Why was there an error in stating Q1 GDP?

Nifty BSE | NSE 09/02/12 Are Mutual Funds betting on Nifty?

The highlight in the Indian economic news space this week was the fiasco regarding the GDP numbers. Probably for the first time the government had to announce a revised GDP number because of obvious errors in the first number.

On August 31, at 11 am, the Central Statistical Organization or CSO, which calculates the numbers, announced that GDP for the first quarter of FY11 grew at a decent 8.8%. But curiously other numbers in that release showed a very poorly growing economy. For instance gross capital formation grew by only 3.7% year on year and private consumption by 0.3%.

The most damning was that GDP counted according to market prices grew by only 3.7% though counted according to factor cost GDP had grown by 8.8% So what's the difference between the 2 GDP nos.

GDP at factor cost takes into account all the goods and services produced in the year and multiplies it by the cost of producing them. GDP at market price looks at the gross domestic product from the consumption side.

•  How much did individuals consume?
•  How much did government consume?
•  How much did we invest in capital formation? 
•  Net exports

Normally, the difference between the two GDPs is very slight. If one adds indirect taxes to GDP at factor cost and subtracts subsides you get GDP at market prices. But there is another factor in GDP calculation. Both these GDP nos at factor cost and at market price are not presented in today's prices. Because in that case one can't capture how much of the higher output is because of higher production and how much because of prices.

So you present both last years and this years output according to prices in the previous years call the base year. In our case it is 2004-05, This means you calculate the GDP at factor cost and at market prices at today prices and then deflate it or reduce it by a relevant inflation index. And this is where the mistake appears to have happened. Someone used the wrong deflator.

Expert statisticians say for instance the personal consumption and govt consumption numbers should be deflated using the consumer price index. Capital formation number should be deflated by using the capital goods index in WPI.

After a lot of angst from economists and the press, a red faced government made the changes and said that the GDP at market prices grew by not 3.7% but by 10.2%

So how did the Central Statistical Organization (CSO) or the Ministry of Statistics let this error slip past? And how reliable are the other numbers put out by these organizations?

In an interview with CNBC-TV18's Latha Venkatesh, the Director General of the Central Statistical Organisation, SK Das spoke about how the error in using the deflator to calculate GDP at market price occurred; and veteran economist Ajay Shah, Senior Fellow, NIPFP gave his perspective on the anomaly.

Below is a verbatim transcript. Also watch the accompanying video.

Q: How did the error in using the deflator to calculate GDP at market price occur?

Das: Actually to work out the constant price, we have to work out the appropriate deflator. That is a mix of, somewhere it is CPI and somewhere it is WPI and somewhere we use a combination of both actually. So if you see the GDP at current prices that growth rate is okay. But while converting into constant price there was a slight mistake actually. That we have also mentioned in our press note that there has been use of inappropriate deflators in converting the estimate at the market prices.

Q: Is it that personal consumption and government final consumption is deflated with CPI and capital formation is deflated with the capital goods index and the WPI, is that how it is done?

Das: This is a combination of both the WPI and CPI actually. If you take this final consumption expenditure for food items we use WPI and some other item, we are using CPI actually. It is a combination, it is not that we are only using CPI.

So in those cases, we have to choose the appropriate deflator. So the mistakes happened in some of these cases. When we looked up carefully after this 3.7% and 8.8% things came up in the subsequent morning. We found that there were some aberrations. So the next day we made a thorough check up of all the items in the market price.

Actually the estimation of GDP at market price is a difficult exercise because quarterly data for these things generally is not available very freely. We have to make a lot of estimations and corrections.

Q: Are there not counter checks? At what level is it done? Is it done at a GM level, at Assistant Director General level, are there not counter checks within your organization?

Das: There are actually. Because it is done at one level and at the level of director and checking done at the level of DDG, at our joint secretary level and additional secretary level also. When the error came to light we were also quite surprised how this could actually happen when there are so many checks and balances.

Q: Let me get a professional economist in. People who were harassed by this wrong reportage, Ajay Shah, the Director General is listening. It's the GDP number the error came to light. Is that now costing aspersions, casting doubts on all the data put out by our statistical machinery?

Shah: It used to be a bit of an insider thing that economist were very conscious about the bad data quality that comes out of the NSSO of the National Accounts Statistics. For example I have never written a paper that uses the annual survey of industries that uses the NSSO that uses the NAS because we consider them to be very low quality data products. So I guess this is just validation of their inner problem. It's good that the problem has come to light.

The question is are we able to execute more fundamental reforms. It's not a matter of correcting one mistake. The question is, are we able to fundamentally change the way the statistical machinery functions?

Q: How is the machinery itself organized? What is the responsibility that the CSO has to this Chief Statistician himself and what is the responsibility of the Chief Statistician to the nation for the mistakes that are committed? Can he ensure that those who made the mistakes are penalized in some form?

Das: First of all, the point made by Ajay Shah actually, the statistics brought out by our National Accounts, if you see the international forum also, these are all appreciated. I don't know anybody not using it. In fact if you see the people around us, everybody is and was happy with the data brought out by NAS.

We are bringing out this quarterly estimates since quite long as per the AGDS of IMF. I don't think that we have ever made a mistake. It is a very rare phenomenon. So I don't know how Ajay Shah has made this point that he is not using the data of NSSO. If you see the amount of users of NSSO data it is quite astonishing that in the last 3-4 years the number of users have gone up tremendously.

Q: Even if you can point out the first two or three problems you have with this data?

Shah: One is a very fundamental point in how a data shop should be organized. I don't know how CSO works, but I have a suspicion. I have a suspicion that they are using spreadsheets and these things are being done manually. So when you say that somebody else checks the data, it actually makes me extremely nervous.

If you want high quality in a data shop there should be complete computerization, the numbers should come out from a computer programme and no human being should be involved in anyway with the functioning of the computer programme. I spent some years inside CMIE and all the statistics that come out, the household survey, information, the firm level databases, they are completely processed by computers. You debug the computer software and then it just works. You don't rely on a human being in the picture.

So I actually get very nervous when you start saying that are there multiple human beings who do the same calculation with a spreadsheet over and over. Spreadsheets are actually not a safe form of calculation. People who do professional work with data will always build a software system and human beings will not interfere with the functioning of the software.

In terms of the deeper questions of data quality, in the National accounts in good countries there is an entire expenditure side which is based on a household survey. In India, it's really a bit fake because we don't have a household survey. The entire expenditure side is mostly just made up.

In the Indian Express, Mr. Raghavan has an article where he quotes an ADB study about data quality, a comparison has been done in all the Asian countries. I seem to remember India is ahead of Pakistan and Bangladesh but for the rest we are at the bottom. I think there are enormous problems with the official statistical machinery, which should not be claiming that all is well.

Q: The issue that both I and Ajay are raising is that when you are applying a deflator to a number that has been accumulated. I am not going to the question of how you accumulated that number itself, the GDP at market prices, at current prices. Let us take that as a given. But when you are applying the deflator, that should be computerized. I mean this should run as a programme. Why is it that you have to have a manual intervention at all over there?

Das: There is no manual intervention, it is computerized actually. I do admit. But while calculating, when you are deflating different items within the private consumption expenditure, what figure should have been used, instead of that some other figures got in actually?

When we work out and have a time target, due to a lot of pressure of work this minor mistake was committed. As I told you earlier we could correct it. But I can assure you in the past we have not made such a mistake. I still don't remember if we have made any mistake earlier. This time mistake has happened and obviously this is a learning experience. We will be extremely careful in the future. I can assure you that we will not be making any such mistakes.