1 Introduction

T

When it was originally inserted into the IDA by an amendment passed by Parliament in 1976, chapter V-B (according to its section 25-K) applied to industrial establishments employing an average of 300 or more workers per working day, excluding establishments “of a seasonal character or in which work is performed only intermittently”. An amendment to section 25-K in 1982 (brought into effect in 1984) reduced this employment threshold to 100. One strand of the earlier literature, following from Fallon and Lucas (1993), treats these years as structural breaks which can be exploited for “before and after” (or “difference in difference”) comparisons of industrial employment and its adjustment to labour demand in order to assess the impact of EPL. Results from these studies are mixed. Another strand of literature, f ollowing from Besley and Burgess (2004) (hereafter BB), uses cross-state as well as inter-temporal variation in the IDA. BB classified state-level amendments (not just to chapter V-B) as proworker, pro-employer or neutral, and assigned them scores of +1, –1 and zero, respectively. If several amendments were passed in the same year in a state, it was assigned a score of +1, –1 or zero for that year, depending on the net direction of change. Finally, a “regulatory measure” (which I call the BB index) for each state was obtained for any year by cumulating its annual scores up to that year. This index was used as a regressor (along with some control variables) to explain various state-level outcomes. Most of these studies find adverse effects of pro-worker legislation on employment, output, investment and productivity in

o rganised manufacturing.

Two recent studies have recognised the centrality of chapter V-B and also the merit of some of my criticisms of the BB index. Among the studies I had criticised in 2006 was one by Ahsan and Pagés which had divided the state amendments of the IDA into different categories, and showed that pro-worker amendments to chapter V-B had adverse effects. They have now come out with a substantially revised version (Ahsan and Pagés 2009) which p urports to take on board my criticism. The first-ever OECD E conomic Survey of India (OECD 2007: ch 4) also acknowledges my arguments, and uses its own index of state-level reform of labour regulation based on questionnaires addressed to employers, trade unions and officials in each state. Although this index is based on changes to the IDA as well as other regulations, the OECD Survey notes that chapter V-B of the IDA is what makes India unusual. Applying a standard methodology for assessing the restrictiveness of labour laws, it shows that for fixed-term and temporary employment, India’s EPL is only slightly more restrictive than the mean for OECD countries. For regular workers it is stricter than all but two, but would fall to the OECD average were it not for chapter V-B (OECD 2007: 122). For collective dismissals, India is far more restrictive than any OECD country (OECD 2007: 125), again as a result of chapter V-B.

In my earlier paper, I had pointed out several problems with the literature as it stood in 2006. Focusing now on chapter V-B alone, I shall elaborate on one of these problems, point out some new ones, comment critically on aspects of Ahsan and Pagés (2009), the OECD Survey, and my earlier paper, and report the results of a limited empirical exercise of my own. In Section 2, I summarise my earlier paper in order to make the present one self-contained. Section 3 expands upon a point made briefly in the earlier paper regarding high court judgments and state amendments that brought about considerable interstate variation in the implementation of chapter V-B, which has not been adequately recognised in the earlier literature. Section 4 discusses some new hypotheses regarding firms’ adaptation to changes in EPL, notably subdivision and outsourcing in order to evade the application of chapter V-B. Section 5 highlights significant data limitations which previous studies have not addressed. Section 6 summarises the results of an attempt to test empirically the hypotheses developed in Section 4 while exploiting the variation highlighted in Section 3 and avoiding the data problems discussed in Section 5. Section 7 concludes the paper.

2 A Summary of Bhattacharjea (2006)

In Bhattacharjea (2006), I had reviewed some studies which showed that labour flexibility in organised manufacturing had actually increased after chapter V-B was introduced in 1976 and also after it was tightened in 1984. I also drew attention to the fact that sections of chapter V-B were inoperative for various periods in different states after being struck down by the Supreme Court and some high courts, and were restored – again at different times in different states – by legislative amendments and r eversal of the high court decisions on appeal. I discuss this issue in greater detail in Section 3 below.

My 2006 paper dealt mostly with BB, and several other studies that had used their index of labour regulation to explain various indicators of industrial performance. I showed that BB misinterpreted several of the state-level IDA amendments; assigned identical scores to minor procedural amendments and major changes in job security rules; aggregated incommensurable pro-worker and pro-employer amendments occurring in the same year to give a summary score of +1 or –1 to a state for that year; used a misleading summation of these scores over time (so that a state that passed amendments in different years was assigned a higher score than one that passed the same amendments

56 s imultaneously); and ignored hundreds of other labour laws, i ncluding some whose provisions overlap with the IDA. Although these criticisms show that the BB index was carelessly constructed, from an econometric point of view measurement errors in an e xplanatory variable bias the estimated coefficient towards zero, and thus do not undermine the results. I went on, however, to critique BB’s use of irrelevant control variables and inadequate tests for robustness; the fragility of their key results when statespecific time trends were included in the regression; and their omission of other important variables influencing industrial l ocation such as the central allocation of industrial licences and the cross-state variation in human capital and industrial r elations.2 I also drew attention to BB’s strange characterisation of Gujarat as pro-worker, and to the entirely unwarranted i nterpretation of their results in the World Bank’s 2004 World D evelopment Report.

