There has been a lot of interest in the factsheet just released by the Ministry of Statistics and Programme Implementation on the latest Household Consumption Expenditure Survey (HCES). The interest is understandable. We haven’t had the HCES since 2011-12. There was one such survey in 2017-18, but data weren’t released. Therefore, any information on poverty and inequality was either based on 2011-12 numbers, extrapolations of those, or data sources not quite robust and reliable.
In case someone mentions the Multi-dimensional Poverty Index (MDPI), reports by UNDP and the Niti Aayog using it, that’s not quite a measurement of poverty as we have traditionally defined and understood it. MDPI is based on three dimensions of ‘poverty’—health, education and living standards. There is no denying the importance of such indicators in capturing development, but health and education indicators (nutrition, child mortality, years of schooling, school attendance) should be explanatory variables for poverty. They don’t quite measure poverty. Living standards indicators are what the government has sometimes dubbed “basic necessities” and, indeed, ensuring that every individual or household has access to these is part of the inclusion agenda pushed by the government.
Having said that, in January 2024, the Niti Aayog released a discussion paper. The last proper MDPI computations for India are based on NFHS-5 (National Family Health Survey) for 2019-21. The discussion paper extrapolated NFHS-5 and estimated an MDPI-based poverty ratio of 11.28 percent in 2022-23. This was an all-India number. It wasn’t separately disaggregated into rural/urban, though there was disaggregation according to states.
Often, greater sophistication in definitions and measurements doesn’t help incrementally, because development indicators are often correlated with one another. A simple measure will provide as much illumination as a more complicated one, even though the latter possesses greater intellectual rigour. That was one virtue of the UNDP’s Human Development Index—it was simple to understand.
Traditionally, poverty measurements involved defining a poverty line and obtaining data (such as the HCES) to compute what percentage of households (separately rural/urban and states) were below that poverty line, the so-called head-count ratio. This too was simple to understand. (We don’t have official surveys on income distributions. Our official surveys are on consumption expenditure, and inequality based on consumption expenditure will be lower than that based on income.) There is an international poverty line (per-capita per-day expenditure) based on purchasing power parity (PPP) prices. As economies develop, the poverty line ought to change, increasing upwards. Our understanding of ‘basic necessities’ changes.
The international poverty line can be $1.9, $2.15, $3.65, or even higher. Most measures will use $1.9 or $2.15 (these are based on 2017 prices). In 2022, the World Bank published a policy research working paper on poverty in India. Using $2.15, it computed a head-count ratio of 10 percent in 2019, a decline from 22.5 percent in 2011. The figure was 11.9 percent for rural India and 6.4 percent for urban India. Notice two points about these numbers. First, MDPI poverty of 11.28 percent in 2022-23 and PPP poverty of 10 percent in 2019. The years don’t correspond exactly, but they do roughly. This is what I meant by different approaches, with differing degrees of sophistication and methodology, validating each other.
We are now in 2024. Four years ago, we can roughly agree that poverty in India was 10 percent. Second, the World Bank found a decline in poverty by about 1.5 percent a year, Niti Aayog by about 2 percent. Very roughly again, not a great difference.
People are excited because the HCES now enables us to gauge poverty and inequality more precisely—sort of. Note that we don’t use the international poverty line. Our poverty line is based on what the Tendulkar Committee recommended. There was a Rangarajan Committee which recommended a different poverty line higher than Tendulkar’s, but that was never officially accepted. The official one still continues to be Tendulkar, which corresponds to $1.9 PPP. Tendulkar hasn’t been replaced by MDPI, not officially.
The numbers in 2011-12 were the following—head-count ratios of 25.7 percent for rural India according to Tendulkar and 30.9 percent by Rangarajan; head-count ratios of 13.7 percent for urban India according to Tendulkar and 26.4 percent by Rangarajan. That’s the baseline, so to speak.
Care is called for. We only have a fact sheet, not unit-level data. This is a 2022-23 survey. There will be another one, back-to-back, in 2023-24, to check for robustness. Methodology has changed, rendering the two sets of numbers non-comparable. Ideally, there should have been another sample, using both the old methodology and the new, to check whether apples are being compared with oranges. But there are people who think fruit juice is fruit juice, regardless of whether it comes from apples or oranges.
Therefore, it is tempting to do the following: use inflation rates from the HCES factsheet to figure out what the Tendulkar poverty line is for 2022-23. Naturally, this will be separately done for rural and urban households. The factsheet gives us the distribution of monthly per capita consumption expenditure across fractile classes (fractile is similar to percentile). The bottom 0-5 percent and the bottom 5-10 percent should suffice. Those are the fractiles where poverty will be concentrated. SBI Research has come out with one such calculation and finds a poverty ratio (in 2022-23) of 7.15 percent for rural India and 4.62 percent for urban India. For rural India, compared to 2011-12, this is a decline of 1.7 percent a year; for urban India, this is a decline of 0.8 percent a year.
These numbers are for 2022-23. For 2020, I mentioned a rough figure of 10 percent. Note that the SBI Research numbers are significantly below 10 percent. A factsheet is based on facts and one can’t quarrel with facts and numerical calculations. But note that consumption (and income) distributions tend to be normal, not symmetric. When the thick part of the distribution passes above a poverty line, declines in poverty ratios are sharp, not linear. But after that, it becomes more difficult. In many countries, poverty ratios thus get stuck at around 5 percent.
I think a decline to around 8 percent in India would have been regarded as plausible, a priori. A decline to around 5 percent is possible, but requires detailed probing and validation. Note also that poverty declines have been relatively less in urban India, highlighting a rural/urban convergence, but also some urban problems. Everyone will wait expectantly for the 2023-24 survey.
Bibek Debroy
(Views are personal)