How Good Are Industrial Statistics

  • November 2019
  • PDF

This document was uploaded by user and they confirmed that they have the permission to share it. If you are author or own the copyright of this book, please report to us by using this DMCA report form. Report DMCA


Overview

Download & View How Good Are Industrial Statistics as PDF for free.

More details

  • Words: 5,805
  • Pages: 6
How Good Are India’s Industrial Statistics? An Exploratory Note R Nagaraj

There is a growing perception of a steady deterioration of the quality of India’s industrial statistics. Is this perception justified? To find out, this study examines the quality of the Index of Industrial Production, and some aspects of the Annual Survey of Industries, and the National Accounts Statistics. The study also examines if (a) the popularly used financial indicators really reflect the underlying investment trends, and (b) the expected association between electricity consumption and industrial output holds. Though exploratory, the findings reported seem to support the growing perception. ACCURATE and up to date industrial statistics are essential for policy, be it public or corporate, in an era of economic ‘planning’ or ‘reforms’. There is a growing perception that the quality of India’s industrial statistics has deteriorated over the years. This exploratory note seeks to find out if such a perception has any basis, with respect to some of the widely used industrial statistics. Section I, examines the quality of the Index of Industrial Production (IIP) – the most widely used leading output indicator. Development finance institutions’ disbursements of long-term credit, and mobilisation of capital in the primary stock market are widely used to forecast corporate investment activity. How useful these are to predict domestic fixed capital formation is examined in Section II. In a modern industrial economy, there is expected to be a close technical relationship between electricity consumption and manufacturing output. Does such a relationship hold in the Indian context, we find out in Section III. With the rapid growth of the unorganised manufacturing, there is a widespread belief that the value added in this sector is significantly underestimated. Section IV provides some indication of this tendency. Section V discusses some evidence of growing problems with the Annual Survey of Industries (ASI). Section VI concludes by summarising the main findings of the study. The questions asked are, how reliable are these leading (and lagging) indicators of manufacturing output and investment? Do they accurately and consistently reflect the underlying trends, given that the production and the organisational structures are becoming increasingly complex?

I Index of Industrial Production Index of industrial production (IIP) – available monthly, with the least time lag – is one of the most widely used leading indicators of industrial production. National Accounts Statistics (NAS) contains annual value added and capital 350

formation estimates, with over one-year lag, separately for the registered and unregistered manufacturing. Disaggregated value added estimates for two-digit industry groups are available with over two-year time lag. The NAS is the only source of estimates for the unregistered manufacturing value added and investment. The ASI Summary Results of the Factory Sector that provide the disaggragated data – at three-digit level and by states – are available with a lag of at least three years.1 Manufacturing sector constitutes over four-fifths of the IIP’s weightage, the remaining being mining and electricity sectors. The index is available for 18 twodigit industry groups; and for five usebased, three input-based and two sectorbased categories [RBI 1986]. Source of the primary data for estimating the index is voluntary reporting of monthly output by firms with equipment investment of over Rs 20 lakh in 1980. However, since in some industries small-scale sector dominates, they are also reportedly included in the index. Last year, after a gap of over a decade, a revised IIP was introduced with 1993-94 as the base year. Reportedly, the number of items included in the 1993-94 series is substantially larger, and it is intended to include even more items from the small sector as and when data become available.2 Does the revised index really represent an improvement? In other words, is it better at reflecting the underlying production trends? We contend that it probably is not, for the following reasons. Periodic revision of any index numbers is desirable to account for the changes in the composition of the basket of goods that they represent. During the 13 years since the last revision, the industrial output has grown annually at over 8 per cent, and with considerable changes in its composition. Therefore, the IIP’s revision is welcome, to the extent the new index better captures the changes in the output composition. In fact, this has been a routine matter with the official agencies as the

