University of Plymouth

PEARL https://pearl.plymouth.ac.uk
Faculty of Science and Engineering School of Geography, Earth and Environmental Sciences

2015-12-15

Micro-determinants of income Inequality

and consumption in rural Bangladesh

Rahman, S
http://hdl.handle.net/10026.1/3953

Journal of Poverty Alleviation and International Development

Yonsei University, South Korea

All content in PEARL is protected by copyright law. Author manuscripts are made available in accordance with
publisher policies. Please cite only the published version using the details provided on the item record or
document. In the absence of an open licence (e.g. Creative Commons), permissions for further reuse of content
should be sought from the publisher or author.
 Journal of Poverty Alleviation and International Development, 6(2)
©2015 The Author. Published by the Institute for Poverty Alleviation and International Development
under open access license CC BY-NC-ND 3.0.

Micro-determinants of Income Inequality and Consumption

in Rural Bangladesh*
1

Sanzidur RAHMAN**

Abstract: The paper examines the extent to which household and

regional characteristics influence income inequality and
consumption/welfare based on an in-depth survey of 406
households from 21 villages in three regions of Bangladesh.
Results show that the overall Gini coefficient for rural incomes is
0.43 but Gini-decomposition revealed that the contribution of
mixed crop production to inequality is just 10 percent while “Green
Revolution” technology contributes almost 29 percent. Land
ownership, farm capital assets, modern irrigation, non-agricultural
income, and household head’s education significantly increase
consumption. Tenants and households with more dependents are
doubly disadvantaged and consume significantly less. Regional
factors also significantly influence inequality and consumption.
Consumption is significantly higher in regions with developed
infrastructure. Comilla is the region with the highest level of
inequality and a significantly lower level of consumption. Thus an
integrated policy of investments in modern irrigation, crop

* This is an open-access article distributed under the terms of the Creative

Commons license Attribution-Noncommercial-NoDerivs 3.0 Unported.
Distribution and reproduction are permitted, provided that the original author
and JPAID as the source are credited. The author gratefully acknowledges
critical and thoughtful comments from two anonymous referees and the editor
which have improved the paper substantially. However, all caveats remain
with the author.
** Dr. Sanzidur Rahman is Associate Professor (Reader) in Rural Development
with the School of Geography, Earth and Environmental Sciences at the
University of Plymouth in the UK. Email: srahman@plymouth.ac.uk
jpaid.yonsei.ac.kr
108 Journal of Poverty Alleviation and International Development

diversification, tenancy reform, mass education and rural

infrastructure is necessary to increase consumption/welfare and
reduce income inequality in Bangladesh.

Keywords: Income inequality, Gini-decomposition analysis,

Consumption or welfare determinants, Bangladesh

Although “eradication of poverty and hunger” has been the main

theme of development in the 2000s, the goal remains elusive. However,
progress in reducing poverty has been impressive and widespread. The
proportion of poor living below the international poverty line (i.e., living
under USD 1.25 a day) has fallen to 25.7 percent (or 1.4 billion persons)
in 2005 from more than 50 percent in 1981 (or 1.9 billion persons).
Nevertheless, there are still large numbers of people living under the
poverty line in Sub-Saharan Africa (51 percent), South Asia (40 percent)
and East Asia (17 percent) (Krishna 2013), and these figures include two
newly emerging economic powerhouses: India and China. Moreover,
Krishna (2013) among others argues that policies which were successful
in reducing poverty in the past have lost their effectiveness and a
business as usual approach is not going to reduce poverty any further.
The importance of non-agricultural income in supporting the
livelihoods of rural households in developing countries has been
increasingly recognized over the past three decades (e.g., Smith et al.
2001; Deininger & Olinte, 2001; Davis 2004; Hatlebakk 2012). Rural
households are commonly involved in diverse income generating
activities in order to cope with adverse factors in agriculture (e.g., Ellis
2000; Barrett et al. 2001; Deininger & Olinte 2001; Ellis & Freeman
2004). However, the influence of such diversified livelihood portfolios
on inequality is not well known or understood.
Bangladesh is a predominantly agrarian economy in which a large
proportion of the population are vulnerable to malnutrition and hunger.
Improvements in food security have relied on the extensive use of a
 Micro-determinants of Income Inequality and Consumption in Rural Bangladesh 109

rice-based Green Revolution technology (i.e., high yielding varieties of

seed, inorganic fertilizers and supplementary irrigation technology) to
feed the fast growing population. Consequently, over the past five
decades, the major policy focus has been directed towards diffusion of
Green Revolution technology aimed at meeting a tripartite objective of
increasing food production, generating employment and increasing the
income of rural households, all of which complement the national goal of
achieving self-sufficiency in foodgrain production and poverty
alleviation. It is worth noting that Bangladesh has made considerable
progress in improving the wellbeing of its population in recent years.
Nevertheless, poverty is still high; 31.5 percent of the population are
living below the poverty line (Bangladesh Bureau of Statistics (BBS)
2011).
The degree to which the policy of widely diffusing modern
agricultural technology has been successful is contentious in the
literature. Several earlier studies based on large-scale sample surveys, for
example, Hossain (1989) and Hossain and Sen (1992) find that poverty
and inequality is relatively lower in villages with a higher rate of
adoption of modern agricultural technology in Bangladesh. However,
these studies do not provide evidence of the effects of modern
technology inputs specifically. Other country studies attempt to identify
the effect of individual inputs. Thapa et al. (1992) showed that the
adoption of new technology did not significantly worsen the distribution
of income. Rather, it was found to substantially increase the rate of return
on land in Nepal. On the other hand, Rahman (2009) noted that the
adoption of modern agricultural technology does not seem to have any
significant influence on poverty in Bangladesh. In contrast, Benson et al.
(2005) noted that agriculture (whether modern or traditional was not
specified) is positively associated with poverty in Malawi. Freebairn
(1995), after conducting a meta-analysis of 300 studies undertaken
during the period 1970–1989, revealed that about 80 percent of these
studies concluded that modern agricultural technology widened both
inter-farm and inter-regional income inequality. However, he further
110 Journal of Poverty Alleviation and International Development

