THE QUESTION of poverty in the 1990s has been hotly debated in India. There have been different views on the extent to which poverty has declined in India and also the extent to which inequality has increased.
Official estimates of poverty level in India are based upon results of sample surveys conducted by the National Sample Survey Organisation ( NSSO). The NSS survey of Household Consumption Expenditure ( HCE) is the sole data source for calculating prevalent measures of poverty. Poverty of the country is measured as the percentage of population living below a pre-specified poverty line (also called head count ratio). In India the incidence of poverty is generally measured by shortfall in per capita consumption relative to a standard poverty line. This poverty line is the monthly expenditure incurred in getting a daily calorie intake of 2,435 kilocalories in rural and 2,095 kilocalories in urban areas. The Planning Commission computes state-specific poverty lines and ratios. The national poverty line is worked out implicitly and the all-India poverty ratio obtained as the weighted average of state-wise poverty ratios.
Sample Surveys on household consumption expenditure have been carried out in many rounds of survey by NSS since its inception in 1950. Major surveys are carried out every 5th year. Such surveys cover around 1,20,000 respondents whereas the minor surveys cover 20,000 only. The latest major survey (the 55th round) was carried out in 1999-2000. The results of this survey showed a drastic reduction in poverty. Official estimates of Government of India indicate that the poverty head count declined by approximately 10 percentage points (from 36% to 26%) between the 50th round (in 1993-94) and the 55th rounds of NSS (in 1999-200). However, official estimates based on the NSS 55th round are not comparable with the estimates of earlier round of surveys due to changes in survey design. There is a common consensus amongst the economists that changes in survey design have led to higher estimates of consumption and thus lower estimates of poverty compared to the 1993-94 round of survey.
The survey has a questionnaire in which various items of consumption are listed. The surveyor takes this questionnaire to various households and fills up the quantity/volume of consumption during a specified period ( called recall period) against each head. The recall period can be past 30 days, past 7 days or past 365 days. The households are selected on random basis.
Surveys designs are modified from time to time to reflect the prevailing economic conditions and to account for the changing structure of economy and consumption. At the same time the changes have to be such that comparability with past estimates is maintained.
NSS used a 30 day recall period for consumption since its inception in the 1950s. However, from 51st to 54th round (1993-94 onwards), the NSS administered two different consumption schedules (questionnaire) to two different households. While Schedule-I used the traditional 30 day recall period for all items of consumption, Schedule-II applied multiple recall periods for different items: 7 day recall for food items (including food, ‘paan’, tobacco and other intoxicants), 30 day recall for high frequency non-food items (fuel and light, miscellaneous goods and services, non-institutional medical items etc.) and 365 day recall for low-frequency non-food items (educational, institutional medical, clothing, footwear and durable goods).
In the 55th round (1999-2000) the design was changed further. In this round, food consumption was obtained by both 7 day and 30 day recall for the same set of households, while spending on low frequency non-food consumption items was obtained only using a 30 day recall period.
The problem with this change was that, typically, consumption reported in a 7 day recall period is higher by 30% compared to that reported for a 30 day recall period. Putting a 7 day and 30 day consumption periods together in a schedule leads to convergence. The respondent will generally tend to reconcile both figures while replying – and if respondent is asked to give a consumption figure for the last 7 days first and the last 30 days later, he will tend to simply multiply the first figure by 4.
Along with the poverty survey, a survey of Employment and Unemployment was also undertaken over the same period by the NSS. Its results showed an increase in the unemployment rate. Further, there has been a downward trend in the calorie intake, especially in the rural areas. If we go by the factors that may reduce poverty like real non-agricultural products per head, average farm productivity, state development spending, inflation rates and state-specific trends, they hardly lead to a conclusion of such a drastic fall in poverty.
There were 4 ‘thin sample’ rounds, i.e. survey with fewer respondents between 1995-96 to 1998-1999 where the recall period was kept at 7 days and 30 days for different households. Taking the results of the 55th round in conjunction with mutually comparable results from the NSS ‘thin sample’ rounds, and adjusting for some correction, it has been concluded that poverty during 1999-2000 was, at best, about 2 percentage points lower than in 1993-94. This means that the pre-55th round academic consensus that poverty reduction has slowed down significantly during the 1990s despite somewhat higher GDP growth, is upheld.
Some economists, hell bent on proving massive reduction in poverty, prefer to get around the problem by relying more heavily on National Accounts Statistics (NAS). Adjusted ‘private consumption expenditure’ (PCE) as per National Accounts is preferred to direct household consumption estimates as per NSS. One reason given for this is that there is an apparent divergence between NSS data and National Accounts. This is concluded because multiplying mean per capita consumption (as per NSS) with population revealed less and less growth in GDP when compared with NAS.
However, the divergence between NAC and NSS in the 1990s has been shown to be deceptive. In the 1990s, the Central Statistical Office changed its methods to be in tune with international standards. This resulted in a sizable upward revision in NAS consumption estimates. Once this effect is eliminated, this divergence disappears. Preference for PCE compared to direct figures from “raw data” is quite odd since PCE is a ‘residual figure’ in National Accounts, i.e., whatever is left after deducting certain items from total output is lumped there.
In such estimates “higher economic growth” as per National Accounts gives higher “private consumption expenditure”. This consumption is distributed among the population. This, when compared with the poverty line shows a drastic reduction in poverty. In this statistical sophistry, the point that one is going around in circles is missed.
It is certain that there has been an acceleration in the growth rate of the economy, but this has not made enough dent on poverty. Poverty in India has been falling at about 1% per year since the 1970s. The impression is that we have again come back to this trend after getting off it in the early 1990s. The problem is that we are needing ever more percent rates of growth in GDP to achieve it. While in 1970s we could reduce poverty with 2-3 % growth in GDP, we are now doing same with 6-7% growth in GDP.
Our pattern of growth in the 1990s has meant less poverty reduction than expected. The sectoral and demographic composition of growth in the 1990s has severely reduced the magnitude of reduction in poverty. The growth has not been found where it would have the greatest impact on poverty. A balanced growth process at the same rate would have doubled the rate of poverty reduction.
-- Sakthivel and Girish