1. A PROFILE OF POVERTY IN PAKISTAN
IFTIKHAR AHMED CHEEMA
SENIOR POVERTY SPECIALIST
NOVEMBER 2005
CENTRE FOR RESEARCH ON POVERTY REDUCTION
AND INCOME DISTRIBUTION PLANNING COMMISSION
ISLAMABAD
Phone: 9202868, Fax: 9210254
www.crprid.org
2. Table of Contents
Table of contents............................................................................................................................. ii
List of tables................................................................................................................................... iv
List of figures.................................................................................................................................. v
Executive summary....................................................................................................................... vii
1. Introduction .............................................................................................................................. 1
2. Data .......................................................................................................................................... 1
3. Methodology ............................................................................................................................ 2
3.1 Choice of welfare indicator ............................................................................................... 2
3.2 Consumption aggregate ..................................................................................................... 2
3.3 Adjustment of consumption............................................................................................... 2
3.4 Price adjustment ................................................................................................................ 3
3.5 Poverty line........................................................................................................................ 3
3.6 Choice of aggegator........................................................................................................... 4
4. Poverty comparison.................................................................................................................. 4
4.1 Spatial poverty comparisom .............................................................................................. 4
4.2 Poverty comparison by household characteristics........................................................... 15
4.2.1 Poverty by household size ............................................................................................... 15
4.2.2 Poverty by dependency ratio ........................................................................................... 17
4.2.3 Poverty by access to amenities ........................................................................................ 17
4.2.4 Poverty by access to land................................................................................................. 17
4.3 Poverty by household head characteristics...................................................................... 18
4.3.1 Poverty by educational attainment of head...................................................................... 18
ii
3. 4.3.2 Poverty by literacy of household head ............................................................................ 20
4.3.3 Poverty by sex of household head ................................................................................... 20
4.3.4 Poverty by status of employment of hhold head ............................................................. 21
5. Characteristics of the poor...................................................................................................... 22
5.1 Household size and dependency ratio ............................................................................ 23
5.2 Ever attended school........................................................................................................ 24
5.3 Gross and net enrolment rates ......................................................................................... 26
5.4 Literacy rate..................................................................................................................... 27
5.5 Highest class completed .................................................................................................. 27
5.6 Immunisation ................................................................................................................... 29
5.7 Maternal health care ........................................................................................................ 29
5.8 Housing conditions .......................................................................................................... 30
5.9 Percentage of income shares by poverty status ............................................................... 31
5.10 Durable goods.................................................................................................................... 31
6. Conclusion.............................................................................................................................. 32
Appendix A: Confidence intervals for poverty measures by region and period.......................... 35
Appendix B: Confidence intervals for poverty measures by province and region ,2001 .......... 36
Appendix C: Hypothesis testing ................................................................................................ ...37
References..................................................................................................................................... 40
iii
4. List of Tables
Table 1: Poverty indices by province and region.......................................................................... . 6
Table 2: Incidence of poverty by household characteristics......................................................... 16
Table 3: Incidence of poverty by characteristics of the household head .................................... . 19
Table 4: Household size, composition, and dependency ratio by poverty status ....................... . 23
Table 5: Education indicators by poverty status,2001 ................................................................ . 25
Table 6: Health indicators by poverty status,2001...................................................................... . 28
Table 7: Housing conditions by poverty status,2001.................................................................. . 30
Table 8: Main source of income by poverty status, 2001 ........................................................... . 31
Table 9: Percentage of households with durable items by poverty status, 2001 ........................ .32
iv
5. List of Figures
Figure 1: Poverty rates over time.................................................................................................. . 5
Figure 2: Headcount ratio and agricultural growth......................................................................... 7
Figure 3: Incidence of povery by region....................................................................................... . 8
Figure 4: Incidence of poverty and contribution to total poverty by region, 2001 ...................... . 8
Figure 5: Incidence of poverty by province.................................................................................. . 9
Figure 6: Incidence of poverty and contribution to total povery by province, 2001 .................. . 10
Figure 7: Cumulative distribution functions of per adult equivalent consumption expenditure . 11
Figure 8: Cumulative distribution functions of per adult equivalent consumption expenditure by
region, 2001 ................................................................................................................................ . 12
Figure 9:Cumulative distribution functions of per adult equivalent consumption expenditure by
province, 2001 .............................................................................................................................. 12
Figure 10A: Cumulative distribution functions of per adult equivalent consumption expenditure
by household size, 2001.............................................................................................................. . 13
Figure 10B: Cumulative distribution functions of per adult equivalent consumption expenditure
with adult and different equivalent scales (equal weight), 2001 .................................................. 14
Figure 10C: Cumulative distribution functions of per adult equivalent consumption expenditure
with adult and different equivalent scales(child=0.6), 2001....................................................... . 14
Figure 11: Headcount by household size, 2001 ......................................................................... . 15
Figure 12: Contribution to total poverty ..................................................................................... . 15
Figure 13: Headcount ratio by dependency ratio, 2001.............................................................. . 17
Figure 14: Headcount ratio by housing condition, 2001............................................................. . 17
Figure 15: Headcount ratio by landownership, 2001.................................................................. . 18
v
6. Figure 16: Headcount ratio by level of education of household head, 2001 ............................... 18
Figure 17: Headcount ratio by literacy of household head, 2001 ............................................... . 20
Figure 18: Headcount raio by sex of hhold head, 2001 ................................................................ 20
Figure 19: Headcount ratio by status of employment of household head, 2001......................... . 21
Figure 20: Headcount ratio by sector of employment of household head, 2001 ....................... . 21
Figure 21: Headcount ratio by occupation of household head, 2001 ...............................................
