PLFS data cannot be a measure of widespread hunger and the rise in poverty

Poverty is estimated on the basis of consumption and not income in part because consumption is easier to measure and smoother than income. (Representative image)

A recent study by Arvind Panagariya and Vishal More has criticised our previous work and claimed that extreme poverty fell in India during the pandemic. The claim is not tenable and makes light of the substantial increase in hardship, food insecurity and indebtedness. Here we clear some misunderstandings and point to data as well as analysis problems which give a misleading picture of the welfare losses endured by the poor.

The authors’ analysis is based entirely on Periodic Labour Force Survey (PLFS) data. As they are aware, this survey is designed to measure employment not consumption. The 2020-21 PLFS report explicitly cautions that “Information on household usual monthly consumer expenditure (UMPCE) was collected in PLFS only to classify the households in different UMPCE classes and it cannot be used to estimate the household consumer expenditure which is generally estimated based on detailed survey." (page 7, emphasis added). Clearly, the same statisticians whose virtues are extolled by the authors when criticising CMIE surveys are cautioning us not to make the measurements that the authors make. The authors do not discuss this caveat and its implications at all.

An added complication is that the change in method in 2020-21 (from one question on household consumption to five separate questions) makes comparisons over time difficult. As is well-known, more detailed questions tend to estimate consumption better and move its reported level upwards. This means that poverty may appear to be lower in 2020-21 for purely statistical reasons.

The 2020 Lockdown

In our work, we used a range of data sources including large datasets such as the Consumer Pyramids Household Survey (CPHS) of the CMIE, as well as our own surveys and many other surveys carried out all over the country. It is a failure of the public statistical system that official consumption data was not available in the time frame needed to estimate the welfare impact of the pandemic.

Based on income (not consumption) data from CPHS we concluded that poverty rates rose sharply during the 2020 lockdown. Usually, poverty is estimated on the basis of consumption and not income in part because consumption is easier to measure and smoother than income. This approach works well in normal times when both income and consumption tend to rise. But the pandemic was a very unusual time. Both incomes and consumption fell. With falling incomes, poor households are forced to maintain a basic level of consumption by drawing down savings or going into debt. Indeed, many reports during the pandemic pointed to a large rise in indebtedness. This aspect of welfare loss is completely missed when we focus only on consumption poverty.

Significant here is the fact that, as the authors report, incomes fall more and stay depressed longer than consumption. Self-employment earnings on the eve of the lockdown (January-March 2020) were 16 percent above the first PLFS quarter (July-September 2017). Earnings collapsed in the lockdown quarter, back to their 2017 levels and had not recovered to the pre-lockdown level as of April-June 2021 (the last quarter for which they report data). We have verified that earnings had not recovered even as of April-June 2022. This is a full two years of depressed earnings, on an already low base.

Poverty Headcount Rate

Another problem is the use of just one single poverty line (Tendulkar) to estimate the poverty rate. In our work, we calculated headcount ratios for many thresholds including Tendulkar, World Bank $1.90 a day and the Anoop Satpathy Committee National Minimum Wage. We did this both because the extreme poverty lines such as Tendulkar or the World Bank line are abysmally low and also because it is well known that a small increase in the line makes a large difference to the poverty rate since a significant proportion of households lie just above the line.

For example, moving the income poverty line up from the Tendulkar (Rs 1,240 rural, Rs 1,482 urban per capita per month in January 2020 rupees) to Anoop Satpathy (Rs 2,900 rural, Rs 3,344 urban per capita per month in January 2020 rupees), raised the pre-COVID poverty headcount rate from less than one percent to more than 32 percent in the rural areas and from less than 0.5 percent to close to 18 percent in urban areas. Note that the event study approach which the authors refer to has not been used to make the above claim. We showed this using simple headcount ratios, as the authors too have done.

Finally, we come to government transfers. There is no doubt that policy support prevented widespread starvation. To what extent this support helped to sustain consumption in face of falling incomes is an important question. But the PLFS is not designed to answer this question. The authors simply assume that the support must have been enough to keep consumption at the level observed. At the very least, they could have observed that in-kind transfers amount to only around Rs 300-400 per month in this data. Further, the inadequate reach of cash transfers as compared to PDS has already been well-reported. It is thus likely that government transfers were far from adequate to bridge the significant gap between consumption and income that developed during the pandemic.

In sum, it is worrying that high levels of suffering and hardship indicated by pre-existing low levels of both consumption and income, the pandemic-induced decline in income (seen across surveys), widespread hunger and rise in indebtedness can be summarily dismissed on the basis of an inappropriate data source and a narrow measure of welfare. One hopes for a much more sensitive and careful debate on this serious matter. Do the people of India deserve any less?

- Akshit Arora provided research assistance.

Amit Basole is Professor of Economics, Azim Premji University and Mrinalini Jha is Assistant Professor of Economics, OP Jindal Global University. Views are personal, and do not represent the stand of this publication.