assessing poverty, risk and vulnerability: a study on the ... · by md. israt rayhan md. israt...
TRANSCRIPT
Assessing Poverty, Risk and Vulnerability: A study on the
Flooded Households in Rural Bangladesh
By Md. Israt Rayhan
Md. Israt Rayhan
Ph.D. Student
Center for Development Research Department of Economic and Technological Change (ZEF-B)
University of Bonn
Walter Flex Str.3
D-53113, Bonn, Germany
Email: [email protected] / [email protected]
Phone: (++49) 228 731852
Fax: (++49) 228 731839
Word count: Abstract (181 words), Introduction to Conclusion (3,200 words).
2
Assessing Poverty, Risk and Vulnerability: A study on the
Flooded Households in Rural Bangladesh
Abstract
Flood is a common catastrophe for Bangladesh. The combination of its geography,
population density, and extreme poverty makes Bangladeshi people vulnerable to flood
risks. This study is set forth to examine the poverty, risk and vulnerability for flood
hazards in the year 2005. Cross sectional household survey was carried out after two
weeks of the flood in four districts and 600 rural households were interviewed through
three stages stratified random sampling. A utilitarian approach is used to assess flood
vulnerability and its components: poverty, idiosyncratic and aggregate risks to capture the
effect of flood on household’s welfare. To estimate the correlates of flood vulnerability, a
set of fixed households’ characteristics are used as explanatory variables. The results
depict that elimination of poverty would increase household welfare and thus lessen
vulnerability mostly amongst its components. Poverty and idiosyncratic flood risk are
positively correlated and highly significant. Households with higher educated members,
male headed and owner of the dwelling place are less vulnerable to idiosyncratic flood
risk. Possession of arable land and small family size can reduce the poverty and
aggregate flood risk.
Key Words: Poverty, Vulnerability, Idiosyncratic risk, Aggregate risk, Flood
3
Introduction
Bangladesh consists mostly of a low-lying river delta with over 230 rivers and tributaries,
situated between the foothills of the Himalayas and the Bay of Bengal. With a population
of 123.85 million and an area of 147,570 square km, Bangladesh is one of the world’s
most densely populated countries (839 persons per square km)1. 40 percent of the
population lives below the poverty line2, and 82 percent live on less than US$ 2 per day.
About 70 percent land of the country is less than 1 meter above sea level. The
combination of its geography, population density, and extreme poverty makes
Bangladeshi people very vulnerable to risks and disasters. Flood is a frequent catastrophe
for Bangladeshi people. In the year 1987, about 40 percent of the country was flooded,
affecting 30 million people and caused about 1800 deaths. The floods in 1988 were even
more serious, covering about 60 percent of the land area, affecting about 45 million
people, and causing more than 2,300 deaths3. In 1998, over 68 percent of the country was
inundated (Ninno D. et al., 2001) and caused about 2,380 deaths. In 2000 and 2002 floods
affected approximately 20 million people. In the year 2004, devastating monsoon flood
submerged two-thirds of the country, 35.9 million people affected, 726 deaths, millions
of people made homeless4.
This study thus is set forth to examine the poverty, risk and vulnerability for flood
hazards in the year 2005, based on the three questions: (a) How vulnerable are the
flooded people? (b) Which sources of risk contribute most to flood vulnerability? (c)
Which types of interventions are most likely to reduce the flood risk and vulnerability?
1Population Census 2001, Bangladesh Bureau of Statistics, July 2003
2 Preliminary Report on Household Income and Expenditure Survey-2005, Bangladesh Bureau of Statistics,
September, 2006 3 Irrigation Support Project for Asia and the Near East (1993: 1). 4 http://www.adb.org/Documents/Economic_Updates/BAN/2004/eco-update-ban.pdf
4
Poverty is an ex-post measure which refers to being deprived of basic levels of economic
wellbeing (absolute income poverty) and human development (Dercon 2001). It also
characterized as the deprivation of capabilities (Sen 1987). Vulnerability, on the other
hand, is an ex-ante measure of household’s wellbeing and concerning about the future
poor (Alwang et al. 2001). Households are vulnerable if a shock (flood) is likely to push
them below a predetermined welfare threshold (poverty line), so vulnerability is a result
of the cumulative process of risk and response. So, the term distinguishes poverty and
vulnerability is risk (Chaudhuri 2003). The risk of a household relates to events possible
occurring, but with less than certainty (Hardarker et al. 2004), which may be upward or
downside for the individuals, households, communities and countries. This study is
focusing more on the downside risk of flood on the households of rural Bangladesh,
albeit very few fishermen or boatmen may increase their income from flooded season.
