In the midst of public and policy attention on the urgent need to create jobs, Surjit Bhalla and Tirthatanmoy Das (‘Jobs data and its discontents’, IE, January 24) assure us that we need not worry as employment has been growing impressively in recent years. In laying out their argument, Bhalla and Das (B&D) attack a recent CEDA-CMIE publication – “The impact of Covid-19 pandemic on people’s economic lives”. Their main target is CMIE, which is apparently part of an unofficial “cottage industry” that presumably spins yarn. Their article is fraught with untenable claims.
B&D’s first claim is that the CMIE Consumer Pyramids Household Survey (CPHS) does not collect or “report monthly data on employment” and they accuse the CEDA-CMIE publication of using “synthetic” data.
To check if their assertion was valid, they could have referred to the CPHS documentation or to any of the academic papers using this data. Even simpler, a cursory look at the landing page of the CMIE website would have revealed a bulletin based on monthly employment data (following B&D, monthly is italicised).
For the benefit of readers, CPHS is a panel survey, where the same households are interviewed three times in a year. Each four-month period is called a “wave”. Each month, roughly one-fourth of the wave sample is interviewed. Unit-level data is available to subscribers (where households surveyed in each month can be easily identified), and summary statistics using monthly data are publicly available.
The second claim is that since CPHS employment estimates do not match the official Periodic Labour Force Survey (PLFS) estimates, the former is incorrect.
Those who get their hands dirty with data know very well that measurement of employment or labour force participation (LFP) is complex in economies such as India where a large part of the workforce is in informal, irregular work. PLFS data contain several measures of employment. They count if individuals were employed during the majority of time during the 365 days preceding the survey (usual status); if individuals were working for any 30-day stretch in the preceding year; if employed over the past week (current weekly status), or the previous day (current daily status). These measures yield divergent estimates, as there are individuals who fall into one category but not another. Individuals are also engaged in multiple activities, some considered primary and others secondary, generating distinct estimates by “principal” and “subsidiary” employment status.
The CPHS definition of employment and LFP differs from these. Amit Basole and Mrinalini Jha compare PLFS and CPHS comprehensively and point out dimensions in which the two surveys are similar as well as different.
The third claim: As further proof that CPHS presents a false picture, B&D point out that their female LFP estimates are significantly lower than those from PLFS.
Female LFP numbers in India are beset by lack of proper measurement. Women’s contribution to economic work is undercounted in developing countries, India being no exception, where income-generating family enterprises run on the backs of free labour by women. In these activities, men are counted as workers, but women are often not. Thus, when surveyors ask household heads whether women “work”, they deny or under-report women’s contribution to the household’s livelihood activities. Since these women are not actively looking for other paid work, they are not counted as unemployed either. Not classified as either working or looking for work, they are deemed to be out of the labour force.
The India Human Development Survey (IHDS), which by B&D logic would also be a part of the “unofficial cottage industry”, is sensitive to this issue. It captures women’s economic work more accurately, yielding significantly higher estimates than NSS. Notably, “unofficial” IHDS estimates do not show a decline in FLFP between 2005 and 2012, whereas the “official” NSS data do. Women’s unpaid economic work on family enterprises, which IHDS captures, but PLFS does not, has not declined over time.
Following B&D, should we reject IHDS estimates because they do not match PLFS estimates? The mere fact that two survey estimates differ cannot be the basis of concluding that one is correct and the other not.
CMIE CPHS data has the lowest estimate of FLFP because of a stringent definition of employment, not because of sleight of hand.
The fourth claim is that a World Bank study “rejected” CMIE weights. That study attempted to reconstruct CMIE data to match NSS based on several debatable assumptions. For our purpose, what is important is that sampling weights (for any sample survey) are not assigned by an external agency but arise from the sampling strategy. The simplest individual-level weights (in a non-stratified sample) are the inverse of the probability of being selected in the sample. As the sample gets stratified, the calculation of weights becomes more complex.
The fifth claim is that monthly employment cannot decline when annual GDP is growing. First, monthly employment can exhibit volatility even when annual output is growing. Two, output growth could be capital intensive and need not be accompanied by employment growth. There is substantial research analysing low employment elasticity of output growth and jobless growth in India.
B&D end by making a plea for regular and timely availability of government data which has been a longstanding demand of the scientific community. However, if we are interested in the generation of good quality and timely data and evidence, we must enable the creation of a robust, healthy, transparent, varied and multi-faceted data ecosystem. This is beyond the capability of a single agency. In the meanwhile, we need to continue to worry about jobs.
The writer is Professor of Economics, Director, Centre for Economic Data and Analysis (CEDA), Ashoka University
CMIE CPHS data has the lowest estimate of FLFP because of a stringent definition of employment, not because of sleight of hand. (Representational Photo) 
Share