index has been revised five times since 1950, roughly once a decade. However, the other problem remains. The inadequate and poor quality of the primary production data used for estimating the index is perhaps far more significant. Unfortunately, the revision does little to correct it. Reportedly, 18 official agencies supply the primary data for estimating the index, though most important of them all is the Department of Industrial Policy and Promotion (earlier DGTD) that provides data on the manufacturing sector. Development Commissioner, Small Scale Industries (DCSSI) is reportedly responsible for supplying data for 18 items of this sector. However, this agency seems to be unable to do so. To quote the press release issued to notify the new IIP, “In the absence of regular monthly production data from the unorganised sector, the item basket has been identified on the basis of data from the registered sector only. Further, the source agency (DCSSI) could not line up the production data for the items of the revised series” (p 5, emphasis added).3 Evidently, the index does not capture the unregistered manufacturing at all – contrary to the official claim and its endorsement by many commentators [Pradhan and Saluja 1998b]. On the face of it, there are reasons to believe that the quality of the primary data has deteriorated over the decades. In a regime of industrial licensing, firms conceivably had an interest in voluntarily reporting their output; and the official agency perhaps had some administrative powers to ensure compliance. In other words, since the data generation process was a by-product of the regulatory regime, the index was perhaps more representative of the underlying production trends. However, since the mid-1980s – and especially since 1991 – with a steady decline and deregulation of output and investment controls, firms have little incentive to report their output to the official agency.4 Moreover, the officials have little leverage to enforce any rule in

Economic and Political Weekly

February 6, 1999

this regard. So, it is likely that nonreporting has gone up, and the index could have become increasingly unrepresentative. To test this proposition, we compute simple correlation coefficient of annual growth rates of the IIP and the NAS series. This assumes that the NAS series – that is, in turn, based on the ASI data (except for the most recent two years) – is a more accurate representation of the underlying production trends. First, the correlation coefficients are estimated between the growth rates of the IIP and (i) the registered and (ii) the total manufacturing.5 These are done for two sets of overlapping time series data: (i) 1970-71 to 1984-85, and (ii) 1980-81 to 1995-96, corresponding to IIP with base year 1970-71, and IIP with base year 1980-81, respectively. Table 1 shows that for Period-I, the IIP growth rate is statistically significantly correlated with both the registered and total manufacturing growth rates. However, for Period-II, the correlation coefficient between the IIP and the registered manufacturing is not statistically significant. Further, if we restrict the time series in Period-II up to 1990-91, then there is no statistically significant correlation between the IIP and either registered or total manufacturing. If we take shorter time-periods, then the association becomes perverse, as illustrated in Table 2, wherein Period-II is divided into three sub-periods. Therefore, there is some basis to believe that increasingly the IIP has become unrepresentative of the underlying output trends, as reflected in the ASI data. How does the association between the IIP and the ASI look at the disaggregated level? To find out, we do a similar exercise, by estimating correlation coefficients of the growth rates at two-digit industry groups. The answer is no, as most of the correlation coefficients are not statistically significant and, there is no systematic pattern to those that are statistically significant (Tables 3 (a) and (b)). Therefore, it is reasonable to infer that neither in the period of licensing (1971-85), nor in the regime of deregulation (1981-95) was the IIP an accurate predictor of value added at two-digit industry level. To summarise the findings of this section: (1) For the period 1971-72 to 1985-86 (Period-I), growth rates of IIP for manufacturing is highly correlated with those of (the registered and total) manufacturing value added. However, this association turned statistically insignificant during 1980-81 to 1995-96 (Period-II). The associations weaken further and Economic and Political Weekly