noted that the authors from Asian countries, who used research data
collected in the Philippines and India, concluded that increasing
inequality is not associated with modern agricultural technology.
Understanding the precise causes and extent of poverty and
inequality has been a major concern of policy makers for some time, and
while regional differences exist, research has focussed on a fairly
constant set of variables. Particular household characteristics such as
education, land ownership, demography and potential sources of income,
as well as regional factors such as level of infrastructure development,
soil quality and fertility and location, are generally accepted as important
influences determining poverty and inequality but with mixed
influences. For example, Achia et al. (2010) noted that although
education significantly reduces the probability of being poor, rural
households are more likely to be poor as compared to their urban
counterparts in Kenya. They also noted that demographic factors such as
age, religion, ethnicity and region influence poverty. Rahman (2009)
noted that while land ownership, farm resource endowments and
non-agricultural income significantly reduce the probability of becoming
poor, the number of dependants and education of female members has an
opposite effect in Bangladesh. Benson et al. (2005) noted the positive
influence of education in reducing poverty in Malawi but highlighted that
the relationship is somewhat more complex than generally understood.
They emphasized increasing access to district level services to address
poverty. In contrast, Anyanwu (2005) noted that household size, primary
or lower level of education and rural occupation are positively associated
with poverty in Nigeria. He also noted feminization of poverty in Nigeria
and differential influence of geographic location on poverty. Wodon
(2000) noted that education, demographics, land ownership, occupation,
and geographic location, all affect poverty in Bangladesh. He further
noted that education influences inequality in urban areas while land
influences inequality in rural areas. Ravallion and Wodon (1999) also
noted that education and regional factors exert significant influence on
poverty in Bangladesh.
 Micro-determinants of Income Inequality and Consumption in Rural Bangladesh 111

In sum, previous studies have made clear that modern agricultural

technology, demographic, and socio-economic and geographical factors
all exert variable influences on poverty. Furthermore, most of the
aforementioned studies examined the influence of these various factors
on poverty and/or the probability of being poor. Only Wodon (2000) and
Ravallion and Wodon (1999) investigated the influence of a wide range
of demographic and socio-economic factors on consumption/welfare of
Bangladeshi households in urban and rural areas. However, they did not
take into account specific impacts of modern agricultural technology
adoption and/or cropping portfolio on consumption/welfare of
households.
Accordingly, the main objectives of this study are to: (a) estimate
the level of income inequality of farming households in selected regions
of Bangladesh; (b) identify the sources of income inequality of these
farming households; and (c) identify the determinants of consumption/
welfare of these farming households. We do so by using an in-depth
farm-level sample survey of 406 households from 21 villages in three
agro-ecological regions in Bangladesh for the year 1996.
The contributions of our study to the existing literature are as
follows. First, it is generally regarded that the value of consumption is a
relatively better measure of capturing a household’s financial situation
than earned income in the context of developing countries because of
co-existence of cash and in-kind transactions, lack of record keeping of
expenditure and income accounts, and difficulty in deriving net incomes
from petty trading and/or business transactions. In this study, we use both
type of measures, i.e., income as well as consumption expenditure to
address our specified objectives, which is not commonly found in the
literature. Second, for countries such as Bangladesh where technological
progress in agriculture is deemed to be a pre-requisite for economic
growth and development, detailed information on the extent and
influence of the adoption of modern agricultural technology is also
crucial to any study of inequality.
The paper is organized as follows. The next, second, section
112 Journal of Poverty Alleviation and International Development

presents the methodology and describes the data including

region-specific income patterns of the households that make up the
sample. The third section presents the results which include contribution
of modern agricultural technology to income inequality using Gini
decomposition analysis, and identification of the demographic,
socio-economic and regional/spatial determinants of consumption/
welfare of these households using a multivariate regression analysis. The
final section concludes and draws policy implications.

Methodology

Measures of income inequality and its sources: A Gini-decomposition

analysis

One of the most common measures of inequality in income

distribution is the Gini-coefficient, which is based on the Lorenz curve.
Moreover, the Gini coefficient has a unique underlying social welfare
function that is based on the rank of individuals (Makdissi & Wodon
2012). Also, the popularity of the Gini as a measure of inequality is that it
can be decomposed by source of income and/or classes (such as region).
Some of the best-known papers on Gini decomposition are Rao (1969),
Pyatt (1976), Fei et al. (1978) and Pyatt et al. (1980). Yao (1997) noted
that the covariance method for Gini decomposition is not appropriate for
unevenly grouped populations and proposed an alternative
decomposition approach that is exact. In this formulation, the
Gini-coefficient for measuring income inequality is given by (Yao 1997):

(1)
with the relation
 Micro-determinants of Income Inequality and Consumption in Rural Bangladesh 113

where, n = number of income groups,

mi=mean income of group i (i–1,2,…n),
m = mean income of the entire population,
pi=population share of group i,
wi=income share of group i in total income.
Qi=cumulative income share from group 1 to group i with pi
and wi following an ascending order of mi (m1≤ m2≤ …≤ mn).
If per capita total income is decomposed into F components, then
the Gini coefficient for component income is given by:

(2)
with the relation

where, n = number of income groups,

mfi = mean component income of group I (i=1,2,…n),
mf = population mean income of component f (f=1,2,…,F),
pi = population share of group i,
wfi = income share of group i in total income of component f.
Qfi = cumulative income share from group 1 to group i with
pi’s and wfi’s following an ascending order of mfi’s (mf1≤mf2≤….≤mfn).
Equation 3 can also be used to calculate the component
concentration ratio if pi’s and wfi’s follow an ascending order of group
mean total income mi’s instead of group mean component income mfi’s as
shown below:

(3)
with pi’s and wfi’s following m1≤m2≤…≤mn.
Substitution of equation 3 in 1 provides the decomposition of Gini
114 Journal of Poverty Alleviation and International Development

coefficient as:

(4)
with the relation

Equation 7 indicates that Gini coefficient is the weighted average of

component concentration ratios. The examination of how each individual
component contributes to total income inequality is given by:

(5)
where, gf = is termed as relative concentration coefficient.
wfgf = share of component f in the Gini coefficient G.
An income component is said to be inequality increasing if its gf>1.
The implication is that if total income remains unchanged, the increase in
the share of a single component will result in an increase in overall
income inequality and vice-versa. In this decomposition method the
relative contribution of an income component to the Gini coefficient will
depend on its share in total income wf as well as on the value of gf.

