Opinion
Poverty can be measured well by means other than money metrics
Summary
The absence of consumption-spending survey data does not mean that we lack knowledge on levels of deprivation in India
Recently, Niti Aayog released poverty estimates for India using a Multi Dimensional Poverty Index (MDPI), an approach which has been promoted by the UNDP. The publication has led to a predictable surge in interest on the issue of poverty in India. The Niti Aayog report, like the UNDP report released last year, shows that India has seen a significant decline in poverty between 2015-16 and 2020-21. The two approaches are broadly similar, with the Niti Aayog report adding a few additional indicators to those used in the UNDP exercise.
The UNDP (along with the Oxford Poverty and Human Development Initiative) developed a Global MDPI in 2010 using indicators across three dimensions—Health, Education and Standard of Living—to measure poverty. The index is a mix of input indicators (access to electricity, safe drinking water, sanitation, clean fuel and housing) and outcome measures (such as nutrition, infant mortality and ownership of fixed assets). A few are difficult to classify, such as participation in and level of schooling. However, a unifying feature of these indicators is their high correlation with poverty. Changes in these indicators require sustained improvements in living standards. This attribute makes these indicators attractive proxies to assess poverty, an approach that is different from the earlier methods of linking poverty to income or consumption.
Money-based measures of poverty are primarily promoted by the World Bank. Their roots lie in an approach to poverty measurement used in India. While our tradition of discussing problems of poverty goes back to the pre-independence era, in the early post-independence years, we did not have any official measure for poverty. In 1971, the Indira Gandhi government came to power on an agenda of “garibi hatao" (end poverty). Notwithstanding the political salience of this campaign, there was no official approach to measure poverty till the Planning Commission constituted a task-force in 1977. Its report in 1979 outlined a framework to estimate poverty. Given the limited data options at that time, it decided to base its estimation on the NSS consumer expenditure survey. This, thanks to the restructuring of the NSS in 1970, had become available on a regular, comparable basis. It should be appreciated that the original objective of the consumption survey was not poverty estimation, but rather to provide data for updating input-output tables and National Account computations.
The task-force calculated the expenditure required for daily per capita calorie needs: 2,400 for rural and 2,100 for urban areas. This expenditure norm was supposed to be subsequently updated using data from the National Accounts. The indexation procedure was revised in 1991. Further refinements were recommended in 2009 and 2013, while retaining the essential character of a calorific norm linked to the consumer expenditure survey. This approach of an expenditure norm based on calorie intake has been criticized on various grounds over the years. Criticisms include the inadequacy of calorie counts to measure malnutrition; changes in consumption patterns over the years (especially on account of basic services like health and education); and improvements in public distribution (resulting in better availability of cheap foodgrain). In addition, there are conceptual problems relating to expenditure or money-metric norms as a yardstick for poverty.
Poverty is primarily a status variable, akin to a stock concept. Consumption or income, however, are flow variables. Measuring status from flow variables requires a large number of assumptions relating to savings behaviour, access to financial markets, etc. To illustrate, consider this analytical example: “In a society where we assume there are no instruments of income smoothening, if all households receive fluctuating income, and therefore consumption (as all income is consumed) of either 100 or 0 with equal probability, then on average, each household will earn 100 every 6 months, and in the remaining 6 months, earn nothing. Further, the social norm for decent living in this hypothetical society requires a consumption of 40 every month. The average monthly income of an household is then 50, i.e., above the poverty norm of 40. So if households were able to save or otherwise transfer resources from earning to non-earning periods, the overall incidence of poverty would be zero. However, given our assumption of zero savings, in any cross sectional settings 50% of the households have zero income and consumption. So the poverty measured in a consumption survey would be 50%. However, from a household viewpoint, since they spend six months being poor, they all perceive themselves as being poor, i.e. a poverty incidence of 100%."
Note that this simple example illustrates poverty being linked only to inter-temporal volatility in consumption, and the inability to save and invest. In an agrarian setting, volatility is further linked to better irrigation, integration of markets, etc. An improved ability to make inter-temporal transfers of earnings would reduce measured poverty. However, until it improves to the point where households never fall below the minimum threshold, some poverty self-perception will remain. This example also highlights how improvements in statistical measures of poverty do not necessarily change household perceptions of deprivation.
Note that the problems were not in the act of measurement, or the application of estimation tools to the proposed measure, but in the dissonance between flow attributes and stock perceptions. Measures like the MDPI seek to address such concerns. Thus, criticisms being made in some quarters that the absence of a consumption survey implies lack of knowledge about poverty in India are uncalled for.
T.C.A. Anant is a former chief statistician of India