turn perverse for sub-periods since 1980-81. (2) At two-digit industry groups, during both the sub-periods, there is no statistically valid association between the growth rates of the IIP and (the registered and total) manufacturing value added. From these, one can reasonably infer that the IIP never accurately predicted manufacturing growth rates at a disaggregated level. Though for the manufacturing sector as a whole the IIP could have been well used as a lead indicator for the 1970s, it cannot be used to predict manufacturing value added in a period of deregulation (in the 1980s and beyond). Clearly, the IIP has deteriorated over the last two decades. This is mainly because the primary data that is used for computing the index has become poorer in quality and probably scarcer in quantity. The recent official press note in fact admits it: “For the registered sector... the quality of production data supplied by the major source agencies suffer from substantial non-response on the part of manufacturing units and consequential estimation resorted to by the source agencies.... The industrial growth based on the revised IIP do not therefore, seem to reflect the perceived ground realities” (p 5, emphasis added). Therefore, no amount of updating and refining the IIP’s weighting diagram can compensate for lack of reliable primary data that are used for computing it. Evidently, the official agency is well aware of the problem. To quote the press release once again, “In order to improve the quality of production data, the Department of Statistics is having regular interaction with the source agencies to improve their system of data collection and estimation procedures. It is expected that the quality of data will improve in the near future”. How will ‘regular interactions’ ensure better data collection? They probably will not, unless the firms face a credible incentive (and a threat) to supply the data.6

II Financial Data and Trends in Fixed Investment Development finance institutions’ (DFIs) sanctions and disbursements of long-term credit have been widely used as

lead indicators of private corporate investment. This is based on Samuel Paul and Rangarajan’s (1973) short-term forecasting model that has been regularly updated for over two decades now.7 Does the flow of long-term credit really predict fixed capital formation in the private corporate sector? To test the proposition, we computed simple correlation coefficient between the annual growth rates of fixed capital formation in private corporate sector (NAS data) and disbursement of long-term credit (both in nominal terms) for the period 1965-66 to 1995-96. Since fixed capital formation is likely to spill over into more than one year, we have also estimated the correlation coefficient with one year lag. Table 4 shows that for none of the time-periods is there a statistically significant correlation TABLE 2: COMPARISON OF THE ASI AND IIP GROWTH RATES OVER THREE SUB-PERIODS Average Total Registered IIP of Years Manufacturing Manufacturing 1981-85 1986-91 1992-96 1981-96

6.2 7.5 6.6 8.8

7.7 7.5 7.1 7.5

5.7 8.9 6.4 7.2

Source: NAS, various issues; Economic and Political Weekly, Vol 29, No 29, July 19-25, 1997. TABLE 3 (a): SIMPLE CORRELATION COEFFICIENTS ANNUAL GROWTH RATES OF THE IIP AND THE NAS VALUE ADDED SERIES, AT 15 2-DIGIT INDUSTRY GROUPS

BETWEEN THE

Average Growth Registered Total Rate for Years Manufacturing Manufacturing 1971-72/1974-75 1975-76/1979-80 1980-81/1984-85 1970-71/1984-85

0.168 0.349 –0.437 0.176

0.480* 0.433* –0.230 0.100

Source: Same as in Table 2. TABLE 3 (b): SIMPLE CORRELATION COEFFICIENTS ANNUAL GROWTH RATES OF THE IIP AND THE NAS VALUE ADDED SERIES, FOR 15 2-DIGIT INDUSTRY GROUPS

BETWEEN THE

Average Growth Registered Total Rate for Years Manufacturing Manufacturing 1980-81/1984-85 1985-86/1990-91 1991-92/1994-95 1980-81/1994-95

0.128 (–)0.223 0.450* (–)0.241

(–) 0.170* 0.694* (–)0.278 0.430

Source: Same as in Table 2.

TABLE 1: SIMPLE CORRELATION COEFFICIENTS BETWEEN ANNUAL GROWTH RATES OF IIP AND NAS MANUFACTURING VALUE ADDED. Correlation Coefficient between IIP and NAS Registered manufacturing Total manufacturing

Period I Period II Period III (1970-71/1984-85) (1981-82/1995-96) (1981-82/1990-91) (1) (2) (3) 0.741* 0.701*

0.440 0.706*

(–) 0.403 (–) 0.007

* Statistically significant at 5 per cent confidence interval, in a two-tailed test. Source: NAS and Economic Survey, various issues.

