determinants of livelihood security in poor settlements in...
TRANSCRIPT
NAF
International Working Paper Series Year 2010 paper n. 10/01
Determinants of Livelihood Security in Poor Settlements in Bangladesh
Sanzidur Rahman and Shaheen Akter
University of Plymouth -UK
The online version of this article can be found at: http://economia.unipv.it/naf/
Scientific Board
Maria Sassi (Editor) - University of Pavia
Johann Kirsten (Co-editor)- University of Pretoria
Gero Carletto - The World Bank
Piero Conforti - Food and Agriculture Organization of the United Nations
Marco Cavalcante - United Nations World Food Programme
Luc de Haese - Gent University
Stefano Farolfi - Cirad - Joint Research Unit G-Eau University of Pretoria
Ilaria Firmian - IFAD
Firmino G. Mucavele - Universidade Eduardo Mondlane
Michele Nardella - International Cocoa Organization
Nick Vink - University of Stellenbosch
Alessandro Zanotta - Delegation of the European Commission to Zambia
Copyright @ Sassi Maria ed. Pavia -IT [email protected] ISBN 978-88-96189-03-0
DETERMINANTS OF LIVELIHOOD SECURITY IN POOR
SETTLEMENTS IN BANGLADESH
Sanzidur Rahman and Shaheen Akter*
Abstract Drawing on livelihoods approach and household survey data this study measures
livelihood security in selected settlements in Bangladesh. Five security domains such as
economic, food, health, education and empowerment were chosen and indices were
computed based on a number of components. Economic security is the dominating
component in the overall livelihoods followed by food. Irrespective of regional differences
in opportunities, people in settlements appear equally insecure. This does not mean that
the same intervention strategy is equally applicable everywhere. There are geographical
differences in the component indicators. Access to assets/capital endowment should be
taken into consideration to design programmes. Areas where land/housing/ponds more
accessible, livestock/fisheries based livelihoods may be supported. Education enhancing
policies are equally suitable for everywhere.
Keywords: Livelihoods security, urban settlements, Bangladesh
JEL classifications: O1; O18; R0
* Contact address: Shaheen Akter, 607A Silbury Boulevard, Milton Keynes, MK9 3AR, UK – e-mail: [email protected]. The authors wish to acknowledge Plymouth University for the financial support. Online dataset was obtained from the IFPRI site.
1. Introduction
No matter how impressive is the economic growth, world hunger is rising (FAO, 2008). The
number of starving people increased from 848 million in 2003-05 to 923 million in 2007.
Furthermore, 65 percent of the 832 million chronically poor people live in only seven
countries including Bangladesh. Recent hike in food prices worldwide has been placing the
poor people even vulnerable. Food security is now identified ‘most off-track MDG’ (EU,
2010). This reality further enhanced the need for rigorous investigation of the extent and
nature of livelihood security.
Early literature considered food security in terms of shortages of food supply due to drought,
flood, cyclone and other natural catastrophes (Frankenberger and McCaston 1998). Later
from the experience of food crisis in the mid 1980s, researchers and development
practitioners realised that food insecurity resulted not only from lack of supply but also
from lack of access to resources and food. This thinking helped developing livelihoods
approach to conceptualising household livelihoods security. The recent recession made
the situation of those lacking resources worse. The challenge ahead is to face growing
problems of not only food shortages but also to ensure access to nutrition, health care,
safe water and sanitation and above all access to resources that help the poor to afford
decent living. The recent FAO estimate says that agricultural production will need to
increase by 70 per cent by 2050 to cope with a 40 per cent increase in world population .
Even if production is adequate at the aggregate level many people are excluded to enjoy
it. Without ensuring poor people’s access increased production cannot resolve the
problems.
One of the world’s populous countries, Bangladesh always finds it difficult to feed its entire
population. Recent MDG analysis identifies Bangladesh as a nation that has attained
some remarkable social and economic successes in terms of per capita income growth,
reduction in population growth, decrease in child mortality, improvements in child nutrition,
expansion of primary and secondary education, reduction of gender inequality in
education, maintaining food production close to self-sufficiency level, and sustained trends
of decline in income-poverty. In spite of impressive progress, 40% of the population is still
living below the poverty line according to Household Income and Expenditure Survey 2005
(BBS 2007), 69 million live less than $1 a day, nearly half the children do not complete
primary education and more than 30 million people do not have access to safe drinking
water. However, this may not be viewed as unusual for a country of more than 140 million
people living on a tiny piece of land with very limited natural resources confronted
frequently by natural calamities. Being trapped in low-wage low-skilled work with little job
security, inadequate food and shelter, deprivations of basic education and health care
services the poor people are extremely vulnerable to ill health, economic dislocation and
natural disasters.
The existing literature on livelihood analysis is skewed towards qualitative accounts and
usually restricted to a geographical area or a particular resource management system and
so conclusions are imprecise, often not possible to generalise them (e.g., Kabeer, 2004;
Toufique and Turton, 2002; Lindenberg, 2002; de Haan et al., 2000; Toulmin et al., 2000;
Ashley, 2000; Carney, 1999). Use of quantitative approach or Q-square approach to
analyse livelihoods is inadequate. Jansen et al. (2006) applied a quantitative approach to
analyse livelihood strategies and their determinants but the application is limited to the
hillside population in rural Honduras.
The primary focus of this study is to analyse livelihood security using household survey
data in two regions in Bangladesh collected by the International Food Policy Research
Institute (IFPRI) and CARE Bangladesh (CARE 2001, 2004).
The main objectives are:
1. To examine the livelihood security status of slum households.
2. To identify the determinants of livelihood security.
3. To discuss policy implications.
The methodology is described in section 2. Data and variables are presented in section 3.