Microdeterminants of consumption: A multivariate regression model

In the previous section, we have developed measures to compute

income inequality and relative contribution of various income sources to
inequality. But what are the determinants of per capita consumption, an
alternative robust measure of inequality that can be analysed using
information at the individual household level whereas the Gini index is
an aggregate measure which can only be computed for a group of
 Micro-determinants of Income Inequality and Consumption in Rural Bangladesh 115

households or region?
We have used “welfare ratios” to measure consumption defined as
the household’s per capita consumption normalized by the appropriate
regional poverty line so that differences in costs of living between
regions are taken into account (Wodon 2000). A welfare ratio equal to
one indicates that the household has consumption at the level of the
poverty line. In other words, if the welfare ratio is below unity, then the
household is deemed to be poor. In the conventional measure, binary
variables are used to define the poor, which takes the value of 1 for
households whose income falls below the poverty line expenditure and 0
otherwise. Such a measure cannot take into account how far a household
is below the poverty line expenditure. Our measure is providing a
continuous measure of the extent of poverty of the household relative to
the poverty line expenditure. A value above unity implies that the
household is above the poverty line and shows the extent to which it is
above the poverty line expenditure.
Following standard practice, we use the semi-log specification
since per capita consumption/welfare ratio (the dependent variable) is
not normally distributed (Wodon 2000). The following regression model
is specified:

(6)

The dependent variable is the log welfare ratios, i.e., log of nominal
per capita consumption divided by the poverty line of the area in which
the household lives (Wodon 2000). The Xi is the vector of regressors, β
is the vector of parameters to be estimated, and µi is the error term.
STATA V10 software is used to estimate the model (StataCorp 2010).

Data

Primary data for the study came from an intensive farm-survey

conducted from February to April, 1997, in three agro-ecological regions
116 Journal of Poverty Alleviation and International Development

of Bangladesh. Twenty-one villages were included, eight from the

Central sub-district of Jamalpur representing wet agro-ecology areas, six
from the Manirampur sub-district of Jessore representing dry
agro-ecology areas, and seven from the Matlab sub-district of Comilla,
representing wet agro-ecology and agriculturally developed areas. A
total of 406 farm households (175 in Jamalpur, 105 in Jessore and 126 in
Comilla) were selected from these villages using a multistage stratified
random sampling procedure.
To identify the sources of income inequality and consumption
determinants, a number of variables were constructed, grouped as
follows: sources of income to the household; production inputs;
household characteristics; and two indices, one to capture soil quality
and the other the degree of local infrastructure development. Regional
dummies are also included.
Income: Household or family income is defined as the return to
family labor, plus those assets owned after the current cost of production
(excluding rent for land and assets) is deducted from the gross value of
production (Ahmed & Hossain 1990). Current costs are those incurred by
individual households in purchasing inputs, hiring labor and animal
power services, and renting services (details of components of income
and their derivation is presented in the Appendix). Income from
agriculture is separated into income from various crops, fisheries,
livestock and lease income from land. This is reported in Table 1. Crop
income is the aggregate of that derived from local and modern varieties
of rice (all season), wheat, jute, potatoes, pulses, spices, oilseeds,
vegetables and cotton. Modern rice varieties account for more than 60
percent of total crop income, while other crops including local varieties
of rice contribute very little. For all crops there are sharp inter-regional
variations. As indicated in Table 1, Jessore, the region with the more
diversified cropping system has the highest income, although this is not
the region with the greatest share of modern varieties (Jamalpur for rice
and Comila for wheat). Nonetheless, it is clear that for two regions, field
crop income is overwhelmingly the dominant source of total household
 Micro-determinants of Income Inequality and Consumption in Rural Bangladesh 117

income and in only one region does non-agricultural income appear to be

significant.

Table 1.
Structure of Annual Total Income (BDT) Per Household
Source of income Share of component incomes to total income by region (%)
Jamalpur Jessore Comilla All regions
Total agricultural income 84.3 73.3 50.2 72.3
Crops 53.3 44.7 34.6 45.9
Traditional rice 3.9 0.3 3.2 2.6
Modern rice 44.5 26.6 22.6 33.3
Modern wheat 0.6 1.9 2.8 1.6
Jute 0.8 4.8 1.4 2.2
Potato 0.9 0.3 2.9 1.2
Pulses 0.0 3.9 0.2 1.3
Oilseeds 0.1 1.5 1.0 0.8
Spices 2.2 0.1 0.6 1.1
Vegetables 0.3 3.6 0.0 1.3
Cotton 0.0 1.6 0.0 0.5
Livestock 14.9 17.3 13.4 15.3
Fisheries 5.5 5.1 1.1 4.3
Lease 10.6 6.2 1.1 6.8
Total non-agricultural income 15.7 26.7 49.8 27.7
Wage 4.6 3.7 8.6 5.3
Business and other 11.1 23.0 41.2 22.4
Total family income 100.0 100.0 100.0 100.0
(31,581) (39,064) (25,314) (31,571)
Note. Figures in parentheses represent total income per household. Exchange rate USD 1.00 =
BDT 81.86 (Bangladesh Bank 2012). Source is Field Survey, 1997.

