Poor Economics: Has India’s poverty really fallen?

0
16

By Santosh Mehrotra & Jajati Parida

Bhalla, Bhasin and Virmani in a working paper (IMF), claimed India’s poverty, per a $1.9 per person per day poverty line (at PPP), was 0.9% of the population in 2020. Thanks to government transfer of free rations of 5 kg per person month, it fell to 0.8% (from 0.9% in 2019). Roy and de Velt, for the World Bank (WB), claim poverty fell from 22.5% to 10.2% between 2011 and 2019. Remarkably, both papers contradict the WB’s official publications, which state that, as a result of Covid, poverty will increase by 88-115 million in 2020, and the largest contributor to this would be India.

Naturally, the media in India has lapped up these conclusions, and purveyed them widely, to the delight of the government. Therefore, the papers deserve a thorough examination, both for their methodology and findings.

First, Bhalla et al use National Accounts Statistics (NAS) to estimate private final consumption expenditure, then used to estimate poverty. Practically no one in the world uses NAS estimates of consumption. It overestimates consumption. That has never prevented Bhalla from using it.

Second, they don’t use household surveys of consumption expenditure (CES), widely used for estimating poverty (including by the WB, relying upon Living Standard Measurement Surveys). Historically, in India, poverty estimation has used the CES. The last one, in 2017-18, followed demonetisation and a hurried GST, which adversely affected the unorganised sector that employs 85% of non-farm workers. That CES showed contraction of rural consumption by 8% and a mere 2% rise in urban, compared to 2011-12.

Third, they ignore data on inequality, rising unemployment, falling employment rates, falling wage rates and rising food inflation. Inequality had fallen between 2012 and 2019, primarily because all consumption was getting compressed and trending downwards. However, the recent PRICE, PEW and Oxfam inequality reports showed income inequality increased during pandemic, as the income of 84% households fell. While unemployment rate increased from 2.2% to 5.8% between 2012 and 2019, the youth unemployment rate increased from 6% to 17%. The share of working-age population with work declined from 38.6% to 35.3%, and, for youth from 42% to 31.5%. The falling wage rates since 2012 are consistent with falling GDP growth rates, especially since 2016. Moreover, a food inflation rate of 31% between July 2020 and July 2021 (RBI Bulletin August 2021), would wipe out the impact of the 5kg of free grain; add to that the fuel inflation, thanks to wilful taxes.Roy and de Velt (2022) find a reduction in poverty between 2011 and 2019, based on CMIE’s Consumer Pyramids Household Survey (CPHS). Their results could be misleading for several reasons. Although CPHS collects detailed expenditure information on about 115 items of household consumption, on a longitudinal basis (covering 174,000 households from 28 states), it suffers from criticisms.

The CPHS adopts a measure of consumption that is not comparable to that of the NSS, due to differences in survey instruments. Furthermore, Dreze and others have questioned the representativeness of the survey compared to the NSS surveys, due to differences in sample design and geographical coverage. These differences will have important impacts on poverty estimates for India.

On sample design, absence of a second-stage stratification puts a question on representation of households from both ends of the income distribution. By contrast, in the NSS, representation of urban households from the 1st to 6th deciles of the distribution is embedded into the sampling design. Homeless people or families living in construction sites are excluded in CPHS survey. This also contributes to under-coverage of the poorest households in CPHS. Then, the differences in instruments. The NSSO uses a more detailed consumption module comprising over 345 items, compared to CPHS’s 114 unique items . Additionally, the NSS expenditure based on uniform recall period captures household consumption over the past 30 days, whereas the CPHS collects consumption based on the past four calendar months. Differences in recall periods across surveys can have large impacts on poverty estimates. The shares of households with access to electricity, water, toilet and ownership of a TV and refrigerator are notably higher in CPHS-2015 and -2019 compared to NFHS from the same years.The undereducated are severely under-represented in the CPHS, with only 2% of the 2018 adult population having not received a formal education. By comparison, the NSSO’s Periodic Labor Force Survey (PLFS) from the same year pegs the share of adults without formal education at 17%. By 2019, adults without formal education are virtually eliminated from the CPHS sample, while the PLFS estimates this at 17%. Basole et al (2021) find that average real incomes in the CPHS of 2018 are about 30% higher when compared to the PLFS from the same year.

Moreover, as Roy notes, the Gini coefficient of inequality using CPHS weights would rank urban India on a par with Sweden, the 25th most equitable country. However, NSS-2011 would rank urban India close to the 60th most unequal.

Thus, comparing expenditure and non-expenditure statistics derived from the CPHS to those obtained from nationally representative benchmark surveys shows that: (1) CPHS under-represents the poorest as well as the richest households in the population; and (2) under-coverage of the poor and the rich is more pronounced in urban areas. Roy adjusts for these differences, but strong differences remain between consumption in estimates based on CPHS and NSSO. Roy then uses regressions to arrive at poverty estimates for India. However, Roy’s pass-through methodology based estimation of poverty is subject to strong assumptions. Hence, the estimates based on this are questionable. Moreover, the regression based estimation of the Log consumption per capita also suffers from problems of endogeneity (due to use of “consumption categories/heads” and “level of education” as the explanatory variables) and excluded regressors (family wage income, number of earning members in the family, and land holdings, etc). These two issues normally produce biased and inconsistent regressor estimates. Hence, the prediction based on this model may not be reliable.

Meanwhile, based on the Tendulkar’s extended poverty line (from 2011-12 to 2019-20) taking CPI inflation into account and after due adjustments between CES and NSSO’s Employment Survey (2011-12) EUS discrepancies, we found that the incidence of poverty between 2011-12 and 2019-20 decreased only about 1 percentage point (from 21.9% to 20.8%). While the incidence of rural poverty declined from 25.7% to 25.2%, in urban India, it declined from 13.7% to 12.4%. But the number of poor actually increased between 2011-12 and 2019-20 due to increased total population. The increase in number of poor in rural areas (from 217 to 235 million) is much higher than the 3 million fall (from 53 to 50 million) in urban India.

The author Santosh Mehrotra is visiting professor, University of Bath, the UK, and Jajati Parida teaches at the Central University of Punjab