Downside risk is defined here as the estimate of the potential that a security, income,
expenditure or overall livelihoods might decline in real value if the area is flooded.
There are several papers which define and measures vulnerability to poverty and risk in
different ways. Amin et al. (1999) use panel data from Bangladesh and detect households
whose consumption tends to fluctuate with income, by controlling for household fixed
effects and aggregate variation in mean consumption. This measure is not suitable for
inter-household comparisons. Glewwe and Hall (1995, 1998) measure vulnerability in
Peru and they are interested in the response of households’ consumptions to aggregate
shocks. It seems also difficult to aggregate this measure over time with long periods of
panel data, containing both positive and negative shocks. Chaudhuri (2003), Chaudhuri et
al. (2002), Christiaensen et al. (2000) and Pritchett et al. (2000) use vulnerability to
5
expected poverty methods which suffer from the same shortcomings of headcount
measure of poverty.
This study applies the utilitarian approach to measure flood vulnerability to poverty and
risks, proposed by Ligon and Schechter (2003). Household’s welfare depends not only on
the average income or expenditure or the value of resources, but also the risk it faces. A
household with low income and facing fewer risk, might be in poverty but future well-
being may be higher than, a household with high level of income but facing higher risk.
Data and Methodology
In the year 2005, Bangladesh was affected by two types of floods, a monsoon flood was
occurred during mid August to September and a flash flood was occurred in the northern
areas during November. A cross sectional household survey was carried out after two
weeks of the floods. Four districts were chosen randomly according to the flood
proneness and damage. A three stage stratified random sampling technique was applied
for the survey, where the first stage was district, second one was the mouza (the smallest
administrative unit in rural area) and the third stage was the households. Flooded
households were detected if at least the home or homestead was submerged by flood
water. Sample size was determined by the estimated proportion formula (Cochran W.G.,
1977, p 75). After the Monsoon flood, three districts (Jamalpur, Shirajganj and
Shunamganj) were randomly chosen, a total of 450 rural households were surveyed (150
households from each district). A flash flood affected the northern part of the country in
the month of November and another 150 rural households were surveyed from the
randomly chosen district, Nilphamari. The total number of flooded households from
6
different rural areas amounted to 600. The questionnaires include some fixed household’s
characteristics such as, age, gender, education, occupation; some inter-household
variables such as, monthly income, expenditure, asset value, number of meals taken, cost
to reach market place for both before and after flood periods by recall memory method;
some inter-community variables such as, availability of primary and secondary schools,
public hospital and electricity, flood height and duration.
It is assumed to be a finite population of households, indexed by i = 1, 2,….., n and
denote the state of the world. Households want to stable their welfare overtime,
even the consequent risks occur. Household’s welfare is defined by per capita monthly
income. The distribution of household i ’s income is denoted by: )(iy . If the household
is risk averse, then the utility function will be concave and its slope will be flatter as the
wealth increases. So, the curvature of the utility function measures the household’s
attitude towards risk. Basically, the more concave the utility function, the more risk
averse the household will be (Varian, 2003 p-225). To measure vulnerability and risk for
each household, a strictly increasing and weakly concave function iU is chosen, such as:
mapping income into the real line. Given the utility function, vulnerability of
household i is defined,
)()()( iiii yEUzUyV
Where, z is some certainty-equivalent income, such that if household i had certain
income greater than or equal to this number, the household will not be regarded as
vulnerable. This study considers z as the poverty line. The poverty line is taken from the
7
nationally representative report5, which is 594.60 taka per capita per month. The
properties of utility function imply that vulnerability estimates will include mean and
variance of household’s income.