Results are discussed in section 4. The study concludes in section 5
2. Methodology The questions are addressed using livelihoods security approach. This approach evolved
from the food crisis in the mid 1980s and Sen’s (1981) theory on entitlement referring to
the set of income and resource bundles (e.g. assets, commodities) over which households
can establish control and protect livelihoods. The evolution of the concepts and issues
related to this theory eventually led to the development of the broader concept of
household livelihood security (HLS). The diversity in the interpretation of livelihoods
approach has documented by Hussein (2002), which presents the key elements of
livelihoods approaches of 15 development agencies, ranging from bilateral and multilateral
to nongovernmental organisations (NGOs). Of them, CARE considers this approach as its
integral part and defines HLS as adequate and sustainable access to income and
resources to meet basic needs (including adequate access to food, potable water, health
facilities, educational opportunities, housing, and time for community participation and
social integration) (Frankenberger 1996). This definition is in fact derived from Chambers
and Conway’s (1992) definition of livelihoods and is linked to basic needs. Chambers and
Conway (1992) conceptualise sustainable livelihoods in terms of capacities and activities:
a livelihood comprises the capabilities, assets (stores, resources, claims and
access) and activities required for means of living: a livelihood is sustainable which
can cope with and recover from stress and shocks, maintain or enhance its
capabilities and assets, and provide sustainable livelihood opportunities for next
generation: which contributes net benefits to other livelihoods at the local and global
levels in the long and short term. (pp 6 – 7) “
CARE views that this livelihood approach can effectively incorporate basic needs and
right-based approaches, which provide an additional analytical lens (Carney et al. 1999).
This concept of HLS embodies three fundamental attributes of livelihoods: (1) the
possession of human capabilities (e.g., education, skills, health, psychological orientation);
(2) access to tangible and intangible assets; and (3) the existence of economic activities
(Drinkwater and Rusinow, 1999). The interaction between these attributes defines what
livelihood strategy that a household will pursue to reach its desired outcomes known as
CARE’s Household Livelihood Security model (Figure 1).
Simply speaking, livelihood security here refers to the ability of the household to meet its
basic needs (or realize its basic rights). These needs include adequate food, health,
shelter, minimal levels of income, basic education, and community participation
(Frankenberger et al. 2000).
The asset box, shown in Figure 1, includes resources to which household members have
access (including access to information and other influences), their capabilities, and ability
to claim from relatives, the state and other actors.
Figure 1. CARE's livelihood security model
Source: Adapted from CARE 2004
The production and income activities are means to improving livelihoods (Drinkwater and
Rusinow 1999). Households live within socio-economic, political and cultural contexts that
influence the livelihood strategies to reach the desired outcomes.
This framework allows investigating a diverse range of indicators of HLS. In this study we
have identified 33 security indicators from the data set and broadly grouped them under
five security domains: economic security, food security, health security, educational
security and empowerment (Lindenberg 2002). Household level HLS indices are then
constructed following Hahn et al. (2009). The framework is discussed as follows:
Indicators are identified and it is assumed that each indictor has equal weight to the overall
HLS index. The indicators are then standardized following the procedure adopted in
measuring Life Expectancy in Human Development Reports (also adopted by Hahn et al.
2009). For example, a standardized indicator j of a household is given by:
jj
jindicatorzind
j
jminmax
min
−
−= (1)
where minimum and maximum values of the indicators are from the same community
within which the household belongs. Once each indicator representing a particular
livelihood security domain is standardised, then the relevant household livelihood security
index for the particular domain is constructed by averaging the standardised indicators:
J
zindj
HLS
J
j
i
∑=
=1
(2)
where J is the number of indicators used to construct the index. Once each HLS index is
constructed, then the composite overall Livelihood Security (LS) index for the household is
constructed by using the formula in equation (3):
∑
∑
=
==n
i
i
n
i
ii
i
w
HLSw
LS
1
1 (3)
where w are the weights determined by the number of indicators used to construct each
HLS index. Weights vary between households because of household level variation in the
number of indicators.
This framework differs from CARE’s, which has developed a composite HLS index using
rapid community appraisal techniques, where a team of 10-12 persons employ about eight
hours in a selected community and interviewed selected households (Lindenberg 2002).
This provides a quick grasp of the constraints faced by the households and/or
communities and the caveats are well known. Particularly, questions are often being raised
on the reliability of the information which usually reflects the views of those involved in the
exercises and so difficult to generalise the conclusions. In contrast our approach uses in-
depth quantitative indicators of different security areas with greater degree of precision
and the tools can be used widely as the required data is increasingly being available world
wide.
3. Analytical framework
We have constructed five livelihood security indices such as: economic security, food
security, health security, educational security and empowerment. In order to examine the
determinants of LS (eq. 3) we have specified the following equation:
i
n
k
ikik
i
iiii XHLSLS εβγα ∑∑==
+++=1
5
1
(4)
where X’s are the exogenous variables representing household’s socio-economic
circumstances as well as community level attributes. The estimation technique should
explicitly take account for possible endogeneity of HLS indices. Based on the test results
we estimated single equation livelihood security model using two stage least squares
(2SLS) method. We also reported the results based on ordinary least squares method
(OLS).
3.1. Data and variables
Data are drawn from the SHAHAR (Supporting Household Activities for Health, Assets and
Revenue) project implemented by CARE-Bangladesh for the purpose of intervention
aiming to establish household livelihood security for vulnerable urban households. The
SHAHAR Baseline Survey was conducted in slums and low-income settlements in August
2000 within the municipal areas of Jessore and Tongi districts (CARE 2001). These two
secondary cities Jessore and Tongi were selected purposively to consider diversities in
city characteristics. Figure 2 shows the locations in a map of Bangladesh. The spatial form
of Jessore is very different from that of Tongi. The latter is characterized by the presence
of large slum areas that have distinct identities and are to a large extent spatially isolated
from neighboring communities. In contrast the slum communities in Jessore are to a large
extent part and parcel of the city, located alongside middle-class and well-off
neighbourhoods.
Jessore is located in the southwest of Bangladesh on the main transport route linking
Bangladesh to India. Administratively Jessore is divided into 9 wards1. Of these 9 wards
some 63 slum communities known as bastis were identified2.
Tongi is an industrial area located 25 km north of Dhaka City. Many of its inhabitants
including women work in the neighboring mills and factories. Some 21 slum communities
from 6 wards were selected for survey.
Study sites in Jessore consist of a mix of rich, middle class and absolutely poor
households living together whereas in Tongi the residents are purely slum dwellers.
1 A ward is the smallest administrative unit in the urban/suburb setting in Bangladesh.
2 A basti is often defined as an unplanned settlement of households typically without secure tenure, adequate sanitation and other urban services needed to maintain minimum environmental health standards.
