The role of training and skills development in active labour market policies


Nigel Meager, Diretor, Institute for Employment Studies, University of Sussex, Brighton BN1 9RF, UK. Email:


There is persistent evidence over several decades that the UK lags behind its international competitors in terms of the skills and qualifications of its workforce, with a detrimental impact on overall economic performance. The most recent attempt by the UK government to address this includes a new strategy aimed at increasing the degree of integration between skills policy and employment policy in the UK. In light of this development, this review paper considers the extensive international evidence on the role and effectiveness of training and skills interventions, as part of a broader portfolio of active labour market policies. The review concludes that while large-scale, ‘broad brush’ schemes have little impact as part of such a portfolio, more targeted programmes addressing specific skill needs may have some impact on employment chances of workless groups.

Introduction and background

This review originated in an attempt to inform current policy debates in the UK with evidence from the international research literature.1 Comparative analyses of the UK's education, training and skills performance, and its links with the country's overall economic performance, have been repeated at intervals for several decades, with uniformly bleak conclusions (Keep, 2008). Policy makers have responded with frequent reforms to the vocational education and training system. The most recent such analysis is the Leitch Review of Skills (HM Treasury, 2006), to which the government response has been the creation of the UK Commission for Employment and Skills, a key task of which is to advise on how the range of skills-related policy measures can be integrated with employment-related measures as part of an ‘integrated employment and skills system’ (the two groups of measures being currently somewhat fragmented between government ministries and their various delivery agencies).

One likely implication of this integration is that skills will play an increasing role in the wide range of supply-side-based labour market measures (the New Deals and related initiatives), which are the key means by which the UK government hopes to meet its key targets in this area (in particular, an overall employment rate of 80 per cent).

It is, therefore, timely to review what is known from the large international literature about the effectiveness of skills interventions within the overall active labour market policy (ALMP) portfolio. The case for doing so, at least in the UK context, is reinforced by the repetitive nature of the policy debate over time, and the apparent failure of policy makers to draw on previous research or international evidence. It is not fashionable in policy circles to acknowledge that policies and measures that existed in previous decades (particularly those introduced under different political regimes), may have lessons to teach about what works and what does not. The pressure for novelty may blind policy makers to the possibility that the problems they face are similar to those faced in previous generations, and that solutions they come up with may also be similar to those tried before (or tried elsewhere). Those who remember the many reviews of the UK skill shortage and productivity problems produced in the 1980s by the National Economic Development Office2 (abolished by the Major government in 1992), may be excused a distinct sense of déjà vu when reading the lamentations on the same topics in the Leitch Review, currently driving policy in this area. Similarly, the policy thrust emerging from Leitch (proposals for a demand-led system, with employers in the driving seat and new employer-led bodies to deliver it) is heavily reminiscent of the almost identically-argued rationale for the Thatcher government's creation of employer-led Training and Enterprise Councils (TEC) in 1989 (the main difference is the TECs' local focus has been replaced in the Leitch era with a sectoral emphasis).

This review focuses on one important aspect of the Leitch policy conclusions: the move towards integrating skills and employment policy. In particular, the emphasis is on that area where current skills policy and employment policy overlap, namely the provision of training and skills development to participants in active labour market measures. The bulk of the literature reviewed consists of empirical evaluation studies of the impact of ALMPs, particularly that evidence, which allows for comparisons between the different components of ALMPs, so that the effectiveness of training measures can be isolated from that of other measures within the ALMP portfolio. The literature examined relates mainly to Organisation for Economic Co-operation and Development (OECD) countries (particularly Western Europe, North America and Australasia) where ALMPs are well established, and with a rich evaluation tradition on which to draw.

Active labour market policies

The literature in this area distinguishes between ‘active’ and ‘passive’ labour market policies. While ‘passive’ policies are concerned with providing replacement income during periods of unemployment or job search, ‘active’ policies emphasize labour market (re)-integration. Passive policies include unemployment insurance and early retirement measures; active measures include training, job creation measures, support for active job search, hiring subsidies and support for enterprise creation among the unemployed.

ALMPs emerged in the post-war period initially in Scandinavia, but were taken up in many advanced economies following the unemployment crises of the 1970s, and this take-up was boosted by the OECD's strong advocacy of ALMPs. The OECD Jobs Study of 1994 was particularly influential here (OECD, 1994). The OECD monitors the relative balance of labour market policy expenditure on active and passive measures in member countries, and has documented a shift from passive to active measures both across the OECD as a whole, and in many member states (see OECD (2005) and various issues of the OECD's Annual Employment Outlook).

