Quantifying vulnerability to poverty in a developing economy.
Haq, Rashida
In recent years, development policy has increasingly been
recognised that a household's sense of well-being depends not just
on its average income or expenditures, but also on the risks it faces.
Hence vulnerability is a more satisfactory measure of welfare.
Application of the methodology to data from Pakistan shows that in 2010
households faced on average a 20.7 percent of poverty level while 34.4
percent chance of becoming poor in the future. When decomposing
vulnerability into poor and non-poor households, it was figured out that
95 percent poor household are vulnerable and 18 percent non-poor
households are vulnerable. The model measuring determinants of
vulnerability indicate that households head with no education, large
family size, high number of dependents, poverty status, lack of
productive assets, agricultural shocks, high cost of shock and rural
residence have significantly higher probability of being poor in future.
Finally, it can be concluded that policy makers should be aware of
social risk management strategies as an integral part of poverty
alleviation programme.
JEL Classification: C21, C23, 132
Keywords: Poverty, Vulnerability, Shocks, Pakistan
1. INTRODUCTION
The concept of vulnerability extends the idea of poverty to include
idiosyncratic as well as aggregate risks which can be defined as the
probability of being in poverty or to fall deeper into poverty in the
future. It can be categorised on the micro-and macro level where macro
vulnerability refers to worldwide threats to social welfare, e.g.
globalisation and recent international financial crises. Conversely,
micro vulnerability refers to the household level risks including health
risks, economic shocks, social shocks, natural disasters, and
demographic shocks [Tesliuc and Lindert (2004)]. To assess and estimate
vulnerability to poverty, various approaches had been proposed. First,
vulnerability can be seen as a probability of falling into poverty in
near future [Chaudhuri (2003); Christaensen and Subbarao (2005)]. The
other ways of measuring vulnerability consider it as low expected
utility [Ligon and Schechter (2003)] and vulnerability as uninsured
expose to risk, i.e., measures of cost, in terms of consumption [Tesliuc
and Lindert (2004)]. The basic idea is that the state of poverty at a
given point actually is not sufficient for assessing poverty and for
drawing results to design poverty reduction programs. Households face
various risks and do not know whether any possible shock will hit them
in future. So the assessment of poverty at a given point in time is a
static approach, not considering possible changes in the future. By
assessing vulnerability it refers to the dynamic perspective, it is
explicitly forward looking and tries to include the risks that may push
people into poverty in future.
Although there are different concepts of poverty and vulnerability
which are closely related due to two established facts; (i) the poor are
typically most exposed to diverse risks, and (ii) the poor have the
fewest assets to deal with these risks. However, the importance of
vulnerability because if policymakers design poverty alleviation
policies in the current year on the basis of a poverty threshold of
income or consumption in the previous year, 'the poor' who
receive income support may have already escaped from poverty and
'the non-poor' who do not receive such support may have
slipped into poverty due to various unanticipated shocks. Hence,
assessing vulnerability helps to distinguish between ex-ante poverty
prevention interventions and ex-post poverty alleviation strategies like
mitigation and coping arrangements [Holzmann and Jorgensen (2000)].
Households and communities in Pakistan face the risks of suffering
from different types of shocks. Some shocks affect communities as a
whole referred as covariate shocks such as economic and financial crises
and natural disasters. Others affect one or a few households noted as
idiosyncratic shocks, such as a death of a household member or loss of a
job. The analysis of vulnerability proposed is crucial for determining
which programs to have in place and when to introduce them or adjust
their levels and coverage. To make these decisions, policymakers need to
have access not only to macro-economic indicators, but also to
indicators that provide an understanding of household-level of
vulnerability and risk profiles and risk management mechanisms,
particularly for the poor. The vulnerability analysis can be useful in
the context of Pakistan, given the large proportion of poor people and
the low level of human capital as Pakistan is categorised in low human
development countries ranked at 146 out of 187 countries with human
development index value of 0.537 [UNDP (2014)].
The main purpose of this paper is therefore to: (a) generate the
poverty vulnerability indices of the households using expected and
variance of consumption expenditures in Pakistan; (b) estimate
incidence, intensity and severity of poverty; (c) contribution of
vulnerability to poverty (d) estimates of vulnerability and poverty
across different socio economic groups; and (e) determinants of
vulnerability. The analysis carried out in the paper uses Pakistan Panel
Household Survey (PPHS)-2010. Expected results from the analyses are
judged to be more relevant to poverty policy formulation in Pakistan.
The rest of the paper is structured as follows. Section 2 provides
review of literature. Section 3 outlines the details of the methodology
and data. The econometric and other relevant results are presented in
Section 4. Section 5 concludes the study highlighting some of the policy
issues for reducing poverty and vulnerability to poverty in Pakistan.
