Economic Outcomes of Adult Education and Training
Active labor market policies – Policies aimed to facilitate adjustment to job loss, assist the jobless to find work, and encourage labor force participation. In addition to training and skills development, these policies include public employment services (counseling, job-search assistance, information, and placement), youth measures to aid the transition from school to work, subsidized employment, and measures for disabled workers.
On-the-job training (OJT) – Training taking place in the firm to learn procedures that are job related, but where learning is an activity separate from regular job duties.
Portability (of training) – The capacity to utilize skills and obtain returns on human capital investments obtained with a previous employer. This concept is linked to that of the specificity of training.
Selection bias – Error committed when estimating the effect of an intervention such as a training program, due to the fact that treatment and comparison groups are intrinsically different, and would have experienced different outcomes even in the absence of the intervention.
Specific training – Training that provides skills that are exclusive to a particular firm, occupation, or industry and cannot be applied elsewhere. The antonym is general training.
Adult education and training have been focal points of labor market studies in developed economies since the 1960s. Increased attention to this topic is part of a shift toward greater emphasis on human capital in both economic and social policy in societies that are increasingly knowledge based. Several factors account for the growing interest in these issues. Technological change – especially advances in information and computer technologies – and globalization of production have resulted in a growing demand for highly skilled workers and changes in workplace skills requirements. Within specific occupations and professions, rapid changes imply a greater need for maintaining and updating skills and knowledge about current best practice. In addition, because of demographic changes and reduced fertility in many countries, the labor force is both aging and growing more slowly than in the past. In an environment in which labor shortages may become more common, there is greater emphasis on the skills of the existing workforce. As the workforce ages, there is a growing risk that older workers who lose their jobs may not have the skills needed to become reemployed.
Skill acquisition of adult workers takes many forms. Learning by doing (acquiring skills during the course of one’s regular duties), for instance, is an important source of growth in workers’ skills. Here, we focus instead on the acquisition of abilities that are job related, but where learning is a separate activity from regular job activities. Within these, we distinguish between formal (learning leading to the completion of diploma or degree) and informal (seminars, interdepartmental visits, etc.) training. A second important distinction relates to who pays for training. From the institutional point of view, individuals, employers, and governments generally differ in their goals when purchasing or providing adult education and training. Government-sponsored training is often concerned with enhancing the employability of displaced or unemployed workers, disadvantaged workers with limited skills, and those at high risk of social exclusion. Recipients of publicly provided training are likely to be unemployed and to come from the bottom of the skill distribution curve. In contrast, employer-sponsored training is provided to those currently employed, as is much of the adult education and training funded by individuals. Furthermore, as many studies have shown, those who are already well educated and highly skilled are more likely to receive work-related training. Employer-sponsored training is also more likely to provide firm-specific skills, which do not raise the value of the trainee to other employers. Governments, on the other hand, prefer general skills that increase individual employability in a variety of workplaces. Finally, these types of training generally differ in their timing. Employer-sponsored training will likely be spread out over the worker’s career but the duration of any particular spell may be short, whereas government-sponsored training is concentrated during spells of unemployment and is usually discontinued when the worker becomes reemployed. Thus, there are numerous reasons to distinguish between publicly financed and employer-sponsored training.
Human capital theory has long emphasized the importance of on-the-job training (OJT) and work experience for productivity and earnings. Recently, additional empirical findings have given rise to a new training literature (Acemoglu and Pischke, 1999). Leuven (2005) provides a useful survey of theoretical developments. Evaluating the impacts of adult education and training on outcomes has generated an extensive literature and a wide range of estimated impacts. In this article, we provide an overview of the methods commonly used to evaluate the outcomes of adult education and training and a summary of the results of this literature.