The literature proposes various ways to categorize ALMPs (see Meager, 2007 for an overview). In simple terms, however, it distinguishes between ALMPs with a supply-side emphasis and those oriented towards the demand side of the labour market. Supply-side measures include:

  • 1Training schemes: these are the classic elements of Scandinavian ALMPs and may cover vocational and/or general skills. The underlying rationale is that the employability and job-finding chances of workless people are enhanced by training.
  • 2Information and job-brokering activities: these are the standard job-matching activities of the public employment service, involving vacancy registration, the provision of vacancy information to job seekers and provision of information on job seekers to employers.
  • 3Information, advice and guidance to job seekers on job search methods, often alongside motivational support.
  • 4Sanctions and incentives aiming to ‘activate’ workless job seekers. The sanctions involve compulsory participation in active measures with the threat of benefit withdrawal in cases of non-participation; while incentives may include a financial bonus when a workless person accepts a job offer.
  • 5Subsidies to individuals who enter self-employment and start their own enterprises (these also have a demand-side component).

On the demand side, measures include:

  • 6Subsidies to employers hiring jobseekers from particular target groups.
  • 7Job creation or ‘make work’ schemes, often in the public or not-for-profit sector, providing work opportunities for the unemployed.

In the UK, demand-side ALMPs have all but disappeared in recent years, and the emphasis is overwhelmingly on supply-side measures, as noted in Meager (2007), who also identifies several other recent shifts in the emphasis of UK active labour market policy, in particular:

  • an increasing focus on economically inactive3 target groups, e.g. lone parents, disabled people and older workers. Several factors explain this shift. First, as ‘mainstream’ unemployment has fallen, it has become clear that reaching the 80 per cent employment rate target requires economically inactive groups to be brought into the workforce. Second has been a concern to reduce benefits expenditure on the economically inactive, particularly incapacity benefits spending, which has grown threefold since the early 1980s. And third is an ideological preoccupation with a ‘work first’ approach, inherited from the USA (and increasingly found in European countries such as the Netherlands: Bruttel and Sol, 2006), according to which, work is the single most important solution to welfare dependency and social exclusion;
  • an emphasis on ‘making work pay’ through benefit reforms and the introduction of in-work benefits or tax credits, aimed at tackling the unemployment trap faced by many benefit recipients;
  • increasing levels of ‘activation’ through mandatory participation of workless groups in the various schemes. While most UK measures for the economically inactive retain a voluntary aspect (unlike measures targeted at the unemployed), the degree of compulsion has grown over time. For example, even where programme participation is voluntary, it is now compulsory for some groups to participate in ‘work-focused interviews’ to discuss the options of programme participation. Further extensions of compulsion to inactive groups have been flagged in the 2007 welfare reform white paper (DWP, 2007);
  • a shift towards individualization in the support packages offered. This trend, observable in many countries,4 is associated with a major shift in the culture of the public employment service and the benefit administration system (merged into a single agency Jobcentre Plus in the UK5), whose frontline staff are gradually having their role changed from one of policing the administration of benefits, to one of ‘personal adviser’; and
  • growing involvement of the private and voluntary sectors in ALMP delivery. While the public employment service retains a key role in service delivery, increasingly it does so in partnership or competition with private/voluntary agencies, operating under a performance-related funding regime. Further acceleration of this process is underway in the UK, following the recommendations of the Freud Report (Freud, 2007).

As Figure 1 shows, the UK spends little on labour market policy (active and passive); the Czech Republic, USA, Korea and Mexico are the only OECD countries spending a smaller share of gross domestic product (GDP) on labour market policies. It is not the purpose of this paper to review the extensive literature on the aggregate relationship between LMP expenditure and employment rates (see the discussion of this literature in Martin, 1998); suffice it to say that, as Figure 2 shows, the UK's low level of LMP expenditure does not simply reflect the currently high employment rate in the UK. In part, this is because the different components of LMP expenditure move differently over the economic cycle, with the share of passive benefits being higher when unemployment/inactivity is higher. In part, however, this lack of a relationship between LMP spending and employment performance reflects fundamentally different policy regimes. Thus, for example, the countries with high employment rates (over 70 per cent), fall into two groups: the Nordic countries, the Netherlands and Switzerland, with high ALMP spending; and the ‘Anglo-Saxon’ economies of the USA, Canada, Australia, New Zealand and the UK, whose low spending reflects less generous welfare states (as measured, for example, by average benefit replacement rates6).

Figure 1.

Total labour market policy expenditure as percentage of GDP (2005)
Source:author's calculations from OECD data

Figure 2.

Labour market policy expenditure and aggregate employment rates (2005)
Source:author's calculations from OECD data

When it comes to active labour market expenditure, however, the UK creeps up the league table a little (Figure 3), although it still spends a much smaller share of GDP on ALMP than most other Western European states.

Figure 3.

