An
econometric analysis of the link between irrigation, markets and poverty in Ethiopia: The case
of smallholder vegetable and Fruit Production in the North Omo Zone, SNNP
Region
By
Tadele
Ferede
Deble
Gemechu
April 2006
Abstract
This paper examines the anti-poverty impacts of irrigation
and markets on the welfare of rural households within the PRISM framework.
Specifically, the paper addresses: the magnitude of anti-poverty effects of
irrigation and conditions for strengthening the poverty-reducing impact of it;
and the market constraints of fruit and vegetable producers. The relationship
between irrigation, market-orientation of smallholders and poverty is examined
using descriptive statistics and multivariate analysis. In the descriptive
analysis, low prices for vegetable and fruit, weak demand, lack of price
information, and inadequate transportation have been identified as the main
limiting factor for output market.
The beneficial effect of
irrigation, literacy rate of household heads, and extra years of schooling is
readily apparent from the regression. A simulation approach is also used to
explore the impact of irrigation and other factors, individually and together,
on poverty. The results show that although irrigation reduces poverty, the
effect is greater when combined with improving the literacy level of
households. This
evidence calls for policy measures that focus on the concurrent interventions
in irrigation,
education, markets, and other supporting inputs, thereby reducing poverty in
the cash growing rural areas.
Key words: Ethiopia, Fruits,
irrigation, markets, poverty,vegetables.
1. Introduction
Ethiopia, like other sub-Saharan African (SSA) countries, is an agrarian
economy, with a very small industrial sector. The agricultural sector, on
average, accounts for about 45% GDP; 90 percent of merchandise
export earnings; 80% of employment; more than 90% of the total foreign exchange earnings; 70% of the raw material supplies for agro-industries, and is also a
major supplier of food stuff for consumers in the country. Smallholders who produce
more than 90% of the total agricultural output and cultivate close to 95% of
the total cropped land dominate the sector. Inter-regional difference in terms
of agricultural production is quite noticeable. For instance, the three
regions, namely Amhara, Oromiya and Southern Nations, Nationalities and Peoples
(SNNP) contribute more than three-fourth of the total agricultural production
in the country. Agricultural production is highly dependent on the vagaries of nature
with significant variability in production and actual production patterns. The majority of smallholders have not
practiced irrigation to mitigate the adverse effects of weather variability and
water is the main limiting factor of agricultural production. As a result, the
poverty situation of the country has not changed over time. For instance, the proportion of people
living under the absolute poverty line in 1999/00 was close to the level five
years earlier (1995/96), estimated at 45% (MOFED 2002). Poverty (on the
aggregate) has, at best, not decreased in spite of improved economic
performances in the 1990s. Moreover, poverty is concentrated in the rural areas
where basic services are in critical shortage to meet the bare minimum demands
(Mulat et al, 2005).
It has been documented that low farm production and productivity
resulting from use of backward technology and other productivity-enhancing
modern inputs are the major reasons for rampant poverty and food insecurity in
rural Ethiopia (Workneh, 2005; Mulat et al, 2005). Given diminishing arable
land per capita and
limited off-farm activities, increasing farm production through improved
technology-based farm intensification can be an important strategic element for
agricultural growth and rural development.
Utilization
of modern technologies is extremely low in Ethiopia (Mulat 1999; Belay 2003;
Mulat et al, 2003). For example, close to 2% of peasants use improved seeds
(Abebe and Mulat 2003). Similarly, less than 5%
of the total irrigable land is utilized so far and use of other modern inputs
is very low. This is
extremely scary but reflects the existence of a huge gap between the actual
performance and the potential that could be attained through improved
technologies (Mulat et al, 2003). By all counts, the
country has experienced very little in terms of productivity-driven
agricultural growth and poverty reduction. Note that whatever growth has been
registered in the agricultural sector, it is mainly driven by area expansion
with little gain in productivity. For instance, based on the past trends of
agricultural performance (specifically, between 1995 and 2002), about 70% of
the increase in crop production has been attributed to area expansion (Diao et al, 2005).
In Ethiopia, roads are extremely
underdeveloped: the average road density is 27 kilometers per 1,000 km2. Close to 70% of Ethiopian farmers live in areas more than half a day’s
walk away from an all-weather road. Poor market access which entails high
transportation costs significantly increases the gap between consumer and
producer prices. This in turn lowers the farm gate prices received by farmers
located in remote areas. It has been documented that the average grain price gap
is estimated to be in the range of 30 to 70 percent across regions. Moreover, domestic marketing costs can
account for more than half of fertilizer prices paid by farmers, which tends to
reduce profitability of modern inputs (Jayne et al, 2003). A recent study also
confirms that agricultural growth in Ethiopia requires concurrent investments
in roads and other market conditions (Diao et al, 2005). Better access to
markets has reduced the cost of inputs and expanded the market for produce.
