The author teaches at the University of Warwick, United Kingdom

The National Sample Survey Organization has just released the latest estimates of its household consumption expenditure survey. This shows that 25 per cent of people were below the poverty line in 2000-01. How significant are these figures?

There are many who feel that poverty levels at a point of time are not particularly important in themselves. They are important only in so far as they reveal something about the trends in the incidence of poverty at different points of time or if they allow observers to compare levels of poverty across different geographical regions at a point of time. So, what do these poverty levels reveal about the changes in the incidence of poverty during the last decade?

There is a saying that if one puts together any ten economists, then one is bound to get eleven opinions since economists can never agree on anything. The situation is not quite as bad in so far as the incidence of poverty in India is concerned — any group of economists will supply at most three answers! Some will assert that poverty has fallen steadily during the Nineties, a second group will claim that poverty has actually increased during this period, while a third group will say that nothing definite can be said. This difference in opinion must come as a surprise to the uninitiated. After all, how can there be any difference in opinion about two numbers — the proportion of people below the poverty line in, say, 1993-94 and the same proportion in 2000-01?

Unfortunately, there may well be good reasons to question the comparability of such numbers. Opinions become more extreme when such numbers are used (or abused) to support particular ideological positions. This has certainly been the story of poverty comparisons in India during the decade of the Nineties. Opinions about the trends in poverty levels in the Nineties have in many cases been shaped by observers’ positions in the ideological debate on the desirability of pursuing market-friendly reforms.

For instance, the pro-reformers point out that there has been a clear fall in the incidence of poverty between 1993-94 and 1999-2000, and conclude that reforms do have “a human face”. Not so, claim those for whom “market-friendly reforms” were conceived in hell (read White House, Washington DC). They point out that the “sampling design” of the NSS survey changed between 1993-94 and 1999-2000 in ways that cast doubts on the comparability of the resulting poverty estimates. In particular, in all earlier rounds, the NSS poverty estimates were based on what people answered to the question about how much they had spent on various items in the previous 30 days. In the 1999-2000 survey, households were simultaneously asked how much they had spent in the previous week as well as in the previous 30 days.

This simultaneous question prompted respondents to cross-check and validate the response based on one recall period with that based on the other. To cut a long story short, it turns out that how one interprets the 1999-2000 NSS data in comparison to those of previous years depends on which sets of estimates are used. If one uses the 30-day recall for food and ignores everything else, then the consumption data for 1999-2000 indicates a fairly sharp fall in the incidence of poverty. However, if one compares the data for the seven-day recall period in 1999-2000 with similar data for the previous four years, then there is some increase in the incidence of poverty.

Of course, it was ridiculous for the NSS to suddenly change its sampling design. Fortunately, it has learnt from its mistakes, and has reverted to its old sampling design. Does this allow us to compare the 1993-94 poverty estimates with those of 2000-01, and so claim that there has been a significant drop in the incidence of poverty between these years?

Unfortunately, there are more hurdles to cross. The NSS does not sample the same number of households across different years. The NSS surveys a relatively large number of households every five years or so. During the intervening years, a much smaller number of households are surveyed each year in order to economize on the cost of collecting this information. These are called the “thin” samples. Now, the reliability of estimates based on sample surveys depends on the size of the sample — the larger the sample, the more reliable are the estimates. It turns out that the 2000-01 survey was a “thin” sample survey, while the 1993-94 results were based on the large survey.

The NSS has released these estimates so recently that the controversy about the comparability of “thin” samples with the five-yearly large samples may not have started. But, of course, statisticians can correctly point out that the estimates for the two years can be made comparable by adjusting the estimates to take account of the difference in the size of the sample — correcting for standard error in statistical jargon. As I write this, I do not know whether the recent NSS estimates for 2000-01 have been corrected for standard error or not. Even if they have not been, more disaggregated data suggests that the anti-reformers will have to work overtime to build up a case against reforms on distributional grounds.

More or less everyone agrees that the reforms process will increase the rate of per capita growth. The controversy is over whether reforms promote a type of growth which is inherently anti-poor. Fortunately, the Nineties have witnessed variations in overall growth rates as well as changes in the incidence of poverty across different states in India. This implies that a detailed analysis of the inter-state variations in poverty and growth parameters is revealing.

There is clear evidence that higher farm yields as well as higher non-agricultural output per person lower poverty in all states. In other words, there is no evidence suggesting that high-growth states have witnessed a worsening distribution of income so much so that the benefits of higher per capita income have been wiped away by a greater degree of inequality. Of course, the suggestion that increased rates of growth help to reduce the incidence of poverty does not mean that there is no “space” for specific anti-poverty measures. Walking on two legs is always better than walking on one.