Poverty line expenditure: The cost of basic needs (CBN) approach

is used to construct the region-specific poverty line expenditure (Wodon
2000, 1997; Ravallion & Sen 1996). In constructing the food poverty
expenditure as a first step, a cost-minimizing long-term diet set with
118 Journal of Poverty Alleviation and International Development

available food items that attain the recommended nutrition level of 2,112
kcal and 58 grams of protein per capita per day proposed by Mian (1978)
is utilized. In addition, expenditure on non-durable goods and/or
non-food allowance is estimated at 30 percent of the food poverty line (a
standard practice in the context of Bangladesh, e.g., Hossain (1989),
Ahmed and Hossain (1990). The region-specific poverty line
expenditures are different across regions with an overall estimate of BDT
5,409 per capita per year (see Table 2).
Consumption: Consumption expenditure was constructed using the
following procedure. First, quantities of food items consumed (both
purchased and home supplied) during seven days prior to the date of
interview were converted into values using market prices within the
village and then multiplied by 52 weeks to compute annual expenditure
on food. Next, monthly expenditure on durable goods, such as,
dress/clothing, education, transportation and debt servicing were
collected and multiplied by 12 months to compute annual expenditure.
Finally, annual expenditure on investment, maintenance of properties,
social and religious works was also collected. All these consumption
expenditure items were summed and then divided by household size to
obtain total nominal consumption expenditure per capita per year. The
region-specific actual consumption expenditure, thus constructed, are
significantly different across regions (F-statistic = 9.83; p<0.01) with an
overall estimate of BDT 6,068.6 per capita per year (see lower part of
Table 2).
 Micro-determinants of Income Inequality and Consumption in Rural Bangladesh 119

Table 2.
Poverty Line Expenditure Required to Fulfill Nutritional and Other
Requirements
Food item Qty. (gm) of food Cost of attaining the optimal diet evaluated at
included in optimal region-specific retail market prices (BDT)
diet Jamalpur Jessore Comilla All regions
Rice 432.6 4.90 4.36 4.46 4.62
Wheat 58.3 0.64 0.58 0.64 0.62
Potato 36.7 0.15 0.14 0.15 0.14
Lentil 25.0 0.53 0.54 0.53 0.53
Fish 38.3 2.11 2.43 2.24 2.24
Meat 1.7 0.11 0.13 0.11 0.12
Milk 31.1 0.48 0.43 0.53 0.50
Dry milk 2.5 0.55 0.55 0.55 0.55
Sugar 27.2 0.70 0.70 0.70 0.70
Oil 12.2 0.70 0.69 0.63 0.68
Onion 8.5 0.09 0.07 0.07 0.08
Non-leafy vege. 86.8 0.38 0.58 0.52 0.53
Leafy vegetable 20.0 0.09 0.09 0.10 0.09
Cost of food per capita per day 11.43 11.29 11.23 11.40
Annual cost of food 4,172.0 4,120.9 4,099.0 4,161.0
Annual cost of non-food items 1,251.6 1,236.3 1,229.7 1,248.3
Poverty line expenditure per year per 5,423.6 5,357.2 5,328.7 5,409.3
capita
Actual estimated consumption 6,453.6 6,648.6 5,050.3 6,068.6
expenditure per year per capita
Welfare ratios (i.e., consumption 1.21 1.24 0.94 1.14
expenditure /poverty line expenditure)
Note. Exchange rate USD 1.00 = BDT 81.86 (Bangladesh Bank 2012). Values are extended from
Rahman (2009).

Production inputs: Two land variables were used in the analysis.

The amount of land (in hectares) owned per capita is an indication of the
wealth of the family, while the area of land under cultivation or farm
operation size per household is a direct production input which is made
up of owned, rented-in and/or mortgaged-in land. In addition, the value
of farm capital assets is included which also serves as production input
120 Journal of Poverty Alleviation and International Development

because it includes the value of livestock resources owned, which is a

major source of draft power in farming. Use of land as a determinant is
abundant in the literature (e.g., Rahman 2009; Benson et al. 2005;
Anyanwu 2005; Wodon 2000; Thapa et al. 1992). Use of farm capital
asset is, however, less common in the literature (e.g., Rahman 2009).
Both of these variables are assumed to positively influence
consumption/welfare.
Cropping portfolios and level of modern technology adoption:
These are the proportion of area irrigated (also a major production input
particularly for producing high yielding varieties of rice) and the level of
cropping diversity. As mentioned earlier, the use of modern agricultural
technology to identify its influence on poverty and inequality are
important but the results are mixed in the literature (Rahman 2009;
Freebairn 1995; Thapa et al. 1992). It is expected that adoption of
modern agricultural technology will be income neutral and/or
consumption/welfare enhancing.
Household characteristics: These variables were the number of
non-working dependents, years of formal education of the head of
household and highest education level of any male members, and age of
the head (a proxy variable representing experience). Use of these
demographic and socio-economic factors is most common in poverty
studies (e.g., Rahman 2009; Benson et al. 2005; Anyanwu 2005; Wodon
2000), though there is no consensus on their influences on poverty and
inequality. We have used two separate indicators of head’s education and
highest education of any male members in the household in order to
identify the existence of centralized decision making (Asadullah &
Rahman 2009) and their corresponding influence on consumption/
welfare. We assume that these variables will be consumption enhancing.
Regional/spatial factors: Two indices to capture the influence of
spatial factors were included in the analysis. These are: soil fertility and
state of infrastructure. The soil fertility index is constructed from test
results of soil samples collected from representative locations during a
field survey for crop year 1996. Ten soil-fertility parameters were tested,
 Micro-determinants of Income Inequality and Consumption in Rural Bangladesh 121

namely soil pH, nitrogen, potassium, phosphorus, sulfur, zinc, texture,

exchange capacity, content of organic matter and electrical conductivity.
Each of these positive characteristics was assigned 1 point and thus a
high index value implies better soil fertility.1 The infrastructure index
was constructed using the cost of access approach. Here thirteen
elements are included, namely primary markets, secondary markets,
storage facilities, rice mills, paved roads, bus stops, banks, union offices,
agricultural extension offices, high schools, colleges, sub-district
headquarters and post offices. A high index value implies poorly
developed infrastructure.2 Use of these two indices, although deemed
quite important, is not usually reported in the literature with a few
exceptions (e.g., Rahman 2009; Anyanwu 2005). We assume soil fertility
status and developed infrastructure to have consumption/welfare
enhancing effects.
Although the data collected for this study are 18 years old, little has
changed with regard to the farming practices and operating institutions
over this period in Bangladesh, except for an increase in the level of
modern rice technology adoption from 38.6 percent of gross cropped
area in 1990 to 62.9 percent in 2011 (Ministry of Agriculture 2008; BBS
2012). Therefore, we argue that our results are capable of providing
valuable information of relevance to policy makers and development
practitioners alike.

