Poverty in Pakistan Context

Earlier Studies: There is abundance of studies on poverty in Pakistan. Most of these have made use of the calorie-based approach and estimated food poverty — the proportion of population unable to afford expenditure needed to meet the minimum calorie intake. A few studies attempted to quantify total poverty — proportion of population whose income or consumption falls short of the expenditure needed to meet the minimum requirements of all basic needs, food as well as non-food. Except for one or two, even the latter studies centred around the calorie approach. While justification for adopting this approach emanates from the fact that calorific requirements are the most important among human needs, its usual problems are well known; fixation of minimum calorie intake, selection of food basket which yields the minimum calories, conversion of this basket into a monetary equivalence, to enumerate a few. A study done by Ahmad (1993), estimated minimum requirements on basic needs independently of the calorie-based approach but on adhoc basis. The approach adopted was that of holding discussions with professional economists and re-checking with heads of different families, regarding the quantum and value of each of the basic needs. Estimates based on such a methodology lack generality as these are based on the response of a limited number of persons. The World Bank, in a recent study (World Bank, 1995a) adopted several judgment-based modifications to Ahmad's estimate to arrive at their reference poverty line. A critical survey of these studies may be seen in the author's recent work (Ali, 1995). Moreover, since each of these studies measures incidence of poverty for one or two or at most three years, the diversity of approaches and techniques adopted in the measurement yields results which are not strictly comparable and thus makes it difficult to trace out historical trends in poverty. There have been attempts to derive such trend by extending poverty line estimated for one particular year to other years through inflation adjustments; but such attempts reflect second best solution to the problem.

The present paper envisages to have three distinguishing features:

  • It deals with total poverty, rather than simply the food poverty;
  • For poverty estimation, it uses a more scientific technique which is based on the extended version, ala Lluch (1973), of the standard Linear Expenditure System originally developed by Stone (1954), hence after called ELES;
  • It uses the same technique for all years starting from 1969-70 up to 1990-91 so as to present the existing poverty situation against the background of its historical trend, derived on comparable basis.

Methodology of ELES: Without spelling out the ELES and its derivation in detail, for which the readers are referred to the original contributions of Stone and Lluch cited above as well as empirical applications particularly by Lluch et al (1974) and in the context of Pakistan by the author (Ali, 1995), it is enough to highlight three basic features of the system:

Firstly, it facilitates estimation of subsistence (minimum) expenditure on all consumption items (which can be termed as total poverty line) without reference to calories, composition of food basket, and even prices (if the system is solved for a cross section data on income and expenditure). This can be shown as follows:

Maximising the utility function implicit in the LES [U(X)=Σfi(xi)- Σai.log(xi-ri)] subject to the constraint that [Σpixi=Y] yields a demand function for ith good as:

vih - piri + bi(Yh-Σjpjrj)                                                                                                      (1)

Where h indicates the hth household, v the expenditure, p the, r the subsistence quantity, b the MFC for ith good with Σb representing aggregate MPC (u),Y the income, and the term (Y-Σjpjrj.) the supernumerary income which is allocated among goods in proportion to bs. Since ri occurs in all equations, the demand system needs to be estimated as a whole by a method which imposes cross equation constraints. In the case of time series, this would call for direct maximisation of the likelihood function associated with the system. For cross section, the term piri becomes independent of the unit of observation since each household faces the same commodity prices. Thus, the term piri can be replaced  by ri* which reflects subsistence expenditure in

prices prevailing at the time of the household survey. This implies a demand function as:

vih = ri* - bi + Σjrj*+biYh                            (2)        

Adding (2) across all goods, yields:

Vh = (l-u) Σjrj* + uYh - D + uYh                 (3)

which gives

Σjrj* = D/(1-u)                                                                (4)

Secondly, the ELES-estimated poverty line is considered robust and closer to reality as this is based on the perception of all those thousands of households whose income and expenditure patterns are reported in the household survey, and not on discussions with a few persons as in Ahmad (1993).

Thirdly, the ELES, being a dynamic approach, provides for a better inter-temporal comparison of poverty. As against the calorie approach wherein the reference point remains static at a given number of recommended calories per person per day, the reference point in the ELES is variable, shaped by the changing perception of the households about the minimum expenditure on different needs in different years.

Poverty Line and Incidence: The ELES technique was applied to the data of households income and expenditure surveys for six different years starting from 1969-70 with 1990-91 being the   latest year for which such data is available (Government of Pakistan, 1973, 1983, 1990, 1993). The results are reported in Table 1.




