How To Improve Industrial Statistics

How to Improve India’s Industrial Statistics It is widely accepted that the quality of official industrial statistics has deteriorated. Why did this happen? This study argues that the administrative mechanism for recording the inception and closure of factories and firms and gathering their production and investment information on a regular and reliable basis has gradually decayed over the years – in response to industrial change and a variety of policy initiatives. This problem needs to be squarely addressed to make any tangible difference. R NAGARAJ

I

n the recent years there has been a growing concern about the deteriorating quality of official industrial statistics. As they are widely used to assess public policy, their reliability and accuracy are intensely debated in the press and parliament. Moreover, as India is increasingly an open economy, there is now a growing international demand for timely and accurate quantitative information on economic performance. After a brief reference to the known shortcomings of the Index of Industrial Production (IIP), and the Annual Survey of Industries (ASI), we examine the process of gathering the primary data for computing these statistics that probably represents the heart of the problem. Section I describes the existing arrangement for gathering data and computing the statistics, Section II looks into their limitations, and Section III suggest some remedies. The conclusions are reported in Section IV.

I Current Status IndexofIndustrialProduction The IIP is a base-year weighted (Laspeyres), quantity index of production of (i) mining, (ii) manufacturing and (iii) electricity – roughly accounting for 17 per cent of GDP. It largely mirrors the manufacturing sector performance as it forms 4/5th of the IIP’s weightage. The index is available monthly, disaggregated by 2-digit industry groups. The currently used IIP has 1993-94 as the base year, available with about six weeks lag. It is perhaps the most widely used leading indicator of short run aggregate economic performance.

966

The IIP is computed using the monthly production data supplied by factories/firms to their respective promotional/ regulatory authorities in the central government – called the ‘source agencies’. Fifteen such agencies supply information to the Central Statistical Organisation (CSO) that is responsible for computing and publishing the IIP. Industrial (Development and Regulation) Act, (IDR) 1951, the national law under which permission is granted to carry out production and investment in the country, mandates factories/firms to supply monthly production data. During the last five decades, the IIP’s base year has been revised five times, to take into account the changing output composition. The weighting diagram for the index is based on the gross value added, as estimated in the ASI.

Annual Survey of Industries Started in 1959, the ASI, as the name suggests, is a survey of production accounts of all factories registered under Section 2m of the Factories Act, 1948. Factories employing 10 or more workers using power (20 or more workers without using power) have to get registered under this law. However, it also includes (i) bidi and cigar establishments, (ii) electricity undertakings, (iii) tea and coffee plantations, and (iv) Warehouses, regardless of their employment. All such factories are mandated to submit their production accounts under the Collection of Statistics Act, 1953. While the field operations division (FOD) of the National Sample Survey Organisation (NSS) collects the data from the factories, the CSO is responsible for processing and publication of the ASI results.

The Factories Act is the cornerstone of the labour legislations in India, intended to protect workers’ safety, health and welfare. Though a national Act, state governments are responsible for administering it. Chief Inspectorate of Factories (CIF) in each state, an agency for administering the Act, maintains the live register of factories that forms the ‘frame’, or the ‘universe’ for conducting the ASI. The ASI consists of two parts – the census sector and the sample sector. Till recently, a census was conducted of all factories employing 50 or more workers using power (100 or more workers without using power). For the remaining – that is, 10 to 50 workers using power (20 to 100 workers without using power) – a 50 per cent sample was conducted every year. Until recently, publication lag of the ASI results was one of its widely perceived shortcomings. However, with the CSO’s modernisation, there seems to be a noticeable improvement. Further, some significant changes were introduced in 1997-98. One, the census sector is now redefined to include factories with 200 or more workers; the rest are covered on a sample survey – consisting of 1/3rd of the registered factories in each year. Two, electricity sector is deleted, as it is deemed a duplication of the information that is available with the Central Electricity Authority.