Figure 22: Dependency ratio by poverty status, 2001 ................................................................ . 23
Figure 23: Percentage of population that has ever attended school by poverty status, 2001 ..... . 24
Figure 24: Left primary school before completing primary level by poverty status, 2001 ........ . 24
Figure 25: Gross enrolment rate by poverty status, 2001 ............................................................ 26
Figure 26: Literacy by poverty status, 2001 ............................................................................... . 27
Figure 27: Highest class completed by poverty status, 2001........................................................ 27
Figure 28: Average number of years of schooling by consumption decile ................................. .28
Figure 29: Percentage of children aged 12-23 months that have been fully immunised by poverty
status, 2001 ................................................................................................................................. . 29
Figure 30: Pre and post natal care consulatation by poverty status, 2001 .................................. . 29
Figure 31: Access to facilities (electricity,gas and telephone) by poverty status, 2001 ............. . 30
vi
7. Executive Summary
The objective of this paper is to present “A Profile of Poverty in Pakistan” which is useful for
broadening and deepening our understanding about different dimensions of poverty. Five
data sets including HIES 1992-93, 1993-94, 1996-97 & PIHS 1998-99 & 2000-01, conducted
by Federal Bureau of Statistics have been used for poverty analysis.
Consumption aggregate has been used as a welfare indicator. It includes both actual and
imputed expenditure. Some expenses such as taxes, fines and expenses on marriage or
funeral and durable items are not included in the consumption aggregate.
While consumption expenditure is recorded at the household level, it needs to be measured at
the individual level. While adjusting household expenditure in order to get per adult
equivalent consumption expenditure, a simple scale has been used. This scale weights 0.8 to
individuals younger than 18 years and 1 for all other individuals. Keeping in view the spread
of survey over a year, Paasche’s price index at primary sampling unit level has been
computed in order to represent welfare indicator in real values.
The poverty line Rs. 673.54 as notified by the Planning Commission vide letter No.
1(41)/Poverty/PC/2002, dated 16.8.2002 has been used as a base poverty line. The base
poverty line has been adjusted over time by inflation rate between household surveys in order
to keep the poverty line constant. Foster, Greer and Thorbecke class of poverty measures
have been used as an aggregating index.
This paper presents poverty estimates for 1992-93, 1993-94, 1996-97 1998-99 and 2000-01.
The poverty estimates have increased over this period with the exception of 1996-97, where a
decline has been observed. The incidence of poverty at national level was around 25 percent
in 1992-93 which increased to 34 percent approximately in 2000-01. Similarly, poverty gap
and severity of poverty increased over time. The factors responsible for this increase in
poverty rates are low GDP growth rate, public finance crisis, problem of external finance,
decrease in formal employment opportunities, cut in subsidies and sharp rise in utility rates.
The difference in the incidence of poverty between urban and rural areas has increased over
this period. Headcount ratio was 20 percent in urban areas in 1992-93 and it rose to
approximately 23 percent in 2000-01 depicting a moderate increase of 3 percentage points
over this period. Rural areas depict a remarkable rise of around 12 percentage points in
vii
8. headcount ratio from 27.63 percent in 1992-93 to 39.26 percent in 2000-01. Poverty gap and
severity of poverty in urban and rural areas almost doubled during this period.
Poor are concentrated in rural areas. It is important that rural livelihoods are improved in
order to ensure that growth benefits the poor. The policy agenda may involve improvement
in access to land, agricultural productivity, price incentives and credit arrangements. There is
also a critical need for major improvement in rural infrastructure.
PIHS 2000-01 has been used to identify the characteristics of the poor. Total dependency
ratio is higher in poor households (115 percent) than non-poor (79 percent). The rate of ever
attended school is 57 percent for non-poor compared to 37 percent for poor. Similarly, a
large number of children in poor households leave school before completing primary level.
Financial costs are more important as a reason for leaving before completing primary level
and never attended school for the poor. Primary gross enrolment rate for poor is about 54
percent and 85 percent for non-poor. Primary net enrolment rate is 31 percent for poor
compared to 50 percent for non-poor. There are also wide differences between poor and non-
poor with respect to gross and net enrolment rate at middle and matric level.
Health indicators are also lower for poor than non-poor. The percentage of children aged 12-
23 months fully immunized is 43 percent for poor whereas it is 60 percent for non-poor.
About 23 percent poor women went for prenatal care consultation compared to 43 percent for
non-poor. Poor women mostly attend government hospitals/clinics in contrast to non-poor
who prefer private hospitals/clinics.
Housing conditions are also worse for poor. Only 16 percent of poor population has access to
piped water for drinking compared to 29 percent for non-poor. Similarly, 27 percent of poor
live in houses having flush toilet compared to around 55 percent of non-poor. In rural areas
65 percent of poor don’t have toilet in their house.
There are three policy messages with respect to education. First, literacy rates may be
improved. Second, policy makers should find ways and means to increase enrolment rates
and reduce dropout rates. Third, simply an increase in enrolment rate will not be sufficient to
reduce poverty level. Level of educational attainment must be improved at least to middle
level in order to reduce poverty significantly.