February 6, 1999

351

between the DFIs disbursement and corporate fixed investment. However, during the period of industrial licensing (1995-80) the correlation is valid with one-year lag. But this ceases to be so in the period of deregulation (1981-96), suggesting that while long-term lending could have been used as a lead indicator in a period of investment licensing, it cannot be used in the liberalised regime. The absence of the association in recent years is widely believed to be due to DFIs’ growing practice of ‘ever greening’: loans disbursed to defaulters that are often used to repay old debts. As the data on disbursements net of ever greening are not publicly available, the widely held belief cannot be tested. Similarly, in recent years, capital raised by initial public offering in the primary stock market is also widely used to predict corporate investment trends in the short to medium term. This measure too has an intuitive appeal. But is it empirically valid? Table 5 shows that these results are similar to the above findings: In the regime of industrial licensing (1962-80), total capital raised in the primary stock market is positive and statistically significantly correlated with corporate fixed investment. This is true even with one-year lag. But the relationship ceases to exist since the 1980-81. Therefore, in the present context, the primary stock market mobilisation has little relation to corporate fixed investment.

III Electricity Consumption and Industrial Output Since almost all modern manufacturing industries use electricity as motive power, and since there is a broad technical relationship between electricity use and value added, growth in electricity consumption, in principle, can be used as a proxy for industrial output growth. To test this proposition, we estimated simple correlation coefficients between annual growth rates of industrial output and electricity consumption (Table 6). None of these correlation coefficients are statistically significant, though all of them have the expected positive sign. Since, in principle, there is a technical relationship between the two variables, lack of correlation suggests incorrect recording of inputs and output.

IV Underestimation of Unregistered Manufacturing Output Over a long period, there is a positive and statistically significant correlation between the growth rates of the registered

352

(ii) benchmark estimates of value added per worker, the underestimation could be because of either variable. A preliminary scrutiny ruled out underestimation of number of workers, as they seem to be based on decennial census and the NSS estimates. Since the estimates of value added per worker are allegedly outdated, this possibly accounts for the underestimation of value added. This seems to be borne out by the exercise reported here, examining the relative movements in the growth in value added per worker, and fixed capital stock per worker during 1981-91 in the registered and unregistered manufacturing sectors. It was found that for 100 units increase in fixed capital per worker in registered manufacturing, value added per worker increased by 156 units. However, in unregistered manufacturing, the corresponding increase in value added per worker was only 88 units. Relatively slower growth of value added per worker in unregistered manufacturing seems to suggest underestimation of value added in this sector due to usage of outdated parameters. The parameters could be outdated (or under-reported) for the following reasons.

and the unregistered manufacturing value added that are reported in the NAS. Therefore, one may believe that the growth rates of unregistered sector are reasonably satisfactory, despite some widely known output underestimation (of level of value added) in this sector.8 However, careful micro studies have repeatedly hinted at the growing manufacturing activity in unregistered sector that escapes the official estimation. While such a criticism has an intuitive appeal, it has been difficult to substantiate it in the aggregate. We now provide some evidence that seems to lend credence to the widely held suspicion. Between 1977-78 and 1993-94, while the unregistered manufacturing sector’s share in total manufacturing value added declined by 4 per cent, its share in total manufacturing employment has increased by 5.1 per cent (Figure 1).9 These changes can be reconciled only under the assumption that the growth rate of value added per worker (labour productivity) in unregistered manufacturing has been growing slower than that in registered manufacturing.10 Since value added in unregistered sector is a product of (i) number of workers and

TABLE 4: SIMPLE CORRELATION COEFFICIENT BETWEEN NOMINAL ANNUAL GROWTH RATES OF DFIS’ DISBURSEMENT AND CORPORATE GFCF Years

No of Observations

Correlation Coefficient

15 16 31

0.0 0.0 0.0

1965-66/1979-80 1980-81/1995-96 1965-66/1995-96

With One-Year Lag No of Correlation Observations Coefficient 14 15 30

0.662* 0.158 0.540*

*Statistically significant at least 1 per cent level. Source: RBI Currency and Finance, and NAS. TABLE 5: SIMPLE CORRELATION BETWEEN ANNUAL GROWTH RATES OF NOMINAL CAPTIAL RAISED IN STOCK MARKET AND GROSS FIXED CAPITAL FORMATION IN PRIVATE CORPORATE SECTOR Years