I Shortcomings IndexofIndustrialProduction – Although the IIP is available for 2-digit industry groups, as also for use-based industrial classification, the sampling

Economic and Political Weekly

March 9, 2002

Per Cent

Figure: Evasion of Factory Registration 60 58 56 54 52 50

58 53

1980

1990 Year

Factories not registered under the Factories Act

design is intended to capture aggregate production only, at one digit of National Industrial Classification (NIC) – that is mining, manufacturing and electricity. – Since the introduction of the base year 1980-81, the IIP is designed to include production of small-scale industries.1 But, in practice it is not so, as reliable and timely information from small manufacturing enterprises is not available on a continuing basis.2 – Although the weighting diagram of the current IIP is drawn on the basis of value added in the total manufacturing, the index is computed using data from the registered manufacturing sector only. – In the IIP’s latest revision, over 500 products were included in the item basket for computing the index, keeping in view the changing output composition. But more than 100 products had to be dropped as the concerned source agencies could not assure timely and regular supply of their production data. Therefore, in effect, the IIP’s latest revision has had little effect in capturing the diversification of output that has occurred since the 1980s. The above shortcomings are, however, relatively minor. For instance, design of the weighting diagram can be easily improved. But these are perhaps symptomatic of a larger problem that is discussed below. As mentioned earlier, the primary data for computing the IIP has all along been a by-product of the government’s regulatory and promotional functions. This, in itself, is not objectionable or unique to India as many countries – including the US and the UK – rely on a similar procedure. However, under the regime of output and investment controls that was broadly in place up to 1985 (or 1991), factories/firms, in principle, had to supply production and capacity data to the concerned official agency. Even to continue in the business; they had to get ‘carry on business’ (COB) licence. Moreover, in principle, firms had an incentive to submit the data, however strategically, to get licence for expansion/

Economic and Political Weekly

diversification, or probably even to prevent potential competition! However, after the deregulation, the existing firms do not require COB licence. New firms only have to file ‘Industrial Entrepreneurs Memorandum’ (IEM) by paying a one-time registration fee. Part I of the memorandum is to be filed to express intention to start a business, and the Part II at the commencement of production. Apparently even this is not mandatory. Therefore, reportedly, while Part I of the IED are filed in large numbers, very few care to file Part II, as it is of little value for conducting business. These changes in policy – and legal enforcement – have considerable implications for the official statistical system. One, after the reforms, the existing firms have no incentive (or a threat of non-compliance) to supply information to the source agencies. Second, as new factories/firms may not even be registered with any regulatory/promotional agency, there is no question of them filing production information. Therefore, the universe of factories/firms available with the source agencies for data collection has become increasingly incomplete. The investment boom in registered manufacturing and acceleration in incorporation of new companies in the 1990s perhaps accentuated the problem, suggesting large-scale entry of new firms [Nagaraj 2000]. In principle, the IDR Act can be invoked for data collection, but it has seldom been done. During the regime of regulation, there was perhaps little need for it, as the licensing requirement was deemed adequate for the purpose. But in the 1990s, in the wake of the official and popular zeal against the ‘inspector raj’, the inability and/or unwillingness of the source agencies to enforce the statutory requirements is understandable. The problem seems to be acute with the Department of Industrial Policy and Planning (DIPP), ministry of industry and commerce that is responsible for supplying

March 9, 2002

data for 62 per cent of the products in the IIP’s items basket, accounting for 52 per cent of its weightage. This source agency represents the statistical section of the erstwhile Director General of Technical Development (DGTD). As per the IDR Act, DIPP represents the ‘residual’ industries outside the purview of most other source agencies that cover relatively homogenous product groups. As the manufacturing sector’s production structure has further diversified after the deregulation, understandably, firms and industries coming under DIPP would have gone up significantly. But, it has little information about the newer factories/firms, as the provisions of IDR Act for registration and submission of production information have not been enforced.3 The results of the deterioration in datagathering mechanism are evident in the quality of IIP. As Nagaraj (1999) showed the correlation coefficient between IIP and value added in manufacturing (as per National Accounts Statistics) has deteriorated in the period 1980 and 1995, compared to that during 1970-85. Singhi and Mishra (1997) had also reached broadly similar conclusion.