Poverty Line

(Rs per month

per capita)

Poverty Incidence

(headcount ratio)

@ (%)



















@ Headcount ratio, defined as the percentage of population whose actual consumption expenditure falls short of the reference poverty line, was computed  from the published data of households income and their share in population  across different income groups. The population share of the group in whose income lies the poverty line was apportioned by linear interpolation. The ratio thus computed may slightly differ from the ratio derived directly from the micro data. This also explains the difference between the incidence for 1990&-91 reported in this table and the one reported in the author's earlier paper (Ali, 1995).

There appears to be a definite trend in the incidence of poverty time. It was found to be significantly high at the close of the ‘60s and the beginning of the '70s, with incidence at around 77 percent in 1971-72, During the following 13 years, i.e., up to 1984-85, it tended to decline; the incidence coming down to around 52 percent in 1979 and further to 49 percent by 1984-85. In the second half of the '80s, the evidence reveals deterioration in the poverty situation, with incidence rising to 54 percent in 1987-88.

Current Position: The estimates for 1990-91 in the series though suggest marginal improvement in poverty over 1987-88, may be seen with caution owing to the reported non-compatibility of this year's household survey with the earlier ones. The compatibility also becomes doubtful as, when compared to 1987-88, the improvement in poverty is found accompanied with substantial deterioration in income distribution (see Table 4). In the absence of household survey for the period beyond 1990-91, the 1990-91 estimates may be taken as a reflection of the existing poverty situation in the country independently of the previous trend. Roughly half of the population is still below poverty line. Poverty gap (indicating shortfall between poverty line and actual expenditure) is 12 percent, though income gap (shortfall between poverty line and income) is around 24 percent. The extent of poverty is little bit more in urban areas (52.6 percent) compared to that in the rural areas (51 percent). Moreover, the poverty incidence appears to be vulnerable, as it may rise if income of some of the middle-income groups falls even moderately. This is because of a narrow difference between the nation’s average expenditure and the minimum requirements for several commodity groups.

How do these results compare with those of the World Bank’s recent study (World Bank, 1995a)? The study is limited to only 1984-85, 1987-88 and 1990-91. Comparison is shown below.




Present Study

World Bank Study











Three observation can be made:

  1. Contrary to the findings of the present study, the bank's study suggests head-count ratio failing in 1987-88 compared to 1984-85, although movement is consistent between 1987-88 and 1990-91. This may be attributed to the static approach adopted in the bank's study wherein the reference poverty line computed for 1991-92 was extended backward for 1987-88 and 1984-85 through inflation adjustment, and thus failed to capture any change that might have occurred in the norms of minimum levels of expenditure owing to the overtime change in the national living standard. The present paper estimated poverty line separately for each year by using that year's income and expenditure data.
  2. The bank's estimates of incidence are lower in absolute terms. This is simply a reflection of the adopted lower poverty line by the bank.
  3. Contrary to the present study, the bank's study shows urban poverty to be higher than rural poverty. This is because the differential between urban and rural poverty tines estimated through ELES (Rs 470 and Rs 341 per month per capita respectively) is 37 percent, almost 3 times that assumed in the World Bank study (13 percent). In our case, the differential captures not only price differentials but also differences in the living styles of the two communities, as reflected in the survey data.

Regional/Gender Disparities: Although social indicators on national level will be covered under international context of poverty (Section III), the regional and gender disparities in these and in certain economic indicators provide an added information on the national dimension of poverty. Some of these indicators are reported in Table 3.









Overall urban rural

literacy rate (%) (1994-95)1






Primary enrolment ratio









Infant mortality rate

(per 1000 live births) (1992)1    





Labor force participation

rate 110+1 (%X1991-92)2








Unemployment rate (%)









Position beyond 1990-91: Though no survey-based evidence for the post-1991 period is available, indications hardly point to any improvement in poverty situation. Two indications are worth noting: (i) the employment elasticity which had already dipped to 0.21 in 1990-91 (see Table 4) declined further to 0.13 in 1991-   92; and (ii) the GDP growth remained depressed (below trend) for three consecutive years after 1991-92, resulting thereby only a marginal growth (0.8 percent per annum) in real GNP per capita during FY 92-95. Of course, one will have to wait for the new survey to confirm this suspicion.

Factors Affecting Poverty: Since poverty has a multidimensional context, it is very difficult to trace the perfect causation of changes in its quantum and pattern. Moreover, non-uniform sample sizes, questionnaires, and sample errors across various surveys further renders the causation difficult. Nevertheless, an attempt is made at least to compare movements in some of the variables which tend to contribute to poverty either positively or negatively, e.g., employment, wages, GNP per capita, and income distribution. Table 4 summarises movements in these variables for the relevant years.