For better understand, vulnerability measure is decomposed into distinct components,
such as: poverty, aggregate risk, idiosyncratic risk and unexplained risk and measurement
error respectively. The household i ’s income at time t is denoted by i
ty , idiosyncratic
variables as i
tx and the vector of aggregate variables as _
tx .
)]()([ i
t
iii EyEUEyUV (Poverty)
))]|()([_
ti
t
ii
t
i xEyEUEyU (Aggregate Risk)
))],|(())|(([__
i
tti
t
it
i
t
i xxyEEUxyEEU (Idiosyncratic Risk)
)]()),|(([_
i
t
ii
tti
t
i yEUxxyEEU (Unexplained Risk and Measurement Error)
For a suitable choice of }{ iU , the poverty term will satisfy all the axiomatic requirements
enumerated in Foster et al. (1984). The risk terms are consistent with the ordinal
measures of risk proposed by Rothschild and Stiglitz (1970). Some additional
assumptions were taken for estimation, first, }{ iU takes the simple form
)1/()()( 1 yyU i for some parameter 0 ; as increases, the function iU
becomes increasingly sensitive to risk. The shape of utility function is characterized by
the preferences of the households. It is reflected by the curvature of the household utility
function which is defined only up to a positive linear transformation. The parameter
can be interpreted as the household’s relative risk aversion. According the
5 Poverty Monitoring Survey-December, 2004, Bangladesh Bureau of Statistics.
8
microeconomic literatures (Hardaker et al. 2004, Ligon and Schechter, 2002), it is
assumed =2. The justification can be derived from the constant relative risk aversion
(CRRA), usually defined by two functions: (1) Logarithmic: )ln( yU and (2) Power:
1
)1(yU . The power function is commonly preferred over the logarithmic functional
form, because it directly incorporates as the constant coefficient of relative risk
aversion for income, where is called the partial risk aversion coefficient. Anderson and
Dillon (1992) proposed a classification of degree of level of risk aversion, based on the
magnitude of the relative risk aversion coefficient, such as: (i) =0.5, hardly risk averse,
(ii) =1.0, somewhat risk averse, (iii) =2.0, rather risk averse, (iv) =3.0, very risk
averse, (v) =4.0, extremely risk averse. This study considered =2.0, assuming
households are rather risk averse to flood, for decision making on their livelihoods, crop
pattern, education, savings, and overall income-expenditure routine from the previous
experience of flood (downside risk) disastrous. It is also assumed that
i
t
i
tt
ii
tti
t vxxxyE ),|(_
, where ,,( t
i ) a vector of unknown parameters
to be estimated. Here, }{ i shows the influence of household’s fixed characteristics on
predicted per capita income and restricted to sum to zero, }{ t captures the effect of
changes in aggregates and }{ is the vector of parameters for household’s idiosyncratic
variables, i
tv is a disturbance term equal to the sum of both measurement error in income
and prediction error. In a stationary environment, the unconditional expectation of
household i ’s income is estimated by,
T
t
i
t
i
t yT
Ey1
.1
For this analysis, is chosen so
9
as to optimally predict i
ty in a least square application. The utility from perfect equality
in a risk-less society is equal to 1. So, the percentage welfare loss from vulnerability is
equal to the size of vulnerability. After estimating vulnerability measure, the percentage
of welfare loss can be divided to some components of vulnerability, such as, poverty,
aggregate risk, idiosyncratic risk and measurement error. To look at the correlates, some
fixed household characteristics are regressed over each component and bootstrap standard
errors for the coefficients are also measured. This study uses STATA (version 9) software
for data analyses.
Discussion
The study results begin with the poverty level measurements at before and after flood
periods to delineate the affect of flood. Table 1 shows the poverty level in overall sample
and four selected districts using before and after flood per capita income. Above 16
percent of flooded households fall into poverty after flood. The drastic change into
poverty occurs in Jamalpur district by the monsoon flood, where head count poverty rate
fluctuates by 29 percent. Households from Sunamganj district face comparatively less
disastrous effect of flood. Initial period (before flood) poverty level was the highest in
Nilphamari district (71.33 percent) and it is augmented by 15 percent due to flood.