No of Observations

Correlation Coefficient

19 16 35

0.446* 0.101 0.246

1961-62/1979-80 1980-81/1995-96 1961-62/1995-96

With One-Year Lag No of Correlation Observations Coefficient 18 15 34

0.424* (-)0.215 0.068

**Significant at 5 per cent level, *** significant at 10 per cent level. Source: Same as in Table 4. TABLE 6: CORRELATION OF GROWTH RATES OF ELECTRICITY CONSUMPTION AND INDUSTRIAL OUTPUT Correlation Coefficient between (i) (ii) (iii) (iv)

Years

IIP manufacturing and energy sales Real GDP in regd mfg and energy sales Real GDP in total mfg and energy sales Real GDP in red mfg and real value of fuel used

No of Observations

Coefficient of Correlation

1981-82/1993-94 1981-82/1993-94 1981-82/1993-94

13 13 13

0.426 0.276 0.284

1973-74/1993-94

20

0.243

Notes:

In (i), (ii) and (iii) above, energy sales refer to public utilities’ sale of electricity to industry in physical quantity. In (iv), it is value of fuel consumed by registered manufacturing industries as reported in the ASI deflated by price index for fuel. Source: NAS, ASI, and Public Electricity Supply: All India Statistics, various issues.

Economic and Political Weekly

February 6, 1999

FIGURE 3: NO OF FACTORIES IN STEEL INDUSTRY (NIC 331)

1983

Employment

1987-88 1993-94 Year Value added

FIGURE 2: SHARE OF FACTORY SECTOR IN CENSUS MANUFACTURING EMPLOYMENT

Per Cent

50 48 46 44 42 40 1980

1990 Year

First, labour productivity could have significantly gone up with diffusion of electricity as motive power that has occurred during the last two decades. Second, unregistered manufacturing has witnessed a steady growth in fixed capital formation, thus indicating a steady growth in potential output. Finally, since unregistered manufacturing, unlike the registered sector, operates under competitive conditions (due to low entry barriers), it is reasonable to argue that investment and employment growth in this sector would have occurred mainly under private profitability considerations. In other words, on the face of it, growth in wage employment and fixed capital formation in this sector is unlikely to have occurred unless the increase in labour productivity more than compensated the cost of capital and labour. Therefore, we have a reasonable basis to argue that parameters of value added per worker used for unregistered manufacturing are likely to be underestimated,

which accounts for the growing underestimation of value added in this sector.

V Annual Survey of Industries In principle, all factories registered under the Factories Act (under section 2m(i) and (ii)) are included in the Annual Survey of Industries (ASI). The universe of the ASI is the live register of factories maintained by the Chief Inspectorate of Factories in each state. Therefore, the ASI’s coverage can only be as good as the factories’ list. Under the Collection of Statistics Act (and related laws), all registered factories are expected to file an annual return. Every year, the CSO conducts a census of all factories employing 50 workers and above (100 workers and above without using power). Sample surveys – covering onehalf of all registered factories employing between 10 and 50 workers (20 and 100 workers without using power) – are conducted every year.11 How good are these estimates? Reviewing the methodology, Pradhan and Saluja (1998a) said, “For the organised manufacturing industries fairly reliable data are available annually, but with a considerable time-lag” (p 1270). This view needs to be re-examined for three reasons: (i) incomplete coverage of factories, (ii) under-reporting of workers in factories covered, especially in small factories, and (iii) under-reporting of value added.12 With the size structure moving towards the smaller sized factories within the factory (Average annual growth rate)

(i) (ii) (iii) (iv)

Finished Steel (in Physical Units)

Hot Metal (in Physical Units)

5.7 5.7

5.1 5.6 5.4

1980-81/1994-95 1985-86/1994-95 1981-82/1994-95 1985-86/1994-95

Real Gross Value of Production (ASI Series)

Real Gross Value Added

4.1 6.7

5.4 5.9

Note: Gross value of output and gross value added include NIC 330, 331, 332. Source: ASI Summary Results and SAIL YearBook, various issues.

