Critics have ridiculed the Index of Industrial Production (IIP) numbers released for January, as these showed a spurt in production of consumer non-durables. T C A Anant, the government’s chief statistician, says IIP numbers are volatile not only in India but elsewhere, too. Those interpreting the numbers, he tells Indivjal Dhasmana & Dilasha Seth, should be cautious enough to not jump on to conclusions from merely a month’s data. Edited excerpts:
The IIP numbers thrown out in terms of some consumer non-durables have prompted many economists to say the data is no longer reliable.
It depends upon what you mean by the word ‘reliable’. In a statistical sense, IIP has been suffering from volatility in the past. The volatility is principally because of the nature of the index. It is a Laspeyres Index, where a fixed set of entities are reporting production figures. In such a thing, volatility is inherent...it’s not limited to the IIP in India. If you look at similar indices in the world, there are similar volatility problems.
Can the problem be rectified?
The answer is, may be. It requires stronger data collection effort. For example, we could try to have a more dynamic mechanism, say, we would move from a Laspeyres Index to a chain base index (which keeps updating the base year and the items in the index). We could also update our entities in which we gather reports more frequently. But that would require a governance system able to approach industries, collect data from them, which has a lot more strength. One difficulty in statistics in India, as with other developing countries is, you are collecting data in systems where governance structures are weak. And, your quality of data is only as good as the structure of the governance through which data is collected.
What governance structures are you really talking about?
Our ability to track which industries are producing what goods, to know how much they are producing, to have a system to be able to go back and check with the industry how much production took place. If you notice, these are not the statistics people (that collect data). (They) only compile data, which are collected from various sources. A lot of this data is reported under administrative arrangements. Those administrative machineries and their ability to track the sectors also need to be looked at.
What efforts are on to streamline IIP data?
We are in the process of revising the base year for IIP. This time, we may constitute a special working group for revision of the IIP, as was done in the WPI (wholesale price index). We have already done some internal work from our standing committee on methodological improvement. There is also a report of Unido and the United Nations Statistical Commission on revised standards for industrial statistics, specifically for IIP. In the last five years, we have put in a number of mechanisms to ensure the ASI (Annual Survey of Industries) results come out more quickly. So, we will examine if it is more feasible to put into place other administrative arrangements to improve the (IIP) index.
Will you revise the base year from 2004-05?
The committee will have to take a call. It has not yet been done. One possibility is that we can look at 2009-10; it is five years after the 2004-05 base. But whether we will use 2009-10 or choose some other method...for example, if they go to a chain base index, they may not suggest a base at all. They can ask for a more dynamic base updating process.
IIP numbers are used for policy making. The media also interpret these. If these figures are so volatile, how could they be analysed with precision?
We have always asked people to be careful about looking at month-to-month variation. Second, we give you a set of numbers; how you choose to interpret is your call. You can look at annual growth rate or can look at monthly growth rate or the series/cumulative effect. A monthly figure is only an indication of any event. If it is sustained over more than a month, then you can say there is a change in trend. If there is not, it is simply an event. Policy makers do not react to a single number; they see the pattern that emerges over a set of numbers. Trying to read too much into a single number is uncalled for.