*' ACTIVE LABOUR MARKET POLICY AND UNEMPLOYMENT -- A FRAMEWORK FOR THE ANALYSIS OF CRUCIAL DESIGN FEATURES by Lars Calmfors Institute for Internationa] Economic Studies, Stockholm University* £ I am grateful for comments from Susanne Ackum Agell, David Coe, Nils Gottfries, Abrar hasan, Richard Jackman, John Martin and Johnny Zetterberg as well as from the participants in seminars at the Swedish Employers' Federation, the Swedish Labour Market Board, the Department of Economics ^Uppsala University, the Institute for International Economic Studies at Stockholm University, and at the joint meeting of the EFTA Economic Policy Committee and DG 2 of the EC Commission in Brussels, October 1993. Joanna Gurney and Helena Matheou provided valuable secretarial assistance. ABSTRACT This paper assesses the role of active labour market policies (ALMPs) as an instrument for fighting persistent unemployment. An analytical framework is developed to examine the effects of these policies on a number of economic outcomes, including: job matching; labour force participation; competition between labour market insiders and outsiders; displacement, deadweight and Subsumtion in the labour market; productivity; and tax effects. While some of the effects converge, others work in opposite directions. The paper develops a framework to assess the net effects and examines available empirical research in this perspective. It concludes with a discussion of key design features of ALMPs which can help improve the positive labour market effects while limiting the negative effects. The role of different types of ALMPs, compensation levels, targeting, co-ordination with unemployment insurance and other social programmes are explicitly examined. RÉSUMÉ Ce document évalue le role des politiques actives du marché du travail en tant qu'instrument de lutte contre un chômage tenace. Un cadre d'analyse y est développé afin d'examiner les effets de ces politiques sur un certain nombre de résultats économiques tels que: ľadéquation de l'offre et de la demande de main-d'oeuvre ; le taux ďactivité de la population active ; la concurrence entre ceux qui ont un emploi et ceux qui en cherchent un ; les suppressions d'emplois, les effets d'inertie et de substitution sur le marché du travail ; les effets sur le plan fiscal. Bien que certains effets aillent dans la méme direction, d'autres divergent nettement. Le document élabore un cadre qui permet ďévaluer les effets nets et analyse les etudes empiriques réalisées dans ce domaine. II se termine sur un examen des principales caractéristiques des politiques actives qui pourraient améliorer les effets positifs sur le marché du travail et en limiter les effets négatifs. Le role des divers types de politiques actives du marché du travail, les taux de remplacement, le choix des groupes-cibles, la coordination entre les regimes ďassurance-chômage et d'autres progŕantmes sociaux y sont également examines. ACTIVE LABOUR MARKET POLICY AND UNEMPLOYMENT -- A FRAMEWORK FOR THE ANALYSIS OF CRUCIAL DESIGN FEATURES TABLE OF CONTENTS THE BASIC ANALYTICAL FRAMEWORK Sags I. DIFFERENT EFFECTS OF ACTIVE LABOUR MARKET POLICES 3 A- Effects on the matching process 3 B. Effects on the labour force 4 C. Competition effects for insiders 5 D. Deadweight and substitution effects 6 E. Reduced welfare losses for the unemployed and crowding-out effects 7 F. Productivity effects 7 G. Work-test effects ; g H. Tax effects jg I. General-equilibrium effects on other policies 10 II. THE NET EFFECT OF ACTIVE LABOUR MARKET PROGRAMMES 11 A. Existing empirical research 12 B. Problems of interpretation 13 V. CRUCIAL DESIGN FEATURES 13 A. Compensation levels 14 B. The extent of targeting 14 C. Types of programmes 16 D. The co-ordination with unemployment insurance and programme duration 19 /. WHAT CAN WE EXPECT FROM ACTIVE LABOUR MARKET POLICY? 20 'ABLES 22 IIBLIOGRAPHY 27 ÜOTNOTES 34 In recent years, there has been a growing interest in so-called active labour market policy as a means of fighting the persistent unemployment in Western Europe. This is easy to understand in view of the disillusionment with more aggregate policies: on the one hand, traditional demand stimulation runs the risk of primarily increasing inflation with only small effects on employment; on the other hand, supply-side structural reforms in order to remove various labour market rigidities have either been difficult to implement or appear to produce results very slowly. In this situation, active labour market policy seems to be regarded by many as the deus ex machina that will ^ provide the solution to the Unemployment problem. It is the purpose of this paper to try to make a realistic appraisal of this option. \ One problem when analysing active labour market policy is that there are i/r/'i> <-' different interpretations of the concept. Here I shall use a narrow definition: measures pi' :l- n, largetted to the unemployed with the aim of improving the functioning of the labour •;,/,, J'j,/: market. Active labour market policy will then comprise three basic subcategories: (i) job ■0 ^broking ,with the purpose of making the matching process between vacancies and job ■•"''\ seekers more efficient; (ii) labour market training in order to upgrade and adapt the I skills of job applicants; and (iii) direct job creation, that may take the form of either I public-sector employment or subsidisation of private-sector work. Another complication is that each type of labour market policy measure may work through several different channels. It is another aim of this paper to try to structure the various effects with the help of a simple analytical framework and to use that to highlight the crucial determinants of policy effectiveness. The discussion is organised as follows. Section I presents the basic analytical framework. Section II discusses the various types of effects of active labour market policy, whereas Section III discusses the possibilities to draw conclusions on the net effect of active labour market policy from existing empirical knowledge. Section IV focuses on the importance of various design features. Section V presents some conclusions. I. The basic analytical framework The traditional way of discussing labour market policy in macroecqnomic analysis has been as a method of "cheaüngjLhe Phillips curve", i.e., to improve the inf^iojv^nemplr^j^e.nt^adej^f3-tfiělreby lowering the equilibrium (natural) rale of unemploymeüT(see, e.g., Baily & Tobin (1977) or OECD (1990, 1993a)). An analysis within such an expACtationsraugmented PWllips-curyc (natural-rate)__framework,. however, tends to become ad hoc in nature, since the fundamental determinants; of the —v/ equilibrium rate of unemployment are usually left unspecified.) What is required is instead an explicit theory of how labour market equilibrium is determined. This paper will take as its starting point the type of labour market analysis developed in, e.g., Layard & Nickell (1986), Johnson & Layard (1986) and Layard et al. IF (1991).' The basic reasoning is illustrated in Diagram 1. The downward-sloping curve is an employment schedule relating employment (assumed to be equal to labour demand) measured as a proportion of the labour force to the real wage.' The upward-sloping V curve is 3 wage-setting schedule, which shows how higher aggregate employment causes pressure for higher reál wages. Such a relationship can be derived from several wage-scliing frameworks: monopoly-union, efficiency-wage and bargaining models/ One argument is that union incentives to avoid wage increases causing lay-offs arc weakened when employment is high and it is easy to get a new job, another that employers have to n.iv mnrf m rnnmi-iľ fur lnhnnr anrl pvlrnrt nntimal p.ffnrl from (heir work forces in Such a umanuli, run employment ot ine laoour lorce is indicated by a vertical line. The equilibrium values of employment and the real wage are obtained as the intersection of the two curves (point A). The amount of involuntary unemployment — on the part of individuals -- can be read off as the horisontal distance between the equilibrium point and the full-employment line. To the above diagram one can — in the spirit of, e.g., Pissarides (1985, 1990) or Blanchard & Diamond (1989) — append the Beveridge curve, which can be interpreted as a measure of the effectiveness_o£TEěTň¥tcKni|nprocess between vacancies^and unemployed. More precisely, the Beveridge curve in DidgTam'2 shows the equilibrium relationship between vacancies and unemployed, along which Wrings exactly match quits, so that employment stays constant. More vacancies will in general be consistent with lower unemployment, since the extra hirings due to more job vacancies need to be offset by fewer job matches due to a smaller number of job applicants if employment is to stay constant. 3 In equilibrium, the unemployment rate in Diagram 2 must, of course, be the one implied by the intersection of the wage-setting and employment schedules in Diagram J. To analyse active labour market policy, this model set-up has to be slightly modified. The reason is the need to distinguish between participation in labour market programmes and regular employment. ""For tffiš'"puŕpô^ and wage- šetEňg schedules arFinstead drawn with regular employment (excluding participation in programmes) on the horisontal axis in Diagram 3ý Participation in training and job creation programmes (as a fraction of the labour force) is measured by the horisontal distance ra between the full-employment line FE and the vertical line RR, showing the proportion of the labour force that is not in programmes. (Programme participation may in principle also include subsidised private-sector jobs that are perfect substitutes for regular jobs from the point of view of the individual employee). Since the intersection between the employment and wage-setting schedules at A now determines the rate of regular employment (as a fraction of the labour force), it follows that the distance u0 between the equilibrium point and the RR-curve measures open unemployment.^ The Beveridge curve also needs to be modified (see Diagram 4). The horisontal axis now measures the total number of job searchers without a regular job, i.e., the sum of openly unemployed and participants in labour market programmes. The vertical axis indicates vacancies for regular jobs. Two implicit assumptions lie behind this illustration. The first is that regular job openings can be filled either from tíiestockjof openly unemployed or from ťhé^TôcklĎf^ográjiími participants.^ The second is that ihejňlítčTnngj^^ much simpler than in the regular job market: the labour market authorities do not usually have to post vacancies"But can directly place unemployed workers in various programmes. The diagrams can be used to illustrate the various effects of labour market programmes. Increased placement in training or job creation schemes can be depicted in Diagram i as a leftward shift of the RR-line (programme participation increases by Ar). ' If nothing else were to happen, the effect would simply be to reduce open unemployment by a corresponding amount (from u0 to u,). This will be referred to as the gross (or bookkeeping) effect of such án expansion of programmes. However, to obtain the Hěí" "effect itis~obviously necessary to lake a number of indirect effects into account. One must analyse how theincentives for wage setting, regular labour demand and labour supply are affected. This is done below. The emphasis will be on how regular employment and the total number of job searchers without a regular job (the sum of open unemployment and programme participation) is affected. The idea is to capture to what extent the gross impact on open unemployment is reinforced or counteracted by the indirect effects. II. Different effects of active labour market policies The analysis will distinguish the following types of effects of active labour market policies, which can be considered additive to each other: (i) effects on job matching; (ii) effects on labour-force participation: (iii) effects on competition in the labour market; (iv) deadweight losses and substitution effects; (v) displacement effects i via wage setting; (vi) productivity effects; (vii) work-test effects; (viii) general- I equilibrium tax effects; and (ix) repercussions on other policies. _J A. Effects on the matching process A traditional rationale for active labour market policy has been to facilitate the matching process in the labour market. This may occur through a number of mechanisms; (i) mismatch between different sub-markets for labour may be eliminated to the extent that the quaiificationsof job searchers can be better adapted to the structure of labour demand; (iijTnore active sejrch^bejiaviour on the part of job seekers can be promoted; and (iii) laboüiFlnarket progrjmimes can substitute for regular work experience in reducing employer uncertainty about the employabiTity of job applicants. The impact of improved matching is to reduce the number of job searchers for a given number of vacancies, i.e., to sjúftjh^^everidge curve in Diagram^Xo the left. This is likely to affect both the employment and the^wage^sěfting schedules in the Layard-Nickell diagram (see Diagram 5). Since vacancies become filled more quickly, they become less costly to firaisJjujdjiLoiejMc.aBcJpljare opened. This is equivalent to an increase of labour^demand, i.e., a rightward _shifl of the employment schedule (CalmfčTOTĽang (1993)). Moreover, to the extent that posting vacancies and offering high relative wages can be seen as substitutes for each other in the hiring process of the individual firm, an increased matching effectiveness weakens the incentives for individual employers to attract labour by pushing up wages. This tends to shift the wage-setting schedule downwards (Johnson & Layard (1986)). As can be seen, both effects work in the direction of increasing regular employment (from A to B), although the realrwage effect is ambiguous.^ ".....'" One should expect intensified placement services that help increase the search effectiveness of job seekers to have unambiguously positive effects on the Beveridge curve in Diagram 4. With_placemeniLj.Q^traJj]mg^oxJob_cjcejU4oji_scJ)ejries, there are, however, also effects working in the opposite, direction, since search intensity is likely to be^reduced during the period of actual programme participation. Edin (1989), Holmlund (1990), and Edin & HoímJuhd (199T) have shown how participants in Swedish relief work schemes appear to search no more intensively than the regularly employed, and that the re-employment chances for those still participating in programmes appear Jower than for the openly unemployed. To get the net effect on job matching, one obviously has to subtract negative locking-in effects on individual search behaviour during programme participation from the potentially positive treatment effects once programmes have been completed. One must also take into account that the prospect of being placed in a programme may reduce search effectiveness already before placement. This risk is greater, the more attractive are the programmes offered. It will not be captured by microeconomic evaluation studies comparing re-employment probabilities of the openly unemployed and earlier programme participants: on the contrary, these may show large differences precisely because of such ex ante effects on search behaviour. V On the whole, micro studies of how the job chances of specific target groups have been affccled by various labour market measures provide little information on the overall effects on job matching, since improved/opportunities for one category may have been bought at the expense of others. Nor are most of the available Beveridg'e-curve studies at the macro level very helpful. The reason is that they have usually focused on the relationship between open unemployment and vacancies (e.g., Jackman el al. (1990), Bourdet & Persson (1990), Layard et al. (1991), OECD (1992, 1993a), and Lehmann (1993)). It is not surprising that one finds an inward shift of the open unemployment-vacancy relationship, as depicted in Diagram 2, when active labour market programmes involving participation during extended periods are expanded: even with a zero effect on matching probabilities, this relationship would shift simply because of the arithmetical reduction of open unemployment when programme participation is increased (the shift from curve 1 to II in Diagram 2). This is why the Beveridgc curve has been drawn here as the relationship between the total number of job seekers without a regular job (the sum of openly unemployed and participants in programmes) and vacancies, in which case an inward shift will only reflect an increased effectiveness of the matching process (the shift from curve I to II in Diagram 4).f In the Swedish case, there is some -- but unfortunately conflicting - evidence on how the total Bcveridge curve in Diagram 4 has been related to active programmes. On the one hand, Bourdet & Persson (1993) find thai there has been no trendwise outward shift of this curve in Sweden, in contrast to other countries, such as France, which have placed less emphasis on active labour market measures. On the other hand, the estimations in both Jackman el al. (1990) and Calmfors (1993a) fail to disclose any lime- \t scries relationship between the variations of programmes and die movements of the total , Beveridge curve.9 Some furlhcr indirect evidence is provided by McCormick & : Skedinger (1991), who find that more programme placements in an individual region ! appear to increase open unemployment, which they interpret as the consequence of a weakening of incentives for labour mobility. B. Effects on Ihe labour force j One of the negative effects of unemployment, especially when it is of long j duration, is that labour supply lends to be reduced. Discouraged workers who do not | find jobs will lend to leave the labour force. The risk of such negative supply e^ffects appears greatest for elderly workers, especially when early retirement is used as a ' method of solving the labour market problems of Ulis group, as has been ihe case in, e.g., Belgium, France, the Netherlands and the Nordic countries. In addition, prospective entrants may not enter the labour force in situations of high unemployment. This reasoning points to another possible channel of influence of active labour market policy, namely to maintain labour force participation. There have not been many quantitative studies of litis link, but one example has been provided by Wadcnsjö (1993) in the case of Sweden. The qualitative effects arc, however, easy to outline. The larger is the labour force, the lower is the proportion of it that is regularly employed at each wage level (since each wage is associated with a given number of employed persons). Hence a positive labour-supply effect of labour market programmes means a ceteris paribus shift to the left of die employment schedule in Diagram 5. The higher supply of labour relative lo demand means more competition for the available jobs, which puts downward pressure on wages. Ä new equilibrium tends to be established at C. As can be seenTThT"läböür-market siluaúaliliéTeTiôTaľes in the sense that a lower proportion of the labour force is now regularly employed, and hence a larger proportion is cither openly unemployed or participating in programmes.'" But situation has improved in the sense that a larger proportion of the population is regularly employed. This illustrates the importance of distinguishing between different goal variables when analysing labour market programmes. C. Competition effects for insiders The labour-force effect discussed above does not presuppose that labour-market programmes are targeted on outsiders in the labour market, such as the long-term unemployed or other marginal groups (young people, women not previously seeking work, immigrants etc.): in principle, programmes for the short-term unemployed may also help counteract the tendency to lower labour supply, since the pace with which these arc turned into long-term unemployed with larger risks of demotivation is slowed down (Calmfors & Lang (1993)). Programmes targeting outsiders should, however, be expected to be more effective in promoting labour-foJce.