Annual Survey of Industries ASI is a better data set, relatively speaking [Pradhan and Saluja 1998]. For that reason, it forms the basis for preparing the IIP’s weighting diagram. There are problems with it, however. Till recently, especially in the 1980s, a growing problem of non-response was widely reported. CSO seems to have addressed this problem. But perhaps the bigger problem is the incompleteness of the frame that arises on two counts: (1) Reportedly, the NSS field staff finds the inclusion of closed factories a major problem, especially in drawing the sample survey of factories employing less than 50 workers. However, CIF does not delete them from their list, since it could jeopardise the interests of various stakeholders in the closed factories – like the workers and the creditors. (2) But perhaps the more significant lacuna lies with the large-scale and growing evasion of registration under the Factories Act. A comparison of factories/establishments in manufacturing employing 10 or more workers was revealing, as shown in the accompanying figure. In 1980, 53 per cent of factories employing 10 or more workers did not register

967

themselves under the Act. This proportion increased to 58 per cent in 1990 [Nagaraj 1999]. In other words, Since the 1980s a majority of factories have evaded the factory registration; and, their proportion went up by 5 percentage points in the 1980s. After the deregulation – when the ‘inspector raj’ was widely perceived as the bane of administration, it is reasonable to expect the extent of evasion to have gone up steeply.4 Admittedly, the bulk of these factories would be in the employment range of 10 -20 workers, and value added per worker in this size class is likely to be modest. But considering the extent and growth of nonregistration, and faster than average employment growth in this size class, exclusion of these factories under the Factories Act has a serious implication not only for output estimation, but also for the analyses of industrial structure, its composition (changes in them over time), and for policy planning.5

III What Should Be Done? Before discussing the possible solutions to the problem, it is perhaps useful to appreciate the nature of the activity that we are dealing with, and the constraints that we face. Economic statistics – quantitative information about the economy in general – is a classic public good whose consumption is non-excludable and nonrival in nature. Therefore, left to the market

forces, their supply is likely to be less than the demand for it, and there is likely to be under investment in this activity. Therefore, the state has an important role to play in preparing economic statistics. Moreover, in a deepening democracy, economic and social statistics can remain credible only if a public agency is responsible for them that is accountable to all the stakeholders in the society. Reportedly, these are the reasons why, even in the developed market economies, such activities are primarily in the public domain. However, in principle, one could consider the possibility of cooperation between public and private agencies to device an institutional arrangement where the industry associations and chambers of commerce would collaborate with the official agencies to create the industrial statistics. This, on the face of it, seems feasible as many industry associations in India, in fact, collect detailed production information from their members – as it widely happens in countries such as the US. However, on a closer examination in our context, we found this proposition infeasible. Hence, we are inclined to believe that the industrial statistics can be satisfactorily produced by an agency of the state alone.

IndexofIndustrialProduction As we have discussed above, there are principally two reasons for lack of suitable

Box 1: Alternative IIP – An Experiment In response to the growing unreliability of the current IIP, the Arun Ghosh Committee set up in 1998 had suggested computing an alternative IIP (AIIP) by using monthly production data collected from factories employing 200 and more workers. If the alternative was found successful, the committee recommended replacing the IIP with AIIP. We now have data for 17 months to make a preliminary assessment of this experiment. Initially NSS collected data for the AIIP by making personal visits to the factories. The experiment was successful in getting response rate of up to 80 per cent of factories in the second month. However, reportedly, this method could not be sustained with the FOD’s existing resources. As additional resources did not come forth as anticipated, personal visits were given up. Then the response rate fell from 62 per cent in the 4th month to 27 per cent eight months later. The average response rate for 12 months during which data were collected by FOD field staff was 46 per cent, which is lower than the response rate of the existing source agencies. Therefore, the data collection by personal visits does not seem any better than that collected by the source agencies. FOD’s attempt to persuade firms to submit information by fax/email/post reportedly failed.