Any solution to
reduce rural poverty must focus on increasing the production and productivity
of smallholder agriculture and speed up the process of structural
transformation. In order to overcome the challenges faced by small farm
households,
comprehensive market-based poverty reduction interventions, which establish a
framework to operationalize integrated market systems for the rural poor, are
required. The intervention would ensure sustainable natural resource
management, reduce poverty and enhance gender equity if based on access and
control of water for crop irrigation. The framework, which focuses on reducing
rural poverty via use of irrigation and smallholder markets, is known as
Poverty Reduction through Irrigation and Smallholder Markets (PRISM). The PRISM
model places a high priority on identifying strategies that enable smallholders
to access, store and control water for crop irrigation via low cost, household
level, micro-irrigation systems which maximize water-use efficiency, minimize
labour-burdens and brings high economic returns to the poor small farm
households.
The main objective
of this paper is to provide empirical evidence regarding the impact of
irrigation and markets on poverty in the cash growing rural area of Southern
Nations, Nationalities, and Peoples (SNNP) region. To do so, both descriptive,
econometric and simulation techniques have been used.
The rest of the
paper is organized as follows. In the second section, we review the link
between irrigation, markets and poverty. Section III presents a model of
welfare determination and develops a simulation framework. Data description,
simulation results and analysis also are
given in this section. The final
section, Section four, concludes.
2. The link between
irrigation, markets and poverty
It has been quite a
while since poverty has posed a serious problem in most developing countries of
the world. During the past two to three decades, a wide array of studies and
researches have been undertaken to understand the root causes of poverty in developing
countries. The results of these studies
indicate that poverty problem in developing countries is complex and
multidimensional and is a result of a myriad of interactions between resources,
technologies, institutions, strategies and actions and others (Hussain, 2003).
It is now well understood that poverty in most developing countries is a result
of lack of resources, information, appropriate institutions and inappropriate
domestic and unfair international policies. In particular, in most developing
countries, poverty is largely a result of low agricultural productivity arising
from very low utilization of modern inputs and technology. Hence, given that
agriculture is largely rain-feed, irrigated water has become very crucial
resource in agricultural production, productivity and poverty reduction.
According to the available
evidences, countries such as East Asia and Pacific and North Africa and Middle
East that have succeeded in poverty reduction have the greatest proportion of
cultivated area irrigated. The poverty-reducing impact of irrigation is
substantial as evidenced in many Asian countries. For instance, about 35-40% of
cropland in Asia is irrigated and poverty reduction in the 1970s was
substantial (Hussain and Hanjra 2003). It should be noted that the
availability of irrigation not only boosts agricultural production but also
make possible the adoption of modern inputs such as improved seeds, fertilizers
and pesticides (Ray, Rao and Subbarao 1988). Note that the transmission mechanisms through which irrigation may lead to poverty
reduction is via increased yields, increased cropping areas and higher value
crops. It also leads to higher employment. The cumulative effect of all these
is that it tends to increase food supplies and raises calorie intakes and
better nutrition levels.
A number of
empirical evidences confirmed the above facts. In many countries, irrigation has been an effective tool for
reducing poverty, increasing cropping intensity, grain production, household
incomes, waged labour employment and livelihood diversification (Angood et al,
2003, 2002;Hussain
et al. 2004; Hussain and Hanjra 2003; Madhusudan et al.2002). Apart from these, there are also stability
effects in agricultural production because of reduced reliance on rainfall.
Farm households who use irrigation will experience lower variability of yield
(reduced climate risk), output and employment compared to those that depend on
rainfall (Lipton et al, undated, Diao et al, 2005; Dhawan, 1988). Comparison of
irrigated versus non-irrigated areas indicate that crop productivity and output
tend to be much higher in irrigated systems than the non-irrigated and
rain-feed areas (Jatileksono and Otsuka 1993; Datt and Ravallion 1998;
Balisacan 1993). Similarly, value of crop production, household income and
consumption are almost double in irrigated settings than the non-irrigated
areas and labour employment and wages are much higher in irrigated areas than
none. In
a comparative study, Hussain et al (2004) indicates that poverty incidence is
about 20-30% higher in rain feed settings than irrigated setting. On the other hand, a study by Haung et al (2005), using a plot-level data
in rural China, indicates that irrigation boosts cropping income and reduces
poverty and inequality.