My 2006 paper also critically surveyed five later studies which had used the BB index to explain various aspects of industrial performance, and showed that they carried over many of the original problems while creating a few more.3 Significantly, one feature common to most of these studies was that their results, including those for the 1990s, seemed to be driven by cross-state variation in the BB index as it stood in 1980, rather than incremental changes in each state. Finally, I reviewed substantial additional evidence to show that although there had been little change in the BB index in the late 1980s and none in the 1990s, there had been a substantial change in the industrial relations scenario during this period. I presented evidence that included a sharp fall in the incidence of strikes and lockouts, stagnant or declining real wages, a decline in the wage share of value added, a trend towards pro-employer judicial verdicts, very different rankings of states’ labour regimes according to other surveys of business conditions, widespread reductions in employment by individual firms and also nationwide, along with substantial increases in employment in states that were classified as “proworker” according to the BB index. All these effects occurred despite little or no formal change in labour laws, making the BB analysis particularly inadequate for the recent period. I concluded with an expression of agnosticism regarding the impact of labour regulation on industrial performance in India, maintaining that there are sound theoretical arguments on both sides of the debate, but the evidence is inconclusive and the quality of a cademic research leaves much to be desired.4

3 Interstate Variation in the Enforcement of Chapter V-B

I now take up in much greater detail a problem that I discussed briefly in my earlier paper. BB (2004: 103) asserted that the i mpact of the central amendments of 1976 and 1982 are captured by year fixed effects; Ahsan and Pagés (2009) follow the same methodology. But the effect of these amendments was not confined to the year of enactment. In order to be consistent with the treatment of the state amendments that went into the construction of the BB index, the central amendments should be captured by dummy variables that switch from zero to one the year after they took effect, and remain at one thereafter. But even this assumes that central amendments of the IDA were enforced simul taneously in all states, which

may 30, 2009 vol xliv no 22

REVIEW OF LABOUR

was not the case. In some states, implementation was delayed by judicial verdicts, while in others amendments almost identical to the 1982 central amendment were enacted by state governments before the central government brought it into force in 1984.

The Supreme Court held as early as 1978, in its Excel Wear judgment, that section 25-O of the IDA, governing closures, was constitutionally invalid because it imposed unreasonable restrictions on the right to carry on business, which was guaranteed by Article 19 (1) (g) of the Constitution. This right was interpreted as including the right to close down a business, and the section as it stood provided neither procedural guidelines, nor any provision for review of the government’s decision.5 In 1980, the Madras High Court struck down sections 25-M and 25-N (governing layoffs and retrenchments) on similar grounds, and the Rajasthan High Court struck down the latter in 1983.6 To overcome these objections, Madhya Pradesh, Maharashtra, Orissa and Rajasthan passed curative amendments to section 25-O, Uttar Pradesh i nserted a section in its own IDA, and finally the central government amended all three sections with effect from 1984, providing procedural guidelines and review mechanisms.

But this was not the end of the story. The amended 25-O was struck down by a single judge of the Karnataka High Court in 1985, on the grounds that the modifications did not remove the problems pointed out by the Supreme Court.7 This decision was reversed on appeal by a division bench of the same court in 1989.8 The Calcutta High Court (with jurisdiction over West Bengal) similarly struck down the amended 25-O in 1988.9 The corresponding section of the Uttar Pradesh IDA was also struck down in 1990 by the Allahabad High Court, having jurisdiction over the state.10 But the Supreme Court eventually reversed all the adverse high court decisions and upheld the original section 25-N in 1992, the amended section 25-M in 1994, and the amended section 25-O and its Uttar Pradesh counterpart in 2002.11 All these developments have a direct bearing on the attempts by BB and others to quantify states’ EPL regimes. This requires some e laboration.