„participatipn. Such programmes may also increase the competitiveness of outsiders relative to insiders and therefore have wage-reducing competition effects in addition to the aggregate labour-supply effect analysed above, as lias been emphasised by, e.g., Layard (1986, 1989, 1990), Layard et al. (1991) and Calmfors & Lang (1993). This is explained by the hypothesis that insiders— in this context the employed and the short-term unemployed — exert a dominating influence on wage setting (Gottfries & Horn (1986), Blanchard & Summers (1986), Lindbeck & Snower (1988), Nickell (1990)). An increase in the competitiveness of outsiders may occur/through several channels: by increasing their productivity, by substituting labour market programmes for regular work experience when employers screen job applicants, or by encouraging more active search behaviour) The competition effect, can be illustrated as a downward shift of the wage-setting schedule in Diagram 5.^ This tends to movefíhe equilibrium to D with lower wages and higher regular employment. ) In principle, all forms of active labour market policy may contribute to a redistribution, p£emplgyment opportunity. As has been found in a number of studies for e.g., Britain, France, the Neüierfänds, Sweden and the United States, intensified ^counselling and encouragement of active search behaviour for the long-term unemployed dor other disadvantaged groups appear to have significant effects on job finding rates lli(Björklund (1990), Haveman & Hollisler (1991), OECD (1993a,b), Lehmann (1993)).12 In fact, one might expect such job search assistance to produce the fastest results with .- respect to competition effects:/ training programmes and temporary job creation will not have an effect until the participants leave the programmesľ)and while in programmes, the participants may even compete less vigorously for regular jobs, as discussed in Section H.A. On the other hand, successful training and job creation schemes might produce more sustained competition effects if they have permanent effects on the skills of the participants. Placement in job creation schemes or programmes giving on-the-job training may also prove necessary to overcome employer reluctance to hiring long-term unemployed because of uncertainty aboi't the job applicants' work capability (Meager & Metcalf (1987), Colbj0rnsen et al. (1992)). , D Deadweight and substitution effects So-called deadweight losses and substitution effects have received substantial attention in the literature (e.g., Haveman & Hollister (1991), Layard et al. (1991), Björklund & Holmlund (1991) and OECD (1993a)). These apply mainly to job creation schemes. The deadweight loss is defined as the hirings from the target group that would jl have occurred in the absence of the programme. The substitution effect is defined as the extent to which jobs created for a certain category of workers simply replace jobs for other categories, because relative wage costs are changed. The deadweight and substitution effects of job creation programmes can be illustrated by a leftward shift ofjhe employment schedule in Diagram 5, which will tend to reduce regulär employment and counteract the fall in open unemployment (the equilibrium tends to move from A to C in this case as well). Such effects are likely to be most important in the case of subsidies: for priyate-sector work, which are common in many countries for young people and long-term unemployed. Similar effects may, however, also occur with public-sector job creation schemes: since these usually are organised at löwer levels of government, there is a risk that they hold back ordinary job creation in order to be able to take advantage of schemes financed by the central governmenLifiscal displacement). In_order_ to. minimise this danger, a principle of laddiiionalityßs joftenjmposed on public work programmes: they are designed to be of stich a character that they would not otherwise have been undertaken (Grubb (1993)). Available empirical evidence indicates that deadweight and substitution effects may indeed be substantial. Studies of subsidisation of private-sector work in Australia (Department of Employment, Education and Training (1989)), Ireland (Breen & Halpin (1989)), and the Netherlands (de Koning et al., (1992)) all point to combined deadweight and substitution offsets of the order of 70-90 percent of the gross number of jobs created. For Sweden, Vlachos (1985) found that the majority of new hirings benefiting from a recruitment subsidy would have been made anyway, whereas subsidies in order to prevent lay-offs achieved a net impact of around 40 percent under the assumption that wage setting was not affected (see Section n.E below). The evidence on fiscal displacement effects in the case of public-sector job creation is more mixed. For the Netherlands, de Munnik (1992) found only small effects in a programme offering permanent municipal work to the long-term unemployed. For Finland (Eriksson;(1993) and Germany (Bellmann & Lehmann (1990)), public-sector job creation was found to Increase the flow out of unemployment. In contrast, Jackman and Lehmann (1990) found no significant impact on the flows out of unemployment from the British Community Programme. For Sweden, a recent study by Ohlsson (1993) comes up with similar results. During their. first_quarter, public-sector relief works are estimated to have a net impact of around half the gross effect, but in subsequent quarters no significant effects on open unemployment can be found. (Incidentally, a similar result is obtained for labour market training). The conclusions in Grämlich & Ysander (1979) and Forslund & Krueger (1993) are more ambiguous. For road building, Grämlich & Ysander find the regular work force to be reduced by more than the increase in the number of relief workers, whereas Forslund & Krueger estimate that total deadweight and substitution effects amount to 60-70 percent of the gross employment in job creation schemes in the building sector. In contrast, there is not much evidence for such displacement effects in the health and welfare sector in the two studies. One can only speculate about the reasons"""for these sectoral differences. Grämlich & Ysander suggest that there are more complementarities in the health and welfare sector (relief workers cannot substitute for the ordinary personnel). Another possible explanation is differences in the budget process: the decisions in the municipalities to apply for relief-work grants in the case of large building projects may often be taken at the same central level as where the overall budget allocations are made. In contrast, the decisions to use relief workers may be taken at more decentralised levels, and with fixed budgets, in the health and welfare sector. E. Reduced welfare losses for the unemployed and crowding-out effects An explicit aim of active labour market policy is to reduce the welfare loss from being out of work. The pursuit of this aim may, however, give-rise also to undesirable side effects, because insider wage demands are accommodated, as has beeristFéššed in IhlTSčlmäinavian discussion öf these policies (see Calmfors and Forslund (1990, 1991), Calmfors and Nymoen (1990) or Holmlund (1990), but also Corneilleau et al. (1990)). The reason is that most wage-setting models postulate a negative relationship between the welfare reduction from a job loss and the wage level. This is the case in union and bargaining models, where unions are assumed to trade off the welfare gains from higher wages for the employed members against the welfare losses for those members that are laid off (see, e.g., McDonald and Solow (1981), Oswald (1986) or Layard et al. (1991)). It is also the case in efficiency-wage models, where employers have to pay more to induce effort on the partôf the employees as well as to recruit and prevent quitting when the welfare for a job searcher is higher (Shapiro and Stiglitz (1985), Johnson and Layard (1986)). There are several reasons why participation in job creation or training schemes may reduce the welfare loss of the unemployed. One is that a labour market programme may offer higher compensation than unemployment benefits. If so, an expansion of labour market programmes will have effects similar to a rise of the unemployment benefi^Caimfors and K (1990)). But the welfare losses from unemployment may also be reduced to the extent that programme participation decreases the risksof future "unemployability" or unfavourable wage developments due to present unemployment spells (Calmfors and Lang (1993), Calmfors (1993a)). Unfortunately, these desired effects of labour market programmes always involve the risk of weakening incentives for wage restraint. In Diagram 5, this is illustrated as an upward shift of the wage-setting schedule, which works in the direction of reducing regularemployment (tending to move the equilibrium from A to E). The effect discussed here thus means that active labour market programmes may crowd out regular employment., It can be seen as a generalisation of the so-called displacement effect, i.e., the possible reduction of jobs elsewhere in the economy because of competition in goods markets, which has been discussed especially in connection with subsidised jobs in the private sector (see, e.g., Haveman and Hollister (1991), Layard et'al. (1991) or OECD (1993a)). Here the correspondence would be the case when participation in a labour market programme is regarded as a perfect substitute for regular work. Then the wage increase from the upward shift of the wage-setting schedule in Diagram 5 would be exactly the same as would have occurred if the employment schedule had been shifted to the right by the amount of jobs created through subsidisation. F. Productivity effects A common argument, dating back to Phelps (1982), is that unemployment may have negative productivity effects because the opportunities, to acquire human capital through on-the-job training are lost and working habits deteriorate. The recent debale has stressed this risk, especially for the longj-tej^unemployed (see, e.g., Layard et al. (1991), Pissarides (1992) or Bea^l'992))" Another possible effect of active labour market programmes is to counteract these tendencies. To raise or maintain the productivity of the unemployed is, of course, the explicit aim of labour market training, but job creation schemes may serve the same end via on-the-job training. Usually, it is just taken foť granted that _lra.iüng,,.by...r|dsing..th.e,„marginal.. productivity of programme participants, contributes to higher employment. If one, as a first approximation, Views labour market training as just one form of general technological progress, the illustration would be a rightward shift of the employment schedule in Diagram 5 (tending to move the equilibrium from A to F). This works in the direction of increasing regular employment. It is well-known, however, that all forms Of technological progress need not give rise to such positive labour-demand effects. For instance, if /technological, development; is labour-augmenting, i.e., if output depends on the amount of labour in efficiency units and labour becomes more efficient, the net employment effect is uncertain. On the. one hand, there is a scale effect lending to increase employment Because of the incentive to expand output by using more efficiency units of labour when the unit cost falls. On the other hand, there is a substitution effect tending lo reduce labour demand, because a given output can be produced by fewer and more efficient workers. The scaleeffect dominatesthe substitution effect only if labour » demand is elastic.^ This uncertainty about the net employment effects of technological progress in general carries over to labour market training as well. Another uncertainty concerns the response of wages to productivity changes induced by labpur:markct training. Although our knowledge on this point is very limited, we can draw parallels to the discussion about the wage and employment effects of productivity change in general. Here there exist two opposing views. On the one hand, insufficient adjustment of wage aspirations to the productivity slow-down in the early seventies has been advanced as a major explanation of the rise in unemployment in most OECD countries (Grubb el al. (1982, 1983), Bruno and Sachs (1985), Sachs (1987)). On the other hand, it has been claimed that the absence of a secular trend in the unemployment rate must be taken to imply that productivity increases in general result in corresponding wage rises (Bean et al. (1986), Blanchard and Summers (1986), Bean (1992)). For instance, Bean's interpretation of the available empirical evidence is that "the productivity slowdown has only a modest role to play in explaining the rise in unemployment". If this interpretation is correct, labour market programmes should not be expected to contribute to higher regular employment via their productivity effects. In terms of Diagram 5, a productivity rise will be associated with a corresponding upward shift of the wage-setting schedule, leaving the employment rate unchanged (with the new equilibrium at G). This would not, however, mean that productivity-raising labour market programmes arc without value: although they will not in this case help increase regular employment, there arc, of course, likely welfare gains from the higher productivity (output) in itself, and from the reduction of open unemployment. A more elaborate analysis must allow for the heterogeneity of labour. The celebrated active labour parket policy in Sweden has traditionally been motivated within a disaggregated framework. The original idea was to alleviate aggregate wage pressure by retraining labour in l0w-productivity sectors with excess supply, so that it could be moved to excess-demand Sectors with high productivity {Fackföreningsrörelsen och den fulla sysselsäilningen (19f 1)). With the help of Diagram 6, the argument can be cast in terms of the analytical framework used here. The Diagram shows a stylised wage-setting schedule, which is horizontal when there is unemployment and becomes vertical al full employment. The employment schedules I and II represent a low-productivity and a high-productivity sector, respectively. \A transfer of labour>from the former lo the latter sector can be illustrated as a rightward shift of the employment schedule I and a leftward shift of the schedule II (al each wage, labour demand increases as a proportion of the labour force in the former sector, whereas it decreases in the latter). The result is a i higher employment rale in the low-produclivity sector and an unchanged one in the high- > productivity one. Since a larger share of the labour force also finds itself in the sector I with a high employment rate, the aggregate employment rate must also increase. This |effect will be reinforced to the extent that the two types of labour are complements in ildemand, since the wage reduction in the high-productivity sector also stimulates ( employment in the other sector. The argument will be qualitatively the same with less extreme wage-setting assumptions. It is enough that wages increase progressively more as the employment rate.....increases (unemployment decreases), as, claimed 'to: be'the . case''by, e.g., Blanchflower and Oswald (1993). This casfc is also illustrated in the diagram. However, the scope for such employment-increasing re-allocations of labour is much smaller in a situation of general excess supply for labour — which would seem to characterise Western Europe at present - when most sectors will find themselves clustered close together on the flat part of their wage-setting schedules. The Swedish re-allocation model was not designed to solve such a problem, but to facilitate a process of structural change. Models have also been constructed in which re-allocation of labour through training programmes may have perverse employment effects. Saint-Paul (1992, 1993) has analysed,fthe case when skilled and unskilled labour are substitutes. If firms are to hire some unskilled workers in this framework, their lower productivity must be balanced by higher unemployment:^ this may either reduce their relative wage (Saint-Paul (1993)), or allow vacancies to be filled more quickly than for skilled workers (Saint-Paul (1992)). Transforming some of the unskilled workers to skilled ones in this analysis raises unemployment among the latter category as an impact effect. As a consequence, the relative wage of skilled workers falls and skilled job vacancies are filled more quickly. This makes it profitable for firms to substitute skilled for unskilled workers, and hence unemployment rises for unskilled workers as well. The net outcome for aggregateünemployrnent is ambiguous: on the one hand/the employment rates increase for both types of labour, on the other hand a larger fraction of workers find themselves in the high-employment category. G. Work-tes t effects There will always be a certain fraction of those that receive unemployment benefits who are not really searching for work. With high unemployment, it is impossible for the labour market administration to test the willingness to work of benefit claimants through regular job offers, placements in labour market programmes may provide an alternative work test for the eligibility of unemployment benefits, since some of those who arc not genuinely inicrcstcd in work will prefer to lose registration rather than to participate in a programme (OECD (1991), Grubb (1993)). This, of course, presupposes that it is compulsory for obtaining benefits to accept offers of programme participation, as is indeed the case in many countries. The work-test effect will tend to reduce unemployment, as measured by the number of benefit claimants. There is not, however, much information on the magnitude of this effect. One would perhaps not expect it to be very large. According to some «studies, more intensive counselling of the unemployed has, however, led to 5-10 per cent Jof the target group leaving tlie register (OECD (1991)). These figures could perhaps be regarded as minimum estimates of the work-test effect of programme participation. There is, however, no reason to expect a fall in the number of benefit claimants to reduce the real amount of involuntary unemployment (as the Labour Force Surveys try to capture), since those affected by the work-test effect were not actively seeking work in the first place. If there is any effect on the effective supply of labour, it ought rather to increase it, by making it less attractive to try to live only on unemployment benefits. Arguably, such an effect — to which the analysis in Section II.B applies — ought to be H. Tax effects 4.odm, 4ww' ?