Government Must Be Transparent About Data Behind New GDP Methodology: Pronab Sen

To my mind, that’s the bigger story.

The base years are the ones where you actually do a full estimate. All the subsequent years are where you make a lot of estimates—you extrapolate, you make assumptions—and that’s where you can go wrong. That happens.

What was surprising about this base revision was the size of the change. Normally, we expect two or three per cent up or down, not very much more than that. Last time, I think, it was about two per cent up. This time, it is seven per cent down.

That’s a large number. What it means, in effect, is that between 2011-12 and now, we have been systematically overestimating growth.

And the question in my mind, which is an important one, is: can we identify what caused that overestimation?

This was not deliberate. There was something in the way things were being done that resulted in this overestimation, and that needs to be identified and fixed.

That really is the big story.

Now, things have become a little confusing because you have that one problem, which is pointing to a problem that already existed earlier. And now you have changed the methodology.

So I have two problems working together: the problems of the past and the problems with the new methodology. Disentangling the two is going to be an issue, and I don’t think it can be done in a short time.

You need to have more data under the new methodology to be able to disentangle this. So the problem is that it has become more complex.

Given the controversy it has raised, I think the more transparent we are, the better off we are.

First, there has to be an analysis. I don’t know whether they have done it or not, because they were in a hurry to do this base change.

But when you do a base change, you really have to look backwards and ask: which are the sectors in which I was having an estimation problem? What was the reason for it? Can I fix it?

That’s the first set of questions.

Then the second set of questions is: how do I make further methodological improvements? That would include things like double deflation.

This is my basic question: you cannot get producer prices the way you get wholesale prices.

For wholesale prices, you go to the market, go to a wholesaler or dealer, and take a quotation. End of story.

Here, the manufacturer must give you the price. It has to come from his accounts.

Now, the problem we had earlier was the manufacturers’ reluctance to provide this information. They would say, “Hey, look, this is top secret as far as we are concerned, because I have many distributors. I don’t sell to them at the same price. Some I will give a discount to, some I will not. If this information gets known, first I would have a problem with my distributors. Then I will have problems with my competitors, because they will know what game I’m playing.”

So this was the nature of the problem: a complete refusal to provide the data.

If they have now managed to get this data, I would give them a thumbs-up. This is an achievement. But please tell us, how did you get it?

This is an old argument. We’ve been using it for seventy years. Every time our people go out to collect data, they make exactly the same pitch: “This is government data. Confidence is guaranteed.” So this is a standard script.

And under the Statistics Act, if any of us—the data collectors—leak the data, we go to jail. There’s nothing new in this.

But people still say, “We won’t give. Do whatever you want.”

There is, however, another question that people ask: if other countries are able to get this data, how do they do it?

The way they do it is by using the Collection of Statistics Act, which we also have. It says that if I, as the government or a statistical agency of the government, ask you for particular data and you refuse to give it to me, I can prosecute you.

In our history, we have never enforced it.

All requests for data go under the Collection of Statistics Act. All of them. But we have never enforced it. We don’t want to get them upset with the government.

So that is the third problem.

Again, you have to get very granular data (for double deflation) because, if you look at any product, the finished good is a single item. The number of inputs that go into it, however, are many. Even if I club them into broad categories, there would still be many.

Take agriculture, which is probably the easiest example. You have one finished-goods price—the price of the crop. Then you have the price of fertiliser, the price of pesticides and the price of seeds. You have three inputs going in. So, for every crop, you have three inputs.

Manufacturing is far more complex. There are multiple raw materials and various other inputs that go into the production process. So what Mr Mohan is saying is absolutely right: to do double deflation, you need granular data on inputs.

We have data on output. That is not the issue. We’ve been collecting that forever. Inputs are the problem.

If you look at the data that we collect, inputs account for about half of the data, which means that, roughly speaking, output and input are equal in terms of the number of items. But the need for input data is much larger.

That would have involved a massive expansion of data collection. So where did this data come from? How was it collected?

There is a pattern of opaqueness across all three problems.

You can do double deflation if my manufacturers are cooperative. They would be able to give me both the producer price as well as the prices of the inputs they are buying. Then I can do the deflation.

But they don’t give it to you.

If you go by the balance sheet, it will give you the rupee value at current prices—how much the manufacturer has spent. Nowhere are the prices mentioned. Look at any balance sheet; you won’t find the prices there.

So you have to get the information from the producers themselves.

You can get both the producer price and the input prices from the same source—the producer. But was this effort made? Why didn’t we know about it? How was this data collected?

There has been no communication on these fronts. Nothing!

But (earlier) we consistently told the IMF: “Look, at the end of the day, the methodology that I adopt for any calculation should be driven by the data I have available—by what I can actually collect. If there is a methodology that requires data that I cannot collect, I shouldn’t use that methodology.”

That was the problem.

So we have always deviated from best practices because we were not in a position to collect the required data. We told the IMF this, and the IMF accepted that.

We’ve had improvements in the data. We now have what is called the Annual Survey of Unincorporated Enterprises.

But that is an annual estimate. It is not an estimate that you can break down quarter-wise. An annual estimate is the best we can do.

And even that comes out about six months after the year has closed. So you can use it only for annual estimates, and even then, not for the provisional or the first estimate, but for the first revised estimate. Before that, you simply don’t have the data.

So, in real time, we are never getting that data.