Households those are not currently poor, may also be counted as vulnerable, some events
(such as, flood, a bad harvest, illness of main earner) could push them into poverty. Jalan
and Ravallion (1999), using a six year panel data from rural households of China,
Table 1 could be replaced here
10
investigate chronic and transient poverty with the classifications: persistently poor
(households whose expenditures in each period below the poverty line), chronically poor
(mean expenditures over all periods less than the poverty line but not poor in each
period), transiently poor (mean expenditures over all periods above the poverty line but
experiencing at least one episode of poverty), and never poor. Authors found that
proportion of transient poor is much higher than chronic poor and never-poor.
Table 2 shows that in overall sample 56.38 percent of households were always poor, but
75 percent of the sample experienced at least one episode of poverty between before
flood and after flood. The individual district also shows the higher proportion of falling
into poverty in one episode than the proportions of always poor and never poor.
Summary statistics of some variables, used to examine the correlates of vulnerability, are
given in the table 3. The after flood average income fall below the poverty line (594.60
taka). The inequality also rises due to flood by 20 percent. The average educational year
of the flooded households is up to primary schooling level. Majority of the households
are male headed (89%). Average land holding is quite low for the surveyed households
(less than one acre per capita). On an average each household possesses five members.
Vulnerability is estimated using Ligon and Schechter (2003) methodology. To assess the
correlates and significance of the components of vulnerability, each component is
regressed on a set of fixed household characteristics, such as, educational year of highest
Table 2 could be replaced here
Table 3 could be replaced here
11
educated household member, dummy variable gender of household head (1=male, 0=
female), age and square of age of household head, arable land per capita, ownership of
the dwelling place and number of household members. Linear relationship is assumed
and Ordinary Least Square (OLS) estimates of coefficients are given in the table 4.
To check whether the omitted variables are significant or not, Ramsey RESET test is
performed using powers of the fitted values of vulnerability (Gujarati, 2003, p-521)6. The
null hypothesis is: Ho: model has no omitted variables and the test result shows that F-
statistic is 2.40 with probability value (p-value) 0.671, so the conclusion would be that
the null hypothesis cannot be rejected at 5 percent level of significance. The next step is
to test for heteroscedasticity because the survey was cross-sectional. Breusch-Pagan-
Godfrey test is executed (Gujarati, 2003, p-411)7 for checking heteroskedasticity with the
null hypothesis, Ho: Constant variance, the p-value of chi-square test statistic comes out
as highly significant (0.00). From the test result, it is depicted that there is
heteroscedasticity in the error variance. The next step is to regress vulnerability and its
factors on the fixed set of households’ characteristics resolving heteroscedasticity
problem. This analysis performs the regression with robustness8 and find out the
bootstrap standard errors with 500 replications. The results are given in table 5.
6 Cited in, Gujarati, Damodar N., 2003, Fourth edition, Basic Econometrics, McGraw-Hill.
Ramsey, J.B., Tests for Specification Errors in Classical Linear Least Squares Regression Analysis, Journal
of the Royal Statistical Society, series B, vol. 31, 1969, pp. 350-371.
7 Cited in Gujarati. Breusch, T. and A. Pagan, A Simple test for Heteroscedasticity and Random Coefficient
Variation, Econometrica, vol. 47, 1979, pp. 1287-1294. Godfrey, L., Testing for Multiplicative
Heteroscedasticity, Journal of Econometrics, vol. 8, 1978, pp. 227-236. 8 White’s heteroscedasticity-consistent variance and standard error test (Gujarati, 2003, p-417)
Table 4 could be replaced here
12
Multicollinearity of the regressed variables also checked by Tolerance and variance
inflation factor (VIF), the values are .83 and 5.62 respectively, which illustrates there is
no collinearity among the explanatory variables9. Table 6 is accumulating the information
on the correlates of vulnerability and each of the components, such as poverty, aggregate
risk, idiosyncratic risk and unexplained risk of flooding with the remedial test for
heteroscedasticity and bootstrapping.