Figure 2. Locations of Jessore and Tongi in Bangladesh.
Also, a few sites in Jessore are located at the fringes of pourashava, which has a complex
mix of urban and rural lifestyles, including extensive crop agriculture.
In Jessore 563 households were surveyed in September 2000, as were 557 in Tongi.
Households were selected randomly from a complete listing done as part of a census in
the areas in April-May 2000. The sample size was statistically representative (CARE
2001). The size was determined as:
2
2
)05.0(
)]1([)645.1( ppn
−×=
where,
1.645 is the standard error associated with 90 per cent confidence level of a
standard normal distribution,
p = proportion of a key variable of interest, estimated prevalence of stunting in this
case, because the survey was a baseline meant for action research to improve food and
nutrition security.
0.05 = error level (5 per cent)3
A structured questionnaire consisting of 17 modules was used for data collection. Topics
comprise household composition, migration and education, status of employment and
earnings, transfers, social assistance and other income, household assets, urban
agriculture, savings, loans, housing, environment, water and sanitation, daily food,
consumption, diarrhoea and other illnesses, health, nutrition knowledge and practice, pre-
school feeding, utilization of health care facilities for pregnancy/birth, anthropometry,
community participation, general household livelihood security.
The enumerators visited each household 2-3 times in September 10-26, 2000 to complete
all sections of the questionnaire. Table 1 shows the sample distribution; number of
households, number of persons in household by gender and average family size. Family
size was slightly higher in Jessore but not different statistically.
Table 1. Sample size, Jessore and Tongi, Bangladesh Locations Households N Male members Female
members Total members
Family size
Jessore 563 1337 1347 2684 4.77 Tongi 557 1292 1289 2581 4.63 Total 1120 2629 2636 5265 4.70
As we discussed in the first sub-section of the methodology, we are interested in five
security areas: economic security, food security, health security, educational security and
empowerment. We assume that these areas are more relevant and directly related to the
welfare of the poor and vulnerable people. Other social, political and environmental areas
could also be given attention. In this study we avoid complexities of measuring so many
indicators that indirectly related to welfare of the poor. Table 2 includes a set of
indicator/component variables that we constructed from the baseline survey data to
calculate the indices. These indicators are assumed to differentiate household status
substantially. For example, income levels differentiate economic status and so it is a
component of economic security. Similarly, dietary diversity distinguishes food security
status. Health security can be distinguished by examining the incidence of sickness and
access to treatment and control. Education security can be differentiated by the level of
3 Rapid assessment was used to estimate the prevalence of stunting (p), which was 38 per cent of boys and 41 per cent of girls. A higher rate of 50 per cent was used to select the sample size to account for any error in the assessment as well as to maximise sample size.
literacy. Empowerment is distinguished by institutional participation and access to services
by such organisations. Some indicators should represent the quality of these components.
Quality component is not given adequate attention in this analysis due to non-availability of
information on quality.
Table 2. Livelihood security indicators, Jessore and Tongi, 2000, Bangladesh Security Components
Indicators
Economic Per person income (TK per person per month) Per person current value of land/house/animal shed/pond (TK) Per person current value of livestock asset (TK) Per person current value of machineries & equipment (TK) Per person current value of other asset (TK) Active population ratio (15-59 yrs population/family size) Proportion of 15-59 population in employment Household income earned by women (TK/person) Per person current savings (TK) Per person current loan (TK) Food Dietary diversity: number of food groups consumed per day Food frequency (number of meals and snacks per day) Household foodgrain stock (TK per person) Number of food convenient months in the year Number of main meals taken by women in household Health Number of days members suffer from diarrhoea last 30 days Number of days members suffer from other sickness last 30 days Number of days unable to work due to sickness Frequency of antenatal consultation Doses of tetanus vaccination Body Mass Index women Body Mass Index children<=5 yrs Education 7+ population read and write (Literacy) Adult male literacy 15+ literate Adult female literacy 15+ female literate Adult members 10 years or more education 6-10 years children enrolled 11-15 years boys enrolled 11-15 years girls enrolled 16-23 years per son in household enrolled Empowerment Community participation/active involvement with organisation
(months) Access to services/organisations that offer services (yes=1,
no=0) Household participation in planning process (yes=1, no=0)
4. Results
4.1. Income and livelihood diversity
Data of regular activities and income of last 30 days for four broad activity groups were
collected in the survey. The activity groups are wage labourer, salaried worker (with and
without salary or pay) and self-employed. Several activities were identified under each of
these broad groups. Data was also collected for seasonal income from enterprises, social
assistance and other irregular sources for the last six months. Income from all these
sources was aggregated and per person monthly income is reported in Table 3. Monthly
average income is slightly higher in Tongi but statistically average income is the same in
two areas. We subdivided all the activities, enterprise income and other sources of income
into 12 groups and calculated Herfindahl-Hirschman Index (HHI)4. The value of this index
ranges from 1 to the number of activities (12 in this case). Livelihoods appeared equally
diversified in the settlements of both the districts. Individuals from different households
engage in many activities but household level diversification in the settlements is low; 1.42
out of 12 activities.
Table 3. Per person monthly income and livelihood diversity, Jessore and Tongi, Bangladesh, 2000. Locations N Per person monthly income
(Bangladesh Taka) 12 activity diversity (HHI)
Average Std Dev Average Std Dev Jessore 559 820.86 664.87 1.44 0.54 Tongi 551 891.88 1314.39 1.40 0.52 Total 1110 856.12 1039.46 1.42 0.53
There are little differences in the aggregate level in terms of per person income and
diversity but from the income shares of each occupation group we find a number of
differences between Jessore and Tongi areas. Wage labour and salaried income are two
dominating sources of earnings in Tongi (Table 4). This is consistent with its industrial
nature and proximity to Dhaka city. Wage labour and self employment (except craft and
related trade workers) are equally important in Jessore. Particularly income share from
trading in Jessore is significantly higher than Tongi. This evidence tends to reflect
Jessore’s convenience of trade with India; many people in this area engage in intercountry
trade. Enterprise income, which includes agriculture, is also higher in Jessore where land
and natural resources are relatively more accessible. Particularly, income from vegetables,
fruits and livestock was much higher in Jessore.