■ OX/% In a complete analysis, one would also have to consider the tax repercussions of í active labour market policies and how they impinge on wage and employment | determination. To' analyse this, it is necessary to distinguish between the jeal consumption wage (the after-tax real wage) of employees and the;feai product wage (the" real wage cost including pay-roll taxes) to employers. There is a general presumption % that tax repercussions may reinforce the net of the effects discussed above (Calmfors and Lang (1993)). For instance, higher regular employment (and output) tends to_jncrease the„taxjaase. and_rcduce_the_sum of the costsfor unempjoyjnent benefits^ and —r programmes, which should allow tax rates to be lower than would otherwisebe the case. If we think of the vertical IKiš in Diagram 5 as denoting the real consumption wage, lpwex.tax_rates will then shJftJh£_empJoyment^.sche^le jo_the_,right, since the real product wage associated with each real consumption wage falls.14 There is an on-going discussion - similar to that about the link between productivity and wages (see Section II.F above) - on how such a tax change will affect wage-setting behaviour in the long-run, i.e., on whether the wage-setting schedule will ultimately shift so much upwards that the real product wage is left unchanged (see, e.g., OECD (1990), Calmfors and Nymoen (1990), Layard et al. (1991), Andersen (1992), Bean (1992)). If the wage-setting schedule shifts less than the employment schedule - as some evidence indicates -active programmes that tend to increase regular employment in the first round also give rise to additional reinforcing tax effects. However, it is also possible that programmes are more expensive than <] unemployment benefits because of higher compensation levels and/or the costs of "; arranging them (from which one has to deduct the value added produced in job creation > ř schemes). To the extent that this is the case, there will be a tendency for tax rates to ,• increase, thereby increasing real product wages and reducing regular employment. There will be similar effects if participation rates increase, with the consequence that more people claim unemployment benefits, but this will be counteracted by lower costs for early retirement pensions and the work-test effects on tlie number of unemployment benefit claimants discussed above. ľ. General-equilibrium effects on other policies In a full general-equilibrium analysis, account must finally be taken of possible interrelations between various policies . f One common claim is that labour market programmes may act as a substitute for .£ long duration of unemployment benefits (e.g., Layard (1991) and OECD (1993c)). It is ~ is well known from a number"ô"f ľštíidies that unemployment in general and long-term f unemployment in particular is correlated with the duration of unemployment benefits JĽ (e.g., Burda (1988), Layard (1990), Layard et al. (1991), Layard and Nickell (1991), / V* f Heylen (1993), OECD (1993a), Zetterberg (1993)). If one interprets this correlation as a ! * h causal relationship, one can argue that a sufficient condition for active labour market '£ policy to increase regular employment in a true general equilibrium that takes all policy t* interrelationships into account is that the "partial" effect without policy responses — i.e., the net of the employment effects discussed in Sections I1I.A-III.H -- is less negative than the net "partial" effect of the alternative with long duration of benefits. Another alternative to active labour market policy may be aggregate-demand policies in order to increase regular employment. It is a common view, especially in the older literature on labour market policy, that an increase in participation in job creation schemes will contribute less to wage pressure than a corresponding increase in regular employment -- even if the same wage is paid and the jobs are also regarded by workers as perfect substitutes in other respects (see, e.g.; Baily and Tobin (1977)). The basic idea is that inflationary effects.,.can„be,,reduGed~by~targeting,»the,,mcrease of labour demand on those who are unej5ploye4„in^^ in general, which will result also in competition between employers for: the already employed. This is likely to be a relevant argument if programmes are tarjgeteďon outsiders, as discussed in Section II.C. However, there are also potential risks if labour market~pölicies are used to counteract unemployment wherever it threatens to occur, as was tried in many countries in the mid-seventies, when it was a common policy objective to try to maintain workers in their current jobs through redundancy-deferring subsidies (OECD 1990)). In fact, such selective accommodation policies are Iikelyjt.ouw.eaken union incentives for wage restraint more than an increase of aggregate labour demand: in the former case, lay-offs resulting from high wage increases for the members of any individual union will be partly accommodated by placements in labour market programmes, whereas in the latter case the individual union's members will have to compete on equal terms for the new job openings.createdI by the expansionjif aggregate labour demand (Calmfors and Horn (1985), Leonard and van Audenrode (1993)). This reasoning has been used to explain why some Swedish studies have found labour market programmes to be more wage-raising than regular employment (Calmfors and Forslund (1991), Forslund (1992)). Indeed, one may see it as a risk that too strong an emphasis on active labour market measures could lead policy makers to neglect other policies. It has, for instance, been claimed that the strong belief in the potential of labour market programmes may have been an important explanation of the restrictive demand policies in Sweden in 1990-92, which seem to have contributed strongly to the dramatic rise in unemployment (Calmfors (1993a)). However, such problems need not be inherent, but may rather reflect slow learning processes. At the same time, it is obvious that properly designed programmes can act as complements to demand policies. To the extent that these are held back because of the fear that increased labour demand will just lead to increased wage pressure, active programmes that attack various persistence mechanisms (by increasing the search effectiveness of the unemployed, by providing them with the skills in demand, or by acting as substitutes for regular work experience in the screening of job applicants by employers) may be a necessary prerequisite for a more expansionary policy stance. The difficulties of drawing precise conclusions about the effects of active labour market policy become even greater once one allows for possible interrelationships with various structural reforms in the labour market with respect to, e.g., wage-setting practices, employment protection legislation, competition policy etc. The great uncertainty surrounding the possibility of such endogenous policy responses does not mean, of course, that these effects need be less important than the ones that are easier to quantify. Indeed, the opposite might be the case. III. The net effect of active labour market programmes The main conclusion from the analysis in Section II is that active labour market policies work vjajjjjuraber jafchannels. Table J represents an attempt to summarise the various possible effects. The discussion has made it'clear that it may sometimes be difficult to sign individual effects. It is impossible to infer this net impact of active labour market policy from theoretical reasoning only. Unfortunately, however, the available empirical evidence on the overall macro effects is still scarce. A. Existing empirical research The area where there exists most empirical research is aggregate wage setting. A number of studies have tried to estimate;how the wage-setting schedule in our diagrams arc affected by active labour market progřammeš^Thěše studies can be interpreted to give estimates on the net of several of the effectslbn the wage-setting schedule discussed above: the matching effect (Section TII.A), the competition effect (Section III.C), the rěduced-welfare-loss effect (Section III.E) and possibly, also, the productivity effect (Section II1.F). Although these effects do not represent the full general-equilibrium ones, they still make up an important subset. Most wage-setting studies made so far are time scries estimations for Sweden, where the traditional emphasis on active labour market policy has made it more natural than elsewhere to analyse the macroeconomic impact. The majority of these estimations, which are summarised in Calmfors (1993a), have found increased programme participation to shift the wage-setting schedule in our diagrams upward (aggregate wage pressure rises when,pcogramme participation increases at a constant rate "of regular employment or, which is the same thing, when agiven number of workers are transferred from open unemployment to programmes). /Sometimes, the wage-raising effects of active programmes appear to be even larger than those of regular employment)^ There are also a few time-scries studies of the relationship between wages and active labour market policy for other countries. For Finland, similar results as for Sweden were obtained by both Eriksson et al. (1990) and Calmfors and Nymoen (1990). The latter also found unemployment-reducing programmes to increase wage pressure in Denmark but not in Norway. A more favourable picture of the wage-setting effects of active labour market programmes is provided in OECD (1993a), where Phillips-curve estimations covering nineteen different countries are máďeľ For the majqrily_of_these, increased programme participation that reduces open unemployment seems to contribute tô^eí£yyjgej.press,ure, aUKöugh the estimates often are imprecise. A dräwíačK~is™aTšo that the number of observations on programmes is very small for each country.16 An attempt to exploit cross-country variations in the size of labour-market programmes has been made by Heylen (1993), who tries to explain differences in the responsiveness of wages to open"ufiemploymenl. He finds that increased expenditures on active programmes ger„uncmp|oyed person (as well as a larger ratio between expenditures on active programmes and on unemployment benefits) tend, tot increaseJhc__ wage responsiveness.tp,changcs in open unemployment. These results are not directly comparable to the ones reported above, since they cannot readily be transformed so as to indicate whether an expansion of labour market programmes that reduces open unemployment is likely to be wage-increasing or wage-reducing./On the one hand, a '/ fall in open unemployment puts upward pressure on wagesjbut on the other hand the i increased responsiveness of wages to unemployment tends to moderate them.*' Finally, it is possible to infer what can be regarded as estimates of the total effects of active labour market policies (excluding only some of the policy interrelationships in Section II.I) from a couple of studies also exploiting cross-country variations. Layard el al. (1991) estimates a cross-country reduced-form equation for open unemployment in the eighties, where expenditures on active programmes per unemployed as a ratio of GDP per capita is one of the explanatory variables. Zcttcrberg (1993) pools cross-country and time series data for 1985-91 and makes his estimations with instead the share of expenditures on active programmes out of total labour market expenditures as the labour-market-policy variable. The two studies come up with similar results: they seem to imply that an increase of programme participation with 1 percentage point of the labour force reduces open unemployment by 1.5 (i.e., the rate of regular employment increases by 0.5 percentage points).*8 B. Problems of interpretation The limited but yet diverse empirical evidence on the net employment effects of active labour market policy suggests the need for more research. However, it appears important to point out a few problems in this context. All the studies mentioned measure unemployment and programme participation in relation to the labour force. Hence they do not take into account the possible job-creating effects of active labour market policy from increased labour-force participation, as discussed in Section II.B. The effects of labour market programmes may very well depend on how much unemployment there is. One can find theoretical arguments why active programmes should be expected to givefbetter results with higher unemployment|(see also Section 1V.C below). There is then'á'smaljer risk of accommodating insideTwage claims. The scope for targeting the long-term unemployed "änd'stréngth^ competitiveness increases, and the task of maintaining labour-force participation becomes more important. Hence the earlier wage-setting studies from Sweden, which refer to situations with very low unemployment (in! the 1-3 percent range) may not be relevant to the typical Western European case of high unemployment (to which also Sweden has recently conformed). There are also statistical problems of interpretation. It has Jiecn pointed out in Grubb (1993) and OECD (1993a) that the typical pattern within thefoECD is that active programme expenditures seem to increase less than proportionally with unemployment. If this is taken to reflect a "government policy rcaction,function"^)lhe results of, e.g., Layard et al. (1991), Zetterbcfg (1993) and van Heylen (1993) may all be subjected to simultaneity-bias that' tends to give too positivTTX"picture of active labour market programmes: the estimated negative effects on, for instance, open unemployment may reflect government reactions to unemployment as well as the response of unemployment to active programmes. The reason is that the above studies all capture the emphasis on active labour market policy by some measure of programme expenditures in relation to unemployment.19 The Swedish wage-setting studies may suffer from an opposite simultaneity problem, since the typical pattern in Sweden seems to have been that programme participation has increased more than proportionally when unemployment has risen (Calmfors and Forslund (1993)). Hence, higher real wages that have led to lower employment may also have caused programme participation to increase relative to open unemployment. Such an effect may have biased the estimated wage effects of an increase in programme participation relative to open unemployment upwards in the Swedish case.20 IV. Crucial design features A possible approach to take may be that the question of the net effect of active labour market policy is perhaps not the most meaningful one. A more relevant issue may be how programmes should be designed and how large they should be in order to make a maximum contribution to employment performance. This Section discusses the following crucial design features: (i) compensation levels; (ii) the extent of targeting; (iii) the type~of programme; and (iv) the duration of programmes and the co-ordination with the unemployment benefit system. A. Compensation levels It is a standard conclusion in the theoretical literature that higher unemployment compensation increases equilibrium unemployment (see, e.g., Oswald (1986), Björklund and HolmlundJÍ99J), Layard et al. (1991), Bean (1992)). An analogous argument can be made with respect to^compěnsatíon levels in labour market programmes. The higher the compensation, the smaller will be the expected income and welfare losses for workers who are laid off or quit (Calmfors and Forslund (1990, 19?T)7Cälmfors and Nymoen (1990)). This strengthens the tendency to wage-raising effects discussed in Section U.E. Moreover, the incentives for participants in programmes to search actively for regular work and accept job offers become weaker, the higher the compensation, with adverse Beveridge-curve effects as a consequence, as discussed in Section H.A. In practice, compensation in programmes ranges from the equivalent of unemployment benefits (usuaíiýln training programmes) to market wages (in many job creation schemes);* The above reasoning seems to suggest that there are strong macroeconomic arguments for setting compensation levels close to unemployment benefits. Indeed, the high compensation levels in public job creation schemes in the Nôr*3ic*countries have been advanced as an explanation of the finding in many studies that labour market programmes there seem to have increased wage pressure (Calmfors (1993a)). It is sometimes claimed that setting compensation levels in programmes that are in excess of unemployment benefits is necessary in order to provide incentives for participation, especially in the case of labour market training. This reasoning appears questionable, especially if unemployment benefits can be withdrawn in the case of refusal to participate in programmes, as happens in several countries. But the argument is not self-evident even in "laissez-faire systems" where programme participation is voluntary. One could argue that a prerequisite for efficient training programmes is that they should be able to attract participants on their own merits, i.e., because of their effects on- future re-employment probabilities and wages, and not because they offer higher short-term compensation. One method of reducing programme compensation levels in, e.g., public job creation schemes, may be to continue paying market hourly wages but ofjM^qnly part^ time work. For instance, in Sweden both working time and compensation in public-sector relief work have been cut to 90 percent of the "market levels". An earlier youth programme offered only half-time work. An additional advantage of this approach is that the spare time is freed for active job search, which can help counteract the tendency for programme participants to be locked in (see Section II.A). IJ. The extent of targeting As already discussed in Section II.B, the extent of targeting is likely to be another crucial design feature. If programmes targeted on outsiders in the labour market stimulates wage reducing competition for jobs, the likelihood of substantial positive employment effects is increased. U, One issue- in this_context is_wheÜiej^one_shOTjdJ^ interventions MtiJ.peopLeJi^^^ try to ideritif^the_di.fficuU-to^ unemployed. This problem relates to the general question of to which extent the "reduction of re-employment probabilities for categories with long™*uTíě"m"ploymeni Ifuräťiôn is caused by heterogeneity (i.e., the fact that the least employable are gradually sorted out and therefore make up a larger proportion of the categories of unemployed the longer they have been out of work) pr.by, slate dependence (i.e., the fact that a given individualXchance of re-employment decreases_over time). Although there is an ongoing discussion on theřěíitive importance of these two mechanisms, there appears to be a con^ensusjhat heterogeneity is important (see, e.g., Jac^an andLayard(-1991) or van den Berg "and"van Oúŕš (1993ä,ŕľj)ľ This provides an argument for targeting easily /* identifiable groups of difficult-to-place, such as imnu^rá^tšTliišabTe^MC^QroiF^with 7 lolTg"earliefjůríemployment spells.