Data war on India lifting 135-140 million out of poverty

On July 17, 2023, the NITI Aayog claimed that “135 million have exited multi-dimensional poverty” during 2015-16 and 2019-21, in its report “National Multidimensional Poverty Index: A Progress Review 2023”. How accurate is this assessment? Not very. Here is how.

To understand its claim, one must look at the source of its data.

The report says its estimate is “based on” National Family Health Survey-5 (NFHS-5) of 2019-2021 – which was released in two phases in 2020 and 2021. The Aayog’s baseline MPI of 2021 – on which the MPI of 2023 is built – was released on November 24, 2021 – and was “based on” NFHS-4 of 2015-16.

The first question that arises is: Why was the baseline MPI of 2021 prepared on the old NFHS-4 of 2015-16 data when the new NFHS-5 of 2019-21 data was already known?

Note, the day (November 24, 2021) the MPI of 2021 was published also saw the publication of phase II NFHS-5 of 2019-21 data – covering 11 states and 3 UTs.[3] The phase I data had been published long ago, in December 2020 – covering 17 states and 5 UTs. Also note, when the NFHS-5 data was released in phases, all data on a particular state/UT was released together. This means the Aayog had the NFHS-5 data on the majority of states/UTs by December 2020. All the rest became available in 2021 but going by the fact that the foreword of the phase II report was signed by Health Secretary Rajesh Bhushan on September 22, 2021, the Aayog could have had this data at least for three months (September to November 2021) in advance to work on.

Why did the Aayog rely on the old 2015-16 data is not known but it certainly is a very unusual practice for an economist.

The second question is about the NFHS-5 data of 2019-21: Why were the years 2019-2021 selected? Why not a single year of 2019-20 or 2020-21 – which is a standard practice in India and abroad? Remember, 2019-20 was a pre-pandemic year and 2020-21 was the first pandemic year (unusual year). Why mix the data of a normal year data with that of an abnormal year? The only logical explanation could be that the Aayog wanted to lessen the true impact of the pandemic on health, education, and living standard -- all three parameters on which MPI is built and all of these were severely hit in the pandemic year.

The third question arises out of the Aayog report (page 4) which reads: “It is important to note that the poverty estimates presented in this report may not fully assess the effects of the COVID-19 pandemic on poverty, since more than 70% of the data (NFHS-5) was collected before the pandemic. At the same time, this report does not capture the economic and social progress the country has made in the last two years.”

This statement is corroborated by the UNDP-OPHDI’s MPI report of 2022 (page 19[5]), released in October 2022, but missing from its abridged version released on July 11, 2023) which reads: “The effects of the COVID-19 pandemic on poverty India cannot be fully assessed because 71 percent of the data from 2019/2021 Demographic and Health Survey for the country were collected before the pandemic.”

The question then arises: Since most of the NFHS-5 data (“more than 70%” or “71%”) is on the pre-pandemic 2019-20, how could this represent the data for 2019-20 and 2020-21? It would only mean, the impact of the pandemic was captured to the extent of 29%. This would mean a highly skewed number for 2019-21.

The fourth question is: Why did the Aayog use only the NFHS data – which is about health – but not on education and standard of living (income/expenditure) – the other two components of the MPI? Another question relating to it is: How reliable then is Aayog’s MPI report of 2023? Or for that matter, its MPI of 2021 report?

Fifth question: Why didn’t the Aayog use the two standard parameters to measure poverty – poverty line and household consumer expenditure (MPCE) which acts as a proxy for household income?

The answers are known: (a) the poverty line was last fixed in 2004-05, called the Tendulkar poverty line – about two decades old and (b) the MPCE was last carried out in 2011-12 – a decade old; the one of 2017-18 was junked in the name of bad data quality but it was actually for showing that poverty went up between 2011-12 and 2017-18 for the first time in 40 years. Another related question arises: Why hasn’t the poverty line been updated and fresh MPCE data collected?

Sixth question: Isn’t this data vacuum giving rise to speculative and flawed estimates on poverty?

UNDP-OPHDI: 140 million lifted out of MPI poverty

While the Aayog claimed 135 million were lifted out of MPI poverty during 2015-16 and 2019-21, the UNDP-OPHDI reports of 2022 and 2023 (abridged version of 2022) said 140 million did so during the same period. (The UNDP-PHDI report said 275 million were lifted out of MPI poverty during 2005-06 (NFHS-3) and 2015/2016 (NFHS-4) also, taking the total to 415 million during 2005-06 and 2019-21).

Both use the same NFHS-5 data. Why does the UNDP-OPHDI use no other data? The answer is data vacuum.

The only difference between these reports is the rate of MPI reduction. The Aayog said it fell from 24.95% in 2015-16 to 14.96% in 2019-21, while the UNDP-OPHDI said it fell from 27.7% to 16.4% during the same period (it was 55.1% in 2005-06, which is NFHS-3). The percentage points matter because India’s population is estimated at 1.425 billion in 2023 (as per UN DESA). Apparently, the two reports used different population bases.