4 The methodology of calculating HHI is included in the Appendix.
Table 4. Share of income from 12 groups of sources Occupation category Jessore Jessore Tongi Tongi Total Total Mean Std
Dev Mean Std
Dev Mean Std
Dev Wage labourer 14.8 31.2 25.7 39.5 20.2 36.0 Salaried worker 13.9 31.0 23.6 36.2 18.7 34.0 Maid servant & other salaried worker 3.9 14.1 2.5 13.7 3.2 13.9 Professional 1.0 7.6 0.7 6.8 0.8 7.2 Craft & related trade workers 8.2 24.7 6.0 20.9 7.1 22.9 Plant & machine operators/drivers 11.5 28.9 10.2 27.3 10.8 28.1 Petty traders 13.8 30.2 9.8 25.6 11.8 28.1 Medium/large traders 13.5 31.2 10.3 28.0 11.9 29.7 Other occupation 3.8 16.4 2.1 11.0 2.9 14.0 Transfer, social assistance, rent, interest etc. 10.7 24.4 7.7 19.8 9.2 22.2 Net income from enterprises 3.2 12.1 0.6 5.3 2.0 9.4 Other irregular sources 1.7 9.6 0.9 8.4 1.3 9.0 Total 100.0 0.0 100.0 0.0 100.0 0.0
4.2. Economic security We selected 10 economic security indicators as in Table 5. Higher values of them imply
households are economically better off and hence more secured economically. Table 5
reports the mean and standard deviation of the indicators separately for two regions;
Jessore and Tongi. The economic security index was calculated using the standardised
values of these indicator variables; standardisation was done using their ward level
maximum and minimum values. Economically, the two regions are the same; economic
security index is statistically the same and low in both regions. We conclude that the slum
population in Bangladesh is economically extremely insecure. Location of the settlements
does not matter to economic security. This indicates policy intervention is equally
necessary in all urban settlements but definitely not the same intervention appropriate in
all locations. For example, households in the Jessore settlements are endowed more with
land based resources as well as machinery and equipment and so intervention with land
based enterprises may be more appropriate for Jessore but may not be suitable for Tongi
settlements.
4.3. Food security Household level food baskets collected on 24 hour recall basis were divided into 8 groups.
These groups are cereals, roots and tubers, pulses, foods of animal origin, vegetables,
fruits, fats and oils, and snacks. Only 2 per cent of the households had diets consisting of
all 8 types of food.
Table 5. Economic security indicators Indicators Jessore Jessore Tongi Tongi Total Total Mean Std Dev Mean Std Dev Mean Std Dev Per person income 820.86 664.87 891.88 1314.39 856.12 1039.46 Per person own land/housing/pond (TK)
25252.17 67261.75 8516.89 25961.28 16929.36 51741.12
Per person livestock (TK)
252.64 1018.12 38.36 275.73 146.08 754.89
Per person machinery & equipment/transport (TK)
1504.96 12435.48 307.94 2021.85 909.66 8947.39
Per person other asset (TK)
2435.58 3357.07 1865.65 2511.12 2152.14 2979.02
Proportion of 15-59 population
0.60 0.19 0.60 0.21 0.60 0.20
Proportion of 15-59 in employment
0.55 0.27 0.62 0.28 0.58 0.27
Per person female income (TK)
65.32 163.93 156.41 269.22 110.20 226.66
Per person current savings (TK)
1419.71 10288.64 431.93 1088.40 928.47 7348.29
Per person current loan (TK)
791.96 2699.52 1093.39 2149.20 941.87 2444.95
Economic security index 0.172 0.074 0.170 0.071 0.171 0.072
Other households missed one or more types. About 66 per cent of the households missed
four types of food other than cereals in 24 hours (Table 6). Missed foods are mainly
protein-rich high value products such as foods of animal origin (milk, milk products, eggs
and meat) and fruits. Data was also available on number of times each type of food was
consumed in a 24 hour period (food frequency).
Food frequency is significantly highly correlated positively to the number of types food
groups consumed (correlation equals 0.78 significant at 1 per cent level). This means that
the peoples, who eat more frequently, also eat more types of food. In other words, food
frequency and dietary diversity are highly correlated variables. Any of these two variables
may be used to represent food diversity; here we have used both food diversity and food
frequency indicators. The frequency of taking food ranges from 2 to 27 times a day. Some
households eat food only twice a day, others eat more frequently up to a maximum of 27
times. Cereals (rice and wheat) are common in the diet of everybody. More than half of the
households in both locations consumed roots and tubers, particularly potatoes. Fish was
also common. Vegetable intake was quite low in Tongi, particularly for the female-headed
households. In general, intake of protein-rich foods (e.g., meat, milk and milk products,
eggs and fruits) was lower in female-headed households than male-headed households in
both areas.
In Table 6 we categorised the households based on frequency of food groups eaten daily.
The most frequent number of taking food is four types of food groups in addition to cereals.
Table 6. Frequency and percentage distribution of sample households by the number of food groups consumed daily (dietary diversity), Jessore and Tongi, Bangladesh (food diversity)*.
Frequency Percent Only cereals 2 0.2 Cereals plus another type 26 2.3 Cereals plus any two types 106 9.5 Cereals plus any three types 273 24.4 Cereals plus any four types 332 29.6 Cereals plus any five types 249 22.2 Cereals plus any six types 97 8.7 All types 23 2.1 Missing 12 1.1 Total 1120 100.0 * Food groups are cereals, roots & tubers, pulses, animal origin foods, vegetables, fruits, fats & oils, and snacks
Average and standard deviation of food security indicators are presented in Table 7.
There are differences between Jessore and Tongi in terms of foodgrain stocks. People in
Jessore have more access to secured food due to the availability of higher foodgrain
stocks. The difference appears small but statistically highly significant.
There is a common tendency of female members to skip meals and eat less after feeding
all other members. Obviously, this has food security implications. So we have included
‘number of main meals took by women in household’ in the indicator list of food security.
There is no difference between Tongi and Jessore with respect to this indicator. Data on
food quantity was not recorded to examine whether female members eat less quantity than
required in each meal.