It is depicted that the correlates of flood vulnerability are apparently similar to the
correlates of poverty (for significant variables) which is also the noteworthy component
for defining vulnerability. Additionally, the significant variables in poverty and aggregate
risk share the same sign of coefficients. Aggregate shocks from flood are the same for all
households, so the poor households may experience greater impact on their utility from
this part of risk. The household’s idiosyncratic risk is measured by three observed
components from two periods (before and after flood), such as: asset value, number of
meal taken and cost to reach market place. To assess aggregate risk some community
based variables are used, such as: availability of primary and secondary schools, public
hospital, electricity and flood shelter.
Education is the most significant variable to define vulnerability. The households with
higher educated member are less vulnerable. The increase of 1 unit educational year of
highest educated member of household will decrease 24 unit vulnerability to flood, most
of this reduction will appear in poverty, idiosyncratic and aggregate risk also decrease
substantially. The gender of household head has no significant effect on vulnerability, but
9 from the rule of thumb used by Kleinbaum et al. (1988), p-210
Table 6 could be replaced here
13
reduces idiosyncratic risk significantly. The reason may be that male headed households
acquire the proper intra-household resource allocation at household-specific flood risk.
Households with older heads face higher idiosyncratic vulnerability but after a certain
point their experience helps them to reduce such kind of vulnerability (negative
coefficient for age square). Arable land holding shows significant relationship with the
poverty and aggregate risk at diminishing rate. Perhaps the more availability of land leads
the households to rotate and diversify their crop choice, hence lessen poverty and
aggregate risk. It is expected that households with more arable land and involved in
agriculture might face risk from unobservable sources, such as inundation of crops by
flood water, this analysis shows the similar pattern but at insignificant way. Ownership of
the dwelling place also has the significant negative relation with each type of
vulnerability, even reduce unexplained risk considerably. With the increase of family
size, per capita income or allocated resource units will be lower, thus vulnerability,
poverty and aggregate risk may be aggravated significantly, but the goods of common
share may help to minimize idiosyncratic risk.
Poverty and aggregate risk due to flood have strong positive correlation. It can be
described by the diminishing marginal utility principle that the poor are mostly affected
by the aggregate flood risk, which is uniformly distributed into the utility of income of
the households. Poverty and idiosyncratic risk are positively correlated and highly
significant. The poor has less asset and selective ways of earning, if flood ruined their
crops or hinders the way of earning then households would be more in poverty and may
Table 7 could be replaced here
14
fall into the vicious circle of debt. Unexplained risk is negatively correlated with
idiosyncratic risk at 10 percent level of significance.
Conclusion
This study adopts a utilitarian approach to assess flood vulnerability, poverty and risks to
capture the effect of flood on household welfare. Flood in Bangladesh is a common
calamity, even for small scales of monsoon and flash floods in the year 2005, the
surveyed households from four districts have drastic changes in poverty levels. The
estimated vulnerability and risks are also extremely high for flooded households.
The results suggest that in the elimination of poverty would increase household welfare
and thus lessen vulnerability mostly amongst its components; this finding supports the
view from Ligon and Schechter (2002). Education and ownership of dwelling place are
found to be the significant variables to reduce poverty and flood risks. Households with
higher educated member are considerably less poor, and significantly less vulnerable to
both aggregate and idiosyncratic sources of flood risk, as Pritchett et al. (2000) show that
average vulnerability rate gets lower if the educational level of household heads is higher.
Perhaps this is because educated members can acquire better coping strategies during
flood. Ownership of dwelling place also significantly reduces vulnerability and its
components, so Bangladesh disaster management authority could authentically look after
this step while they rehabilitate the flood victims. Male headed households are facing
higher aggregate risk than the female headed households, which resemblance with the
results of Glewwe and Hall (1998). Households which have larger family size are
significantly more vulnerable and poorer, confronting higher level aggregate risk but
15
lower level of idiosyncratic risk having benefited from the share of common goods. This
contrasts with the result of Ligon and Schechter (2002), who find that households with
smaller family size experience lower level of idiosyncratic risk.