Total active labour market policy expenditure as percentage of GDP (2005)
Source:author's calculations from OECD data

Despite the low absolute level of spending on labour market measures in general and active measures in particular, it is striking that the share of UK labour market spending on active measures (Figure 4) is the highest among the OECD countries. This partly reflects the UK's lower than average passive spending, because of low benefit levels. It is clear, however, that although the UK spends little on labour market policy, this spending has a more ‘active’ orientation than most other countries, and in this respect, is closer to some of the Nordic welfare state regimes, than to its usual comparators in the other Anglo-Saxon nations.

Figure 4.

Share of active labour market policies in total labour market expenditure (2005)
Source:author's calculations from OECD data

How do training measures fit into ALMPs?

Turning to the role of training, the UK is again very different from the Nordic model: not only does it spend a small share of GDP on training measures (Figure 5), but such measures also account for a relatively small share of ALMP expenditure (Figure 6). It should be stressed that these data refer only to training measures for the unemployed and inactive, and do not include public expenditure on the training of the employed workforce.

Figure 5.

Expenditure on training-related ALMPs as percentage of GDP (2005)
Source:author's calculations from OECD data

Figure 6.

Share of training measures in total ALMP expenditure (2005)
Source:author's calculations from OECD data

The UK's approach to ALMP is dominated by: job-matching activities of the public employment service; job-search advice and support; and, to an increasing extent, sanctions for non-participation. Training measures are a relatively minor part, and from OECD data (Figure 7), it seems that expenditure on such measures has fallen in the UK in recent years (in 2005, this spending fell below 0.1 per cent of the GDP, compared with 0.15 per cent in 1999–2000). Similarly, Figure 8 indicates that expenditure on training fell from 45 per cent of ALMP spending in 1999 to 28 per cent by 2005.

Figure 7.

Expenditure on training-related ALMPs as percentage of GDP (UK: 1999–2005)
Source:author's calculations from OECD data

Figure 8.

Share of training measures in total ALMP expenditure (UK: 1999–2005)
Source:author's calculations from OECD data (no data available for 2000)

This declining emphasis on training within active measures is not confined to the UK; De Koning (2007) presents ALMP expenditure data for 18 OECD countries over 1991–2003, showing an increase in the share of mediation, counselling and similar measures, and an increased use of subsidies, alongside a declining spend on training measures (from 31 to 24 per cent). De Koning argues that this shift is consistent with the evaluation evidence on the relative effectiveness of different types of measure (explored in subsequent sections of this paper).

This finding that the UK share of ALMP spending on training is small and falling is of particular interest, at a time when the UK is aiming for a greater integration between its training and employment policies.

How effective are training/skills-based active labour market policies?

Whether this is a desirable and appropriate direction of policy depends on whether, in comparison with alternative active measures, training interventions are (cost-)effective tools for integrating the unemployed and inactive into the labour market. The notion that skills-based measures might have an important role to play is intuitively plausible. At an aggregate level, there is a strong cross-country relationship between levels of initial education and continuing vocational training on the one hand and employment performance on the other (OECD, 2004b). Similarly, at the individual level, there is a strong relationship between training experience and the probability of being in work. In both cases, there are difficulties in establishing causality through econometric methods, but much evidence is at least consistent with a causal relationship (Brunello, 2007). Less clear, however, is whether there is an impact of training on unemployment. As OECD (2004b), who failed to identify such a relationship, point out, this may reflect ‘crowding-out’ effects, i.e. those who are trained may simply displace those with lower skill levels. OECD concludes, however, on balance, that:

appropriate [training] policies can improve the labour market position of specific targeted groups. Such policies can be an important component of a general strategy geared at reducing non-employment traps (OECD, 2004a, p. 185).

It does not, however, follow that training measures are the most effective intervention for disadvantaged, workless groups, particularly in the UK context, with evidence of a low-skills equilibrium in at least part of the labour market (Wilson & Hogarth, 2003), and the existence of a significant proportion of low/unskilled jobs. In such a context, cheaper, shorter-term interventions focused on job entry (job search support with related sanctions and incentives) may give a greater pay-off. The balance of measures in the ALMP portfolio described previously suggests that the UK government has taken such a view, although some recent evaluations of UK measures with a skills component have offered positive results (e.g. Speckesser & Bewley, 2006).

Looking more broadly at this question, a wealth of international research (of varied methodological sophistication) throws more light on the relative effectiveness of training interventions. Comparative reviews of this evidence include: Björklund (1991, 1994), Calmfors et al. (2002), De Koning (2005, 2007), Estevão (2003), Fay (1996), Forslund and Krueger (1997), Kluve (2006), Martin (1998), Meager and Evans (1998), OECD (1992, 1993), Robinson (1995a) and Speckesser (2004).

Early evaluations

It is difficult to draw universal conclusions from this literature. However, the growing volume of research within and between countries has moved the situation on from that reported by Björklund (1991) who, surveying the early Swedish evaluations and the disappointing results of studies of general training schemes for the unemployed, noted:

the results obtained are too uncertain to allow firm policy conclusions (Björklund, 1991, p. 90).

