6 Work-based skill acquisition
Worker skills are determined by endowed ability, family background factors, accumulated knowledge from formal education, and learning in the workplace. Work-based learning can occur through learning-by-doing, formal employer-funded or government-funded training, investments by earlier employers, and informal knowledge transfers from other workers. Since a significant proportion of knowledge accumulation occurs on the job, or is required by continuous learning in response to technological obsolescence, the adequacy of work-based training is important. The standard model of human capital formation suggests that training will be provided in sufficient quantities by market forces, and that there is no basis for government intervention. Other commentators have pointed to evidence on labour market imperfections which suggest that employers and workers will tend to under-invest in training. The section considers the rationale and scope for government intervention in work-based training, and the factors that should influence the design of suitable interventions.
In the knowledge accumulation process, the knowledge acquired in one period complements the skills available in next. The stock of knowledge in one period permits further rounds of skill development and product innovation, with consequential productivity effects. Thus, skills can be expected to be more productive in environments where there is an existing high stock of knowledge. This effect can influence a firm’s policies around recruitment, the building up of firm-specific knowledge, information retrieval processes, or the use of teams. One important implication is that the likelihood of getting training appears to rise with the prior level of educational attainment. Firms spend more on training those who already have qualifications. Conversely, those with few educational qualifications are likely to face more limited training opportunities in the workforce. Thus, the observed patterns of provision and participation in training tend to amplify skill gaps, rather than compensate for low levels of prior educational attainment.
6.1 Analysis of work-based training
In decentralised labour markets, the dominant model for thinking about firm-based human capital investments is that of Becker (see Acemoglu and Pischke (1999)). Skill acquisition is viewed here as an investment decision. In the absence of capital and labour market imperfections, this model substantially relies on workers and employers to finance the optimal level of work-based training. Workers evaluate investments in education, based on likely improvements in future earning streams, the cost of training and any earnings foregone. They may also accept lower or no wages early in their career while their skills are being built up. Employers, faced with risks of labour turnover or poaching by other employers, will base a worker’s wages on their current productivity. They will invest mainly in firm-specific knowledge (which by definition is not valued elsewhere), and require training wages to be paid to fund the acquisition of generic skills that can be used by other employers. Workers will, in effect, be required to bear the full cost of general training. In the Becker model, no externalities are involved, and there is no case for government intervention on efficiency grounds.
The OECD (1999) and Long et al (2000) identify a number of factors that are systematically associated with employment-based training, which are broadly in line with the predictions of the Becker model. Those in jobs longer are more likely to receive formal training, while the extent of training falls off with age. Employer funded training increases with tenure, while higher labour turnover lowers the incentives on businesses to invest in training. Men tend to receive more financial support, while women receive less training in total because of their less continuous employment. The level of on-the-job training increases with the size of the firms, possibly because larger firms have larger internal labour markets and career paths, and hence greater opportunities to reap the benefits of the training (see Long et al. (2000)). Employer funded training reinforces skill differences, with more educated workers being trained more, although NZ is one of several OECD countries where there is a more even distribution of training.
A substantial proportion of training is informal. The OECD (2003a) cites Australian and US evidence that informal training accounts for at least half of all training, although it should be noted that informal on-the-job training is more difficult to measure. This training may take the form of general work experience, the upgrading of technical skills, the acquiring of firm specific information, or the development of a wider range of skills. The OECD (1998, p 60) finds that the earnings effects of work experience are significant, and that the effects of additional years in work since formal education have effects of a similar order to years of schooling. However, Loewenstein and Spletzer (1998) found that off-site training paid for by the firm had larger effects on wages than on-the-job training. This could be because informal training contains larger levels of specific skills, while formal training allows accreditation of the skills received. Thus, if worker expertise is unclear, recruiting firms may reduce wage offers. These factors seem likely to increase the wage effects from formal training.
(i) limitations of the Becker model
One difficulty with the Becker model is that firms are observed to fund a substantial amount of general training. The OECD (2003a, p246) notes that firms fully pay for more than 70% of formal vocational training, and that most of this training is quite general in nature. In the Becker model, employers would not pay for such training, but would require workers to accept a training wage. Recouping training costs by paying workers wages below their productivity is not possible, as it would encourage workers to change jobs. There have also been studies of the extent to which employers pay training wages. Loewenstein and Spletzer (1998) using US longitudinal data, note that employers almost always pay the explicit costs of training. Barron, Berger, and Black (1999) found that training lowers starting wages, but that the estimated effects are small. They also found, when controls are made for unobserved worker ability, that the effect of training on wages was small relative to the large and robust effects on productivity growth. They suggested that firms pay most of the cost and reap most of the benefits from training.