2. REVIEW OF LITERATURES
The existing literature, which intends to estimate the aggregate
vulnerability of households, has been pioneered by Townsend (1994) and
Udry (1995), who were some of the first using panel data to analyse,
whether households are able to insure their consumption against
idiosyncratic income fluctuations over space and time. In this spirit
several studies followed analysing consumption fluctuations over time
[e.g. Dercon and Krishnan (2000); Jalan and Ravallion (1999); Morduch
(2005)], concluding that households are partly but not fully capable of
insuring consumption against income fluctuations. A severe drawback of
this literature is that it relies on panel data, which is very limited
for developing countries.
The second strand of empirical literature on vulnerability, which
estimates the impact of selected shocks on households' consumption,
has its limitations because information on idiosyncratic and covariate
shocks is in most households' surveys very limited or sometimes
missing [Gunther and Harttgen (2005)]. As a consequence most authors
have only been able to focus on the impact of selected shocks on
consumption [Dercon and Krishnan (2000); Gertler and Gruber (2002)].
Moreover, these studies have rarely been able to analyse the impact of
these shocks on the vulnerability of households. Several measurements to
analyse vulnerability to poverty have recently been proposed, empirical
studies are still rare as the data requirements for these measurements
are not met by the surveys that are available for most developing
countries.
In recent debates on poverty in Pakistan, the issue of
vulnerability have been mentioned frequently [Pakistan (2010); World
Bank (2002)]. Furthermore, the poverty incidence in KP province is
higher and agriculture is more risky than in other parts of Pakistan.
Kurosaki (2010) showed that most of the vulnerability measures summarise
micro-level information on consumption and income, since the welfare of
an individual depends not only on consumption but also on other
non-monetary aspects such as education and health in case of KP
province.
The literature on risk and vulnerability by using a cross-section
survey to map and quantify shocks from all sources, ex-post responses
and outcomes for a sample of relatively poor Pakistani households was
explored by Heltberg and Niels (2009). They found high incidence and the
cost of shocks, with health-related shocks easily the worst. These
findings add to the evidence that health shocks often dominate and
impose severe coping costs in terms of medical expenses while relying
mostly on informal and ad hoc responses: informal borrowing, spending
savings, and working more were the most frequently used responses. The
extent of household vulnerability to poverty in Pakistan was also
estimated by Jamal (2009) who found that about 52 percent population was
vulnerable to poverty during 2004-05 while rural headcount ratio in
terms of household vulnerability is relatively high as compared to the
vulnerability incidence in urban areas. Although monetary poverty has
declined during the period 2001-05, the relative incidence of
vulnerability has increased from 50 percent in 2001 to 52 percent in
2005.
This review of literatures on shocks, poverty and vulnerability
indicating direct implications for welfare loss due to health shock,
agricultural shock and natural disaster etc., ultimately, translated in
income shock.
3. DATA AND METHODOLOGY
(a) Data Collection
Ideally, vulnerability measurement would require a long panel data.
However, for many developing countries, reliable panel data are scarce
and only cross-sectional survey data are available. Pakistan is no
exception in this regard. The absence of panel data obliges us, in our
assessment of vulnerability to poverty in Pakistan, to adopt the
approach proposed by Chaudhuri (2003) which is particularly designed for
cross-section data.
This study is based on a cross-section data from 'Pakistan
Panel Household Survey (PPHS)-2010' conducted by Pakistan Institute
of Development Economics financed by the World Bank. The
households' sample of PPHS was selected on the basis of a
multistage stratified sampling procedure. The survey consists of 16
districts from four provinces (Punjab, Sindh, Khyber Pakhtunkhwa (KP)
and Balochistan) with their urban and rural counterparts. The districts
included are Attock, Faisalabad, Hafizabad, Vehari, Muzaffargarh, and
Bahawalpur in Punjab; Badin, Mirpur Khas, Nawab Shah, and Larkana in
Sindh; Dir, Mardan, and Lakki Marwat in Khyber Pakhtunkhwa (KP); and
Loralai, Khuzdar, and Gwadar in Balochistan.
The total sample size of PPHS-2010 was 4142 households; 2800 in
rural and 1342 in urban (Punjab1878; Sindh 1211; KP 601 and Baluchistan
452). After cleaning the data (deleting outliers, no responses and
missing cases) a sample of 3500 households was selected for final
analysis. The analysis was based on this information together with other
information concerning characteristics of the head of the household
(i.e., individual characteristics such as sex, age, education) and
household characteristics, like household size (taken as adult
equivalent), dependency ratio, (1) poverty status, (2) quality of
house-whether mud house or brick house, agricultural land ownership,
livestock ownership (large or small animals), log per adult equivalent
consumption expenditure, in addition to community characteristics likes,
regions and provinces. The shocks variables included are divided into a
number of broad categories: natural/agricultural; economic; political/
social/legal; and demographic/life-cycle shocks that inflict welfare
loss. Natural/ agricultural shocks include flooding, drought, fire,
earthquake but also erosion and pestilence affecting crops or livestock.