Clearer policy conclusions began to emerge by the mid-1990s: large scale, broad-brush training measures for the unemployed/inactive were expensive, with little or no effects on employment chances or subsequent earnings. Indeed, some studies showed negative impacts of such schemes, which may even have made things worse by ‘locking participants out’ of the labour market while they did the training. As the current author noted (Meager & Evans, 1998):

It is rapidly becoming conventional wisdom in the policy evaluation literature that labour market training and re-training schemes for the unemployed have not lived up to expectations. As many evaluation studies [ . . . ] have shown, such schemes (including those in countries such as Germany and Denmark, where the quantity and quality of workforce training in general is regarded as high), often appear to make little difference to the employment (or earnings) chances of participants (Meager & Evans, 1998, p. 49).

Indeed, these negative findings were used in the UK to provide support for a shift in the balance of the UK policy portfolio. As Robinson (1995a) argued:

The recent movement of resources in Britain away from training and work programmes towards the initiatives run by the Employment Service is overwhelmingly backed by the results of research from across the OECD. Simple initiatives which offer improvement placement services or assistance with job search can be shown to significantly boost participants' job prospects. The evidence for the effects of training and work programmes is much more mixed (Robinson, 1995b).

Some authors (e.g. Lange & Shackleton, 1998) went further, arguing against more or less any types of ALMP interventions, other than job search support coupled with benefit incentives/sanctions.

Evaluations conducted in the 1990s

During the 1990s, labour economists struggled with the mismatch between the strong theoretical justification for enhancing job seekers' employability through increasing their human capital, and the poor performance of active measures designed to do just that. Some authors argued that the resolution of this mismatch lay with the scale and focus of the training/skills measures: where the schemes were smaller in scale, and targeted on groups with specific needs, and where they were (1) coupled with interventions to address other barriers which individuals faced (including social barriers); and/or (2) took place in a ‘market’ environment (e.g. training in a real workplace, in combination with a job placement or work trial, rather than a classroom), then better results were possible.

As Jackman (1995) observed:

Given the increasing disparities in wages and employment opportunities between skilled and unskilled workers, one might expect training schemes to offer a high return to those taking them. A remarkable, but consistent finding is that the microeconomic evidence on general training schemes for unemployed adults provides very little support for such a view.

OECD (1993) examines the effects of numerous types of training schemes. A general conclusion is that programmes targeted on a small number of individuals, with a relatively high cost per head, often appear quite effective in improving the wage and employment prospects of at least some of the individuals involved. On the other hand, broader programmes covering a larger number of people at relatively low cost per head typically seem to have little if any effect on the prospects of participants.

A good example is the effects of programmes carried out under the Job Training Partnership Act (JTPA) in the US. Studies of the effects of JTPA show that the training component had no significant effect on the earnings or employment prospects of unemployed people generally, but special schemes, e.g. for those with poor education, did tend to improve their labour market prospects [ . . . ]

The overall conclusion is that training schemes need to be targeted to meet the needs of specific groups and supported by adequate resources. A corollary is that training of this quality cannot be made generally available to all unemployed people. Thus effective training schemes cannot be a large part of a labour market policy which seeks to place every unemployed person on a scheme or in temporary work (Jackman, 1995, pp. 15–16).

Consistent with these conclusions, a review of the US and European evaluation evidence on ALMPs targeted at disadvantaged young people (Auspos et al., 1999) found generally negative results, but the exceptions were of this small scale, targeted nature. Thus, among six US training programmes for disadvantaged youth (some combined with other interventions such as benefit sanctions or job placements), there were two exceptions to the general finding that the schemes were ineffective. One was the Job Corps programme, which ‘produced large and consistent benefits across a range of employment and social outcome measures’. It was, however, a high-cost intervention, being a residential, long-term intensive training programme targeted on extremely disadvantaged groups (the non-residential variant was ineffective). The other was delivered by the US Center for Employment and Training, which achieved high earnings gains for participants through intensive instruction over a substantial period (typically 6 months), with open entry and exit and a strong emphasis on early employment outcomes.

The European evidence considered by Auspos et al. was, however, less consistent and more positive than the American evidence. The authors attributed this to more rigorous evaluation (particularly random assignment methods) used in the US research, as well as differences between labour market institutions between the US and Europe. Auspos et al. note that the evidence on large-scale youth training programmes in Sweden indicated that such measures generally had no effect, or even a negative effect on employment and earnings (Björklund, 1991, 1994; Forslund & Krueger, 1997). Results were mixed in Norway (Raaum & Torp, 1996; Torp, 1994, 1995), but with some small evidence of a positive earnings effect. In Denmark, for the most disadvantaged long-term unemployed groups, there were small but statistically significant gains in subsequent employment entry (Jensen et al., 1993), and Auspos et al. concluded that ‘In view of its short duration, the training was probably cost-effective when focused on those with the greatest employment problems.