I begin with the curative amendments passed by certain states to revive section 25-O, which had been struck down in the Excel Wear judgment, before the central amendment along the same lines took effect in 1984.12 BB consider only specific sub-sections of these amendments, coding them as +1 on the grounds that they deviated in a pro-worker direction from the central amendment. In Madhya Pradesh, the deviation applied to construction activities, which is irrelevant to a study of manufacturing. Picking up on my earlier article, Ahsan and Pagés (2009) recode this amendment as zero. But I had not noticed that BB had coded the Maharashtra and Orissa amendments as pro-worker because they deviated from the central amendment by giving “power of appeal to workers to overturn decision to close down firm” (BB 2004: 131). In fact, the Orissa amendment did nothing of the sort, and was identical to the central amendment. The Maharashtra amendment gave both employers and workers the right to appeal, and should be regarded as neutral rather than pro-worker. Even if we regard it as pro-worker, it took effect from 1981, rather than 1983 as erroneously tabulated by BB, and thus according to their aggregation rule should not have affected Maharashtra’s score in 1981 because other pro-worker amendments took effect

Economic & Political Weekly

EPW

the same year. Rajasthan’s amendment was followed only a few months later by the central one, and was not significantly different from it. All four amendments should thus be disregarded.

On the other hand, it is possible to step back and view these amendments in a different light. Coding deviations from the central amendment amounts to missing the wood for the trees, for the whole purpose of the state amendments, with or without deviations, was to reanimate all of section 25-O, which was undeniably pro-worker. However, Rajasthan’s amendment remains disqualified on the grounds stated above, as does Maharashtra’s once we correct the date. The others can still be treated as proworker for the brief period until the central amendment took effect in 1984 – provided that for the entire country, the creation of chapter V-B in 1976 is treated as pro-worker, and the period d uring which section 25-O remained unavailable because of the Excel Wear judgment is treated as pro-employer.

Further, Maharashtra, Orissa, Rajasthan and West Bengal amended section 25-K so as to reduce the employment threshold at which chapter V-B applies, even before the central government did so with effect from August 1984. BB code all these state amendments as “pro-worker” and summarise them in identical terms, ignoring crucial differences in the coverage of these amendments. West Bengal reduced the threshold to 50, and the others to 100, but Maharashtra applied this threshold only to section 25-O. But these reductions were undertaken simultaneously with other proworker amendments, so according to the BB aggregation rule, they do not further alter the states’ scores in the respective years.

On the other hand, the BB index completely overlooks an amendment passed by Uttar Pradesh to its own IDA in 1983, inserting two new sections 6-V and 6-W, identical to sections 25-K and 25-O of the central IDA as amended in 1982, except that the employment threshold for permission for closures was set at 300. The Supreme Court later held that this would prevail over the central IDA, noting that the state legislature had passed it conscious of the 1982 central amendment that had lowered the threshold to 100, and the president had given his assent.13 Effectively, the reduction in threshold was legislatively blocked in Uttar Pradesh, making the state deviate significantly from others, but this is not reflected in the BB index, or in the Ahsan and Pagés index of changes in EPL.

While the curative amendments discussed above have been imperfectly incorporated into the BB analysis, none of the court judgments striking down various sections of chapter V-B has even been recognised, although they too have direct implications for the index. In principle, as high courts are constitutional courts, the invalidation of a central law by any of them is operative across the country. But in each instance in which a section was struck down, at least one other high court upheld it. In such cases, the relevant section remains operative in jurisdictions where it has been upheld, but not in those where it has been struck down (unless the high court judgment is stayed). Its status in jurisdictions where high courts have not pronounced upon its validity remains a grey area, but there is a presumption of constitutionality.

These legislative and judicial developments also show that amendments to chapter V-B have not been accurately captured in the BB approach. But even taking them into account does not allow us to carry out the BB analysis by changing the scores of the affected states in the relevant years from zero to one or vice versa. Judgments operate retrospectively from the date the legislations were enacted. So if a court strikes down a section, it becomes invalid from the start. This might not be of much consequence for actions taken in the interim: if permission for retrenchment or closure is denied while the section is in force, it cannot be granted retrospectively a fter it is struck down. But if a high court judgment striking down a section is later reversed by the Supreme Court, the section is revived from the date of enactment, as if the high court had never struck it down. Then retrenchments undertaken without permission become retrospectively illegal, and can prove extremely costly.

To appreciate this point, consider an epic legal battle in Rajasthan between J K Synthetics and the Rajasthan Trade Union K endra (RTUK). In 1983, under the impression that section 25-N was no longer operative after the Madras High Court had struck it down three years earlier, the company retrenched 1,201 w orkers without seeking government permission. Later the same year, the Rajasthan High Court allowed a writ petition by the company and held that the relevant clauses of section 25-N were un constitutional, simultaneously dismissing the RTUK petition c hallenging the r etrenchments.14 The RTUK then appealed to the Supreme Court. While the appeal was pending, in 1992 the S upreme Court held in its Meenakshi Mills judgment (n 11 above) that section 25-N as enacted in 1976 was constitutionally valid. In 1993 it remanded pending matters back to the Rajasthan High Court to decide on merits in light of that judgment. In 1994 a s ingle judge of the latter court held that the 1,201 workers retrenched in 1983 would now be entitled to reinstatement with full back wages.15 This part of the single judge’s order was upheld on appeal by a division bench of the same court, which sternly reprimanded the company for arguing that it had effected the retrenchments u nder legal advice that section 25-N was unconstitutional.

