Poverty in India is yet again being overstated

Progress on poverty has a large political dimension—if growth in gross domestic product (GDP) matters for elections in advanced economies, poverty decline matters most in emerging economies. India has a national election in a year, and several important state elections ahead of it. Hence, the need for a proper understanding of what has happened, is happening, and will likely happen to poverty in India is a subject matter for an economic and political discussion. G-K’s contribution needs to be seen in this objective and positive light; however, as I document in some detail below, their results and inferences are more than problematic.

Non-Release of NSS 2017-18 data

Let us start by getting some cobwebs out of the way. Almost all non-government “experts” criticized the government for not releasing the 2017-18 Consumer Expenditure Survey (CES). I remember writing that the government should release the survey data, and I added, if only so that the world (and India) would find out how bad the data were. There will be time for a forensic study of the unreleased CES17-18 data.

Is there a reasonable explanation as to why the government did not release these data? Leaked reports on the 2017-18 data suggested that the data was, at a minimum, spooky. Rural consumption growth of -9.2 %, urban at +2 %, for an average decline of -5.6 %. Rural India, according to leaked CES (hereafter LkCES), which accounts for 70% of the population and most of the poor, witnessed a 10% fall in food consumption.

One defense of the government decision to not release the data could be that once released, the data would obtain an aura of respectability and authenticity, with international organizations (like the World Bank) forced, by their own rules, to accept the data as a “true” reflection of consumption and poverty in India in 2017-18, just a year after the controversial demonetization in November 2016! It is very likely that history will reveal LkCES to be a deeply flawed survey; hence better to release the data and be confident that competition among researchers and international organizations will reveal the reality, sooner than later.

Improvements in ew 2022-23 CES

If newspaper reports are to be believed, the government is undertaking two back-to-back consumer surveys (for agricultural years July-June 2022-23 and 2023-24) so that the economy, and polity, is better informed of the changes in livelihoods. However, even before the results are out, or have been tabulated, the critics of the non-release of the 2017-18 data (e.g. Himanshu, Mint) have started to raise some doubts as to whether we will ever see an end to the poverty debate. The reasoning? The 2022-23 survey is based on a completely revised schedule (Himanshu, Mint, 10 April).

This Himanshu “objection” or classification of 2022-23 survey is both peculiar and surprising. The questionnaire remains pretty much the same as 2011-12, except, to get more accurate reporting, each household will be visited thrice instead of once; once each for questions pertaining to seven-day consumption (fruits and vegetables), once for 30-day consumption (most items), and once for more durable items (recall period of 365 days). What could possibly be wrong with this procedure?

The recommendation to have food questions on a weekly basis recall (rather than monthly) was started in the late 1990s by World Bank and academic experts, including Nobel laureate Angus Deaton. The statistical authorities in India (NSSO, now MoSPI) listened, and post 2011-12, the Modified Mixed Recall Period (MMRP) method has become the official, and only, method of data collection by NSS. Is Himanshu suggesting that improvements in data collection should not be made because it might offend the purists’ desire to have a “consistent” series of poverty estimates? Around the world, changes in methodology of data collection are constantly made for important policy questions (e.g. should women be considered employed if they are working as unpaid labour in a family enterprise? Should women be considered employed if they take care of the children or the elderly?). Given that we have all criticized the NSSO for not releasing the 2017-18 data, and delaying release of the Periodic Labour Force Survey (PLFS) data (rectified now), let us congratulate the NSS for the important innovations they are making in collection of data.

Data and conclusions in the G-K study

This is a relevant and all-important detour into some methodological concerns I have with regard to the G-K study. In a chart entitled “Estimates Aplenty” that showed varying estimates of poverty, they point to a Mint estimation based on LkCES of poverty in India between 2011-12 and 2017-18 that it is stuck in a flat line around 22% of the population. In the concluding paragraph, the authors find support for the conjecture that the poverty rate has stayed constant—flat—for the last decade, and at around a high mid-twenties level. “Bibek Debroy, the chairman of the Economic Advisory Council to the Prime Minister of India, has calculated the poverty headcount to be 18% in 2020-21 (using the PLFS). This is closer to our overall assessment that headcounts from a wide range of estimates puts India’s poverty at between 20% and 25% prior to the pandemic. In effect, not much appears to have changed over the last decade.”

Let this “conclusion” sink in. Over the last decade (2011-12 to 2021-22), per capita GDP in India has increased by a cumulative 52%. As contrast and as a comparator, real per capita consumption (national accounts data) grew at a healthy 3.13% CAGR, for an aggregate increase of 37%.

So average consumption increases by 37% and none of it is obtained by the bottom 25%? None?! When consumption surveys are not available, the World Bank takes 67% as a “reasonable” pass through of private final consumption expenditure (PFCE) growth to household consumption growth. I will leave it to another occasion to question the veracity of this somewhat unsubstantiated assumption/conclusion. (the interested reader is referred to the IMF Working paper authored by this author, Karan Bhasin and Arvind Virmani (BBV), April 2022, which contains a detailed discussion of several of the results discussed in this article). Even if 67% is taken as correct or informative, it still means that per capita consumption increased by 25% (given as 0.67*37), and none of it reached the bottom 25%?

