Opinion Think

Shooting in the Dark

To understand India’s jobs crisis, which is more nuanced than politicians make it out to be, we need reliable data. No such luck.

One of the important ongoing debates in the country has been on the topic of job creation. The purported ‘withholding’ of NSSO’s report on Employment for the year 2017-18 has only intensified this debate. However as is the norm these days, the supporters of the Government find no issue with either the Government withholding the report or on the question of whether adequate jobs have created in the last five years. And the detractors of the government find wrong in everything the government does. The reality is more nuanced. Far more fundamental issues over policy are involved than the level of debate in the country currently allows a discussion on.

But first, what should be unequivocally criticised is the withholding of any report by the government – whichever government. Any economic or social or economic or demographic data that is generated as part of routine periodic surveys should be released to the public on an automatic basis. There shouldn’t be any discretion. This withholding of reports or data that perhaps does not show the government in a positive light can in no time transform into a complete mistrust in any data the government releases. And that is a very slippery slope to go down onto.

One way to ensure that this does not happen is to have a pre-announced schedule for all data releases. The same way that there is a pre-announced schedule for all high-frequency data releases – CPI, IIP, GDP, etc. Currently, there is no fixed schedule for the release of NSSO reports – be it on employment or consumption or something else. Once a ‘pre-commitment’ to release a report has been made, it will be hard to not release the said report on that date.

In fact, there is a need to go one step further and to ensure that there should be no prior access to any ‘public’ report or its findings to anybody other than those people directly working with compiling that report. The Government at large should have access to a report/data the same day as the public at large. No preferential access. No information asymmetry between the government and the public at large.

With this, as we get into the debate over jobs and the purported ‘leaked’ report, there are 3 distinct issues with the way we focus on the jobs issue in India – we focus excessively on the absolute number of jobs rather than unemployment, we obsess over quality of jobs which is like chasing a mirage and we focus on the absolute level rather than the trends.

The misplaced focus on jobs rather than unemployment

There is an excessive amount of focus on the outright number of jobs that the economy is creating, and the way this is measured is by the difference in the number of people employed at two points of time. That is an incredibly flawed measure of analysing the state of the labour market. There are two distinct trends playing out in the economy currently. The first and widely appreciated trend is that millions of new people are entering the workforce every year looking for a job. But the second and opposite trend is that as the economy gets richer, the hitherto ‘distress’ employment is starting to wane. So, if a 16-year old child had to work because the family needs to feed itself, and now, he does not have to work because the family can feed itself without his income, and if he can now go to a school or a college, that is a reduction in employment. But this reduction in employment is a positive trend.

There is a fair amount of distress employment in the country. And this is indeed what happened between 2004-05 and 2011-12 as per the NSSO’s reports. There was a sharp fall in labour force participation in children in the age group 15-19 and 19-24, and an increase in the proportion of children attending educational institutions. So just focussing on a headline number of people employed or the number of jobs created can give us a very misleading picture of the state of the labour,

Focus on quality of jobs is useless beyond a point

Ask any person whether he is happy with the job he has or with the pay he gets and most people (I suspect) would want a job that is less stressful, less demanding, less hectic, involves less effort or which pays more. This is standard human behaviour. My point is that focussing beyond a point on whether people are getting a high-quality job is pointless. For what is a high-quality job? Is a job in government a high-quality job? Or is a job in a fast-evolving start-up, which could shut down in a few weeks or months, a high-quality job? Is the job of an Uber driver a high-quality job? If a high-quality job is one which pays more relative to the amount of physical or mental effort involved, then such a job does not exist.

The argument that there should be more ‘salaried’ jobs rather than unorganised or self-employed type jobs presumes that there is optimal, pre-determined, order in the way the labour market must organise itself. There is no basis for that. What the state of equilibrium in the nature of jobs should be is for the market to figure out for itself, once the distortions that stem from the stringent labour laws are done away with. It goes without saying that if labour laws are strict and compliance onerous, people would want to avoid compliance by remaining below the threshold of compliance. So, one way to improve the quality of jobs is for policy to remove the distinction between a part-time employee or a contract employee and an ‘on-payroll’ employee and let businesses decide, on commercial logic, what kind of staff they need.

Focus on trends and not the level

What the unreleased report apparently shows is that unemployment in India in 2007-08 was at a multi-decade high. And hence that is a damming report for the current government. However, firstly, if a data is so out of sync with what has historically been the case, the first question to ask should be whether the data makes sense. Because, prima facie at least, the employment situation does not seem as bad as the headline data from the report suggests it is. As they say in medicine, if clinical findings do not match blood tests, then one must double check the blood tests. However, questions over the reliability of the data should not result in its withholding. Indeed, it is even more important to put it out in the public domain so that analysts can pore through the data and report possible areas of concerns.

