ISBN 92-2-111766-9
ISSN 1020-5322
First published 1999

Download the document in Adobe Acrobat's Portable Document Format (PDF)

Contents

Foreword

It is often argued that differences in employment protection play an important role in explaining differences in labour market outcomes. In particular, the poor employment performance of European countries with respect to the North American one is often attributed to the strictness of employment protection in Europe. Economic theory provides little guidance when assessing this statement insofar as most models show that employment protection tends to negatively affect both layoffs and hirings, job creation and destruction, unemployment inflows in outflows, one effect dominating the other depending on the values of the parameters. It follows that the role played by employment protection in aggregate labour market outcomes is mainly an empirical matter. However, the empirical literature on the macroeconomic effects of employment protection has to rely on highly imperfect measures of the strictness of these regulations. While previous research has circumvented measurement difficulties by using qualitative rankings of Employment Protection Legislation (EPL) stringency, recent developments, notably ongoing reforms of employment protection in most countries and the expansion of non-standard forms of employment, not only have rendered obsolete existing information, but have also called into question the methodological basis for such empirical exercises.

This report illustrates the successes and shortcomings of existing work in light of simple empirical evidence. While EPL rankings developed in the early 1990s are rather strongly correlated with employment stability in the 1980s, more recent evidence indicates that new measurement efforts are called for. What is needed is not only an update of EPL rankings capturing new legal provisions in the various countries, but also measures reflecting the increasing complexity of legal provisions in this areas, their interactions and/or inconsistencies. Against this background, this report develops an important and hitherto neglected aspect, namely enforcement procedures as a source of EPL heterogeneity across countries and over time: available information on this aspect, despite its limited and rough nature, appears highly relevant to recent empirical evidence. Hence policy recommendations should be formulated with caution, and should not be based on indicators available to date. Moreover, governments should try to exploit some of the meaningful linkages identified in the report between EPL and other institutional features.

Gek-Boo Ng
Chief
Employment and Labour Market Policies Branch
Employment and Training Department

Acknowledgements

We wish to thank P. Tergeist from the OECD, C. Vargha and P. Auer for providing detailed comments on a previous draft. Many ILO colleagues from the Headquarters and the field also helped us collecting and interpreting the data. However, any errors of facts or judgement must be laid most firmly at the authors' door.

1. Executive summary

This report studies analytical and empirical issues encountered in the assessment of the influence on aggregate labour markets of legal provisions, institutional arrangements, and jurisprudence in the area of employment protection. It summarises the current state of knowledge in the field, outlines useful research directions, and offers a preliminary discussion of new relevant information.

Section 2 surveys existing theoretical work and empirical evidence on the relationship between Employment Protection Legislation (EPL) and broad labour market performance indicators. (The more complex issue of whether employment-reduction costs can directly or indirectly increase the productivity of employment relations is not directly addressed in this Report. We do note, however, that such effects and/or insurance concerns must have motivated existing legislation, and should be taken into account by any reform process). Formal models of dynamic labour demand do not yield clear-cut implications as to the relation between employment protection and main labour market aggregates, such as employment and unemployment rates and labour force participation. However, these models unambiguously indicate that employment should be more stable and individual employment relationships more durable when employment reduction is costly for employers. The set of rules governing unfair dismissals, layoffs for economic reasons, severance payments, minimum notice periods, administrative authorization for dismissals and prior discussion with representative of unions and/or labour market administrations certainly implies that employment reductions are costly for employers; it does so, however, in ways that are hard to represent quantitatively. Previous research has circumvented measurement difficulties by relating observable labour market performance indicators to qualitative rankings of EPL stringency, rather than to (unavailable) quantitative measures of firing costs. The evidence uncovered by such empirical efforts, while not as univocal as theoretical models would predict, offers much useful information as to the implications of EPL for employment dynamics and its interaction with other institutional and economic features of industrialised economies. But recent developments, notably ongoing reforms of employment protection in most countries and the expansion of non-standard forms of employment, not only have rendered obsolete existing information, but have also called into question the methodological basis for such empirical exercises. The report illustrates the successes and shortcomings of existing work in light of simple empirical evidence. While EPL rankings developed in the early 1990s are rather strongly correlated with employment stability in the 1980s, more recent evidence indicates that new measurement efforts are called for. What is needed is not only an update of EPL rankings capturing new legal provisions in the various countries, but also measures capturing the increasing complexity of legal provisions in this area, their interactions and/or inconsistencies. Reforms of EPL rarely have addressed the whole set of provisions, but addressed only specific contractual types, e.g. have expanded the scope of various kind of fixed-term contracts without reducing the protection of those under permanent contracts. This increasing dualism of labour markets (the coexistence of a large group of workers with low employment security and of a core of workers still protected against the risk of dismissals) and institutional complexity (multiplication in the number of contractual types and ad-hoc provisions) requires substantial revisions of the methodology used in the past to compute EPL rankings. Appropriate indicators should not only try to encompass these various features of EPL, but also take into account their interactions, e.g., the fact that increasing shares of employment under fixed-term contracts may also be a consequence of strict employment protection for regular workers rather than a sign of greater labour market flexibility per se.

Against this background, Section 3 focuses on an important and hitherto neglected aspect, namely enforcement procedures as a source of EPL heterogeneity across countries and over time. The discretion of policy-delivery mechanisms, such as local labour market administrations and the Public Employment Service (PES), and of jurisprudence in interpreting the rule of law is augmented by the increasing complexity of EPL provisions. In other words, too many rules would almost seem to play the same role as the absence of rules because they create many exemptions and legislative vacuums (e.g., because the new contractual types are not fully regulated and norms applicable to regular contracts cannot be readily extended to non-standard forms of employment). If greater institutional complexity endows administrations and judges with more degrees of freedom in the enforcement of EPL, it also increases their social responsibilities vis-à-vis workers and employers. The relevant empirical issue is whether the predominance of enforcement over legislative norms alters the effects of EPL and makes EPL dependent on local labour market conditions, e.g., because judges and PES officials feel that under conditions of severe labour market slack in a region or during cyclical downturns, workers should be more heavily protected against dismissals than in buoyant labour market conditions. Unfortunately little information is available on the enforcement of EPL. We find that available information on this aspect, despite its limited and rough nature, appears highly relevant to recent empirical evidence. Rankings based on the vague notion of "difficulty of dismissals" are more closely related than other available ones to job-termination probabilities, i.e., to the indicators of labour market performance which, on the basis of theory and previous evidence, are most importantly affected by EPL. Moreover, the nature and stringency of EPL enforcement does vary across countries in meaningful ways. Its behaviour over time, however, appears strongly influenced by labour market conditions in the countries for which the relevant information is available. We proceed to analyse in some detail the conceptual and practical difficulties encountered when constructing more refined indicators of EPL enforcement across countries and over time, and we collect in several Annexes much relevant information gathered in the context of this project.