™"However, a necessary requirement is that the >■' programmes are_reasonablyeffective_in_raising^fhelcgmpetitiveness ofthe^difficulkto-píaceľ HereThe results are somewhat mixed. On the one hand, Haveman and Hollister TÍ991) conclude that targeting seriously disadvantaged groups seems to be the most effective (although the impact does not appear very strong). On the other hand, a few of , ("the studies surveyed in OECD (1993a) seem to indicate that the best results are obtainéŕľ" » j by targeting groups with "moderately severe, easily identifiable" problems (such as) ) women re-entering the labour market). ' <_1 There are, however, also likely to be limits to the degree of targeting on those I that are regarded as difficult to place. Exclusive targeting on this group is not likely to I be appropriate, since programmes then mn"thl:^|l'„ijQ^^ í p£oblemj:äses and hence signal to employers^that híaňg-4hěm~should-be-avoidecL, I (Burtless (1985)). This would seem to be a strong argument against focusing programmes only on disadvantaged groups. From a theoretical point of view, young entrants to the labour market clearly form a group of outsiders that compete with the current insiders. Youth programmes, such as are common in many countries, would therefore be expected to reduce wage pressure (Calmfors and Lang (1993)). Surprisingly, however, according to some Swedish studies, programmes targeting young people seem to have been less successful than one would expect from theoretical reasoning. Skedinger (1991) finds that active labour market programmes for this group appear to have been more wage-raising than programmes for other age groups. Wadensjö (1987) also points to large crowding-out effects on regular employment for young people. Calmfors and Skedinger (1994) find that the extent of targeting on this group does not influence the overall employment impact of active labour market programmes. As concerns targeting according to the duration of unemployment, i.e., the issue of optimal timing of labour-market-policy interventions during a typical unemployment career, there has, somewhat surprisingly, been almost no empirical research. It is, however, easy to outline the main trade-offs involved. There are a number of advantages with grogramme placements late in an rf unemplQymenLspeJ], 'f^a'dwJfghOoIsH^w'iíl be smaller to the' extent that oňeavôids programme participation by many of those who will airy way find a new job. This tends >! to hold down „costs so that unfavourable tax repercussions with possible adverse employment effects, as discussed in Section II.H, are reduced. A smaller number of job jj> applicants are locked in by programmes. One is also more certain to target outsiders. 4 Finally, the benefits of programme participation for the individual will be discounted >- more heavily, which dec£easej>.jhe-rislcs of accommodating insider wage demands, as stressed in Section II.E"(see aíso Calmfors andXarig (1993)). The above benefits, must however, be traded off against the disadvantages. The later programme placements occur, the smaller is the potential number of outsiders who can compete more effectively with insiders. It is also likely to be more difficult to restore lost competitiveness on the part of the unemployed, the more it has been allowed to dcterionuey**This has"""led, e.g., Layard et al. (1991) to recommend targeting the ''medium-term unemployed" (those having been unemployed axp^d^i^jnonlhs) rather than the really long-term unemployed. To analyse the appropriate timing of labour-market-policy interventions, one would need more knowledge oh the duration dependence of both re-employment probabilities and the effects of active programmes. On the latter point research is almost non-existent. On the former, there is a fair amount of evidence on negative duration dependence, i.e., that re-employment-probabilities fall over time, but there is also material that does not support this conclusion (see, e.g., Björklund (1990), Steiner (1990), Layard et al. (1991), OECD (1991) and van der Berg and van Ours (1993a,b)). Also, to the extent that there is negative duration dependence, the speed of deterioration of re-employment chances is very important, e.g., if there are "jumps" in the development at certain points of time. Here we have only circumstantial evidence. For instance, a Norwegian study indicated only minor changes in employer attitudes toward hiring an unemployed during the first half-year of unemployment, but significant changes when duration exceeded six and twelve months, respectively (Colbj0rnsen et al. (1992)). Macroeconomic studies of wage-setting that have found a higher proportion of long-term unemployed to increase wage pressure — given the total rate of unemployment — have drawn the dividing line between short-term and long-term unemployment at twelve months (see, e.g., Layard and Ňickell (1986), Franz (1987) or OECD (1993a)). In contrast, Calmfors and Nymocn (1990) found no evidence on differential effects of short-term and long-term unemployment on wage formation in the Nordic countries when six months was used as the dividing line. When analysing the impact of targeting, one should in general be careful when , drawing conclusions from wage equations with the proportion of long-term ä unemployment as an explanatory variable. Although this variable has been shown to be negatively related lo the ratio between active expenditures and unemployment benefits across countries (OECD (1993a)), earlier estimated wage equations reflect labour-market processes where active programmes have been of limited importance.J Variations in measured long-term unemployment arc likely to have come about mainly through variations in regular employment and outflows from the labour force. Hence the equations may provide little information on the extent to which wage pressure can be reduced by a large-scale expansion of active programmes in order to cut long-term unemployment. This may be seen as an example of the general Lucas critique: the statistical relationships that hold under one policy regime may not be relevant under another when variations in the explanatory variables arise from different policies. Unfortunately, the dala necessary to test whether a reduction of long-term unemployment through labour market programmes arc likely lo have the desired effects do not exist today. The reason is that in most countries no statistics are readily available on the proportion of job seekers — including both the openly unemployed and programme participants ~ who have been out of a regular job more than a certain period of time and qnJiow large shares of the outflows from long-term unemployment, are into rjrggrammcs.Clhc collection of such data on an international scale ought to receive high priority if active labour market policy is to be evaluated properly^) C. Types of programmes Another crucial issue is how to best allocate resources between the various types of programmes discussed here, i.e., placement services and job search assistance, labour-market training, public-sector job creation and subsidised work in the business sector. Put differently, how does the optimal labour-market-policy portfolio look? The sad answer is that we, despite an impressive amount of research, know very little. Already to go through only existing micro studies or surveys of such micro "studies is a very distressing experience, because of the difficulties to generalise (see, e.g., Haveman and Hollister (1991), Björklund (1993), Dolton (1993), Johannesson and Zetlerbcrg (1993) or OECD (1993a)). There seem to be as large differences in results wiilurTpjogiimme categories as between them. Sometimes one finds substantial effects or both future earnings and employment, sometimes one does not. It is often not possible to explain variations in results by the differences in programme design. Nor can one usually judge to which extent resource costs have differed between programmes that appear to have produced different results. It also happens frequently that various evaluations of the same programme, made with different methods or for different time periods, give widely diverging results. Still, a few general comments can be made. As discussed in Section II.C, there appears to be a broad consensus between most studies that intensified f^^^^^~M^jc^§ssi^f^isianai_á6Tšise re-employment probabilities substantially for the target group in question, and especially so for the long-term unemployed. It is not surprising that the studies also indicate large sübsTitution cTféctl'of^sucrľmeasures. One may indeed consider this to be exactly the point, i.e., to contribute to lower wage pressure by subjecting insiders to more competition. This does, of course, presuppose careful targeting. Unlike other programmes, intensified counselling and job search assistance do not involve any risks that participants are locked into programmes with reduced search activity as a consequence. On the contrary, the whole idea is to put the unemployed involved at the immediate disposal of the labour market. There will, however.^be no "book-keeping" reduction of open unemployment as with placements in training ofjob creation schemes (see Diagram 3)p Nevertheless, there may still be welfare increases to the extent that demo^iyjition jrij^^rojyiagcment of long-term unemployed and other outsiders are c^mcňctcd, although these effects are perhaps smaller than with the other programmes. ~~~ Although much of the recent labour-market-policy discussion has stressed the ,jnentsj^training as opposed to direct job creation (e.g., OECD (1990)), few systematic comparisons of the macroeconomic effects seem to have been made. For Sweden, there exists some evidence that participation in training programmes ~ in contrast to direct job creation — exerts downward pressure on wages (Forslund (1992), Edin et al. (1993)). Hcylen (1993) reports ä similar result for the wage responsiveness to unemployment from cross-country regressions. When trying to explain regular employment in pooled cross-region and time series regression for Sweden, Calmfors and Skedinger (1994) also find training programmes to give consistently belter results. So did Jans (1992) in an aggregate lime scries analysis. Although'TCese results provide some support for the view //that training programmes may be more beneficial than job creation schemes, one can still i i raise some caveats: (i) One possible explanation of the findings may be that compensation in training programmes is usually lower than in job creation schemes rather than the fact that the former are intrinsically more effective. '!,(l(fú//-/.Jo)'c/f- (ii) Micro evaluations of various training programmes have produced mixed ^ J..^ results (e.g., Haveman and Holister (1991), Björklund (1989, 1990), OECD (1991, /T.' 1993a)). In general, there appears lo be a tendency for narrowly targeted programmes \j ^Jt„prj3yid.e...sp5cjfic=,s^Ilsinsjder wage claims and/locking large groups into programmes as well as rising-substitution and deadweight effebfsTTo the extent that one tries to avoid the latter by job creation in the public sector or non-profit organisations based on the atói7/onfl/ř^.pjjhidple_j[Grubb_ (1993)), i.e., when the projects undertaken are confined to such that would not otherwise tiaveTaken place, the result is instead likely to be a low marginal value of the output produced. Mfa J)Q 'tf/QfaOM /Decreasing returns/to scale are likely tOeaserimisprobIejn_also withjraining pjogrammes (see, e.g., Cälmfors (1993b) or Grubb(í§93))r^íejarger the programmes, the more~difficult it will be to tailor them to the specific needs orTn¥pMi*ôpTntr*(see point (ii) above). The possibilities of identifying and increasing the supply of those categories of labour that are likely to be in short supply (and thus serve as bottlenecks) in the coming upswings may soon be exhausted. The supply of high-quality training capacity is limited, and the difficulties of monitoring the effectiveness of the programmes increase as they expand. There will be more participants who are poorly motivated, and who are likely also to have negative external effects on die more motivated ones. Usually, the methods used for evaluating the effects of various programmes do not allow average and marginal effects to be distinguished. An attempt to make such a distínction"was7howeveŕ made by Bjj^klun^lj^^OTfjM1987) in their study of Swedish training programmes in the early 1980s. They found negative marginal effects on future wages with a programme size of around 1 percent of the labour force. Forslund and Krueger (1993) use information on the eariůngsjyife£ts_of Swedish training programmes from a number of studies and seem to be able to reject the hypothesis that they have been socially profitable.21 .1 ■# if; t. The obvious conclusion is that the optimal mix of programmes should be a "balanced portfolio". This may need emplíaslsing,*liiňée there is always a risk that čl!an™gíng"Tasruoňs due to earlier disappointments lead to excessive swings in policies. Today, the generally favourable attitude towards education and training might involve such a danger. The greatest risks of decreasing returns to scale for these programmes are likely to be associated with rapid H^ň^íohs^béTčTě^tRFněčessary investments in training capacity have been made. Since these will largely involve investment in human capital through acquiring organisational knowledge, the process'of expanding capacity may very well be a slow one. Another risk to be avoided is that large placements in training — as well as job creation schemes — are allowed tó'^train the resources of the labour-market-administration to the point that basic placement services and job-search assistance are crowded out) D. The co-ordination with unemployment insurance and programme duration Although there has been an extensive discussion about the duration of unemployment benefits (see Section IIJ above), much less interest has been devoted to the co-ordination between active labour market policy and the unemployment insurance system as well as the appropriate length of programmes. There appears to be a general tendency in much of the policy discussion to view unemployment insurance and active labour market programmes as distinctly different systems, where "passive" unemployment benefits are regarded as "bad" and active measures as "good"*.' However, a more fruitful approach may be to analyse the whole system of support for the unemployed, and to recognise that the incentives for lower unemployment depend to a large extent on theíô^oTffinátlôn ôf ifs various parts. The most obvious aspect of co-ordination between unemployment insurance and active programmes concerns benefit eligibility. It is common that programme participation qualifies the participants for new* benefit periods. In (Denmark, for example, this has been one of the main aims of labour-market programmes: a job offer guarantee — later on combined with an education offer — has been in effect for the unemployed approaching the termination of benefits (after 2Vit years; see OECD (1991a, 1993d) or Grubb (1993)). Such a use of programmes should be expected to lead to similar problems as long duration of unemployment benefits, the main difference being that the long-term unemployed will instead be permanently going back and forth between programmes and open unemployment. Indeed, this way of co-ordinating unemployment insurance and programmes may be an important explanation of why long-term unemployment in Denmark has been considerably lower than in other Western European countries with similar rates of unemployment. Ifproerammes come to be regarded mainjv_as,a..means of prolonging the duration ot benefits, there is likely to be a serious weakening -- among placement officers, among organisers and among participants -- of the incentives to strive for maximum efficiency in terms of enhanced re-employment probabilities. Although there are few evaluations of Danish labour market programmes, there is some evidence that this may indeed have occurred. Spells in job-offer schemes have often been recurrent - 60 percent of the participants in 1984-88 had more than one placement up to 1991 (and around 25 percent more, than two) according to Langager (1992). Among the unemployed, the education offer seems widely to have been seen as a method for benefit renewal — as an "unemployment benefit generator" to use the Danish term -- rather than as a step to a regular job (.Rapport om arbejdsmarkedets struHiurproblemer (1992)). The effects on hiring probabilities have been found to be negligible (Aarkrog el al. (1991) and Pilegaard et al. (1991)). The finding by Thaulow and Anker (1992) that,lhe participants' evaluation of tlie job and education offer schemes was not related to the subsequent, employment situation, might be considered indirect evidence of the risks involved. In a recent study for Sweden, Regner (1993) found significantly lower future incomes for participants in labour market training than for those that were openly unemployed in the 1989-91 period. This is in contrast to some studies of earlier periqds (see Politik mot arbetslöshet (1993)). The suggested explanation is that training programmes had increasingly come to be used as a means of renewing benefit eligibility. Leaving the co-ordination with unemployment benefits aside, the duration of programmes is an important issue in itself, about which die empirical knowledge is scant. On the one hand, programmes of long duration are a risk from the point of view that participants may be locked in and theíeíorc search less actively for regular jobs. On the other hand, a certain length of programmes may be necessary in order to achieve substantial results. An indication of this may be that a recent cross-country study found [higher programme expenditures per participant - which is likely to be correlated with "programme duration - to contribute to lower aggregate wage pressure (OECD (1993a)).") Somewhat surprisingly, however, Axelsson (1989) and Axelsson and Löfgren (1992/ were not able to confirm any relationship between duration of training programmes and their social return in the case of Sweden. Korpi (1992), however, found that the f probability of finding a permanent regular job seems to increase with programme j duration, but to decrease with the number of programme placements. >v A specific issue concerns temporary versus permanent job creation measures. Temporary job crcajjon is the prevailing pattern in most OECD countries. But in some countries, notablylBelgium, the Netherlands and Spain, permanent jobs -- usually in the public sector - are* created for long-term unemployed.