Corps de l’article

Introduction

The issues of an older workforce are gaining importance for policy and practice for a variety of reasons. Because the workforce is aging, living longer and retiring later, work life is being extended with a growing proportion of workers in older age brackets. Older workers are gaining importance not only numerically but also in their potential to mentor and synergize with younger workers.[1] Evidence from personnel economics (Lazear and Freeman 1997), for example, indicates that a mix of young and old workers will most likely produce the most productive work environment.

Older workers may be working longer because the recession and financial crisis of 2007- 2008 dissipated the savings and increased the debt load of workers nearing retirement (Marshall 2011). As well, the longer time spent in acquiring higher education is providing an incentive to work longer to amortize education costs. Older workers will also be working longer because of growing uncertainty about their employer-sponsored pension plans and because incentives for early retirement in defined-benefit pension plans are no longer prominent. Finally, mandatory retirement policies have been largely banned by legislation (Conference Board of Canada, 2005).

In their transition to retirement and increasingly back from retirement, older workers are often leaving their career jobs and engaging in alternative “bridge” jobs, many of which are non-standard (e.g., part-time, limited-term contracts, self-employment, telecommuting) and quite different from their earlier career jobs (Cohen 2008; Monette 1996; OECD 2019a, 2019b). Although our research deals with the training of all older workers, including those who have been dismissed or lost their lifelong jobs, their human capital and skills are often industry-specific and tend not to fit the requirements of the new knowledge economy (Neal 1995). For example, many older workers have been displaced through mass layoffs from their earlier career jobs in declining industries like steel, pulp and paper and auto manufacturing, with harmful effects on their life expectancy (Morissette, Zhang and Frenette 2007; Sullivan and von Wachter 2009). Such displaced older workers are often considered in a state of limbo—too old to begin a new career, but too young to retire. Consequently, it will become increasingly important to understand how this aging workforce is utilized, given the growing knowledge economy and the decline in physically arduous blue-collar work (Beach 2008).

The literature on the effects of permanent job loss tends to focus on job loss from plant closings and mass layoffs. More recently, the pandemic has also led to massive job losses. The hope is that much of this is temporary and that individuals will eventually return to their former jobs when the pandemic is over. However, Barrero, Bloom, and Davis (2020) estimate that 42 percent of recent layoffs from COVID-19 in the United States will result in permanent job losses, and that there are only 3 new hires for every 10 layoffs caused by the pandemic. This pandemic-induced restructuring and reallocation of labour clearly will affect older workers, whose health is also more at risk because of the pandemic.

In addition to the issues of an aging workforce and the effects of economic restructuring, training issues have attracted more attention in the current economy. Employers are often facing skill shortages due to retiring workers in the large baby-boom cohort (Cohen 2008; Conference Board 2005; OECD 2019a, 2019b). With the decline in lifetime employment provided by the old standard employment contract, individuals can expect to change jobs more often, with obvious implications for training needs. This change has also put a premium on continuous lifelong learning—and relearning—with retraining being an important component of that process (OECD 2019a, 2019b; Steffens 2015). For example, vocational rehabilitation and workplace accommodation requirements often involve training components (Campolieti, Gunderson and Smith 2014). Training is generally regarded as a key component of active labour adjustment programs that help reallocate labour from declining sectors and regions to expanding sectors and regions, with the twin benefits of reducing unemployment in the declining sectors and decreasing skill shortages in the expanding ones (Cohen 2008). Active adjustment programs, like training, are generally preferred to passive income maintenance programs that can support the “stay” option and exacerbate unemployment and labour shortages. With the dramatic increase in higher education, and the realization that such education is no longer a ticket to secure employment, increased attention is being paid to vocational training that can make individuals “job ready.”

In addition to equipping workers to adjust to technological change, training can also foster or induce endogenous technological change, as has likely occurred with the computer revolution since the mid-1970s (Beaudry and Green 2005). This may be a case of Says Law: a growing supply of skilled labour will drive technological change (Acemoglu 1996, 1998, 2002). Training has also driven the innovation that is regarded as crucial to sustain productivity in a high-wage economy.[2] In the literature on high-performance work practices that foster competitiveness, the emphasis is on the importance of “bundling” training with other complementary workplace practices, such as employee involvement, job rotation, multi-tasking, broader-based job classifications and performance-based compensation.[3]

Clearly, the issues of an aging workforce and training are each of increasing importance for policy and practice. Their importance is compounded where both intersect: the training of older workers. That intersection is the focus of our study.

We will begin by discussing some of the theoretical issues in training of older workers. Particular attention will be placed on how those issues shed light on the training incentives of employers and older workers, on the barriers that older workers face and on what organizations can do to overcome them. We will then discuss the data that we will use, followed by a discussion of the empirical framework and estimation procedures. Next will come the empirical evidence on three relevant dimensions. First, we will profile how 54 different training indicators differ between older and younger workers, without controlling for any of the other factors that may influence such indicators. Second, we will provide an econometric analysis of the effect of being an older worker, as opposed to being a younger worker, after we have controlled for differences between the two groups in other determinants of training across the 54 training indicators. Finally, we will use a decomposition analysis to illustrate the extent to which differences in the mean probability of receiving training between younger and older workers are due to differences in the means of worker characteristics (explanatory variables) that affect training indicators, as opposed to differences between older and younger workers in the propensity to undertake or receive training (i.e., regression coefficients). Our paper will conclude with a summary and policy discussion.