In general, the results of these econometric studies
indicate that crop output and productivity, farm income, consumption,
employment and rural wages tend to be much higher irrigated areas and
irrigation is a positive and significant determinant of income and consumption
and a negative determinant of poverty. Note that irrigation alone may not lead to poverty reduction. Rather,
the poverty-reducing impact of irrigation will be stronger if it is supported
by use of other yield-enhancing inputs. It is often argued that even though
irrigation with other modern inputs are used to enhance production, this may
not entail the intended result if farm households don’t have access to markets
for their produce. A combination of
irrigation, other modern inputs and access to markets are critical for poverty
reduction and this will eventually lead to accelerate agricultural growth. For
instance, reducing marketing costs primarily benefits smallholders via better
prices for their produce and raises farmers’ income. Moreover, there is also
another effect of improving market conditions: it stimulates the trading
sector, which itself can generate greater non-agricultural income.
3. Modelling the effect of irrigation and
markets on Poverty
In this section, an attempt will be
made to quantify the link between irrigation, markets and household welfare,
measured in terms of consumption per capita. In the process of modelling such
linkage, simulations will also be carried out to examine the impact of some
policy interventions and other socio-economic factors on the well-being of
rural households.
3.1 Econometric models and methods of estimation
The
objective of specifying the model is to assess to what extent irrigation and
market affect the well-being of rural households. To answer questions about the effect of these variables, conditional on
the many other potential determinants of poverty, multivariate analysis is
required (Gibson and Rozelle, 2002; Ravallion, 1998). In this regard, econometric models of the
determinants of poverty where key modern agricultural inputs such as irrigation
and market access variables would be
entered explicitly as an argument in the model.
The usual approach in the multivariate analysis of poverty is to
classify households as poor and non-poor based on consumption per capita (Datt,
1998; Gibson and Rozelle, 2002; Mulat et al,
2003). Denoting the ith household’s per capita expenditure by Ci,
then a household is classified as poor if the ith household’s Ci is less than the poverty line (Z). Accordingly,
a binary variable is constructed to classify households as poor and non-poor.
Then the probit estimation assumes the following functional forms:
[1]
where F is the
standard cumulative normal distribution function, X is a matrix of explanatory
variables such as agricultural technology and market-related variables and
other determinants of consumption, and b is a vector of parameters to be
estimated. However, the probit estimation, as indicated in equation (1), focuses
only on incidence of poverty and ignores the poverty gap and severity. A more
generalized poverty measure for household i can be specified as (Foster, Greer and
Thorbecke, 1984):
[2]
Where 
is the estimated
poverty measure of household i, Z refers to the poverty line and a is a
non-negative parameter taking integer values 0, 1 and 2. It should be noted
that aggregate poverty of a given population is simply the weighted mean of the
above poverty measure, where the weights are given by household size. When a assumes
values of zero, one and two, the aggregate poverty measure corresponds to the
incidence of poverty or head-count index, the poverty gap and squared poverty
gap (which is sensitive to inequality amongst the poor), respectively.
Instead of using poverty probits, the approach of this
paper is to model the determinants of consumption per capita, and then derive
from the regression model estimates of the various poverty measures following
simulated changes in certain variables. More specifically, the model is of
(log) nominal consumption expenditure per adult equivalent, deflated by a
poverty line, which gives a ratio often known as the “welfare ratio” (Gibson
and Rozelle, 2002; Blackorby and Donaldson, 1984) and is given by:
[3]
Where Ci denotes per
capita consumption of household i, Di refers to demographic
characteristics, human capital variables are given by Hi, Fi
denotes farm characteristics, Ai is a matrix of technology-related
variables such as irrigation , Z is the poverty line and vi is a
stochastic term with zero mean (0) and constant variance (σ2).
In a more compact form, equation (3) can be expressed as:
(4)
Where Xi is a matrix of explanatory
variables indicated above. Since the consumption model estimates are independent of the chosen
poverty line, it is potentially attractive to model household consumption
level, and then link it to household poverty level (Mulat et al, 2004). After
normalizing consumption per capita by poverty line, it is possible to classify
households into poor and non-poor, i.e. if the logarithm of the normalized
welfare ratio (ln (Ci/Z)) is less than zero, then a household is
deemed to be poor, otherwise non-poor. The probability of the ith household
being poor can be derived from the estimated parameter
and standard error 
of the regression.
Formally, the probability of the ith household’s
logarithm of welfare ratio being less than zero is given by:
(5)
Equation (5) gives the weighted average of the
predicted incidence of poverty for the ith household (P0,i) where the weights are household sampling
weights in terms of adult equivalent household size. Similarly, the methodology
can easily be extended to derive the simulated poverty gap denoted by (P1,
i) and poverty severity (P2, i) as:
(6)
(7)
Equations (5), (6) and (7) are
employed to generate predictions of poverty following various policy simulation
exercises.
3.3 Description of explanatory variables of the model
The set of variables that is
hypothesized to determine the level of consumption, and hence poverty, may be
categorized into:(a) household characteristics; (b) human capital; (c) farm
charactersitics; (d) access to market and modern technology. Among the set of
potential determinants of poverty, an attempt is made to choose those variables
that are arguably exogenous to current consumption.