Some youth programmes in France (Bonnal et al., 1994), Ireland (Breen, 1991), the UK (Green et al., 1996), Austria (Winter-Ebmer & Zweimüller, 1996) and Belgium (Cockx et al., 1996) had positive effects, but it was difficult to distinguish the impacts of training interventions from other measures delivered as part of the same programme (temporary work placements, job search support, etc.).

Other studies from this period in different countries also support the general conclusion that the scale and degree of targeting of the training matters, as does the extent to which it is embedded in a broader programme including elements such as work experience: see, for example, the evidence from Austria (Biffl et al., 1996), Ireland (O'Connell & McGinnity, 1997) and the Netherlands (De Koning et al., 1991) cited in Meager & Evans (1998). Particularly interesting amongst these is the Irish study, which used a survey-based follow-up of participants across the full range of active measures in place, and a matched group of non-participants to compare the impacts of the different measures on employment probabilities, job duration and earnings. The research compared general training schemes, a specific skills training programme, employment subsidies and direct job-creation programmes. The analysis generated a useful typology of measures, building on the traditional demand/supply-side distinction, to incorporate the degree of ‘market orientation’ of different measures. Thus, for example, on the supply side, training programmes involving private sector placements with on-the-job training are seen as having ‘strong’ market orientation, while classroom-based schemes have ‘weak’ market orientation. Similarly, on the demand side, traditional direct job-creation measures are weakly market oriented, while indirect measures subsidizing jobs in the private sector are strongly market oriented. Their dominant conclusion was that what made the difference was the degree of ‘market orientation’ of the programme. In particular, programmes with a stronger ‘market orientation’ led to higher placement rates, longer job durations and higher earnings, than schemes with weak market linkages. Thus, the specific skills training programme had far more positive impacts than the general training programmes or the job-creation schemes. Their results also highlighted a need for targeting on the most disadvantaged groups (to avoid creaming and high deadweight7). While the results do not wholly militate against schemes with weak market linkages, they suggest that they should be appropriately targeted (on the least ‘job ready’ groups), and lead to progression to schemes with a clearer market orientation. The authors raised the question of whether the positive impacts of skills training may have a peculiarly Irish aspect given that, at the time of the evaluation (early 1990s), Ireland had a low incidence of in-company training (O'Connell & Lyons, 1995). It was, therefore, possible that specific skills training of the unemployed through ALMP acted as a functional equivalent of the training of employees or new recruits, which elsewhere (in Scandinavia, Germany, the Netherlands or the UK), might be more likely to occur within companies.

In the UK, the older evidence (reviewed in Auspos et al., 1999) also suggests that the smaller schemes, targeted on disadvantaged groups, and/or particular occupations, had the most impact, especially where the training was customized to employers' needs and delivered in conjunction with practical work experience. It is notable, for example, that the impact of the Employment Training programme for the long-term unemployed was greatest where the training was linked to a private sector work placement (rather than work experience on a job-creation project) and where it led to a formal vocational qualification (Payne et al., 1996). It is not always possible in such cases to distinguish between the relative impacts of the work experience component and the training component. It is possible, for example, that participants who got jobs did so, not because of the training per se, but because their placement providers kept them on, or because the work experience made their curricula vitae more attractive to future employers.

On the evidence available up to the end of the 1990s, the consensus was that job-matching services, information advice and guidance measures, along with some kinds of targeted subsidy schemes, had the most positive impact on outcomes such as employment rates and subsequently were most cost-effective, while both direct job-creation schemes and training/skills programmes performed poorly, except where they were small-scale, targeted and linked to practical job experience. This consensus seems to have contributed in many countries to a shift, within ALMPs, away from training measures towards job broking, advice and guidance, and incentives/sanctions. At the same time, an increasing scepticism can be found in the policy-oriented literature regarding the value of ALMPs of any type. The conclusions of Calmfors et al. (2002), who examined the Scandinavian evidence, are typical of this consensus and worth listing here (author's emphasis):

The Swedish experiences of the 1990s provide a unique example of how large-scale active labour market programmes (ALMPs) have been used as a means to fight high unemployment. [ . . . ]. The main conclusions are: (1) there is hardly any evidence for a positive effect on matching efficiency; (2) there are some indications of positive effects on labour force participation; (3) subsidised employment seems to cause displacement of regular employment, whereas this appears not to be the case for labour market training; (4) it is unclear whether or not ALMPs raise aggregate wage pressure in the economy; (5) in the 1990s, training programmes seem not to have enhanced the employment probabilities of participants, whereas some forms of subsidised employment seem to have had such effects; and (6) youth programmes seem to have caused substantial displacement effects at the same time as the gains for participants appear uncertain.

