Sudipto Mundle interview: ‘Nobody has questioned integrity or methodology of back series GDP data’

In an interview with P Vaidyanathan Iyer, Mundle talks at length about the second best option the committee adopted to arrive at back series data. The production shift method employed by the committee produced back series data that showed double-digit GDP growth for two years during the UPA regime, which of course, gave enough ammunition for the Congress to take a swipe at the NDA government. “Despite this, politicians have not questioned the integrity of the method adopted by the committee. Let someone else come up with a better solution, and we can always discuss that. But for now, there is a back series data that we have arrived at,” he said.

READ | GDP back series was second best solution: Sudipto Mundle

Why was the need felt to set up this committee? Wasn’t the CSO supposed to arrive at the back series data on GDP?

The committee was set up by the National Statistical Commission (NSC) as one of five different technical committees looking at different aspects of the database of Indian economy. Ours was one of them. And this was set up three years ago by the NSC.

OPINION | GDP Report – misreading the facts

C Rangarajan, the first chairperson of NSC, had an overarching and detailed agenda of reforms for improving the statistical system. His commission was followed by several others — all of which worked on strengthening the database. Ours was the committee on real sector. The mandate was to look at different real sectors — agriculture, services, industry. How to use micro-macro linkages, and also see if we could link the old and new GDP data series, which was to be looked at by the CSO, which it generally does whenever the base year is changed. This helps evidence-based policy design. For the entire corporate sector, the GDP series is very critical. A lot of market projections are based on the backbone of the GDP series. We took it up as one out of half-a-dozen other aspects, which are all part of the committee’s report.

What was the primary challenge? What were your considerations?

First of all, when the new series came out, the base was made 2011-12, as opposed to 2004-05. For three years, the CSO gave a number to the old series. So some comparability was provided thereby. The year 2011-12, the first year of the new series, the new series actually has a slightly higher growth rate, compared to the old series. The question was how to link up the two. Traditionally, one would go backward using the data sources you have, but that was not possible because the new series is based on a more robust database than the old series. The new series uses MCA-21, which gives you data on the audited balance sheets of lakhs of companies, an unbelievably rich data source. There are many other improvements too. There are some disadvantages, which hold you back. For instance, you cannot use this new data source backward, because it didn’t exist. You couldn’t go beyond a few years. This is the main reason the CSO couldn’t do it either. They were struggling. For a few years, they could use MCA21, then for a few more years, they could use a private source, CMIE, which has a limited database.

READ | Back series GDP figures are not official estimates: MoSPI

Was there a better way?

Another way, which was my suggestion, is that you don’t have the new data source going backwards. But the old databases which were being used, those still exist going forward. One is to do a back series. The other is to do a forward projection of the old series for a few more years and then see how best you can splice it, see whether there is a pattern. This requires much more work, and has to be done internally by the CSO because in that traditional method, there are n number of detailed sources they would use, some they would use indirectly for carrying forward data. But it could be done only internally. I have left that suggestion in the report.

Why and how did you zero in on the production shift methodology?

Production shift method was feasible and could be done relatively easily econometrically. You start with a certain base year, whatever the GDP number in the old series, take it as correct. Then for the first year of the new series 2011-12, take that as also correct. Try and econometrically project, so that you start with the base year of the old series and end up getting exactly the same number of the new series in 2011-12. The base is changed because the structure of economy changes. Typewriters give way to computers, etc. But these changes do not happen overnight. The best way to mimic this econometrically is a gradual change. Gradual adjustment for the old number for the base year 1993-94 to 2011-12. Every year, only a little adjustment is done to end up with the 1.5-2 per cent difference between the old base and new base.

READ | As per new GDP measure, UPA decade growth revised up to 8.12%, hit 10% in 2006-07: NDA panel

Surely, you looked at sub-sectors too…

This was applied at sub-sector levels in manufacturing, services and agriculture. It is econometrics at that level. What it did not do is go back to basic data sources.

Why did it (the CSO) not dig into data using the old data sources then?

That’s what CSO does and can be done internally only by the CSO. It is so much detailed, arcane work that you need a whole army of people. That is why my suggestion of the forward projection, the committee could not do it, but left it for the CSO to explore. But what you could do, and is perfectly a reasonable thing to do (after all, econometrics is a recognised discipline) is to econometrically link the two as we have done in a way that was consistent with the base year as in the old series and the new common year. We did this at the sector level, and aggregated it to the overall GDP.

Would you say that the production shift method used to produce GDP back series raises the confidence level of market players?

First of all, the new series base on 2011-12, as per my view, is based on a more robust database. Secondly, there have been some conceptual adjustments to bring the Indian national accounts more in line with International Systems of National Accounts. But because we used a much better data source, it didn’t exist earlier. So, what can we do? We had to go for the second-best solution. We had to find an indirect way of connecting the two, without 1) in any way tampering with the new series because this is what is accepted, and 2) keeping the old series undisturbed as far as possible, at least at the aggregate level. Each year, the change is marginal; at the end of it, it covers this 1.5-2 per cent gap. That was the logic.

Did you seek or receive affirmation from other agencies, for instance, multilateral institutions?

The short answer is ‘no’. This has been on the public domain since mid-to-end July. Nobody has got in touch with the committee members or me so far. Whatever I have seen, as reported in media, virtually everybody – 95 per cent of professionals – have appreciated this idea. Also, I have to say, when the politicians started discussing this, all that I have noticed in newspapers, nobody questioned the integrity of the new series or its methodology.

If one could explore back data using older sources, would it give a remarkably different set of numbers?

The digging of data is possible only going back a few years, not more, because the data doesn’t exist. This entails huge amount of detailed work. With all that effort, you can maybe go back to 2005-06, i.e. 5-6 years back beyond what was available. But you cannot generate a 15-year series. The data (new series) came out only in 2015. It has been there for three years. Obviously, they (the CSO) had been trying, and found it difficult. They would have to use the traditional method for the earlier years, and forward cast then data sources such as Annual Survey of Industries. The other is that, some of them had written papers, suggesting the production shift method. I am not sure if the CSO tried using that.