Though health indicators have improved over time still they are low particularly for poor
people. This situation demands public policy to focus particularly in rural areas on raising
viii
9. health expenditure and aiming on prevention and control programmes, especially in the area
of reproductive health, child health, nutrition deficiencies and communicable and infectious
diseases.
ix
10. A Profile of Poverty in Pakistan
1 Introduction
Reducing poverty has the remained main objective of policy makers but it has gained more
importance since the adoption of MDGs. In order to design poverty reduction strategies, it is
very important for policy makers to know who are the poor? How many poor are there? Where
do they live and what is their social and economic profile? The main objective of this paper is to
provide baseline poverty profile by answering these questions. Poverty profile is useful for
broadening and deepening our understanding about the different dimensions of poverty and
measuring the degree to which the government has been successful in reducing the poverty over
time.
2 Data
This paper uses five data sets for poverty analysis i.e., HIES 1992-93, 1993-94, 1996-97 & PIHS
1998-99 and 2000-01. The sample size of these household surveys is substantial enough to allow
representative estimates to be obtained for each province and region.
Name of survey Sample size
Urban Rural Overall
HIES 1992-93 5586 9006 14592
HIES 1993-94 5632 9036 14668
HIES 1996-97 5447 8814 14261
PIHS 1998-99 5527 9152 14679
PIHS 2000-01 5536 9169 14705
These household surveys conducted by Federal Bureau of Statistics provide comprehensive
information about household consumption expenditure and income. PIHS 1998-99 and 2000-01
provide rich information about different socio-economic indicators that are essential for poverty
profiling. This paper uses PIHS 2000-01 for identifying different socioeconomic characteristics
of the poor and non-poor.
Minor data cleaning has been made. In some cases data entry errors were observed. These data
entry errors were corrected simply by moving the decimal point in either direction through
computer programming in order to make data meaningful. Moreover, there were a few cases
where quantities for food items were missing. In missing cases new values were obtained by
dividing the value of that item by median unit price of that item within the primary sampling
unit.
1
11. The structure of the paper runs as follows. The next section describes the methodology of
estimating poverty rates, followed by section 4 wherein poverty comparisons are discussed.
Section 5 explains socioeconomic characteristics by poverty status. The last section concludes
the paper.
3 Methodology
3.1 Choice of welfare indicator
In developing countries, consumption is more appropriate than income as welfare indicator.
First, income is interpreted as a measure of welfare opportunity while consumption as a measure
of welfare achievement. Second, it is generally believed that survey respondents are more willing
to reveal their consumption pattern than their income. Third, consumption is measured better
than income in developing countries because of difficulties in defining and measuring income for
self-employed. Finally, income is subject to seasonal variability while consumption tends to be
less variable. This paper uses consumption as a welfare indicator.
3.2 Consumption aggregate
Consumption aggregate is comprehensive and consists of both actual and imputed expenditure. It
includes not only actual purchases but also self-produced and consumed items or consumption of
items that were received as gift or assistance or wage and salary in kind. Thus consumption
aggregate includes food items, fuel and utilities, housing (rent, imputed rent and minor repair),
frequent nonfood expenses (household laundry and cleaning personal care products and services)
and other nonfood expenses (clothes, footwear, education, and health related expenses).
However, some expenses such as taxes, fines and expenses on marriage or funeral and durable
items are not included in the consumption aggregate because these are not related to living
standards.
3.3 Adjustment of consumption
While consumption expenditure (food and nonfood) is recorded at the household level, welfare
needs to be measured at the individual level. The general practice is to divide household income
or consumption expenditure by the total number of household members without making any
2
12. adjustment for household composition. This methodology is not much convincing as it gives
equal welfare ranking to two households with same total income/consumption and with same
number of household members even if one of the households is dominated by adults and the
other by children. Nutrition based adult equivalent scales, which differentiate, between
households on the basis of sex and age are also used to transform the number of persons in a
household to adult equivalents. The application of nutrition based equivalent scales to any
expenditure other than food expenditure is questionable. Its use is defendable when food
expenditure occupies a comprehensive share of total expenditure. While adjusting household
expenditure in order to get per adult equivalent consumption expenditure, this paper uses simple
equivalent scale that weights 0.8 to individuals younger than 18 years and 1 for all other
individuals as food expenditure represents only about 50 percent of the total consumption
expenditure.
3.4 Price adjustment
It is necessary to represent the welfare indicator in real values as households face different prices
during the year of the survey. Laspayer’s price indices as calculated by the Federal Bureau of
Statistics are not suitable for using price adjustment because these indices don’t consider
differences between urban and rural areas or between provinces. Paasche’s price index calculated
at the primary sampling unit has been used in order to convert the welfare indicator in real
values. Though, household income and expenditure surveys don’t always provide information on
prices but it is still possible to calculate spatial price index by means of unit values that are
obtained by dividing expenditure per food and fuel items by quantity consumed. Paasche’s price
index has been calculated at the primary sampling unit level by using the median prices and
average budget shares in each primary sampling unit.
3.5 Poverty line
Using PIHS 1998-99, Federal Bureau of Statistics estimated absolute poverty line Rs. 673.54 on
calories 2350 per adult equivalent per day with calorie based approach. This poverty line was
notified, as national poverty line by Planning Commission vide Letter No. 1 (41)
Poverty/PC/2002, dated 16.8.2002. This paper uses Rs. 673.54 as a base poverty line for the
3
13. whole country (urban and rural areas) for poverty analysis. This poverty line has been adjusted
for other years in the analysis by the inflation rate between the two household surveys so that the
base poverty line remains constant and poverty measures are consistent and comparable over
time and across regions. It is highlighted that the level of inflation between two household
surveys is calculated using monthly official consumer price index computed by the Federal
Bureau of Statistics. While conducting household survey, different percentages of interviews
take place in different months and these facts need to be considered when inflation rate between
two household surveys is computed.