Section 4 outlines the report's policy implications. In summary, the research reported here indicates:

• that EPL is indeed an important determinant of various aspects of labour market performance, and notably of the employment adjustment issues on which our report focuses;
• that while assessing the relevance of EPL is a complex endeavour, since both EPL and its effects are intrinsically multi-dimensional, measurement efforts are far from futile, since meaningful linkages can be identified along many relevant dimensions;
• that theoretical perspectives on the relevant issues deserve to be tested on the basis of improved statistical information
• that there are relevant policy implications of further work, based on enhanced statistical information.

2. Existing work

2.1 The nature and relative stringency of EPL

Legal restrictions on dismissal of redundant employees differ widely across European and American labour markets. The nature of employment protection, however, is similar in all countries. Legislation typically requires that termination of contracts of individual employees be motivated, and that workers be given reasonable notice or financial compensation in lieu of notice. In practice, enforcement is based on the worker's right to appeal against his or her termination in the case of individual dismissals. Rules regarding dismissal of individual employees can interfere with firms' decisions to adjust overall employment levels. Even in the relatively unregulated American labour market, for example, empirical evidence indicates that legal provisions meant to protect individual employees become more binding during cyclical downturns (Donohue and Siegelman, 1995).

As regards collective dismissals, legislation often mandates administrative procedures, involving formal negotiations with workers' organizations and with local or national authorities. The US labour market again offers a useful benchmark. Unemployment insurance contributions of US employers are experience-rated, that is, are higher for employers who have reduced employment in the past. Hence, firms do face cost increases when reduce employment, albeit small ones in relation to those imposed by the more stringent EPL of European countries (Card and Levine, 1994; Anderson, 1993). Some redundancy costs also arise from the Worker Adjustment and Retraining Notification Act (WARN) of 1988 requiring covered firms to provide employees with 60 days' advance notice of plant closures and large-scale layoffs.

Most European countries feature more stringent regulation of individual and collective dismissals. Procedural details do vary substantially across countries, sectors, and time, and so does the economic impact of EP institutions on different labour markets. Quantitative measures can be readily computed for some EPL aspects, such as the number of months' notice required for individual and collective redundancies. Others aspects are more difficult to measure precisely, for example the willingness of labour courts to entertain appeals by fired workers and the interpretation placed by judges on the notion of "just cause" for termination (we analyse these issues in detail in below).

Such problems have been circumvented in previous work taking advantage of the fact that even partial and qualitative indicators of EPL provisions make it possible, when available EPL indicators are positively correlated with each other, to form qualitatively unambiguous cross-country rankings of EPL. Work along these lines found that, in the late-1980s, countries mandating longer notice periods also tended to specify larger redundancy payments and more complex procedures for authorization and implementation of collective dismissals. Such available quantitative measures were also consistent with survey assessments indicators of EPL stringency, especially those made available by the EC ad-hoc Survey of employers (Emerson, 1988 and reported in various issues of European Economy). Bertola (1990) used this latter evidence to form a ranked list of ten industrialised countries. Grubb and Wells (1993) developed rankings for a larger cross section of OECD countries on the basis of various aspects of individual-dismissal regulations for regular contracts, fixed term contracts, and temporary work agency (TWA) regulations. Table 1 reports several such rankings.

Table 1. Ranking of employment protection legislation indicators by 'strictness'

2.2 EPL and labour market adjustment: theory and evidence

This report focuses on the effects of EPL on broad labour market performance indicators, most specifically on theoretical and empirical relationships between EPL and the level and dynamics of employment and unemployment.

We should acknowledge at the outset that, of course, our narrow focus neglects several important and interrelated issues. In many models, and in reality, market interactions are not such as to ensure that laissez-faire employment relationships achieve complete efficiency. Thus, EPL and other institutional features of labour markets can in principle enhance productivity and efficiency. To mention but one example, stable employment relationships can foster investments in job-specific human capital, both by employers and employees. Laissez faire contractual arrangements might theoretically ensure an appropriate degree of job stability, but appropriate contracts are typically hard to draft and enforce, and explicit legislation may give better incentives to the accumulation of such human capital. A compressed wage structure (which, as we argue below, typically accompanies strict EPL provisions) can also foster productive investments, through similar channels: see, e.g., Acemoglu and Pischke (1998) and their references for a review of the role of search frictions, asymmetric information and unobservable training investments in the interaction of labour market institutions and labour.

An even more obvious role of Employment Protection Legislation (in conjunction with wage-equalizing provisions, and with unemployment insurance schemes) is, indeed, the protection of individual workers against unfair labour market developments. The efficiency-enhancing effects of employment stability, or its "protective" role, or more likely both must be at the root of most EPL provision, and are certainly essential for any evaluation of reform prospects. In what follows, however, we shall not discuss such deeper issues, to better focus on the implications of EPL for employment determination rather than on its motivation or on its effects along other dimensions.

Theory suggests a number of implications in this respect (see e.g. Bertola, 1998 for more formal and detailed arguments):

1. Firing costs stabilise employment in downturns but also lead employers to refrain from hiring in upturns for a constant (and any other given) cyclical wage pattern. Hence, more stringent EPL should be associated to smoother dynamic employment patterns.

2. Since EPL has contrasting effects on employers' propensity to hire and fire, its net effect on longer-run relationships between wage and employment levels is a priori ambiguous. It may increase or decrease average employment, depending on such subtle features of formal models as the functional form of labour demand functions, the persistence of labour demand fluctuations, and the size of discount and attrition rates. A general insight holds true: since higher turnover costs reduce both hiring and firing, their effect on average employment levels over periods when both hiring and firing occur is an order of magnitude lower than that on hiring and firing separately. Such issues are studied in some detail by Bentolila and Bertola (1990) and Bertola (1990, 1992), who find that average employment effects are indeed small and of ambiguous sign in reasonable parameterizations of dynamic labour-demand problems.

3. To the extent that firing costs prevent dissolution of existing employment relationships, sharp employment reduction is less likely in countries with stringent job security provisions. At times when employment would increase in the absence of EPL, however, employers are less inclined to hire when they fear that future firing costs shall make it difficult to reverse current decisions. Hence, EPL reduces job creation as well as job destruction, and results in smoother employment dynamics. More subdued turnover implies that individuals who - like new entrants to the labour market - happen to be unemployed at any given point in time are less likely to exit into employment, and more likely to experience long-term unemployment.

In summary, theoretical models suggest that EPL need not bear on medium-and long-term employment and wage levels. Rather they indicate that employment should be more stable and individual employment relationships more durable when EPL is stricter. In other words stringent EPL reduces hirings and firings. It may also affect the character of unemployment experiences.

Empirical work has explored these implications using the above mentioned overall "rigidity ranks" as indicators of EPL, and a variety of cross-sectional indicators of labour market performance.

