The use of apprenticeships by employers dates back hundreds of years. More recently, the perception that apprenticeships are also effective in improving labour market outcomes have made apprenticeships increasingly popular with policymakers and employers. In the UK, for example, government has run apprenticeship programmes since 1994. The UK has historically followed a market-based approach, in which firms take the lead; apprenticeships are a form of temporary employment contract, lasting between one and four years, and there is no formal link to the education system.
Following the 2012 Richard Review, the previous system, which involved over 250 different apprenticeship frameworks, is now being simplified and formalized, with some graduate-level apprenticeships introduced that are designed to offer a formal alternative to a full-time degree. The number of apprenticeships is also increasing. For example, since 2010, the number in England has doubled, with accompanying increases in funding; however, there are still fewer apprentices as a share of the workforce in England (11 per 1,000 employees) than some other countries (39/1,000 in Australia, 40/1,000 in Germany and 43/1,000 in Switzerland).
The Government is committed to further increasing the number of apprenticeships. The 2015 Summer Budget announced the creation of 3 million new apprenticeships by 2020, funded by a levy on large employers. The exact details of this levy, including rates and implementation, have yet to be established but firms will have access to the funds raised to cover the costs of post-16 Apprenticeships in England. Money to pay for the training will continue to go straight to providers rather than employers, with employers accessing the funding via an apprenticeship voucher system. The overall aim is to increase employer investment in training.
In addition to these national level changes, devolution will also have implications for policy, with Greater Manchester now having control over the employer apprenticeship grants budget, and other cities potentially following suit in the future.
Our review focuses on the economic impacts of apprenticeship programmes on participating individuals and firms (and by implication, the local areas in which they are based).
Large claims are sometimes made for these economic impacts. For example, a recent BIS report suggests that the net present value of Level 2 and Level 3 apprenticeships begun in 2013/14 are £12bn and £10bn respectively. These are substantial numbers. In addition to helping with programme design, the impact evaluation evidence reviewed in this report is one element in determining whether or not such predictions are plausible.
What can we expect apprenticeship programmes to achieve? There is a large existing literature on the economic returns to training and education, including some studies that look directly at apprenticeships. These evaluations typically find large, significant positive wage and employment gains to individuals who have participated in apprenticeships, versus individuals who did not. Not all of these studies meet our quality thresholds, but we include any suitably robust evaluations in our shortlist.
What might drive these effects? Apprenticeships aim to improve individuals’ employment and wage outcomes by raising their human capital. Apprentices should be more employable after the programme, and have mastered specific skilled roles. This should raise lifetime wages, reduce the risk of future unemployment and improve prospects of career progression.
Apprenticeships may also help employers, by providing them with more skilled (and thus more productive) employees. Apprentices may also be more loyal and contribute more effort than regular entry-level workers. Of course, after completing the programme apprentices may be poached by other employers, who then reap the benefits. For this reason firms will tend to underprovide apprenticeships, which is why governments provide support to firms to cover some of the cost of the programme (through grants, wage subsidies or some combination of these). These policy features have the additional benefit (to employers) of making apprentices cheaper than other, equivalently qualified, members of their workforce.
As explained further below, figuring out the ‘causal’ impact of apprenticeships on firms and workers is not straightforward. Issues include agencies or firms ‘cherry picking’ the most able apprentices; the ‘best’ or ‘worst’ firms choosing to participate in a programme; identifying the reasons for post-apprenticeship moves in the labour market, since the most able apprentices may be the most likely to move; and tracking participants (and control group members) over time, particularly for studies that seek to pick out the long term effects of apprenticeship programmes.
Figuring out the impacts of apprenticeships on a local economy is even harder. Even if we know the individual benefits, worker (and firm) location decisions affect the geographical distribution of those benefits. For example, to the extent that apprentices in a given area stay in that area after completing the programme, workforce quality in the area rises; however, if participants move to another area that human capital is lost to the new location. Of course, human capital across the UK rises either way.
These complexities also make it more difficult to estimate the net benefits of apprenticeship schemes, even if cost data is available (which is only the case in one of our shortlisted studies).