968

monthly production data to compute the index. One, the administrative framework that formed the basis for data collection has deteriorated, especially after the reforms. Two, the frame of factories/firms has become increasingly incomplete, as the registration with the concerned agency is not mandatory, or required for conducting business. This can be remedied in two ways: (1) Revamp the existing set-up, or (2) Create an alternative data collection mechanism. Although the licensing role of most of the source agencies has been eliminated, reportedly many of them have been redesignated as promotional agencies. In an increasingly open economy, firms in many industries, like steel, do seem to find it useful to keep in close touch with the concerned official agency to protect and promote their interest, and to guard against external threat to their businesses. If such an assessment is correct, then the source agencies can still perform the datagathering role, provided the registration of firms and the submission of data are made mandatory. Alternatively, the CSO/NSS could be empowered to collect the monthly production statistics under the Collection of Statistics Act. This would entail (i) extending the scope of the Act, and (ii) expanding the field operations division (FOD) of National Sample Survey (NSS) by augmenting its resources. Arguably, since FOD staff keeps track of factories for the collection of ASI data, gathering monthly statistics would, in principle, not pose a major problem. However, the recent experiment with the ‘alternative IIP’ does not give us confidence about this option (Box 1).

An Assessment of Options Before assessing the merits of the suggested alternatives, it is perhaps important to stress the fact that irrespective of the choice, we need to know the universe of factories/firms from which to draw a sample to collect production data. If the CSO/NSS is to collect the data, then it has no option but to use the ASI frame, which, as we have shown above, is growingly unrepresentative. If, on the other and, the source agencies are to collect the data, then they will have to make considerable effort to update the frame, without which there is little hope of improving the representativeness of the sample of firms to draw the data

Economic and Political Weekly

March 9, 2002

from. On balance, we are inclined to favour the first option of strengthening the exiting source agencies, for the following reasons: – With all their limitations, the existing source agencies nevertheless have a better appreciation of the industry. They seem to have, over the years, acquired some industry specific knowledge as they often interact with the industry associations and the bigger firms on a regular basis. Firms apparently seek the source agencies’ assistance to put forth their viewpoint to the concerned government departments. Smaller firms, in particular, seem to seek source agencies’ assistance in trying to deal with the threats and opportunities posed by the external markets. Therefore, it seems to make sense to use the source agencies to gather the monthly production data. – In the 1990s, after the abolition of licensing for most industries, Part I of Industrial Entrepreneurship Memorandum that have been filed with DIPP can be used to trace working factories/firms to rebuild the DIPP frame. Similarly, the list of inward foreign direct investment (FDI) approvals could be used to rebuild a frame of foreign owned/operated factories/firms in India. – Admittedly, the organisational infrastructure of many of the source agencies seems weak. There is a need for modernisation of their resources, including perhaps upgrading of their statistical capabilities. – Our suggestion does not presume that firms will be forthcoming to voluntarily supply the production data. Therefore, we have to devise a system of incentives and penalties, to ensure compliance – a practice widely used in all market economies. – The best conceivable penalty is the invoking the provision of the IDR Act to mandate (i) registration of firms with the source agencies and (ii) supply of production data. Each firm should be assigned a unique registration number (more about it later). Use of this number should be made mandatory for all their transactions with the government agencies, banks, financial institutions and the stock markets. – Firms could be given suitable incentive for providing data in terms of promotional assistance.

Annual Survey of Industries As discussed above, the ASI’s main problem is gross and increasingly incomplete coverage of factories employing 10 or more workers. Therefore, there is, in our

Economic and Political Weekly

view, an undisputed need for getting the factory frame right. How to do it? As this Act is intended to protect workers’ interest, there are, in principle, stringent regulations regarding the safety, health and welfare of workers; violation of the Act is a criminal offence. As often is the case with such laws in developing economies, they tend to be inspirational in nature, with poor implementation due to numerous (perhaps deliberately built-in) loopholes. From the viewpoint of an entrepreneur, registration under the Factories Act is an avoidable cost, considering the additional work that it involves. Apparently, its evasion is easy, as the practical costs of doing so are marginal. If this assessment is correct, then the only way to get a more complete frame of factories is to simplify the Act. However, this may not be a practical suggestion in the medium term. But, as a first step in this direction, the Act can be modified to introduce a clause of factories deemed closed for statistical purposes only. In other words, factories that ceased to produce output for a certain number of years continuously (say three years) can be deemed closed for statistical purposes. Such a modification will eliminate the problem that the NSS faces with respect to the CIF frame. Another step is to conduct a census of factories. As it is not practical to conduct a census of only those employing 10 or more workers in manufacturing industries, such a census should use the frame provided by economic census or the house listing operations of the decennial census. At least, a one-time census would go a long way in building a frame of manufacturing factories/firms. Admittedly, such a census would soon be outdated, given sizeable entry and exit of factories, especially in the lower end of the size distribution. But considering our other suggestions of capturing registration of factories/firms and plant closures discussed above, we believe a census can nevertheless be of considerable value.