Growth in



Real Growth in Wage bill


Real Growth in GNP per


Gini Coefficient











































1970-91 (annual average)





1981-91 (annual average)





The elasticities were computed through a Quadratic regression equation of employed labour force (L) and real GDP (Y) with 13 observations covering selected years between 1968-69 to 1991-92 for which Labour Force Survey reports are available. The estimated equation was:


L                =          b0               +blY                       +b2Y2


= 6.91                     + 9.037E-05 Y    -8.553E-11 Y2

t-ratio (6.9)                                 (11.6)                      (-6.4)

                                                                           R2-R2 = 0.99

                                                                           D.W. = 2.1

Growth rates are annual averages for the period covered between the war under reference and the previous year as recorded in the table, e.g., 3.4 for 1979 under col. 3 represents annual growth rate between 1971-72 and 1979.

The employment trends appear to support poverty trends at least during the '70s. The growth in employment was fast between 1972 and 1979 which is the period wherein poverty incidence dropped significantly. This was partly supported by a relatively high employment elasticity during this period (around 0.5). In 1987-88, high growth in wage bill helped containing deterioration in poverty despite substantial fall in employment elasticity.

During the entire period under review (FY 70 to FY 91), real GNP per capita rose by 2.2 percent per annum with rise during the 70s and first half of '80s when poverty declined, being more than the average (3.1 percent). The latter was made possible by huge inflows of remittances from abroad particularly from Gulf countries, which touched $2.9 billion mark in 1982-83. This had a significant impact on reducing of poverty as majority of the labourers who went abroad belonged to low-income category. The changes in income distribution patterns were found in line with poverty trends at least during the first half of the '80s though not during the '70s.

Current Position of Anti-Poverty Policies: There have been very few policies specifically targeted towards alleviation of consumption poverty, though indirect effects of the general economic policies particularly in respect of structural adjustment in agriculture, industry and physical infrastructure cannot be denied. The policies adopted under the social safety nets, however, were specifically anti-poverty in nature. Three such policies have been more important.

First, the zakah system: Introduced in 1980, the system envisaged to make compulsory deduction of zakah at source on various profit-earning accounts/assets and disburse it to poor (mustaheqeen) through zakat committees. At present, about Rs 2.8 billion are collected under the system, of which roughly two-thirds are disbursed through neatly-classified three heads, namely, subsistence cash payments, rehabilitation grants and scholarship/stipends to students in educational institutions. For 1992-93, over two million mustaheqeen benefited from zakah under these heads.

Net effect of zakah on  poverty  depends  on (i)  to what extent the disbursement is efficient, that is, recipients are really poor and leakages are minimum; and (H) whether and to what extent the compulsory deduction of zakah at source by the government replaced voluntary disbursement of zakah by individuals themselves. If there has been a strong crowding-out, that is, the compulsory deduction replaced voluntary zakah disbursement by the individuals, then the net impact on poverty would not be significant especially in a situation when government distribution channels are not very efficient. On the other hand, if compulsory deductions are considered entirely new (no crowding out), it had anti-poverty impact despite inefficient distribution. To know the net impact, there is a need to ascertain the substitution effect through some kind of survey investigations and research.

Second, the Baitul Maal Fund: Established in January 1992, the fund envisages to provide assistance to poverty groups, e.g., destitute and needy widows, orphans, invalids, infirm, etc., who are not covered under zakah safety net. The purposes for which funds are utilised include: rehabilitation in various professions or vocations; providing residential accommodation and necessary facilities; free medical treatment; setting up free hospitals, poor houses and rehabilitation centres; giving financial aid to charitable institutions, including industrial homes and other educational institutions established specially for the poor and needy; providing stipends for education, and finally sponsoring and promoting self-employment scheme. The fund also provides money for flood relief.

Baitul Maal gets funds from federal budget, amounting to roughly Rs 1 billion each year. The funds are disbursed in two forms; direct financial support to individuals; and through foodstamps. Food Stamps Scheme was introduced recently under which wheat flour, pulses and essential items are being provided to the poor at subsidized rates through the Utility Stores Corporation. Presently, about half a million persons are benefiting from Baitul Maal.

Third, self-employment schemes: These are schemes operated by public sector financial institutions to provide loans/credit at concessional rates to promote self-employment. Prime Minister’s Self-Employment Scheme is one such scheme providing loans ranging from Rs 50,000 to Rs 300,000 to both educated and uneducated unemployed persons. Recently, Federal Cooperative Bank has launched a similar scheme to be implemented through 40,000 registered primary cooperative societies, provincial cooperative departments and banks. Agricultural Development Bank advances loans for self-employment projects in agriculture, like, dairy farming, fruits/vegetable processing, animal breeding, etc.


Source: Poverty Alleviation in Pakistan: Present Scenario and Future Strategy, Mohibul Haq Sahibzada. Republished with permission.


Copy URL