However, assuming the data to be correct, even the high absolute unemployment rate does not tell us how bad things really are because we only have data points this decade – 2011-12 and this report for 2017-18. We do not know for instance whether things were worse in 2016-17 and have gotten better or whether things are getting worse. We do not know how good or bad things were in 2012-13 or 2013-14, the years that saw very weak economic growth and significant macro-economic stress in the form of high inflation, currency depreciation etc. The policy response will be very different depending on whether things are improving or getting worse! This practice of doing surveys every 5 years means that we have snapshots at different points of time but not enough data to build a complete picture. And thus the resultant data is largely unhelpful as an input for policy formulation.

A related issue is about how we measure unemployment. The fact that at 6.1%, the unemployment rate is the highest in the surveys conducted in more than last 4 decades suggests a problem with how we measure unemployment. A lot of economists consider 4% to be ‘full-employment level’ and as per NSSO’s reports, we were at this level at several points of time in the last 4 decades. That is rather odd, no? And this is a conceptual problem rather than a statistical one.

The standard metric of unemployment that the NSSO uses is something called ‘Usual status adjusted for subsidiary status’. According to this, a person is deemed to be unemployed if he didn’t pursue any activity during the 365 days prior to the survey for a period longer than 30 days. So if a person had some work for 30 days in a year or more then he is not considered as unemployed under the standard unemployment definition followed by the NSSO. That’s a remarkably shallow definition of unemployment that we have been following for decades. Sure, the NSSO does worker classification at weekly and even daily basis, but using a worker’s employment status for 4 random weeks in a year and using that to generalise the state of the labour market for the entire year seems pretty arbitrary and inadequate. The point is that the way we measure unemployment itself is unhelpful.

To mitigate some of these concerns over data, the government has been releasing data on PF accounts as a proxy for employment growth – that data is both timely and high frequency. But also suffers from significant limitations making it practically useless. Private sector agency CMIE also publishes monthly job market data but even that data has credibility issues.

The EPFO data is incomplete at best, misleading at worst

The Government has started to release data on new PF accounts as a proxy for a formal payroll data. That data suggests impressive job creation – in the last three months number of net new PF accounts with EPFO has grown over 100% YoY. However, this data suffers from two major flaws:

  • First, this data only covers the formal sector of the economy which covers only a small part of the labour force. The EPFO for example as just over 6 crore members as against a total labour force of over 50 crores. The formal sector labour force is thus just 10-15% of the total labour force in India and thus the EPFO data gives us a picture into the state of a very small part of the labour market.
  • Secondly, there are monetary incentives (tax and non-tax) for new hires some of which expire at the end of the current financial year. A precondition for availing these incentives is that new hires ought to have a retirement account with EPFO or the equivalent. It is quite likely that the formal sector payroll data released by the EPFO is being distorted (read: overstated) due to this.

The bottom line is that given the above limitations and the fact that the EPFO data has just 15-months history, this data is not useful to gauge the state of the labour market.

CMIE data raises serious questions as well

Private company CMIE also provides data on the state of the labour market. However, their data also raises credibility questions. For instance, as per CMIE, unemployment fell in the immediate aftermath of demonetisation because things became so bad that millions of people stopped looking for a job (effectively they gave up any hope of finding any job) and thus dropped out of the workforce. And this conclusion does not correspond with other high-frequency data (including corporate performance, which cannot be ‘managed’) around that time, each of which suggests that the economic activity saw a sharp fall in the immediate aftermath of demonetisation, but activity started recovery within 3-4 months thereafter. And in any case, even if we accept CMIE’s explanation, their own data suggests that unemployment in FY18 was 300bps lower than in FY17 – so the labour market improved dramatically in FY18.

A final point to note is that the lack of credible labour market data will cascade to other statistics. India follows a path whereby the core statistics (National Accounts, Inflation, IIP) are revised to a new base every 5-7 years. Thus, the past ‘base year’ for National Accounts were 1993-94, 1999-2000, 2004-05 and the current 2011-12. This is as well because of the rapid changes that are underway in a fast-growing economy like us that ‘base-year’ which establishes baseline data gets outdated quickly. And each of those years is associated with a major labour market survey and it is a ‘necessary condition’ for this.

The 2011-12 base year National Accounts data is discredited due to multiple issues, and if the labour market survey for 2017-18 gets discredited also, either due to fundamental issues with the data or due to political factors, it means we will be stuck with a discredited data on National Accounts for another few years by which it will become even less representative of the economy. And this will mean that other statistics which are linked to the same base year also become outdated till a new acceptable labour survey is done. Thus, the 2017-18 survey getting discredited will cascade to other statistics.

The bottom line is this: The objective of collecting an important data set such as on employment should be that it can feedback to policy. However, collecting such data once in five years, and with such vague definitions, means that the date is of very little use for policymaking. Its only use is for scoring political points, which is what is happening currently. Data should not just be used to explain what has happened in some academic dissertation but it should be capable of being used for policy in a proactive manner. Sadly, successive governments in India have given very little to generating and using data as a tool for policy making. We continue to shoot, mostly,  in the dark.

About the author

Ashutosh Datar

Ashutosh Datar was the Economist at Institutional Equities research team of IIFL from 2007 till a few months ago. He is now an independent blogger.