Weekly estimates of labour statistics during April had suggested that the unemployment rate would rise during the month. Weekly rates had risen to 7.9 percent in the first week of April after having averaged around 6.4 percent through the weeks of March. The first week of April therefore saw a sudden and substantial jump in the unemployment rate to 7.9 percent. Then, it inched up further to 8.1 percent in the second week and continued to rise to 8.4 percent in the third week.
Weekly estimates can be volatile but the trend in these three weeks were a lot more decisive to draw an inference that unemployment was rising. During the fourth week, the unemployment rate dropped suddenly to 6.5 percent. But, this fall was not sufficient to change the fact, the unemployment rate had risen sharply during April 2019.
Weekly estimates are volatile but, they are not unreliable. It may be useful at this point to provide some additional statistics on these estimates to get a sense of how reliable they are. First, these estimates are based on a sample of about 30,000 individuals from randomly selected households across rural and urban India. This is a large sample by any standard. Second, and more importantly, the standard errors of these estimates are extremely small.
All statistical measures carry margins of error. The effort is to minimise these through robust sample selection, sound execution and appropriate estimation methods. While CMIE has done this, it is perhaps, instructive for users to know of the margins of errors in the estimates. We explain these below.
In the first week of April, the unemployment rate was estimated at 7.91 percent with a standard error of 0.006195. This means that there was a 5 percent chance that the true unemployment rate was less than or more than 1.96 times 0.006195 of 7.91 percent i.e. the estimated unemployment rate of 7.91 percent was most probably between 7.90 percent and 7.92 percent.
During the second, third and fourth weeks of April, the unemployment rate and the corresponding standard errors were: 8.06 and 0.005309; 8.38 and 0.007408; and 6.65 and 0.005139. As is evident from these, the standard errors are so low that the estimates are likely to be correct within a probability of 95 percent upto the first place after decimal. We therefore, mostly report the unemployment rate upto the first place after the decimal point.
The average weekly standard error is around 0.0075. In comparison, the monthly standard error is much smaller at around 0.0025. We therefore consider the monthly estimates to be a lot more reliable than the weekly estimates.
The month of April 2019 ended with an unemployment rate of 7.60 percent and there is a 95 percent chance that the true unemployment rate is between 7.59 percent and 7.61 percent. If these two estimates are rounded to their first decimal place, they yield 7.6 percent, which is exactly what the mean estimate is anyway.
Monthly estimates are based on a sample of about 100,000 adults with good representation from every major state and rural and urban regions within these. These are therefore a lot more robust than the weekly estimates.
The average of four weekly estimates of a month is a good approximation of the monthly estimate of unemployment. But, small variations are to be expected. The average of the weekly estimates of April was 7.7 percent while the monthly estimate turned out to be 7.6 percent. The difference arises because first, weeks are not always co-terminus with months and secondly, because monthly estimates are adjusted for non-response while weekly estimates cannot be adjusted for this.
An interesting fast frequency measure of unemployment is the 30-day moving average. Here, the sample is of the order of 100,000 and this is therefore almost as robust as the monthly estimates.
This tells us that the unemployment rate had peaked at 7.61 percent on April 29 and has since dropped a tad to 7.59 percent on April 30 and May 1.
It is useful to keep a track of fast frequency measures such as the 30-day moving average and the weekly estimates to understand the direction of change in the unemployment rate. Weekly and 30-day moving average estimates are also produced for the labour participation rate and the employment rate. The 30-day moving average series gives us a little extra insight into the performance of the labour markets during April.
First, the unemployment rate was rising steadily almost every day. Second, the labour participation rate rose steadily till April 14 and then, it began a decline. It rose from 42.44 percent to 43.36 percent and then fell back to 42.8 percent. Third, the employment rate rose similarly steadily from 39.56 percent to 40.25 percent by mid-April and then dropped to 39.56 percent by the end of the month. April 14 was the Hindu New Year in many parts of the country – Vaisakhi in the north to Vishu in the south and Bihag Bihu in the east. This seems to have been a tipping point in labour participation.