Results

Level of income inequality and its sources

The Gini-coefficients (G and Gf), income shares (wi and wfi),

component concentration ratio (Cf), relative concentration ratio (gf) and

1 For details of the construction procedure of the soil fertility index, see
Rahman (1999).
2 For details of the construction procedure of the infrastructure index, see
Rahman (1999) and Ahmed and Hossain (1990).
122 Journal of Poverty Alleviation and International Development

inequality weights (wfgf) for the income components modern agricultural

technology income, other field crop income, non-crop agricultural
income, and non-agricultural income classified by region are presented
in Table 3.

Table 3.
Measures of Income Inequality and Its Sources
Income Per capita Share in total Gini Concentration Contribution Relative Inequality
income income coefficient ratio to total Gini concentration weight
components % ratio %
mf wf G and Gf Cf wfCf gf=Cf/G wfgf
Comilla region
MV technology 1003.5 23.9 0.512 0.156 0.037 0.340 8.1
Other crop 413.7 09.8 0.757 0.409 0.040 0.889 8.7
Non-crop agri. 658.0 15.7 0.513 0.278 0.044 0.604 9.5
Non-agriculture 2129.2 50.6 0.750 0.670 0.339 1.457 73.8
Total income 4204.4 100.0 0.460 0.460 100.0
Jamalpur region
MV technology 2837.5 45.4 0.464 0.351 0.159 0.887 40.2
Other crop 442.3 07.1 0.746 0.329 0.023 0.829 5.9
Non-crop agri. 2030.0 32.4 0.599 0.463 0.150 1.167 37.9
Non-agriculture 946.8 15.1 0.820 0.425 0.064 1.075 16.3
Total income 6256.6 100.0 0.395 0.395 100.0
Jessore region
MV technology 1761.6 27.6 0.514 0.353 0.097 0.873 24.1
Other crop 950.8 14.9 0.646 0.457 0.068 1.131 16.8
Non-crop agri. 1835.8 28.8 0.562 0.398 0.114 0.986 28.3
Non-agriculture 1837.2 28.8 0.662 0.431 0.124 1.066 30.7
Total income 6385.4 100.0 0.404 0.404 100.0
All regions
MV technology 1990.1 35.2 0.534 0.354 0.125 0.817 28.8
Other crop 564.9 10.0 0.737 0.442 0.044 1.019 10.2
Non-crop agri. 1554.0 27.5 0.617 0.459 0.126 1.060 29.1
Non-agriculture 1544.0 27.3 0.757 0.507 0.139 1.170 32.0
Total income 5653.05 100.0 0.433 0.433 100.0

Analysis of the Gini coefficients reveal that the degree of income

inequality is highest in Comilla (0.46), lower in Jessore (0.404) and
 Micro-determinants of Income Inequality and Consumption in Rural Bangladesh 123

lower still in Jamalpur (0.395). Estimates of Gini indices, based on

in-depth farm surveys, are 0.35 for the year 1982 (Hossain 1989) and
0.35–0.37 for the year 1987 (Hossain et al. 1990). The current estimates
indicate that inequality has increased over this time period. However,
Wodon (2000), using Bangladesh Household Expenditure Survey
1995/96 data, reported an overall rural Gini index of 0.26 instead.
The impact of modern agricultural technology adoption on income
distribution is complex. It is evident from Table 3 that the contribution of
modern agricultural technology to income inequality is substantial and
accounts for 28.8 percent of total inequality (last column) for the regions
in aggregate. However, the regions vary in the rate of adoption of these
technologies and this is reflected in the extent to which the inequality
enhancing effect is apparent. In Jamalpur, where the share of modern
technology income is very high (45.4 percent), the contribution to
income inequality is also high at 40.2 percent. In comparison, the other
two regions have adopted modern technologies on a small scale, and
subsequently show a lower impact on their inequality weighting.
However, it is encouraging to note that, generally, modern agricultural
technology is inequality decreasing as shown by the relative
concentration ratio (column 6). This term for the aggregate sample is
0.817 (< 1.00) while other field crop income is neutral (1.019 ~ 1.00).
Non-crop agricultural income and non-agricultural income is inequality
increasing (1.06 and 1.17 > 1.00) as expected. The implication is that the
promotion of modern agricultural technology as well as diversified
cropping patterns will reduce income inequality relative to non-crop
agricultural and non-agricultural income sources, given that total income
remains unchanged.