1. The cyclical volatility of employment is much more pronounced in the United States and the United Kingdom than in Germany, Italy, and especially France. Aggregate wages are ambiguously related to employment fluctuations in all countries (see Brandolini, 1995), and the volatility of aggregate production is similar across industrialised countries (see, e.g., Bertola and Ichino, 1995a). Hence, differential stringency of EPL is relevant to evidence of much wider cyclical employment and unemployment fluctuations in the relatively less regulated labour markets of the US (and of the UK since the 1980s) than in the continental European countries, and especially in France.

2. Markets where EPL is more stringent feature more stable employment and unemployment around levels which, in the long run, are not clearly correlated to the stringency of job security provisions. European unemployment series are closely related to increasing wage trends, but their average long-run level is much less clearly related to their EPL ranking. In formal empirical regressions, EPL indicators are statistically significant but their coefficient is small (see, e.g., Scarpetta, 1996).

3. Unemployment is qualitatively different in labour markets characterised by different EPL stringency. In European labour markets with more stringent EPL, a larger percentage of the unemployed experiences long-term spells of joblessness; many of the unemployed are young labour market entrants; and relatively few are job losers.

Empirical evidence gives some support to theoretical implications as to the behaviour of standard macro-indicators of labour market performance: aggregate employment and unemployment levels are not strongly affected by cross-sectional indicators of EPL stringency, but seem more stable. Other evidence also supports the relevance of job-security provisions. To the extent that hiring and firing are inhibited by EPL, employers have incentives to exploit other sources of (costly) flexibility, such as overtime: indeed, aggregate employment fluctuations are relatively subdued in Europe, but hours per worker are more variable there (Abraham and Houseman, 1994). As argued by Davis and Henrekson (1997) with reference to Swedish and American evidence, labour market institutions and other forms of regulation appear relevant to a host of other empirical features in cross-country comparisons.

Empirical work has also explored the effects of EPL on the stability of employment relationships from a disaggregated perspective, collecting and interpreting evidence on various measures of such stability from the employees' and the employers' points of view (see Box 1 for a definition of such measures, and Table 2 for a selection of empirical results).

Table 2. Job turnover and labour turnover

Box 1. Concepts used in the measurement of disaggregated employment-duration evidence (Davis, Haltiwanger) A number of concepts are used in the analysis of job creation and job destruction. Although some of them are easily understood, meaningful measurement and interpretation of statistics require careful definitions. A job is defined as an employment position filled by a worker. An enterprise (firm) is an economic entity (that encompasses one or more plants). An establishment/plant is a physical location where production takes place. Establishments may be classified into four categories: Openings: establishments with zero employment at the beginning of the period and employment at the end. Expansions: establishments with employment in both periods, with a higher level at the end. Total (1) + (2) = Job Creation (or Gains). Contractions: establishments with employment in both periods, with a lower level at the end. Closures: establishments with employment recorded at the beginning and none at the end. Total (3) + (4) = Job Destruction (or Losses) Job turnover (or reallocation) is the sum of these 4 components (without regard to sign): Job turnover = Job creation + Job Destruction. Put another way, job turnover at time t is the sum of all plant-level employment gains and losses that occur between t-1 and t. Net employment change is equal to Job Creation - Job Destruction. Excess job turnover is the difference between total job turnover and the absolute value of net employment change: Excess job turnover = Job turnover -  Net employment change. Survival rate of new firms: number of firms still present in a given year, as a percent of the total number of new firms at the starting date. Labour turnover (or reallocation) measures the movement of individuals into and out of jobs over a given period. Labour turnover = Hirings + Separations. Put another way, labour turnover at time t equals the number of persons who change place of employment or employment status between t and t-1. It is also the sum of job turnover and flows of workers into and out of existing jobs in establishments or firms. Hirings: one measure of the share of positions in which there has been hiring is given by the tenure of workers (the number of persons with less than twelve months (one year) of tenure corresponds to positions in which at least one new hire has been made during this year) Separations (quits or layoffs): one measure of separations is given by the currently jobless persons who were employed twelve months earlier. Minimum labour turnover equals the larger of job creation or job destruction. It represents a lower bound on the amount of labour turnover required to match job turnover.

Most relevant for the current study's perspective is the fact that measures of labour turnover, especially job-loss probabilities, tend to be negatively related to EPL rankings. The information in column (2) of Table 2 displays transition probabilities computed from surveys of individual workers. These worker-based estimates should in principle offer an accurate measure of turnover intensity, summarised in the table by the "gross" indicator which sums separations and new hires during the sample period as a percentage of average employment levels. Even though these statistics do not distinguish voluntary from involuntary separations, they do give some indication of strong associations between the stringency of EPL and labour market dynamics: in USA and Canada, for example, worker turnover is about twice as intense as in most European countries. Further, the evidence of Table 3 does indicate that job tenures are significantly longer in countries with more stringent EPL, such as Italy and France.

Table 3. Tenure length distribution of existing jobs, 1995

*1991 for Canada
Source: Eurostat, OECD.

The evidence reviewed and illustrated above is usefully interpreted in light of theoretical results, but certainly not particularly strong. In other respects, the evidence does not readily conform to theoretical predictions. Recently, employer-based measures of labour market turnover have become available following Davis and Haltiwanger's (1992) work on US Census data. The statistics of Table 2, column (1) measure "job creation" as the size-weighted percentage employment increase at establishments which are expanding during the sample period; "job destruction" as the similar percentage employment decrease at establishments which are contracting during the same period; and "gross job turnover" as the sum of absolute employment changes over sampled establishments normalised by employment stocks (see Box 1 for further details). There are well-known conceptual and practical problems with this procedure. The definition of an "establishment" is theoretically unclear, as different types of jobs presumably coexist within real-life production units, and empirically ill-defined, as data are available for different definitions of "plants" and "firms" across countries. As discussed in Boeri (1996), the interpretation of these data is also made difficult by statistical problems, such as different frequency of observations and sample composition (especially with respect to the age and size of establishments) across countries.

Gross-job turnover statistics, however, may be viewed as a rough index of labour market flexibility. Recalling that stringent EPL should reduce both hiring and firing, it is quite surprising to find that job turnover statistics are very loosely related to EPL rankings. Table 4 reports some relevant evidence. Perhaps most remarkably, Italian and French estimates at 21 to 24 per cent are not only very large in absolute terms (one every five jobs is either created or destroyed every year), but also extremely close to the US and Canadian estimates despite much heavier regulation of dismissals in the European labour market. EPL does appear more relevant, however, if data refer to continuing establishments only (see Garibaldi et al, 1997).