Some policy implications to the target groups may mitigate future flood risks. The aid
programs could intend to reach the transient poor rather always poor and provide the
opportunities not only to the current poor but to those households that experience flood
shocks. Social protection, social insurance or micro credit schemes for the landless
households might motivate them to start small scale business or farming. Food for
education policy already implemented in Bangladesh, which supposed to be monitored
properly to enhance the efficiency of flooded households, hence reduces future risks.
16
References
Alwang, J., P. B. Siegel, and S. L. Jørgensen (2001) Vulnerability: A view from different
disciplines, Social Protection Discussion Paper No. 0115, Washington, D.C.: World
Bank.
Amin, S., A. S. Rai, and G. Topa (1999) Does microcredit reach the poor and vulnerable?
Evidence from northern Bangladesh, Working Paper 28, Center for International
Development, Harvard University.
Anderson, J.R. and Dillon, J.L (1992) Risk Analysis in Dryland Farming Systems,
Farming Systems Management Series No. 2, FAO, Rome.
Baulch, B. and J. Hoddinott (1999) Economic Mobility and Poverty Dynamics in
Developing Countries, Journal of Development Studies, vol. 36, issue 6, pp. 1-24.
Breusch, T. and A. Pagan (1979) A Simple test for Heteroscedasticity and Random
Coefficient Variation, Econometrica, vol. 47, pp. 1287-1294.
Chaudhuri, S., J. Jalan and A. Suryahadi (2002) Assessing Household Vulnerability to
Poverty from Cross-sectional Data: A Methodology and Estimates from Indonesia,
Discussion paper no. 0102-52, Department of Economics, Columbia University, New
York.
Chaudhuri, S. (2003) Assessing Vulnerability to Poverty: concepts, empirical methods,
and illustrative examples, Columbia University, New York, mimeo.
Christiaensen, L. J., R. N. Boisvert and Hoddinott (2000) Validating Operational Food
Insecurity Indicators against a Dynamic Benchmark: Evidence from Mali, Policy
Research working paper no. 2471, Poverty Reduction and Social Development Unit,
World Bank.
Cochran, W. G. (1977, third edition) Sampling Techniques, New York: John Wiley &
Sons.
Dercon, S. and P. Krishnan (2000) Vulnerability, Seasonality and Poverty in Ethiopia.
Journal of Development Studies, vol. 36, issue 6, pp. 25–53.
Dercon, S. (2001) Assessing Vulnerability to Poverty, Center for the Study of African
Economies, Department of Economies, Oxford University.
Foster, J., J. Greer and E. Thorbecke (1984) A class of decomposable poverty measures,
Econometrica, vol. 52, issue 3, pp. 761–766.
17
Glewwe, P. and G. Hall (1995) Who is most vulnerable to macroeconomic shocks?:
Hypotheses tests using panel data from Peru, Living Standards Measurement Study
Working Paper no. 117, The World Bank, Washington, D.C.
Glewwe, P. and G. Hall (1998) Are some groups more vulnerable to macroeconomic
shocks than others? Hypothesis tests based on panel data from Peru, Journal of
Development Economics, vol. 56, issue 1, pp. 181–206.
Godfrey, L. (1978) Testing for Multiplicative Heteroscedasticity, Journal of
econometrics, vol. 8, pp. 227-236.
Gujarati, Damodar N. (2003, Fourth edition) Basic Econometrics, New York: McGraw-
Hill
Hardaker, J.B., R.B.M. Huirne, J.R. Anderson and G. Lien (2004, second edition) Coping
with Risk in Agriculture, Oxfordshire, UK: CABI publishing.
http://www.adb.org/Documents/Economic_Updates/BAN/2004/eco-update-ban.pdf (Last
access 5th August, 2007)
Irrigation Support Project for Asia and the Near East (1993) Flood response and
guidelines on planning flood proofing, Bangladesh Flood Action Plan FAP 14/ FAP 23.