Microdeterminants of consumption/welfare: A multivariate regression

analysis

In the previous section, we have examined the level of inequality

and relative contribution of various income sources to inequality. In this
124 Journal of Poverty Alleviation and International Development

section, we examine the determinants of per capita consumption/welfare

using the multivariate regression model (Eq. 6).
The explanatory variables included in the model are: per capita land
owned (ha); total farm operation size (ha); value of farm capital assets
(‘000 BDT); modern irrigation (share of cultivated area under modern
irrigation); tenancy dummy (1 if tenant, 0 otherwise); Herfindahl index
of crop diversity (number); number of dependents in the household
(persons); share of non-agricultural income in total income (proportion);
education level of household head (completed years of schooling);
maximum education level of any male member in the household
(completed years of schooling); age (years); index of infrastructure
underdevelopment (number); index of soil fertility (number); and
regional dummies (Comilla and Jessore). Choice of these variables is
based on the literature discussed at length in the Methodology section
above.
Result shows that the welfare ratios or consumption expenditures
are also significantly different across regions (F-statistic 8.89; p<0.01)
with an average of 1.14 while for Comilla region, the figure is below one
(see last row of Table 2). These welfare ratios are comparable to the one
reported by Wodon (2000) estimated at 1.08–1.29 for the rural sector in
1995–1996 based on Household Expenditure Survey data.
The results of the Ordinary Least Squares regression with robust
standard error are in Table 4. A large number of coefficients are
significant at the one, five or ten percent level. The F-statistic further
indicates statistically that these variables contribute significantly as a
group to the explanation of the determinants of consumption of rural
farm households. About 45 percent of the variation in consumption is
explained by these characteristics variables as indicated by the value of
the Adjusted R-squared. The null hypothesis of no influence of
socio-economic factors jointly on consumption/welfare is strongly
rejected at the one percent level. Similarly, the null hypothesis of no
influence of geographic/spatial factors jointly on consumption/welfare is
also strongly rejected at the one percent level. Coefficients in Table 4 can
 Micro-determinants of Income Inequality and Consumption in Rural Bangladesh 125

only provide direction of the influence but not the correct magnitude of
influence. Therefore, to obtain a measure of change in
consumption/welfare with respect to changes in the characteristics
variables, consumption/welfare elasticities at the sample means are
estimated and reported in Table 5.
In general, studies investigating the influence of land resources,
whether measured as owned or simply land under cultivation, find a
positive effect on income levels or input demand. In this study, two
variables are used to represent land resources. Intuitively, the major land
factor determining income levels and hence consumption will be land
owned. Results reveal that land ownership significantly increases
consumption/welfare and is the most dominant variable consistent with
our expectation. The elasticity value is estimated at 0.14 implying that a
one percent increase in land owned per capita will increase consumption
by 0.14 percent. Wodon (2000) also reported significant influence of land
ownership on consumption in rural areas. Although farm size has no
influence on consumption, farm capital assets significantly influence
consumption, again consistent with expectation. The implication is that
wealth, measured in terms of land ownership and/or capital assets,
significantly enhances consumption. This is expected in a land scarce
country like Bangladesh where land is a major source of wealth.
Modern irrigation, a major pre-requisite and input for modern
agricultural technology, also significantly increases consumption as
expected and is the second strongest determinant amongst the
socio-economic factors after land ownership, with an elasticity value of
0.07. This is because irrigation opens up opportunities to adopt modern
rice technology which provides significantly higher yields than
traditional rice, and therefore enhances consumption.
Return to education of the head on consumption is also significant
and is the third most dominant variable with an elasticity value of 0.04.
The coefficient on the maximum level of education of any member in the
household is also positive and significant at the 15 percent level. Wodon
(2000) also reported significant returns to education on consumption.
126 Journal of Poverty Alleviation and International Development

Similarly, Achia et al. (2010) and Benson et al. (2005) noted a significant
influence of education on reducing poverty. The implication is that
education for members of the household is positively associated with
enhancing consumption. The combined influence of these two education
variables is 0.09 which means that a one percent increase in educational
attainment at the household level will increase consumption by 0.09
percent.
Non-agricultural income source also significantly increases
consumption as expected and is consistent with the literature. A one
percent increase in non-agricultural income will increase consumption
by 0.03 percent. Rahman (2009) also noted a significant poverty
reduction influence for non-agricultural income.
Tenants have a significantly lower level of consumption.
Bangladesh is an economy where functionally landless households
account for over 50 percent of agricultural production units. Since the
land rental market is not very effective, generation of income through
farming is difficult which ultimately affects consumption adversely.
Rahman (2009) also noted that tenants are more likely to be poor.
The number of dependents in the household significantly reduces
consumption and the effect is highest with an elasticity value of –0.28. A
higher number of dependents exerts pressure on the household with
respect to consumption of goods and services which are to be provided
by fewer earners. The finding is consistent with those reported in the
literature (e.g., Anwanyu 2005).
The regional/location factor also exerts significant influence on
consumption. There is a pronounced positive influence of rural
infrastructure on consumption. Consumption is significantly higher in
developed regions. The elasticity value indicates that a one percent
increase in the index of rural infrastructure will increase consumption by
0.13 percent. In general, the developmental effect of infrastructure is
indirect and complex (Ahmed & Hossain 1990), although its influence
on poverty and inequality is emphasized in the literature. For example,
Benson et al. (2005) and Anyanwu (2005) recommended improving
 Micro-determinants of Income Inequality and Consumption in Rural Bangladesh 127

access to services to reduce poverty. We have demonstrated that the

development of rural infrastructure has a clear positive influence on
increasing consumption/welfare of the households. Consumption is
significantly lower in Comilla region where inequality is also highest
(Table 3).

Table 4.
Microdeterminants of Income Inequality
Variable list Coefficient Robust z-values
standard errors
Constant 0.348* 0.217 1.65
Socio-economic factors
Per capita land owned 1.208*** 0.192 6.30
Farm operation size 0.001 0.024 0.01
Farm capital assets 0.002* 0.000 1.67
Herfindahl index of crop diversity -0.035 0.075 -0.49
Irrigation 0.108* 0.067 1.65
Tenants -0.117** 0.053 -2.21
Share of non-agricultural income 0.123** 0.059 2.07
Dependents in the household -0.070*** 0.010 -6.90
Education level of household head 0.010* 0.005 1.82
Highest education of male member in the 0.009 0.006 1.47
household
Age of the head 0.001 0.001 0.84
Village level factors
Index of infrastructure underdevelopment -0.004*** 0.001 -2.60
Index of soil fertility -0.117 0.136 -0.86
Comilla region -0.175*** 0.047 -3.70
Jessore region -0.043 0.071 -0.61
Model diagnostic
Adjusted R2 0.45
F-statistic (15,390df) 25.98***
H0: No influence of socio-economic factors on 25.27***
consumption/welfare (F-statistic (8,390df)
H0: No influence of spatial/geographic factors on 6.45***
consumption/welfare (F-statistic (4,390df)
Number of observations 406
*** = significant at 1 percent level (p<0.01); ** = significant at 5 percent level (p<0.05); * =
significant at 10 percent level (p<0.10).
128 Journal of Poverty Alleviation and International Development