^ A serious drawback of such a policy is that the workers concerned are likely to withdraw permanently from active search from regular jobs. Such permanent job creation schemes seem hard to justify except possibly for elderly and disabled workers, where early retirement may be the main alternative. | In e.g., Belgium, Denmark and Sweden, job offers for the long-term unemployed ätake the form oľ explicit job guarantees. One should expect these to have more adverse effects on search intensities than when discretionary decisions arc taken about placements in job creation schemes. V. What can we expect from active labour market policy? The main conclusion from this analysis is that active labour market policy may give rise to a diverse set of effects, some of which are favourable and some of which are not. One cannot from a theoretical analysis evaluate the net impact of these policies. There are also severe problems of interpretation with much of the empirical macrocconomic research and a lot of conflicting evidence from the microeconomic studies. There arc crucial areas such as the optimal timing of labour-market-policy interventions as well as the optimal mix and size of programmes that appear more or less unexplored. Still one can draw some tentative conclusions on how active labour market programmes should be designed so as to maximise the chances of success. Obviously one must try; to strike a balance between providing the unemployed with better opportunities and at the same time maintaining sufficient incentives for individual job search as well;as for responsible collective wage-setting behaviour. A tentative list of requirements would be as follows. C Compensation levels ought to be set well below market wages. Programmes should be'carefully targeted so as to provide insiders with more competition for the available jobs. This is^not,,however, likely to mean, exclusive targeting on the most long-term unemployed and on the otherwisc-most-diffičult to place," since this may give programmes a bad^reputation, making it impossible to substitute them for regular work experience as a scfeening mechanism for employers. Targeting medium-term unemployed in order to prevent them drifting into long-term unemployed may also be more efficient than trying to counteract the negative effects of long-term unemployment once it has occurred. At the same lime, it is obvious that the very concept of targeting in order to redistribute employment opportunities puts severe restrictions on. the overall size of programmes. It seems to be important that programmes arejjiot of too long a duration in order to avoid locking-in effects?! Programme placements -- especially in the form of guarantees - in order to make the participants eligible for prolonged unemployment benefits are likely to have detrimental effects on the effectiveness of the programmes. To avoid such risks, limits may have to be imposed on the extent to which programmes may qualify the participants for extensions of unemployment benefits. It may also be wise in many countries to combine an increased emphasis on active labour market programmes with a reduction of the maximum duration of unemployment benefits in order to avoid undesircd incentive effects. As to the type of programmes, one should expect the optimal mix to be a balanced portfolio involving all forms of active labour market policy. Although an expansion of labour-market training may be appropriate in most countries, there may be a risk that ,the current discussion overemphasises the benefits and neglects the importance ofjjecreasing returns to scalep These risks are, however, likely to be smaller with on-therjob-training than with other training schemes. It appears important that the administration of large training and job creation schemes is not allowed to crowd out counselling activities and job search assistance for the long-term unemployed, which have often proved to be quite effective. : What contribution can active labour market policy make to fighting unemployment? On the basis of this exposition, my guess would be that most countries in Western Europe could do better with more active programmes -- if carefully designed ~ but not a lot better. The proper perspective appears to be to view active labour market policy as only one ingredient of many in a general programme against unemployment. !eäi wage Full-employment schedule Wage-setting schedule Un Employment schedule Employment rate i\ Diagram 2: The Beveridge Curve Vacancy | rate Unemployment rate m Diagram 3: A revised Layard-Nickell framework Real wage WS WS = Wage-setting schedule RES= Regular-employment schedule FE = Full-employment schedule RR = Schedule indicating full employment less programme participation Uo RES Rate of regular employment ncy II U0+ r0 Unemployment and programme participation as a proportion of labour force Diagram 5: Various effects of labour market policy on wages and regular employment ws RES Rate of regular employment Diagram 6: Reallocation of labour Real wage I WS WS' Rate of regular unemployment ■ Table I: Various effects of active labour market policy EFFECT Real wage Regular Regular Effective labour Measured employment as employment as force labour force proportion of proportion of labour force population MATCHING "LABOUR FORCE COMPETITION FOR INSIDERS + (?) -(0) "(?) SUBSTITUTION AND DEADWEIGHT LOSSES REDUCED WELFARE ■ŕ _ _ 0 LOSS PRODUCTIVITY + (0) ?(+) ?(+) 0 WORK TEST O(-) 0(+) 0<+) 0(+) TAXES 7 ?(0) ?(0) ? OTHER POLICIES •> ? ? ?(+) ?(+) Parentheses indicate possible but uncenain effects. The last columns do not take into account the secondary labour-tořče effects that may occur because of the induced wage and employment chances according to the first three columns. it BIBLIOGRAPHY AARKROG, V. (1991), "Uddannelsetilbud for langtidsledige", Danmarks La;rer-h0jskole & Udviklingcentret for folkeoplysning og voksenundervisning, September. ANDERSEN, P.S. (1992), "Taxes and Labour Market Developments", Ministry of Finance, Copenhagen. AXELSSON, R. (1989),"Svensk arbetsmarknadsutbildning — en kvantitativ analys av dess effekter", PhD dissertation, University of Umeá. AXELSSON, R. and LÖFGREN, K-G. (1992), "Arbetsmarknadsutbildningens privátoch samhällsekonomiska effekter", EFA-rapport nr. 25, Arbets-marknadsdepartementet, Stockholm. B AIL Y, M.N. and TOBIN, J. (1977), "Macroeconomic Effects of Selective Public Employment and Wage Subsidies", Brookings Papers on Economic Activity, no. 2. BEAN, C. (1992), "European Unemployment: A Survey", Discussion Paper No. 71, Centre for Economic Performance, London School of Economics. BEAN, C, LA YARD, R. and NICKELL, S. (1986), "The Rise in Unemployment: A Multi-Country Study", Economica, vol. 53, no. 2. BELLMANN, L. and LEHMANN, H. (1990), "Active Labour Market Policies in Britain and Germany and Long-term Unemployment: An Evaluation", report presented for the European Association of Labour Economists (EALE) in Lund (mimeo). BEHRENZ, L. (1993), "Effekt- och effekiivitetsanalyser av 1987 ärs personalför-stärkning till arbetsförmedlingen", mimeo, Department of Economics, University of Lund. VAN DEN BERG, G. and VAN OURS, J (1993a), "Unemployment Dynamics and Duration Dependence", Research Memorandum 1993-22, Vrije Universitet, Amsterdam. VAN DEN BERG, G. and VAN OURS, J. (1993b), "Unemployment Dynamics and Duration Dependence in France, the Netherlands and the UK", Research Memorandum 1993-38, Vrije Universitet, Amsterdam. BJÖRKLUND, A. (1989), "Klassiska experiment inom arbetsmarknadspolitiken", Forskningsrapport No. 37, The Industrial Research Institute, Stockholm. BJÖRKLUND, A. (1990), "Unemployment, Labour Market Policy and Income Distribution". In Persson, I. (ed.), Generating Equality in the Welfare State: The Swedish Experience, Norwegian University Press, Oslo. < BJÖRKLUND, A. (1993), "The Swedish Experience", in Kongshoj-Madsen, P. and Jensen, K. (eds.), Measuring the Effects of Labour Market Measures, Jennerop offset. BJÖRKLUND, A. and HOLMLUND, B. (1991), "The Economics of Unemployment Insurance: The Case of Sweden". In Björklund, A. et al. (eds.), Labour Market ■ Table I: Various elľecis of active labour market policy KFFECT Renl wage Regular Regular Effective labour Measured employment as employment as force labour force proportion of proportion of labour force population MATCHING •) + (?) + (?) 0 0 LABOUR FORCE - -(0) + ' + + COMPETITION FOR „ + + 0 0 INSIDERS SUBSTITUTION AND . 0 0 DEADWEIGHT LOSSES REDUCED WELFARE ~ . 0 0 LOSS PRODUCTIVITY + (0) ?(+) ?(+) 0 0 WORK TEST O(-) 0(+) 0(+) 0(+) - TAXES ? ?{0) ?(0) ? 7 OTHER POLICIES 7 7 7 ?(+) ?(+) Parentheses indicate possible but uncertain effects. The last columns do not take into account the secondary labour-force effecis that may occur because of the induced wage and employment changes according to the first three columns. BIBLIOGRAPHY AARKROG, V. (1991), "Uddannelsetilbud for langtidsledige", Danmarks Laerer-h0jskole & Udviklingcentret for folkeoplysning og voksenundervisning, September. ANDERSEN, P.S. (1992), "Taxes and Labour Market Developments", Ministry of Finance, Copenhagen. AXELSSON, R. (1989),"Svensk arbetsmarknadsutbildning - en kvantitativ analys av dess effekter", PhD dissertation, University of Umeá. AXELSSON, R. and LÖFGREN, K-G. (1992), "Arbetsmarknadsulbildningens privátoch samhällsekonomiska effekter", EFA-rapport nr. 25, Arbets-marknadsdepartementet, Stockholm. BAIL Y, M.N. and TOBIN, J. (1977), "Macroeconomic Effects of Selective Public Employment and Wage Subsidies", Brookings Papers on Economic Activity, no. 2. BEAN, C. (1992), "European Unemployment: A Survey", Discussion Paper No. 71, Centre for Economic Performance, London School of Economics. BEAN, C, LA YARD, R. and N1CKELL, S. (1986), "The Rise in Unemployment: A Multi-Country Study", Economica, vol. 53, no. 2. BELLMANN, L. and LEHMANN, H. (1990), "Active Labour Market Policies in Britain and Germany and Long-term Unemployment: An Evaluation", report presented for the European Association of Labour Economists (EALE) in Lund (mimeo). BEHRENZ, L. (1993), "Effekt- och effektivitetsanalyser av 1987 árs personalför-stärkning till arbetsförmedlingen", mimeo, Department of Economics, University of Lund. VAN DEN BERG, G. and VAN OURS, J (1993a), "Unemployment Dynamics and Duration Dependence", Research Memorandum 1993-22, Vrije Universitet, Amsterdam. VAN DEN BERG, G. and VAN OURS, J. (1993b), "Unemployment Dynamics and Duration Dependence in France, the Netherlands and the UK", Research Memorandum 1993-38, Vrije Universitet, Amsterdam. BJÖRKLUND, A. (1989), "Klassiska experiment inom arbetsmarknadspolitiken", Forskningsrapport No. 37, The Industrial Research Institute, Stockholm. BJÖRKLUND, A. (1990), "Unemployment, Labour Market Policy and Income Distribution". In Persson, I. (ed.), Generating Equality in the Welfare Slate: The Swedish Experience, Norwegian University Press, Oslo. BJÖRKLUND, A. (1993), "The Swedish Experience", in Kongsh0j-Madsen, P. and Jensen, K. (eds.), Measuring the Effects of Labour Market Measures, Jennerop offset. BJÖRKLUND, A. and HOLMLUND, B. (1991), "The Economics of Unemployment Insurance: The Case of Sweden". In Björklund; A. el al. (eds.), Labour Market FORSLUND, A. (1992), Arbetslöshet och arbeismarknadspolilik, Bilaga 7 till Längtidsutredningen, Finansdepartementet, Stockholm. FORSLUND, A. and KRUEGER, A. (1993),. "An Evaluation of the Swedish Active Labor Market Policy: New and Received Wisdom", Paper Presented at the NBER-SNS Conference in Stockholm, December 2-4, 1993. FRANZ, W. (1987), "Hysteresis, Persistence, and the NAIRU: An Empirical Analysis for the Federal Republic of Germany", in Layard, R. and Calmfors, L. (eds.), The Fight Against Unemployment -- Macroeconomic papers from CEPS, MIT Press, Cambridge, Massachusetts. GOTTFRIES, N. and HORN, H. (1987), "Wage Formation and the Persistence of Unemployment", Economic Journal, vol. 97. GRÄMLICH, E.M. and YSANDER, B-C. (1979), Relief Work and Grant Displacement in Sweden", in Eliasson, Holmlund and Stafford, Studies in Labor Market Behavior, The Industrial Research Institute, Stockholm. GRUBB, D. (1993), "Some Indirect Effects of Active Labour Market Policies in OECD Countries", mimeo, OECD, Paris. GRUBB, D., JACKMAN, R. and LAYARD, R. (1982), "Causes of the Current Stagflation", Review of Economic Studies, Supplement, vol. 49, no. 5. GRUBB, D., JACKMAN, R. and LAYARD, R. (1983), "Wage Rigidity and Unemployment in the OECD Countries", European Economic Review, vol. 21. HAVEMAN, R. and HOLLISTER, R. (1991), "Direct Job Creation: Economic Evaluation and Lessons for the United States and Western Europe", in Björklund, A. et al. (eds.), Labour Market Policy and Unemployment Insurance, FIEF Studies in Labour Market and Economic Policy, Oxford University Press. HEYLEN, F. (1993) "Labour Market Structures, Labour Market Policy and Wage Formation in the OECD". Labour, vol. 7. HOLMLUND, B. (1990), Svensk lonebildning - teori, empiri, politik, Bilaga 24 till Lángtidsutredningen, Finansdepartementet, Stockholm. JACKMAN, R. and LAYARD, R. (1991), "Does Long-Term Unemployment Reduce a Person's Chance of a Job? A Time Series Test", Economica, vol. 58, no. 229. JACKMAN, R. and LEHMANN, H. (1990), "British Policies - A Fresh Evaluation", Working Paper No. 13, London School of Economics. JACKMAN, R., PISSARIDES, C. and SAVOURI, S. (1990), "Labour Market Policies in the OECD Countries", Economic Policy, vol. 5, no. 11. JANS, A-C. (1992), "Vad bestämmer den svenska arbetslösheten?", mimeo, Department of Economics, Stockholm University. JOHANNESSON, J. and ZETTERBERG, J. (1993), "Effekter av arbetsmarknadspolitiska átgärder — mikroutvärderingar", in Politik mot arbetslöshet, Betänkande av EFA, SOU 1993:42, Stockholm. JOHNSON, G.E. and LAYARD, R. (1986), "The Natural Rate of Unemployment: Explanation and Policy", Handbook of Labour Economics, North-Holland, Amsterdam. de KONING, J., GRAVESTEIJN-LIGTHELM, L., THOEN, N. and VERKAIK, A. (1992), Werkt de KRA?, Arbeidsvoorziening, Rijswijk. KORPI, T. (1992), Two Papers on Youth Unemployment Dynamics, mimeo, Stockholm University. LALONDE, R. (1992), "The Earnings Impact of U.S. Employment and Training Programs", mimeo, University of Chicago. LANGAGER, K. (1992), "Dagpengekarrusel eller jobgenerator? -- En forl0bsanalyse af deltagerne i arbejdstilbudsordningen", Samfunds0konomen, no. 6, Copenhagen. LAYARD, R. (1986), How to Beat Unemployment, Oxford University Press. LAYARD, R. (1989), "European Unemployment — Cause or Cure?", Discussion Paper No. 368, Centre for Labour Economics, London School of Economics. LAYARD, R. (1990), "Wage Bargaining and Income Policy: Possible Lessons for Eastern Europe", Discussion Paper No. 2, Centre for Economic Performance, London School of Economics. LAYARD, R. (1991). "Varför överge den svenska modeilen?", Ekonomisk Debait, no. 4. LAYARD, R. and NICKELL, S. (1986), "Unemployment in Britain", Economica, vol. 53, np.,5. LAYARD, R. and NICKELL, S. (1991), "Unemployment in the OECD Countries". Paper presented at the "Conference of Unemployment and Wage Determination", October 17-18, Cambridge, Massachusetts. LAYARD, R., JACKMAN, R. and NICKELL, S. (1991), Unemployment, Oxford University Press. LEHMANN, H. (1993), "The Effectiveness of the Restart Programme and the Enterprise Allowance Scheme", Centre for Economic Performance, Discussion Paper No. 139. LEONARD, J. & VAN AUDENRODE, M. (1993), "Corporatism Runs Amok: Job Stability and Industrial Policy in Belgium and the United States", Economic Policy, no. 17. LINDBECK, A. and SNOWER, D. (1988), The insider-Outsider Theory of Employment and Unemployment, MIT Press, Cambridge, Massachusetts. MCCORMICK, B.. and SKEDINGER, P. (1991), "Why do Regional Unemployment Differentials Persist?", in Skedinger, P. Essays on Wage Formation, Employment and Unemployment, Department of Economics, Uppsala University. MCDONALD, I. and SOLOW, R. (1981), "Wage Bargaining and Employment", American Economic Review, vol. 71. MEAGER, N. and METCALF, H. (1987), "Recruitment