The principal outcomes that have been examined – especially for government-sponsored training – are individual labor market outcomes such as wages, employment, and earnings. Earnings is a useful summary measure because it captures both the price dimension – the wage rate – and the quantity dimension – hours, weeks, and years of work. For employers, the key outcome isworker productivity and thus firm profitability, although the provision of training opportunities may also influence other dimensions of performance such as innovative activity as well as recruitment and retention of employees. The impacts on earnings and productivity are key ingredients in any assessment of the costs and benefits of adult education and training.
Methodological Issues in Evaluating Outcomes
Obtaining credible estimates of the impacts of adult education and training programs is a difficult problem. In modern approaches to evaluation, the nature of the challenge is usually illustrated using the potential outcomes model (Angrist and Kreuger, 1999). Each individual in the population of interest – which in this case consists of individuals who may or may not receive training – has two potential outcomes: (1) the outcome that would be experienced if he/she receives training, and (2) the outcome that would be experienced if he/she does not receive training. One of these two potential outcomes is inherently unobservable. For each participant, we can never know what outcomes that individual would have experienced had they not participated. Similarly, for each person who does not participate we cannot know what outcomes would have been experienced had they participated. The problem is thus one of obtaining credible estimates of the counterfactual outcome – the potential outcome that cannot be observed. For participants, this means obtaining an estimate of the outcomes they would have experienced had they not participated in adult education or training.
The gold-standard approach to the evaluation problem is to use random assignment to treatment and control groups. Random assignment ensures that the treatment group that receives training and the control group that does not receive training are statistically equivalent in their pretreatment characteristics; that is, that there are no statistically significant differences in the characteristics of the two groups. Since the two groups are statistically equivalent prior to treatment, any differences in outcomes experienced by the two groups can be attributed to the causal impact of the treatment.
When random assignment is not feasible or desirable, there are a number of nonexperimental methods that can be employed, including regression-based methods, natural or quasi-experiments, instrumental variables, matching estimation, and panel data methods. Each of these methods has advantages anddisadvantages, and has different data requirements. Furthermore, no one method dominates other methods in all circumstances. Determining the methodological approach to employ in a particular setting requires a detailed understanding of institutional features relating to program operation and the nature of available data. Simply put, there is no magic bullet in nonexperimental methods.
All nonexperimental methods for estimating the impacts of education and training programs use a comparison group of nonparticipants to estimate the outcomes that participants would have experienced had they not participated in the program. However, in the absence of random assignment, participants and nonparticipants will generally differ along several dimensions, some that are observed by the researcher and some that are unobserved. Because of these differences between the two groups, the outcomes experienced by the comparison group – in the absence of statistical or econometric modeling – are unlikely to provide a reliable estimate of the counterfactual. In particular, since individuals choose whether or not to apply for and/or enroll in a training program, the individuals who receive training constitute a self-selected sample and may differ in various ways from the individuals who do not apply for and/or enroll.
Employers also choose which employees receive training and which do not. Similarly, for governmentsponsored training, there may be some selection by those who administer or operate the programs. Simple comparisons of participants and nonparticipants are thus likely to be subject to selection bias and do not provide credible estimates of program impacts. The various nonexperimental methods provide different ways of dealing with the selection bias problem.
In the past two decades, there have been significant advances in nonexperimental evaluation methods. Equally important have been improvements in the quality of the data available for analysis. As a consequence, recent studies using state-of-the-art techniques to assess the impacts of adult education and training generally provide more credible evidence than earlier studies.
Outcomes of Adult Education and Training
As noted previously, there are several reasons for distinguishing between government-sponsored adult education and training and private-sector-sponsored training. Although the expenditure on private-sector-sponsored training far exceeds that by governments – even in countries such as Sweden that devote considerable public expenditure to training – more is known about the outcomes of publicly supported training.This situation reflects a greater desire by governments (and taxpayers) to evaluate the impacts of their programs, as well as data challenges in assessing the impacts of employer-based training.
