Modi 2.0 presents a new window of opportunity to usher in some fundamental reforms for the Indian economy. A modern dynamic economy requires a robust statistical system to provide precise and real time estimates of several critical indicators. One of these is unemployment — which has been at the heart of prolonged acrimonious public debate in India for several years. Now is the time to move beyond the politics of unemployment to the real and pressing issue of measurement of unemployment. In 2016-17, the government of India, quite rightly so, was of the view that the existing frequency of measuring key unemployment and employment indicators was not adequate for the design of economic policies. There was a need for more timely and periodic measures — a quarterly estimate in urban areas and an annual estimate in rural areas — the “Periodic Labour Force Survey (PLFS)”. A committee of senior bureaucrats, economists and business leaders was constituted to look into this and an all-India PLFS survey was undertaken.
The results of this survey have not been officially released. However, the leaked reports of an historically high unemployment rate and the subsequent resignation of two members of the National Statistical Commission (NSC), who were involved with the PLFS, created a furore and heightened the politicisation of unemployment. The Opposition used this as an opportunity to malign the government, while the government representatives at NITI Aayog resorted to the view that the survey results have not been reviewed by experts, and therefore the report was not deemed reliable enough to be released. The truth of the matter, however, is that there is neither credible evidence of a job crisis in India, nor credible evidence of the absence of it. The problem requires a serious effort by the government to address issues of measurement. In this essay, I will outline what the government needs to do in this regard.
First, measurement of economic indicators, for example the unemployment rate, is an apolitical issue that requires statistical expertise of the highest standards. Before the release of any figure, it is imperative to discuss, debate and deliberate the methodological issues around the measurement. For example, to measure the unemployment rate, it is practically impossible to conduct a periodic census of all citizens above 15 years. Therefore, we have to rely on the second-best option of conducting sample surveys, and the natural question is then about the size of the sample survey. The sample size critically depends on the question of interest, for example, the sample size in the US is determined “on the requirement that a difference of 0.2 percentage points in the unemployment rate for two consecutive months be statistically significant at the 0.10 level”. If the objective of the PLFS was to measure changes in unemployment from quarter-to-quarter in urban areas and year-to-year in rural areas, then we need to discuss sampling precisely to gauge these estimates. Therefore, there can be no credible discussion on changes in unemployment from one period to another in the absence of a paper that outlines in detail the underlying sampling methodology. Proprietary databases like the CMIE, which claim to estimate these time trends, have marketed themselves with a complete absence of such methodological requirements, and are hence non-credible.
Second, even if the sample size issue is addressed to minimise what statisticians call sampling errors (the sample size might not be large enough to address the question of interest), there are issues relating to non-sampling errors. This includes non-participation in the survey or refusal to answer some questions or misunderstanding of survey questions. Each of these are important issues and needs to be addressed in the methodology. For example, suppose there is a job boom in the economy and the employed overwhelmingly refuse to participate in such surveys or do not answer all questions, then it is possible for the survey to indicate high unemployment. Therefore, non-participation is an important issue and methodological rigour requires for a survey to have transparent strategies to prevent or minimise these errors.
Third, India is a large, complex and diverse economy that is undergoing structural transformation. Hence, we are moving towards precision policy-making which requires local and real time socio-economic indicators. We know that the nature and incidence of unemployment, for example, differs from state to state. This requires local measures of unemployment so that economic policies can be tailored depending on local conditions. For instance, unemployment is a rural phenomenon in several states, while in others it is concentrated in urban areas. The state governments will have to participate along with the central government to have comparable uniform measures of periodic unemployment. This requires state governments to conduct periodic surveys on the same lines as the central government. Unfortunately, at present, several state governments do not have the capacity to conduct such regular surveys. This, however, is imperative for a large dynamic economy such as ours. Robust statistical systems will require that we begin to create such local capabilities urgently. It is time to move beyond one-size-fits-all solutions to more inclusive solutions that take into account local conditions.
In the end, one cannot help but recall eminent statistician John W Tukey’s words in 1949: “It is far easier to put out a figure than to accompany it with a wise and reasoned account of its liability to systematic and fluctuating errors. Yet if the figure is to serve as the basis of an important decision, the accompanying account may be more important than the figure itself.”
To enhance India’s statistical capabilities, we have to move beyond the politics of it and focus on measuring with precision. Timely estimates are the cornerstones of policy-making in a modern dynamic and an aspirational economy like ours.
The writer is research director, Brookings India and member, Economic Advisory Council to the Prime Minister