of the Long Term Unemployed", Institute of Manpower St'jidies Report 138. dc MUNNIK, R, (1992), Recent Evaluation Studies. Part Two, The Employment Service, Rijswijk, The Netherlands. NICKELL, S. (1990), "Unemployment -- A Survey", Economic Journal, vol. 100. OECD (1990), Labour Market Policies for the 1990s, Paris. OECD (1991a), Employment Outlook, Paris. OECD (1991b), Evaluating Labour Market and Social Programmes, Paris. OECD (1992), Employment Outlook, Paris. • OECD (1993a), Employment Outlook, Paris. OECD (1993b), "The Role of Structural Policies in Reducing Unemployment", Paper to Working Party No. 1 of the Economic policy Committee. OECD (1993c), "High and Persistent Unemployment: Assessment of the Problem and its Causes", Paper to Working Party No. 1 of the Economic Policy Committee. OECD (1993d), Economic Survey of Denmark, Paris. OHLSSON, H. (1993), "Sysslesättningsskapande átgärder som stabiliserings-politiskt medel", in Politik mot arbelslöshet, Belänkande av EFA, SOU 1993:42, Stockholm. OSWALD, A. (1986), "The Economic Theory of Trade Unions: An Introductory Survey", in Calmfors, L. and Horn, H. (eds.), Trade Unions, Wage Formation, and Macroeconomic Stability, London: Macmillan. PHELPS, E. (1982), Inflation Policy and Unemployment Theory, London, MacMillan. PILEGAARD, T. (1991), "Indsatscn for langtidslcdige. -- En Unders0gelse af administration of cffckl af arbejds- og uddannelsetilbud", Amternes og Kommunernes Forskningsinstitut, Copenhagen. PISSARIDES, C. (1985), "Short Run Equilibrium Dynamics of Unemployment, Vacancies and Real Wages", American Economic Review, vol. 75. P1SSARIDES, C. (1990), Equilibrium Unemployment Theory, Basil Blackwell. PISSARIDES, C. (1992), "Loss of Skill During Unemployment and the Persistence of Employment Shocks", Quarterly Journal of Economics, vol. 107. Politik mot arbelslöshet (1993), Betänkande av EFA, SOU 1993:42, Stockholm. Rapport om arbejdsmarkedels strukturproblemer (1992), Del II, Finansministerie«, Copenhagen. REGNER, H. (1993), "Choosing Among Alternative Non Experimental Methods For Estimating the Impact of Training: New Swedish Evidence", mimeo, SOFI, Stockholm University. SACHS, J. (1987), "High Unemployment in Europe: Diagnosis and Policy Implications", in Siven, C-H. (ed.), Unemployment in Europe, Timbro, Stockholm. SAINT-PAUL, G. (1992), "Are the Unemployed Unemployable", CEPR Working Paper No. 689. SAINT-PAUL, G. (1993), "Unemployment, Wage Rigidity, and the Returns to Education", Paper Prepared to the Congress of the European Economic Association in Helsinki, 1993. SHAPIRO, C. and STIGLITZ, J. (1984), "Equilibrium Unemployment as a Worker Discipline Device", American Economic Review, vol. 74. SKEDINGER, P. (1991), Essays on Wage Formation, Employment, and Unemployment, PhD Dissertation, Department of Economics, Uppsala University. STEINER, V. (1990), "Long-Term Unemployment, Heterogenity, and State Dependence: New Microcconomic Evidence on Unemployment Persistence", European Association of Labour Economists, Conference Compendium IIB. THAULOW, I. and ANKER, N. (1992), ArbejdslOshedens psykiske og sociale konsekvenser — En oversigt over nyere nordiske forskningsresultater, SFI Arbejdsnotat, február 1988, Copenhagen. VLACHOS, V. (1985), "Temporära lönesubventioner -- En studie av ctt arbctsmarknadspolitiskt medel", Lund Economic Studies No 34, Department of Economics, Lunds University. WADENSJO, E. (1987), "The Youth Labour Market in Sweden - Changes in the 1980s", Economia & Lavoro, no. 1, Marsilio Editori, Venezia. WADENSJO, E. (1993), "Arbetsmarknadspolitikens effckter pá löner och priscr". In Politik mot arbelslöshet, Betänkande av EFA, SOU 1993:43, Stockholm. ZETTERBERG, J. (1993), "Arbelslöshet, arbetsmarknadspolitik och löneför-handlingssyslem". In Politik mot arbelslöshet, / in footnote 6. With c = /, the revised Beveridge curve will not shift, although du = -dr for a given v. 9 This cannot be seen directly in Jackman et al. (1990). A recalculation is done in Calmfors (1993a). 10 In the special case of a vertical wage-setting schedule, the proportion of the labour force that is regularly employed remains unchanged (Layard et al. (1991)). So does the sum of open unemployment and programme participation as a proportion of the labour force. But since the labour force has increased, both regular employment and the sum of open unemployment and programme participation rise in relation to the population. 11 This case requires a more complex model of wage setting than in the earlier examples which introduces heterogeneity in the labour force and thus distinguishes between the reemployment probabilities for laid-off insiders and outsiders (Calmfors & Lang (1993)). 12 For Sweden, a few recent studies from the second half of the eighties have found less favourable results of intensified placement services than earlier studies (Delander & Niklasson (1987) and Behrenz (1993)). One interpretation is that such policy measures contribute less in situations of low unemployment in general, such as was the case in Sweden in these years {Politik mot arbetslůshet (1993)). 13 Suppose that the production function is Y - F(eN, K), where Y = output, N = the number of employed workers, e = the efficiency of an individual worker and K = the capital stock. The profit-maximising level of employment is then given by w = eF^eN.K), where w = the real wage. Only if the implicit labour demand function implied by this condition has an elasticity with respect to the real wage exceeding unity, will a rise in e increase the number of employed persons. 14 Letwt = the real consumption wage, wp = the real product wage, 0 = the tax wedge, W= the nominal wage, P = the price level, / = the income tax rate and r = the pay-roll tax rate. Then wc = W{\ - t)/P, wp = W{\ + r)/P and ß = (l + r)/(l - /). Hence wp = 9wc and a reduction of the tax wedge lowers the real product wage for a given real consumption wage. 15 If w = the real wage, r = the share of the labour force in programmes, and u ™ the unemployment rate, a transfer of openly unemployed to programmes (dr = -du) increases the Footnotes 1 The employment schedule is negatively sloped provided that the ordinary labour-supply curve is not more backward-bending than the ordinary labour-demand schedule. The employment schedule can also be thought of as indicating excess demand ia the labour market, measured as the ratio between demand and supply. 2 If w = the real wage, s = the re-employment probability of an unemployed worker, and a - other factors, we have in most wage-setting models that w = w(J,a) with dw}ds>0. The reemployment probability can be expressed as the ratio between the number of job openings and the number of unemployed job-seekers. If q = the probability of a quit, N = employment, U = unemployment, L = the labour force, n - N/L - the employment rate (as a fraction of the labour force), u - 1-n - U/L - the unemployment rate (as a fraction of the labour force), we have s » qN/U - qn/u - qn/(l-n). 3 If, in addition to the symbols in footnote 1, H = the number of hirings and V = the number of vacancies, the usual assumption is that H - H(V, U) wilhdH/áV > 0 and dH/äJ > 0. For employment to stay constant, hirings must equal quits, i.e., H{y,U)=qN. If the hiring function exhibits constant returns to scale, and v = V/L= the vacancy rate, one obtains H[v,u) = qn = q(l-u), which implicitly defines the Beveridge curve in Diagram 2. 4 As shown in Calmfors & Lang (1993), one can under certain conditions derive that w = w(j,o) also in a model where participation in a labour market programme is an alternative to open unemployment, if s is interpreted as the expected probability to find a regular job for a representative worker without one. Let, in addition to the symbols in footnote 1, c indicate the search effectiveness of a programme participant relative to an openly unemployed person, J, the re-employment probability of the latter, and R the number of programme participants. We then have s„=qN/(cR + U). Suppose that the proportions of those without a regular job in programmes and open unemployment arc y and (1-y), respectively, /.e.» y = R/(R + U) and (l-r) = U/(R+U). Then it holds that ^ =qN/[rc + (l-r)]iR + U) or j = jcj +(1-/)* =qN/(R + U). Letting r = RjL - the share of the labour force in programmes, and remembering that now r + u = l-n, it follows that s = qn/(r + u) = qn/(\-n) in this case as well. 3 Note that I have implicitly assumed that all programme participants belong to the labour force. • 6 Using the notation of the earlier footnotes, the hiring function for regular jobs is now assumed to be H(V,cR + U), where c again measures the relative search effectiveness of a programme participant. From H(V,cR + U) = qN, it is straightforward to derive the equation for 19 See footnotes 17 and 18. The point is well illustrated in OECD (1993a), where it is shown that the unemployment-reducing effect found by Layard et al. (1991) is no longer significant, if active expenditures as a proportion of the wage bill is substituted for the Layard et al. measure of policy activism. However, since programme expenditures do seem to increase with unemployment, this alternative measure is likely to introduce a simultaneity bias in the opposite direction. 20 In general, to the extent that employment reacts only with a lag to wages, the simultaneity bias is likely to be less severe in wage than in un(employment) regressions. Unfortunately, there appears to be no early way of handling the simultaneity problem, since this requires finding instruments that shift the government policy reaction function but not the employment (or wage-setting) schedule. 21 They can reject the hypothesis of a 3 percent earnings increase (over a 20-year period), which according to a rough calculation would have been required for the social gains to balance the costs. wage if a, > a, in a wage equation written as w = at-alu-a2r +... . Alternatively, since u = \-n-r, where n = regular employment as a share of the labour force, the equation can be written w = (a0 - a,) + a,n + (a, - a, )r +.... It follows that a1>a1 is also the condition for the wage-setting schedule in our diagrams to be shifted upwards. If a, <0, a2 <0 means that programmes create more wage pressure than regular employment. Two later studies, not summarised in Calmfors (1993a) are OECD (1993a) and Ohlsson (1993). The former conforms to the majority of earlier Swedish studies, whereas the latter does not find any significant wage effect of increased programme participation. 16 The estimations are of the form &]nw = ßt-ßlu-ß7k+..., where w = the real wage, u = the unemployment rate, k = brRJwN = expenditures on active programmes relative to the wage bill, br = expenditures on programmes per participant, R = the number of programme participants and N - regular employment. If we let br/w « 1 and RJN « RJL = r, where L = labour force, and r *» the share of the labour force in programmes, the equation can be reformulated as Alnw » ßt -ßp-ßj- +.... The equation is estimated on pooled time series and cross-country data for the 1985-90 period, but separate ß% coefficients are estimated for each country. For most countries, the point estimates indicate that ßz > /?,. 17 Heyien starts out from a wage equation like w = a„ -atu -a2r +..., as in foot-note 15, but where a, = 0. It is then assumed that or, depends upon, La., the size of labour market programmes. For instance, in one equation he lets a, = sc + ejr + s2A, where y = brrjbu u, is the ratio between active and passive expenditures, br = programme expenditure per participant, bu = unemployment compensation, and A = a vector of other explanatory variables. Hence w = at-{sí+sly+s1Á)u+.... This can be rewritten w = a0-(eB+elA)(r+u) + (st + s1A-slbr/bll)r. The wage effect of a transfer of openly unemployed to programmes (du = -dr) depends upon whether e^ + s^ 1 Etbjbu. Since van Heyien transforms the wage responsiveness parameter £■, to a variable giving the rank of different countries, it is not straightforward to calculate dwjdr from his regressions. 18 A typical estimate by Zetterberg is u = -0.13^-100+..., where in addition to the symbols in the earlier footnotes y = brr/(b/ + bmu). If we let br = bu, u = 7.0 percentage points and r = 3.0 percentage points, we can derive dujdr a -1.5 or dnjdr = d(l-u-r) a 05, i.e., open unemployment falls by 1.5 percentage points and regular employment increases by 0.5 percentage points when programme participation is increased by 1 percentage point of the labour force. br/bu > 1 gives even more favourable employment effects. Layard et al. instead estimate u = -0.13f • 100 +.,.., where y = brrjuy and y = GDP/capita. Letting bjy = 0.5 and u = 7.0 and r = 3.0 as above gives dujdr » -0.67 and dnjdr = if(l - u - r)jdr a -0.33. European Journal of International Relations The European journal of International Relations pays special attention to conceptual, normative and formal theories, seeking in particular to foster an awareness of methodological and epistemological questions in the study of International Relations. It reflects research and developments of both a conceptual and an empirical nature in the major sub-areas of the field, including Foreign Policy Analysis, International Organization, Strategy and Conflict, International Law, and International Political Economy. Your 1995 subscription ivill include.. • Neorealism and the English School: A Methodological, Ontological and Theoretical Reassessment Richard Little • The European Human Rights Regime in Comparative Perspective Andrew Moravcsik • Signals, Signposts and Symptoms: Interpreting Change and Anomalies in World 1'olilics jamcsN. Rosenau • Instability and the New Multidimensional Politics of Security: A Rational Choice Argument fór US- EU Cooperation John Peterson and Hugh Ward • 'Hegemonic' Leadership: Naked Emperors or the Worship of False Cods? Jarrod Wiener • The Limils of Self-Reform: Institution Building in the European Union Gerald Schneider ISSN: 0354*0661) Published quarterly in March, Junr, September and December Dnu 'I Ittrgcl. wc $iiiiiaiilcc lhal if yiiir arc iíissiiiisficit with i/oiir jimnitit in any luny, uv will rť/ulii/ the ami 11/vwir Mil'.-cripYmii. Newin 1995! Ediled by Waller Carlsnaes Uppsala University "The new European Journal of International Relations represents a landmark development." Professor Barry Buzan, University of J&arwkk Order Form —————— Send this order form lo: SACE Publications, 6 Uonhill Street, London EC2A 4PU, UK Tel: +44 (01171-374 06-15 US Orders to: SAGE Publications, I'O Uox 5096, Thousand Oaks, CA 91359, USA O Yes! I wanl lo enlcr my new subscription to Ihe Euru/iMti /»umdf 11/ l/ilrmaliiiirui Rtiiifiimi at the introductory rale Q Individual Kale al £24(Ľ«n/ 5381SW) •IhUMÍMStulT Q Institutional Kale al £90 / SH4 Daylimu Tel:__„------------------------------------- Ql enclose a cheiiuc {made payable' tu Sage Publications) Ql'lcase charge my credit card Q Mastercard Q Visa Q American Express Q Diner's Club Q Eurocard C.uJNuiiitvrl I I I Expiry Dale / Signature------------- Dale / l/'-V) Article PITFALLS AND DILEMMAS IN LABOUR MARKET POLICIES FOR DISADVANTAGED GROUPS - AND HOW TO AVOID THEM Ides Nicaise, HIVA. University of Louvnin, Belgium Joust Bollens, HIVA Untveristy of Louvain, Belgium Len Dawes, CLiYIS Uiiiversily of Leicester, UK Sliuhin Laghaei SFI, Copenhagen, Denmark Ivan Thaulow, SFI, Copenhagen, Denmark Minelle Verdié, TF.N, Paris, France Alain Wagner, CUPS, Luxembourg Summary This article summarizes the findings of an explorative research project commissioned by the European Commission's Third Poverty Programme, on inefficiencies and^iindesired side-effects of targeted labour markerpolicies:' These 'pitfalls and dilemmas' are discussed in relation to (hree types of socioeconomic effects that are cxpecteilfrom thesi polices: partners ami hcal_tion-profit qrganiz- tump'""""" imwo-ecnmmiic effects: it is argued that 'active labour market policies' are on the whole more effective at redistributing opportunities than at creating employment. In a situation of excess labour supply, one can hardly expected siipply-jjtj^measures to restore the equilibrium. Therefore, activating labour supply should lo in tandem with more structural policies. distributional effects: a number of explicit or implicit mechanisms of discrimination against vulnerable groups in existing policies are described: legal or administrative barriers to entry, .creaming off mechanisms, inconsistencies between measures. Even positive discrimination in favour of the most vulnerable groups appears to have its drawbacks. The article concludes that the abolition of existing discriminations, and the establishment of guaranteed, jcfa ' viccs^tojilj, unemployed people are to be preferred over positive discrimination, labour market outcomes for beneficiaries: low or even negative performance of policies are manifested in failures to meet the needs of unemployed people or of Ihe labour market, segmentation of provision, or other kinds of 'shortcomings' including stigmatization. Remedies are sought in a better design of policies for particularly disadvantaged groups, a rigorous application of the 'routine' principle, and partnerships between 'public agencies, social Résumé PliiGES ET DILUMMUS DANS LES POLITIQUES DU MAItCHF. DU TRAVAIL VIS-A-VIS DES UROUl'ES DÉFAVOUISÉS - liT COMMENT LUS É V ITER Cel article resume les résidtats ďiin projet de recherche exploratoire realise dans le cadre du troisiéme programme pauvreté de la Cominii-nauté sure les effets non desires el enlrainant une inefficacitc des politiques ciblées du mar che du travail. Ces 'pěges el dileinmes' soul discutes en function de trois types d'effets socio-econonii-ques qui son! attendus de telies politiques. I Les effets de repartition: nous decrivons 1111 certain nombre de měcanismes implicites et explicites de discrimination envers les Journal ofťiHufiniii Social Policy l*?S 5 (J| I V>-217 © Liirl^iii.ni Ciruup Lirtliled IHM 1MH-'J2»7M.S/U53u2m/S0j.i0 200 IDES NlCAISE ET AL. groitpes vulnerables tliws les pulitiques acluelles: les barriěres legales im administratives d'aecis, les méatiiisiiies tľécré-inage', les contradictions entre tlifféreiites mesítres. Les procedures tle discrimination positivu envers les groiipes les plus vulnerables semblent avoir égaleinent ties incon-vénients. l.'arlide conclm que ľabolitum ties discriminations existantes et ľc'tablisse-mement de services garantis ä lautes les personlies Sans emploi tloivent ilre préférés ä des mesures de discrimination positive. 