LIBERALISERS like Manmohan Singh can take heart from a new Princeton study which shows, contrary to the government’s statistical methodology, that rural poverty in India declined sharply between ’87-88 and ’93-94. This may influence the debate within the Congress Party on economic policy.
The new methodology, using data from the National Sample Survey, has been developed by two distinguished economists, Angus Deaton and Alessandro Tarozzi. It suggests that the proportion of people below the poverty line fell from 38.29 per cent in ’87-88 to 32.84 per cent in ’93-94, a decline of 5.45 percentage points. The official estimate, based on the methodology of the Lakdawala Expert Group, shows only a marginal decline of 1.97 percentage points, from 39.18 per cent to 37.21 per cent.
The official estimates show a steep fall in urban poverty from 40 per cent to 32.62 per cent. The Deaton-Tarozzi methodology shows a much milder drop from 22.45 per cent to 18.11 per cent. Note that this methodology shows urban poverty levels as being far lower than rural ones. By contrast, the official Lakdawala methodology shows urban poverty as higher than rural poverty in ’87-88.
Many layfolk will find the Deaton-Tarozzi version more credible. Migration from villages to towns looks like a sure indicator that income prospects are better in towns.
The official Lakdawala methodology relies on the Consumer Price Index for Agricultural Labour (CPIAL) to estimate changes in the rural cost of living. But NSS household data themselves contain information on quantities and outlays of households for a variety of goods.
From this, Deaton and Tarozzi have been able to calculate prices actually paid by households. These prices suggest a significantly lower level of inflation than official indices like the CPIAL or the Consumer Price Index for Industrial Workers (which, with modifications, is used officially to estimate urban poverty).
The official price indices belong to a family of indices which statisticians call Laspeyres Indices, which are weighted to reflect household consumption in the base year. For the CPIAL, the base year is ’60, making it one of the most obsolete indices in use anywhere in the world. A Laspeyres Index always tends to exaggerate inflation because it overemphasises goods used in the distant past and fails to capture increased use of new goods or goods whose prices have fallen, thus encouraging consumption. This problem is not restricted to India: the USA officially admits to exaggerating increases in the cost of living by relying on a Laspeyres Index.
Another standard statistical tool is the Paasche Index, using the weights of the final year instead of initial year, but this tends to understate inflation. So Deaton and Tarozzi have used a more neutral index called the Tornqvist Index, and created a Tornqvist price index for rural India based on prices emanating from the NSS data.
They have also estimated rural-urban price differentials, and used these to create a separate price index for urban areas. In recent years, economists have worried about the huge difference in living standards suggested by the national income data of the Central Statistical Organisation (which show vibrant growth) and those of the NSS (which show virtual stagnation in per capita consumption).
The Deaton-Tarozzi study suggests that the difference may not be so large after all, once we use an appropriate price index.