Table 4. Job turnover: Average annual rates as a per cent of total employment

This can be rationalized theoretically by wage setting institutions since, as noted by Bertola and Rogerson (1997), labour-demand fluctuations are more likely to generate hiring and firing when institutional features make it difficult or impossible for wages to accommodate them. From a theoretical point of view, it is indeed far from surprising that relative wage variation should be heavily constrained in the same markets where job security provisions are most stringent. Quantitative firing restrictions, in fact, could hardly be binding if wages were completely unrestrained and employers could reduce them so as to make stable employment profitable, or to induce voluntary quits. Limiting the freedom offered to employers and workers in setting wages gives force to quantity constraints. As briefly mentioned at the beginning of this section, such labour market institutions may well address important imperfections of laissez-faire market interactions. The combined policies may be rationalised by "equal pay for equal work" principles, or by the belief that freely contracting parties may not be sufficiently rational or informed as to correctly evaluate the ultimate consequences of arrangements that might appear optimal at a particular moment (with, for example, detrimental effects on human-capital investments in training).

Wage and quantity rigidities, however, may also reflect a desire by organised labour to enforce monopolistic wage-setting practices by preventing underbidding by the unemployed. As noted above, firing costs do not generally reduce average employment at given wages; symmetrically, EPL per se need not increase the bargaining power of "insiders" relative to outsiders, since outsiders could and should in principle be able to bid down wages so as to "buy" themselves a job. If contractual arrangements make it possible to do so, competitive pressure on equilibrium wage and employment patterns should make turnover costs next to irrelevant in wage determination in a dynamic labour demand model with ongoing fluctuations. The combination of institutional wage compression and job security provisions is a powerful source of insider power, however, and their apparent association in the data with high wages and low employment is far from surprising.

Some empirical work has also related labour market performance to the time series behaviour of quantitative EPL indicators (even though such indicators are a very partial measure of EPL, rankings are, by their nature, essentially cross-sectional and cannot be used in time-series analysis). The results of such exercises are mixed. Lazear (1990) finds evidence of a positive relationship between EPL and unemployment, but Addison and Grosso (1996) find no significant evidence when using a similar but more precise set of data. This is not surprising, since while theory predicts that a given set of EPL provisions should affect the dynamic behaviour (rather than the average level) of employment, the effects of expected and unexpected changes in EPL provisions are generally ambiguous and certainly different from each other. The theoretical prediction of small average effects hinges on the fact that the firm's current and future actions offset each other along the optimal path. A more stringent employment-protection legislation implies smaller employment reductions in the face of a given labour-demand downturn or wage increase. This does not mean that employment will be higher than it would in the absence of EPL: to the extent that labour-demand downturns or wage hikes are not completely unexpected events, in fact, a rational employer should have hired less during previous upturns, and therefore avoid excessive overmanning during cyclical downturns. This offsetting mechanism can be absent, however, if legislation changes unexpectedly. For example, an unexpected increase in the stringency of EPL should be associated with ceteris paribus higher employment or wages, as firms find it difficult to shed the ex post redundant labour they have hired without expecting employment reductions to be costly or difficult. Symmetrically, an unexpected relaxation of EPL might at least initially reduce employment if firms shed labour taking advantage of currently low firing costs but B fearing future firing costs B still implement restrained hiring policies in the expectation that EPL will be tightened again (Bertola and Ichino, 1995). Available information does not generally make it possible to disentangle these effects, and time-series work on EPL and labour market performance can easily be misled by reverse-causation channels of interaction between the two if, as is likely, weak labour market performance leads political authorities to increase the stringency of EPL. This may have indeed occurred in the 1970s when, before and after the first oil shock, EPL was strengthened by a wave of synchronous restrictive reforms by European countries. In more recent times, the British labour market has undergone a flexibility-oriented institutional transformation. The conservative governments of the 1980s tried and largely succeeded to weaken unions and labour market regulations. Not surprisingly from a comparative institutional perspective, the British labour market's performance is now similar to its American counterpart in many respects (but not all, see Blanchflower and Freeman, 1993).

In summary, empirical work provides mixed results in the evaluation of the influence of labour market regulation on labour market adjustment. Studies using "rigidity rankings" as indicators of EPL does not give clear-cut results. This may be due to the elusive and complex nature of available information, and of the EPL concept itself.

Figure 1. Relationship between EPL rankings and variability rankings of employment growth rates ⁽¹⁾

r_c = rank correlation.
⁽¹⁾ EPL rankings are based on late 1980s information (Bertola, 1990)
Source: "Total Dependent Employment", OECD; data for Germany refer to West Germany until 1995, to the united Germany's growth rate thereafter.

2.3 The need for new evidence

To the extent that it is possible to assess unambiguously (if only qualitatively) the relative stringency of job security constraints, theoretical and empirical work has successfully correlated aspects of labour market performance to EPL indicators. Available EPL measures, however, are based on institutional information dating back to the end of the 1980s, and no longer display cross-country covariation with various aspects of labour market performance in line with the predictions of theoretical models.

For example, Bertola (1990) found a negative correlation between the variance of employment growth and job security rankings using his data, which ranged from the 1960s to the mid 1980s. This is consistent with theoretical predictions if the countries considered are similar in all respects other than the stringency of EPL (in particular, if the dynamic volatility of labour demand and wages is similar in all countries). In diagrams like those of Figure 1, similar to those in Bertola (1990), country-specific points should be aligned along a downward-sloping line if a country with stringent EPL also features low employment-growth volatility: Italy, for example, has the most stringent EPL (ranked 1) and has also a low employment-growth variability (ranked 8 for the whole period, graph 1 of Figure 1).

In the whole period 1969-97, the empirical relationship is indeed negative, but not very pronounced. Interestingly, the evidence displayed in the graph 2 of Figure 1 for the 1969-86 period is fully consistent with theoretical predictions, and with previous findings. In that period, when the countries considered experienced similar shocks (from the oil shocks to restrictive monetary policies), employment volatility is very significantly and negatively related to the stringency of EPL as measured in the mid-1980s.

In the more recent 1985-97 period the correlation is essentially absent instead (graph 3 of Figure 1). This quite possibly reflects more varied shocks across countries (note, for example, the obviously peculiar case of German unification). More importantly for our purposes, however, lack of correlation in the more recent period also might also indicate that EPL rankings developed on the basis of late-1980s information are obsolete. Reforms in the 1970s were synchronized across countries, so the rankings were to a large extent unaffected by these reforms. In the 1980s, and even more so in the 1990s, reforms may have changed the extent to which individual components of EPL covary with each other and, especially, are much more country-specific - consider, for example, the UK sudden and for a long time isolated move from stringent EPL and strong unionization to a largely deregulated labour market. Table 5 summarizes other recent changes in legislation.