3.6 Choice of aggregator
After the finalization of poverty line, one has to choose the aggregating index. The most popular
used measures of poverty are the Foster, Greer and Thorbecke class of poverty measures. These
include headcount ratio, poverty gap and severity of poverty. Headcount ratio is defined as the
proportion of population below the poverty line. It is easy to calculate but it does not take into
account depth of poverty. Poverty gap reflects the average short fall of the incomes/consumption
of the poor expressed as a share of the poverty line. It considers depth of poverty but does not
take into account the distribution of income amongst the poor. Severity of poverty is the square
of poverty gap. It is sensitive to distribution of among the poor as more weight is given to the
poorest below the poverty line.
4 Poverty comparison
4.1 Spatial poverty comparison
National poverty
It is useful to examine how poverty rates vary across regions and over time. Comparison of
poverty across regions helps in targeting poverty alleviation programmes to meet the needs of the
poor while over time it is useful for policy makers to monitor the effectiveness of past
programmes in reducing the intensity of poverty among various socioeconomic groups of the
poor.
4
14. Though this paper presents poverty estimates for 1992-93, 1993-94, 1996-97, 1998-99 and 2000-
01 but the main focus is on 2000-01.
There are three main poverty measures that are used for poverty analysis. They include
headcount, poverty gap and severity of poverty.
The poverty estimates have increased over Figure 1: Poverty rates over time
this period with the exception of 1996-97 40.00 34.46
31.08
28.17
where a decline has been observed. The 30.00 25.46 25.78
% 20.00
incidence of poverty at national level was
10.00 4.27 5.22 4.38 6.58 7.03
1.10 1.44 1.14 2.06 2.13
around 25 percent in 1992-93 which has 0.00
1992-93 1993-94 1996-97 1998-99 2000-01
increased significantly over time,
Head count Poverty gap Severity of poverty
amounting to 34 percent approximately in
2001, indicating an increase of 9 percentage points.
The last two surveys (PIHS 1998-99 & 2000-01) reveal that the proportion of population under
poverty line has increased from 31.08 percent to 34.46 percent depicting an increase of 3.38
percentage points. Changes in headcount ratio between years are statistically significant except
between 1992-93 and 1996-97 when, they are statistically insignificant (Appendix C).
Similarly, a substantial rise in poverty gap and severity of poverty occurred between 1992-93
and 2000-01. Poverty gap was 4.27 percent in 1992-93 compared to 7.03 percent in 2000-01.
Severity of poverty has increased from 1.10 percent to 2.13 percent over this period. Changes in
poverty gap and severity of poverty between years are statistically significant except between
1992-93 & 1996-97 and 1998-99 & 2000 (Appendix C).
Between 1998-99 and 2000-01 poverty gap increased form 6.58 percent to 7.03 percent and
severity of poverty from 2.06 percent to 2.13 percent. A sharp rise in headcount compared to the
poverty gap and severity of poverty over the last two surveys reveals that a substantial proportion
of poor population lies close to the poverty line and they are most likely to cross it in the
presence of better performance of the economy.
5
16. In absence of any formal modeling of determinants of poverty due to very few discontinuous
observations, one can conjecture that the following factors may be responsible for an increase in
poverty during this period are low GDP growth rate, public finance crisis, problem of external
finance, decrease in formal employment opportunities due to privatization of government owned
enterprises and downsizing in public sector, cut in subsidies and sharp rise in utility rates.
When we analyze the dip in poverty rates in 1996-97 we get some piece of information that may
be helpful in understanding the
Figure 2: Headcount ratio and agriculture growth
situation. Given the primary role of
agriculture sector in the rural economy, 40 15
30 10
its performance is likely to be critical in %
5
%
20
0
explaining the observed trends in rural 10 -5
0 -10
poverty. Agriculture growth rate
2 93 -94 -95 -96 -97 -98 -99 -00 -01 -02
1-9 2- 3 4 5 6 7 8 9 0 1
appears to affect poverty with one-year 1 99 199 199 199 199 199 199 199 199 200 200
lag as is evident from the Figure 2. A Headcount(left scale) Agriculture growth rate (right scale)
good performance of agriculture in
1995-96 may be one of the causes leading to reduction in poverty in 1996-97. Moreover, wage
rates also have an effect on poverty. Weighted average of nominal growth of labour wages
(unskilled labour) in the construction sector in the 12 biggest cities of Pakistan was remarkably
higher in 1996 and 1997 compared to other succeeding years. The rate of growth was around 17
percent in 1996 and 1997 followed by 5.79 percent in 1998, 6.52 percent in 1999, and 4.12
percent in 2000 and 1.70 in 2001. Furthermore, household size in 1996-97 was lower than 1993-
4 by 0.2 person. Apparently it is underestimated in rural Sindh where very few households
indicated servants and other relatives as household member when the definition of household
member is the same for all these household surveys. Rural poverty in Sindh is underestimated as
it is 19 percent in 1996-97 compared to 30 percent in 1993-94 and 34 percent in 1998-99.
Poverty comparison by urban and rural areas
Poverty reduction policies aim to reach disadvantaged groups and backward areas effectively and
efficiently. Poverty profile is useful in targeting development resources towards poor areas.