Government-sponsored training is part of the broader category of active labor market policies (ALMPs) that facilitate adjustment to change, assist the jobless to find work, and encourage labor force participation. These policies include public employment services (counseling, job search assistance, information, and placement), youth measures to aid the transition from school to work, subsidized employment, and measures for the disabled. Our focus here is on training programs, which are the most important active measure in many Organization for Economic Co-operation and Development (OECD) countries.
There is a significant amount of research assessing the labor market outcomes of government-sponsored training and other active programs. The comprehensive survey by Heckman et al. (1999) summarizesmore than 40 studies from the US and Europe. Other useful surveys of this literature include LaLonde (1995), US Department of Labor (1995), Martin and Grubb (2001), and Kluve and Schmidt (2002). Our treatment is brief because several excellent surveys are available and there are space constraints.
Relative to other OECD countries, the US has traditionally displayed a stronger commitment to rigorous evaluation of government programs, including substantial use of randomized trials. As a consequence, the state of knowledge about the effectiveness of these programs is dominated by US evidence. However, as the recent surveys by Martin and Grubb (2001) and Kluve and Schmidt (2002) indicate, there is growing use of serious evaluations in several European countries, resulting in a rapidly expanding body of evidence on the impacts of active programs.
Several salient features emerge from reviews of the US evidence (LaLonde, 1995; US Department of Labor, 1995; Heckman et al., 1999). First, the impacts on earnings of disadvantaged workers are mixed, being positive for some groups and zero or even negative for others. In this regard, it is worth noticing that training programs for displaced workers (who typically have considerablework experience and are highly motivated) have in general, a better track record than those for the disadvantaged ( Jacobson et al., 2005b). Other generalizations include:
- government-sponsored training significantly raises the earnings of economically disadvantaged adult women;
- the effects of these programs on disadvantaged adult men are often smaller than on women, and not always positive;
- the impacts of training on out-of-school youth are generally zero or negative;
- classroom training can be effective for adult women, but has limited success for adult men, especially those with low education; and
- best results are obtained when classroom training or OJT have a strong work experience component linked to local employers.
Greenberg et al. (2004) offer evidence from meta-analytic techniques on the persistence of these effects. Impacts on earnings seem to diminish over time for adult males and youth, but remain stable over time for females.
A second feature of the US evidence is that estimated impacts, even when they are positive, are generally modest in size. As noted by LaLonde, this outcome is not surprising, given the limited duration and cost of these programs:
The best summary of the evidence about the impact of past programs is that we got what we paid for. Public sector investments in training are exceedingly modest compared to the magnitude of the skill deficiencies that policymakers are trying to address. Not surprisingly, modest investments usually yield modest gains. . . (LaLonde, 1995: 149)
A third salient feature of the US evidence is significant heterogeneity in estimated effects. Evaluation of experimental programs shows earnings effects that range from negative impacts for individuals who never worked before, to large positive impacts on economically disadvantaged female household heads or economicallydisadvantagedmale household heads. Substantial variation in program impacts across sites is common in training programs using random assignment. Variability of estimated impacts on earnings is also evident in nonexperimental programs. However, when studies most susceptible to selection bias are removed, the qualitative evidence from nonexperimental programs is similar to that of experimental programs (Heckman et al., 1999).
In Canada, the published literature on government training programs is considerably sparser – likely due to the fact that program evaluations are carried out internally by government agencies and the findings are not published in academic or policy journals (Riddell, 1991). Riddell (1995) reviews employment and training programs that operated in the 1980s and early 1990s. Programs targeted on the economically disadvantaged had modest effects similar to those found in US studies, but impacts of programs serving clients with fewer barriers were larger. Subsequent analysis by Park et al. (1996) found significant gains from training programs for some groups, including women reentering the workforce and individuals trained in areas with identified skills shortages.
European programs are more focused on speeding the transition to work and reducing unemployment, especially among youths. The difference in emphasis reflects the facts that European youth are generally less economically disadvantaged than US youth, but are more likely to be unemployed for extended periods of time. A common finding of European studies is that training results in substantial gains in employment but has little impact on wages (Heckman et al., 1999; Kluve and Schmidt, 2002).