Uniform Numbering System Currently, each administrative department, official data collecting/processing agency assigns a unique number for each factory/firm for identification. Often a large proportion of information in these numbering systems is common, like the state code, district code, industry code, and so

March 9, 2002

on. But the existing numbering systems are not useable across agencies, leading to considerable duplication of work and delay in data processing and retrieval. This also makes it impossible to verify existence and functioning of a factory/firm. In principle, a unique business numbering system that is mandatory all factories/ companies, and for use in all official transactions can overcome this problem. This is akin to the social security number system used in the developed countries for their citizens. The Department of Company Affairs has recently initiated a computerised, 21-digit code for registered companies, called the Corporate Identity Number (CIN). This can, in principle, be extended to make it a general business identification number (BIN). Admittedly, use of such a big alphanumeric number has practical difficulties. For instance copying the number, in a semi-literate country like ours can lead to mistakes. Mandatory bar coding of the numbers could eliminate many of these practical problems. Moreover, with the recent advancement in technology, bar coding can facilitate instant transferring of information and communication.

IV Conclusions Economic statistics are usually a byproduct of public administration, and the enforcement of the legal system of a nation state. Even independent national statistical authorities largely rely on the administrative and legal framework to the draw the universe of factories, firms and households. In India, since the early 1950s, the Industrial (Development and Regulation) Act, 1951 (IDR Act, 1951), and the Factories Act, 1948 provided such a framework for the legal, regulatory and promotional activities for industrialisation. A manufacturing firm had to register under the IDR Act if belonged to a scheduled industry, and also under the Factories Act if it employed 10 or more workers using power. However, firms in the traditional industries and modern small enterprises with capital investment less than Rs 7.5 lakh were generally outside the purview of the IDR Act, but had to register with specialised promotional agencies meant for their development, if they sought official assistance. Expectedly, these two acts provided the framework for the collection of industrial statistics. Monthly production data for

969

computing the Index of Industrial Production (IIP) was collected by various agencies responsible for administering the IDR Act. Such information was collected for issuing licences for production and investment. Although the submission of statistics was mandatory under the law, it was seldom invoked as the licensing provision reportedly ensured compliance, however strategically. Similarly, the live register provided the universe of factories that were to be covered under the Annual Survey of Industries (ASI); submission of statistics was enforced under the Collection of Statistics Act, 1953. This ‘idealised’ set-up underwent considerable changes over the decades, as the regulations grew complicated and their legal enforcement got diluted. To illustrate, businesses found it easy and profitable to evade the factory registration. Even those registered under-reported employment, as larger factories attracted more and stringent provisions of the Act. Expansion of the coverage of the modern small-scale sector and exemptions offered diluted the scope and rigour of the IDR Act. The situation perhaps deteriorated with the deregulation initiated since the mid1980s that accelerated in the 1990s, as the mandatory registration of factories/firms were overlooked in the liberal regime. Therefore, the existing firms had no reason to supply production information to the concerned source agencies, and even the registration of newer firms is rendered voluntary. Hence, we now do not know the universe of factories/firms and their distribution across industries, regions and by size with any reasonable accuracy. Further, timely, representative and reliable data on industrial production has become increasingly scarce. If this diagnosis is correct, then the solution lies in (1) Obtaining the universe of factories/ firms, and (2) Devising a mechanism to enforce a legal provision of their registration, and to collect their production statistics. How to go about it? As the Factories Act is the basic building block for industrial statistics, measures must be initiated to simplify the Act, at least for statistical purposes. As a first step, the Act needs to be amended to delete – for statistical purposes – closed factories that ceased production for specified number of years continuously (say three years). Census of working factories needs to be conducted.