But there are even more serious problems with the G-K conclusion. Let us look more closely at the stylized constant figure of 20-25% poor and the mid-point of that range—22.5%. It may be just a co-incidence, but 22.5% happens to be the 2011-12 estimate by the Uniform Recall Period (URP—all questions on consumption for the last 30 days—now discontinued by NSSO) method. This conclusion is wrong on several counts. First, 22.5 % was one of three estimates of poverty contained in the 2011-12 survey. According to the Mixed Modified Reference Period (MMRP) method, poverty in India in 2011-12 was 10 percentage points lower at 12.4%. Second, and more importantly, the 2017-18 survey did not collect data for any method other than MMRP. Hence, the conclusion of constancy of poverty rate at 22.5% post 2011-12 just does not arise.

Of course, one can argue that the leaked 2017-18 survey gave us the right estimate of poverty. But what is that estimate? On the basis of decile means provided by Subramaniam for 2011-12 and 2017-18, one can make a reasonably accurate estimate of MMRP poverty in 2017-18. Compared to 2011-12, and World Bank PPP $1.90 poverty line, we obtain the same level of poverty in 2017-18 as 2011-12—12.4 %. (The constancy is obtained due to a large increase in rural equality—a Gini co-efficient of 0.255 in 2017-18 compared to 0.287 in 2011-12—which compensates for the 9.2% decline in rural consumption). Constancy at 12% poor can be stated—but then one has to offer why the survey results are credible, and all other information about the Indian economy is incredible.

In conclusion, the LkCES survey is a world record—there is no other instance in world survey collection of consumption that shows a large negative decline with a very large increase in comparable national accounts data. The LkCES conclusion is similar to CPHS-CMIE surveys on female labour force participation in India (lowest in the world, lower than war-torn economies like Yemen, and at 8%; the official PLFS data reports the figure three times higher at 27% in 2021-22!).

Governments, politics and the trend of poverty decline

G-K start their article as follows: “As the 2024 elections approach, taking stock of the economy after a decade-long rule of the government led by Prime Minister Narendra Modi becomes a natural talking point.” We agree. In this context, it is important to explicitly compare the pace of reduction in poverty between the UPA years—2004-13—and the Modi years—2014-21. The World Bank has recently updated their poverty estimates for India and other countries (available at www.data.worldbank.org ). The reader can choose her own time-period of comparison (or contrast). No matter which approximate time-period is chosen to represent the two political regimes, the pace of poverty reduction, for two different poverty lines, is considerably higher during the post-2013 period. (Note that as pointed out in BBV, the pace of reduction is even higher during the Modi years if the consumption of free or subsidized food is incorporated into the analysis.)

Sticking with the World Bank estimates, the pace of poverty reduction in the Modi years (2015-2019) was almost twice the rate of reduction between 2004-05 and 2011-12, show calculations based on the compounded annual growth rates in the poverty headcount ratio. Of course, the World Bank estimates of poverty post 2011-12 are synthetic estimates, as are all other estimates. Several other conclusions follow. There is no year in which poverty did not decline, and note the large poverty decline of 5 percentage points between 2016 (year of demonetization) and 2017. Second, the pace of decline in whatever combination of years is taken, is in the narrow range of 15-17% decline per annum. Third, all World Bank estimates refer to the URP method (both prior to 2011-12 and synthetic estimates post 2011-12). As noted above, the URP method is history and will be especially so once exclusive MMRP results for 2022-23 are published.

Three organizations (NSSO, NCAER and CMIE) have ongoing surveys on consumption and therefore estimates of poverty. I suspect we have reached the end of conventional debates on poverty in India. Much like Mr Praline’s dead parrot in Monty Python’s Flying Circus, the argument over Indian poverty will soon cease to be. Those that have been swimming statistically naked will be identified. And the dictum “let us publish any poverty estimate” will no longer be advisable.

I was privileged to be part of the original “Great India Poverty Debate”. To term the present discussion as continuation of the same (as done by G-K) is grossly incorrect; it is now time for all of us to progress onto the next stage of the discussion—which policies work best for the alleviation of absolute poverty; and is it time to graduate to a relative concept of poverty (as has happened in advanced economies). And is there any other country, or society, rich or poor, which is as primitive, or argumentative, as ours in discussing non-policy issues?

To facilitate a transparent debate nee discussion, my colleagues (Arvind Virmani, Tirtha Das, Abhinav Motheram and Karan Bhasin) and I have developed a website, www.living-standards.org. We plan to periodically update this website with state-specific information so any researcher can easily, and transparently, indulge in the discussion, and come-up with her own estimates of poverty.

Surjit S. Bhalla is former executive director of IMF and former member of Prime Minister Modi’s Economic Advisory Council.