Seventh question: Why is this difference in fall percentage (9.99 percentage for the Aayog and 11.3 percentage for the UNDP-PHDI) when the data source is the same?

The answer: (a) different methodology and (b) two additional indicators (taking the total to 12) used by the Aayog: maternal health and bank account. Two eminent economists with extensive working knowledge of the Indian economy have objected to the Aayog’s methods and indicators. Pronab Sen, who heads the new Standing Committee on Statistics, said: “The way the MPI is structured, it is almost impossible to show a decrease in these indicators over a period of time”. C Rangarajan said: “The MPI takes into account bank accounts, which does not indicate welfare. Bank accounts will keep increasing over a period. Therefore, it needs to be discussed how relevant some of these indicators are.”

One must remind here that the Aayog has a questionable reputation in such matters. Its pandemic year report “SDG India-Index & Dashboard 2020-21” showed a dramatic fall in poverty, hunger, and inequality – of 28 states/UTs it mapped, poverty fell in 25, hunger in 23, and income inequality in 13. This was a dramatic reversal of pre-pandemic 2019-20 when poverty had gone up in 22, hunger in 24, and income inequality in 25 of those states/UTs.

It never explained how did poverty, hunger, and inequality went down during the two devastating pandemic waves, national lockdowns, and the economic collapse.

Data vacuum

The latest round of poverty debate, however, didn’t start in July 2023. It started in April 2022.

In April 2022, the IMF’s “Pandemic, Poverty, and Inequality: Evidence from India” and the World Bank’s, “Poverty in India Has Declined over the Last Decade But Not As Much As Previously Thought” provided fresh estimates of “extreme poverty” in India (per capita per day expenditure of $1.9).

The IMF report covered 2011-2020 and said: extreme poverty in 2019 (pre-pandemic FY20) was 1.4% – declining by 10.8 percentage points since FY12 – and in 2020 (pandemic fiscal of FY21) it declined to 0.8% due to food “transfers” (“free” over and above “subsidised” ration to 62.5% households), without which it would have been 2.48%.

The World Bank report covered 2011-2019 and said: poverty (at $1.9) declined from 22.5% in 2011 (FY12) to 10.2% in 2019 (FY20) – a drop of 12.3 percentage points but it didn’t explain why this happened.

The lead authors of these reports were Indian economists: Surjit Bhalla for the IMF and Sutirtha Sinha Roy for the WB. What data did they use? Both took the MPCE of 2011-12 as a base for their statistical projections into FY21 (IMF) and FY20 (WB). That is, these were statistical constructs – not based on actual data.

There are several problems with the IMF estimates:

· IMF justifies junking of MPCE of 2017-18 on the ground that an expert committee had questioned its data quality (but unbeknownst to it, the said committee had refuted this and upheld the data quality). At the same time, it used the very same NSSO’s MPCE of 2011-12 as its base.

· It’s the consumption data for post-2011-12 years taken from the 2011-12 GDP series (PFCE) which maps the consumption of all Indians, rich or poor, except government consumption (GFCE) and doesn’t tell which income segment group consumes how much. Besides, if then Chief Economic Advisor (CEA) Arvind Subramanian didn’t trust the 2011-12 GDP series (introduced in 2015), stating that it overestimated the GDP growth by 2.5 percentage points during 2012-2016, it is tough for others to trust it either (add to that frequent retrospective revisions).

· It assumes that all 62.5% of households received the full quota of PDS supply (money value was used for the calculations). This is wrong and the repeated tongue-lashing by the Supreme Court for not providing ration to millions of migrants in both 2020 and 2021 is a testament to that. The Global Hunger Index report of 2022 shows India is not only home to the maximum undernourished population (224.3 million or 27% of the world) but its progress reversed during 2014-2022 (hunger score went up from 28.2 to 29.1) – unlike others like Rwanda, Nepal, Pakistan, Myanmar, Bangladesh who continued to reduce hunger and all of them have a lower level of hunger in 2022 than India. The UNDP-OPHDI of 2022 (its 2023 is an abridged version, not a new estimate) said: “India still has the highest number of poor children in the world (97 million, or 21.8 percent of children ages 0–17 in India).”

· Free and/or subsidised ration can avert hunger – not poverty. Poverty gets reduced only by raising income. So, using PDS supply underestimates poverty.

The World Bank report is also questionable: It uses the MPCE of 2011-12 as a base and then uses private sector CMIE’s CPHS data to estimate poverty in FY20, even while admitting that it is not comparable with the NSSO’s MPCE (and hence, tweaks it). The CPHS data is known to underestimate poverty vis-à-vis other surveys and is also known for its bias for the well-off (urban population) – which the CMIE admitted and promised to look into in 2021.

How reliable then are the IMF and WB estimates on poverty?

But why do the IMF and WB estimate poverty statistically? The answer is the same: data vacuum.

The lesson in all this?

Indian government must collect relevant data – to rebase the poverty line and update MPCE. Better if it starts collecting (household) income data – as the US, for example, does periodically. The MPCE is for India in which the economic structure didn’t allow collecting income data directly and hence, MPCE was used as a proxy.