Jalan, J. and M. Ravallion (1998) Transient Poverty in Post-Reform China, Journal of
Comparative Economics, vol. 26, pp. 338-357.
Jalan, J. and M. Ravallion (1999) Is Transient Poverty Different? World Bank,
Washington, D.C., mimeo.
Kamanou, G. and J. Morduch (2002) Measuring vulnerability to poverty, Discussion
paper no. 2002/58, United Nations University (WIDER).
Kleinbaum, D. G., L. L. Kupper and K. E. Muller (1988, second edition), Applied
Regression Analysis and other Multivariate Methods, PWS-Kent, Boston, Mass.
Ligon, E. and L. Schechter (2002) Measuring Vulnerability: The Director’s Cut,
Discussion paper no. 2002/86, United Nations University (WIDER).
Ligon, E. and L. Schechter (2003). Measuring vulnerability, The Economic Journal, vol.
113, issue 486, pp 95-102.
Ninno, C.D. Dorosh, P.A. Smith, L.C. and Roy D.K. (2001) The 1998 floods in
Bangladesh : disaster impacts, household coping strategies, and response, Research
Report, International Food Policy Research Institute Washington, D.C.
Population Census (2001), Bangladesh Bureau of Statistics, Dhaka, Bangladesh.
18
Poverty Monitoring Survey (2004) Bangladesh Bureau of Statistics, Dhaka, Bangladesh.
Preliminary Report on Household Income and Expenditure Survey (2005) Bangladesh
Bureau of Statistics, Dhaka, Bangladesh.
Pritchett, L., A. Suryahadi and S. Sumarto (2000) Quantifying vulnerability to poverty:
A proposed measure, with application to Indonesia, SMERU Working paper.
Ramsey, J.B. (1969) Tests for Specification Errors in Classical Linear Least Squares
Regression Analysis, Journal of the Royal Statistical Society, series B, vol. 31, pp. 350-
371
Ravallion, M. (1988) Expected poverty under risk-induced welfare variability, Economic
Journal, vol. 98, issue 393, pp. 1171–1182.
Rothschild, M. and J. E. Stiglitz (1970) Increasing risk: I. A definition, Journal of
Economic Theory, vol. 2, pp. 225–243.
Sen, A.K (1987) The Standard of Living, Cambridge: Cambridge University Press.
Suryahadi, A., Sumarto, S. (2001) The Chronic Poor, the Transient Poor, and the
Vulnerable in Indonesia Before and After the Crisis, SMERU Working Paper, SMERU
Research Institute, Social Monitoring & Early Response Unit, Jakarta.
Townsend, R. M. (1994) Risk and insurance in village India. Econometrica, vol. 62, issue
3, pp. 539–591.
Varian, Hal R. (2003, sixth edition) Intermediate Microeconomics: A Modern Approach,
New York: W.W. Norton & Company.
von Neumann, J. and Morgenstern, O. (1944) Theory of Games and Economic Behavior,
Princeton, NJ: Princeton University Press.
