Why India needs a 2024 poverty line

The "free" ration entitlement ("@5 kg food grain per person per month") is for "67% of the population," as per the National Food Security Act (NSFA) of 2013. Going by the population estimate of 1,392.3 million in 2023 (Lok Sabha answer of July 25, 2023), 923.9 million Indians (or 67%) should get the "free" ration. But exclusion and inclusion and non-updating errors have denied 119.4 million Indians this benefit. The exclusion is so huge that on March 19, 2024, the Supreme Court directed the Centre and states to issue ration cards to 80 million migrant workers, registered on the eShram portal but excluded from the NFSA within two months – which the Centre had promised in its affidavit of April 26, 2020 (four years ago).

Going by this fact alone, should India's poverty estimate be 67% (923.9 million)? The basis of the claim that poverty fell to "single digit," "closer to 5% or less" and "4.5-5%", respectively, in 2022-23 is the household consumption survey (HCES or MPCE) of 2022-23. But what does this data actually tell?

Reading HCES 2022-23

The HCES of 2022-23 shows the average monthly household expenditure (with imputation, at 2011-12 prices) is ₹2,054 in rural and ₹3,544 in urban areas. That is, per capita per day living expenditure of ₹68.5 in rural ($0.8 at exchange rate of ₹83) and ₹118.1($1.4) in urban areas.

Since India's own poverty line estimate is two decades old, called the Tendulkar's poverty line of 2004-05, estimating poverty on the basis of this data is a wild-goose-chase. That this is grossly underestimates poverty can be gleaned from the following facts (more of it later):

(i) Harvard study, published in the JAMA Network Open in February 2024, says India "accounted for almost half of the zero-food children" among 92 low- and medium-income countries during 2010-2022. It found India had 19.3% (6.7 million) children of 6-23 months who "did not consume animal milk, formula, or solid or semisolid food during the last 24 hours." India refuted this, saying that the study excluded breast-milk feeding. That is right, but the study cited the WHO to say why it did so: "After age 6 months, breast milk alone is insufficient to meet the growing nutritional requirements of infants and young children."

(ii) India's own (latest) national health survey, NFHS-5 of 2019-2021, says, Indian children below 5 years had stunting rate of 35.5%, wasting rate of 19.3% and underweight of 32.1% – due to high-level of malnutrition and unhealthy living conditions.

How poverty fell below 5%?

NITI Aayog CEO BVR Subramanyam said his estimates were based on the fall in household consumption of food items below that of non-food items for the first time since 1999-2000 – a sign of growing prosperity. He used this data to claim that the impact of inflation might have been overstated. Household expenditure (with imputation) on food indeed fell from 52.9% to 47.5% in rural areas and from 43% to 39% in urban areas during 2011-12 and 2022-23 (a decade).

The SBI Research report was more elaborate. It took into account: (a) 2011-12 poverty line estimated at ₹816 for rural and ₹1,000 for urban areas; (b) adjusts it for "decadal inflation" factor (1.89 for rural and 1.85 for urban); and (c) "imputation factor" (1.05 for rural and 1.04 for urban) to arrive at a new poverty line for 2022-23 at ₹1,622 per month (₹54 per day) for rural and ₹1,929 per month (₹64 per day) for urban areas. It then estimated poverty at 7.2% in rural and 4.6% in urban areas – for which it used; (d) two poverty line estimates as the base – (i) Tendulkar's 2011-12 poverty line and poverty estimates (25.7% poverty in rural and 13.7% in urban areas) and World Bank's 2018-19 poverty estimates (11.6% and 6.3%, respectively).

Making sense of HCES and poverty reduction

A few caveats are needed to bring clarity and perspective to the above arguments/data.

Caveat 1: HCES of 2022-23 does not estimate poverty line; it provides household expenditure. When poverty line is available, HCES data can be used to estimate poverty percentage or number of poor. The poverty line was estimated in 2004-05 and 2011-12 – when the earlier HCES were carried out. But not in 2022-23. The claim of fall in poverty is being measured from the 2011-12 poverty line – which itself is an updated version of the 2004-05 poverty line (Tendulkar's), not a new one.

Caveat 2: HCES of 2022-23 comes a decade after that of 2011-12 and it lists four "changes" in methodology to caution against "comparability" with that of 2011-12: (i) expansion of items covered from 347 to 405 through inclusion and merger of items; (ii) "three separate questionnaires" were used for food, consumables and durable goods (against a single questionnaire earlier) which were asked, (iii) in "three separate monthly visits" (as against single visit). Anyone familiar with surveys will vouch that all the three changes will change the responses (outcomes). The 2017-18 HCES, which showed a fall in 'real' consumption expenditure, was junked on the claim that an expert committee had objected to "data quality," which turned out be false. On the contrary, the said committee had validated it, stating that it was in line with other relevant data.

Caveat 3: The HCES 2022-23 data is partial – only the value data has been released, not the volume data. Hence, the role of inflation can’t be accurately measured. SBI Research uses only value data, while the Tendulkar’s used both value and volume data – for 2004-05. The 2011-12 poverty line was an extension of the Tendulkar’s and was done by the Planning Commission as interim one which showed overall poverty at 21.9% even as it waited for the Rangarajan's estimates because the Tendulkar's had come under widespread criticism for being "too low".

Caveat 4: Rangarajan’s poverty line came in August 2014, fixing it for 2011-12 – at ₹972 for rural and ₹1,402 for urban areas – and estimated poverty at 29.5% in 2011-12. This was far higher than the Planning Commission's 21.9% for the same year. But by then the government had changed and the new government didn't accept it. Nor has the government re-estimated it after 2014.