It is not understood who opined that the writ of Madras High Court would run even in the territory of the state of Rajasthan. There is always a presumption that a validly enacted provision of law is in accordance with the constitutional provisions … However despite this settled legal position, the company ventured to act upon the advice that the provisions of Section 25N of the act were unconstitutional. If the company took risk, it cannot save itself from the consequences flowing from the breach of the p rovisions.16

What this case tells us is that there is no obvious answer to the seemingly simple question, “Was section 25-N of the IDA in force in Rajasthan in 1983?” The company did not think so, and the high court validated this belief later the same year, but a decade later the section was deemed valid from its enactment in 1976, the r etrenchments were illegal, and the company was liable to pay wages with retrospective effect. Economists would assume that a fter the Madras High Court struck down the section in 1980, private agents in Rajasthan would have assessed some probability of its being revived, and acted accordingly. If one were to adhere to the BB approach, Rajasthan’s score should be incremented by some fraction between zero and one. But this probability would perhaps decrease after the Rajasthan High Court also struck down the s ection in late 1983, and continue to vary until the Supreme Court settled the issue in 1992. Presumably agents in other states would have assessed different time-varying probabilities to the e nforceability of each of the contested sections of chapter V-B, as legal challenges made their way through the judicial system. The complications for the BB index are obvious.

Many of these problems carry over into the otherwise pathbreaking study by Ahsan and Pagés (2009). They subdivide the 113 state amendments listed by BB into those that affected the ability of parties to initiate or sustain industrial disputes on the one hand, and EPL amendments that increased job security and reduced labour flexibility on the other. They construct a BB-type index for each of these two types of amendment separately, assigning scores of +1 to amendments that increased the costs of settling disputes and –1 to those that reduced it. All EPL amendments were in the direction of greater restrictions on labour flexibility, and are coded +1. A similar index is constructed for amendments to chapter V-B alone. The original BB index turns out to be strongly correlated with the disputes index, but only weakly with those for EPL and V-B. Correspondingly, the adverse effects of the disputes index on various performance indicators for organised manufacturing were much stronger than that of EPL, although the two types of amendments mutually reinforced each other.

Ahsan and Pagés then show that their principal results continue to hold after recoding some amendments in conformity with observations in Bhattacharjea (2006). The coefficients on the BB index become larger, confirming that the original coding was erroneous, but recall that I had acknowledged that correcting errors in measuring regressors could strengthen the results. Ahsan and Pagés also admit that after the corrections the very large coefficient on the index of amendments to chapter V-B now rests on amendments in only three states (Maharashtra, Orissa and West Bengal), and could be partially driven by reverse causality (2009: 72). But the problem is actually much deeper. As I showed above, even these amendments were misinterpreted by BB, and differed considerably in coverage. Moreover, Ahsan and Pagés ignore the interstate variation caused by the 1983 amendment of the Uttar Pradesh IDA, the central amendment of chapter V-B which was enforced from 1984, and the court judgments. F inally, my critique of BB was not limited to pointing out a few misinterpreted amendments; I had also questioned their methodology on the various other grounds discussed above (aggregation, cumulation over time, omitted variables, and so on), and Ahsan and Pagés continue to follow it in all other respects.

To sum up this section, legislative amendments and judicial decisions in different states in different years show that 1976 and 1984 cannot be treated as natural breaks that changed EPL across the country, as in the Fallon and Lucas approach. But I would argue that these state-level changes are also too heterogeneous to be uniformly quantified as unit increments (or decrements) in an index which is then used as a regressor in a cross-state panel regression. Each change should be treated as a separate case. Section 6 s ummarises an attempt at empirical analysis along these lines.

4 Modes of Adaptation to EPL

A tightening of EPL can be viewed as a negative shock to a firm’s discounted present value. Its response can take various forms, of which the existing literature considers the following possibilities: doing nothing, closing down, relocating to another state,

may 30, 2009 vol xliv no 22

REVIEW OF LABOUR

d ownsizing production, and replacing regular with contract labour. The literature has so far not considered the option of subdividing production horizontally among nominally independent factories under common control, or outsourcing some inputs previously produced in-house. The latter option may involve vertically subdividing the production process among nominally independent units or subcontracting to genuinely independent ones.