Economic and Political Weekly

1993-94

1992-93

1991-92

1990-91

1989-90

1988-89

1987-88

Year — No of Factories

TABLE 7: GROWTH IN STEEL INDUSTRY Years

1986-87

1985-86

1984-85

1983-84

1982-83

1980-81

1977-78

1981-82

90 80 70 60 50 40 30 20 10 0

4000 3500 3000 2500 2000 1500 1000 500 0

No of Factories

Share of Total Manufacturing

FIGURE 1: UNREGISTERED MANUFACTURING SECTOR’S SHARE IN TOTAL MANUFACTURING EMPLOYMENT AND VALUE ADDED

February 6, 1999

sector and into the unorganised sector, (i) and (ii) are believed to have increased significantly [Nagaraj 1994]. To illustrate the extent of non-recording of factories and changes in them, we compared the number of factories in ASI in manufacturing with number of establishments in manufacturing employing 10 or more workers in economic censuses of 1980 and 1990.13 In 1980, number of factories in ASI formed less than one-half (48 per cent) of manufacturing establishments in the census. Even if a variety of manufacturing establishments are exempt from the Factories Act, the fact that over 50 per cent of them have not registered under the act suggests a gross extent of under-reporting of factories (Figure 2). This is consistent with evidence from many micro level studies. More significantly, the proportion of factories registered under the Factories Act fell by 5 per cent, to 43 per cent in 1990 suggesting a rapid growth of underreporting of factories.14 This finding can be corroborated with other evidence as well. During 1980-90, when registered manufacturing value added grew annually at over 8 per cent, with a steady delicensing of investment and output controls resulting in considerable new entry into manufacturing industries, yearly trend growth in number of factories was as low as 0.9 per cent. Under-reporting of value added is another important problem that has been repeatedly pointed out by careful studies. Raj (1986) suggested serious underestimation of value added in registered manufacturing due to growing tax evasion.15 More recently, T N Srinivasan (1994) reiterated the same point: “... given the incentive for evasion of excise and other taxes, there are reasons to believe that value added data may be biased and the extent of the bias could be varying over time” (p 9). The above mentioned problems of increasingly poor coverage and probable under-reporting of value added can be 353

illustrated by the following example of steel industry. As shown in Figure 3, number of factories in 3-digit industry 331 (‘manufacture of semi-finished iron and steel products in re-rolling mills, coldrolling mills and wire drawing mills) has sharply fallen from about 3,200 till 1988-89 to about 1,400 thereafter. On the face of it, it could be due to a reclassification, to accommodate a change over from NIC 1971 to NIC 1987. But a close perusal of data did not suggest any corresponding increase in other steel related 3-digit industry groups. Therefore, we suspect that enumeration has become incomplete, unless there is evidence of a large-scale plant closures. The sharp fall in number of factories in this industry is in contrast to other evidence, mainly from the corporate sector.16 Since mid-1980s, in response to delicensing, there has been considerable expansion of existing firms, and new entry into the industry. Moreover, there seems to have been a change in the product mix in favour of ‘flat’ products and technological upgradation leading, in principle, to greater value addition per unit of output – for example, automobile grade flat products of thinner gauge and greater width.

354

Table 6 shows estimated growth rates of output from SAIL Year Book, and real gross value of production and gross value added using ASI data. Evidently, growth rates reported by different measures are roughly of same orders of magnitude. However, given the changes in the industry since delicensing in mid-1980s, many indications suggest an increase in value added to value of production ratio. Since this is not revealed in the growth rates reported above, one suspects that value added estimates may be underestimated.