7
17. Which region should be given priority in targeting? Poverty estimates provide an easy answer to
this question. Incidence of poverty indicates
Figure 3: Incidence of poverty by region
that rural areas have more poverty than urban
50
areas and rural poverty has increased faster 39.26
40 33.54 35.13
27.63 30.17
than urban poverty. The difference in the 30
%
20 21.37 22.69
19.99
10 15.36 15.84
incidence of poverty between urban and rural
0
areas was around 8 percentage points in 1992- 1992-93 1993-94 1996-97 1998-99 2000-01
93, which doubled in 2000-01. Headcount ratio Urban Rural
was about 20 percent in urban areas in 1992-93
and it rose approximately to 23 percent in 2000-01 depicting a moderate increase of 3 percentage
points over this period. Even between 1998-99 and 2000-01 incidence of poverty in urban areas
increased marginally from 21.37 percent to 22.69 percent. Similarly, poverty gap and severity of
poverty experienced slight rise over the period.
Rural areas depict a remarkable rise of around 12 percentage points in headcount ratio from
27.63 percent in 1992-93 to 39.26 percent in 2000-01. Poverty gap and severity of poverty
almost doubled during this period. Changes in poverty rates between urban and rural areas are
statistically significant over the period (Appendix C).
The magnitude of regional contributions to national poverty depict that about 81 percent of all
the poor are concentrated in rural areas while the share of rural areas in total population is 71
percent (Figure 4).
Poor are concentrated in rural areas. It is important that rural livelihoods are improved in order
to ensure that growth benefits the poor. Figure 4: Incidence of poverty and contribution to
total poverty by region, 2001
This implies that agriculture growth is a
precondition in order to improve the 120 100
100 80.94
livelihoods of the poor in rural areas 80
%
60 39.26
34.46
because agriculture accounts for about 40 22.69 19.06
20
23 percent of GDP, 41 percent of labour 0
Overall Rural Urban
force and provides a livelihood for Headcount Contribution to total poverty
about two-third of population. The
policy agenda may involve improvement in access to land, agricultural productivity, price
incentives
8
18. and credit arrangements. There is also a critical need for major improvement in rural
infrastructure – improved water supply, better irrigation and road facilities, communications and
more effective research and extension activities. Moreover, the inequality in land distribution and
thinness of agricultural labour markets suggest that the non-formal sector is very important for
enhancing the earning power of the poor. So public policy should also improve opportunities in
the labour-intensive non-formal sector.
Poverty comparison by province
Poverty increased in all provinces from 1992-93 to 2000-01. Headcount ratio was highest in
NWFP (33.62 percent) in 1992-93 and it maintains its position in 2000-01 with 41.47 percent.
Incidence of poverty increased sharply in Figure 5: Incidence of poverty by province
Sindh and Balochistan between 1998-99 50
and 2000-01. In Sindh headcount ratio 40
30
%
increased from 26.01 percent to 35.32 20
10
depicting a rise of 9 percentage points over
0
this period. Balochistan experienced a steep Punjab Sindh NWFP Balochistan
rise of 14 percentage points in headcount 1992-93 1993-94 1996-97 1998-99 2000-01
ratio over this period from 21.55 to percent 35.49 percent. This sharp rise in poverty in Sindh and
Balochistan is the outcome of drought that affected badly the rural areas of these two provinces
during this period. Rural poverty rose steeply in Sindh and Balchistan between 1998-99 and
2000-01 from 34 percent to 45 percent and 21 percent to 37 percent respectively. Changes in
headcount ratio in rural areas between 1998-99 & 2000-01 are statistically significant in Sindh
and Balochistan (Appendix C).
Drought had more significant effect on rural areas in Sindh and Balochistan than on Punjab and
NWFP. Between 1998-99 and 2000-01 production of wheat and rice decreased by 17 percent
and 13 percent respectively in Sindh, that affected rural poverty (Pakistan Statistical Yearbook
2002). Balochistan was also badly affected by drought and it experienced steep rise in rural
poverty from 21 percent in 1998-99 to 37.45 percent in 2000-01.
9
19. Poverty remained stable in Punjab and NWFP between 1998-99 and 2000-01. Between these two
years production of cotton, wheat and rice increased by 29 percent, 17 percent and 18 percent
respectively in Punjab. In NWFP wheat production decreased but this reduction was offset by
increase in maize crop (Pakistan Statistical Yearbook 2002). Changes in poverty rates are
statistically insignificant in Punjab and NWFP between the last two surveys (Appendix C).
As far as the contribution to
total poverty is concerned, Figure 6: Incidence of poverty and contribution to
total poverty by province, 2001
Figure 6 indicates that Punjab
120
province contributed around 52 100.00
100
percent to total poverty, 80
52.48
%
60 41.47
followed by Sindh (26.16 40
34.46 32.24 35.32 35.49
26.16
16.80
percent), NWFP (16.80 20 4.56
0
percent) and Balochistan (4.56 Overall Punjab Sindh NWFP Balochistan
percent). The evidence Headcount Contribution to total poverty
suggests that poverty reduction
strategies should focus on rural Punjab and rural Sindh that are the habitat of majority of the poor
population.
Checking the robustness
In order to design poverty reduction strategies, policy makers are interested in whether poverty
has increased or decreased and which areas/regions have more poverty. Sensitivity analysis
answers these questions. Sensitivity analysis is undertaken by comparing the ranking of
cumulative distribution functions of welfare indicator over time or region. The lines show the
fraction of population on the vertical axis whose per equivalent adult consumption is less or
equal to the amount indicated on the horizontal axis. If a cumulative distribution function for
example “A” in one year or one region attaches a higher proportion of the population to each per
adult equivalent consumption level compared to another cumulative distribution function for
example “B” then distribution “A” has more poverty than distribution “B” independent of any
poverty line. The finding of more poverty in distribution “A” is robust.