19
Table 1: Income poverty (in percentage) for flooded households
Indicators Head count poverty
Before Flood After Flood
Overall 57.38 73.83
Districts Shirajganj 47.06 66.01
Jamalpur 58.67 87.33
Shunamganj 52.67 56.0
Nilphamari 71.33 86.0
n=600, Sample is weighted by household size, Poverty measures followed the formula
proposed by Foster et al. (1984) with =0, Poverty line is the absolute poverty line
based on food energy intake (FEI) method, the changes between before and after flood
unit price of most commonly consumed items do not show significant t-statistic, so per
capita incomes are not weighted by any consumer price index
20
Table 2: Classification of transient and chronic poverty (in percentage) for flooded
households
Chronically poor (mean per capita
income below poverty line)
Transiently poor only
(mean per capita
income above the
poverty line)
Never
poor
Always poor
(before and
after flood)
Not persistently
poor
Overall 56.38 13.27 5.05 25.30
Shirajganj 47.06 11.76 7.2 33.98
Jamalpur 57.33 25.34 6.0 11.33
Shunamganj 50.0 6.0 2.67 41.33
Nilphamari 71.33 10.0 4.67 14.0
21
Table 3: Summary of Variables
Variables Value
Monthly income per capita before flood (mean) in Taka*
750.01
Gini coefficient for income before flood .396
Monthly income per capita after flood (mean) in Taka
545.45
Gini coefficient for income after flood .596
Educational year of highest educated member (mean) 5.02
Male headed households (percentage) 89%
Age of household head (mean) 43.60
Cultivated land per capita in acres (mean) 0.078
Ownership of house (percentage) 53.57%
Family size (mean) 5.24
*80 Taka (Bangladesh currency) =1 Euro (at field survey time, 2005)
22
Table 4: OLS Estimation of vulnerability on some fixed characteristics of households
Covariates Coefficient Standard
Error
t-
statistic
P>t [95% Confidence
Interval]
Education -23.82 5.04 -4.72 0 -33.73 -13.90
Male 35.80 61.94 0.58 0.56 -85.84 157.45
Age -3.03 7.25 -0.42 0.67 -17.27 11.20
Age square 0.008 0.07 0.11 0.91 -0.13 0.15
Arable land per
capita
-24.11 71.83 -0.34 0.73 -165.21 116.96
Ownership of
house
-80.39 37.74 -2.13 0.03 -154.52 -6.26
Family size 10.68 8.74 1.22 0.22 -6.50 27.86
23
Table 5: Correlates of vulnerability in income with bootstrap standard errors and robust
estimates
Covariates Observed
Coefficient
Bootstrap
Standard
Error
z-
statistic
P>z Normal-based [95%
Confidence Interval]
Education -23.82 5.08 -4.68 0 -33.78 -13.85
Male 35.80 52.25 0.69 0.49 -66.60 138.21
Age -3.03 6.74 -0.45 0.65 -16.25 10.19
Age square 0.008 0.06 0.12 0.90 -0.11 0.13
Arable land per
capita
-24.11 38.11 -0.63 0.52 -98.82 50.59
Ownership of
house
-80.39 40.80 -1.97 0.04 -160.78 -0.41
Family size 10.68 6.05 1.76 0.07 -1.18 22.55
24
Table 6: Correlates of vulnerability, poverty, risks
Vulnerability Poverty Aggregate
Risk
Idiosyncratic
Risk
Unexplained
Risk
Variables Coefficient Coefficient Coefficient Coefficient Coefficient
Education -23.82***
-11.80***
-7.89***
-0.31***
-3.81
(5.08) (1.74) (1.33) (0.04) (4.23)
Male 35.80 -13.61 14.85 -2.01**
35.01
(52.25) (20.66) (16.71) (0.89) (44.63)
Age -3.03 2.15 -1.64 0.14**
-3.68
(6.74) (2.23) (1.58) (0.06) (4.99)
Age square 0.008 -0.02 0.02 -0.001*
0.001
(0.06) (0.02) (0.01) (0.001) (0.04)
Arable
Land per capita
-24.11
(38.11)
-48.33*
(25.44)
-23.22**
(11.25)
-0.29
(0.47)
47.73
(48.34)
Ownership of house -80.39**
-33.48***
-24.92***
-1.20***
-20.77*
(40.80) (10.45) (7.97) (0.39) (10.52)
Family size 10.68*
5.86**
4.10**
-0.17*
0.88
(6.05) (2.68) (1.99) (0.09) (5.56)
2R .63 .58 .59 .74 .31
Numbers in parenthesis are bootstrapped standard errors, ***-significant at 1% level, **-
significant at 5% level, *- significant at 10% level, n=600
25
Table 7: Correlations between elements of vulnerability in per capita income
Poverty Aggregate Risk Idiosyncratic
Risk
Unexplained
Risk
Poverty 1.00
Aggregate Risk 0.747***
1.00
Idiosyncratic
Risk
0.388***
-0.291 1.00
Unexplained
Risk
0.042 0.018 -0.148*
1.00
Spearman rank correlations technique is chosen for above table. ***- significant at 1%
level, **- significant at 5% level, *- significant at 10% level