
veyed the role of cattle/livestock as a buffer for income
shocks. However, farm households, due to the lack of
other alternatives, are forced to protect consumption
from idiosyncratic income shocks through relatively
costly methods [2].
This research aims to find out the consumption
smoothing strategies of maize farmers in Kebumen as a
response to income shocks. The study also evaluates the
effectiveness of the government policies in smoothing
their consumption. The nationwide policies such as
cheap rice, cash transfer and health insurance are distrib-
uted aiming to protect the poor from vulnerability in
consumption. These policies have been implemented for
quite long time. Cheap rice, basically food price subsidy,
started in 1998 followed by cash transfer and health in-
surance in 2005. Thus, the research outcome of the cur-
rent study is not only useful for discussion of coping
strategies adopted as a response to income shocks, but
also relevant to policy makers to construct an effective
policy to protect farmers from vulnerability especially in
Indonesia.
This article consists of three main parts. The first part
describes the maize production in Kebumen and the sec-
ond part reviews the methodology and data utilized in
this research. The last part analyzes empirical results
from OLS estimation.
2. Overview of Maize Production in
Kebumen
According to Kebumen in Figures 2008, Kebumen ad-
ministratively consists of 26 districts with a total area of
128,111.50 hectares or 1,281.1 km². Most of area is dry
lands (68.96%). By 2008, the population is 1,241,437
and 52.85% of them depend mainly on agricultural ac-
tivities as sources of income. Thus, they are seriously
vulnerable from income shocks such as sickness, dead of
family member, crop loss, bad weather, rodent and dis-
aster.
The local government of Kebumen in collaboration
with the central government supported farmers to culti-
vate maize through some incentives in order to improve
farmers’ welfare and utilize the dry fallow lands. As a
result of this policy, maize cultivation area sharply
increased from 2,714 hectares in 2002 to 4,717 hectares
in 2003. Moreover, the maize production increased from
7,537 metric ton in 2002 to 15,382 metric ton in 2003
(Figure 1).
However, in 2004, many farmers did not continue to
cultivate ma ize, therefore, both maize cultivation area and
maize production significantly decreased. We observed
that there are two main reasons why farmers stopped to
cultivate maize. First, government was inconsistence in
supporting farmers to cultivate maize. Second, a massive
land conversion from other crops into maize had created
abundant supply of maize thus the maize price dropped.
Both price fall and inconsistent policy have created disin-
centive for farmers to cultivate maize.
Figure 1 shows that maize production fluctuated since
2003. However, following an increase in the interna-
tional price of maize in 2007, farmers massively substi-
tuted other crops into maize and utilized the fallow dry
lands to cultivate maize. Consequently, the maize pro-
duction sharply increased by 123% during the last two
years. We also observed that the productivity increased
almost 86% from 2.78 metric ton/hectare in 2002 to 5.17
metric ton/hectare in 2008. The highest productivity was
Figure 1. Land Utilization of Maize Production and Pro-
ductivity. (Source: Authors’ calculation based on Ke bumen
in Figures 2004, 2005, 2006, 2008.)2
2Central Statistic Agency (BPS) Kebumen, “Kebu-men in Figures,
Several Publications of 2004, 2005, 2006 and 2008,” Kebumen: Bap-
p
eda dan BPS. http://www.bappeda.kebumenkab.go.id/data/dda_2008.
p
df

T. DARTANTO ET AL.
151
observed in 2005 when one hectare could produce 5.29
metric tons. The main sources of the high maize produc-
tivity are the use of hybrid seeds, the application of pro-
duction process in an appropriate and measurable way
and the learning by doing process.
3. Methodology and Data
We propose an econometric model to quantitatively es-
timate farmers’ consumption smoothing strategies to re-
cover from income shocks. This model is based on [8-
10], in this research, we propose a two-step-calculation.
First, we calculate the household consumption gap which
is derived from the difference of consumption expendi-
ture between those reported shocks and those in the ab-
sence of shock s.
ˆii
h
h
C
C
i
 
