9 March 2017

All You Wanted To Know About GDP Numbers But Were Too Lazy To Ask


Seetha

How on earth could the economy have grown 7 per cent during a quarter that saw 84 per cent of currency in circulation being yanked out of the system?

The economy appears to have airily shrugged off what were widely predicted to be the disastrous effects of Operation Demonetisation. Predictably, the demonetisation sceptics-cum-Narendra Modi critics have expressed complete disbelief in figures for third-quarter (Q3) gross domestic product (GDP) growth as well as the projections for 2016-17. The mildest reaction is that the data released by the Central Statistical Organisation (CSO) does not reflect ground level reality; the strongest is that the figures are fudged.

Such questioning is not new. Every finance minister has faced jibes of getting GDP numbers bumped up during scheduled data revisions to show the fiscal deficit in a better light. Many in the present government who are affronted by the disbelief over the current set of numbers have themselves mocked growth numbers achieved under other governments. It’s just that this time around, the jibes are more toxic; they are actually allegations.

In this tit-for-tat game, the hapless victims are the faceless statisticians of the National Accounts Division. Drawn from the Indian Statistical Service, to get into which they have to clear a tough competitive examination, they are finding their credibility and competence being questioned like never before.

So, does the CSO pull GDP numbers out of a hat? Can it dress up numbers following a phone call from some higher up in the government of the day? Can some ideologically aligned number-crunchers among the 200-odd, who make up the National Accounts Division, manipulate numbers to help whichever party is in power?

Those who have convinced themselves that the current numbers have been fudged can stop reading from this point. Those who have an open mind and want to understand how GDP numbers are arrived at can continue reading. As for those who think the latest numbers only show that demonetisation had no negative fallout at all, hang on and read till the end.

Since the Q3 figures have set off a storm, let’s start with that. How on earth could the economy have grown 7 per cent during a quarter that saw 84 per cent of currency in circulation being yanked out of the system?

This is something even the demonetisation non-sceptics are finding a bit hard to swallow. But this mismatch may have more to do with data inadequacy than data fudging.

The disbelief is mainly on two counts. One, how could private consumption have risen 11 per cent between Q2 (July-September) and Q3 (October-December)? Two, when the national accounts do not capture informal sector activity (which was most adversely affected), how can this be a correct picture of the economy?

Chief Statistician T C A Anant has pointed out in several interviews that there are no direct measures for both of these and the CSO has to rely on proxy measures and extrapolations to arrive at figures.

Consumption, Anant told The Times of India, is inferred from value added. So it is derived from data from producers. Now, value-added data comes from corporate sector balance sheet filings. Anant’s predecessor and former chairman of the National Statistical Commission, Pronab Sen, told Business Standard in an interview that what companies may show as sales on their balance sheets may be sales to dealers and retailers and not necessarily final consumers. Hence, the collection of indirect taxes went up, making GDP grow faster than gross value added (GVA), though people may not have been buying.

Measuring informal sector activity in the national accounts is a vexed issue, not just in India but most developing economies, where this sector is large. There is a lot of work done on this globally. In another interview to the Indian Express, Anant has explained that the informal sector is assessed through correlated indicators in agriculture, trade, construction and manufacturing. Enterprise surveys, as well as the five-year employment-unemployment surveys, are also used. And because of the very nature of the sector, none of this will be entirely accurate.

Even the data that has come in for the formal sector may not be complete, and as more information trickles in, the GDP numbers will be revised. The full year figures that will be released on 31 May will see revisions in quarterly estimates as well.

What about the mismatch between the index of industrial production (IIP) and GDP data? The latter shows manufacturing doing much better than what the former shows.

First, IIP measures only the volume of production from factories and that too from the organised manufacturing sector. GDP is now about gross value added, which includes taxes and subsidies. The GDP data also includes figures on the unorganised sector (imperfect as it is).

Second, and most important, the IIP, as it stands today, is highly flawed and in dire need of an overhaul. “This is what people should be getting worked up about – why is the IIP not being updated,” Sen told Swarajya. The base year is still 2004-05, which means the basket of goods it tracks is hopelessly outdated (the base year for GDP is 2011-12). Sen points out that the highest growth in manufacturing happened between 2004-05 and 2010-11. “The entire structure of manufacturing has changed over this period, but the IIP is capturing the sunset sectors and not the sunrise sectors.”

