The World Bank has been using the standard of $1.90 per person per day (in 2011 Purchasing Power Parity US dollars) to monitor levels of extreme poverty around the world. By this standard, 22% of India’s population was deemed to be in extreme poverty in 2011-12. Though a decade has now passed by, more recent estimates of poverty in India remain embroiled in controversy. The reason is the absence of comparable data from the National Sample Survey (NSS) Organization’s Consumer Expenditure Surveys, which have been the mainstay of poverty measurement in India for decades.
The last publicly available round of the NSS consumption survey was for the year 2011-12. Another round of the survey was conducted for the year 2017-18, but neither the results nor the data from this survey were released with the government citing “data quality issues” with the survey as the reason for its suppression. But leaked tabulations from the survey, which indicated a stagnation in national poverty and an increase in rural poverty since 2011-12, made many an observer wonder whether the decision (made months before the 2019 general elections) was politically motivated.
Be that as it may, the data vacuum has led to a number of attempts to estimate post-2012 poverty levels in India using a variety of imputation methods and/or alternative data sources. Most of these indicate a decline in poverty since 2012 though at a rate slower than that observed for the preceding decade, which seems consistent with the slowdown of India’s growth especially since 2016-17. One exception to this is the recent IMF Working Paper by Bhalla, Bhasin and Virmani (hereafter BBV), which has declared the virtual elimination of extreme poverty in India. To put it in their own words: “By this standard [PPP$1.90 poverty line], India can reasonably claim that in pre-pandemic [sic] India was on the verge of eliminating extreme poverty” (p. 15).
There are other claims in the paper too regarding the fall in inequality and the negligible poverty impact of the pandemic, but let’s focus on the claim of pre-pandemic virtual elimination of extreme poverty. Does it hold up to reasonable scrutiny? To answer this question, we need to unpack the claim.
BBV present two sets of poverty estimates, with or without taking account of the food transfers through the Public Distribution System (PDS). Without food transfers, they claim poverty to have fallen from 21.8% in 2011-12 (not very different to the aforementioned estimate of 22%) to 3.4% by 2019-20; with food transfers, the claimed fall in poverty is from 19.9% to 1.9% (Tables 2 and 5 in BBV). Thus, the claim of virtual elimination of extreme poverty rests mostly on the claimed progress without food transfers; the valuation of the subsidy component of food transfers is only a cherry on top. So, let’s focus on the claimed fall in poverty from 22% in 2011-12 to 3% in 2019-20.
BBV’s methodology of arriving at these numbers is relatively straightforward.
Step 1: Begin with the consumption distribution from the last NSS consumption survey for 2011-12.
Step 2 (and Assumption 1): Assume mean per capita consumption in nominal terms has grown at the same rate as the per capita private final consumption expenditure (PFCE) from the National Accounts (NA).
Step 3 (and Assumption 2): Assume that every household’s per capita consumption has grown at the same rate as the mean per capita consumption to arrive at the “updated” distribution for every subsequent year after 2011-12 (which is the same as the 2011-12 distribution except for the mean).
Step 4: Estimate poverty measures for every year subsequent to 2011-12 using the “updated” distribution from Step 3 and the 2011-12 poverty line updated by the Consumer Price Index (CPI).
Thus, effectively BBV assume distribution-neutral growth since 2011-12 at a rate given by the NA growth in private consumption per person. (BBV also present a version assuming distribution neutrality at the state level.) BBV defend Assumption 1 by arguing that “survey capture [ratio of survey-to-NA consumption] in India has “normalized” to around 50 % in 2004-5, 2007-8 (a small-sample NSS survey), 2009-10 and 2011-12 according to the URP [Uniform Recall Period] method” (p. 8). The trouble is that there is no evidence of this.
Figure 1 shows the survey capture ratio (also referred to as the “pass-through” rate) over nearly sixty years based on the work of Datt, Ravallion and Murgai (2020). It has been falling rapidly since the mid-1980s, and there is no sign of stability since 2004-05 either. Even using the new revised back series of National Accounts, the survey capture ratio was 48, 45 and 46 percent for 2004-05, 2009-10 and 2011-12 respectively. This is hardly the evidence on which to hang the projected constancy of the ratio for the next decade, especially in light of its secular decline over the preceding four decades.
The declining survey capture ratio is of course closely related to the violation of Assumption 2. The survey capture ratio has been declining in good measure precisely because the surveys fail to capture the top tail of the income distribution, while top income shares have been rapidly rising. Estimates by Chancel and Piketty (2019) indicate that missing top incomes account for a large and increasing share of the NA-survey gap in mean incomes, about 45% in 2014-15 (Figure 2).
There is thus little justification for Assumptions 1 and 2.
Give up on these assumptions, and the entire house of cards built up by BBV collapses. Alternative estimates that relax Assumption 1 and use survey capture ratio of 0.67 put the extreme poverty rate for 2017-18 at about 10%. Other estimates that use alternative (reweighted) data from the Consumer Pyramid Household Surveys (CPHS) since 2014, and survey-to-survey imputation methods to infer consumption levels comparable to the NSS surveys, also arrive at a similar estimate of extreme poverty of about 10% for 2019-20. These alternative estimates are not perfect but they cast significant doubts on the claim of elimination of extreme poverty in India.
Way back in 1988, B. S. Minhas had warned against pro-rata adjustment to the survey-based distribution of consumption in the Seventh Plan documents:
"…it is hazardous to carry out pro-rata adjustment in the observed size distribution of consumer expenditure in a particular NSS round by multiplying it with a scalar derived from the ratio between the NAS estimate of aggregate private consumption (for some financial year) and the total household expenditure available from the NSS round. This kind of mindless tinkering with the NSS distribution, as practiced by the Planning Commission in the Seventh Plan documents, does not seem permissible either in theory or in light of known facts."
It is unfortunate that the gaps in India’s statistical system and the lack of comparable data has created an environment where such mindless tinkering has continued in our times.
Gaurav Datt
23 May 2022
Σχόλια