What's wrong with India's economic data?

Farm and factory data collection methods urgently need to be expanded and electronically integrated to provide accurate and meaningful numbers.

Irony by definition is laced with dark humour, but it would be more like black humour when one views data systems in India.

Less than a week after the Reserve Bank of India (RBI) expressed concern over the quality of data, the Central Statistics Office drastically reduced the Index of Industrial Production (IIP) growth number for capital goods for April to 7.3 per cent from 14.5 per cent when it was initially announced.

Such revisions are quite bizarre and have a deeper implication because such high-frequency data is used for high-frequency monetary policy stances.

The RBI's frustration is palpable because the April number would have provided the justification for increasing interest rates based on robust investment growth. But, now it turns out that capital goods production was, at best, stable.

One may recollect that for a long time we had spoken about our base years being anachronistic at 1993-94, and bringing forward the base year to 2004-05 was pragmatic.

For some reason, the GDP, IIP and Wholesale Price Index (WPI) indices were changed sequentially.

A statement often made was that there was high volatility in the growth numbers that would be addressed by the new series.

Though having a new series is necessary, the larger issue is whether it addresses the question of volatility. Volatility in the jargon of financial markets economics means the standard deviation of the growth rates.

Now, the annualised volatility for the IIP old series was 16.4 per cent and it has increased to 22.6 per cent in the new series.

For the WPI series, the annualised volatility was 11.5 per cent and 10.2 per cent respectively.

So there has been improvement in the WPI, but not the IIP.

Curiously, when the same volatility is reckoned on annual data, then the IIP volatility changes from 2.96 per cent to 4.76 per cent and 1.67 per cent to 2.42 per cent for WPI, meaning that the older series fares better!

Three conclusions can be drawn here. First, the volatility argument does not hold.

Second, the fluctuations in growth rates will vary based on seasonal trends as well as base year effects. Third, the problem is with our data collections systems and processes.

What are the problems with our data? There are basically two areas that need to be addressed.

The first concerns farm prices. Today prices come from the mandis, where both transactions and prices are opaque.

Agricultural Marketing Research & Information Network (AGMARKNET), the official source of data shows prices within a rather wide range that is not helpful.

Further, the single numbers that are displayed vary significantly from what commodity exchanges like NCDEX collect from the ground level.

In fact, the volatility of prices is much higher for exchange prices compared with the WPI because the latter are based on modal values that do not vary very often.

Further, since farm products are seasonal, they do not enter the mandis every month but are traded widely all the same, which moderates the WPI numbers because they are dependent on the quotes received from mandis.

The solution is to have electronic mandis where all transactions are recorded so that we have actual prices that can be weighted by the trades that take place.

The other pertains to manufactured goods. If one looks at metals, our WPI shows that prices are increasing when globally prices are falling based on World Bank data.

India is a price taker in all metals except iron and steel.

Therefore, there should not be such differences in price changes.

The problem in manufacturing is exacerbated by the fact that one-third of total manufacturing comes from the unorganised sector where it is difficult to get timely information.

Even for the organised sector, the WPI index for a particular product often does not change for weeks because of non-availability of data after which there is a sudden spike in prices.

One way out is to make it mandatory for all firms to record every transaction at the factory gate.

By linking this to tax filing, firms will be compelled to report accurately the true picture.

Admittedly, this is a challenge considering that even unaudited results of firms are not always filed on time.

This being the case, what should be our approach?

First, we need to move away from providing high-frequency data if we are unable to vouch for its sanctity.

While revisions happen everywhere, changing growth rates by 50 per cent is dangerous for policymakers.

Second, we need to electronically connect all the mandis and have a database in which all firms registered with the Registrar of Companies have to mandatorily enter their production and price numbers.

Third, we need to look at other leading indicators when taking policy decisions based on monetary data that is generally more accurate because it comes from a smaller universe of commercial banks.

Fourth, whenever we interpret data, we should never look at single month data points to eschew the trap of base year and seasonal influences.

It would be better to look at cumulative numbers, especially for the real sector.

When we look at annual IIP growth rates, we do not look at March over March, but the average of 12 months over the same of the previous year. This automatically factors in the so-called volatility due to inaccuracies or seasons.

Finally, the RBI should seriously think of going back to two policies with need-based Keynesian intervention.

While adopting global practices like the Federal Reserve is progressive, other authorities do not have distorted images in the form of inaccurate data like we do.

We will have to wait some more time to reach these levels.

The writer is Chief Economist, CARE Ratings. These views are personal.