In any case, he says, much of the GDP growth is due to efficiency improvement and once this is factored in, the dissonance between the two sets of data becomes insignificant. Between 2011 and 2015, the growth in manufacturing according to GDP numbers, he points out, was 7 per cent, of which 5 per cent came from efficiency improvement and 2 per cent from production increase. The growth in manufacturing in the IIP during this period was 1.8 per cent.

But why assume data can’t be fudged?

Simply because of the way national accounts computing is done. This isn’t a bunch of amateurs fooling around with numbers. It is a sophisticated econometric exercise. The data sets and methodology have been laid down by the advisory committee on national accounts. The CSO is left completely free to crunch numbers but within the framework of the methodology that is laid down; there is no room for deviation.

Data is collected for eight sectors – agriculture, forestry and fishing; mining and quarrying; manufacturing; electricity, gas, water supply and other utility services; construction; trade, hotel, transport, communication and services related to broadcasting; financial, insurance, real estate, and professional services; public administration, defence and other services.

The data comes from diverse sources: agriculture ministry, the ministry of corporate affairs, departments of state governments, the Reserve Bank of India, among others. Other data sets are also mined – sales tax collections, figures on rail, road, air and shipping transport, offtake of inputs (cement and steel for construction, for example).

There are 15-odd divisions in the National Accounts Division that divide this data and work out their own estimates. All this is then coordinated by a coordination unit. Then it goes for a first round of senior-level vetting by the head of the National Accounts Division. After he okays it, the data is presented to the Chief Statistician, who gives it a final once-over before signing off on it. The head of the National Accounts Division and the Chief Statistician may point to data mismatches or problems caused by technical glitches or human error.

Nobody from outside the CSO is involved in this exercise, both Sen and R P Katyal, a retired head of the National Accounts Division, told Swarajya. There was a time when the data was shown to the Planning Commission for its inputs before being released, but this practice stopped in 1997.

What about revisions?

Data collection systems in India are admittedly weak. Data comes in late and is often incomplete, so figures need to be revised as it trickles in. Revisions are not unique to India. They happen even in advanced economies with far more robust statistical and data collection systems.

GDP data is revised four times over two years – there are the provisional estimates of a financial year that come out on the last working day of May every year. Then there are the first, second and third (and final) revised estimates. The first revised estimates, released on the last working day of January every year, grab headlines but the second and third go unnoticed.

Take the example of the 2014-15 numbers. The provisional estimates for the full year were released on 29 May 2015; the GDP for that year was estimated at Rs 125.41 lakh crore. The first revised estimates, released on 29 January 2016, pegged the GDP lower at Rs 124.88 lakh crore. The second revised estimates, released on 31 January 2017 along with the first revised estimates of 2015-16, revised it downward further to Rs 124.33 lakh crore. The third revised estimates will be released on 31 May, along with the provisional estimates for 2016-17.

Quarterly data is also revised on a quarterly basis. The advance estimates released on 28 February had an upward revision of the figures for Q1 (April-June) and Q2 (July-September) of 2016-17 and a downward revision of figures for Q3 of 2015-16 (which is another factor for the higher-than-expected growth in Q3 of this fiscal).

It is this that has given rise to charges of manipulation. But revision is also not a casual exercise. According to Katyal, new data is subjected to careful vetting, after which it is either accepted or rejected.

So, there’s no scope for political interference?

“There is always scope for pressure; the CSO is, after all, a department in a ministry,” says Sen. But the point, he says, is simply this: what is the evidence that pressure has been put on the CSO? Knowing the data and knowing the methodology, he says he was not surprised by the 7 per cent Q3 growth figure. Indeed, even in end-November, Sen (who is a vocal critic of demonetisation) had maintained that Q3 growth would not be affected. “People read more into the data than data permits them to. The data can mask the demonetisation effect. But that doesn’t make it wrong. It also doesn’t make it totally reflective of reality.”

Will the critics and champions of demonetisation both take heed?

No comments: