Policy will always be about as good as the information that goes into it. There are many grouses with our data ? sudden changes, little explanation, and general opacity. After two months have elapsed, we still do not know quite how the axe was taken to GDP originating from non-bank financial institutions beginning 2000-01, as a result of which the annual growth slumped to 4 per cent. Whenever a new number is released, it has become crucially important to check whether the driving force is in revisions made to the data for the reference period.

But it is not this bane, serious as it is, of our statistics that we address today. All readers will have noticed that the data always informs as to what the growth (or an occasion decline) was with reference to some past period: such as, this year versus the last year. Quite fine thus far. The problem arises when we look at time segments that are smaller than a year ? one quarter or a month. Through the year, as everybody knows, there are seasonal variations in levels of economic activity ? originally caused by the inherent seasonal property of agricultural practices. As the share of agriculture declines in an economy, other forms of seasonal characteristics remain in the data, either as legacies from the past, or on occasion the outcome of new developments.

In India for instance, for a variety of reasons the first quarter of the financial year (April-June) has been when annual shuts are taken in industrial plants. The reason originally was because most of the workforce still had their roots in the countryside and would take leave to help with the harvest and preparations for the summer crop. The monsoons as we know bring much needed rain, but also disrupt movement of goods, and the difficulties of working in wet conditions are always greater. So, the uptick in industrial and commercial activities follows upon the end of the monsoon and continue into the end of the financial year ? the last two quarters.

Clearly, it makes little sense to directly compare the data for the fourth quarter with that of the third, of for that matter data for December with that for November. That is why we have followed the practice of measuring the change ?with respect to the comparable period of the previous year?. Thus, we also get an annualised rate of change.

But there is a problem with this. Say a slump in the level of activity occurs in Q1 of a year, followed by an improvement in the next. The year-on-year measure that we use will give us a low or negative growth for Q1 and maybe better numbers in Q2 ? which might tell us that there has been an improvement in Q2. But this indirect signal might not be so clear ? for much would depend on what went on in the last year as between the first two quarters, and maybe even in the year before that. As more and more time periods enter our analytical frame, we rapidly lose information and the likelihood that nothing will emerge is distinctly higher than the more fortunate outcome.

The obvious solution is to find a way to make a direct comparison between the respective quarters. There exists a technique called seasonal adjustment, which generates a series from the original data that eliminates the seasonal variation, permitting quarter-on-quarter comparisons. It has been used by the federal statistical agencies in the United States for many decades now, and the computation tools have been continuously upgraded. The European Union and Japan also report seasonally adjusted series.

For a good example of the practical utility of this technique look at the accompanying chart. We have plotted the quarterly rates of growth of GDP in the US on seasonally adjusted annual rates and on the ?comparable period last year? system that we use. Look at the dip in 1999-Q2, the rapid expansion in Q3 and Q4 and the slump in 2000-Q1. The way we measure (QoQ), we would not have spotted a thing. The same applies for most periods, and most recently in the slump of the second half of 2001 and the sharp recovery in Q1 of 2002. Good, responsive policy is predicated on quality information. There is much that we can readily improve upon, and easily.

Saumitra Chaudhuri is economic advisor to ICRA (Investment Information and Credit Rating Agency) and editor of Money and Finance, the ICRA bulletin