Economic shocks include business closures, mass layoffs, job loss, wage
cuts, loss of remittances and other reasons. Social shocks in Pakistan
include court cases and bribery, as well as long duration general
strikes, violence, crime and political unrest. Health/life-cycle shocks
include death, injury and illness of household members. The survey
distinguished between death of the primary income earner and death of
other household members. Similarly, the respondents were also asked
whether the household was affected by idiosyncratic or covariate shocks
and with the value of cost of burden.
So finally, in addition to these questions about specific shocks,
households were also asked about the most important coping strategies to
manage the reduction in income such as (i) asset-based strategies (sale
of assets including land, livestock and stored crop); (ii)
assistance-based strategies (help from friends and relatives); (iii)
borrowing-based strategies (from friends and relatives, banks, NGOs and
money lenders); and (iv) behaviour-based strategies (decrease
food/non-food consumption, increase labour supply particularly of women
and children, dropped out of school and beggary).
(b) Methodology
In this section the detailed estimation procedure of the analysis
of vulnerability to poverty in Pakistan is delineated as follow:
Model Specifications
(i) Foster, Greer, Thorbecke (1984) Poverty Measure
The methodology that is used in this analysis is the class of
poverty measures by Foster, Greer and Thorbecke (FGT) which is widely
used because they are consistent and additively decomposable [Foster, et
al. (1984)].
[P.sub.[varies]] = 1/n [summation].sup.q.sub.i=1[([Z -
[C.sub.i]/Z).sup.[varies]] (1)
Where Z is the poverty line measured as per adult equivalent
consumption expenditure of Rs 1671.89, Ci is the ith welfare indicator
measured in terms of per adult equivalent consumption expenditure, n is
the total households, q is the number of households who are below the
poverty line, and [alpha] [greater than or equal to] 0 is a
'poverty aversion' parameter. The FGT poverty measure formula
delivers a set of poverty indices, i.e. incidence (a = 0), intensity or
poverty gap ([alpha] = 1) and severity of poverty ([alpha] = 2).
(ii) Vulnerability--the Probability of Being Poor in the Future
Modeling the Probability of Becoming Poor: In order to analyse the
effect of some idiosyncratic and covariate shocks on households'
consumption expenditures, the approach proposed by Chaudhuri (2003) and
Chaudhuri, et al. (2002) and Suryahadi and Sumarto (2003) developed
particularly for cross-section data is used. Vulnerability in this
context is defined as expected poverty, or in other words as the
probability that a household's consumption will lie below the
predetermined poverty line in the near future. Following Chaudhuri
(2000), for a given household h, the vulnerability is defined as the
probability of its consumption being below poverty line at time t + 1:
[V.sub.ht] = [P.sub.r](In[C.sub.h,t+1] < In z | [X.sub.h]) (2)
where [V.sub.ht] is vulnerability of household h at time t,
[C.sub.h t+1] denote the consumption of household h at time t+1 and Z
stands for the poverty line.
For generating per adult consumption expenditure ([C.sub.h]) of h
household is given as:
Ln[C.sub.h] = [X.sub.h] [beta] + [e.sub.h] (3)
where [X.sub.h] represents individual and household
characteristics, [beta] is a vector of parameters, [e.sub.h] is the
error term.
Suppose the variance of [e.sub.h] is given by:
[[sigma].sup.2.sub.e,h] = [X.sub.h] [theta] (4)
To measure the parameters of [beta] and [theta], a three-stage
feasible generalised least square (FGLS) procedure is used. In the first
stage, equation 3 is to be estimated with ordinary least square (OLS)
method and the square of the generated error terms are to be regressed
against the independent variables to generate the predicted values of
the error terms.
[[??].sup.2.sub.OLS,h] = [X.sub.h][theta] + [[eta].sub.h] (5)
The predicted values of the error terms in Equation 5 are used to
transform the same equation in a manner specified below:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (6)
Equation 6 will be estimated using OLS method to obtain an
asymptotically efficient FGLS estimate denoted as
[[theta].sup.[??].sub.FGLS]. In this case,
[X.sub.h][[theta].sup.[??].sub.FGLS] is a consistent estimate of,
[[sigma].sup.2.sub.e,h] variance of the idiosyncratic component of
households' consumption expenditures. Therefore, equation 3 is to
be transformed with FGLS
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (7)
To obtain
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (8)
Equation 8 is to be estimated using FGLS method, and it yields a
consistent and asymptotically efficient estimate of [beta]. The expected
log consumption can be estimated by using the estimates of
[[beta].sup.^] and [[theta].sup.^] In this case, it can be noted as:
[E.sup.^] [In[C.sub.h] |[X.sub.h] | = [X.sub.h [[beta].sup.^] (9)
The variance of log consumption expenditure for each of the
[h.sup.th] household is given as:
[V.sup.^][In[C.sub.h] | [X.sub.h] = [[??].sup.2.sub.h] =
[X.sub.h][[theta].sup.^] (10)
The vulnerability level of h household, which is the probability
that household h with characteristics [X.sub.h] will be poor in the
future can be estimated by assuming that households' consumption
expenditures are log normally distributed. Therefore, suppose (.) [PHI]
denotes the cumulative density of the standard normal distribution,
probability of vulnerability can be computed as:
[V.sup.^.sub.h] = [P.sup.^.sub.r] (In[C.sub.h] < In z |
[X.sub.h]) = [PHI] [In z - [X.sub.h][[beta].sup.^]/[square root of
[x.sub.h] [[theta].sup.^] (11)
According to Chaudhuri, et al. (2002), a vulnerability threshold of
0.5 is applied, indicating that a household with a 50 probability of
falling into poverty is vulnerable to poverty at least once in the next
years.