VI Summary and Conclusion Reliable and up to date statistical information is vital for economic decisionmaking, both at the micro and at the macro level. This exploratory note tried to assess the quality of some of the widely used industrial statistics. IIP is the most widely used leading indicator of output trends, as it is available monthly, with least time lag, and with analytically meaningful disaggregation. NAS is the only source of data for the unregistered manufacturing value added and capital formation. ASI provides detailed information on registered manufacturing, though with considerable

time lag. Long-term credit by development finance institutions (DFIs) and the initial public offerings in the primary capital market are also widely used to predict fixed investment trends in the private corporate sector. How reliable are these data sources? Have their quality deteriorated over time? This note sought to answer these questions, using simple correlation coefficient method to time series of annual growth rates. The following are the main results. (1) Annual growth rates of the IIP for manufacturing and value added in manufacturing (registered and total) sector are highly correlated, for the periods1971-85 (Period I). But the association turns statistically insignificant for 1981-96 (period II), and parts thereof. Correlation between the IIP and the ASI for crosssection of 2-digit industry groups is not statistically significant for both the time periods. Since the IIP is a lead indicator, it could have been used to predict valueadded trends in Period I, that is, during the regime of investment and output licensing. However, it cannot be used in the same way in the liberalised regime (Period II). The study supports the view that the quality of IIP as deteriorated since the 1980s with gradual industrial

Economic and Political Weekly

February 6, 1999

deregulation, as much of the primary data for estimating the index was a by-product of the regulatory regime. (2) Widely used information on (i) development finance institutions’ sanctions and disbursements of term loans and (ii) capital raised in the primary capital market is not correlated with fixed capital formation in the private corporate sector since 1980-81. Therefore, this financial information is a poor predictor of the real trends in the deregulated regime. (3) In principle, though there is a strict technical relationship between electricity use and manufacturing value added, in reality this association was not found to exist in India. This finding questions the quality of the recorded information on electricity input and value added estimates. (4) During 1977-78 and 1993-94, while the employment share of unregistered sector in total manufacturing has gone up, the corresponding value added share went down. These inverse movements can be reconciled only under the assumption that the labour productivity growth in unregistered manufacturing is lower than that in the registered sector. A closer examination seems to strengthen the suspicion that the parameters of value added per worker used in computing output could have been seriously underestimated. (5) In recent times, the ASI seems to underreport number of factories and hence value added even in a well-organised industry like steel. It, therefore, raises suspicion that the quality of the ASI data is declining in recent years. Admittedly, results of this exploratory effort have yielded only bits and pieces of evidence on the quality of the data. They nevertheless seem to tell a reasonably consistent story: India’s industrial data system has weakened over the years, and therefore the information may not reliably capture the underlying real tendencies. Many of the widely used indicators and presumed technical relationships have little empirical validity. This finding supports the popular perception of the deteriorating data quality. If this inference is correct, then there is an urgent need for a thorough re-examination, and revamping of the statistical system.

Notes [Following the usual disclaimer, the author thanks K V Ramaswamy, M H Suryanarayana and A Vaidyanathan for their detailed comments and suggestions on earlier versions of this study.] 1 For a detailed account of the strengths and limitations of all these sources of data, see Pradhan and Saluja (1998a). 2 Pradhan and Saluja (1998b) gives details of