Table 7. Food security indicators Indicators Jessore Jessore Tongi Tongi Total Total Mean Std
Dev Mean Std
Dev Mean Std
Dev Food frequency (meals and snacks per day) 11.7 3.5 12.0 4.2 11.9 3.9 Dietary diversity: number of food groups consumed per day 5.1 1.2 4.8 1.3 4.9 1.3 Household foodgrain stock (TK per person) 57.6 383.7 28.9 177.7 43.3
299.7
Number of food convenient months in the year 9.6 2.7 9.5 2.3 9.6 2.5 Number of main meals taken by women in household 2.9 0.3 2.9 0.3 2.9 0.3 Food security index 0.555 0.135 0.525 0.121 0.540 0.129 Peoples in the settlements are much better in terms of food security relative to economic
security but still the average is in the middle of the scale of 0 to 1.
4.4. Health security We have used seven component measure of health security and identified that people of
Jessore and Tongi were equally health insecure but there are differences in terms of some
individual components. Sickness is significantly higher in Tongi (Table 8). An estimated 81
per cent of households in Jessore and 83.3 per cent in Tongi had at least one member
who was sick during the 30-day recall period. Consistently, body mass index is
significantly lower in Tongi. Further analysis of data shows that in Tongi close to 49 per
cent of girls and 41 per cent of boys under age 5 were stunted while in Jessore 33 per
cent of girls and 40 per cent of boys were stunted. Another 20 per cent of the children in
Tongi and 15 per cent in Jessore were underweight for their height. This indicates the
existence of alarming malnutrition.
Table 8. Health security indicators Household level indicators
Jessore Jessore Tongi Tongi Total Total
Mean Std Dev
Mean Std Dev
Mean Std Dev
Per person incidence of diarrhoea in last 30 days 0.5 2.0 0.8 2.1 0.7 2.1 Per person days of other sickness in last 30 days 7.3 8.0 7.4 7.6 7.4 7.8 Number of days an active person unable to work due to sickness 3.6 4.4 5.6 6.1 4.6 5.4 Per women frequency of antenatal consultation 4.1 2.0 4.2 2.2 4.2 2.1 Per women doses of tetanus vaccination 2.2 0.9 2.2 1.0 2.2 0.9 Body Mass Index per women 21.2 3.5 20.2 3.1 20.7 3.4 Body Mass Index per children<=5 yrs 15.3 6.7 15.0 5.3 15.1 6.0 Health security index 0.235 0.125 0.233 0.125 0.234 0.125
4.5. Education security Seven indicators in Table 9 were used to measure education security. It is lower in Tongi
than in Jessore. All the indicators have lower average value in Tongi in spite of its
proximity to capital city Dhaka. These may be the combined effects of a number of factors.
Both cities comprise population with a majority rural migrants but Tongi holds more. As the
literacy rate is lower in rural areas, this is reflected in the education indicators in Tongi.
Tongi is more congested and so basic services are extremely poor . Nearly two thirds of
the households in Tongi and more than half of the households in Jessore are struggling in
absolute poverty. Female-headed households, which account for 21 per cent of
households in Tongi and 11 per cent in Jessore, with 85 per cent and 70 per cent,
respectively, are not able to meet basic needs.
4.6. Empowerment
Empowerment has the lowest values among the five domains of livelihood security (Table
10). People are slightly more empowered In Jessore but not statistically significant.
Table 9. Education security indicators Household level indicators
Jessore Jessore Tongi Tongi Total Total
Mean Std Dev
Mean Std Dev
Mean Std Dev
7+ population read and write (Literacy)
2.43 1.87 1.91 1.84 2.17 1.87
Adult male literacy 15+ literate
1.00 1.08 0.74 0.91 0.87 1.01
Adult female literacy 15+ female literate
0.70 0.78 0.49 0.72 0.59 0.76
Adult members 10 years or more education
0.36 0.88 0.15 0.53 0.26 0.74
6-10 years children enrolled
0.44 0.62 0.36 0.60 0.40 0.61
11-15 years boys enrolled
0.17 0.42 0.14 0.39 0.16 0.40
11-15 years girls enrolled
0.19 0.44 0.14 0.37 0.16 0.40
16-23 years per son in household enrolled
0.16 0.53 0.08 0.34 0.12 0.45
Education security index
0.15 0.20 0.10 0.17 0.12 0.18
Table 10. Empowerment indicators Household level indicators
Jessore Jessore Tongi Tongi Total Total
Mean Std Dev
Mean Std Dev
Mean Std Dev
Community participation /active involvement with organisation (months)
2.12 5.23 6.51 15.78 4.31 11.92
Access to services/ organisations that offer services (yes=1, no=0)
0.13 0.34 0.10 0.30 0.12 0.32
Household participation in planning process (yes=1, no=0)
0.06 0.24 0.06 0.25 0.06 0.24
Empowerment index 0.11 0.15 0.09 0.13 0.10 0.14
Empowerment index was calculated based on three indicators such as community
participation, access to services and participation in the planning process. Community
participation is measured by number of months of active involvement with any organisation
that deliver community services. Access to services is measured based on whether
households received any service (yes=1 and no=0) such as training, credit, health
awareness, water and sanitation, sports, culture and other urban amenities from any
provider. Household participation in the planning process was measured from the answers
(yes=1 and no=0) to question that ‘Have any of the household members ever participated
in any planning process with the pourashava regarding future of your community?’ Only 6
per cent of the households reported participation in the Pourashava (city council) planning
process. In spite longer involvement with different organizations, Tongi people had lower
access to services, perhaps because the area is overcrowded.
4.7. Overall livelihood security
Overall livelihood security index comprises five major livelihood security domains:
economic security, food security, health security, educational security and empowerment.
On an average, overall security is higher in Jessore (Table 11). The difference is small but
statistically significant at 1 per cent level. This variation arises from the significant
difference in food, education and empowerment. The other two areas such as economic
security and health security have the same average statistically in both regions. In both
regions education and empowerment involve much lower median than average indicating
that the distribution is skewed towards the lower values of the indices. This means that the
majority are less secured than the average.