10
20. Figure 7: Cumulative distribution functions of per adult equivalent consumption
expenditure
Pihs01 Pihs99
Hies97 Hies94
.8
.6
F ctio o co su p n
ra n f n m tio
2001
.4
1998-99
1993-94
.2
1996-97
0
250 500 750 1000
Per adult equivalent consumption expenditure
The Figure 7 presents the cumulative distribution functions of the welfare indicator for different
years where per adult equivalent consumption expenditure is expressed at the prices of 2000-01.
The cumulative distribution functions present robust results as the curves don’t cross each other.
Ranking of these distributions don’t change what ever poverty line is used.
It shows that the distribution of per adult equivalent consumption expenditure for 1996-97 lies
entirely below and to the right of other distributions followed by 1993-94, 1998-99 and 2001. In
other words, poverty was lowest in 1996-97 and highest in 2001 whatever poverty line is set.
Focussing on inter-region distribution in 2001, distribution of rural areas dominates the
distribution of urban areas indicating more poverty in rural areas than urban areas (Figure 8).
Cumulative distribution functions of provinces (Figure 9) cross each other at several points and it
is not possible to distinguish their welfare rankings.
11
21. Figure 8: Cumulative distribution functions of per adult equivalent consumption
expenditure by region, 2001
Urban Rural
.6
Rural
Fraction of population
.4
Urban
.2
0
250 500 750 1000
Per adult equivalent consumption expenditure
Figure 9: Cumulative distribution functions of per adult equivalent consumption
expenditure by province, 2001
Punjab Sindh
NWFP Balochistan
.8
Balochistan
.6
Fraction of consumption
NWFP
.4 Sindh
Punjab
.2
0
250 500 750 1000
Per adult equivalent consumption expenditure
12
22. It is argued that use of different equivalent scales change correlation between poverty and
household size. To answer this issue, we compare the cumulative distribution functions of
household of different composition and again compare them by applying different equivalent
scales. As expected Figure 10A presents robust results indicating poverty increases as the
household size rises. Households with 1-2 persons are richer than other households. On the other
side households with persons 8 and/or above have higher poverty. These results don’t change
what ever poverty line is used.
In Figure 10B households with two adults and varying number of children are compared and
each individual has equal weight. While in figure 10C children under age 18 have weight 0.6
meaning children below 18 years need 60 percent of the consumption of an adult.
The Figure 10B and 10C depict that household with two adults and no children are richer. As the
number of children increases, poverty also increases. The results hold true throughout the
distribution. The distributions don’t cross each other suggesting that larger households are poorer
and use of different scales don’t reverse the ranking.
Figure 10A: Cumulative distribution functions of per adult equivalent consumption
expenditure by household size, 2001
hhsize1 hhsize2
hhsize3 hhsize4
.8 8 + Persons
5-7 Persons
.6
Fraction of population
3-4 persons
.4
.2
1-2 Persons
0
250 500 750 1000
Per adult equivalent consumption expenditure
13
23. Figure 10B: Cumulative distribution functions of per adult equivalent consumption
expenditure of households with adult and different equivalent scales, 2000-01
(Equal weight)
hhsize1 hhsize2
hhsize3 hhsize4
.8
2 adults + 5 children
.6 2 adults + 4 children
Fraction of population
2 adults + 3 children
2 adults + 2children
.4
.2 2 adults + 1 child
2 adult
0
250 500 750 1000
Per capita consumption expenditure
Figure 10C: Cumulative distribution functions of per adult equivalent consumption
expenditure of households with adult and different equivalent scales, 2000-01
hhsize1 hhsize2
hhsize3 hhsize4
.8 hhsize5
hhsize6
.6 2 adults + 5 children
Fraction of population
2 adults + 4 children
2 adults + 3 children
.4
2 adults + 2 children
.2
2 adults
0
250 500 750 1000
Per adult equivalent consumption (children=0.6 )
14
24. PIHS 2001 is the latest household survey available and it provides rich information about
different socioeconomic indicators. This data set has been used to identify the different
dimensions of poverty and characteristics of the poor.
4.2 Poverty comparison by household characteristics
Table 2 presents head count ratio by household characteristics, which are useful for poverty
comparison.
4.2.1 Poverty by household size
The impact of household size on poverty is well established. As the household size increases so
does poverty. PIHS 2001 corroborates that larger households have higher incidence of poverty
than smaller ones. The incidence of poverty is highest (49 percent) in households consisting of
11 persons or higher and it is lowest (2.72 Figure 11: Headcount ratio by household size, 2001
percent) if household size is one. Head Urban Rural Overall
60
count ratio increases monotonously as the
50
household size increases. There are wide 40
%
30
differences in the incidence of poverty when
20
compared by urban and rural areas. Larger 10
0
households are likely to have more young 1 2 3 4 5 6 7 8 9 10 >=11
children, which pose financial burden on the Household size
households due to high cost of education
and health.
Another way of looking at the distribution of
Figure 12: Contribution to total poverty
poverty is in terms of the contribution of
35
different size of households to national
30
Poverty. Table 2 indicates that households 25
20
consisting of 11 persons or greater %
15
contribute 34 percent approximately to total 10
5
national poverty whereas households with 0
1
2
3
4
5
6
7
8
9
10
>=11
size 6 or lower contribute around 16 percent.
Household size
Since larger households tend to have higher
15
26. Proportion of children so poverty might be reduced significantly if households were to have
fewer children.