On the whole, ALMPs have probably reduced open unemployment, but also reduced regular employment. The overall policy conclusion is that ALMPs of the scale used in Sweden in the 1990s are not an efficient means of employment policy. To be effective, ALMPs should be used on a smaller scale. There should be a greater emphasis on holding down long-term unemployment in general and a smaller emphasis on youth programmes. ALMPs should not be used as a means to renew unemployment benefit eligibility (Calmfors et al., 2002, pp. 1–2).

The current picture from the evaluation evidence

Moving on to more recent literature, there has been a gradual build up of an empirical base of methodologically rigorous evaluations. From this literature, while the broad conclusions established in the 1990s have not been fundamentally altered, and the results vary by national context and the precise design of the measures, it is possible to detect some more positive conclusions emerging with regard to some types of training initiatives.

In particular, Kluve (2006) undertook a comparison of 73 microeconomic evaluation studies of European ALMPs (published since 2002), with an emphasis on more rigorous approaches, which, although generally falling short of the random assignment ‘gold-standard’, do at least use matching estimators and control for sample selection effects. He concluded:

Training programs are the most widely used active labour market measure in Europe. The assessment of their effectiveness shows rather mixed results; treatment effect estimates are negative in a few cases, and often insignificant or modestly positive. Still, there are several indications that training programs do increase participants' post-treatment employment probability, in particular for participants with better labour market prospects and for women. However, this pattern does not hold for all studies. Locking-in effects of training are frequently reported, though it remains unclear to what extent these are really entirely undesirable, and not rather a necessary element of this type of program' (Kluve, 2006, p. 10).

Kluve went on to undertake an innovative meta-analysis of data from some 95 evaluation studies (yielding 137 programme observations), enabling him to assess the differential effects on programme effectiveness of the type of programme (training, subsidies, etc.), the institutional and national context, the macroeconomic environment, and the decade in which the programme was implemented. The results suggest that programme type is by far the most important variable influencing post-programme employment probabilities. In particular, the meta-evaluation suggested that training measures sit in the middle of the hierarchy: neither the most effective measures, nor the least. Subsidies, however, along with job search support and incentives, remain at the top of the effectiveness league. As Kluve puts it:

Traditional training programs are found to have a modest likelihood of recording a positive impact on post-program employment rates. Relative to these programs, private sector incentive programs8 and Services and Sanctions9 show a significantly better performance. Indeed, we find that evaluations of these types of programs are 40–50 per cent more likely to report a positive impact than traditional training programs. By comparison, evaluations of ALMPs that are based on direct employment in the public sector are 30–40 per cent less likely to show a positive impact on post-program employment outcomes. Also the target group seems to matter, as programs aimed specifically at young workers fare significantly worse than programs targeted at adults, displaying a 40–60 percentage points lower probability of reporting a positive effect.

The general policy implications that follow from these findings are rather straightforward. Decision makers should clearly focus on the type of program in developing their ALMP portfolio: Training programs should be continued, and private sector incentive schemes should be fostered. Particular intention should be paid to Services and Sanctions, which turns out to be a particularly promising and, due to its rather inexpensive nature, cost-effective type of measure. A well-balanced design of basic services such as job search assistance and counseling and monitoring, along with appropriate sanctions for non-compliance, seems to be able to go a long way in enhancing job search effectiveness. If further combined with other active measures such as training and employment subsidies, this effectiveness could be increased . . . (Kluve, 2006, p. 27).

De Koning (2005) presents an even larger review of 130 evaluation studies, amounting to 161 measures being evaluated. Although, unlike Kluve, he does not undertake a meta-analysis; his findings broadly mirror Kluve's, concluding that the evidence on incentives for job seekers (job search monitoring, sanctions, bonuses, etc.) ‘overwhelmingly points to a positive effect on job entry chances’ as does, to a lesser extent, the evidence on counselling, placement and wage subsidies for regular jobs. However, the evidence on training is much more mixed, and ‘the number of training studies that point to positive effects is more or less the same as the number of studies showing insignificant or significantly negative effects’. Job-creation measures fare least well, with most studies showing insignificant or negative effects on job chances.

Once again, it is difficult to avoid the conclusion that, while there may be a role for training interventions in a balanced ALMP portfolio, they are not a panacea, nor universally relevant, and much depends on the type of training and the target group in question. Most evaluations of training measures relate to schemes targeted at young people, and De Koning's results indicate that training interventions are largely ineffective for workless young people. The results for adults are more ambiguous and differ between studies. Although, as noted previously, there exist examples of positive evaluations of training programmes for workless adults, when the different target groups are considered according to the personal characteristics of the participants, or the nature of the training, there are too few examples of each type to draw any general conclusions. De Koning et al. (2005) do, however, cite earlier Dutch evidence suggesting that training interventions may be particularly effective for older unemployed adults (Heyma et al., 2003 is a good example).