970

As it is not practical to consider census of factories employing 10 or more workers only in manufacturing, such a census could be combined with the population census’s house listing operations or with the economic census. The frame that is provided by such a census would not only be of value for the ASI, but also for the IIP. The universe of factories/firms with source agencies has become increasingly incomplete with the liberalisation. This problem is perhaps acute with respect to DIPP – the statistical wing of the erstwhile DGTD, accounting for over one-half of manufacturing output. How to ensure factories/firms submit the production data? This problem is mainly affects the IIP. To ensure firms submit monthly production data, there are two options. One, entrust the task to the CSO/ NSS with additional resources, and to empower it under the Collection of Statistics Act. Perhaps, a better solution is to revamp and modernise the existing source agencies. They should invoke the IDR Act to ensure submission of the data. Further, a unique, business identification number (BIN) needs to be devised for all business enterprises, including all factories and companies. Use of this numbering system should be made mandatory in all their dealings with government agencies and financial entities likes banks, stock markets and financial institutions. Once such a computerised numbering system is place, and it is shared by all the concerned agencies, mandated bar coding would go a long way in hastening retrieval of the computerised database of business enterprises. -29

Notes [This is a revised version of a study conducted for the National Statistical Commission. I thank to C Rangarajan, the commission’s chairman, for his keen interest in this study, and for his consent to its publication. However, it represents my personal views. I am grateful to a large cross section of concerned officials in the public and private sectors who spared their time for discussions. I am indebted to R B Burman, Bernard D’Mello, K L Krishna, M S Maulik, M Neelakantan, V R Panchamukhi, Kirit Parikh, Nilachal Ray, Y V Reddy, Vaskar Saha, N S Sastry, S L Shetty, M C Singhi, C B Singh and M H Suryanarayana for their valuable responses and suggestions.] 1 Modern small-scale industries (SSI) and traditional industries like Khadi and village industries, coir, etc (6 in all) are outside the purview of the IDR Act. SSIs are defined as industries with investment in machinery

2

3

4

5

equipment below a certain financial limit. This limit has been steadily enhanced over the years; the current one is Rs 1 crore. Development Commissioner, Small Scale Industries (DCSSI) is responsible for supply of production data for 18 items of this sector. However, the source agency has failed to so. This is evident from the official press release issued to notify the introduction of the new series of IIP. It said, “In the absence of regular production data from the unorganised sector, the item basket has been identified on the basis of data from the registered sector only” (emphasis added). Ministry of Industry (1995) has candidly admitted the problem. To quote its study, “… all establishments falling within the scope of IDR Act are required to file a monthly production return to the department of Industrial Development… However, since there is hardly any evidence of prosecution having taken place for the defaulting units, supply of information by establishments even in respect of items that are included in the index [IIP] may not have been complete or timely. Since the elimination of licensing for most products and winding up of the erstwhile DGTD, the quality of data may become even further deteriorate, despite the fact that the format of the return has been substantially simplified from its earlier complicated version (which itself must have been a major contributor to the poor quality of data in the past)” (p 11) This seems to be reflected in the fact that despite a boom in fixed investment in registered manufacturing and registration of new companies, growth in number of factories in the 1990s has been relatively modest. One may dismiss this concern as undue focus on accuracy. But there are many reasons why I believe this is a serious matter. Value added by the non-included units gets seriously underestimated as the Economic Census data for estimation of value added are, reportedly, seriously flawed. Besides, the lack of this data distorts the structure of production, by size class of factories/ their industrial/locational distribution. As much of our macroeconomic, sectoral and inter-industry analyses are based on close study of ASI data our understanding of structure and changes in the economy gets distorted.

References Ministry of Industry (1995): ‘Industrial Production Statistics’, paper presented at the seminar of Indian Association for Research in National Income and Wealth, March, New Delhi. Nagaraj, R (1999): ‘How Good Are India’s Industrial Statistics? An Exploratory Note’, Economic and Political Weekly, Vol 34, No 6, February 6-12. – (2000): ‘Organised Manufacturing Employment’, Economic and Political Weekly, Vol 35, No 38, September 16. Pradhan, Basanta K and M R Saluja (1998): ‘Industrial Statistics in India: Sources, Limitations and Gaps’, Economic and Political Weekly, Vol 33, No 21, May 23. Singhi, M C and J P Mishra (1997): ‘Industrial Production Statistics’, Ministry of Industry, Office of the Economic Advisor, Paper No 20, New Delhi.

Economic and Political Weekly

March 9, 2002