In summary, government-sponsored training has a mixed record, with rising earnings of some groups (adult women and displaced workers) but having little or no impact on earnings of others (disadvantaged adult men and youths). Even when they are positive, the impacts on earnings are modest and are usually not large enough to substantially reduce poverty rates. Training does, however, have a better track record in improving the transition to employment.
Two general responses have emerged in response to the mixed performance of public training programs. One, illustrated by Carneiro and Heckman (2003), is to argue that problems associated with low skills should be addressed much earlier in the life cycle. The other, illustrated by Martin and Grubb (2001), argues that public training programs can be made more effective by improving their design. Design features that they regard as crucial include careful targeting on participants most likely to benefit from training, keeping programs small in scale, including a strong work experience component to establish links with employers, and having programs that produce a certification that is recognized in the labor market.
Private Sector Training
Private-sector training refers to forms of skill acquisition that are job related and where learning is an activity separate from regular job duties. This type of skill acquisition may include both formal and informal learning, be general or specific and either employer- or individuallysponsored. In general, employer-financed training tends to be firm specific and informal in nature. However, firms may finance the acquisition of (formal) vocational training, such as apprenticeships, that is more general in nature. Studies of work-related training focus on one or the other type of training, depending on data availability.
Methodologically, the literature surveyed here relies on nonexperimental data. The selection problems are acute, as employers are likely to select for training the most promising individuals, who may have higher-thanaverage earnings regardless of training. Similarly, individuals who choose to undertake training are likely to be more productive or to have unobserved characteristics, such as motivation, associated with higher wages. The papers we consider here use different methods to resolve this problem.
Interest in the training policies of firms and the skill acquisition choices of workers originated in the US during the 1980s when slow productivity growth (relative to Japan) was at the forefront of the economic agenda. However, there was little information about training in available data (Barron et al., 1997). During the next decade, research produced many studies on the determinants of training and on the effects of training on wages, thanks to the availability of the National Longitudinal Survey of Youth (NLSY), which follows a representative sample of youth over the period since 1978, gathering exhaustive information about their labor market experiences. The survey asks about any type of training, other than schooling, military training, and government-sponsored training, and further categorizes it as: (1) company training or OJT, (2) apprenticeship programs, and (3) off-the-job training (OFT). The richness of information regarding skill-related characteristics contained in the survey allows for adequate measurement of the variables involved, including usually unobservable variables such as those measuring ability. In addition, the longitudinal nature of the data allows researchers to address the methodological issues regarding selection bias.
Most of the US literature is hence based on the NLSY. One of the earlier studies by Lynch (1992) focuses on a subsample of non-college-educated youth to estimate the effect on wages of weeks spent in company training, OFT, or apprenticeship training. The estimated coefficients translate into approximately 25% higher wages for an average apprenticeship of 63 weeks and 15% for an average spell of company training of 31 weeks. Using all workers in the NLSY sample, Veum (1995) finds smaller returns of 7.5% for an episode of company training, 6% for an apprenticeship, and 11% for an occurrence of OFT. In addition, he estimates that company training and OFT increase wage growth by similar magnitudes. Both these studies are confined to episodes of training lasting more than 1 month. Later, Parent (1999), using a more sophisticated methodology and not restricting the sample to long training episodes, estimates that 1 year of OFT or OJT training increases earnings between 12% and 16%.
Further issues arise within this literature. One relates to the portability of training. Lynch (1992) separates training into that received with a current employer and that received with a previous employer, finding high returns to apprenticeships and OFT received during a previous job, but no returns to company training received during a previous job. The results in Parent (1999), on the other hand, suggest that employer-based training and OFT are the most portable forms of training. Given the small fraction of individuals undertaking apprenticeships (around 1%), results regarding this form of training should be treated with caution. A second issue is who pays for private-sector training. Parent (1999) finds that workers partially pay for OJT with lower initial wages. Barron et al. (1999) use a different data set and find that although the initial wages seem to be smaller for workers undergoing OJT, the difference is small. The general perception in this matter is that firms pay for OJT training and reap the benefits through higher productivity of the firm.