Although the trend towards greater subcontracting from the registered (“organised”) to the unregistered (“unorganised”) manufacturing sector has been studied by earlier authors (Ramaswamy 2003; Unni and Rani 2008), they have not examined it specifically in the context of firms’ adaptation to EPL changes of the 1980s. The literature on labour regulation does examine whether its adverse effects on registered manufacturing were mirrored by the distribution of registered factories being skewed towards those with less than 10 or 20 workers (OECD 2007: 128), or expansion of the unregistered sector (BB 2004: 105), both of which are outside the purview of most labour legislation.17 But Ahsan and Pagés (2009: 68) show that their disputes index, rather than EPL, is driving the BB result. In any case, the hypothesis of this literature is that pro-worker regulation encourages firms to stay small so as to avoid registration. The literature does not recognise the possibility that an observed increase in size of the unorganised sector could also be a consequence of subdivision by registered factories above the employment threshold for chapter V-B, although fragmentation into such small units would entail a huge sacrifice in terms of feasible technologies and scale economies.

Fragmentation or outsourcing within the organised sector itself would not entail as large a sacrifice, but this possibility has also not been recognised in the earlier studies. Chapter V-B, covering factories employing 100 or more workers, requires official permission for layoffs, retrenchments and closures, and also notice and compensation to the workers. Factories with less than 100 workers are covered by chapter V-a, which does not require official permission, imposes shorter notice periods, and provides a cap on compensation in cases “where the undertaking is closed down on account of unavoidable circumstances beyond the control of the employer”. Undertakings employing less than 50 workers are not required to give notice of closures to the government, or to pay compensation for layoffs. Fragmenting or outsourcing production to smaller factories within the organised sector may therefore be a viable strategy if the gains from escaping enhanced restrictions on labour flexibility outweigh the resulting increase in transactions costs.18

This has an immediate implication for a key hypothesis of the earlier literature. BB (2004), Aghion et al (2008) and Ahsan and Pagés (2009) show that states that enacted more pro-worker amendments to the IDA witnessed a relative decline in the number of factories. From this they infer a higher rate of plant closures and reduced capital formation in such states. I point out in the next section that the reported number of factories does not accurately capture exit. But even at a theoretical level, these authors ignore the possibility that larger units were subdivided vertically or horizontally into smaller ones specifically to drop below the radar of V-B. Similarly, fresh entrants might set up several small registered units rather than one large one. Hence, states in which

Economic & Political Weekly

EPW

EPL became more stringent might witness an increase in the number of factories.

Any assessment of whether firms actually resorted to fragmentation as a strategy to cope with a tightening of EPL runs into several possible complications. First, individual factories are not identified in the data, so entry, exit and subdivision can only be indirectly inferred from the number of factories. Second, the a dverse effects on investment noted by the earlier studies might simultaneously be at work, so the effect on the number of factories could go either way. Third, the newly spun-off units may have been established in other states with weaker EPL, in which case we would not observe an increased number of factories in a state with stricter EPL. Fourth, subdivision and relocation would necessarily involve the creation of new factories, to which some of the labour and machinery of the old plant would be transferred. The then-pervasive investment licensing regime, which required a licence for setting up new units and relocating old ones, would have dampened this effect. Starting in the mid1980s, however, such relocation and subdivision would have b ecome easier as a result of the liberalising reforms that d elicensed selected industries in 1985-86 and almost all the r emaining industries in 1991.

A further complication arises from the fact that the threshold in section 25-K of the IDA and the layoff and retrenchment provisions (in sections 25-M and N) are all stated in terms of employment in an industrial establishment, while section 25-O applies to closure of “an undertaking of an industrial establishment.” S ection 25-L of the IDA states that for purposes of chapter V-B, i ndustrial establishments include, apart from mines and plantations, factories as defined in clause 2(m) of the Factories Act which covers “any premises…in any part of which a manufacturing process is being carried on.” The omnibus 1984 amendment of the IDA also inserted section 2(ka), according to which if only part of the activities of an establishment were industrial, and the rest were severable, then they would not be taken to be part of the establishment.

Judicial interpretation of these clauses has interesting implications for firm adaptation to EPL. A 1960 precedent held that a cement factory and a nearby quarry owned by the same firm and supplying it with limestone constituted a single establishment, though their workers were regulated by different authorities.19 A 1986 case involved closure of a marketing division which also handled the accounts, purchases, and payroll for a manufacturing unit belonging to the same owners, but the bulk of whose business came from marketing the products of another firm. The manufacturing unit was registered under the Factories Act, while the marketing division was located in a different area and r egistered under a separate act governing shops and commercial establishments. Nonetheless, the Supreme Court, reversing the decision of the state’s industrial court, held that their employment would be aggregated to determine whether the closure required permission under section 25-O.20 It held that a factory can be located in more than one set of premises, an establishment can have more than one undertaking, functioning in different premises, and an undertaking of an industrial establishment need not be a factory or industrial establishment. The closure of any such undertaking, even though it employed less than 100 workmen, could therefore t rigger section 25-O if the total employment in all undertakings having a “functional integrality” as part of an industrial establishment was not less than 100. The very next year, the Supreme Court held that two factories owned by the same firm, carrying on the same kind of business and l ocated within 200 metres of each other, but having independent licences, accounts and payrolls, were separate undertakings, and one of them, employing 32 workers, could be closed down i ndependently.21