Table 5. Major changes in employment regulations

2.4 Directions of progress

In the late 1980s and through the 1990s, European labour markets reforms have been relatively frequent, but piecemeal in character. Such evidence of institutional variability along the time-series dimension calls for economic and political studies of reform processes, rather than of institutions at each point in time. As shown in Coe and Snower (1997), various labour market policies strengthen each other's effects on the labour market's productive efficiency, to imply that comprehensive reforms should be preferred to marginal adjustments. The timing and credibility of reforms are also important in a dynamic and heterogeneous environment, where expectations have an important role and policy reforms affect different groups of market participants in different ways. Saint-Paul (1997a) notes that introduction of more flexible contracts should increase employment along a two-tiered adjustment path: as employers take advantage of new, more flexible hiring opportunities at the same time as they still hoard protected employees, employment increases during the transition to the new steady state. Job security provisions have generally remained in force for standard employment contracts; to decrease overall EPL stringency, less protected (part-time and temporary) forms of employment have been allowed. This presumably reflects the political importance of distribution considerations, and the role of such atypical forms of employment is not always captured by the standard EP measures.

Eligibility requirements have often become tighter for unemployment benefits and, especially, for early retirement and invalidity pensions. Not only EPL but also benefit systems, wage determination, pensions, extent of part-time work matter for labour market performance (see Annex A for a presentation of some of these different schemes and institutions in the selected countries). Some of these aspects are substitutable to each other; others are complementary, as in the already mentioned case of administrative constraints on relative-wage variability. Labour market institutions aim at protecting workers from dismissal and wage losses, and/or at offering unemployment compensation to job losers. Protection from job loss is all the more desirable when only scant unemployment insurance is available, and unemployment insurance is very much appreciated when weak job security provisions make joblessness likely. Indeed job security, notably jurisprudence in favour of the employees, does appear, in some countries, to be inversely correlated to the coverage and level of unemployment insurance (for example, in Denmark, Italy or Spain) or other adjustment tools such as early retirement provisions (see also Auer, 1999).

Ideally, one should try to develop monetary-equivalent measures of the various aspects. This is obviously impossible, but our own analysis below of particularly hard-to-measure EPL aspects (such as administrative requirements and enforcement processes) and of the character and scope of reform processes sheds some light on possible directions of progress in the wider area of interest.

3. The enforcement of EPL

The existing indicators of EPL are ill-suited to track asymmetries across countries and over time in the degree of enforcement of employment protection. Yet, there are several important indications that such asymmetries may be more marked than differences in regulations per se and that they may play a crucial role in affecting the work of labour markets, notably the extent of job loss and the incidence of unemployment.

3.1 Does enforcement matter?

The more or less restrictive interpretation given by the jurisprudence to employment protection regulations - e.g., to the definitions of "just cause" and "justified reason" for a dismissal - and the credibility of the threat to employers of being forced to reinstate workers involved in unfair dismissals, would seem to have been the single most important feature of EPL affecting labour market flows. As recalled in the previous section, economic theory unambiguously predicts that EPL should be negatively correlated with the incidence of unemployment (inflows as a percentage of the labour force) and with the extent of job loss (workers becoming non-employed as a result of dismissals). Table 6 displays correlation coefficients between, on the one hand, various features of EPL and, on the other hand, the two above mentioned flow magnitudes (INCIDENCE and JOBLOSS) for which there is a clear-cut prediction of economic theory. In particular, three sub-rankings of the overall OECD (OECD, 1994) measure of employment protection are provided, which capture respectively:

1. procedural inconveniences (translating into a ranking a qualitative assessment of the complexity of procedures to be activated when issuing a dismissal notice);
2. notice and severance pay requirements (the number of months elapsing between a decision to dismiss and the time at which the notice becomes effective plus the number of months of severance pay to be granted to individuals at different tenures in the firm), and
3. difficulty of dismissal (translating a qualitative assessment of the strictness of the legal definitions of unfair dismissal, the frequency of verdicts involving the reinstatements of employees and of the monetary compensations typically required in the case of unfair dismissals). ⁽¹⁾

Clearly the third feature (difficulty of dismissals) is the one offering the closest approximation of the interpretation given by jurisprudence of EPL. Interestingly, the sub-index "difficulty of dismissal" bears the strongest (and negative, as predicted by economic theory) correlation with JOBLOSS and INCIDENCE. As the correlation between the other two features of EPL and labour market flows is not significant, it would seem that the measure which best captures the enforcement of EPL drives the overall negative correlation between, on the one hand, employment protection and, on the other hand, unemployment inflows and the percentage of the workforce having lost a job.

Table 6. Employment protection and labour market performance (rank correlation coefficients)

Moreover, the role of jurisprudence on terminations of employment contracts in the work of labour market may have increased in importance over time. As discussed in Section 2 and documented in table 5, partial reforms of EPL undertaken in most European countries in the last decade significantly increased the institutional complexity of labour markets. The very fact that such reforms were piecemeal has amplified the duality of labour markets, creating a large segment with short-tenures and low protection which goes hand-in-hand with a stable-jobs segment offering long-tenures, and high employment security. On the one hand, the protection on regular contracts is still there and, on the other hand, a widespread use of non-regular contracts is being made ⁽²⁾. New contracts and, more broadly, atypical forms of employment often do not have a well-defined juridical status and the process defining "charters of rights" for atypical workers is still far from being completed. Under such conditions of increasing institutional complexity (a broader range of contractual types) and legislative vacuum (rights of workers under new contractual types not yet fully defined), national administrations and the labour courts have objectively a more determinant role in the enforcement of employment protection.

3.2 How is EPL enforced?

There is significant cross-country variation in the enforcement of employment protection legislation (some administrative procedures for individual dismissal are presented in Table 7). Some of the cross-country differences have to do with the source itself of regulations. Legal protection against unjustified dismissal is provided:

• by laws of general scope -- such as labour codes, labour acts, or civil codes;
• by specific legislation -- e.g., dealing with individual employment contracts or individual and collective dismissals; and
• by constitutional provisions (in particular regarding non-discrimination or equality of opportunities)⁽³⁾.

While these three layers of legislation do provide employment protection in most countries, collective agreements tend to supplement basic legislative provisions. In some countries -- such as Denmark, ⁽⁴⁾ the United States or Canada -- collective agreements are the main source of protection against dismissal ⁽⁵⁾. However, their scope is often limited to a part of the workforce, with the remaining part being subject to the principles of common law, custom and practice. Indeed, collective agreements are most relevant for collective dismissals, while constitutional or legislative provisions on the protection of human rights and protection against unfair labour practices provide protection in case of individual dismissals.

Depending on the source of the rules, enforcement will rely more on administrative bodies (e.g., labour inspectorates), or on voluntary dispute settlement procedures, or on court proceedings. A common denominator of the various countries is, however, the reliance on case-law, particularly in defining grounds for dismissal; in countries like Italy, jurisprudence plays a very important role in the interpretation of laws and agreements in the case of disputes over contract termination; in Ireland and the United Kingdom, case law is a fundamental characteristic of the legal system.