Another difficulty with Becker’s model is that the evidence points to firms paying wage increases following training that are less than the gains in firm productivity. The growth in the worker’s wages do not equate to the increase in their productivity from training. The gains in productivity also accrue to the firm in terms of increased profitability. For instance, Loewenstein and Spletzer (1998, p142) note evidence that “the existing empirical evidence suggests that employers realize an inordinate share of the returns to training”, with productivity growth about five times as large as wage growth. Similarly, Dearden, Reed, and van Reenen (2000) explore the direct effects of firm-based training on industry productivity using longitudinal UK industry data on wages, training and productivity, and find that a 5 percentage point increase in workers trained is associated with a 4% increase in worker value added and a 1.6% increase in wages.
One approach to explaining why general training is so prevalent without training wages utilises the imperfect competition model. With employers having monopsony power in the demand for labour, workers are paid less than their marginal productivity. Where this occurs, workers will not gain the full benefits of training and will tend to under-invest. Imperfect competition could be occurring for several reasons (see David (2001, pp9-15) and Acemoglu and Pischke (1999)). The first relates to the problems associated with information asymmetry and monitoring, where different levels of information are held by the transacting parties. This occurs in labour contracting matters, such as recruitment when the expertise of the applicant is incompletely known, or in assessing work effort where the level of shirking is unknown. Incomplete information tends to turn general skills into specific skills, and less useful to future employers when recruiting. The second explanation for imperfect competition arises from mobility or exit costs associated with moving between employers. The search and mobility costs associated with finding a better wages with another employer generate rents for the present employer. This allows the current employer to pay wages below the worker’s productivity and recoup training costs. Stevens (1999) suggests that imperfect competition can also arise where skills are only partially transferable, and are not useful for many other employers.
Acemoglu and Pischke (1999) and Acemoglu (2001, Chapter 5) note, however, that one consequence of the compression of wages is that firms have an incentive to increase training back towards the competitive equilibrium. This is because they are able to reap the productivity gains not passed on to workers in higher wages. Acemoglu (2001) sets out evidence on the relationship between wage compression and training in several contexts where proxies for wage compression can be found, namely minimum wages, union effects, and the cross-country effects of labour market institutions. He cites some evidence of positive effects of wage compression on training levels, especially in cross-country comparisons.
Another interpretation of this evidence on training is that a firm’s wage structure may be influenced by incentives to encourage work commitment, reward effort, and reduce worker turnover. For example, “efficiency wages” may be paid that are initially low but increase over the worker’s tenure. Such a practice could limit shirking because of the threat of firing. In this case, the presence of general training or wages initially below worker productivity is less of a problem. Thus, Cappelli (2002) provides evidence that US employers routinely provide financial support for a significant percentage of their employees to pursue post-secondary education, but that a high proportion of those employees have to wait for a year before gaining access. He suggests that the payment of assistance for part-time study acts as a screening device that helps to identify above average ability workers, while delaying employee access to assistance reduces turnover. Firms recoup their investment from retaining higher productivity workers and reducing recruitment costs. In both human capital and efficiency wage models, wages can be expected to rise over time. David (2001) argues that distinguishing these two stories empirically is difficult. Training practices may also be used to screen workers, where information on the worker’s ability and motivation are hard to obtain directly.
(ii) poaching and the distribution of benefits between employers
Where general training is provided, scope exists for workers to leave their current employer before the full benefits of training have been reaped. If the pay scales do not reflect training-based productivity improvements, this reduces the incentives on employees to invest, and increases their incentives to change jobs to get a better wage/productivity match. As a consequence, much work-based training appears to be funded by employers (OECD (2003a)). This provides scope for “poaching externalities” and the under-provision of training by the current employer.
A variety of evidence is available on the effect of training on labour mobility. Blundell et al (1999b) reports studies that have found employer-provided training is quite portable. In the UK, the returns from on-the-job training undertaken with a previous employer are similar to the return from that training with the current employer. Similarly, Blundell, Dearden and Meghir (1999a) find that employer provided training leading to qualifications gives positive returns for both current and earlier employer. Further, Loewenstein and Spletzer (1998) cite US evidence where past spells of general training have larger effects on wages than current spells of general training. However, Dearden et al (1997) cite evidence that firms who train also tend to retain those workers. They examine the relationship between training and job mobility in Britain using two longitudinal data sets. They find that employers who provide training have a lower than average probability of losing workers who received training in previous periods, especially if the training was employer funded.
There is a tension here between the benefits of better matching flowing from labour turnover, and the disincentives on employers to invest in training. As in the Becker model, increasing competition in the labour market increases the scope for “poaching”. If positive or similar returns accrue to future employers, as the evidence suggests, poaching would be beneficial. Conversely, as post-training wages rose towards the level of the worker’s productivity, the benefits to firms from funding training would reduce, and so would the level.