These cases suggest first, that a firm might be not be able to evade chapter V-B through vertical division, but horizontal division seems more likely to succeed, especially after the liberalisation of investment licensing. Second, they also suggest that firms might have changed their strategies after these judgments in the mid-1980s. In particular, subdivision might have to give the impression that the separate units have separate owners. OECD (2007: 128) cites anecdotal evidence of fictitious partitioning of industrial units among different owners (usually members of the same family) to avoid having to register.

Apart from these additional modes of adaptation to EPL, the timing of adaptation can be different from what is assumed in the earlier literature. Both the “before and after” and BB approaches look for changes in industrial performance indicators in the years following IDA amendments, applying a lag of one year and sometimes more. As some of the authors acknowledge, stricter EPL discourages both hiring and firing, so the theoretical effect on employment is uncertain. The adverse effects on employment that some researchers have found can then only be attributed to hiring being retarded relative to natural separations. However, if a tightening of EPL is anticipated, the potentially affected firms may begin to cut back on hiring and also resort to retrenchment or closure before the restrictions take effect. Stronger effects would appear before the amendment is brought into force. The two-year lag in the notification of the 1982 amendment provides a good example of an anticipated tightening of restrictions, even though employers could not have anticipated the date of notification with certainty.

5 Data Limitations of Previous Studies

(A) Inappropriate Factory Counts

BB (2004), Aghion et al (2008) and Ahsan and Pagés (2009) use the number of factories registered in each state to capture the effects of labour regulation on net entry. There are several problems with this. First, as Figure 1 shows, the vast majority of such factories employs less than 50 workers, and is not subject to most of the job security laws discussed here. Further, many factories shut down and yet may remain registered for years. In 1982-83, the Annual Survey of Industries (ASI) changed the way in which it dealt with “non-existing” factories, that is, factories registered under the Factories Act which were included in the ASI frame but which did not submit returns and could not be traced. As Figure 1 shows, the number of factories consequently fell by about 11%. The decline is attributable almost entirely to the numerically predominant 0-49 employment size class. This is what one would expect, not only because of the higher mortality of such units, but also because up to 1982-83, all “nonexisting” factories would have been reported as having zero e mployment and therefore be assigned to this size class.

60

The problem this poses for authors who use the count of registered factories is threefold. If factories are not deregistered when they cease production, changes in the number of factories cannot capture exit. Second, if entry increases the number of registered factories but exit does not reduce it (at least for several years), the factory count will over time increasingly represent such “ghost” units. Third, ratios formed by dividing aggregates by the number of factories to obtain indicators such as average employment or capital per factory (as in Ahsan and Pagés 2009: 70) will be distorted by having these ghosts inflating the denominator while contributing nothing to the numerator.

(In an earlier version of this paper (Bhattacharjea 2009), I mistakenly criticised BB and Ahsan and Pagés for using ASI data on the number of factories, which would give the impression that the apparent mass exit of factories in 1982-83 was a consequence of the pro-worker IDA amendments of the early 1980s. As bb did not use ASI data, this criticism was misconceived, but the data on registered factories that they do use is plagued by the errors d escribed above.)

Figure 1: Number of Factories 1973-2001

140000 120000 100000 80000 60000 0-49 Employees Total

40000 All others

20000

0

1973 1977 1981 1985 1989 1993 1997 2001

-74 -78 -82 -86 -90 -94 -98 -02

Source: ASI data as compiled in EPW Research Foundation (2007).

(B) Inappropriate Employment Thresholds

To see how establishments responded to the reduction of the threshold of applicability of chapter V-B from 300 to 100 workers, one should ideally look for differential behaviour by establishments employing 100 to 299 workers. However, only the ASI gives a distribution of factories by employment size, and it presents several insuperable obstacles to making this comparison. First, it does not identify individual factories, so monitoring their adaptive responses by tracking them over time is not possible. We, thus, have to infer these responses from changes in the characteristics of different size classes of factories in different states. But this too is problematic, because the distribution of factories by employment size is not available at the state level. Second, the definition of employment in ASI data includes supervisors and managers, whereas the threshold in section 25-K of the IDA is in terms of “workmen”, explicitly defined in section 2(s) to exclude these categories.22 The ASI does report the aggregate number of workers and other employees separately, but the size distribution is reported only by number of employees, making it impossible to match the IDA thresholds. Third, ASI employment data include contract workers, while factories can exclude them for

may 30, 2009 vol xliv no 22

6 An Empirical Test

In an earlier version of this paper (Bhattacharjea 2009), I reported the methodology and results of a limited empirical