4.2.2 Poverty by dependency ratio
Dependency ratio is defined as the
Figure 13: Headcount ratio by dependency
number of household members ratio, 2001
70
60.43
divided by the number of earners in 60
50
the household. Poverty increases as 40 34.29 34.91 36.65
%
26.76 23.06
30
dependency ratio rises. It is lowest 20 10 12.18
10 3.98
(10 percent) in households where 0
0.34
dependency ratio is 1 and highest = 1.0 =1
.5
=2
.0
=3
.0
> 3.0
tio o< o< o< tio
Ra ati ati ati Ra
around 37 percent with dependency < R < R < R
1.0 1.5 2.0
ratio greater than three. Headcount Contribution to national poverty
4.2.3 Poverty by access to amenities
It is argued that households having access to amenities are most likely less poor compared to
those without them. This argument is
Figure 14: Headcount ratio by housing condition, 2001
supported by PIHS 2001. Poverty is
about 30 percent in household that 100
90 Electricity
have access to electricity compared to 80 Gas
70
around 49 percent having no 60 Telephone
% 50 No
electricity. Similarly, households with 40 Yes
30
access to gas and telephone 20
10
connection (land) have lower poverty 0
U R T U R T U R T
rates than houses without them. There
are stark variations in the incidence of poverty between urban and rural areas. Moreover, poverty
is about 23 percent in those households where main source of drinking water is piped water and
it rises to 46 percent where the main source is well. Further the percentage of population in
poverty is about 46 percent where household has no toilet in the house compared to 14 percent
approximately if house has a flush connected to public sewerage.
17
27. 4.2.4 Poverty and access to land
Poor households typically own less land
Figure 15: Head count ratio by landowership,
than non-poor. About 37 percent of the
2001
poor people live in households, owning
Headcount Contribution to national poverty
land 1.5 acres or less and contributing 30 36.87
40
percent approximately to total poverty. As 35 29.53
30 29.66 25.91 24.95
22.64
25 22.76 22.74
the size of land ownership increases, 20 16.42
%
15 16.47
poverty declines monotonically but with 10
5 5.77
2.6
0
less margin. It is important to highlight that
re 3 7 1 2 re
.5 ac s <= cres <= s <= cres <= >25 ac
<= 1 .5 < acre a acre a
variation in quality of land is important 1 3.5 < 7.5 < 15 <
while analysing poverty by land
ownership. Moreover, poverty results may be interpreted cautiously due to drought affect during
this survey period.
4.3 Poverty by household head characteristics
Table 3 shows headcount ratio by the characteristics of the head of household.
4.3.1 Poverty by educational attainment level of household-head
Education plays an important role in accelerating economic growth and reducing poverty. So the
relationship between education and poverty demands much more attention. There is an inverse
relationship between poverty and education of the household head. The higher educated are more
likely to have greater incomes and thus, have lower chances to be poor.
Figure 16: Headcount ratio by level of education of household head,
2001
70 63.49
60
50
43.43
40 37.14
%
33.27
30
20 14.37 22.96
8.74 17.31 14.56
10 5.09 4.99
1.48 1.228.72 0.068.1
0
Never attended Less than Completed Completed Completed Completed Completed Diploma/other
school primary Primary middle Secondary higher higher level
secondary
Contribution to national poverty Headcount
18
29. The Figure 16 shows that poverty rates are relatively high for households in which the head of
household has either no schooling (43.43 percent) or completed less than primary (37 percent) or
completed primary level (33 percent).
At higher levels of education, the likelihood of being poor becomes much lower. The result
suggests that the greatest gains accrue to education beyond primary level. Poverty rate decreases
from 33.27 percent with head having primary education to 23 percent approximately when
household head has completed middle level, showing a big fall of about 10 percentage points.
Poverty rates fall further with more education and it is around 9 percent where the head is
graduate or higher.
4.3.2 Poverty by literacy of household head
The data suggest that poverty is about 43
Figure 17: Headcount ratio by literacy of household
percent in illiterate household head head, 2001
compared to 24 percent in literate 68.12
household-head. Contribution to total 31.88
100
poverty by households with illiterate head
% 50
43.22
is 68 percent approximately against their 24.03
0
population share of 64 percent. Literate Illiterate
There are three policy messages. First Headcount Contribution to national poverty
literacy rate may be improved. Second, policy makers should find ways and means to increase
enrolment rate and reduce dropout rates. Third, simply an increase in enrolment rate will not be
sufficient to reduce poverty. Level of educational attainment must be improved at least to middle
level to make a big dent in poverty.
Figure 18: Headcount ratio by sex of household
4.3.3 Poverty by sex of household head head, 2001
96.09
In developing countries the women are
100
disadvantaged in comparison with men.
35.27
50 21.97
One indicator of gender gap is whether 3.9
female-head households are worse than 0
Male Female
those headed by male. PIHS 2001 suggest
Headcount Contribution to national poverty
that the incidence of poverty is higher in
20
30. households having male head than female-headed household. The percentage of population
below poverty line is 35 percent in male-headed households compared to female-head
households where this rate is around 22 percent. The people living in female-headed households
account for 6 percent of total population and contribute only 4 percent approximately to national
poverty. The lower rate of poverty in female-headed households does not mean that women are
earning better than men. In female-headed households, most of adults are either working in other
areas of the country or overseas. The female-headed households receive domestic/foreign
remittances. This finding is reversed in cases where female-head is the only person working in
the household. However, it constitutes a small proportion of all female-headed households.