Apprenticeship systems across countries
A final complication arises from the fact that not all apprenticeship systems are the same. A lot of discussion in the UK focuses on the German system, in which apprenticeships are an integral part of the national education system, with centralized provision and chambers of commerce closely involved in regulating content and quality. Over 40% of general secondary school leavers in Germany complete an apprenticeship, which can last up to three years. Other European countries such as Switzerland, Austria, Norway and Denmark also follow this approach. Countries such as the US take a hybrid approach, with some centralized programmes, but these are very small and focused on sectors such as construction or manufacturing. Other programmes are devolved and designed by individual states, or through public-private partnerships.
As a result of this complexity, it is unsurprising that the evaluations that we end up considering cover countries that approach apprenticeships in a slightly different way. As noted, this makes them both harder to define and more difficult to compare. That said, the programmes we review can be grouped into three overarching categories: the dual educational system, employment contracts and a combination approach.
Dual Educational System
This approach is particularly strong in Germany, and has been followed by some of its European neighbours, including Switzerland, Austria, Norway and Denmark. The key feature of this model is centrally co-ordinated, government-led provision, with apprenticeships formalised as part of the national education system (Cooke, 2003 – Paper 847). Provision is typically facilitated through a ‘dual educational system’ which offers different schooling routes for young people: the academic route, typically leading to university and higher education (Matura in Austria; Gymnasium in Germany) and the vocational track, which comprises compulsory vocational training and leads to apprenticeship (Hauptschule in Germany and Austria), although this format varies from country to country (Cooke, 2003 - Papers 847; Fersterer et al., 2004 - Paper 878). Apprentices divide their time between on-the-job training (65-70%) and educational training (1 to 2 days a week), provided typically by the state (Parey, 2009 - Paper 986). In this model, the government is normally responsible for covering the cost of all apprentices’ classroom training. A key element of this apprenticeship model is the formalised, institutionalised nature of apprenticeships, and the regard with which apprenticeships are held nationally; as a mark of the popularity of the apprenticeship route in these countries, over 40% of general secondary school leavers in Austria complete an apprenticeship (Fersterer et al., 2004 - Paper 878).
The approach in other European countries, such as the UK, is very different from that outlined above. Rather than a government-led system, the onus is on the private sector to provide apprenticeships. In these systems, the apprenticeship is seen more as a form of temporary employment contract, which allows the firms to pay less social contributions, and in return asks that they provide apprentices with certified on-the-job and classroom training (Capellari et al., 2012 – Paper 897).. This is traditionally not linked to the education system; a student may choose to leave school and commence an apprenticeship, but this is not a formal track set by government (ibid). There is no centrally controlled apprenticeship programme and as such there time split between work and training varies from apprenticeship to apprenticeship. Training is often provided in firm, but external courses can also be used – this could link to courses provided by government education bodies (ibid). Funding for training is typically split between government and businesses in this model; in the UK, Government only contributes 50% of the cost of training for apprentices aged 19 – 24 (Skills Funding Agency, 2015).
Countries such as the US – which has historically had a very small apprenticeship system with little coverage outside construction and manufacturing – operate a hybrid approach.
The US has a growing apprenticeship system, with President Obama announcing a new $100 million grant program to support the development of innovative apprenticeship programs across the country in 2015 (White House Office of the Press Secretary, 2015). Centralised apprenticeship programmes exist, which are government-led, such as the Registered Apprenticeship which is administered by the Office of Apprenticeship and State Apprenticeship Agencies at a national level. In addition, the US have implemented School to Work reforms which focus on high school students, with some elements of the academic / vocational tracking seen in Germany (Neumark and Rothstein, 2005 – Paper 964). In addition, however, there is also an emphasis on facilitating more private sector provision of apprenticeships. The US department of labour runs an ‘American Apprenticeship Grant’ which invites public-private partnerships to implement apprenticeship systems (US Department of Labour, 2015). Programmes for apprentices are therefore organised on a state by state basis by private, public and third sector bodies, as well as at a national government level. (Veum, 1995 – Paper 976).
