Let’s get our data right first: Financial Express 10th Febuary 2012

In a lighter manner, it is said that published data is as whimsical as forecasts made by a plethora of experts and organisations, including several government arms. The recent revisions of GDP and exports data cast a shadow on the robustness of our data. The downward revision of GDP growth by 0.1% points implies around R5,000 crore in real terms. In fact, even the balance of payments data, which is normally not susceptible to change as it is based on actual dollar inflows, has shown inconsistencies, leading to wider debate on what should be done about it.

Forecasts today are taken with a full glass of salt, given that there are variances in what even various officials project or forecast or target or hope. The foreign organisations tend to understate their numbers while Indian banks strike a cautious balance. One is not quite sure how seriously these are taken. But what is a concern is the actual data put out by the concerned agencies, which changes frequently and is used by all.

Official data is sacrosanct and used by analysts to take their own views, which are sent out to their clients. The media makes hay with various views of experts and analysts, which then spooks the markets—stocks, money and forex. Then, ironically, various other government bodies, such as the ministries and RBI, have to formulate their policies based on these numbers and could end up taking an incorrect call in retrospect when data is revised.

One can look at some really drastic changes that have taken place in data sets that affect policy. For FY10, we had started of with advance estimates of 7.2% growth in GDP, which is mediocre compared with 8.4%, which is the final rate. Similarly, in case of capital goods this year, in April 2011 we had the provisional estimate of 14.5% growth, which caused optimism to overflow. But with a series of changes, it is now 6.6%. In the same period, the overall IIP has been revised downwards from 6.3% to 5.3%. Clearly, any policymaker will be flummoxed with such revisions, especially so as decisions are taken with the first number in mind. The problem is with collection of data.

Let us look at the frequency of data releases. WPI on primary and fuel products used to be a weekly affair with a lag of a fortnight, while manufactured products were monthly. Industrial growth numbers and CPI inflation come on a monthly basis, while GDP, balance of payments are quarterly handouts. Banking data comes fortnightly (credit, deposits and money supply), while others come weekly, such as forex reserves. Agriculture data comes out as advance estimates with up to four rounds, with the last one coming 4-5 months after the year. Banking data is usually more robust based on returns filed by banks periodically. The others are provisional data and tend to change drastically at times.

A higher inflation number causes bond yields to rise and stocks to fall. Forex rates move when there is news of a high current account deficit, though this data too comes with a lag. There is a cost involved in terms of market

capitalisation, GSec trades or forex transactions. Also, such numbers create expectations of RBI policy moves, which in itself cause volatility in the call market.

Why does this happen? The first is that we are forcing data to come out with higher frequency without updating our systems. We have changed the base year with the hope that numbers become more reliable. But when the underlying systems remain the same, the revisions will remain.

Second, a large section of our economy is unorganised, like agriculture, small industry and services such as transport, retail trade, real estate, storage and hospitality. The problem then is that accounting systems are not maintained and data is based on imputations, causing the GDP numbers to change.

Third, the APMCs are still traditional in operations and the hopes of linking them electronically are still several decades away. Farm prices are reported in a haphazard manner and often there is a major variance between prices recorded in the mandis and the prices disseminated by commodity exchanges such as NCDEX, which are based on a more scientific MIBOR like polling system.

What are the solutions? The first is that we should abandon the system of having high frequency data until such time we are sure that the numbers are robust. Second, we have to ensure that data reporting takes place at the right time and manner. Electronic linking of mandis is absolutely essential so that all transactions are automatically captured. Third, companies should be compelled to report their factory gate production dispatches just the way the tax department ensures tax payments. In fact, given the spread of telecom in the country, every production unit should perforce be made to submit such returns to the concerned department and it can be linked to the provision of public utilities, especially for the small sector. Admittedly, this is a huge task as even the tax department has not been able to get them in the net.

Given these issues, it is quite appropriate that we have decided to settle for only monthly WPI numbers. Maybe even for others we should wait before releasing data. And, if we must, we should also alongside have the error probability, as this will help in using such numbers for taking business decisions.