How dodgy is the data

Written by Sunil Jain | Updated: Oct 11 2010, 15:15pm hrs
Its been a pretty rough ride for Indias chief statistician, TCA Anant, and by all means things look like theyre going to get a lot worse. To begin with, there was the embarrassing goof-up on the quarterly estimates of GDP when Anant was away in Japan at the end of August, and his predecessor Pronab Sen had to be called in to mount a rescue operation, to find that the wrong deflators had been used to convert the current prices data to 2004-05 prices data. The IIP surge in July looked so funny that, in mid-September, RBI felt constrained to say the high volatility over the past two months raises some doubts about how effectively the index reflects the underlying momentum in the industrial sector.

Around the same time, at a seminar in the capital, CMIE managing director & CEO Mahesh Vyass presentation had a slide showing the growth in nominal sales of manufacturing companies for 2009-10 was a mere 1.63%. This was, in itself unremarkable, except that Vyass next slide superimposed that bit of data on the IIP manufacturing, the index that measures growth, in absolute terms without worrying about inflation. This showed the 2009-10 growth was a whopping 10.92%. (10.92% in real terms versus 1.63% in nominal terms!) Indeed, while Vyass data showed a reduction in growth, from 15.82% in 2008-09 to 1.63% in 2009-10, the IIP manufacturing showed a surge from 2.75% in 2008-09 to 10.92% in 2009-10. In other words, not only were the numbers dramatically different, they showed a completely different direction as well. Given that the index is ultimately a collation of what firms in the sector are selling, the sharp difference is inexplicable. And since CMIE is talking of a 1.63% hike in nominal sales, once you remove the impact of inflation, it goes into the negative arena.

A fortnight after Vyass presentation, a top broking house Centrum put out a report on the July IIP data that was even more damming. Most newspapers carried editorials saying the IIP looked funny since, while growth was slowing till June, the July data suddenly showed a surge thanks to a spike in capital goods. Centrum actually went into the sub-components and found the surge was due to a 308% hike in insulated cables and wires on a month-on-month basis (and 517% on a year-on-year basis). It concluded, rightly, that the data was spurious, removed it, and found that instead of growing 63% like the CSO data said, capital goods growth was a mere 0.3%. The IIP growth for July, instead of rising by 13.8%, rises by just 6.2%. This distortion in insulated cables, Centrum says, has been there since May 2007while the official data shows the May 2007-July 2010 growth was 195%, the 1995-2007 average was 12%.

What is Anant to do He has ordered an audit into the collection of various economic data, beginning with the IIP. How long the audit will take is anyones guess since, as Anant told this newspaper last month, one of the problems with the statistical system is that it just takes too long. It was always dodgy, he said, to do forecasting based on IIP data since the weights in it went all the way back to 1993-94, but it took 3-4 years to get various experts/committees to agree to using newer weights. He also said the criticism that NSS data systematically underestimated the consumption of the rich may also be correcttheres also a seminar in January on the NSS data where the usual lot of experts have been invited.

Whatever Anant does, or doesnt, its a good idea to use private data sources to validate the official numbers with, or at least to get a sense of where they could be incorrect. The HSBC India Composite Purchasing Managers Index, it is true, also has survey results of service sector companies, but it is uncanny that, over the last few months, the PMI and IIP have never moved in the same direction. IIP growth fell from April to May and from May to June; the PMI got stronger and stronger in this period. The PMI got weaker in July, and thats when the IIP soared. The top few hundred firms in the country account for the bulk of manufacturing output, so why not at least look at their nominal sales data while finalising IIP Look at sales data for firms who file taxes that are available with the finance ministry. CMIEs CapEx database has data on the investment plans of firms; why not look at this while putting out investment data NCAER has income surveys, why not look at these while making poverty estimates No dataset is perfect, but it may just help to see what the other data says.

There is a huge irony in Vyass data comparison though. In the late 1990s and early 2000s, CMIE used to publish its own variant of the IIP and that varied dramatically from the official IIP. At one point, when the discrepancy became really huge, CMIE just stopped publishing its own variant.