4.8. Determinants of livelihood security
In order to estimate the model 4 we include the variables as defined in Table 12. In
addition to livelihood security indices, we specified 10 variables to represent household
circumstances (X variables in equation 4).
For example the household which have higher level of family size and dependency ratio,
their demand for basic needs is also higher. We would expect these variables to affect
livelihood security negatively, other things being equal.
The security variables would affect overall security positively. Other variables included are
the characteristics of the household heads.
Table 11. Livelihood security in urban slums and settlements in Jessore and Tongi, Bangladesh, 2000. Mean Median Skewness Kurtosis Jessore Economic security index 0.170 0.160 1.296 5.986 Food security index 0.555 0.560 -0.746 4.880 Health security index 0.235 0.220 0.305 2.657 Education security index 0.146 0.06 1.802 6.178 Empowerment index 0.108 0.050 1.708 5.875 Overall livelihood security index 0.248 0.242 0.675 4.109 Tongi Economic security index 0.172 0.160 1.065 4.919 Food security index 0.526 0.540 -0.656 4.166 Health security index 0.233 0.230 0.290 3.055 Education security index 0.098 0.020 2.412 9.646 Empowerment index 0.090 0.040 1.840 6.163 Overall livelihood security index 0.238 0.235 0.549 4.280 Both regions Economic security index 0.171 0.16 1.190 5.518 Food security index 0.540 0.55 -0.662 4.539 Health security index 0.234 0.23 0.297 2.854 Education security index 0.122 0.04 2.064 7.506 Empowerment index 0.099 0.04 1.786 6.129 Overall livelihood security index 0.243 0.238 0.679 4.372
We would expect their age and education to associate positively with the security level but
other variables such as marital status, gender etc. may be associated with security level
either positively or negatively depending on the circumstances.
4.9. Test results
We first conducted specification tests. For example we have to decide whether we carry
out aggregate model combining Jessore and Tongi together or model them separately.
Here we applied Chow test and identified that they should be modelled separately or use
independent set of dummy variables to represent their differences. Separate models are
preferable due to fewer numbers of parameters to be estimated in each model.
We also carried out tests for normality, constant variance and endogeniety in the two stage
least squares (2SLS) framework.
Table 12. Variable definitions Variable Description Hlseco4 Economic security index hlsh1 Health security index hlse1 Education security hlsempo Empowerment index hlsfood Food security index famsize Family size (number of persons in the household) landhpc Per person own land/housing/pond (millionTK) deprati Dependency ratio (dependant members/active members) ageh Age of household head classcomh Highest education (years) of household head femh1 Female headed household (female head=1) mstd1 Head is currently married and living with spouse = 1 slitd1 Head is able to only read or only write or only sign name=1 litdummy4 Head is able to both read and write=1 tdum2 Head received business related training=1 Instruments zasopc Standardised value of household assets other than
land/housing/pond zactpro Standardised value of 15-59 member proportion in household zsick Standardised value of per person days of sickness in last 30 days zl7plus Standardised value of literacy (7 years and above can read &
write zserv Standardised value of household access to organisation/services zfdiv1 Standardised value of household dietary diversity Breusch-Pagan / Cook-Weisberg test for heteroskedasticity (STATA 2008) on the OLS
models identified that the variance was not constant (test results are reported below the
Tables A1 to A3 in the appendix). Cameron and Trivedi's decomposition of IM-test also
showed heteroscedasticity and non-normality problems (Cameron and Trivedi 2009).
Models for each security areas can be simultaneously estimated by 3SLS but it requires
normality, constant variance assumptions to be hold. So we applied two stage least
squares (2SLS). Results are shown below.
4.10. Jessore analysis All security variables (economic, food, health, education and empowerment) are highly
significant determinants of household livelihoods security (HLS). Signs of the coefficients
are as expected. The contribution of economic security at the margin is the highest to
overall livelihoods security. This is followed by food security. In proportionate term, for
example, with a 10 per cent increase in economic security overall livelihoods security rises
by 4 per cent (Table 13). Among other variables, dependency burden has significant
negative effect on livelihoods security. Given that the average level economic, education
and empowerment securities are extremely low it is necessary to implement programmes
targeting to improve economic condition, education and empowerment. It may be much
easier to implement programmes to enhance, for example, education security, which could
also improve overall livelihoods security. However programmes targeting economic
security would translate into higher livelihood outcomes.
4.11. Tongi analysis
In Tongi, results are consistent for all security variables and dependency ratio. The effect
is lower here relative to Jessore other than the effect of 2SLS estimate of food security.
This implies that economic, health, education and empowerment enhancing programmes
would produce slightly higher livelihood impact in Jessore than Tongi. For education, a 10
per cent increase would result in the rise in overall security by 0.94 per cent in Tongi in
place of 1.13 per cent in Jessore. So education enhancing policy would even produce
better outcomes in Jessore. The effect of dependency ratio is extremely consistent;
virtually the same in both regions. The goodness of fit of the Jessore model is slightly
better than the Tongi model. Some changes may be necessary in the Tongi model to
improve the results.
5. Conclusions and Implications This study measures livelihood security in selected settlements (slums and squatters) in
Bangladesh. Five security domains such as economic, food, health, education and
empowerment were chosen and indices were computed based on a number of
components under each domain. From the results we conclude that irrespective of
regional differences in opportunities, people in urban squatters appear nearly equally
insecure. This does not mean that the same intervention strategy is equally applicable
everywhere. There are geographical differences in the component indicators. Access to
assets/capital endowment should be taken into consideration to design programmes.
Areas where land/housing/ponds more accessible, livestock/fisheries based livelihoods
may be encouraged. Education enhancing policies are suitable for everywhere.
Table 13. Determinants of livelihood security in Jessore, Tongi and Pooled models. Variables* OLS
Jessore 2SLS Jessore
OLS Tongi
2SLS Tongi
OLS Pooled
2SLS Pooled
hlseco4 0.3742 0.3798 0.3707 0.3227 0.3724 0.3458 0.0109 0.0215 0.0110 0.0304 0.0077 0.0181 hlsfood 0.1904 0.1857 0.1902 0.2198 0.1905 0.2024 0.0047 0.0068 0.0064 0.0118 0.0040 0.0065 hlsh1 0.1567 0.1380 0.1558 0.1527 0.1573 0.1447 0.0052 0.0094 0.0067 0.0106 0.0043 0.0067 hlse1 0.1162 0.1271 0.1035 0.0944 0.1113 0.112 0.009 0.0083 0.0088 0.0091 0.0063 0.0063 hlsempo 0.1414 0.1340 0.1485 0.1266 0.1450 0.1369 0.0047 0.0090 0.0047 0.0111 0.0034 0.0068 famsize 0.0002 -0.0002 -0.0005 -0.0003 -0.0002 -0.0003 0.0007 0.0006 0.0009 0.0011 0.0006 0.0007 Landhpc 0.0071 0.0089 0.0433 0.0632 0.0099 0.0171 0.0108 0.0112 0.0257 0.0333 0.0090 0.0101 deprati -0.0034 -0.0026 -0.0018 -0.0037 -0.0027 -0.0033 0.0010 0.0010 0.0010 0.0014 0.0007 0.0008 ageh -0.0001 -0.0001 -0.0001 -0.0001 -0.0001 -0.0001 0.0001 0.0001 0.0001 0.0001 0.0000 0.0000 classcomh -0.0005 -0.0005 -0.0003 0.0000 -0.0004 -0.0003
0.0002 0.0002 0.0003 0.0004 0.0002 0.0002 femh1 -0.0024 -0.0027 -0.0014 0.0002 -0.0016 -0.0007 0.0027 0.0026 0.0026 0.0027 0.0019 0.0019 mstd1 -0.0007 0.0005 -0.0002 -0.0012 -0.0007 -0.0003 0.0028 0.0026 0.0027 0.0027 0.0020 0.0020 slitd1 -0.0016 -0.0013 -0.0012 -0.0007 -0.0015 -0.0012 0.0019 0.0020 0.0016 0.0018 0.0012 0.0013 litdummy4 -0.0067 -0.0070 -0.0108 -0.0121 -0.0090 -0.0094 0.0023 0.0023 0.0024 0.0025 0.0016 0.0016 tdum2 -0.0031 -0.0044 0.0125 0.0131 0.0034 0.0032 0.0030 0.0029 0.0069 0.0072 0.0035 0.0035 Constant 0.0232 0.0301 0.0294 0.0245 0.0259 0.0279 0.0057 0.0064 0.0049 0.0074 0.0037 0.0048 N 563 563 557 557 1120 1120 R2 0.9606 0.9585 0.9361 0.9279 0.9493 0.9473 * Robust standard errors are shown below the row of each variable coefficients. Significance levels of the coefficients are included in the appendix Tables A1-A3.
Not only overall livelihood security status is relatively poorer in Tongi but also the impact of
individual security domain on overall status is lower as compared to Jessore. This may be
due to the fact that the survey sites in Tongi are contextually different, more crowded and
are subject to more livelihood constraints as these are purely slum areas. Any poverty
reduction strategy should take care of this difference. Failure to do so would cause areas
like Tongi to be less benefitted from an intervention.
References
- Allen, E.W. 2008. International wheat outlook for 2008/09. Economic Research Service
Report #WHS-2008-01. United States Department for Agriculture (USDA), Washington,
D.C.
- BBS 2007. Report of the Household Income & Expenditure Survey 2005. Bangladesh
Bureau of Statistics, Government of Bangladesh, Dhaka.
- Cameron A C and Trivedi P K 2009. Microeconometrics Using Stata, A Stata Press
Publication, StataCorp LP, Texas.
- CARE 2001. Baseline Survey Report: Livelihood Security Analysis of Vulnerable Urban
Households Jessore and Tongi Pourashavas, CARE – Bangladesh in collaboration with
the International Food Policy Research Institute
- CARE 2004. Measuring Livelihood Impacts: A Review of Livelihoods Indicators,
Livelihood Monitoring Unit (LMU) Rural Livelihoods Program CARE Bangladesh, Prepared
by TANGO International, Inc.
- Chambers R & Conway G (1991) 'Sustainable rural livelihoods: practical concepts for the
21st century'. IDS Discussion Paper 296. IDS, Brighton.
- Ellis, F. 2000. Rural Livelihoods and Diversity in Developing Countries. Oxford University
Press.
- EU 2010. A twelve-point EU action plan in support of the Millennium Development Goals,
Communication from the Commission to the European Parliament, the Council, the
European Economic and Social Committee, and the Committee of the Regions, Brussels,
21.4.2010 COM(2010) 159 final.
- FAO, 2008. The State of Food Insecurity in the World, 2008. Food and Agriculture
Organization, Rome. ftp://ftp.fao.org/docrep/fao/011/i0291e/i0291e00.pdf
- Frankenberger T., Drinkwater M and Maxwell D. 2000. Operationalizing household
livelihood security: A holistic approach for addressing poverty and vulnerability Program
Document, CARE USA.
- Hahn M B, Riederer A M & Foster S O 2009. The Livelihood Vulnerability Index: A
pragmatic approach to assessing risks from climate variability and change—A case study
in Mozambique, Global Environmental Change, 19(1): 74-88.
- Hussein K 2002. Livelihoods Approaches Compared: A Multi-Agency Review of current
Practice, Department for International Development and ODI, UK.
- Lindenberg, M 2002. Measuring Household Livelihood Security at the Family and
Community Level in the Developing World, World Development, 30(2): 301-318.
- Mehra R and Rojas M.H. 2009. Women, food security and agriculture in global
marketplace, International Center for Research on Women, USA, 2009
- Sen A K 1981. Poverty and famines. Oxford, UK, Clarendon.
- Stata 2008. Stata 10.1 Statistics/Data Analysis, StataCorp, Texas, USA.