4.3.4 Poverty by status of employment of household head
Households can escape from or fall
in poverty depending upon their Figure 19 : Headcount ratio by status of
employment of head, 2001
earnings from employment. So it is 60
50
useful to look at the relationship
40
%
between poverty and employment 30
20
status of the head of household. 10
0
The Figure 19 indicates that poverty
or
r c ker
r
y)
r
er
er
ed
ed
...
pe
to
nl
at
th
oy
ac
oy
oy
va
or
op
(o
tiv
O
pl
is highest among sharecropper (50
pl
pl
lti
w
ly
cr
ul
Em
k
cu
m
em
al
oc
ily
e
fe
ic
ar
m
ct
st
ne
om
id
l
fa
tra
Sh
ve
Se
Pa
percent) followed by livestock (46
w
on
Li
d
on
O
ai
ec
C
np
ot
U
N
percent) and paid employed (38
Headcount Contribution to total poverty
percent). It is lowest among households
Figure 20 : Headcount ratio by sector of
where head is not economically active. employment of household head, 2001
This category includes pensioners and 90
80
those receiving income exclusively 70
60 45.2
13.3
50
%
6.31
from property such as landowners. The 40 7.93
10.62 13.85
30 55.69 2.79
20 38.05 35.99
people who receive remittances are also 10
32.74 27.63 27.44 22.65
0
included in this category. The ltu
re ing tio
n
ad
e
ag
e
ce
s he
r
cu tur uc l tr tor rvi Ot
gri fac nstr tai ds se
A nu Co &r
e an ity
distribution of poor with respect to the Ma le ort un
sa sp mm
ole Tr an Co
Wh
status of employment of household Headcount Contribution to total poverty
21
31. head indicates that paid employed contribute about 40 percent to total poverty. It is therefore,
useful to examine what is the sector of employment and occupation of paid employed head. Most
of them are employed in agriculture and construction sector. Figure 20 reveals that poverty rate
is around 56 percent, adding 13 percent to total poverty where sector of employment of head is
construction and it is 38 percent in case of agriculture sector, accounting for around 45 percent of
the total poor. Both agriculture and construction sector contribute around 58 percent to total
poverty. Another useful piece of
Figure 21: Headcount ratio by occupation of
information from the occupation household head, 2001
of head is that poverty rate is Elementary occupations
about 52 percent if the head has Skilled agricultural workers
Plant & machine operators
elementary occupation1 and it Craft and related trade workers
Service workers
contributes 32 percent to the total Clerk
Technicians
poor (Figure 21). It can be Professionals
Legislators & senior officials
inferred that informal sector is
0 20 40 60 80 100
main habitat of the poor. It %
suggests that income-generating Headcount Contribution to national poverty
policies in order to reduce poverty
should focus on the agriculture and construction sector. It should also not ignore those who are
doing elementary job. The poor can reap benefit from policies designed to improve the
functioning of rural markets, including those of agricultural commodities and inputs. Land
cultivation arrangements need to be improved keeping in view of the interests of land cultivators
(sharecropper and contract cultivators) where poverty rates are significantly high. Policies
directed at the formal labour market will not be much productive in reducing poverty. As most of
the poor people work outside of the regulated labour market, so strategies may be devised to
increase their incomes by improving their income generating capabilities.
5 Characteristics of the poor
Although consumption based poverty measures provide us with an easy tool for measuring the
distribution of living standards of population but they don’t fully capture other characteristics of
the poor such as literacy, health, and access to water and sanitation. This section presents an over
22
32. view of the distribution of selected non-monetary indicators of household living standards using
PIHS 2001.
5.1 Household size and dependency ratio
Household composition in terms of the size of the household and the characteristics of its
members is often quite different for poor and non-poor households.
Table 4: Household size, composition, and dependency ratio by poverty
status, 2001
Region Average Number of household members in Dependency ratio
and household the age group
status size 0-4 5-14 15-64 >=65 Total Child Aged
Years Years Years Years
Urban
Poor 8.95 1.40 2.81 4.46 0.28 100.67 94.39 9.90
Non-poor 6.43 0.75 1.59 3.85 0.24 67.01 60.78 15.09
Overall 6.87 0.87 1.80 3.96 0.24 73.48 67.42 13.33
Rural
Poor 8.42 1.49 2.73 3.84 0.36 119.27 109.90 13.19
Non-poor 6.31 0.85 1.75 3.40 0.31 85.59 76.47 17.71
Overall 7.00 1.06 2.07 3.54 0.32 97.46 88.42 15.46
Overall
Poor 8.52 1.48 2.74 3.96 0.34 115.15 106.57 12.41
Non-poor 6.34 0.82 1.70 3.55 0.28 78.87 70.99 16.47
Overall 6.96 1.00 1.99 3.67 0.30 89.65 81.47 15.08
Table 4 shows average household size and dependency ratio. The poor do tend to live in larger
and younger households. They have larger
Figure 22: Dependency ratio by poverty
household size than non-poor. Poor households status, 2001
Poor Non-poor
comprise, on average, 8.52 members which
compare with mean of 6.34 for non-poor 78.87
Overall 115.15
85.59
Rural
indicating on average more than two extra 119.27
67.01
Urban 100.67
persons in poor households relative to non-poor.
0 20 40 60 80 100 120 140
Like household size, household composition also %
differs by the status of poverty. This relationship
holds for both urban and rural areas. The key finding is that dependency is higher in poor
households than non-poor. The total dependency ratio is defined as the number of members in
1
It includes street vendors, porters, shoe cleaning, agricultural labourers, mining and construction labourers etc.
23