India’s IIP is being revised but fixing the accuracy of this indicator isn’t a patchwork job

India’s current IIP series is unrepresentative of underlying production trends, as widely admitted. There are two reasons for this, contends a discussion paper recently released by the NSO: (i) non-reporting of data by factories that are closed; and (ii) lack of data when the selected factories change their production lines. These constitute a significant share of the selected factories. The discussion paper suggests ways to overcome these problems.

To quote the paper: “Presently, the compilation of the Index of Industrial Production (IIP) relies on a fixed panel of factories selected based on the base year to represent industrial activity across various sectors. Over time, certain factories in the sample may experience operational disruptions, such as: Permanent closure, Change in the production line …. Generally, the base year of the IIP is revised periodically; however, when there is a significant gap between revisions, such issues tend to arise.... e.g. over time, in the current series the weight of the closed factory comes to about 8.9% of the index. ….To address these challenges, substitution of factories has become necessary… [by a set] that produces the same item/item group, ensuring continuity in the data series." (Page 2)

Is this diagnosis of the IIP’s estimation problems correct and complete? Would the suggested solution help capture India’s underlying production trends truthfully? Is the non-reporting of output/changes in the selected factories’ production lines the principal or binding constraint? We examine these questions by using available official reports and academic analyses.

To appreciate the extent of the shortcomings, the graph alongside reports the annual growth rates of production by the IIP and the Annual Survey of Industries’ (ASI) gross value added (GVA) at constant prices for 2012-13 to 2023-24 (the best available production statistics ).

The two series move broadly in the same direction, with the IIP showing much lower annual growth rates and lower yearly fluctuations. The average annual growth rates over these 11 years are starkly different: 5.9% by the ASI and a mere 3.3% by the IIP. Evidently, the IIP grossly under-reports industrial growth.

But would the 8.9% of selected factories flagged by the discussion paper explain such a lower IIP growth rate? No. There are probably bigger problems at hand.

The IIP is a ‘leading’ indicator of short-term economic performance, available monthly with a 6-week lag. It represents physical production of mining, manufacturing and electricity generation, which accounted for 22% of GDP in 2023-24.

Manufacturing accounts for 78% of the IIP’s general index in the current series. Its efficacy in describing underlying production rests on (i) the weights assigned to individual items in the base-year (the weighting diagram of the index), (ii) the size and composition of the sample of factories selected for reporting output, and (iii) the timeliness and accuracy of production data used for the estimation.

The items (i) and (ii) are statistical issues, addressed by periodic re-basing of the index and a sample selection of factories from the census of manufacturing that follows the principles of sampling. However, access to timely and accurate production data depends on our administrative and legal setup. In other words, the legal provisions or administrative procedures under which factories/firms furnish production data to ‘source agencies’ are of critical significance.

Until 1991, under India’s industrial licensing regime, factories/firms had to submit monthly production data to official agencies for licences and renewals to run their businesses or expand capacity. Enterprises had an incentive to comply with administrative demands. After economic reforms, while the reporting mandate stayed in place, they could get away without meeting it.

Over time, as production reporting dwindled, primary data available for estimating the index became scarce and infrequent. Efforts at using information technology to get the data did not go far.

The Collection of Statistics Act of 2008 made data submission a legal must, though it is seldom invoked for preparing the IIP. This law seems to function better for the ASI, for which the NSO field staff visit factories to collect data.

The problem was deliberated upon at length by the Report of the Working Group for Development of Methodology for Compilation of the All-India Index of Industrial Production with Base Year 2009-10/ 2011-12, produced in April 2014 by a group chaired by Soumitra Chaudhuri for the ministry of statistics and programme implementation.

The report said, “…most of the Government of India organisations… which supply monthly IIP data suffer from problems of non-response in primary data collection, which may be due to lack of legislative or regulatory control over the production units. In such cases, they are left with no option but to estimate the production of the non-responding units. In the absence of trained statistical personnel and clear-cut guidelines, most of the agencies are unable to carry out statistically sound procedures of estimation—the data quality being the obvious casualty as a result." (Page 10). Further, the report said: “The Working Group is of the view that a significant step towards improving the data collection system would be the implementation and strict enforcement of the Collection of Statistics Act, 2008." (Page 54).

In sum, we need a candid analysis of the root of the IIP problem. It seems to lie in inadequate production data from factories and firms. The NSO discussion paper, though, looks for solutions in the professional domain of statistical methodology.

This seems like a proverbial case of looking for one’s keys under a lamp, not where they were lost.

The author is a former professor, Indira Gandhi Institute of Development Research (IGIDR), Mumbai.