In Canada, private-sector training has been much less analyzed and the results are nonconclusive. Parent (2003) uses the Follow-Up of the School Leavers Survey (FSLS) and finds that participation in employer-provided training raises male earnings by 10% but not significantly so for females. Results from this study also suggest that training may increase employment. Hui and Smith (2003), using the Adult Education and Training Survey (AETS), find positive effects of self/employer-financed training on employment for women. In addition, small positive effects on weekly earnings from employer-financed training are apparent for both genders, while self-financed training seems to have small negative impact on female wages. These findings seem to be contradicted by Havet’s (2006) study, which uses matched employer–employee data and finds a positive impact of firm-provided training on the wages of women, but not of men.
Studies from Europe also tend to find positive effects of training on labor outcomes. In Great Britain, Greenhalgh and Stewart (1987) find positive effects of vocational training on occupational status. Booth (1993) estimates that 1 week of training raises the earnings of graduates by 1%. Both studies find significant gender differences. More recently, Blundell et al. (1999), use the National Child Development Study (NCDS) to study the impact of work-related training on the earnings of UK workers. They find positive returns (5–6%) for employer-provided training and also that this type of training seems transferable across employers. These findings are corroborated in a later study by Arulampalam and Booth (2001) that uses the same data.
For countries other than North America or the UK, evidence on the returns to training is less abundant. Goux and Maurin (2000) estimate wage returns to employerprovided training for France. They report around 7% higher wages among those who received training, but the estimates become insignificant when they account for firm selection of workers on the basis of posttraining mobility. Similarly, Pischke (2001) finds that, in Germany, 1 year of full-time work-related training (training during leisure time) increases wages between 3% and 4%. When he accounts for selection based on wage growth, the returns are higher in magnitude, from 3% (males) to 6% (females), but not statistically significant. Schøene (2002) finds that employer-financed training participation is associated with 1% higher wages in Norway when controlling for selection based on seniority and job complexity (down from a 5% return in the absence of selection controls).
Overall, the returns to private-sector training seem quite high – higher, for instance, than the returns to schooling. In contrast, the incidence of training is not large, which is puzzling if there are such high returns. A leading explanation for the high returns to training is selection bias – that is, high-productivity workers are more likely to receive training and to have higher wages and/or wage growth. All papers find that addressing this selection problem is important. Estimates produced with a selection correction procedure are generally smaller than those obtained without, corroborating the heterogeneity of the returns to training. The effect of accounting for selection varies depending on the method used. In general, estimates are reduced by 40–50%. In some cases, the results are quite extreme. For instance, when selection is based on worker’s mobility, since firms are less likely to train workers at high risk of leaving in the near future, the effect of training practically disappears (Goux and Maurin, 2000). A potential explanation could be that most of the benefits of training accrue to job movers and not to job stayers. Gerfin (2003) provides some evidence for this hypothesis in his study of Swiss work-related training. He shows that training undertaken during the last year generates the highest returns to individuals who changed jobs (around 5% increases in monthly earnings) versus those who stayed at the firm (1.2%). Other studies find no returns to training once selection is taken into account. For instance, Leuven and Oosterbeek (2007) find that when comparing the returns to training of Dutch workers who took training with the returns of workers who intended to undertake training but did not do so because of a random event (sickness or family circumstances), earnings differences are not statistically significant. There is also cross-country evidence from the European Community Household Panel indicating that the effect of training is significantly smaller when individual heterogeneity is accounted for (Bassanini et al., 2007).