Appendix Calculation of inverse Herfindahl-Hirschman Index measuring livelihood diversity
The Herfindahl-Hirschman Index was originally being developed for measuring the degree
of market concentration that takes into account both the relative size and distribution of
each source, increasing as the number of firms in the market falls and the disparity in the
size of those firms increases. The inverse of this can be used to measure the degree of
livelihood diversity that takes into account both the relative size and distribution of each
source of livelihoods, increasing as the number of sources increases and the disparity in
the share of those sources in income/livelihood output falls. For example, a share of
livelihood source j in income (I) of a household is given by: i
j
jI
II =
The inverse of the Herfindahl-Hirschman Index (IHH) for this household is then calculated
as:
∑=
=J
j
i
Ij
IHH
1
1
Household income sources are first categorised on the basis of flow of income into three
categories. First regular occupation consists of either employment or self employment;
second category consists of net income from farming (crop, livestock, fisheries and agro-
forestry), which are seasonal in nature and third category consists of transfer, social
assistance, pension, rent, interest, income from pawning assets etc.
Table A1. Determinants of livelihood security in Jessore. OLS 2SLS Livelihood security
Coef. Robust Std. Err.
P>t Coef. Robust Std. Err.
P>t
hlseco4 0.3742 0.0109 0.0000 0.3798 0.0215 0.0000 hlsfood 0.1904 0.0047 0.0000 0.1857 0.0068 0.0000 hlsh1 0.1567 0.0052 0.0000 0.1380 0.0094 0.0000 hlse1 0.1162 0.0090 0.0000 0.1271 0.0083 0.0000 hlsempo 0.1414 0.0047 0.0000 0.1340 0.0090 0.0000 Famsize 0.0002 0.0007 0.7940 -0.0002 0.0006 0.7740 Landhpc 0.0007 0.0108 0.5110 0.0089 0.0112 0.4250 Deprati -0.0034 0.0010 0.0010 -0.0026 0.0010 0.0130 Ageh -0.0001 0.0001 0.0590 -0.0001 0.0001 0.0100 Classcomh -0.0005 0.0002 0.0300 -0.0005 0.0002 0.0310 femh1 -0.0024 0.0027 0.3780 -0.0027 0.0026 0.3140 slitd1 -0.0007 0.0028 0.7960 0.0005 0.0026 0.8530 mstd1 -0.0016 0.0019 0.3890 -0.0013 0.0020 0.5290 litdummy4 -0.0067 0.0023 0.0040 -0.0070 0.0023 0.0020 tdum2 -0.0031 0.0030 0.3070 -0.0044 0.0029 0.1340 Constant 0.0232 0.0057 0.0000 0.0301 0.0064 0.0000
Number of obs 563 OLS R-squared
0.9606
2SLS R-squared 0.9585
Breusch-Pagan / Cook-Weisberg test for heteroskedasticity: chi2(1) = 41.94 , Prob>chi2 = 0.0000
Table A2. Determinants of livelihood security in Tongi. OLS 2SLS Livelihood security
Coef. Robust Std. Err.
P>t Coef. Robust Std. Err.
P>t
hlseco4 0.3707 0.0110 0.0000 0.3227 0.0304 0.0000 hlsfood 0.1902 0.0064 0.0000 0.2198 0.0118 0.0000 hlsh1 0.1558 0.0067 0.0000 0.1527 0.0106 0.0000 hlse1 0.1035 0.0088 0.0000 0.0944 0.0091 0.0000 hlsempo 0.1485 0.0047 0.0000 0.1266 0.0111 0.0000 Famsize -0.0005 0.0009 0.5630 -0.0003 0.0011 0.7670 Landhpc 0.0043 0.0257 0.0920 0.0632 0.0333 0.0580 Deprati -0.0018 0.001 0.0640 -0.0037 0.0014 0.0070 Ageh -0.0001 0.0001 0.0080 -0.0001 0.0001 0.1060 Classcomh -0.0003 0.0003 0.3430 0.0000 0.0004 0.9320 femh1 -0.0014 0.0026 0.5780 0.0002 0.0027 0.9500 slitd1 -0.0002 0.0027 0.9450 -0.0012 0.0027 0.6660 mstd1 -0.0012 0.0016 0.4490 -0.0007 0.0018 0.6890 litdummy4 -0.0108 0.0024 0.0000 -0.0121 0.0025 0.0000 tdum2 0.0125 0.0069 0.0720 0.0131 0.0072 0.0700 Constant 0.0294 0.0049 0.0000 0.0245 0.0074 0.0010
Number of obs 557 OLS R-squared
0.9361
2SLS R-squared
0.9279
Breusch-Pagan / Cook-Weisberg test for heteroskedasticity: chi2(1) = 19.55 , Prob>chi2 = 0.0000
Table A3. Determinants of livelihood security in Jessore and Tongi (pooled). OLS 2SLS Livelihood security
Coef. Robust Std. Err.
P>t Coef. Robust Std. Err.
P>t
hlseco4 0.3724 0.0077 0.0000 0.3458 0.0181 0.0000 hlsfood 0.1905 0.0040 0.0000 0.2024 0.0065 0.0000 hlsh1 0.1573 0.0043 0.0000 0.1447 0.0067 0.0000 hlse1 0.1113 0.0063 0.0000 0.1120 0.0063 0.0000 hlsempo 0.1450 0.0034 0.0000 0.1369 0.0068 0.0000 Famsize -0.0002 0.0006 0.7080 -0.0003 0.0007 0.7010 Landhpc 0.0010 0.0090 0.2750 0.0171 0.0101 0.0920 Deprati -0.0027 0.0007 0.0000 -0.0033 0.0008 0.0000 Ageh -0.0001 0.0000 0.0050 -0.0001 0.0000 0.0040 Classcomh -0.0004 0.0002 0.0340 -0.0003 0.0002 0.1170 femh1 -0.0016 0.0019 0.3810 -0.0007 0.0019 0.7000 slitd1 -0.0007 0.0020 0.7460 -0.0003 0.0020 0.8750 mstd1 -0.0015 0.0012 0.2070 -0.0012 0.0013 0.3750 litdummy4 -0.0090 0.0016 0.0000 -0.0094 0.0016 0.0000 tdum2 0.0034 0.0035 0.3260 0.0032 0.0035 0.3560 Constant 0.0259 0.0037 0.0000 0.0279 0.0048 0.0000
Number of obs 1120 OLS R-squared
0.9493
2SLS R-squared 0.9473
Breusch-Pagan / Cook-Weisberg test for heteroskedasticity: chi2(1) = 68.53, Prob>chi2 = 0.0000