I asked someone what her income was last month; she gave me a figure and produced a pay-slip. I then sought to know how she had spent the money; she thought of various items but couldn’t account for Rs 15,000. I told her, I believed neither her income figure nor her pay-slip. I am sure you agree mine was a stupid statement.
Anyone with even minimal exposure to economics and national income accounting knows there are three ways to compute national income: Summing production, summing income (value of inputs) and summing expenditure. Conceptually, whichever way these are aggregated, the answer should be identical. Because of multiple reasons (different sources of data being one), however, they differ. This isn’t an Indian problem alone; it occurs in every country, the US included.
There was a huge hue and cry about “discrepancies” in the recent Q4 (Jan-March) 2015-16 GDP estimates. If those discrepancies hadn’t been there, real growth would have been 3.9 per cent, not 7.9 per cent; as a result, the growth discourse was littered with the unflatteringly colourful terms — “fake”, “spin” and “jumla”.
So what are these discrepancies? GDP was computed through the product approach, which relies on the production or supply-side; it is now determined using the expenditure approach, a bit like the demand-side. The gap between the two figures is referred to as “discrepancies”. In a country like the US, such a gap is alluded to as a “statistical discrepancy”. Perhaps, the adjective “statistical” lends legitimacy in the eyes of our commentators.
Notice that just because I have been unable to explain how some income was spent, that income doesn’t vanish into thin air; any more than Rs 15,000 from a pay-slip becomes “spin”. There is no dispute about those goods and services (their value) having been produced.
So how large were these discrepancies in the fourth quarter of 2015-16? The answer: Rs 1,43,210 crore in constant prices and Rs 1,72,106 crore in current prices. Suppose I were to ask that woman what her income was in the same month last year, do you think she would have used some kind of deflator to derive her real monthly income? I doubt it. In all probability, she would have reported her nominal income; indeed, national income is calculated at current prices and then converted into constant price numbers (and real growth) using deflators. Therefore, if I am going to spin a tale about cooked up figures, I should use the current price numbers. Apart from everything else, current price figures are higher than constant (2011-12 prices) ones.
Commentators, however, have all worked with constant price numbers, which further substantiates the proposition that they know precious little about national income accounting. An impression has also been conveyed that this fourth quarter is special because of the discrepancies, but they have always existed. Discrepancies in current prices were Rs 139,540 crore in the Q4 of 2011-12 and Rs 130,419 crore in Q4 of 2012-13. I have a series that goes back to 2006-07 (It can be dragged back earlier still). Naturally, it’s best to express discrepancies as percentage of GDP, since nominal figures are involved. For Q4 of 2015-16, that high current price figure converts to 4.7 per cent of GDP; in quick estimates in Q4 of 2012-13, the share was 6.2 per cent and in revised estimates in Q4 of 2013-14, the share was 5.9 per cent. This is hardly a case of 2015-16 being an outlier.
But why do we have such high discrepancies? After all, developed countries may have discrepancies, but they aren’t this large. That’s because our expenditure data is bad. And, moreover, “discrepancies” simply mean a residual category.
Expenditure has categories such as private final consumption expenditure, government final consumption expenditure, gross fixed capital formation, change in stocks, valuables and net exports. Of these, we have some information on government final consumption expenditure, gross fixed capital formation and net exports. The rest of it, including private final consumption expenditure, is pure guesswork, particularly when quarterly data is concerned. (We aren’t really equipped, statistically speaking, to start a quarterly GDP series.) Therefore, as we go through the cycle of revised estimates and quick estimates under the old GDP series, and provisional estimates and first revised estimates under the new GDP series, those “discrepancy” numbers themselves change.
For instance, under revised estimates, in the fourth quarter of 2013-14, discrepancy/GDP ratio was 5.9 per cent and 0.7 per cent in quick estimates. It is bound to be no different in Q4 of 2015-16 too.
Once we have a full year’s data, we shouldn’t bother about quarterly GDP numbers any longer; they are not robust. In current (and constant) prices, we also have 2015-16 GDP data for the full year, not just quarters; this is the one that showed 7.6 per cent real GDP growth for the entire year. This is also the one which shows discrepancies of Rs 9,135 crore for the full year (0.1 per cent of GDP). If a full year is superior to quarterly data, why didn’t commentators pick 2015-16, instead of just the fourth quarter? Clearly because that headline wouldn’t have grabbed eyeballs.
“It is a capital mistake to theorise before one has data. Insensibly one begins to twist facts to suit theories, instead of theories to suit facts.” That’s a Sherlock Holmes quote from A Scandal in Bohemia. But the evidence suggests that even if one has data, one can twist facts to suit theories. Either commentators are ignorant (they don’t know economics), or they are spin columnists.