Economic and Political Weekly

the revision that the IIP has recently undergone. 3 Brief Note on the Revision of Base Year of Index of Industrial Production from 1980-81 to 1993-94 (undated). 4 In fact, the CSO officials have admitted this. To quote Kulashrestha and Kolli (1995): “After liberalisation, some of the major units including some of the PSUs have not been furnishing returns. This makes the estimation procedure for non-responding units very difficult in the absence of information on whether the unit is in existence or closed, or on strike or on partial operation. Government’s efforts to persuade the units to furnish returns now met with little success. The GOI have subsequently issued a Press Note ... reiterating the requirement of submission of returns by the industrial undertakings to concerned technical agencies. Despite this, the coverage of units has been steadily declining”. (p 125) 5 Unless otherwise mentioned, all variables in this paper are in real terms. 6 A report in The Times of India (January 16, 1999) said, ‘The Department of Statistics has not released the industrial production figures for November 1998 on the ground that the data provided by the department of industrial policy and promotion (DIPP) on manufacturing sector ‘suffers from lack of quality’ ... ‘Despite repeated efforts, the DIPP ... has not furnished the information regarding itemwise response rate as well as the method of estimation or non-response’ Mr Asthana [Secretary in the department of Statistics] said (p 17). 7 Till 1982, Rangarajan’s forecast of corporate investment was reported in the Economic and Political Weekly. In the recent years, these were officially estimated, and reported in the Reserve Bank of India Bulletin. 8 There have been many studies on the ‘black economy’ that in fact looked carefully at specific unorganised manufacturing industries like power loom weaving, dyestuffs, etc. Most of them estimate the extent (level) of underestimation without saying if the black economy is growing faster than the measured economic output. 9 This method of looking at the relative employment and value added shares to detect output underestimation is an old one, widely used in studying the long term trend by Arthur Burns and Simon Kuznets. 10 Sources of data for Figure 1 are, NSS employment and unemployment surveys, and NAS, various issues. 11 For a careful and fuller description of the ASI’s methodology, see Pradhan and Saluja (1998a). 12 Growing non-response to ASI is widely believed to be yet another reason for deteriorating data quality. On a closer examination, we did not find any statistically significant deterioration in the extent of non-response over the period 1980-81 to1994-95. 13 These figures refer to all-India, excluding Assam, as the censuses were not conducted in that state. 14 This evidence on the growing extent of nonregistration of factories under the factories act seems to reinforce findings of many fieldwork based micro studies [Nagaraj 1989; Singh 1990]. But what is more surprising as

February 6, 1999

we discovered during our field-work in Bangalore city in mid-1980s, was that many medium sized factories that we had personally visited were missing from the factories’ list. Though we do not have documentary evidence to support our case, we do believe there is a case for examining the quality of the factories’ list maintained by the Inspectorate of Factories. 15 To quote Raj (1986:11), “The number of ‘registered firms’, i e, those registered with income tax authorities, has been increasing at a phenomenal rate from about the middle of the 1960s. Many of them are known to be used by manufacturing enterprises as commission agents for purchase of inputs and sale of products, thereby siphoning away profits through various forms of transfer pricing. Underestimation of value added in this manner has been therefore probably increasing in scale through this period”. 16 After delicensing of steel industry in 1985, there was considerable new entry into the industry as evident from capital mobilised in the primary stock market and term loans granted by development finance institutions. Of the 48 listed ‘mini steel’ companies listed in Bombay Stock Exchange in 1997, half entered the industry after 1985. For detailed statistical information, see the annual report of the Department of Steel, 1997-98.

References Ahluwalia, Isher Judge (1985): Industrial Growth in India: Stagnation since the Mid-1960s, Oxford University Press, Delhi. Kulashreshtha, A C and Ramesh, Kohli (1995): ‘Impact of Liberalisation on Data Collection’, The Journal of Income and Wealth, Vol 17, No 2, July. Nagaraj, R (1989): Sub-Contracting in Manufacturing Industries: The Bangalore Experience, Ph D thesis (Centre for Development Studies, Thiruvananthapuram, Jawaharlal Nehru University, New Delhi. – (1994): ‘Employment and Wages in Manufacturing Industries: Trends, Hypothesis and Evidence’, Economic and Political Weekly, Vol 29, No 4, January 22. Paul, Samuel and C Rangarajan (1974): ShortTerm Investment Forecasting, Macmillan, Delhi. Pradhan, Basanta K and M R Saluja (1998a): ‘Industrial Statistics in India: Sources, Limitations and Data Gaps’, Economic and Political Weekly, Vol 33, No 21, May 23. – (1998b): ‘Revised Index of Industrial Production: A Note’, Economic and Political Weekly, Vol 33, No 28, July 11. Reserve Bank of India Bulletin (1986): ‘Index Numbers of Industrial Production – Revision of Weights’, Reserve Bank of India Bulletin, Vol 40, No 10, October. Singh, Manjit (1990): The Political Economy of Unorganised Industry: A Study of the Labour Process, Sage India, Delhi. Raj, K N (1986): New Economic Policy, V T Krishnamachari Memorial Lecture, Oxford University Press, Delhi. Srinivasan, T N (1994): ‘Data Base for Development Analysis: An Overview’, Journal of Development Economics, Vol 44.

355

Related Documents