1 Les résultats stir le niarché du travail pour les bcnéficiairest Les performances faibles voire negatives des politique! menčes sout attcstées par 1'écbec de recontrer les besoins des personnes sans emploi on du marché du travail, ce qui se traduit par exemple par line segmentation de l'offre de formation (on autre) ou par d'aulres formes ď imperfection y compris la Stigmatisation des iiucrcssés. Les remčdes sont ä rechercher au travers ďune meilleure conception des poliliques visant particuliěrc-ment les groupes défavorises, par line application rigoureuse du principe de 'guidance' et par le partenariat enlre les agences publiques, les partenaires sociaux et les associations locales. .i Les effets macw-economiques: les politi-ques actives du marché du travail sont dans l'ensemble plus efficaces pour redistribuer des opportunités de travail que pour créer tie I'emploi. Dans tine situation d'offre de travail execdentaire, on pent difficilement attendre des mesures se concentraut sur l'offre qu'elles restaurant ľéquilibre. Var consequent, des mesures actives doivent s'accompagiwr davanlage de mesures striiclurelles. Introduction Our article describes a number or mechanisms that cause inefficiencies and undesircd side- effects in targeted labour market policies for disadvantaged groups. It summarizes the results of a transnational and multidisciplinary - though still explorative — research project commissioned by the Kuropěni) Commission's Third Poverty Programme.^ The 'poor' as a target group are very sensitive as a touchstone for evaluation, because of the gap that separates them from the average job seeker whose needs are mostly used as a yardstick in the design of policies)) Our study is focussed mainly on five EU-member states: Belgium, Denmark, France, Luxembourg, and the United Kingdom. It is based on case studies and a literature review. Throughout this article, we will make flexible iisejif terms such as 'poverty', ^Hf-" advantage', and 'social exclusion'. Generally tí speaking, wc refer to 'tlie poor' as individuals and groups whose material, cultural and social >, -' " resources are so limited as to exclude them jf from the minimum acceptable way of life in ,-^ the country which they live (see the definition 's' .given by the Council of the European Communities in its Council decision of 19 > December 1984). The more recent shift in lex- J minology from 'poverty' to 'social exclusion' is essentially meant to emphasize the multi-dimensionality, the duration, and the structural causes of poverty. However, in a policy context, one has to acknowledge the gap between the theoretical definition of concepts and their operational use. Target groups of policies, or administrative categories, are never defined in such a holistic way as to represent exactly the" 'poor' or the 'socially excluded'. Moreover, it is indeed useful and necessary to consider the poor in connection with related social groups (for example low-skilled or\ long-term unemployment, minimum income | recipients, school dropouts, workers in the ( secondary labour market, single-parent faiiiil-, ies, disabled persons, and migrant workers) \ because these are the categories targeted by policy makers and with which they either over-hip or compete on the labour market. Our analysis is confined to policies aimed to (re)inicgrate unemployed people into the reg- Juttntiilof Ettiapean Social ľolicv \ť)S S (i) LABOUR MARKET POLICIES 201 ular labour market, leaving aside, for example, preventive actions to maintain jobs, or policies focused on the upgrading and regulation of working conditions. Within the (rc)iutegration strategies, we concentrate on selective programmes targeted at disadvantaged groups, but we also consider general programmes as far as they are relevant for the poor. The analysis covers a wide range of actions, including (pre)vucational education and training, recruitment incentives, direct employment programmes, 'social economy initiatives', orientation and guidance, as well as all kinds of mixed programmes. The expected economic effects of targeted labour market measures can be roughly classified into three categories: 1 distributional effects: does a measure redistribute employment in the right direction? To what extent is the target group covered? What are possible side effects for other groups? 2 labour market outcomes for beneficiaries: does the measure enhance the employment opportunities, earnings, quality of jobs for tlicrarger group? 3 macro-economic effects: can one observer effects on the general level of unemployment, wages, the government's budget? Our discussion of pitfalls and dilemmas will be ordered accordingly. 1 Distributional effects Prom a distributional point of view, the so-called Mjttxliswjjffect,1 or the 'law of perverse disiribution' is a well-known phenomenon in the literature on social expenditures (Deleeck et al. 1983). In the context of labour market policy," it refers for example to the over-representation of young, short-term, highly skilled job seekers in training programmes. However, this phenomenon should not be con- sidered as a religious, nor as a social or natural law; it is also the consequence of wrong policy design, and hence, can be avoided. Legal and administrative barriers to entry A lirst possible explanation is that the most disadvantaged groups are almost by definition not eligible for participation in mainstream labour market programmes (such as wage subside schemes, enterprise allowance schemes, or training). Most programmes arc confined either to registered unemployed people, or even in sónuTčase (o those entitled to unemployment insurance benefits. In Belgium in 1993, only -175,000 individuals who were 'remunerated full-time unemployed actives sj^kjng work'had access to programmes such as Guidance Plan, Youth Recruitment Plan, or the Enterprise Allowance Scheme. Therefore 3 12,000 other - registered — unemployed people (including those part-time unemployed, disabled people, luiiiimuin income recipients, other unemployed not entitled to benefits, and the unemployed aged oyer 50 who are automatically 'exempted from control') had no access to the programmes - quite apart from several tens of thousands of unregistered unemployed people.3 '" Governments and administrations have an obvious interest in narrowing thc^definitiou of uneniphíýineut, and indirectly, in setting artificial boundaries to the labour market. The differences between official cnüiTts'aild the Labour Force Survey, which uses the International Labour Organization definition of unemployment, are sometimes substantial. An estimate for the UK (based on l.awlor and Kennedy 1992) is that the number of people looking for work but who were not registered in the official unemployment Louut, over the years 1984 to 1991 averaged 841,000 Poluy makers should realize that in this way they tend to reinforce the exclusion of marginal segments from the labour market. Imtmattit l ttiope.iii Smi,!! /Vlil IWMÍ 5 I)) 202 li)ĽS NlCAISE ET M.. Recipients or a iniuintum income guarantee MIC) are a particularly vulnerable group in this context. In Denmark as well as in Belgium, they have no access 10 mainstream labour market programmes. Single parents in the UK, the long-term unemployed (I.TU) classified as 'disabled' in the Netherlands, and the older unemployed in Belgium and ľrance arc other examples ol disadvantaged groups that are considered as being 'out' of the labour market in some countries (yet not in other countries) and therefore ineligible (or eligible only as an exception to the norm) for programmes designed to combat exclusion from employment^ Analyses of the living conditions and socio-economic profiles of these groups indicate that many individuals from these groups live in relative poverty. 11 Creaming off: the false dilemma between : leqnity and efficiency? 1 Sibille and Vcrdič (1992) carried out a detailed case study of the ľrench AIF-scheme (Actions d'lnsertion et de formation). This scheme is targeted specifically at the long-term unemployed (LTU) and within this category, those designated as priorities were: people who had been out of work for a very long period of time (more than three years), the over-50s, and people in unstable situations, such as those in receipt of a minimum income allocated ior integration purposes (RMI - Reuenu Minimum d'lnsertion). The study reported a large percentage of people not officially entitled to bendit from the initiative. Among the five departments studied, the most 'dynamic' department can boast the greatest number of people who should have been disqu.ihriLtl Iroin the initiative because they had been registered for too short a tunc with the employment office (one m two), it also had the lowest percentage of 'over-SOs' (5 per cent) and the highest proportion of advanced training course (for more than half of the trainees). The evaluators identify a number of explicit as well as implicit mechanisms that explain this bias. TJjejjfjicers from the employment agency themselves claim their selection procedures arc based on; T i efficiency (priority is given to people who árcihoiigliT ti) be most 'motivated' and liávc a goal in mind, as thcy'are thought m have the best employment prospects after participation). Or then again when the officers choose people from the short-term unemployed category, they feel they are helping the people to avoid 'joining the army of long-term unemployed later on' 2 recruitment difficulties ' J 'humanitarian' reasons: it would be ini-póšsjble not to^allow someone to participate in a scheme on the grounds that he or she has not been out of work long enuugh. Another possible criterion, one that is less explicit, is based on the fact that the All'" initiative includes on-the-job training and the c^edibjlit^of the cmpbynientsctvice pfficcrs is ať stake in the eyes of the employers to whom the officials might be induced™"«) propose candidates later on (this primarily concerns the employment offices). It is therefore tempting to select trainees who are most likely to be employable, who already have some skills and have not been out of work for too long a period. Training bodies that are awarded.the 'qiul-.-. ity clause' by the public employment service ;ľre~em1[ľed to larger subsidies and sometimes the clause is conditional upon the number of trainees that lind a job immediately after completing a scheme. It is obvious, then, that this criterion does not encourage the training bodies (who in many instances select the trainees themselves) to pick unemployed people who are the most disadvantaged and those who have been out of work for the longest periods of time. Similar funding rules are applied in the UK, where the large programmes of centrally funded training (Youth Training and Employ- hmw.ll iif tiurupi;m Soci.il ľtilicy IVS-S 5 (.1) Laiiour Market Policies 203 ment Training, now called Training for work) are delivered through local Training and Enterprise Councils (TECs and LLCs). Their funds are ipure deliberately related to the 'output' they can deliver ('output related funding'), specifically qualifications and olfers of employment; they are therefore implicitly encouraged to select and favour the better qualified/ t adjusted and speimness^ň^lieJ1mMíe-wLMf! Sua.il I'ulicy liiyi 5 (J) 28 204 IDÜS NlCAlSEtT/tL. tiinities that would otherwise be opened for better situated job seekers. With reverse substitution, employment opportunities are diverted from those at the end of the queue towards more privileged groups, and poverty actually increases. To illustrate the real danger of such effects, consider the discussion about 'preventive' versus 'curative' policies vis-ä-vis long-term unemployment. The problem is particularly topical in Belgium, where the regional govcriimuir'~of~FľYmirŕs has taken several measures tu f tvour of the most disadvantaged groups, whereas the tederal government has focused us str.HLgy on 'preventive' incisures The pohucs are obviously complementary, but to some extent also contradictory hor example, the Flemish Wecr-Werh nctie ('back to work' action) aims to reactive long-term unemployed on a voluntary basis through a ^combination in counselling, training and work experience measures.^The federal 'Guidance plan' interferes to some extent with this scheme by providing (compulsory) extra assistance to the 'almost LTU', including,counselling, training and wage sub'sidies^ln 1993, in dffi"1S5ňTexTÔT very" sluggish economic growth and upsurging youth unemployment, together with the abolition of military service, the federal government launched a Youth Recruitment Plan (YRľ) with wage subsidies for youngsters after six im>Wľrs3?f3|ľSH|?!oyin5ítr'TKHe* wagT~síibsidics arc nearly unconditional and im^lnHirrcuiftťŕif£iiVelhäinhecxištri^'ŕecrui(-ment incentives for LTU. Whereas the Youth Recruitment Plan has had an unexpected success, labour market analysts fear that resources and opportunities are, to some extent, diverted away from the hard core of the unemployed. Since the start of the YRP, within six months, the share of vacancies reserved exclusively to youngsters has risen from 6 to 36 per cent (fioldcrbeke 1994). Moreover, some of die most disadvantaged among the unemployed youngsters, those who arc not entitled to unemployment benefits, have no access to the YRP; even young LTU who are entitled to benefits and who had the opportunity to get some work experience in the Weer-Werk action, are (temporarily) excluded from the plan. In December 1994, criticism of these inconsistencies persuaded the Minister of Employment of the need to harmonize different types of wage subsidies. A similar problem has been noticed in sonic northern EU-member states, with ESH-iuterventions focussed equally on youngsters ('ĽSF-objective 4') and LTU ('objective 3'), with a real danger of reverse substitution at the expense ofLTU. This danger was accentuated by the fact that the relatively well-educated í young unemployed people are more casify j reached by policy measures than the LTU, who ? arc often discouraged by previous failures and poor employment expectations. As the budgets for objective 4 were exhausted mote quickly than those for objective 3, a tendency to shift budgets from LTU to youngsters was observed in some countries. Insufficient supply and within-group substitution To explain the relation between insufficient supply and substitution within a target group, let us consider a simplified (theoretical) case: suppose that a new job search assistance scheme is targeted at the LTU, but that, due to budget constraints, current resources are merely reallocated to assist half of the target group (selected at random) more intensively, instead of serving the whole population in a usual way. In this setting it can reasonably be expected that the How of vacancies to the LTU remains unchanged. It is obvious that the employment probabilities will be merely redistributed within the group of LTU: beneficiaries will improve their position, whereas the position of non-participants from the same target group deteriorates. To our knowledge, the concept of substitution within target groups líášo'eéh'Téčôg-TYized w"butiiever mcasureti in practice^ However, obvious cases of within-group sub- jutitnal of t'.utopean Social ľttticy 19Ü5 5 (J) Labour Market Policies 205 stitution have been observed in some mandatory programmes. In the UK, it has been argued that changes in legislation which were designed to ensure that young people (aged 16 to 18) are either in education,jur in employment or vocational training have resulted in numbers of young people becoming homeless or otherwise marginalized.JThis is because income support was withdrawn from those aged 16-18 unless they were registered as attending a Youth Training (YT) scheme (those in employment receive wages and those in education are assumed to be supported by families). Since there was a shortfall in provision of YT places this left a number of young people with no income. The number of places for YT fell from 389,224 in March 1988 to 274,000 in 1993. It is estimated that around 50,000 homeless are living in London alone and about 150,000 youngsters become homeless each year (The Guardian, I June 1994). There has been an attempt to ameliorate this situation recently by changes in the rules on income support payments, but the basic situation remains unchanged. Similar drawbacks have been noted in the Netherlands with the introduction of the Youth Guarantee Act (JWG - jcugdwerkgar-antiewet) in 1992, which caused the exclusion of thousands of youngsters from social security due to a lack of (suitable) provision, and in Denmark, where the most vulnerable unemployed lost access to AMU-training courses (AMU->* Atbejdsmarkedsuddannelse 'Labour market education') due to insufficient supply. I'ositive discrimination has drawbacks too There are many examples of exceptions to the normal eligibility criteria for programmes, which result in a form of positive discrimination: for example in the UK exceptions may be made for women who are returning to economic activity after child-rearing, for recently released prisoners, for people with disabilities, for people recently quitting the armed forces, for those with difficulty with literacy or numeracy, and in other special cases such as in areas with large-scale redundancies. In Belgium MIG-rcccivcrs with the status of head of household', and those under 25 years of age enjoy priority treatment in some cases, while in France there is a general category ot exemption tor"~"those~''experiencing problems with exclusion, áTwčlľ as priontv status tor those Iffiemployed for loiigerjhan three years, for MIG-recipients who have been unemployed for more than one year, for unemployed people of over 50T"|üld~for people with disabilities. ~....." While there are clearly examples of good practice aimed at positive discrimination, one question which must be asked is how far these priorities are substantiated by objective studies of the relative disadvantage of different groups. For example, in many countries, eligi-bility for lalSouTTíúTŕkinľjäŕogrlunmeš[is closely linked to the iluřalioii of one's employment sl^élf^Sč\\~íp\SiSAe'óíläňéma\oyment begins a "hew time period, so a person who repeatedly leaves the unemployment register, perhaps to undertake shorr spells of seasonal work, will never reach the duration threshold which provides access to a scheme. In other words, intermittent workers with an equally long Imemploynienlľ'experience are implicitly discriminated against by the mere definition of unemployment duration. This amounts to a special case of 'within-group substitution' or even 'reverse substitution'. AnotKríjííestufň is'luiw well the rules are publicized to employers as well as to the population groups who would benefit most by taking advantage of them. It must also be acknowledged that t|etailcd priority rules tend to make legislation and administration more complex - sometimes even incoherent - and that in sonie casesthey.arcicathcr designed for 'Treasons of demand management than for söaaTrcasons. AsTcônscquencc, the question "Třřiiaňls"1 "Wlféilíer positive discrimination encouragcsHiiidcrutilizaiion or indeed a better targeting of measures. JtwrHtil o/iMitilKiiit Sucijl I'tilkv mW 5 til 20í I m; s Nk.ai.si-: htm.. A final drawback of positive discrimination is the risk ol siigmati/.atinn of bencliciaric's. litis itupíTes iliat (iulilěíí handicaps arc rcl-egatcd to employment services or to employers by the mere fact of 'benefitting' Írom a special programme. For example, in Luxembourg, the hencliciarics or the RMG (iniminnm gu.iran-tced income), received, during the lirst years ol implementation ol the law, an introduction card to the employment services coloured differently from thai of other job seekers. Needless to say, this scheme proved to be highly discriminatory. It has, in fact, siiilľ been abolished. 