r
C
i
(1)
ˆh
C is per capita consumption expenditure gap of
households-h. s consumption expenditure of house-
holds-h with family members-i who have reported
shocks in Kebumen, h = 1, …, 220; i is number of family,
i = 1, …, i; Meanwhile,
i
h
C
i
r
C is average consumption
expenditure of poor rural households in absence of
sh ocks with famil y me mbers –i in region-r, r = Kebumen
and the surrounding regency in Central Java Province.
We only consider the sample with negative gap for fur-
ther analysis. It is assumed that the positive gap means
the households are not affected by the shocks. We ob-
served that those who have a positive consumption gap
are categorized as non poor based on the ownership of
asset and endowment indicators and the consumption
expenditure. Therefore, the sample used in this research
is the poor-maize-farmers in Kebumen.
We utilized the different data set drawn from the 2008
National Socio-Economic Survey (SUSENAS) to proxy
the rural consumption expenditure in the absence of
shocks. It is because most of the surveyed respondents
experienced shocks. It is justified that since the questions
related to the household consumption expenditure on the
surveyed questionnaire referred to the questions on
SUSENAS, thus, utilizing the average rural expenditure
as a proxy of consumption in the absence of shocks can
create unbiased approximation. This approximation is
calculated from the 7,441 samples of rural households.
We made some adjustments on the surveyed data of
household consumption in which the value is adjusted
into the 2008 valu e by using the price index.
Second, the econometric model shown by Equation 2,
calculates the farmers’ strategies in order to smooth their
consumption. The dependent variable is calculated from
first step, while the independent variables included in
this model refer to the previous researches done by
[2-10]. We expect that all coefficients will be negative
which means chosen strategies are effective to narrow
the consumption gap.

01 2
34 5
67
89
ˆ
log log
hh
h
hh h
hh
hhh
CLANDOTH
LOANREMIT CATTLE
SALELAND RASKIN
TRANSFER ASKES
 
 


JOB

 
 


(2)
where,
 ˆ
C

: absolute per capita of consumption expendi-
ture gap,
 LAND: land ownership of household in squared
meters,
 OTHJOB: dummy variable of side jobs; 1: having
side jobs, 0: ot her wi se ,
 LOAN: dummy variable of access to loan; 1: hav-
ing access, 0: otherwise,
 REMIT: dummy variable of receiving remittance; 1:
receiving remittance , 0 : otherwise,
 CATTLE: dummy variable of selling cattle; 1: sell-
ing cattle, 0: otherwise,
 SALELAND: dummy variable of selling land; 1:
selling land, 0: otherwise,
 RASKIN: dummy variable of receiving cheap rice;
1: receiving cheap rice, 0: otherwise,
 TRANSFER: dummy variable of receiving cash
transfer; 1: receiving cash transfer, 0: otherwise,
 ASKES: dummy variable of receiving poor health
insurance, 1: receiving poor health insurance, 0:
otherwise,


: error term i.i.d () 0E

, 22 ()E

 ,
 h: household-h, h = 1, …, 220.
The coefficients in the model were estimated using
Ordinary Least Square (OLS) by dividing samples with
four sub samples based on reported shocks such crop loss
(Model 1), price falls (Model 2), sickness and death
(Model 3) and Customs (Model 4). Separating sample
helps to show how farmers respond to each shock.
However, in Model 1 and Model 2, we deliberately did
not include the variables of land sale (SALELAND) and
health insurance (ASKES). It is too costly for the house-
holds to sell their own land as an alternative to smooth
consumption from crop loss and price fall shocks.
Moreover, the reason for excluding health insurance is
that this policy is distributed only to the poor aiming to
ease access to health services. Moreover, in Model 4, we
also deliberately did not include all variable of govern-
ment policies because these policies are not purposed to
cope the expenditure shocks.