Cross-country comparisons of the evaluation evidence
Clearly, this complexity creates challenges when comparing the results of evaluations of schemes in different countries. As with our other evidence reviews, we still think that it is possible to further our understanding of apprenticeships by looking at, and comparing, the available impact evaluation evidence from across the OECD. But this discussion highlights some of the caveats that must apply to the findings emerging from our review. Where necessary, we highlight specific issues further in the full review.
Definition of Apprenticeships
Apprenticeships are not easy to define, and delivery models vary across countries, making definition harder. However, we can identify certain common features. For the purposes of this report, apprenticeships are defined as:
Paid employment within a firm, alongside theoretical training that is usually provided by government, the employer, or a trade union, targeted specifically at school leavers (level of education varies by type of apprenticeship scheme). The apprentice often acquires a formal qualification by the end of the apprenticeship.
Apprenticeships are not simply a) programmes which take place either entirely in the classroom (i.e. vocational education); or b) programmes that take place entirely in the firm (as discussed in the Employment Training Review). Rather, it is the combination of these two elements that matters. Although apprenticeship schemes do not necessarily need government support, such support is very common. In line with this, for all those programmes for which the evaluation provides details, there is always an element of public sector support.
Impact evaluation for apprenticeships
Evaluating the causal impacts of apprenticeship programmes is not straightforward. Ideally we would want to randomize participants into a programme, and then compare changes in their post-programme labour market outcomes with changes for similar people who did not participate. In principle randomisation is feasible for programmes like apprenticeships - however, we only found one example in this review (we found rather more in our review of other employment training programmes).
Apprenticeships are, in many ways, a partnership between apprentice and employer, and unlike many employment training programmes, employers choose to get involved. As noted in the introduction, survey evidence suggests that participating firms are very positive about apprenticeships. This may reflect real benefits of such programmes, but also suggests that participating employers may not be representative of all employers (i.e, they ‘select into’ providing apprenticeships). In principle randomising the firms who participate in apprenticeships is possible, but we found no real-world examples of programmes where this happens.
In the absence of randomization (of participants and/or firms), we worry about the ‘selection into treatment’ problem. As with active labour market programmes as a whole, selection into treatment occurs when individuals – or firms – participating in the programmed differ from those who do not participate in the programme, in ways that can be hard for researchers to observe.
For example, employers or programme delivery agencies may ‘cherry pick’ participants with the most skills or motivation to succeed in an apprenticeship. Agencies may also select the firms who provide the best opportunities. This means that the average effects of the apprenticeship are biased upwards, since the programme would not deliver the same benefits to other participants or in other businesses. Conversely, we may see downward bias, if programmes are targeted at the ‘hardest to help’ – such as people who are very long term unemployed or who have chaotic lifestyles, which make it harder for them to complete the apprenticeship.
A second set of challenges arises post-programme. A number of the studies here set out to explore the impact of ‘post-apprenticeship events’, such as moving jobs or career paths. Just as participants may select into an apprenticeship, however, their decision to stay or move jobs afterwards may be related to unobservable characteristics of those individuals. For instance, firms may actively seek to ‘lose’ the ‘worst’ apprentices.
A third set of challenges is more prosaic. Apprenticeships can have long timescales – sometimes running for three years – which means that some participants may drop out during the programme. Keeping track of participants post-programme raises similar issues. It is not clear how long any effects of apprenticeships might last – in our shortlist we have studies tracking immediate impacts (such as the first job gained), as well as lifecycle effects (wage gains over the following 40 years). Understanding the longer-term impacts of apprenticeships is crucial for policymakers – but is challenging to do in practice. Non-participants – in control groups – may be even harder to track, especially over long time periods.
The most robust studies in our shortlist adopt imaginative strategies to deal with these challenges, and to establish treatment and control settings. One paper (study 994) uses a Randomised Control Trial approach and scores 5 on the Maryland Scale. This paper looks at the US Community Restitution Apprenticeship Focused Training initiative (CRAFT), a 6-month employment programme designed to train and place high risk youths and juvenile offenders in employment in the construction industry.