2 Effects for ihc beneficiaries A number of recent evaluation studies of selective labour market policies have pointed at poor, and sometimes even zero or negative employment" el^frčíT'Isľee'fíň^'exälnpli^ÔĽC.'íy l^?i!I71'^?řSí""''I"'l ' WeSreřgllíd-Níělšen Í99.1). in the absence of deeper analysis, one would lie tempted to attribute the causes of the failure of the target group or to the concept of the measures as such Yet there is some evidence ol poor implementation or management of the measures, which can be avoided in the future Mismatches between provisions and V needs of disadvantaged groups In Denmark, the mandatory training courses for long-term unemployed people within the framework of the job offer scheme have been subject to much discussion in recent years. Not more than 1.3 per cent of the participants appear to have a job within six months alter leaving iheéóúrše (Thaulow and Anker 1992). According to JenseneTu/. 'f í 992)the employment effects of the courses are even negative, compared with a control group of non-participants with the same profile. Wcstcrgárd-Nielsen (1993) found positive effects for Itmnliilit/ ľ.Hriiľťtiit .\'nťi.l/ ľii/ii v IWS 5 111 employed participants in AMU-courscs, while the same courses yielded negative employment results for I.TU. Among the many reasons put forward, tine possible explanation is a mismatch between the scheme and the needs and cäplflnTutélľoTĽTUnin the survey"by Thaulow and Anker (1992), about 25 per cent of tlie trainees declared th.it the courses did not suit at all their abilities or needs; consequently, about -II) per cent of the individuals in public loh-olfers said they had no experience whatsoever with the work they were supposed to do 1 Ins in turn caused demotivation on the . ""parr ol tu.inv beneficiaries. * " V There is surprisingly little empirical evidence about the needs and motivations of disadvantaged juli seekers. However, the conclusions ; converge On the following points. Some long- ' term unemployed have had problems at school i > and thcir^fcar of repeatedTáilifre^čan prevent iKcm from attending courses. The best way of overcoming this problem is that training courses match the qualifications and stimulate the self-confidence of the unemployed (rather TnluimalTňig tliem compulsory). One shortcoming of many training schemes is that they are too short for groups who lack basic skills such as literacy and numeracy, or baste work attitudes Or in some cases, pre-quatifving courses exisr but are not linked to subsequent stages of training so tfiat the process stops halfway Trauung is generally considered as an investment with an uncertain outcome. The 'costs' tif this investment on the part of the unemployed imiyTie'iaTree io \a:r._ 1 direct material constraints, like the linan-číäPcost (for example: transport, child care, material) or suuplv the absence of facilities to respond to these practical problems Z indirect financial costs, participants have tefposipone or reduce jjieir job search activities while they attend courses and hence are increasing their expected unemployment spell in the short run. Policymakers should acknowledge that neither Labour maiikli Poi icuis 207 the direct material cosis, nor the indirect costs can be borne by poor participants: in other words, adequate remuneration of participants is an essential condition of success 3 it is also necessary to remove the psychological constraints, in Tinier"tif enable tmempioyed people to participate in programmes and achieve good results. Obstacles such as a negative self-image, ŕc^r^ärrfailure's7tJireä'räPexciu'siäii^"fafá,l-ťsm should be rerií3ve3"By~íinkiiíg_ provision with social assistance in a holistic approach. Also, positive support seems to be a better incentive than pressure and sanctions. Some of the most disadvantaged unemployed people (older, totally unskilled, long-term unemployed people) are less keen to participate in training than the average job seeker, lor them, training is not i«^ssarily,..tji!ej)ip.sjt. strategy to šwR^wiřlí. When the expected rcturíľôTf training is lower - for example, for reasons of agc_or fojmerschool^ failure - or when the financial rieedjssojargent jliiit it does not allowT—fiSr^ptHtponed incomes, direct employment seems to be a better response in the short run, possibly supplemented by training in a second stage. This justifies the strategy of many local employment initiatives, some-tnrrt's-Siíed"TsócíäTeconouiy projects', which give priority to (non-profit) employment over training. Í" .«,„. Considerations of this type have induced the Danish government to modify its job and training offer scheme through a diversification of services, with different options at each stage. In this way the pressure on the mandatory scheme is also mitigated. Mismatches between provisions and needs of the labour market There are different, yet all unsatisfactory, ways of measuring the degree ofmaiching bctyveeii^ provisions and needs of the labour market. A lairlv sini|)le method is to ask ex-participants |e g. ex-trainees) who had lound a |oh alter the programme how lar thev jiercetve a correspondence between the content ol the programme and their subsequent emplovmenr. A recent evaluation stink (which used this method) ol the Flemish public training agency revealed a 'complete non-correspondence ratio' ol 10-40 peF cent tor \ oc.itiun.il training courses. Although these hgures cannot be compared to similar studies in other countries, they do not seem very uncommon since on the whole, these courses appear to be fairly effective (see Bollens and Nicaise 1994). Despite the difficulties encountered in measuring the phenomenon of (mis)matching, there is much 'anecdotal' evidence of provisions for disadvantaged groups being geared towards weak labour market segments. This applies to training as 'well"liš"job creation projects, especially in small scale, local initiatives. Hie economic jiotentiiil^ for local initiatives of this sort is_a_snialj_iine. Attempts to develop businesses and ucntc jobs through? initiatives the 'classic' economic sector often run aground because of comnjniioti and the resistance of companies active in the sector in question. On the other hand, new economic sectors (high-tech industries, information technology, electronics and Communications) are still underdeveloped in Europe and they are often beyond the reach of local initiatives (because of a lack of expertise, professionalism and funding on the part of the instigators of the schemes, plus the inadequate qualifications of those taking part in the initiatives). Thus, local initiatives tend to work well in areas^haTanT not brand new but h ive been lbandoncd by companies for failing to give an adequate return on mvesiliient Several studies of local projects in turopc show that a good number of them tend to offer training or employment in traditional activities such as recycling, clearing and rcnuva}ion, gardening, ŕésttíratioiClíčTglT-bourhood jicryiccs, promoting local products, or activities carried out in the context of farm co-operatives. tmtniiil u( littiutimm Stuial t'oticv iyv5 5 til 28 208 iDHS NlCAISE ET AL. The choice of offered activities may also have an impact on the financial situation of the projects - as illustrated in the evaluation study oT'söciäl workshops' in Flanders (Vanhuyssc aiiťTniaíEeTTWJOrpour outcomes (in lin.ni-cial terms) in some local projects arc not sokly attributed to the personal characteristics ol the target group but also to poor management, lack of experience, and an inappropriate choice of proposed activities (like gardening, maintenance, recycling activities, and subcontracting) whereas activities within the nidus trial sector bring better results. Inappropriate funding methods may them selves affect íhé~labour"inarket"oiitcoiiies of provisions. A study of ET (Employment Training - now Training for Work') in the UK by Alan Felstead (1994) indicates that this programme is largely inclined to provide skills which are not in much demand - paradoxically, due to the output related funding of the local training companies (TECs - see above). When the cost per output in-some area such as construction and engineering trades is higher, rhc~ TECs have more economic incentives to provide output in low-cost areas such as clerical and secretarial occupations. This will result lirstly in a relative excess supply of these areas. A second cônšct|UĽUCc is that training providers are being segmented into providers of specialist (Colleges of Further Education or dedicated workshops) ami large providers of low skill trainees. Segmentation and sidetracks The recent movement towards decentralization of efforts and resources aimed at Lonibatiug unemployment may have made it possible m many cases to differentiate the pro-iisiou and to respond m a more effective way to the needs of the more vulnerable groups of uuemplnved. Nonetheless, this tendency implies, to some extent, a rúk_ix(,ft^roducing the existing scgmcnta_ti;^>jui.^_iJtt^J^boj!,t_ market. There are examples where die most Tfišádváutaged candidates are put on sidetracks ' which in one way or another cut off the way to the regular labour market. 1 raining markets are typically split up „ t among three different types ol providers. 1 employers or social partners are mainly luTcrcstcd in updating or renewing skills of their (potential) employees according to the changing needs of the labour market 2 public employment agencies cover an intermediate held their primary aim is to reduce unemployment by retaining job süükerS, rlius-eiisiiŕiug' a better inajdi between labour demand and supply i dually, local non-profit organizations offer /." training programmes that reflect "the spccijic nepds_or~disadvantagcd groups, v rather than needs of the labour market. All three types of providers have their strengths and weaknesses. However, in some ; j f „., cases, unhealthy competition leads to discriiní- í i milion against the private non-profit-sector, which then faces problems of insecure and irregular funding, and hence, a higher turnover of staff, discontinuities in programmes, weaker management, and poorer outcomes. Similar segmentation processes can be observed among local_job crcation; initiatives, where again three models can be distinguished: 1 'work experience initiatives ' offering just a tem^ffipjobroftčnconibincd with on-the-job training, with a view to transferring flie worker to the regular labour market within a well-defined time limit. The 'enterprise d'insertion' in France are the typical example 2 'social . co-operatives' (Germany, Italy) luňEfloTinTg^u^^i^sj^ujiím^řcial basis, whose aim is to create new and permanent jobs in a 'nicjie^of the market J 'social workshops' (Netherlands, Flanders, (lermanv) that work on a non-profit basis, offering permanently subsidized jobs to the least 'employable' |ub seekers. Itmntal »I í iinf/itiui Son.it ľiilicy W95 5 [1] Labour Market Policies 209 Every type ol initiative seems to respond to the needs of a specific subgroup. But social workshops, which depend more heavilylm subšíílíešj tend to become more vulnerable themselves when the siibsidies^í^nnľrdeyuäfS____ As"a conš«pTéňce7 tličír poorer economic performance tends to be traiisTefrctTfo the target gřo"Up~and to act as a self-fulfilling prophecy, thus indeed limiting the mobility of participants towards the regular labour market (sec the section on somatization below). Dead ends The solidarity for work (CES - 'Contrats emphi solidaritě' contracts represent one of the major instruments of labour market policy in France (660,000 contracts in I99J) offering half-time jobs for a limited time in the non-market sector (including a training offer in some cases) to the vulnerable unemployed (like LTU, MIG-receivers, young unemployed with no vocational qualification). While this instrument represents a first offer of integration to a large number of people suffering from social exclusion, it has to be noted that for a large group of the participants, this does not go beyoncUa temporary solution. In too many peases CES contracts are followed by a new f period of redundancy. A regular work contract 'as a result of a CES contract is certainly not the rule. The problem with the integration of disadvantaged groups in public employment programmes is also recognized in the Netherlands, namely in the 'labour pools' (banenpools) for the LTU, as well as in the Flemish 'back to work' programme (Weer-Werk), and ill the Luxembourg ATI-Scheme (Affectation Tem-poraire d'lnsertion) for MiG-recipients. In the present context of mass unemployment, temporary jobs in non-profit activities normally end with backsliding into unemployment, yielding at most a renewed contract with the labour market and a reactivation of job search activities. On the other hand, the discouragement ar the end of the employment pcnočľís all thTgrcliterr"''~J'"'''J'~**""J"" "M~ ~""~'' It would be worthwhile examining whether — such work experience programmes would yield more stable employment in the private sector. But in many countries there seems to be a reluctance on the part ol pohev makers to subsidize private employers m the profit sector for luring disadvantaged ]ob seekers, mainly for fear of substitution and displacement elfcctiTX^\?^íeÍííiitalecms to emerge here, between 'additionälifý' and 'integration in the regular láBiuTrliiaŕkeť, The pniicíplíToŕ \ul-ditionality' of job creation programmes is indeed sometimes operatinnalizcd through the deliberate creation of separate employment circuits, well outside the regular labour market, but possibly with fewer prospects til stable contacts. In the light of the above analysis, this approach would need to be reconsidered: if 'addirionality' implies further segmentation of provision, lower transition to regular jobs, and hence, less stable employment results, it would perhaps be preferable to accept more substitution. Carousels This mainly refers to the idea that different arms or the administration may attempt to shift the responsibility tor unemployed clients from oneuf another, asthe desire to reduce their costs, reduce their degree of responsi-bility, or achieve administrative targets takes precedence over the needs of clients. As a result, the effect of integration programmes is sometimes íery'téniporary and does not imply any long-teřm^přogresT on the part of the beneficiaries. In Denmark there arc examples of carousels in the Job Offer scheme, and in the employment projects run by the municipalities, MIG-clients are employed just long enough to render them eligible for the (nationally funded) Jtmrn.ilof liiim/HiMSua.il Policy IW 5 {.11 99 5 8422 34 68 210 IDĽS NlĽAlSE ET AL. unemployment insurance scheme. As the income maintenance of die uninsured unem-ploved puiple is expensive for the municipalities, some of (lie municipal provisions are used, not tu integrate those unemployed, but to make them eligible to claim hcnelits from the insurance system. As soon as the unemployed become eligible, ihev arc no longer emitted to henelits trom the sucinl assistance system. In this way part of the measures are managed to serve institutional interests, which results in "tlie unemployed bHiigpušhcd from one system to another. In Uclgium a similar carousel is thought to be the temporary work programme designed [to move clients from social assistance back 'into social security (art. 60#7 of the law on social welfare centres). This is illustrated by some striking findings from an evaluation study carried out by Van de Vcldc (1990); {retrospective interviews with Jľ73 individuals fwho had participated in the scheme three years, earlier revealed that nearly all of them had indeed gamed access to unemployment insurance. However, 91.5 per cent had fallen back into unemployment immediately after the scheme; 61.4 per cent were unemployed at the time of the interview, while 48.5 per cent had remained unemployed for the whole of the period following the scheme. Those who did fail back into long-term unemployment run the tisk, of then being excluded from unemployment insurance, and having to teturn to social assistance. Two important reasons for these poor results wcriTflic absence of training offers (only 5 per cent had .mentlSii a formal training course) and the lack ul guidance and mediation. As a consequence ofthese findings, the scheme was reorganized so as to include training and a guidance for all participants. A second evaluation, three years after the reform (Woutcrs vl al. 1994) revealed that fewer participants (42 per cent) had fallen back to persistent unemployment despite the economic recession of the early 1990s. In a third phase, the scheme was again improved by including intensive guidance and counselling services. In Luxembourg, another kind of carousel is Jouniill of liuwpnitl Sna.it t'olicy 1*95 5 (.11 observed by field workers, who call it 'social and training tourism'. Among young jrcoplé taking part in initial vocational training in 1990, .10 per cent had participated in one or more programmes before; and 60 per cent had at least one job befote with an average employment duration of more than 20 mouths. Only 24 per cent of participants had had no work experience and no training. Stigma effects We talk^about stigma effects, in the narrow sense,