In all countries the right to lodge complaints is indeed an essential element of a worker's protection: a worker who considers that her/his employment has been wrongfully terminated may present a grievance to an impartial body. In most countries, trade unions may provide assistance to their members or act on behalf and in place of their members. In many countries the worker may appeal to labour courts ⁽⁶⁾ or an arbitration committee or arbitrator. In other countries, it is generally the ordinary courts which are competent to hear appeals against unjustified dismissal.

Yet, the competent body may vary according to different criteria:

• the category of worker (private or public sector; white-collar or blue collars);
• the type of dismissal (with or without notice, with or without prior authorization) or
• the provisions invoked (labour law, civil law, basic rights ⁽⁷⁾, rights resulting from a collective agreement).

For countries in which the provisions or practices do not require justification in all cases of dismissal, the situation varies depending on whether the worker may invoke specific forms of protection, the clauses of a collective agreement or the principles of common law: in the United States for example, which has not adopted general legislation against wrongful dismissal, proceedings based on common law concerning wrongful dismissal are brought before the courts; employees covered by collective agreements may use arbitration procedures before private arbitrators (after internal grievance procedures), whose decision is final; if protected by the National Labour Relations Act, American workers may also present a complaint to an administrative law judge; yet, other dispute settlement procedures are still available for workers protected by other statutes (for example, they may present a complaint before the Equal Employment Opportunities Commission, EEOC, for discriminatory actions at the workplace). It may also be the case that, depending on the nature of the complaint, the complainant may choose the appeal procedure he or she considers more appropriate: in Ireland, the worker may apply first to a Rights Commissioner, then to the Employment Appeals Tribunal for settlement, or refer directly to the latter. Finally, a worker may generally, after the first instance, refer the dispute to the appeal bodies, such as the appeal courts or the Supreme Court. This right to appeal must generally be exercised within a certain time limit. Such time limits should not be too short to allow the workers to find out their rights (in particular when they are not helped by any representative). Yet, deadlines for challenging dismissal or for taking legal proceedings vary considerably from one country to another. In some countries, the courts may even allow an appeal after the time limit if there is a valid reason for the late application (e.g., in the United Kingdom).

According to national laws and practice, provision may be made for recourse to a procedure of conciliation before or during appeal proceedings against dismissal. Conciliation gives each party an opportunity to review, in presence of a third party, the question of justification of the dismissal, to assess the probability of winning or losing the case before the competent court and the possibility of reaching an agreed solution (withdrawal of the complaint, reinstatement or compensation). This enables the number of cases to be heard by the competent court to be reduced. Again, it is crucial to understand institutional differences across countries; differences may arise from: a) the role and timing of the conciliation procedure: it may take place before or instead of a contentious procedure (for example, conciliation is a stage of the appeal procedure, in France and Germany); b) the nature of the conciliation procedure (it may be compulsory, for example in New-Zealand or voluntary, with sometimes differences within the country: in Italy, for example, conciliation is compulsory prior to the proceedings, in enterprise with less than sixty workers while it is voluntary within the framework of judiciary dispute settlement); c) the conciliation service may be provided by the State, by an agency independent of the State but funded by the State such as, the Advisory Conciliation and Arbitration Services (ACAS) in the United Kingdom, the Federal Mediation and Conciliation Services (FMCS) in the United States or the Australian Industrial Relations Commission (AIRC), or by private institutions: for example the American Arbitration Association (AAA). The importance and the nature of the conciliation procedure may give information on the uncertainty and the costs involved in the process: as conciliation is a low risk process, usually expeditious, generally with a very high rate of compliance with the outcome, parties may prefer it -when choice is available-to court's settlement. Data on conciliation are therefore undoubtedly important, but quite scarce. So it is important to interpret the data in light of the different litigation procedures, since differences may affect the outcome of the process.

In many EU member states the burden of proof is incumbent to the employer, which is in line with the rationale of the requirements of justification or grounds for the employer's decision to dismiss; in other countries, it is placed on the complainant or on neither the employer nor the worker. Again, this may differ within a country, according to the category of worker involved or the reason invoked for dismissal ⁽⁸⁾.

3.3 Measuring the enforcement of EPL

Once established that enforcement matters, notably that jurisprudence on unfair dismissal is often more important than the nominal strictness of regulations per se, it still remains to decide how to properly measure this crucial feature of employment protection. This task is complex for at least three reasons.

First, there is little information on the jurisprudence concerning the termination of employment contracts. As discussed more in detail in Annex A, some data are often (though by no means always) available from administrative records on the number of cases brought before the competent tribunals and on court rules. Even if labour court cases are not necessary the best indicator of enforcement, they constitute a good proxy variable. Yet, like all administrative statistics, such data are affected by changes in regulations: hence, their meaning changes along with the features they are supposed to measure. To give an example, the coverage of labour disputes offered by administrative records is likely to increase when arbitration services are offered by national administrations, as there may be a stronger incentive in such a case to notify the existence of a labour dispute from both parties. Moreover, data are affected by confidentiality rules, especially when employees choose to use private arbitrators for litigation procedures, as in the United States. Finally, information is sometimes not centralised especially in countries -- such as Australia, Canada or the United States -- where several institutional levels interfere (provincial, state, federal) and/or different sources of information should be referred to (ministries, labour dispute settlement institutions, etc.). For instance, there are no systematic sources available in the United States.

Second, such statistics are seriously affected by selection bias. For example, only relatively clear-cut cases may be brought in court, to imply that data sampled from court records are not representative. Selection bias operates in directions which are often not predictable even on the basis of in-depth analyses of procedural obligations to be fulfilled before appealing to courts in the various countries; of existence of impartial bodies specialised in appeals; of the extent and effectiveness of litigation procedures; of the average delay between the start of the procedure and the verdict; and of the likelihood of rules favourable to employees.

For instance, when conciliation and arbitration procedures are compulsory, court ruling may be rather infrequent and hence provide a very inaccurate basis to measure the costs of unfair dismissals. The issue is that procedural obligations vary substantially across countries and sometimes within each country. Some countries give more importance to procedures to be followed before the contract termination while others rely on workers' claims against the termination. To give a few examples, the right to defence of the worker before contract termination does not apply in Germany or in the United States, which therefore seem to rely much on rulings after termination. However, in Germany the work councils must be informed and consulted before termination of employment. This is also the case in Austria, where work councils may also take legal action. In some countries, the employment of a worker should not be terminated, unless the employer has given the worker appropriate (repeated) warning(s), generally in writing and specifying the reasons for dismissal. Clearly, in all countries relying on procedures intervening before the termination of contracts, any information on labour disputes settled by conciliation of the parties is hardly available. Finally, the likelihood of rules favourable to employees may induce the employers to go for conciliation prior to referring to the court, if not to give up altogether to the dismissal, in which case court rules are also not representative of the actual costs of unfair dismissals ⁽⁹⁾.

