The official statistical system has three major indicators that are used to report on industrial activities: the Index of Industrial Production (IIP), Annual Survey of Industries (ASI) and manufacturing-related indicators within National Accounts statistics.

Also read: Centre working on multi-pronged approaches to reform its statistical systems

These regular databases are supplemented by surveys of the National Sample Survey covering the labour force and employment, and the unincorporated sector. These operate on an annual basis and the National Statistical Office has indicated that they will be strengthened to release monthly estimates.

The National Accounts and ASI are used for major analytics on the industrial sector uses, so it is worth asking how good these are from the viewpoint of assessing industrial performance.

The ASI is the main data source for analysts doing a range of industrial studies. The ASI schedule has a wealth of information for this purpose. Of late, the Central Statistical Office has not only improved the timeliness of releases, but also made anonymized unit-level data available. Despite this, there are problems associated with the sampling frame and design of the survey.

The sampling frame of the survey is based on registrations with the chief inspector of factories under Sections 2m(i) and 2m(ii) of the Factories Act, 1948. This requires industrial establishments employing 10 or more workers using power or 20 or more workers without power to register (recently some states have loosened this requirement to 20 and 40 respectively).

Many studies over the years have shown that this coverage suffers from many loopholes. For example, the increased use of outsourcing, information technology and computer-aided manufacturing have reduced regular employment in factories, which implies that many high value-added establishments are excluded from ASI coverage.

These gaps are sometimes not filled even by NSS surveys on unincorporated enterprises, as these establishments may belong to registered companies.

A second problem is that the ASI is an establishment-based approach, requiring a NSS surveyor to visit the premises to collect information based on records maintained on the factory premises. Resource limitations in any such exercise constrain the NSS, and limits visits to around 60,000-70,000 establishments in a year.

This limit is managed by splitting the frame into census and sample sectors. This is primarily based on the number of workers in the establishment, but also on the level of industrialization in the state and/or district.

This approach makes the selection dynamic, and establishments can be classified as belonging to 'census' or 'sample' on the basis of changes in the underlying parameters. Such changes in coverage add volatility to estimates, leading to cautionary notes being regularly observed in ASI reports.

Also read: Data users need clarity on numbers from the statistics ministry

Despite these design limitations, the ASI used to serve as the primary source of estimating organized sector manufacturing value added in the National Accounts till 2011-12. In 2011-12, it was replaced by use of the MCA database as the primary source of measurement for organized activities.

The reasons for this change included the problems alluded to above, but also because of inadequate coverage of large multi-establishment and multi-state enterprises with activities in both manufacturing and services. In these cases, the value added in some of their establishments that were either below the ASI threshold or in non-manufacturing domains could get excluded from National Accounts coverage.

The MCA database comprehensively covers all active companies. The approach taken was to allocate a company's entire value addition to its major sector of operation.

However, this creates a problem for accurate sectoral assessment because value added depicted in the manufacturing sector will include some element of service activities and will conversely exclude some manufacturing activity being carried out in primarily service enterprises.

The rise of Industry 4.0 will make such misallocation even larger. A consequence of the 2011-12 revision was that while it made the overall estimate of gross value added (GVA) more comprehensive, it had an unfortunate side-effect of making sectoral estimates more imprecise.

In simple terms, the ASI's coverage of value added is incomplete, but precise in its focus on manufacturing. By contrast, the MCA's coverage of value added is complete, but its allocation of value added to manufacturing is imprecise.

These problems can be fixed by a dual strategy. GST registration, by design, is an establishment-based process, and it has a clear link between the enterprise and the establishment (as the enterprise identifier is embedded in GST registration).

The GST register, with a registration threshold based on turnover, is simpler and more complete. Using this register to conduct ASI surveys will eliminate many of the gaps noted above and will permit better coverage of multi-state and multi-establishment enterprises.

A more robust possibility is to use detailed GST filings, which will offer a complete picture of production and input usage. Admittedly, GST data will not have employment information. However, this can be obtained by separate focused surveys of these entities.

Also read: India's statistics system has been a mess. The PMO is looking in.

In addition, gaps in the MCA data can be filled by a follow-up survey under the Collection of Statistics Act (2008), which asks enterprises to report establishment-based value-added information. If this survey is directed at large listed companies, it is likely that the results will be more complete, as the factory level records of these entities have less information than what is available in their management information systems.

While the above suggestions require major restructuring of the statistical office, a start can be made by shifting the sampling frame to the GST register and using GST data to produce reports (monthly/quarterly/annual) on sectoral value addition. This can be done by simple query operations on tax filings and does not compromise any principle of confidentiality.

The author is a former chief statistician of India.