T. DARTANTO ET AL.
152
Table 1 shows a descriptive analysis of data used in
this econometric model. The data shows that the average
per-capita consumption gap varies depending on the
shocks. The average per-capita consumption gap of
farmers experienced crop loss was IDR 131.8 thousands
(14 USD) while that of farmers experienced sickness-
death was IDR 106 thousands (11 USD). Table 1 also
shows that maize farmers in Kebumen are dominated by
small and subsistence farmers. Most farmers are aver-
agely holding land around 0.15–0.169 hectares. If the
productivity is 5.17 metric ton/hectare, farmers can only
produce approximately 0.77–0.87 metric ton of maize.
This amount might not enough to cover their daily cost.
However, we found that most farmers have side jobs,
thus their income is not solely depending on agriculture
activities.
On the other hand, consumption smoothing strategies
chosen by farmers are quietly different depending on
shocks. Selling cattle was the most favorite strategy to
cope all income shocks and the second strategy was ask-
ing family members who are working either inside or
outside Indonesia to send remittances. The third strategy
was access to loan from financial institutions which was
chosen by 17.6% of farmers in order to cope price falls.
However, most farmers combined many strategies re-
sponding to shocks because a single strategy might not
enough to cover the consumption gap. Moreover, farmers
faced demographic shocks related to sickness and death
and also custom shocks related to birth, family marriage
and religious events were forced to sell land.
Table 1 also summarizes the distribution of govern-
ment policies related to social safety nets. The house-
holds receiving cheap rice was 87% (crop loss), 72.5%
(price falls), and 90% (sickness-death) and the house-
holds receiving cash transfer was 63% (crop loss), 51.4%
(price falls), and 54.8% (sickness-death). The percentage
of those receiving cash transfer was lower than that of
those receiving cheap rice since the strict conditions
must be satisfied to receive cash transfer. On the other
hand, we found that more than one third of the house-
holds experienced sickness-death shocks received poor
health insurance. Th is insurance is distributed to the poor
in order to improve their acce ss to health facilities.
4. Empirical Results
Table 1 has already shown that maize farmers respond
differently to income shocks. However, the descriptive
analysis could not able to statistically determine the most
referred strategy chosen by farmers for narrowing the p
Table 1. The Result of Survey in Kebumen, Central Java, Indonesia.
Variable Crop Loss Price Falls Sickness-Death Customs
Dependent Variable
Average Per-capita Consumption Gap (IDR) 131,801.6 116,363.0 106,016.5 114,969.6
Independent Variables
Control Variables
Average Land Owning (Square Meter) (LAND) 1,541.3 1,608.1 1,691.9 1,549.3
Having Side Jobs (OTHJOB) 88.6% 90.1% 96.7% 97.0%
Smoothing Strategies - - - -
Access to Loan (LOAN) 7.0% 17.6% 3.2% 9.9%
Remittances (REMITT) 7.9% 9.9% 9.7% 10.9%
Cattle Sales (CATTLE) 79.8% 73.2% 90.3% 85.1%
Land Sales (SALELA ND) 3.2% 2.0%
Government Policies
Cheap Rice (RASKIN) 87.0% 72.5% 90.3%
Cash Transfer ( BLT) 63.0% 51.4% 54.8%
Poor Health Insurance (ASKES) 38.7%
Source: Authors’ calculation based on Survey Data .

T. DARTANTO ET AL.
153
Table 2. Regression result.
Log Consumption Gap
Variables Crops Loss
(1) Price Falls (2) Sickness-Death (3) Customs (4)
14.000*** 14.828*** 15.482*** 14.835***
Constant 19.386 14.742 6.308 16.999
–0.264** –0.439*** –0.436 –0.435***
Land Ownership (LAND) –2.539 –3.130 –1.573 –4.055
–0.399** –0.065 –0.573 0.136
Side Job (OTHERJOB) –2.270 –0.451 –1.217 0.471
–0.742 –0.391 –3.246*** –0.591**
Access to Loan (LOAN) –1.109 –0.717 –18.974 –2.453
–0.394** –0.270 –0.968*** –0.125
Remittances (REMIT) –2.367 –1.275 –3.101 –0.587
–0.338* –0.780*** 0.536 –0.329*
Cattle Sales (CATTLE) –1.936 –3.278 1.445 –1.861
- - –0.330*** –2.466**
Land Sale (SALELAND) - - –4.155 –2.245
–0.079 0.575* –0.927*** -
Cheap Rice (RASKIN) –0.307 1.636 -2.818 -
0.311* 0.154 0.342 -
Cash Transfer (TRANSFER) 1.904 0.748 1.632 -
- - –0.171 -
Poor Health Insurance (ASKES) - - –0.621 -
R-Squared 0.372 0.385 0.788 0.421
F-Statistic 8.979 11.962 8.649 11.374
Observation 114 141 31 101
Source: Authors’ estimation. Notes: Figures in italic are t-statistic. The standard errors are corrected due to heteroscedasticity. ***,
**, * are significant at 1% level, 5% level and 10% level, respectively.
consumption gap. Therefore, we apply OLS to statistic-
cally estimate the consumption smoothing strategies.
Generally, all models are suitable for analyzing the con-
sumption smoothing strategies of maize farmers’ in re-
sponse to income shocks in Kebumen. It is shown by a
statistically significant of F-statistic in all models.
Moreover, R-squared of four models are 0.37 (Model 1),
0.39 (Model 2), 0.79 (Model 3) and 0.42 (Model 4).
These values could be categorized as high enough for
cross section estimation. The estimation results of four
models are shown in Table 2.
Model 1 measures farmers’ consumption smoothing
strategies to cope with cro p loss. The estimation sign ify-
cantly confirms that farmers who experienced this shock
relied on remittance and cattle sales as a buffer for
smoothing consumption. About 8% of households ex-
perienced this shock reported receiving remittance from
family members who are working either inside or outside
Indonesia while about 80% of households reported sell-
ing cattle to cope with this shock. The farmers receiving