Third, jurisprudence may be affected by the underlying labour market conditions. There is evidence, for example, that in western Germany court rules have been particularly unfavourable to employers during downswings as if jurisprudence was playing the role of a stabiliser (Berger, 1997). By the same token, there are some indications that court ruling has been more on the part of employees in the high unemployment Mezzogiorno than in the northern part of Italy (Ichino et al., 1997). Put in another way, jurisprudence may not be fully exogenous, but may be itself affected by labour market conditions. Two crucial dimensions likely to affect the verdict - namely the costs of job loss for the employees and the cost of reinstatement for the employer - may deeply be affected by the way in which the labour market operates. The more dynamic the labour market is, the easier is it to find another job after the dismissal, the less weight is (more or less implicitly) given by the Courts to the case of the worker. Another way in which the labour market affects jurisprudence is via the degree of unionization of the workforce. The stronger the union within the firm, the more likely that disputes are solved before going to the tribunals.

Thus, in principle, when using administrative records one should control for cyclical and regional labour market conditions, the degree of unionization of the workforce and institutional features related to the way in which labour courts operate.

Surveys of employers and employees are likely to offer better measures of the enforcement of EPL. Employers form expectations about the actual costs of dismissals whenever they decide whether to layoff or not a worker. The first question that they are likely to ask themselves is how much such an action will ultimately cost to the firm. Similarly, a worker has to weight the pros and cons of going to the court rather than accepting the offers made by the employer in the course of the pre-litigation procedures. Unfortunately, we are not aware of the existence of surveys designed to offer such information. The European Commission's ad-hoc questionnaire asks employers to provide a qualitative assessment of the overall costs of dismissals. No monetary assessment are required (e.g., questions on the willingness to pay of employers to avoid going before a tribunal). Another important survey on these issues - the survey administered by the ISR (International Survey Research) - collects information on employees' perceptions about job security, but again no question specifically deals with jurisprudence and there is no attempt to discern evaluations of the likelihood of favourable outcomes of labour disputes taken to courts.

3.4 Preliminary evidence

With the above caveats in mind, Table 7 provides information on the number of cases brought before the tribunals and on the percentage of verdicts favourable to the workers in the countries for which data are available. Standard indicators of the strictness of EPL and measures of the coverage of unemployment benefits are also reported on the last three columns on the right-hand-side in order to ease the interpretation of data.

Data on jurisprudence come from national sources collected by ILO correspondents. In particular, a questionnaire was sent (see Annex B) to experts operating in thirteen OECD countries (mostly European countries, the United States, Canada, Australia and New Zealand). The most striking fact revealed by the table is the very wide cross-country variation in the number of cases brought to courts. In Spain 1 employee in 200 hundred appealed to the courts in 1995 compared with 1 employee in 15 thousand in Austria. There is also significant cross-country variation in the percentages of cases won by the workers with Spain again and France at one extreme and Ireland at the other. Cross-country differences of a similar magnitude can only come from a variety of factors, and some of the most serious ones may simply be related to the coverage of statistics. Yet, it is tempting to try to make some inferences on economic factors which may also have played a role in the cross-country variation in the incidence of jurisprudence in labour disputes.

Significantly, the countries where tribunals are the most frequently involved in labour disputes arising from the termination of contracts tend also to be those to have the highest percentage cases favourable to employees. Spain is a case in point. Here almost 72 per cent of cases were in 1995 won by the workers compared with less than 50 per cent in North-American countries and a low 16 per cent in Ireland, ⁽¹⁰⁾ all countries where tribunals seem to intervene rather infrequently in labour disputes concerning contract termination. The high incidence of judicial procedures in France may also be partly explained by a large share (74 per cent) of cases favourable to the workers ⁽¹¹⁾. As argued above, the likelihood that the court rulings are favourable tends to play an important role in inducing workers to bring their case to the courts, although it may, on the other hand, also encourage employers to reach extra-judicial agreements. Insofar as available statistics on cases brought to courts well capture the first steps of a judicial procedure, we would, however, expect appeals to courts to be more frequent in the country with a tradition of sentences favourable to the employees and this is consistent with the evidence produced in Table 7.

Table 7. The role of tribunals in the enforcement of EPL: Preliminary evidence (1995 data unless otherwise specified)

Rather vague legal definitions of unfair dismissals, attributing many degrees of freedom to Labour Courts in interpreting employment protection regulations, may also have been an important factor behind the quite impressive caseload of Tribunals highlighted by Table 7 in countries like France, Germany and Spain. In Germany, the Protection Against Dismissal Act of 1951 gave much discretion to Labour Courts in interpreting existing regulations, quite in striking contrast with the German legal tradition which is often placing strong restrictions on the discretionary power of judges (Berger, 1997). Indeed Labour Court activity would seem to have declined in more recent years, due to reforms of this Act, which reduced the degrees of freedom of Labour Courts, e.g., by establishing that the applicability of social plans (allowing for rather generous compensation of workers made redundant) were to be related to objective parameters like the size of firms and the number of employees affected by the redundancies. Similarly, in Spain the extension and clarification of definition of unfair dismissals (allowing, inter alia, for economic, technological, organizational as well as cyclical factors to be considered as reasons for "justifiable dismissals") has been de facto provided only with the labour reform of 1997; before that date, Labour Courts had more discretion than in other countries to set arbitrarily when the grounds for an unfair dismissals were met. Unfortunately, available EPL rankings consider legal definitions only along one-dimension, namely whether or not workers' capability is considered a valid ground for dismissals, e.g., as in the ranking displayed in the third column of Table 7. The above argument suggests that the precision, the transparency and the consistency used by legislators in defining justified reasons for dismissals are also very important (see Tables 1 and 2 in Annex B).

Sanctions applicable to employers in the case of unfair dismissals are also candidates to explain the cross-country variation in the incidence of court rules, although the toughness of sanctions may play a twofold role: on the one hand should encourage workers to appeal (in which case the caseload increases), but, on the other hand, it should also encourage employers to reach extra-judicial agreements before the worker's appeal (which should play against an involvement of Tribunals in Labour Disputes). The fourth column of Table 7 summarises information on perhaps the strongest deterrent to dismissals for employers, namely the option offered to the employee to request the reinstatement in the firm. In many countries, the employer may be empowered to refuse reinstatement, choosing instead to pay compensation ⁽¹²⁾; in some other countries, the competent body can decide to award both reinstatement and financial compensation simultaneously. The information provided in the table is, however, based just on national regulations, while it should be noted that the effectiveness of the reinstatement depends on several factors, such as the length of proceedings (if there are years of delay between the start of the procedure and the verdict, reinstatement probably does not occur), the size of the undertaking (being reinstated in a larger enterprise is easier than in a small unit), and on the sanctions on employers who do not reinstate the worker. There are also moral damages, as distinct from compensation, which may be awarded in some countries: for example, in New Zealand, compensation can be paid for "humiliation, loss of dignity and injury to feelings". When the body is free to set the amount, it plays a particularly important role in determining the criteria to be taken into account: in the United States, discriminatory reasons may, for example, involve high compensation. So, the criteria on which the choice between reinstatement and financial compensation is based vary widely as well as the amount of monetary sanctions. The decision may refer to legislation, which sometimes details a list of cases in which reinstatement can be ordered or proposed (the application of the principle of reinstatement is particularly important for staff representatives), or a list of cases in which it is not feasible or appropriate; decision may also be established by way of collective agreement.

