Please bear with my fluid physics analogy as much as you can (It was something of a first love for me in college, one I can never forget...and somewhat highly relevant too given a fluid is a macro-entity of many much more individual constituents and the forces that govern them):
All "stats" are not the same rubbish. Rather they are different levels of rubbish (and have to be qualified, sorted as to which ones are less toxic to play around in)...some even stop being rubbish all together if certain criteria can be met. If you go to a landfill, you can sometimes find stuff that should never have been thrown away for example and are worth huge amounts of value still etc....but again you need a criteria to go about finding such things.
I for one very much find the ones specified by the GDDS (IMF standards) to be fairly credible because they are quite upstream (laminar) data that is highly standardised (and cross-verified)....i.e the good relevant balance of enough distance from the source introduction (which is hidden from us and we want to measure in some way)... but not too far down with data disturbances/entropy effects.
Downstream (turbulent, disturbed) data has some level of signal loss (no longer laminar flow). Also streams that are not even that developed and broad do not have the requisite developed flow to sample effectively in first place (the signal just was not good and coherent enough to begin with)....so it would be foolish to compare apples to apples with those that have developed flow and much higher SNR. This would be why comparing unemployment data of India is silly compared to unemployment data of say France, UK, US, Japan etc.
This is why I say the data streams have to be broadened (before our dippy stick strategy to measure elements of the flow becomes sufficiently relevant and broadly standardised and comparable with the original concept/reason of measuring it)....but in many ways that needs the source flow (economy) to also broaden much more. Something of a catch 22 situation....which is why we have to debate and discuss which stats are most relevant to India's particular stage of development right now and what are the best priorities for measuring and debating on. Its broadly the more simple, directly sampled stuff that India is already big/coherent enough on right now...to measure and standardise with others and for policy goals. The more indirect and micro-derived level you go....the worse the signal loss and thus relevance. They will broaden and gain relevance down the road....its not a case of India is a country that should focus on every statistic possible....we must be choosy....we are growing the cake still....rather than deliberating on the icing and presentation and number of candles to put on it and what our next cake should be like etc.
This applies to every country/macro-entity in the world.
@VCheng @GeraltofRivia @Chak Bamu @Oscar @Jungibaaz @RescueRanger @farhan_9909 @That Guy @Indus Pakistan @Joe Shearer @waz @Arsalan @WAJsal
While this is a brilliant piece of popular science writing, and makes the idea of national statistical data easy to comprehend, it also misses engaging with the fundamentals of employment statistics in India. Here we are dealing with arithmetic: the number of apples gathered and counted in our collection baskets vs. the number on an average per apple tree, with the difference forming the number unemployed.
Specifically because Indian employment in the organised sector is so limited, it is easy to enumerate: the armed forces, the Railways, the PSUs, which are far more effective at gathering statistics than at making planes, earth movers or wireless sets (to name the products of the three that form the units under Ministry of Defence Production), two or three of the biggest private sector segments (the largest in recent years being software services, employing between 150,000 to 300,000 a year) and we're done. That leaves the informal, or the unorganised sector, and there, too, it isn't punishingly hard to estimate the number of shop assistants (guess how the number of shops is identified), drivers of private transport vehicles, and casual labourers working under contractors. These can be measured in very broad brush strokes, disturbingly broad compared to the estimation techniques available in better organised economies, but there are such brush strokes.
The fundamentals are not difficult; I've been dealing with these for the last five years or more. The average number coming into the Indian job market is 13 million a year. The number in the organised sector is half a million; the unorganised sector takes up nearly ten times more, 4.5 million. It doesn't take too many sheets of paper or too many pencils to estimate the balance, the jobless, given that we already have pretty good estimates of the population of our major urban agglomerations, and of the number of families in them, the sizes of the families and the employed members in these. Simply a matter of offsetting these from the population growth per year. The saving grace is that our National Sample Surveys are backed up, or rather, corroborated by our Census figures.
Any way that we cut these, and slice and dice them, the figures make grim reading. There has been no impact on unemployment during the NDA; the key breakaway element that swept it to power in the last General Elections, the urban yuppie descended from new entrants into cities or large towns, and trained to this side of literacy, just sufficiently to be able to handle tele-calling and other call centre jobs, hasn't got to the pot of gold promised in the election manifestos.
Lower down, throughout the unorganised sector, there is havoc. Most of the damage was done by demonetisation. You mentioned the increased propensity to file returns, therefore, the broadening of the tax net, and the broad measurement of improvement in digitalisation; however, the increase appears to be very close to the statistical increase of previous years, leading to rather baffling conclusions: it didn't matter whether there was a compulsion to file returns or not; people generally joined the ranks of tax payers at the same rate every year.
That leaves us with not very much of the balm of Gilead to spread over our lacerated limbs. But we digress.
We had an optional paper on Econometrics in our second year. We learnt then that the forming of estimates of certain factors was better done through estimates than omitted. Here, too, the same situation applies; it is better to estimate and chalk in some figures than to omit them altogether. And here, too, there is a lesson for the amateur at national income statistics: it is better to apply a range to every data count than to risk a spot estimate. While it is unfortunately true that the outcome is of an increasing degree of complexity, this is the way to ensure that there is a reduction of uncertainty of a measure.
My kaamwali, who is from the unorganised sector, is due - overdue! - to make an appearance, so I must leave behind this note, a Russian peasant to be thrown to the wolves, especially the lupine Nilgiri, and say a brief prayer for it, and move on to emptying the rubbish bucket.