(iii) Measuring Determinants of Vulnerability
Risk signifies the possibility of adverse effects in the future. It
derives from the interaction of social and environmental processes, from
the combination of physical hazards and the vulnerabilities of exposed
elements. The negative evens are not the sole driver of risk, and there
is high confidence that the levels of adverse effects are in good part
determined by the vulnerability and exposure of societies and
social-ecological systems [UNISDR (2004)]. A logistic regression model
as specified by Gujarati (2012) is used to examine the determinants of
vulnerability to poverty in Pakistan. The vector of independent
variables measuring individual characteristics for household head
(gender, age, education), households characteristics (household size,
dependency ratio, poverty status, housing quality, productive assets),
transitory components (agricultural, social, economic and health shocks,
four types of coping strategies) and place of residence (urban/rural and
four provinces).
4. ANALYSIS
In order to determine the effect of some idiosyncratic and
covariate variables on households' consumption expenditures, the
approach of Chaudhari (2000) to generate vulnerability indices with
single point consumption data is used. At the same time to get three
poverty estimates, a class of decomposable poverty measures by Foster,
Greer and Thorbecke (FGT) has been adopted. It is worth noticing that
the target of policy in this paper is a population of households, not
individuals, simply because the data on consumption expenditures are
obtained from the household surveys. This means that if we consider a
household as poor or vulnerable every member in this household is deemed
equally poor or vulnerable. The poverty and vulnerability estimates
based on micro data from PPHS-2010 are presented in Table 1. It was
analysed that 20.6 percent households in Pakistan are poor while 34.4
percent households are vulnerable to poor in future which is much higher
than the point-in-time estimates of poverty, thus, signifies the
importance of forward looking poverty analysis. The distribution of
population by poverty status can be decomposed in vulnerable and
non-vulnerable households indicating that 95.2 percent household will be
remain poor in future while only 4.8 percent will be nonvulnerable in
next year. It is important to note that 57.1 households are both poor
and vulnerable while 42.9 percent are non-poor and vulnerable. Jamal
(2009) estimated poverty level at 29.8 percent in 2005 and 33.7 percent
in 2001 while vulnerability is 51.62 percent and 49.88 percent,
respectively for these two periods. A measure of vulnerability was also
developed using a five-year panel for rural Pakistan, which had
illustrated that on average 67 percent households are vulnerable between
1986 to 1990 [Mansuri and Andrew (2002)]. It can be concluded that both
poverty and vulnerability had decreased in Pakistan while a sizable
fraction of non-poor households are vulnerable to poverty.
As shown in Table 2, three estimates of poverty are given as
headcount, intensity of poverty and severity of poverty across region. A
high incidence of rural poverty (22.7 percent) is observed as compare to
urban poverty (14.7 percent). In this manner a high intensity of poverty
(the average poverty gap, or the amount of income/expenditure necessary
to bring everyone in poverty right up to the poverty line, divided by
total households) also observed in all areas of Pakistan. This can be
thought of as the amount that an average household in the economy would
have to contribute in order for poverty to be just barely eliminated.
Furthermore severity of poverty is income inequality among the poor. A
high intensity and severity of poverty is seen in rural areas as compare
to urban areas and overall Pakistan indicating a high risk of future
poverty. In addition to, some districts with high average poverty and
vulnerability level can be seen in Figure 1.
[FIGURE 1 OMITTED]
Table 3 presents a cross-distribution of the percentage of
vulnerability and poverty for the households who had suffered a welfare
loss due to a shock during last five years, 2006-2010. The welfare loss
is measured in terms of income loss which has resulted in consumption
variability of food and non-food expenditure. Households and communities
in Pakistan face the risks of suffering from different types of shocks
that affect communities as a whole referred to as covariate shocks such
as natural disasters while others affect one or a few households denoted
as idiosyncratic shocks such as a death of household member or loss of a
job. Even though, any household can be affected by these shocks, not all
of them have the same probability of recovering from the consequences of
suffering from them. Poor households are more vulnerable because they
lack the necessary physical and human capital to recover from it. All
households who are hit by any type of shock have high poverty rates than
average poverty level in Pakistan, as demonstrated in Table 3.