Longer-term impacts

It is possible that, to a greater extent than other ALMP interventions, training measures are ‘slow burn’ interventions, and that it takes time for the impacts of human capital acquisition on employment rates or earnings to emerge. In this light, the relatively poor performance of training measures could simply reflect the fact that most evaluations have not collected outcome data over a sufficiently long post-programme period. It is notable that the few studies that consider impacts over a longer period do indeed pick up more positive effects. Thus, Harkman et al. (1996), in their controlled evaluation of Swedish labour market training programmes, found no impact 6 months after scheme participation, but a positive impact on employment rates and wages after 2.5 years (especially among those with low initial educational backgrounds). Similarly, the UK evaluation of the Employment Training scheme (Payne et al., 1996) reported increasing impacts over time from participation. Likewise, Hotz et al. (2000) reported higher longer-term returns (over 9 years) from a US intervention, and argued that some earlier conclusions about the skills component of that intervention (the GAIN programmes in California) relative to the ‘work-first’ variants (emphasizing early work placements) should be reassessed:

We [ . . . ] find that the stronger impacts of Riverside County's work first program tend to shrink, whereas the weaker impacts for the human capital programs in Alameda and Los Angeles Counties tend to remain constant or even grow over time. [ . . . ] On a substantive level, our re-examination of the GAIN experiment leads us to conclude that although the work first programs were more successful than the human capital accumulation programs in the early years, this relative advantage disappears in later years. (Hotz et al., 2000).

Kluve (2006) also notes, from his review of recent evaluations, that

The more recent literature on the evaluation of training emphasizes the need to consider long-run impacts. Such an assessment has become increasingly possible due to extended data. There are indeed indications from these studies that positive treatment effects of training exist in the long run (Kluve, 2006, p. 10).

Kluve and others have also noted that such positive longer-term effects might be sufficient to outweigh the impact of negative ‘locking-in’ effects observed in the short term among some training interventions. Despite Kluve's optimism, however, the number of studies tracking participants for longer than a year after participation remains small, and it is too soon to draw a strong generalizable conclusion from these studies, particularly because, as De Koning (2007) notes, some studies also exist that show increasingly negative results with time:

. . . the relatively few studies that measure long-term effects of training do not point at more favourable long-term effects. In fact, most studies dealing with long-term effects find that the effects get smaller over time [ . . . ], while only one [ . . . ] finds that the long-term effects are larger than the short-term effects (De Koning, 2007, pp. 39–40).

Studies yielding negative longer-term returns to training-based ALMPs include: an evaluation of a Dutch training programme in De Koning et al. (1988) and Hujer and Caliendo (2001), who review a range of German microeconometric evaluations of training programmes (most of which do not find significant or better impacts in the longer term than in short-term estimations); Fitzenberger and Prey (2000), whose evaluations of East German vocational training programmes show considerable ‘locking in’ effects, which are not offset in the longer-term; and Lechner (2000), who finds, again for Germany, negative effects on the risk of unemployment in the short term, and no positive long-term effects on employment probabilities or earnings.

Bergemann et al. (2004) undertook a more sophisticated analysis allowing for repeated participation in programmes, and over a longer period (again for East German schemes during the 1990s), with only slightly more positive results:

Overall, our results are not as negative as previous results in the literature and it is unlikely that training on average reduces considerably the future employment chances of participants. We also find noticeable differences among different treatment types. At the same time, it remains questionable whether on average training programs are justified in light of the large costs incurred (Bergemann et al., 2004).

It is, of course, possible that the negative results from such studies, conducted at times of relatively high unemployment, and in countries (the Netherlands, Germany) with extensive systems of initial and continuing vocational training, may be specific to those countries. Arguably, if the main problem is slack labour demand, and if the overall level of skills training in companies is already relatively high, then it is not surprising that training interventions do not increase employment rates. In other economies with higher levels of skill deficiency, lower levels of company training and/or tighter labour markets, such programmes might have more impact.

Overall, as with the short-term findings, the results from studies that look at impacts over a longer time frame are extremely mixed. On this evidence, it would be hard to argue that training measures in ALMPs can be justified by their longer-term employability impacts. In this context, it should also be borne in mind that, given the typically much higher cost of training interventions than some of the other ALMP measures, the fact that a (small) impact may emerge after a number of years may still render such measures unattractive to policy makers.


In light of the growing emphasis on greater integration between skills policy and employment policy in the UK, this review has examined international evidence on the effectiveness of training and skills interventions as part of a broader ALMP portfolio. Compared with its main competitors, the UK invests little in ALMP in general, and in training measures in particular. As in other countries, moreover, the share of ALMP resources devoted to training measures has declined in recent years, and the share devoted to ‘work-first’ interventions (job broking, job search support, advice and guidance, benefit incentives/sanctions) has increased. This shift was partly influenced by evaluation evidence from the 1990s, suggesting that large-scale training measures for the unemployed or inactive were ineffective (in increasing employment rates or earnings), and might even keep people out of the labour market for longer than would otherwise be the case.