Table 5.
Consumption/Welfare Elasticities
Variable list Elasticity estimates
Socio-economic factors
Per capita land owned 0.135**
Farm operation size 0.001
Farm capital assets 0.012*
Herfindahl index of crop diversity -0.021
Irrigation 0.067*
Tenants -0.017**
Share of non-agricultural income 0.027**
Dependents in the household -0.279***
Education level of household head 0.037*
Highest education of male member in the household 0.054
Age of the head 0.047
Regional/spatial factors
Index of infrastructure underdevelopment -0.132***
Index of soil fertility -0.196
Comilla region -0.054***
Jessore region -0.011
*** = significant at 1 percent level (p<0.01); ** = significant at 5 percent level (p<0.05); * =
significant at 10 percent level (p<0.10).

Conclusions and Policy Implications

Factors influencing inequality in rural households are complex.

The present study clearly demonstrated that the total income derived
from crop production is higher in a diversified cropping system and
cultivation of modern varieties of rice alone does not necessarily
translate into high total income. Gini-decomposition results support this
intuition. It revealed that the contribution of diverse crop production to
existing inequality is lowest (only 10 percent) while modern agricultural
technology adoption contributes about 29 percent, even though both have
an inequality decreasing effect while total income remains unchanged.
 Micro-determinants of Income Inequality and Consumption in Rural Bangladesh 129

Among the socio-economic characteristic variables, education levels of

the head of household significantly increase consumption/welfare
whereas the number of dependent persons has a doubly disadvantageous
effect as it significantly reduces consumption by a large margin. Turning
to the impact of land, land ownership significantly increases
consumption and is the most dominant positive variable.
In general, it is encouraging to note that factors within the control
of household decision making processes, such as education levels and
sources of income, particularly from diverse crop and modern variety
cultivation and non-agricultural income, reduce inequality in the
distribution of income and/or significantly increase consumption/
welfare. Therefore, the inherent disadvantage posed by location and
underdeveloped infrastructure in increasing inequality and/or reducing
consumption can be somewhat offset by promoting crop diversification
and modern agricultural technology diffusion in addition to mass adult
literacy improvement. Also, government has an important role to play to
improve the factor equalization role of land rental markets because
farming is still the dominant source of livelihood in Bangladesh and our
results reveal that consumption/welfare are significantly lower for
tenants. However, the conventional land reform measure of equalizing
land ownership among farmers, which is a common policy suggestion in
land scarce economies, is not feasible in the case of Bangladesh because
of technical and economic limitations, as well as the political economy of
its agrarian structure (Rahman 2010). The key policy thrust here should
be to facilitate the operation of the land rental markets instead, as well as
to improve ownership of the farm-capital assets that are also essential in
farm operations, which is shown to significantly improve consumption.
Therefore, an integrated policy of decentralized crop
diversification, incorporating the balanced adoption of modern
agricultural technology (e.g., one main season in a crop year cycle), mass
adult literacy promotion, tenancy reform to enable land rental market to
operate effectively, and rural infrastructure development to promote
economic diversification and non-agricultural income, is recommended
130 Journal of Poverty Alleviation and International Development

in programs designed to reduce income inequality and increase

consumption/welfare of rural households in Bangladesh.

References

Achia, T.N.O., Wangombe, A., & Khadioli, N. (2010). A logistic regression

model to identify key determinants of poverty using Demographic
Health Survey data. European Journal of Social Science, 13(1), 38–46
Ahmed, R., & Hossain, M. (1990). Developmental impact of rural
infrastructure in Bangladesh (Research Report No. 83). Washington,
DC: International Food Policy Research Institute.
Anyanwu, J. C. (2005). Rural poverty in Nigeria: Profile, determinants and
exit paths. African Development Review, 17(3), 435–460.
Asadullah, M. N., & Rahman, S. (2009). Farm productivity and efficiency in
rural Bangladesh: The role of education revisited. Applied
Economics, 41(1), 17–33.
Barrett, C. B., Reardon, T. & Webb, P. (2001). Nonfarm income
diversification and household livelihood strategies in rural Africa:
Concepts, dynamics, and policy implications. Food Policy, 26(4),
315–331.
Bangladesh Bank. (2012). Economic Trends (monthly publications) 2012.
Dhaka: Bangladesh Bank.
Bangladesh Bureau of Statistics (BBS), (2012). Yearbook of agricultural
statistics of Bangladesh, 2012. Dhaka: Author.
Benson, T., Chamberlin, J. & Rhinehart, I. (2005). An investigation of the
spatial determinants of local prevalence of poverty in rural Malawi.
Food Policy, 30(5-6), 532–500
Davis, J. R. (2004). The rural non-farm economy, livelihoods and their
diversification: Issues and options. Chatham, UK: Natural Resources
Institute, University of Greenwich. Retrieved from:
http://www.nri.org/projects/rnfe/papers.htm.
Deininger, K., & Olinte, P. (2001). Rural non-farm employment and income
diversification in Columbia. World Development, 29(3), 455-465.
Ellis, F. (2000). Rural livelihoods and diversity in developing countries.
Oxford: Oxford University Press.
Ellis, F. & Freeman, H. A. (2004). Rural livelihoods and poverty reduction
 Micro-determinants of Income Inequality and Consumption in Rural Bangladesh 131

policies. London: Routledge.

