The recent revision of industrial production (IIP) numbers revives the debate about the quality of quick estimates of macroeconomic parameters. More importantly, it brings the final revision of these estimates and its implications for policymaking back into focus. Recently, the Reserve Bank of India (RBI) too has highlighted the implications of frequent and large data revisions on policymaking. Data revisions take place due to two reasons. First, is the periodic revision of the base every 10 years to better reflect the changing economic structure. The second, and more worrying, aspect is the data revision due to the errors and/or delays in accurately capturing the economic activity in India.

Let us take the example of industrial production data. The IIP data for February 2012 at 4.1% should have lifted market sentiments. Growth during the month was above the 3.2% average growth in the previous three months. The data also showed positive growth in the mining sector, which had been contracting for the preceding eight months. Instead, what caught attention was the stark downward revision in the January 2012 IIP estimate to 1.1% from 6.8%, released earlier. At the time of its release, the close to 7% IIP growth for January had cheered the market participants. Sensex gained some lost ground due to the release of strong IIP data, while proponents of a rate cut in the the March 15 monetary policy review temporarily took a back seat. The latest revision in IIP, therefore, has once again raised questions over the reliability of industrial production estimates.

In June 2011, the Central Statistical Organisation (CSO) released the IIP with a new base of 2004-05, replacing the 1993-94 series, which had been in use since May 1998. The revision was pivotal as it aligned the IIP series with the same base as GDP and WPI, allowing for a more meaningful comparison with these macroeconomic aggregates. The new series also had wider coverage as its composition was more relevant in terms of the current economic structure and assigned weights to sectors based upon their contribution to GDP.

As per the new IIP series, industrial growth during 2007-08 was almost double at 15.8% as compared to the previous estimates of 7.8% based on the old series. So, the old series significantly underestimated industrial production. This slowdown in industrial activity suggested that RBI?s monetary policy tightening stance, which it had adopted in 2006-07, had begun to show effect. The assumption that real GDP growth was slowing to 8.5% from 9.6% in the previous year, as suggested by the then available IIP data, had in part influenced the central bank?s decision to pause rate hikes during 2007-08. Subsequent revision in real GDP growth to over 9% during the year, in addition to rising inflation pressures, however, triggered strong monetary policy actions in 2008-09.

A similar episode emerged with the upward revision of industrial growth estimates in 2010-11 and the downward revision in 2009-10. The new series revised industrial production growth in 2010-11 to 8.3% from 7.8% earlier. It also halved the growth in 2009-10 to 5.3% from 10.5% as per the old series. There was a marked divergence between the new and old IIP series, not only with regards to estimates, but also in the identification of underlying trends in the economy. This analysis has been detailed further in a previous note1 released in July 2011.

Since the release of the new IIP series, in addition to the divergence from the old series, the high volatility in the monthly numbers has been a cause of concern. Within the IIP, the capital goods production index is extremely volatile. This index has a relatively low weight, of 8.8%, in overall IIP but has the highest volatility. The lack of stability in the capital goods index is a cause for concern. Capital goods data under IIP is the closest proxy of investment activity in the economy given its higher frequency of release. Since GDP data is available with a lag of nearly one quarter, analysts and policymakers look at the IIP capital goods data to gauge the investment scenario. Hence, high volatility in monthly numbers hinders accurate assessment of trends in investment.

Data revisions are not unique to India. Across countries, even advanced economies, most estimates of macroeconomic parameters undergo revision. Since policymakers don?t have advance information about the possibility and extent of future data revisions, they must decide their policy stance on the basis of information available at that time. In the Indian context, accurate estimation of industrial production growth is of utmost importance. While final GDP numbers are based on output data from the Annual Survey of Industries (ASI), these figures are available only with a considerable time lag in India. Since IIP data is more recent, and there is a noticeable similarity between ASI and IIP manufacturing data, advance and quick estimates for GDP are drawn from IIP data. Since growth and inflation are the two important parameters that govern policy changes, accuracy in estimating these is crucial. If the revisions are very stark, then initial data would have misled policymakers and analysts in hindsight.

The launch of a new IIP series on a revised base, increased volatility in the numbers, and revision in the monthly data has caused quite a stir and questioned the reliability of the index. The solution lies in quick revisions of the base with which economic data is estimated and strengthening the statistical machinery to capture data accurately. Meanwhile, policymakers and analysts would do well to adopt a cautious approach to interpreting the IIP numbers, as high volatility and sharp revisions in these can mislead decisionmaking.

Policymaking in light of data revisions, Dharmakirti Joshi and Dipti Saletore

The author is economist, CRISIL. Views are personal