More recent evidence, with more rigorous methodologies, has yielded more finely nuanced findings. While some studies confirmed that traditional training measures were indeed generally ineffective, they also indicated that smaller-scale measures, targeted on particular groups (such as older workers) or on particular skills and occupations, might nevertheless yield positive results, particularly where such measures are delivered alongside ‘real’ work experience. Also, while it is not generally the case that greater impacts emerge over a longer period after the training intervention, again, there are examples where this has happened in particular programmes.

Overall it would be hard, given the extremely mixed evidence, to justify a major shift towards skills measures within an overall welfare-to-work strategy, particularly given the more unambiguous, cost-effective impacts associated with measures such as job search support, advice and guidance, and benefit incentives. There may, however, be limited value to be found in smaller, targeted training interventions for groups and circumstances where it is clear that (easily remedied) skill deficiencies are the main barrier to labour market (re-)entry.

Martin and Grubb (2001), in a review of ALMP evidence, set out six principles to guide policy makers in their selection of active measures; their key message is to promote in-depth counselling, job-finding incentives and bonuses, and job search assistance (coupled with a tight monitoring and sanctions regime). Their second principle, however, relates to the use of training interventions and remains valid today in the light of the evidence considered in the present review:

Second, keep public training programmes small in scale and well targeted to the specific needs of both job seekers and local employers. Build in as much on-the-job content to training programmes as possible (Martin & Grubb, 2001, p. 33).

Before concluding, however, it should be noted that this review has covered micro-level evaluations of individual ALMP interventions. The picture becomes more complex when aggregate cross-country evaluations are considered, looking at the macro-level impacts of different ALMP regimes. Usually, such evaluations take measures of overall ALMP participation or expenditure as a key independent variable; rarely do they separate out the effects of different types of ALMP intervention. One recent study that does this, however (Boone & van Ours, 2004), shows that job training is the most effective intervention in reducing the aggregate unemployment rate and increasing the employment rate. This result contrasts strongly with the findings emerging from the micro-level studies considered in this paper. This apparent paradox remains to be conclusively explained, but Boone and van Ours (2004) hypothesize that increased training levels improve matching between workers and jobs, thereby reducing labour turnover, which, by reducing inflows into unemployment, generates the positive impact. In contrast, micro-studies pickup only any impact on outflows from unemployment.

What does this review suggest for future research in this area? It has highlighted some remaining gaps in the evidence base: there is a need for more longitudinal evaluations of training interventions for the unemployed/inactive, allowing sufficient time for longer-term impacts on employability to emerge. In addition, there is a case for more aggregate level, cross-country comparisons taking account of different components of expenditure and scheme participation to build on the early findings of work such as that of Boone and van Ours (2004).

As far as policy development is concerned, the case against mass, ‘broad brush’ training programmes within the overall ALMP portfolio seems convincing. If greater integration between employment and skills policies is to be achieved in the UK and elsewhere, the evidence suggests that flexible, small-scale training interventions, targeted on specific groups for whom skill deficiency has been identified as a real barrier to work entry, are most likely to be effective, particularly where such interventions are coupled with real work experience or job placements.


  • 1

    An earlier version of this review was prepared for the Sector Skills Development Agency (SSDA), while the author was a member of the SSDA expert panel. The views in the paper are those of the author, and should not be attributed to the SSDA or its successor body (the UK Commission for Employment and Skills).

  • 2

    See, for example, National Economic Development Office (1984), and see also Keep (2008) for a trenchant account of the repetitive cycles of research and policy-making on this topic.

  • 3

    The labour market policy literature divides workless people of working age into two broad categories: the unemployed (who are actively seeking work, and available for work); and the economically inactive (who are either not available for work or do not want work and are not actively seeking it; because they are in the education system, because family or health circumstances prevent them working, or because they have ‘retired’ early from the labour market).

  • 4

    See Carstensen and Pedersen (2008) for a discussion of the trend towards individualization in UK, Dutch and Danish ALMP.

  • 5

    Similar mergers have occurred in several European countries (including, for example, the Netherlands, France and Norway).

  • 6

    The ratio of out of work benefits levels to previous or average earnings in work (OECD, 2004).

  • 7

    In the policy evaluation literature, ‘deadweight’ occurs when the measured outcome would have occurred in the absence of the policy, and cannot therefore be attributed to the policy as an ‘impact’.

  • 8

    Kluve is here referring mainly to subsidies to employers or to workless individuals.

  • 9

    In ‘Services and Sanctions’, Kluve includes job search advice and support, counselling and monitoring, as well as financial or other sanctions for non-participation in active job search measures.