(a) Household characteristics: This includes household size, age
and sex of household head. In order to take into account non-linearities in the
relationship between consumption and household size, a quadratic term has been
introduced in the regression model.
(b) Human capital: Included in this category are
literacy of household head and years of schooling for adults.
(c) Farm characteristics: include holding size and quality
indicator of land. The number of plots (as a proxy for the degree of crop
diversification) has also been included in the model. It should be noted that
the number of plots indicates the land covered by different crops, hence serves
as a proxy for crop diversification.
(d) Access to modern technology and markets: with regard to market access variables, distance to the largest buyer
(output market), distance to the most important input supplier (input market),
and the proportion of sales from the total output are included. Similarly, a number of variables have been
identified that reflect use of modern agricultural technology: Irrigation
practices and experience, the proportion of irrigated land, soil conservation
and water harvesting practices.
3.4 Estimation of the model
3.4.1
Description of dataset
The dataset
used in the estimation of the model is obtained from a household survey in two
woredas of North Omo zone: Arbaminich and Mirab Abbaya woredas in the Southern
Nations, Nationalities and Peoples (SNNP) region. These woredas are the major producers of
fruits and vegetables. The woredas have been
purposively selected from the woreda Agricultural Office since the focus of the
study is on cash crop producers using irrigation. A list of Peasant Associations (PAs) that
mainly produce fruit and vegetable using irrigation was obtained from the
woreda Agricultural Office and then households have been randomly selected from
those PAs. Accordingly, a total of 216 households have been included in the survey. The survey
provides data on a wide spectrum of socio-economic variables including
household composition and structure, education, use of modern technology,
household assets, employment and income, consumption expenditure (both food and
non-food), health status and other welfare indicators. More importantly, the questionnaire included a module which
is designed to capture plot-level information such as whether a plot is
irrigated, the area of irrigated land, type of crop grown on a plot, crop
yield, land quality and slope of land. In addition, a market participation
module has been included in the questionnaire, which intends to capture key
market variables.
3.4.2 Results and Discussions
Descriptive statistics
(a)
Household demographics and Farming characteristics
Before going directly to the model results, it
is important to give some basic background information regarding the sample
households. Of the sample households, the majority (90%) are male-headed where
only 10% are female-headed households.
Farming provides the primary source
of livelihood for the sample households. The average holding size is about 1.1ha
for the sample households.
This means that, with an average family size of six persons, per capita holding
size would be about 0.18ha in the study area. 48.4% of farm households have
less than a hectare of land while 18.4% have landholding size greater than
1.5ha (Table 3.1). As for the farm
characteristics, about 77.1% reported that their farmland is of fertile quqlity
while medium quality is indicated in 19.2% of the cases. Only 3.7% reported
that the land is of poor quality.
It seems that, on average, the land is suitable for agricultural production in
the study area. Note also that the majority of households have flat farmland,
i.e. less steep and hence not susceptible to soil erosion. 97% of the sample households reported that
soil erosion is not a problem in the village.
Table 3.1: Distribution of landholding size
|
Description
|
Number of households
|
% of
households
|
|
Landholding size
<0.50
|
38
|
18.4
|
|
0.50-1.0
|
62
|
30.0
|
|
1.0-1.5
|
69
|
33.3
|
|
1.5-2.0
|
20
|
9.7
|
|
>=2
Land quality
Leum (Fertile)
Leum-teuf (Medium)
Teuf (Not fertile)
|
18
165
41
8
|
8.7
77.1
19.2
3.7
|
Source:
Own calculation from survey data
As for the education level of
household heads, about 49.3% don't read and write while 42.7% have some primary
education. Only 7.5% of the sample household heads have completed secondary
education (Table 3.2). Almost all female-headed households are either
illiterates (90%) or have some primary education (10%).
|
Table
3.2: Education level of household head by gender
|
|
Education
category
|
Male
|
Female
|
Total
|
|
Illiterates
|
45.08
|
90.00
|
49.30
|
|
Primary education
|
46.11
|
10.00
|
42.72
|
|
Secondary education
|
8.29
|
|
7.51
|
|
Post-secondary education
|
0.52
|
|
0.47
|
|
Total
|
90.61
|
9.39
|
100.00
|
Source: Computed from survey data
With regard to the use of modern
inputs, it is indicated that 3.6%, 81.3% and 17.6% of the sample households use
chemical fertilizers, improved seeds and other chemicals such as pesticides,
respectively (Table 3.3). It should be noted that about 90% of the cultivated
land is irrigated. The average irrigation experience of the sample households
is 13 years, indicating that irrigation has been practised quite a long period
in the study area. More than half of the sample farm households have more than
15 years of experience in irrigation. River or stream diversion is the main
source of water for irrigation, and pump irrigation is nearly non-existent.