Fei, J., Ranis, G. & Kuo, S. (1978). Growth and the family distribution of
income by factor components. Quarterly Journal of Economics.
92(1), 17–53.
Freebairn, D.K. (1995). Did the Green Revolution concentrate incomes? A
quantitative study of research reports. World Development, 23(2),
265-279.
Hatlebakk, M. (2012). Regional variation in livelihood strategies in Malawi.
South African Journal of Economics, 80(1), 62-76.
Hossain, M. (1989). Green revolution in Bangladesh: Impact on growth and
distribution of income. Dhaka: University Press Ltd.
Hossain, M. & Sen, B. (1992). Rural poverty in Bangladesh: Trends and
determinants. Asian Development Review, 10(1), 1–34.
Hossain, M., Quasem, M. A., Akash, M. M., & Jabber, M. A. (1990).
Differential impact of modern rice technology: The Bangladesh case.
Dhaka: Bangladesh Institute of Development Studies.
Krishna, A. (2013). The mixed news on poverty. Current History, 112(750),
20-25.
Makdissi, P., & Wodon, Q. (2012). Gini decomposition and gini income
elasticity under income variability. Bulletin of Economic Research,
64(2), 184–191
Mian, A. J. (1978). A least cost nutritional diet for Bangladesh: An
alternative approach to food supply (Ph.D. Dissertation). Ohio: The
Ohio State University.
Ministry of Agriculture. (2008). Agricultural handbook of Bangladesh, 2007.
Dhaka: Ministry of Agriculture, Government of Bangladesh.
Pyatt, G. (1976). On the interpretation and disaggregation of Gini coefficients.
Economic Journal. 86(342), 243–255.
Pyatt, G., Chen, C-N., & Fei, J. (1980). The distribution of income by factor
components. Quarterly Journal of Economics. 95(3), 451–473.
Rahman, S. (1999). The impact of technological change on income
distribution and poverty in Bangladesh agriculture: An empirical
analysis. Journal of International Development, 11(7), 935–955.
Rahman, S. (2009). Micro-determinants of poverty among farming population
in Bangladesh. Outlook on Agriculture. 38(4), 349-355.
Rahman, S. (2010). Determinants of agricultural land rental market
132 Journal of Poverty Alleviation and International Development

transactions in Bangladesh. Land Use Policy, 27(3), 957–964

Rao, V. M. (1969). Two decompositions of concentration ratio. Journal of
Royal Statistical Society, Series A. Series A(132): 418 – 25.
Ravallion, M. & Sen, B. (1996). When methods matter: Monitoring poverty in
Bangladesh. Economic Development and Cultural Change, 44(4),
761 – 792.
Ravallion, M., & Wodon, Q.T. (1999). Poor areas, or only poor people?
Journal of Regional Science, 39(4), 689–711
Smith, D.R. Gordon, A., Meadows, K., & Zwick, K. (2001). Livelihood
diversification in Uganda: Patterns and determinants across two rural
districts. Food Policy, 26, 421-435.
Stata Corp, (2010). Stata Statistical Software: Release 10.0. College Station,
Texas: Author.
Thapa, G., Otsuka, K., & Barker, R. (1992). Effect of modern rice varieties
and irrigation on household income distribution in Nepalese villages.
Agricultural Economics. 7(3-4), 245–65.
Wodon, Q. T. (1997). Food energy intake and cost of basic needs: measuring
poverty in Bangladesh. Journal of Development Studies. 34(2),
66–101.
Wodon, Q. T. 2000. Microdeterminants of Consumption, Poverty, Growth,
and Inequality in Bangladesh. Applied Economics, 32(10),
1337–1352.
Yao, S. 1997. Decomposition of Gini coefficient by income factors: A new
approach and application. Applied Economics Letters. 4(1), 27–31.
 Micro-determinants of Income Inequality and Consumption in Rural Bangladesh 133

Appendix

Components of Household Income

The disaggregation of total family income into the following

components provides a first-hand picture of sources of income:
1. Income from crop production (CROPI)
2. Income from livestock (LIVEI)
3. Income from fisheries (FISHI)
4. Income from land leased-out/rented-out (LEASEI)
5. Income from wage (WAGEI)
6. Income from business and miscellaneous sources (BUSI)
7. Total agricultural income (AGI) = CROPI + LIVEI + FISHI +
LEASEI
8. Total non-agricultural income (NAGI) = WAGEI + BUSI
9. Total household income (INC) = AGI + NAGI

Derivation of Income

Income derived from crop production (CROPI) is straightforward.

As the present study covers information on all types of crops produced
by the households in one year, so the total income from producing
various crops are computed directly after deduction of all input costs
including purchased and family supplied items. Costs of family supplied
inputs were imputed with the respective market prices as appropriate.
The income is net income from crop production.
Income from livestock sources are estimated from direct questions
to the respondents on various products and by-products produced from
livestock resources, such as from milk, meat, egg, sale, value of
consumed product, etc. Also, information on weekly expenditure on
livestock raising is collected which is then multiplied by 52 to arrive at an
annual expenditure and deducted from total gross income to yield net
income from livestock.
134 Journal of Poverty Alleviation and International Development

Incomes from fisheries resources are estimated from direct

questions on costs and returns of fish production in one year. Costs
include excavation, liming, fertilizing, feeding, renting (if multiple
owned) and harvesting costs. Incomes include revenue from sale of
harvest, imputed value of fish consumed by the family and value of stock
in the pond. The total cost is then deducted from gross income to yield
net income from fisheries.
Income from all other categories are estimated from direct
questions on type of activities, in which individual working members of
the household are involved for one week preceding the day of survey,
number of days worked and income earned from these activities. This
weekly income derived from various sources is then multiplied by 52 to
arrive at the annual income.
It should be noted that though such computation is highly
subjective, a cross-examination of annual expenditure incurred by the
household (based on similar method which is reported in lower panel of
Table 2) and the derived total income revealed a discrepancy of about 10–
15 percent only.