Two joblessness narratives in India from two large-scale surveys5 min read . Updated: 23 Jun 2020, 07:38 AM IST
A Mint analysis suggests that the CMIE figures should not be considered a proxy for the official unemployment and labour force participation rates
As the country went under lockdown, and everything, including the collection of statistics, came to a halt, analysts and economists were left scrambling for data to seize up just how badly the economy was doing. One dataset that has come to the rescue is the nationally representative employment-unemployment database of the Centre for Monitoring Indian Economy (CMIE).
The high-frequency data on unemployment published by CMIE have been widely used to gauge the state of joblessness, and as a proxy for the state of economic activity across the country during the lockdown period.
But how far does the unemployment rate captured by CMIE match the unemployment rate tracked by India’s official statisticians through the periodic labour force survey (PLFS)? A Mint analysis of the closest comparable indicators from the two datasets shows striking levels of divergence in unemployment and labour force participation rates across states over the period for which data is available from both surveys. The analysis suggests that the CMIE figures should not be considered as a proxy for the official unemployment and labour force participation rates.
While CMIE quarterly figures are available till the last quarter, the quarterly urban employment rates reported by PLFS are available only from the Apr-Jun 2018 to Apr-Jun 2019 quarters. Over this period, the levels and trends (direction of change) in unemployment as reported by the two surveys have been different. The divergence has been particularly marked for female unemployment rates.
The divergence is also marked if we consider the labour force participation rates. The employed and the unemployed (who are willing to work) together make up the labour force. CMIE reports much lower female labour force participation rates and much higher female unemployment rates compared to PLFS.
Such divergences are starker at the state level. States that saw the sharpest rise in unemployment between the June 2018 and June 2019 quarter according to PLFS don’t figure in the list of the states that saw the sharpest rise in unemployment according to CMIE, with the sole exception of Uttarakhand. In fact, two states, Assam and Andhra Pradesh, which saw the sharpest rise in unemployment over this period according to the PLFS figure in the list of states that saw the biggest drop in unemployment according to CMIE. And two other states, Uttar Pradesh and Chhattisgarh which saw sharp declines in unemployment over the same period according to PLFS figure in the list of the states which saw the sharpest rise in unemployment according to CMIE.
Such divergences in the unemployment trends reported by the two surveys also extend to the reported levels of unemployment. For the four quarters between Jul-Sep 2018 to Apr-Jun 2019, the average unemployment rates reported by PLFS and CMIE differ significantly across most states.
For some of these states, the differences are driven by trends in female unemployment rates. But for several states which show a marked divergence in unemployment levels over this period --- such as Haryana, Himachal Pradesh, Assam, Uttarakhand, Odisha, and Telangana --- the reported differences in male unemployment rates are also quite stark.
An analysis of the labour force participation rates also suggests similar patterns, with very little match in either levels of labour force participation rates across states or the direction of change in those levels over time.
The divergence between the two datasets could be driven by the differences in how these surveys are designed, according to independent experts. PLFS focuses exclusively on the job market trends while CMIE’s employment questions are part of a broader set of questions on consumption and investment behaviour. The reference period, even in the quarterly surveys, is not identical. CMIE’s survey has a recall period of a day while PLFS’ has a reference period of a week. CMIE also has a larger urban sample and a different stratification strategy compared to the PLFS.
How questions are worded may also be impacting these results, particularly for female employment and labour force participation, said demographer Sonalde Desai, director at NCAER-National Data Innovation Centre, and professor of sociology at the University of Maryland. Desai’s research shows that when enumerators take care to ascertain various activities performed by different members of the household as opposed to simply asking whether they are employed or not, the estimates for women workers can be strikingly different.
“Women in households that run very small enterprises like a shop are likely to get classified as employed with an easy definition of being employed such as that used by PLFS," said Mahesh Vyas, who heads CMIE. “But, in our surveys, a person needs to be employed for a better part of the day on the day of the survey to be classified as employed. Now, if a person is easily considered to be employed then the labour participation rate is inflated and the unemployment rate is deflated from what I would call a more meaningful definition of employment."
CMIE data would be useful to figure out the changes in unemployment levels but to gauge the absolute levels of unemployment, PLFS may be a better bet, said Pronab Sen, the former chief statistician. CMIE relies on its panel of (fixed) households, and after a while, a panel can become non-representative (of the population), said Sen.
However, it is this feature of having a fixed panel that may have allowed CMIE to continue its surveys in recent months, at a time when official surveys have been hamstruck, first because of fears related to the NPR (National Population Register), and then because of the lockdown.
CMIE enumerators did not face much difficulty in their work because of NPR-related issues, said Vyas, because of their ‘strong relationship’ with the respondents. Wherever there were problems in accessing households, such visits were rescheduled. During the lockdown, house-to-house surveys were promptly converted into telephonic surveys, thereby ensuring that their surveys continued, he added.
What has also helped CMIE’s popularity is the real-time data dissemination. The quarterly PLFS surveys were supposed to be high-frequency datasets that would enable policymakers in real-time decision-making but India’s statistics ministry has managed to convert the publication of even the quarterly surveys into annual affairs. The last quarterly survey report, released earlier this month, pertains to the year-ago period, Apr-Jun 2019. CMIE provides quarterly, monthly, and even weekly updates. It will be aeons before our statistics mandarins beat that kind of timeliness.