At the 40th conference of The Indian Econometric Society held at the Institute for Social and Economic Change in Bangalore, B.S. Minhas — the distinguished economist, statistician and battler for data quality — spoke on the unreliable, variable and uncertain quality of Indian data used to create projections of the Indian economy. He pointed out that this data forms the basis on which many institutions make forecasts of gross domestic product growth and its composition.
At the National Council of Applied Economic Research, I pushed for the creation of forecasting models for short- and long-term forecasts not only of the economy as a whole but also of its major sectors, and also in depth, of selected sectors, as for instance with cement. The purpose was not to give certainties but one more tool to be used with other tools. But they were not intended to replace application of informed judgment of the user. In the open economy of today, an informed understanding of the way in which the economy is likely to go is a great competitive advantage. Merchant bankers, commercial banks and economic research institutes give their own forecasts of the economy. In many cases, there is no clarity about the basis for the forecasts. They often vary by as much as 2 per cent between each other. The Central government, through the Central Statistical Organization, also supplies forecasts. Many regard these forecasts as being almost accurate since they are from the government.
Even forecasts based on forecasting models have assumptions built into them that often reflect the ideology or bias of the maker of the model. That is more reason to use such forecasts with caution and mitigated by many other observations. But many of the forecasts made today have no real basis and reflect individual opinion, of which we must be cautious.
For example the “India Shining” campaign is based on the CSO’s estimate of the growth in GDP in the third quarter of 2003-2004 by 8.4 per cent and the subsequent “advance” forecast (started recently) that next year the economy would grow by 8.1 per cent. This is a weak foundation for such a claim, even when we accept that there is a high level of business confidence today and even among many in urban India.
The CSO is under the administrative control of the government. There is always a suspicion that the CSO tailors its estimates and forecasts to suit the needs of the party in government. It has been many times recommended that statistics should be neutral and reflect the best reality that can be estimated. But no government in India has accepted this argument.
The ISEC deliberations raised the issue as to the basis for the assertion of quarterly growth by 8.1 per cent. A quarterly estimate is perhaps made by CSO because multilateral financial institutions like the International Monetary Fund want it. They want it to complete their own forecasts about the future of the world economy. But it is well known that there is little reliable data about the economy for a quarter. Our data gathering system keeps revising its measurements of the different sectors many years after a year has ended and has no mechanism to get comprehensive and reliable measurements at the end of a quarter. Further, the estimate for a quarter may have to smooth out the cyclical variations in many sectors of the economy, for example in agriculture.
If growth in the harvest quarters and not those in between were given due regard, the off-harvest quarters would show substantial decline. Similarly, demand for many industrial products is also seasonal. This is also true of services like hotels and even some financial services. If there were to be wide variations in each quarter, the resulting figures could cause panic.
CSO’s estimates of national income are changed over the next three or four years as better measurements come in. The final estimates have many times been very different (usually lower) than the first one. But by the time the final revision is made, many years have passed and there is little attention paid to the downward revision.
Further, almost 60 per cent of the national income estimate is based on hunch and guess (“proxies and ratios”) because there is no basis for them. This is so not only for estimates of services and small-scale industry (neither of which has quarterly or annual actual measurements). Even industrial production and organized manufacturing have extreme unreliability in their annual measurements since only a fraction of the producers report each year on their production for the year. Further, the same producers are not reporting each year, so that the aggregates in different years are not strictly comparable.
Thus the data on which estimates are made is based on incomplete and inconsistent measurements. The estimate guesses the numbers for many sub-sectors. Forecasters base their forecasts of the future on these estimates. When measurements for a whole year, after it is over, are so unreliable, can we honestly expect them to be even of the same quality for a quarter' All estimates and forecasts of GDP are seriously biased by the poor quality of the data on which they are based. Individuals and institutions that base their investment decisions only on such data do so at their risk.
In NCAER, I got our forecasting team to probe regularly into the minds of businessmen. Businessmen base their investment and expenditure decisions on their expectations of the future. The Business Confidence Index is a way to get short-term expectations of businessmen to supplement forecasting models. Perhaps this exercise might be extended to other experts so that a Delphi-type estimation is possible to check the forecast from the model.
There are also regular and large-scale sample surveys that give information about consumption. The National Sample Survey of the CSO, the Market Information Survey of Households by NCAER, and other smaller surveys by market research agencies like the ORG and others supply such information. They are very useful in looking at trends in household consumption and purchase, ownership of durable products and in the case of MISH, even the changing income distribution. This data is available for long periods and hence can be useful in developing forecasts for the future. But they must still be moderated by other observations and judgment about changes in lifestyles that influence consumption and technologies that make newer ways possible of satisfying human needs.
This is not to say that we should not use the estimates given by the government agencies. This is the best information that we have. While we should make every attempt to improve them, we must also use them, but moderated by other observations. We must know the error in such information and in the forecasts trotted out by different agencies.
The conference at ISEC accepted that estimates and forecasts in other countries are also based on incomplete and unreliable measurements. It advised that they must be used with caution, be questioned at every stage, errors built into them be calculated, and forecasts must define the models on which they are based and their assumptions.
India may be shining just now because of a good monsoon, an improving world economy, rising foreign exchange reserves, an appreciating rupee and a thriving IT industry. We cannot say yet that this is an indicator of the future. But the mood is good and if the gods smile, we might do well (though not as well) next year. We will need luck, apart from good management. We have still to do a lot to make India shine in the long-term.