154 T. DARTANTO ET AL.
remittance were able to narrow their consumption gap by
0.394 while those selling cattle narrowed their co nsump-
tion gap by 0. 3343.
Model 2 evaluates farmers’ consumption smoothing
strategies in coping price falls. The OLS regression con-
firms that selling cattle played a central role to cope with
price falls by narrowing consumption gap up to 0.78.
Approximately 73% of househ olds reported selling cattle
to deal with this shock. Moreover, selling cattle might be
enough to cover the gap because price falls affects
consumption smaller than crop loss. The degree of
impact is different since farmers suffering from price
falls are still able to harvest crops meanwhile those suf-
fering from crop loss are not. This estimation also
showed that households may not need to utilize either
remittance or loan as an alternative to smooth consump-
tion. In similar with Model 1, land ownership signify-
cantly reduced the consumption gap. Moreover, neither
Model 1 nor Model 2 showed access to loan as an alter-
native to cope this sho ck. The difference between Model
1 and Model 2 was the ineffectiveness of side jobs and
remittance in narrowing the consumption gap shown in
the second model.
Model 3 analyzed farmers’ consumption strategies to
deal with demographic shocks related to sickness and
death. The death of a productive family member reduces
income as well as consumption due to loss of labor input
in agricultural activities or selling labor to othe r activities.
Since the health insurance system in Indonesia is unde-
veloped yet, sickness is closely related to an unexpected
expenditure withdrawing a larger share of household
income. Therefore, farmers respond to death and sick-
ness quiet differently from our previous models. Selling
cattle, land ownership and side jobs which previously
play a central role, are replaced by access to loans, re-
mittance and selling land . The reason is that selling cattle
might not be sufficient to cover the gap due to a larger
reduction of i n come.
The coefficients of Model 3 indicate the condition
where access to loan is the first alternative chosen, while
remittance and selling land is the next alternatives. Under
these alternatives, the consumption gap narrows by 3.25,
0.97 and 0.33, respectively. This confirms that selling
land is a costly option selected only when other alterna-
tives are not feasible. Surprisingly, even though the
magnitude of both land ownership and side jobs in nar-
rowing the consumption gap is consistent with Model 1
and Model 2, neither land ownership nor side jobs is sig-
nificant. The main reason is that those reported for these
shocks were only 31 samples, thus the standard error (SE)
would be high due to large variance. In contrast to Model
1 and Model 2, government policies seem to be effective,
which is shown by negative coefficients of both cheap
rice (RASKIN) and poor health insurance (ASKES), in
assisting maize farmers to cope with the demographic
shocks associated with sickness and death. Unfortunately,
the negative coefficient of ASKES is not statistically
significant in narrowing the gap.
Lastly, Model 4 investigates the farmers’ consumption
strategies in coping to expenditure sho cks related to cus-
tom such as birth, marriage, culture and religious even
such circumcision. As well as the demographic shocks,
these shocks take account a quite large share of income.
However, the managing strategies are different from
Model 3. The farmers choose to selling land, making
loan and selling cattle in priority. The coefficients of
Model 4 indicate the condition where selling land is the
first alternative chosen, while access to loan and selling
cattle are the next alternatives. Under these alternatives,
the consumption gap narrowed by 2.47, 0.59 and 0.33,
respectively. This confirms that custom shocks need a
large source to finance the gap. Thus, the selling land
becomes the first alternative. However, farmer experi-
enced selling land might have a serious future cones-
quence since all models show that farmers holding large
land size are relatively resilient to any type of income
shocks. Moreover, in contrast to Model 3, remittance is
statistically insignificant to narrow the consumption gap
in this model.
5. Concluding Remark
Farmers respond differently to income shocks depending
on their ownership of assets, access to loan, family as-
sistance such as remittance and the typ e of shocks. In the
case of maize farmers in Kebumen, consumption
smoothing strategies vary in accordance to the type of
shocks and the magnitude of their impact on household
income as well as consumption. If a shock for example
price falls, has only a little impact on income, farmers
choose to sell cattle to protect their consumption. In ad-
dition, during other shocks with greater impact than price
fall, such as crop loss, farmers not only sell their cattle
but also need remittance as an additional coping strategy.
An opposite smoothing strategy from previous strategies
is chosen when an income shock occurs due to sickness
and death. Farmers who experienced this type of shocks
face difficult choices to protect their consumption. In the
worst case, they are forced to sell their land even though
it is costly. However, widening access to loan market
enables them to easily protect their consumption. Unfor-
tunately, a consistent statistical evidence in all models
does not exist to support the hypothesis that government
policies such cheap rice, cash transfer and poor health
insurance are effective as an instrument of consumption
3The consumption gap m ea ns t he l og arithm consumption gap.