In Pakistan covariate shock in the form of flood is a common
phenomenon whereby informal insurance mechanism become fail, resulted in
high poverty (26.3 percent) and vulnerability (36.5 percent). These
estimates indicate that the observed incidence of poverty underestimates
the fraction of the population that is vulnerable to poverty. The level
of underestimation is revealed by the vulnerability to poverty ratio,
which is greater than one for all households in 2010. Although incidence
of poverty had decreased but the vulnerability and vulnerability to
poverty ratio had increased during 2001-05 [Jamal (2009)]. The level of
poverty had further decreased in urban and rural areas of Pakistan but
the vulnerability and vulnerability to poverty ratio had deteriorated in
2010. This analysis also documented that health shocks are more costly
than agriculture ones as can be seen with high poverty and
vulnerability. For example, Kenjiro (2005) found that in rural Cambodia,
the economic damage caused by sickness is more severe than that caused
by a crop loss. Gertler and Gruber (2002) explored evidence that in
Indonesia the economic costs associated with major illness are high and
cause a severe reduction in household consumption. Moreover, health
shocks may prevent households from using some coping strategies, for
example households are less likely to use the labour supply to cope with
health shocks because they may affect the ability of the households to
provide labour [Kochar (1995)]. However, the experience of shock by rich
households might have negatively affected other households that depend
on them for livelihoods and economic survival.
In developing countries households face substantial risks, which
they handle with risk-management and risk-coping strategies, including
self-insurance through savings and informal insurance mechanisms but
despite these mechanisms, however, they remain vulnerability to high
risk. In Table 4 distribution of vulnerability and poverty is given by
risk management strategies when a household is encountered by a shock.
These ex-post coping strategies can be divided into four main
categories: (i) asset-based strategies; (ii) assistance-based
strategies; (iii) borrowing-based strategies; and (iv) behaviour-based
strategies. The level of vulnerability and poverty is higher for those
households who had adopted behaviour-based strategies such as consuming
less, increasing labour supply or taking children out of school for
work. These types of coping strategies are practiced more often for
natural/agricultural shocks than for economic shocks [Haq (2012)]. In
previous studies these finding are also supported by Beegle, et al.
(2006) who found that in Tanzania households respond to transitory
income shocks by increasing child labour and the extent to which child
labour is used as a buffer is lower when households have access to
credit. Jacoby and Skoufas (1997) also showed that in India
unanticipated income shocks have a significant effect on children school
attendance and that school attendance appears to play an important role
in the self-insurance strategy of poor households. It is often noted
that some households had to adopt more than one strategy for consumption
smoothing. The findings in Table 4 demonstrate that those households who
had adopted more than one coping strategies are relatively more
vulnerable and poor. In addition, these households reduced food
consumption, non-food consumption and increased labour supply of
children or women in response to shocks as a second coping strategy
because they had limited asset base, and face missing formal insurance
and finance markets to smooth consumption expenditure.
An individual can be vulnerable to falling below a threshold across
several dimensions, such as education, quality of housing,
household's productive assets such as land or livestock, and place
of residence etc. The estimates given in Table 5 demonstrate that
household heads that had up to primary level education are more
vulnerable and poor as compared to higher level education As a
consequence, this low level of human capital does not allow them to earn
enough income to reduce vulnerability. Similarly, households living in
mud houses are more vulnerable to poverty as compared to residence of
better quality of houses. At the same time households with low level of
physical capital are the most vulnerable and hence remain chronically
poor. Vulnerability to shocks can be seen a cause of chronic poverty
(poor in both periods). However, Okidi and Mugambe (2002) state that
vulnerability to shocks is not just a cause of poverty but is also a
symptom of poverty. This is highlighted by Baulch and Hoddinot (2000)
who state that "households with greater endowments and greater
returns will tend to be less vulnerable to shocks. Furthermore, in this
analysis those households who are affected by any type of shocks have
high incidence of poverty and vulnerability as compared to households
with average poverty level in Pakistan.
Table 6 presents the results of logit regression of the
determinants of vulnerability, where vulnerable households are those who
are expected to be perpetually poor or expected to either fall into
poverty. The results show that the models produced a good fits of the
data as revealed by the statistical significance of the pseudo
coefficient of determinations and Wald Chi-square parameters. The
analysis shows that as the age of the household head increases, the
probability of being vulnerable significantly decreases while large
household size significantly increases this probability. Increasing
aged/child dependency ratio is another significant variable to increase
the probability of vulnerability.