Economic & Political Weekly

EPW

e xercise to test the hypotheses advanced earlier regarding firms’ adapting to EPL by subdivision, outsourcing and employing contract labour, while avoiding the problems described above. To capture exit and subdivision, I used the number of factories reported by the ASI in the post-1981 period, when non-functional factories were dropped. The ratio of value added to gross value of output was used to capture vertical subdivision as well as outsourcing, both of which should cause a fall in the ratio. The proportion of contract to total workers was used for obvious reasons. I looked for atypical changes in these variables in the years immediately preceding and following legislative and judicial interventions affecting chapter V-B in a state, relative to other states in which no such change took place in the same year. I distinguished between changes that loosened EPL for employers (high court judgments that invalidated sections of chapter V-B), and those that tightened it (by reversal of these judgments on appeal, or by legislative amendments that overrode them and/or reduced the employment threshold at which the chapter applied). For reasons discussed in Section 4 above, I allowed for adaptive behaviour in anticipation of amendments, but not judgments.

Unfortunately, the results of this exercise were very mixed, and hence are not reported here. In some cases, the changes were in the direction predicted by my hypotheses, but not in others. This is not altogether surprising. First, the data were perhaps too crude. I have already discussed (see Section 4 above) the limitations of using the number of factories to capture subdivision. The state-level value added ratio would be affected by changes in industrial composition, prices of inputs and outputs, and taxes which would differ across states. This would make it too “noisy” to capture inter-firm outsourcing. Second, it is possible that some of the high court judgments were stayed while they were under appeal; it was impossible to get any information on this. Finally, in view of the pervasive uncertainty about whether a particular section would be struck down or revived, firms may not have responded to EPL changes in either direction in ways that were costly to reverse. For example, it is costly to reconsolidate a factory after subdividing it; so is the reverse process. As regards the other adaptive measures, setting up a new factory or hiring more regular workers for an existing one are of course more difficult to reverse when EPL is tightened, but even in periods when it is l oosened, closure or replacement of regular by contract workers can also prove costly if judicial reinterpretation with retrospective effect forces employers to reinstate the retrenched workers with back pay.

7 Concluding Comments

In this paper, after summarising my earlier critique of the literature on labour regulation and industrial performance in India, I undertook a detailed assessment of state-level legislative and judicial changes which showed that attempts by earlier authors to quantify state labour regimes were riddled with errors. The earlier literature also overlooked some possibilities regarding the mode and timing of firms’ adaptation to EPL, and also limitations in the data it relied upon. The results of an alternative empirical exercise were however mixed. Given the limitations of the data and methodology, perhaps this was only to be expected. The main contribution of this paper therefore lies in its critique of the tests of the various new hypotheses advanced in this paper, must earlier literature and its suggestion of alternative modes of adap-await another occasion. As in my 2006 paper, I can only conclude tation to EPL. A proper empirical test of the hypotheses advanced that the evidence is inconclusive and my own views on the imby the earlier authors, after correcting their mistakes, as well as pact of EPL remain agnostic.

Notes

1 These terms are used idiosyncratically in the IDA, so definitions might be helpful. A layoff “means the failure, refusal or inability of an employer on account of shortage of coal, power, or raw m aterials or the accumulation of stocks or the breakdown of machinery or natural calamity or any other connected reason to give employment to a workman whose name is borne on the muster rolls of his establishment” (IDA Section 2 (kkk)). Layoffs are limited to 45 days on half pay. R etrenchment means the permanent termination of a worker’s service, other than on account of punishment, retirement, ending of a contractual period, or continued ill-health [IDA, section 2 (oo)].

2 I would now like to add to this list the geographical distribution of public sector manufacturing units and the location, size and diversity of preexisting/existing industrial agglomerations. The importance of both these determinants of industrial expansion has been nicely demonstrated by Chakravorty and Lall (2007) in their application of the “New Economic Geography” (NEG) to I ndia. NEG, like some of the development literature of the 1950s, shows that economies of scale result in cumulative patterns of industrialisation and deindustrialisation which cannot be reversed by small policy changes (see Bhattacharjea 2008 for an introduction). This could partly explain why the BB index of incremental changes in l abour regulation becomes insignificant when state-specific time trends are included in the r egressions.

3 These included earlier versions of Hasan et al (2007), Aghion et al (2008), and Ahsan and Pagés (2009). Hasan et al used a modification of the BB index that has now been used by Mitra and Ural (2008). While I discuss Ahsan and Pagés (2009) in more detail below, as it deals specifically with EPL, my reservations about the other papers that use the BB index remain as they were in Bhattacharjea (2006).

