Consumption-based poverty estimates have relevance Premium

Multi-dimensional poverty estimates are not substitutes for National Sample Survey consumption-based poverty ratios

‘There is ground to believe that the rate of reduction in the poverty ratio must have slowed down. This is at best a guess. We need to wait for consumption expenditure survey data’ | Photo Credit: Getty Images/iStockphoto

A recent report by NITI Aayog on multidimensional poverty shows that the percentage of the poor has gone down from 25% in 2015-16 to 15% in 2019-21 and around 135 million people were lifted out of poverty during this period. The Global Multidimensional Poverty Index report of 2023 of the United Nations Development Programme (UNDP) and the Oxford Poverty & Human Development Initiative (OPHI), which was released recently, also shows that the incidence of the multidimensional poverty index declined from 27.5% in 2015-16 to 16.2% in 2019-21.

In this context, we briefly examine the issues, particularly on methodology relating to the multidimensional poverty index, and argue that consumption-based poverty estimates are still very relevant. Multidimensional poverty estimates are not substitutes for National Sample Survey (NSS) consumption-based poverty ratios. In the end, we also flag some concerns about consumption expenditure surveys and the need to correct them.

Comparison of results

The report of the Global Multidimensional Poverty Index (MPI) 2018 says: “India has made momentous progress in reducing multidimensional poverty. The incidence of multidimensional poverty was almost halved between 2005/06 and 2015/16, climbing down to 27.5 per cent. Thus, within ten years, the number of poor people in India fell by more than 271 million — a truly massive gain”. This is high praise indeed.

Is the conclusion of global MPI a new revelation? No, as far as the 2015-16 estimates are concerned. The estimates of poverty based on consumer expenditure and using the Tendulkar committee methodology show (over a seven-year period between 2004-05 and 2011-12) that the number of poor came down by 137 million despite an increase in population. According to the Rangarajan Committee methodology, the decline between 2009-10 and 2011-12 is 92 million, which is 46 million per annum. For a decade, it will be larger than that of global MPI. However, in absolute terms, the poverty ratios based on the Tendulkar and Rangarajan Committee methodologies are lower than as estimated by global MPI.

The search for non-income dimensions of poverty possibly stems from a view that in terms of the capabilities approach to the concept and measurement of poverty, some of these ‘capabilities’ may not be tightly linked to the privately purchased consumption basket in terms of which the poverty lines are currently drawn. Therefore, poverty based on income or consumption is different from deprivations based on education or health.

As pointed out by the Expert Group to Review the Methodology for Measuring Poverty (2014), there are reservations on using multiple indicators as these multidimensional indicators/measures raise several issues regarding their measurability, aggregation across indicators, and, crucially, of databases that provide the requisite information at reasonably short intervals. These need to be considered and evaluated carefully. For example, there is a problem with the child mortality indicator as it is for population groups and not for households.

Aggregation is another problem. In principle, they should be independent. Access to safe drinking water, for example, cannot be aggregated with indicators such as child mortality. Even in respect of independent indicators, analytically appropriate rules of aggregation require that all of them relate to the same household. More generally, this requirement poses several data constraints.

It may be noted that we are not against multidimensional poverty or deprivations. One can analyse the progress of non-income indicators such as education, health, sanitation, drinking water, and child mortality over time with income or consumption poverty. But, converting all of them into an index poses several problems. Deaton and Drèze (2014) also indicate that “it is important to supplement expenditure-based poverty estimates with other indicators of living standards, relating for instance to nutrition, health, education and the quality of the environment”.

On multidimensional issues, Srinivasan (2007) says viewing public services as another dimension besides consumption in a multidimensional conceptualisation of poverty is more fruitful. However, he is critical of multidimensional indices. He says that “collapsing many relevant but not necessarily commensurate dimensions into a single index defined as an arbitrarily weighted sum of disparate indexes makes little sense. The Human Development Index pioneered by the United Nations Development Programme is an example of an arbitrarily weighted sum of non-commensurate indexes. It certainly is not a multidimensional conceptualisation in any meaningful sense but simply yet another arbitrary unidimensional index”.

In the minds of most people, being rich or poor is associated with levels of income. The various non-income indicators of poverty are in fact reflections of inadequate income. Defining poverty in terms of income or in the absence of such data in terms of expenditure seems most appropriate, and it is this method which is followed in most countries.

We do not have official data on consumer expenditure after 2011-12 to make a comparison with trends in the multidimensional poverty index. The survey data on consumption expenditures done in 2017-18 have not been released officially. In the absence of such data, there have been several studies on poverty using indirect methods and using Centre for Monitoring Indian Economy (CMIE) and Periodic Labour Force Survey (PLFS) data sources — and they have come up with differing conclusions.

Need for changes in surveys

The consumption expenditure survey is being conducted in the current year. For purposes of comparison, we need to follow one method. Therefore, it is best to wait for the survey results to be published. Earlier surveys clearly indicate that the poverty ratio comes down strongly during a period of high growth. If you look at recent years including the COVID-19 period, the growth rate has come down. There is ground to believe that the rate of reduction in the poverty ratio must have slowed down. This is at best a guess. We need to wait for consumption expenditure survey data.

An important issue is the differences in aggregate consumption estimates between National Accounts Statistics (NAS) and NSS data. These two estimates of consumption (NSS and NAS) do not match in any country; India is no exception. What is perplexing is that the difference in India between the NSS and the NAS consumption is widening over time. From a difference of less than 10% in the late 1970s, it has come to 53.1% in 2011-12, i.e., the Survey Estimate is only 46.9% of NAS estimates. The difference is too big to be brushed aside. The National Statistical Office must study the problem and come out with possible suggestions to improve the collection of data through both routes.

In addition, there is a need to supplement the results of consumption surveys with a study of the impact of public expenditure on health and education of different expenditure classes.

C. Rangarajan was Chairman, Expert Group to Review the Methodology for Measuring Poverty (2014) S. Mahendra Dev was Member, Expert Group to Review the Methodology for Measuring Poverty (2014)