Furthermore, compared to base category 'illiterate head of
household', the household heads that had secondary level or high
level of education significantly reduces the probability of falling into
poverty in future. This is also expected since education is expected to
increase capacity for escaping poverty [World Bank (2002)]. The model
shows that poor households significantly increases the probability of
being poor in future thus remain chronically poor or remain poor in both
periods. Physical capital which is related to productive assets such as
agriculture land and livestock are important in risk management. The
households who had land ownership and large animals are less likely to
be vulnerable because, possession of assets leads to an increase in the
expectation of future consumption and provide a secure source of income
in the face of negative shocks to income. However, households having
small animals and residing in lower quality of housing significantly
increases the probability of being poor in future. It is also elaborated
that households with natural /agricultural shocks, including flood,
earthquake, drought, crop failure and loss of livestock, etc. are likely
to be more vulnerable as compared to those households who are affected
by social shocks while households with health shocks and economic shocks
have less probability of being vulnerable.
When economic hardships occur in developing countries, people
resort to various risk coping strategies to smooth consumption, since
formal credit and insurance markets are less developed. In this analysis
different types of risk management strategies are adopted by households
showing that households who had adopted assets, assistant and borrowing
based strategies are less likely to be vulnerable as compared to those
who had adopted behaviour based strategies. These strategies includes
decrease food and non-food consumption, increase labour supply
particularly of women and children, beggary/ prostitution, children drop
out from school and sent to work and bonded labour arrangements, thus
had inter-generational transmission of poverty and vulnerability. The
final set of results concerns the geographical locations that also play
an important role in determining a household to be vulnerable to
poverty. Rural households are more vulnerable as compared to urban
residents while households living in province of Punjab, KP and Sindh
are less vulnerable as compared to Balochistan.
5. CONCLUSIONS
The well-being of poor households depends not only on
households' current consumption or expenditures, but also on risk
and uncertainty about their future welfare. This paper has developed a
measure of vulnerability to poverty that takes both permanent and
transitory household characteristics into consideration to forecast
vulnerability. The methodology involves a three stage FGLS method for
generating vulnerability indices by employing PPHS-2010 data.
The analysis highlights that total vulnerability is found to be
34.4 percent as opposed to the poverty of 20.7 percent. Vulnerability in
rural areas is even higher which is estimated to be 35.9 percent as
oppose to 29.4 percent of urban statistic. When decomposing
vulnerability into poor and non-poor households, it was figured out that
95 percent poor households are also vulnerable while only 18 percent
non-poor households are vulnerable. Risks to livelihood are particularly
important in Pakistan where there is generally high dependence on
agriculture sector. Households who had suffered a welfare loss due to a
shock particularly covariate shocks are more vulnerable to poverty. High
vulnerability and poverty is found for those households who suffered
from agriculture, social, economic and health shocks during the last
five years. In addition the ratio of vulnerability to poverty is also
high for agricultural, social, health and idiosyncratic shocks. When
these shocks occur, household resort to various risk coping strategies
to smooth consumption, since formal credit and insurance markets are
less developed. Households who had adopted borrowing or behaviour based
strategies are more likely to be vulnerable to poverty. The more one is
vulnerable, the less one has the capacity to cope, the more one tends to
adopt multiple coping mechanisms hence these households are more
vulnerable to poverty as observed in this analysis. The study also
revealed that households may be vulnerable across several dimensions,
such as no schooling, low quality housing, no productive assets and
geographical location. The model measuring determinants of vulnerability
indicate that households head with no education, large family size, high
number of dependents, poverty status, lack of productive assets,
agriculture shocks, high cost of shock and rural residence have
significantly higher probability of being poor in future.
Finally, a clear observation in this analysis is that vulnerability
and poverty is still concentrated in Pakistan. It is important to build
productive assets for the poor and vulnerable households, increasing the
coverage of education and health, and strengthen the disaster management
and relief mechanisms. In addition, this paper argues that despite the
limitations of purely cross-sectional data, an analysis of this data can
potentially be informative for poverty alleviation programmes.
REFERENCES
Arif, G. M and Shujaat Farooq (2012) Rural Poverty Dynamics in
Pakistan: Evidence from Three Waves of the Panel Survey. Pakistan
Institute of Development Economics, Islamabad. (Poverty and Social
Dynamics Paper Series, PSDPS-2).
Baulch, B. and J. Hoddinott (2000) Economic Mobility and Poverty
Dynamics in Developing Countries. Journal of Development Studies 36:6,
1-24.
Beegle, K, Rajeev H. Dehejia, and Roberta Gatti (2006) Child Labour
and Agricultural Shocks. Journal of Development Economics 81, 80-96.
Chaudhuri, S. (2000) Empirical Methods for Assessing Household
Vulnerability to Poverty, Columbia University. (Draft).
Chaudhuri, S. (2003) Assessing Vulnerability to Poverty: Concepts,
Empirical Methods and Illustrative Examples. Columbia University.
(Mimeographed).