4 I have recently come across another example. After a terse two-line summary of my paper, Djankov and Ramalho (2009: 5) state that “the overall result that rigid labour regulation results in net job losses remain”, giving the misleading impression that this was my conclusion.

5 Excel Wear vs Union of India (1978) 4 SCC 224. 6 See, respectively, K Gurumurthy vs Simpson & Co Ltd, Madras (1981) 2 Labour Law Journal 362-65 (Madras) (DB); K V Rajendran vs Deputy C ommissioner of Labour, Madurai, (1981) Labour and Industrial Cases 799: 805-06 (Madras); J K Synthetics v Union of India, 1984 Labour and Industrial Cases (NOC) 40 R ajasthan. 7 Stumpp Scheule and Somappa Ltd vs State of K arnataka (1985) 2 LLJ 543 (Karnataka). 8 Union of India vs Stumpp Scheule and Somappa Ltd (1989) 2 LLJ 4-6 (Karnataka) (DB). 9 Maulins (I) Ltd vs State of West Bengal (1989) 1 Calcutta Law Journal 338. 10 Jay Shree Tea and Industries Ltd vs Industrial Tribunal (I) Allahabad (1990) Labour and Industrial Cases 1411-21 (Allahabad). 11 See respectively Workmen vs Meenakshi Mills (1992) 3 SCC 336; Papanasam Labour Union vs Madura Coats Ltd (1995) 1 SCC 501; Orissa Textile and Steel Ltd vs State of Orissa (2002) 2, SCC 578. 12 Details of all the amendments are from Radhakrishnaiah (2003). 13 Engineering Kamgar Union vs Electro Steel Castings, (2004) 6 SCC 36. This judgment was based on Article 254(2) of the Constitution of India,

23 This finding stands in sharp contrast to that of Ahsan and Pagés, who found that “industrial dispute laws lead to a decline in the capital/labour ratio within plants” (Ahsan and Pagés 2009: 70), although they were dealing with a much longer time span and did not distinguish between factories above and below the 100 worker threshold. Both studies use the book value of capital reported by ASI, whereas the standard procedure is to use the perpetual inventory method to obtain an economically meaningful figure for capital stock.

24 I thank R Nagaraj for this last observation.

References

Aghion, P, R Burgess, S J Redding and F Zilibotti (2008): “The Unequal Effects of Liberalisation: Evidence from Dismantling the License Raj in India”, American Economic Review, 98 (4): 1397-1412.

Ahsan, A and C Pagés (2009): “Are All Labor Regulations Equal? Evidence from Indian Manufacturing”, Journal of Comparative Economics, 37 (1): 62-75. Earlier version available at http://ideas.repec.org/p/iza/izadps/dp3394.html.

Besley, T and R Burgess (2004): “Can Regulation Hinder Economic Performance? Evidence from India”, Quarterly Journal of Economics, 119 (1): 91-134.

Bhattacharjea, A (2006): “Labour Market Regulation and Industrial Performance in India: A Critical Review of the Empirical Evidence”, The Indian Journal of Labour Economics, 49(2): 211-32. Available at http://ssrn.com/abstract=954908.

Chakravorty, S and S V Lall (2007): Made in India: The Economic Geography and Political Economy of Industrialisation (New Delhi: OUP).

Djankov, S and R Ramalho (2009): “Employment Laws in Developing Countries”, Journal of Comparative Economics, 37 (1): 3-13.

EPW Research Foundation (2007): Annual Survey of Industries 1973-74 to 2003-04 (Vol II): A Data Base on the Industrial Sector in India.

Fallon, P and R E B Lucas (1993): “Job Security Regulations and the Dynamic Demand for Labor in India and Zimbabwe”, Journal of Development Economics, 40: 241-75.

Hasan, R, D Mitra and K V Ramaswamy (2007): “Trade Reforms, Labor Regulations, and Labor-demand Elasticities: Empirical Evidence from India”, Review of Economics and Statistics, 89 (3): 466-81.

Mitra, D and B P Ural (2008): “Indian Manufacturing: A Slow Sector in a Rapidly Growing Economy”, Journal of International Trade & Economic Development, 17 (4): 525-59.

OECD (2007): OECD Economic Surveys: India, Volume 14.

Radhakrishnaiah, P M (2003): The Industrial Disputes Act, 1947, revised and enlarged by S A Chary (Hyderabad: Asia Law House).

Ramaswamy, K V (2003): “Liberalisation, Outsourcing and Industrial Labour Markets in India: Some Preliminary Results” in S Uchikawa (ed.), Labour Market and Institution in India: 1990s and Beyond (New Delhi: Manohar).

Unni, Jeemol and Uma Rani (2008): Flexibility of Labour in Globalising India: The Challenge of Skills and Technology (New Delhi: Tulika).

may 30, 2009 vol xliv no 22