T. DARTANTO ET AL.
155
smoothing policy. Like many previous research findings,
this research also confirms that maize farmer households
holding large land size are relatively resilient to any type
of income shocks.
6. Acknowledgements
We would like to thank University of Indonesia for
funding this research through the National Research
Strategic Fund 2009. We also thank Mrs. Lily Yunita
(University of Indonesia), Mr. Yoshua Wisnungkara and
Mr. Mark Rebuck (Nag oya University) for their valuab le
comments. Any remaining errors are our responsibility.
7. References
[1] BPS-BAPENAS-UNDP, “National Human Development
Report 2004, the Economics of Democracy: Financing
Human Development in Indonesia,” Jakarta: BPS, 2004.
http://www.undp.or.id/pubs/ihdr2004/ihdr2004_full.pdf
[2] K. Kochar, “Explaining Household Vulnerability to Idio-
syncratic Income Shocks,” The American Economic Re-
view, Vol. 85, No. 2, 1995, pp. 159-164.
[3] M. Eswaran and A. Kotwal, “Credit as Insurance in
Agrarian Economies,” Journal of Development Econom-
ics, Vol. 31, No. 1, 1989, pp. 37-53.
[4] M. R. Rosenzweig, and K. Wolpin, “Credit Market Con-
straints, Consumption Smoothing, and the Accumulation
Of Durable Production Assets In Low-Income Countries:
Investments In Bullocks In India,” The Journal of Politi-
cal Economy, Vol. 101, No. 2, 1993, pp. 223-224.
[5] J. Morduch, “Income Smoothing and Consumption
Smoothing,” Journal of Economics Perspectives, Vol. 9,
No. 3, 1995, pp. 103-114.
[6] H. Kazianga and C. Udry, “Consumption Smoothing,
Insurance and Drought in Burkina Faso,” Journal of De-
velopment Economics, Vol. 79, No. 2, 2006, pp. 413-
446.
[7] J. McPeak, “Contrasting Income Shocks with Asset
Shocks: Sales in Northern Kenya,” Oxford Economic
Papers, Vol. 56, No. 2, 2004, pp. 263-284.
[8] M. R. Rosenzweig, “Risk, Implicit Contracts and Family
in Rural Areas Of Low-Income Countries,” The Eco-
nomic Journal, Vol. 98, No. 393, 1988, pp. 1148-1470.
[9] A. Kochar, “Smoothing Consumption by Smoothing
Income: Hours of Work Responses to Idiosyncratic Ag-
ricultural Shocks in Rural India,” The Review of Eco-
nomics and Statistics, Vol. 81, No. 1, 1999, pp. 50-61.
[10] G. Berloffa and F. Modena, “Income Shocks, Coping
Strategies and Consumption Smoothing: An Application
to Indonesian Data,” Discussion Paper No.1-2009, Uni-
versita degli Studi di Trento, Dipartimento di Economica,
Italy.