DEVIL IS IN THE DETAILS
CURRENT SURVEYS PROVIDE A BROAD EMPLOYMENT SCENARIO BUT MISS OUT ON INSIGHTS REQUIRED FOR INCLUSIVE POLICYMAKING.
Current surveys provide a broad employment scenario but miss out on insights required for inclusive policymaking.
For the longest time, we simply have not had a method for frequent collection of jobs data. The main source of data is the National Sample Survey (NSS), which is collected every five years. The NSS used to have an annual round, which represented the country as a whole but could not break the data down to states as the sample size was small (40,000). Regarding jobs, there is more interest in state-wise data than the national statistics. That is why the data was not used very much, and my mid2000 annual survey was stopped. Since then, we only have five-year jobs data.
Along with that, the Labour Bureau used to collect data essentially on wages. But after the global financial crisis in 2009, the bureau was told to start a survey on job losses. So, it selected nine industries (across labourintensive sectors) to measure job losses. As it started collecting jobs data, the Labour Bureau stopped collating wage data in 2009.
Then the bureau started an annual survey like the NSS survey sometimes in 2011 to measure job creation. The first couple of annual surveys were not good but then it was revised, and now the survey is good enough.
So, once the periodic labour force survey by the National Sample Survey Office (NSSO), which gives annual statistics of employment state-wise (urban and rural) and quarterly estimates for urban areas, stabilises, the Labour Bureau would stop its annual labour survey and go back to wage survey. Wage data is important because it reflects the bargaining power of the labour force, and in a way, indicates the employment situation.
Besides government data, the Centre for Monitoring Indian Economy (CMIE) is the only private agency that comes out with labour data. The CMIE survey is not designed for measuring employment but it is a household survey, and its main focus is on consumption. Because it is doing a household survey, it is getting employment data as well.
The problem with the CMIE data set is that it is surveying the same households and that kind of sample becomes non-representative after a couple of years. CMIE has undertaken the survey for four to five years. So, by now, the sample has become non-representative.
Variations and Shortcomings There would be variations in the results of all the surveys mentioned earlier. There would be variations regarding samples, but most importantly, there would be variations regarding concepts.
The Labour Bureau has now started following the same concept as the NSS so that the results are comparable. CMIE, however, is somewhat different.
You can measure employment in different ways. The simplest way is to ask if one has been employed most of the time during the year. That gives you an idea of the larger employment picture, but it does not tell you if people are working full time.
There are more sensitive measures, which are actually better. One measure is that you use weekly data. You take previous week’s data and break it into 14 halfdays and ask the person if he/she was employed during those periods.
There are many shortcomings of a survey data. The first, of course, is that the sample has to be representative. For a sample to be representative, you must have some way of knowing the full population. The only place where we get that kind of data is the census. In fact, almost all these surveys are based on the last census data. Now, as we start moving away from the census, the quality of sampling drops. But there is nothing you can do about it; you cannot have a census every year.
These surveys are based on responses from individuals, and when you ask them questions, there could be what you call recall error. When you ask anyone if he/she was employed for 183 days, the person may not remem-
ber the exact number of days. So, the shorter the period, the more accurate the responses could be. But the problem with shorter time span is that you do not capture the seasonality.
The third problem is how people interpret questions. It is most common among the women surveys. Women, who consider themselves as homemakers, could be carrying out a lot of economic activities such as being involved in animal husbandry or doing embroidery. But when you ask them if they have been working, they would say no because they think these activities are part of their household work.
Current Status We do not have the data to determine what is the correct employment scenario today. The only data we have is the Labour Bureau and the CMIE data. The two do not match, but both are showing negative growth. So, to that extent, at
Almost all surveys are based on the last census data. As we start moving away from the census, the quality of sampling drops. But you cannot have a census every year
least the direction seems to be consistent.
Another problem with the jobs data is that only knowing the overall picture does not serve the entire purpose. The overall picture hides the differences, which can be sectoral and geographical.
I can easily have a situation where there is a massive loss of jobs in one state and a gain in another state. If I have an overarching national level policy, it may not address the problem because the problem can be a localised one. So, you do need a certain degree of granularity in data to be able to frame policies correctly.
I think the periodic labour force survey that the NSSO is doing should be able to solve the problem to a substantial extent because it is an all-India exercise, and gives you state-level and occupation-wise estimates.