Three papers score four on the Maryland Scale, and use a range of approaches for identification. One uses instrumental variables (IVs). Study 878 looks at wages, investigating the effect of the Austrian apprenticeship system, which is a one-year vocational training scheme aimed at secondary school leavers. It focuses on firms who go bankrupt after taking on apprentices, so that apprentices’ training time varies in ways participants cannot control. Study 897 uses spatial and time variation in policy rollout in Italy to identify effects: apprenticeships legislation was introduced in different regions at different times, unrelated to regions’ underlying economic trends. As a result, the evaluation is able to compare the change in performance of similar firms ‘treated’ with apprenticeships at different points in time. Study 939, another Italian evaluation, uses a regression discontinuity design that exploits different regional age cutoffs in eligibility for apprenticeships. As a result of these variations in eligibility similar people under the age of (say) 27 will get apprenticeships in some regions and not others.
Studies look at post-apprenticeship job moves tend to focus on particular classes of movers, where the decision to move is (more or less) random. For example, Study 867, for Germany, looks only at apprentices who move job when their employer shuts down or conducts a mass layoff. In some cases these movers also have to look for work in different industries to which they trained, which helps test the transferability of their skills and human capital.
These examples show a number of ideas that UK policymakers could adopt to evaluate the apprenticeship system in this country. In addition, access to administrative datasets that track individuals over time – such as data held by the HMRC, DWP and DfE – will be important in identifying the long term impacts of apprenticeship programmes.
The returns to education literature
The papers we review are those where apprenticeships are the main focus of the evaluation. This helps ensure that the studies we consider explicitly tackle the challenge of trying to identify the causal impact of apprenticeships on outcomes such as employment and wages.
The apprenticeships studies we review here complement the wider literature on the returns to education. The latter typically regresses wages on educational achievement controlling for personal characteristics (i.e. estimates a Mincerian wage equation). As with the studies we consider on apprenticeships, these studies struggle with selection bias: there is selection into both type and length of education based on unobservable individual characteristics such as motivation and ability. In order to deal with the selection problem, a large body of ‘returns’ studies use instrumental variable approaches. Card (1999) summarises some of these earlier IV studies, finding the wage return to a year of schooling to typically lie between 8 and 13%.
In contrast to the extensive general returns to education literature, there are far fewer studies where the main focus is apprenticeships. To the extent that the strategies used in the return to education studies convincingly deal with selection in to apprenticeships they would provide additional evidence on wage effects. But assessing this would be a time consuming task – hence our decision to focus on studies explicitly looking at apprenticeships and to cover a wider range of outcomes than in some of our other studies.
Most of the available evidence that we review uses a before-and-after comparison against a control group (SMS 3). A before-and-after study of the wage returns to apprenticeship is challenging for two reasons. First, it is difficult to find a pre-apprenticeship wage as most apprentices are school leavers and did not work before. Second, the before-and-after comparison should cover a long study period since apprenticeships are a long term investment in which the pay of materialise over the lifecycle of earnings.
For these reasons we only find relatively few studies (five in total) that estimate the wage returns to apprenticeships. Three of these use an instrumental variable approach, as in the wider returns to education literature (837, 878, 986). Since a pre-apprentice wage is not available, the two studies that use a before-and-after approach (847, 976) use wage changes at a later stage in the apprentice’s career. This is not a strict before-and-after comparison but does eliminate the effect of fixed unobservables. The drawback here is that it identifies only the effect on wage changes not levels; thus only a partial impact. These wage return studies typically study a longer period than evaluation of other outcomes. For example a dataset covering over 20 years is used in study 837.
Some studies examine a different aspect of wages by looking at the change in wages following a job move for apprentices who move out of their profession against those who stay in the profession. This provides a measure of transferability of skills rather than general returns to apprenticeships.
Many of the studies we review, however, do not examine wages at all but instead consider outcomes such as employment and further study. These are more easily implemented as a before-and-after policy evaluation with shorter evaluation periods. More than perhaps any of our other reviews, the existence of a substantial evidence base on the returns to education suggests that the focus in this policy area needs to shift to assessing the effectiveness of different elements of policy design. We return to this issue below.