Caveat 5: SBI Research apparently uses (not disclosed) the World Bank's outdated poverty line of $1.9 (at 2011 PPP) – which was changed to $2.15 (at 2017 PPP) in 2019 to account for inflation. It quoted poverty at 11.6% in rural and 6.3% in urban areas for 2018-19. But the World Bank database shows, India's overall poverty (at $2.15, 2017 PPP) was 12.7% in 2019, which fell to 11.9% in 2021.

The World Bank’s 2022 report says, India saw "an increase of 56 million poor people" in 2020 (at $2.15) due to the pandemic. Pew Research Institute's March 2021 report (before the second wave) said the number of Indian poor ($2) increased by 75 million, its middle class shrinking by 32 million and India contributing "60% of the global retreat in the number of people in the middle-income tier" in 2020. But India never acknowledged that poverty increased due to the pandemic or due to pre-pandemic economic shocks of demonetisation of 2016 and GST of 2017 (more of it later).

Poverty estimates fail smell test

Economist Ashoka Mody analysed the HCES 2022-23 data and other indicators to write ("India's Poor Will Not Be Wished Away") that it failed the "smell test"; "in reality, poverty remains deeply entrenched in India and appears to have increased significantly." He flagged "the lack of comprehensive consumption – and inflation – data," making "it impossible to get an accurate picture." While volume data for HCES of 2022-23 is not released (hence, incomplete), about inflation he wrote, "regrettably, Indian authorities do not provide inflation data segmented by household income."

What is his own poverty estimate?

He wrote: "Adjusting the Rangarajan and Subramanian estimates using plausible current inflation rates leads to a stark conclusion: Urban poverty rates range between 40% and 60%, which means that 30-40% of all Indians are poor." Subramanian's reference here is to poverty expert S. Subramanian’s 2012 estimates.

The gaps in HCES and other official data sets are too glaring to miss.

Here are two examples. The PLFS data from 2017-18 to 2022-23 show a surge in informal and low-paying agricultural jobs to 45.8% (sectoral distribution); a surge in self-employment to 57.3% due to a fall in quality jobs and unpaid jobs to 18.3% (category-wise distribution). Wages growth is negative for regular wages/salaried and self-employment in these five years. There is also a surge in MGNREGS's menial, below statutory minimum wages jobs, engaging (average) 64 million rural households and 93 million individuals during FY19-FY24 (up to February 2024). They point to a surge in poverty, not reduction.

More flawed poverty estimates in 2022 and 2023

The Aayog first claimed a dramatic fall in poverty in 25 states/UTs during the pandemic 2020-21 – without explaining why or how – after its surge in the pre-pandemic in 22 states/UTs in 2019-20 – in its SDG reports!

More recently, on July 17, 2023, the Aayog claimed "135 million have exited multi-dimensional poverty" (MPI) during 2015-16 and 2019-21. MPI is calculated on the basis of three parameters with equal weightage – health, education and living standard (income).

But this estimate was based only on data source, health survey NHFS-5 of 2019-21. No other data source was used for education and living standard (income) – thereby excluding two-third measures of MPI. The NFHS-5 data is also questionable for using the pre-pandemic 2019 survey (70% data) and the unusual year of the 2021 pandemic (the rest 30% data), missing 2020 completely, and averaging 2019 data with that of 2021 – an unheard economic data collection. Its baseline index was “based on” the old and outdated NFHS-4 of 2015-16 when the NFHS-5 data for 2019-20 was already available.

Then, on January 15, 2024, the Aayog released another MPI report claiming that "24.82 crore" Indians (248.2 million) escaped MPI poverty during 2013-14 and 2022-23 – with MPI poor continuing "to decline from 29.17% in 2013-14 to 11.28% in 2022-23." There was no additional survey or data to warrant this estimate. The HCES of 2022-23 data will be released in February 2024.

The Aayog’s January 2024 estimate is purely statistical and it said so (“owing to lack of data for the years between 2005-06 and 2015-16 and after 2019-21”). The report says: “This requires interpolation of estimates for the year 2013-14 and extrapolation for the year 2022-23. Thus, 2013-14 figure was computed from the 2005-06 estimate using annual compound rate of 7.69% reduction. Similarly, 2022-23 estimate was projected by applying compound rate of reduction achieved during 2015-16 to 2019-21.”  The references to 2005-06 and 2019-21 are to the NFHS-4 and NFHS-5 data, respectively.

Note Aayog claimed decline in poverty in just 40 days – from 11.28% on January 15, 2024 to “closer to 5% or less” on February 26, 2024. True, MPI poverty and simple poverty are different but that nuance is lost in the splashed headline numbers of 11.28% and 5%.

How to estimate poverty better?

There are certainly better ways to estimate poverty than base it on HCES.

Firstly, the HCES data are known to be widely off-the mark, these being self-reported or voluntary disclosures. Unlike in India, household surveys are widely studied subject in the US and non-response and/or denial of benefits (imputation value) by both the poor and the rich households (2015, 2017, and 2021) are well known. In India, the poor has incentives to under-report or lie about government support (imputation) to get more of it; the rich to evade exclusion and tax liability. Hence, as many have argued, India must address such issues.

Secondly, it is time to shift to collecting income data. Today’s India is very different from the early decades of independence when getting income data was difficult. Like the US, India can now collect such data. Such data can be supplemented with income tax returns (ITRs), digital transactions, property tax, EPFO, PLFS, MGNREGS data etc. to get a better picture of income levels and hence, poverty.

The starting point, however, should be fixing a poverty line relevant to 2024 – not that of 2004-05.