Chaudhuri, S., J. Jalan, and A. Suryahadi (2002) Assessing
Household Vulnerability to Poverty from Cross-Sectional Data: A
Methodology and Estimate from Indonesia. Columbia University.
(Mimeographed).
Christiaensen, L. J. and K. Subbarao (2005) Towards an
Understanding of Household Vulnerability in Rural Kenya. Journal of
African Economies 14:4, 520-558.
Dercon, S. and P. Krishnan (2000) Vulnerability, Seasonality and
Poverty in Ethiopia. Journal of Development Studies 36:6, 25-53.
Foster, J., J. Greer, and E. Thorbecke (1984) A Class of
Decomposable Poverty Measures. Econometrica 52, 761-776.
Gertler, P. and J. Gruber (2002) Insuring Consumption against
Illness. American Economic Review 92:1, 55-70.
Gujarati, D. N. (2012) Basic Econometrics (Fifth Edition). New
York: Tata McGraw-Hill Companies. Inc.
Gunther, I. and K. Harttgen (2005) Vulnerability to Poverty in
Madagascar. Background Paper prepared for the World Bank Social
Protection Unit.
Haq, R. (2012) Shocks as a Source of Vulnerability: An Empirical
Investigation from Pakistan. Pakistan Institute of Development
Economics. (Poverty and Social Dynamics Paper Series PSDPS: 6)
Heltberg, R. and L. Niels (2009) Shocks, Coping, and Outcomes for
Pakistan's Poor: Health Risks Predominate. Journal of Development
Studies 45:6, 889-910.
Holzmann, R. and S. Jorgensen (2000) Social Risk Management: A New
Conceptual Framework for Social Protection and Beyond. World Bank.
(Social Protection Discussion Paper No. 0006).
Jacoby, H. and E. Skoufias (1997) Risk, Financial Markets, and
Human Capital in a Developing Country. Review of Economic Studies 64:3,
31 1-335.
Jalan J. and M. Ravallion (1999) Are the Poor less Well Insured?
Evidence on Vulnerability to Income Risk in Rural China. Journal of
Development Economics 58:1, 61-81.
Jamal, H. (2009) Assessing Vulnerability to Poverty: Evidence from
Pakistan. Social Policy and Development Centre (SPDC). Islamabad.
(Research Report No.80).
Kenjiro, Y. (2005) Why Illness Causes More Serious Economic Damage
than Crop Failure in Rural Cambodia. Development and Change 36: 4,
759-783.
Kochar, A. (1995) Explaining Household Vulnerability to
Idiosyncratic Income Shocks. American Economic Review 85:2, 159-164.
Kurosaki, T. (2010) Targeting the Vulnerable and the Choice of
Vulnerability Measures: Review and Application to Pakistan. The Pakistan
Development Review 49:2, 87-103.
Ligon, E. and L. Schechter (2003) Measuring Vulnerability. Economic
Journal 113, C95-C102.
Mansuri, Ghazala and Andrew Healy (2002) Vulnerability Predictions
in Rural Pakistan, IFPRI-World Bank Conference on Risk and
Vulnerability: Estimation and Policy Implications.
Morduch J. (2005) Consumption Smoothing across Space: Testing
Theories of Risk-Sharing in the ICRISAT Study Region of South India. In
S. Dercon (ed.) Insurance against Poverty. Oxford: Oxford University
Press.
Okidi, J. and G. Mugambe (2003) An Overview of Chronic Poverty and
Development Policy in Uganda. Chronic Poverty Research Centre,
Manchester, UK. (CPRC Working Paper 11).
Pakistan, Government of (2010) Economics Survey of Pakistan,
2009-10. Economic Advisor's Wing, Finance Division, Islamabad.
Suryahadi, A. and S. Sumarto (2003) Poverty and Vulnerability in
Indonesia before and after the Economic Crises. Asian Economic Journal
17:1.
Tesliuc, and L. Kathy (2004) Risk and Vulnerability in Guatemala: A
Quantitative and Qualitative Assessment. Social Protection Unit, Human
Development Network. The World Bank.
Townsend, Robert (1994) Risk and Insurance in Village India.
Econometrica 62:3, 539-591.
Udry, C. (1995) Risk and Saving in Northern Nigeria. The American
Economic Review 85:5, 1287-1300.
UNDP (2014) Sustaining Human Progress: Reducing Vulnerabilities and
Building Resilience. Human Development Report, 2007/2008. United Nations
Development Programme.
UNISDR (2004) Living With Risk. United Nations International
Strategy for Disaster Reduction, Geneva, Switzerland.
World Bank (2002) World Development Report 2002. Washington, DC:
World Bank.
Rashida Haq <rashida@pide.org.pk> is Senior Research
Economist, Pakistan Institute of Development Economics, Islamabad.
(1) The dependency ratio takes the sum of the population under the
years of 15 and over 64 and divided by the population in the
intermediate range of 15-64.