Finally, other factors may explain the correlation between high wages and training. Promotions, for instance, are a factor generally overlooked. These positively influence both pay and training. In addition, the returns to training are likely to differ by type of job, with more productive jobs (managerial/professional) requiring more training and higher pay than blue-collar jobs. Failure to account for these factors is also likely to overestimate the effect of private-sector training (Frazis and Lowenstein, 2005).
In summary, studies of informal private sector training seem to agree on the following:
- returns to training are generally positive and significant, particularly employer training and OFT;
- there are gender and ethnic differences in the returns to training, although the sign of the difference seems to be country specific; and
- there is substantial evidence that unaccounted-for factors lead to overestimating the returns to training. In general, those who undergo training seem to be those whose productivity, occupation, or other characteristics would lead them to have high wages regardless of training.
The papers examined so far in this section do not formally distinguish between formal and informal education, although it is expected that some OFT may be conducive to a degree or diploma. However, there are studies analyzing the impact of adult education and training that focus specifically on the effect of acquiring formal schooling later in life. Although the notion of adult education is country specific, it is generally accepted that formal schooling is associated with the completion of courses toward the achievement of a degree or diploma. The skills acquired with formal schooling are also more likely to be general rather than trade or profession specific.
The North American literature finds substantial returns to formal certification for older workers. Leigh and Gill (1997) and Jacobson et al. (2005a) find returns to community college in the US for workers over age 28 that are around 8% (9%) for males (females). In Canada, Zhang and Palameta (2006) use longitudinal information from the Survey of Labor and Income and Dynamics to evaluate the impact of formal schooling on earnings for individuals who have been out of school for more than a year and then enrolled in formal education. They find substantial earnings gains for those individuals who obtained a certificate (7% increase in earnings for men, 10% for females). In this study, younger workers who switch firms after obtaining a certificate gain the most from their studies, whereas workers over 35 reap higher benefits staying with the same firm. Ferrer and Menendez (2007) also find evidence for Canada of large earning effects from acquiring formal education later in life.
However, the European findings about the impact of adult formal education show that formal qualifications obtained later in life seem to have little impact on earnings. With the exception of the Blundell et al. (1999) study, which finds that all forms of work-related training leading to formal qualifications enhance earnings by 5–10%; other studies do not find such positive returns. Egerton (2000) uses 10 years of the British General Household Survey, which incorporates a rich data set of covariates including father’s social class, to examine this issue. Her results show no significant difference in earnings between full-time mature students and early graduates, although she does not address selection bias. Jerkins et al. (2003) use panel data from the NCDS to analyze the impact of adult education on employment and earnings. Their findings reveal that episodes of adult education, particularly in occupational training, have positive effects on employment but a limited effect on wages, except for the leastqualified individuals. Part of the difference in results with Blundell et al. (1999) is likely due to differences in the goals of study, which lead them to consider different control groups. In Sweden, Ekstro¨m (2003) analyzed the impact of adult secondary education on annual earnings during the early 1990s, finding a negative effect for males and only weakly significant positive effects for females. Later, Albrecht et al. (2004) follow the large expansion of Swedish adult education programs during 1997 through 2002, called the Knowledge Lift, to estimate the impact on annual earnings and employment of increasing formal schooling for the low skilled. Their results show no effect of Knowledge Lift programs on earnings or employment, with the exception of an increase in the employability (but not earnings) of young men.
In summary, the literature on employer-based training finds that this type of human capital investment generally has positive effects on labor market outcomes such as wages. The magnitude of the returns varies depending on the type of training undertaken and the methodology employed. As is the case with the literature on the returns to education, careful econometric analysis is needed to take into account the possibility of selection bias arising because training is undertaken by individuals who are more productive and would have commanded higher wages regardless of training.
See also: Apprenticeships; Evaluation of Adult Education and Training Programs; Lifelong Learning; Participation in Adult Learning; Training and Learning in the Workplace.
A Ferrer, University of Calgary, Calgary, AB, Canada
W C Riddell, University of British Columbia, Vancouver, BC, Canada
© 2010 Elsevier Ltd. All rights reserved.