Another factor which is likely to affect - this time unambiguously - the incidence of Courts rulings relates to the costs of legal procedures. A combination such that prevailing for many years in Germany (Berger, 1997) of low court fees, possibility to call unions before the courts, and low legal costs charged on the defeated party tends to clearly play in favour of a stronger involvement of Tribunals in labour disputes.

Finally, the coverage offered by unemployment insurance may play a role in affecting court rules. The last column on the right-hand-side of Table 7 documents that the countries with a high percentage of cases favourable to workers tend also to be characterised by a low coverage of unemployment benefits (beneficiaries over unemployed according to the Labour Force Survey definitions). Conversely, in countries like Ireland and the UK where courts are favourable to employees in significantly less than half-of-the cases, unemployment insurance covers a very large portion of the unemployed. These are just hypothetical questions; more data points would be needed to assess the correlation between these variables.

The above suggests that there is much that we can learn from the cross-country variation in the degree of involvement of Tribunals in labour disputes. Other relevant information on the enforcement of EPL can be obtained by analysing the time-series properties of data on jurisprudence. Unfortunately this was possible only for two countries, namely Germany and Spain.

Figure 2. Total closed labour court cases, unemployment rate and growth rate of GDP, Germany (1970-1993)

Figure 2 plots the series on the number of closed labour court cases as a proportion of the labour force in Germany in the period 1970-93. The unemployment rate and the yearly growth rate of GDP are also displayed in the top and bottom panel, respectively. The striking fact is the very marked covariation of the incidence of jurisprudence and of unemployment. Cyclical conditions, captured by GDP growth rates, would seem to be relevant only insofar as they significantly affect unemployment.

Figure 3. Total closed labour court cases, total cases favourable to the employee and unemployment rate, Spain (1980-1997)

Comovements of indicators of jurisprudence and unemployment can also be observed in Spain, especially when the focus is on cases ended with sentences favourable to the employees. All this suggests that labour market conditions play indeed an important role in the magnitude and nature (more or less favourable to employees) of jurisprudence. In other words, the evidence points to an endogeneity of jurisprudence which should perhaps be duly acknowledged when assessing the causal relation between EPL and unemployment.

4. Conclusion and policy implications

The effects of EPL, while fairly clear from a theoretical standpoint, are difficult to study in practice because of the elusive and complex nature of available information, and of the EPL concept itself. Hence, policy recommendation should be formulated with caution, and should not be based on the largely unsatisfactory information available to date. In particular, available rankings of employment protection cannot be used for the purpose of surveillance of structural reforms in the labour market area, as envisaged in the European context of the Luxembourg process. A broad conclusion of the report is that simplification of rules in this area is needed for Governments to regain control over the enforcement of EPL. If a greater complexity of rules is a by-product of political constraints, notably the unfeasibility of comprehensive reforms of EPL, then precise guidelines should be given to administrations concerning enforcement, and legislative vacuums concerning the new contractual types should be possibly avoided. Moreover, Governments should exploit the interactions between EPL and other institutional features. The documented negative correlation between, on the one hand, strictness of employment protection and, on the other hand, coverage of unemployment insurance, suggests for example, that the extension of the coverage (which does not necessarily imply an extension of the duration!) of unemployment insurance could ease reform of EPL. It would also reduce pressures on judges deriving from the social responsibility they feel when processing labour disputes in areas and time-periods characterised by high levels of unemployment. Similarly, decentralization in wage setting allowing employers in firms facing transitory adverse shocks to adjust wages or hours of work rather than reducing the workforce, could also make it easier to reform EPL and reduce the resistance of policy delivery mechanisms to the liberalization of labour markets. It should be finally stressed that the costs of labour mobility are not only related to EPL, but also to the adaptability of the workforce to different tasks, the infrastructure supporting regional mobility, etc. Thus reforms of EPL cannot substitute for reforms of the education systems, enhancing the fungibility of school curricula, as well as for the build-up of networks of training (and retraining) providers and for measures promoting greater regional mobility of workers.

Notes

1. The methodology behind the OECD ranking of employment protection is discussed in detail in Grubb and Wells (1993).

2. There are indications that this duality is increasing in the strictness of the protection of regular contracts. If one plots the share of employment under temporary contracts vs. measures of the protection for regular contracts for the countries which have reformed their EPL, one gets a strikingly positive association between the two measures. This suggests that countries which liberalised temporary contracts have had a growth of temporary employment that is proportional to the rigidity of the employment protection on regular contracts (see Boeri, 1999: "Enforcement of employment security regulations, on-the-job search and unemploymnet duration" , which shows that protection of regular contracts gooes hand-in-hand with duality in the tenure distribution: the first two columns of table 3, for example, indicate that the share of fixed-term contracts in total employment is positively correlated with the duality of the tenure distribution).

3. This, clearly, makes it very difficult to get a full picture of EPL. For instance in Germany, dismissals are regulated by the "Protection against Dismissal Act" (1969), the Civil Code, various individual Acts and the Works Constitution Act (1972). Moreover, there are approximately 4500 framework collective agreements, 13000 agreements for most branches and regions and 3200 works agreements dealing with this issue (ILO, 1995).

4. Main agreement between the Danish confederation of Trade unions, LO, and the Danish Employers Confederation, DA.

5. The US have not adopted any general legislation against unjustified dismissal and there is no general statutory prohibition against unfair dismissal in Denmark.

6. This is the case of France, Germany, New-Zealand and the UK.

7. In the US appeals against discrimination are very frequent.

8. For example, in Quebec, two types of appeal are available: in the case of "wrongful dismissal" (pratique interdite), burden of the proof is placed on the employer, but not in the case of "just cause" (cause juste et suffisante).

9. See Priest and Klein (1984), Eisenberg and Farber (1996) for a thorough discussion of these forms of self-selection.

10. First year unfair dismissal cases referred was only 1977.

11. This is likely to be the case of Germany where anecdotal evidence suggests that most cases are won by the workers. Unfortunately, it was not possible to collect statistics on the number of cases won by workers in this country.

12. For example, in Italy, this choice depends on the establishment size and on the process of dealing with the dispute.

13. In the UK, under the Employment Protection Act 1978, workers employed between eight and sixteen hours a week must work for the same employer for five years to get protection against unfair dismissal, rights to compensation or appeal against unjustified dismissal, bef