Table 1
Vulnerability and Poverty at Household Level: 2010
Poverty Status
Vulnerability Poor Non-poor Overall
Vulnerable 57.1 42.9 34.4
(95.2) (18.6)
Non-vulnerable 1.5 98.5
(4.8) (81.4) 65.6
Overall 20.6 79.4 100
Source: Author's computation is from the micro data of PPHS-2010.
Table 2
FGT Poverty Estimates: 2010
Poverty Measures
Region Headcount Intensity of Poverty Severity of Poverty
Urban 14.7 2.60 0.7
Rural 22.4 5.11 1.75
Overall 20.6 4.54 1.5
Source: Author's computation is from the micro data of PPHS-2010.
Table 3
Estimates of Vulnerability and Poverty by Type of Shocks: 2010
Vulnerability and
Poverty Status Ratio of
Vulnerability
Shocks Vulnerable Poor to Poverty
Incidence of Shock 36.2 23.7 1.53
Idiosyncratic Shock 34.2 21.4 1.60
Covariate Shock 36.5 26.3 1.39
Natural/Agriculture 35.5 23.0 1.54
Economic 37.8 38.2 0.99
Social 37.0 23.3 1.59
Health 36.7 23.5 1.56
Source: Author's computation is from the micro data of PPHS-2010.
Table 4
Vulnerability and Poverty by Risk Management Strategies: 2010
Vulnerability and
Poverty Status Ratio of
Risk Management Vulnerability
Strategies Vulnerable Poor to Poverty
Assets based 30.8 21.9 1.41
Assistance based 38.1 25.9 1.47
Borrowing based 40.6 24.6 1.65
Behavior based 43.9 26.0 1.69
One strategy 36.2 23.7 1.53
Two strategies 39.8 27.8 1.43
Three strategies 39.5 28.5 1.39
Source: Author's computation is from the micro data of PPHS-2010.
Table 5
Estimates of Vulnerability and Poverty across Groups: 2010
Vulnerability and
Poverty Status Ratio of
Vulnerability
Groups Vulnerable Poor to Poverty
No Schooling 39.8 24.0 1.65
Primary education 39.1 24.7 1.58
Secondary education 26.4 14.4 1.83
Higher education 10.3 9.4 1.09
Mud house 47.4 30.8 1.54
Mixed (mud and brick house) 27.7 15.3 1.81
Brick house 24.1 12.6 1.91
Landless households 37.6 22.7 1.66
Small landholders (up to 3 acres) 37.7 24.0 1.57
Medium landholders ([3.sup.+] -10 26.2 14.8 1.77
acres)
Large landholders ([10.sup.+] 20.9 10.6 1.97
acres)
No livestock 37.6 22.5 1.67
Large animals (No.) 29.5 17.0 1.74
Small animals (No.) 35.4 20.7 1.71
Urban 29.4 14.7 2.0
Rural 35.9 22.4 1.6
Overall 34.5 20.6 1.67
Source: Author's estimates based on the micro data of PPHS-2010.
Table 6
Determinants of Vulnerability to Poverty in Pakistan
Vulnerable / Non Vulnerable
Exp
Correlates Coefficient Std. Error ([beta])
Intercept -3.25 1.418 0.039
Male headed households -0.88 0.677 0.413
Age of HH Head (years) -0.12 *** 0.007 0.88
Household size (No) 0.067 * 0.027 1.069
Primary Education -0.050 0.104 0.95
Secondary Education -0.23 *** 0.13 0.79
Higher Education -0.463 * 0.119 0.63
Dependency ratio 0.202 *** 0.106 1.22
Poverty status 4.32 * 0.25 75.17
Land ownership (acres) -0.039 * 0.01 0.96
Large animals (No.) -0.413 * 0.055 0.662
Small animals (No.) 0.13 * 0.022 1.139
Housing quality (mud house) 1.26 * 0.562 3.53
Agriculture shocks 1.054 * 0.274 2.86
Economic shocks -3.33 * 0.499 0.036
Health shocks -1.36 * 0.279 0.26
Covariate shock -0.016 0.25 0.98
Cost of shock (Rs.) 0.366 * 0.082 1.44
Asset based strategy -2.08 * 0.246 0.13
Assistance based strategy -2.14 * 0.357 0.12
Borrowing based strategy -1.36 * 0.284 0.26
Multi strategies 0.203 0.047 1.23
Region (Rural=l) 0.397 * 0.110 1.48
Punjab -0.443 *** 0.24 0.64
Sindh -0.734 * 0.25 0.48
KPK -1.48 * 0.295 0.227
Chi-square 2874.37
-2 Log likelihood 4807..32
Pseudo [R.sup.2] (Cox and Snell) 0.152
Observations 3500
Source: Author's estimates based on the micro data of PPHS-2010.
(a.) The reference category is: Non Vulnerable households.
* significant at 1 percent, ** significant at 5 percent and
significant at *** 10 percent.