Theoretical issues in the training of older workers

A multidisciplinary perspective is important because training of older workers can be studied within various disciplines and from different perspectives. The different theoretical perspectives and their interrelatedness are discussed below, with an emphasis on the implications for training of older workers.

The basic human capital framework of economics suggests two points: older workers are less likely than younger workers to be trained and they receive less training. The benefits are likely to be smaller for them and the costs higher. Specifically, older workers are likely to benefit less because they have a shorter remaining work-life over which to amortize the costs of training (Picot and Wannell, 1987). This is so whether employers or employees bear the costs (Xu and Lin, 2011). As workers grow older, they are more likely to be matched with the requirements of their job and do not go through the frequent job turnover that younger workers go through (and which requires re-orientation or re-training) as they search for a good job match (Park 2012). The accumulated experience of older workers may also act as a substitute for training.

In addition to the reduced benefits of training, older workers may also face higher costs in terms of lost wages, i.e., opportunity costs, during time spent on training. As well, there may be higher psychological and learning costs to the extent that the new training is harder for them to absorb, in part because more time has passed since their earlier period of formal education.

On a related note, the health and safety literature documents a strong positive relationship between age and disability (Arin 2015; Cossette and Duclos 2002) and between age and days lost due to injuries (Dillingham 1981), as well as longer absences due to illness and longer recovery times (Rosen and Jerdee 1985, p. 27). These facts show the need for vocational rehabilitation training and workplace accommodations for older workers.

The discrimination literature shows that older workers are subject to discrimination and age stereotyping, and there is little reason to believe that such discrimination would not affect training of older workers, as evidenced by the phrase “you cannot teach an old dog new tricks.” Such stereotypes have been documented in various reviews of the literature[4] and in résumé studies, which show that older workers receive fewer call-backs compared to equally qualified younger workers (Baert et al. 2016; Postuma and Campion 2009; Riach 2015; and Richardson et al. (2013). However, Kunze et al. (2013) provide evidence that, contrary to stereotypes, older workers are less resistant to change than are younger workers.

According to the literature on how productivity changes with age, there is little or no clear relationship between productivity and age.[5] The heterogeneity among individuals of the same age group is greater than the heterogeneity between individuals across age groups. Some skills like strength, dexterity, memory and reaction speed decline with age; however, older workers often compensate through increases in other inputs, such as institutional knowledge, firm-specific human capital, wisdom, diligence and experience as well as the ability to mentor younger workers.

According to the organization behaviour/psychology literature, retirement, and especially involuntary retirement, has negative effects on the cognitive functioning of older workers and their health and well-being (Bonsang et al., 2013; Mazzonana and Peracchi 2012; Rohwedder and Willis 2010). Training may facilitate not only continued employment but also cognitive learning.

Closely related to the organization behaviour literature, the psychology and training literature on cognitive and non-cognitive skills does suggest that older workers perceive themselves as having less need for training and having concerns over their ability to absorb and utilize training (Guthrie and Schwoerer 1996 and references cited therein). Importantly, the literature also finds that older workers have more difficulty in absorbing training. They take longer to be trained and may have limited productivity gains from training.[6] This difficulty reflects a variety of factors: declines in cognitive, physical, memory and motor skills; difficulty in keeping up with the pace of instruction; difficulties with conceptual learning as opposed to hands-on learning; lack of foundations in computer skills and IT; lack of familiarity with new equipment and technology; and awkwardness in being retrained with younger workers.

Importantly, these difficulties can be overcome if training is redesigned to meet the needs of older workers in several ways:[7] by making instruction slower and self-paced to allow sufficient time; by assigning hands-on practical exercises; by ensuring that the training is relevant; by building on the trainee’s current knowledge base; by using modular training components that can build on previous components to go from the simple to the complex; by providing feedback; by familiarizing the trainee with new equipment; by emphasizing experiential and practical learning, as opposed to conceptual learning; by minimizing required reading and the amount of material to cover; by training in small groups; and by training older workers separately from younger workers.

While these various perspectives can shed light on the issue of older worker training, we find the human capital perspective and the psychology and training literature to be most useful for interpreting our subsequent empirical results. The human capital perspective shows the incentives for both employers and employees, while the psychology and training literature shows the barriers to older workers and how employers can design training to overcome those barriers.

To understand the features of training that may be barriers to the productive training of older workers, we used a worker-firm matched dataset that has 54 training indicators. These indicators may also show ways to accommodate the training needs of older workers.

While our 54 training indicators pointed to the usual suspects (training costs and benefits, barriers to older workers), there were also some surprises. Older workers did not refuse training primarily because of health reasons or a perception of their courses as being too difficult. They refused primarily because they felt the courses were not suitable. The onus is thus on employers and training institutions to design courses that suit the needs and capabilities of older workers.

WES data

Our empirical analysis was based on Statistics Canada’s worker-firm matched dataset: the Workplace and Employee Survey (WES). Because the survey has (unfortunately) been discontinued, we used 2003 data. Only odd-year WES data contained information on organizational factors that could affect training decisions.

On the employer side, the target population was defined as all business locations in Canada that had paid employees in March, except for employers in the Yukon, Nunavut and the Northwest Territories and employers in crop or animal production, fishing, hunting and trapping, private households, religious organizations and public administration. On the employee side, the target population was composed entirely of employees working or on paid leave in March in the selected workplaces that received an income tax form. The WES drew its sample from the Business Register (BR), a monthly updated list of all businesses in Canada. To reduce the response burden, this list may combine the information the businesses provide with data from other surveys or administrative sources (Statistics Canada, 2021).

Our analysis is based on the individual WES file. It is restricted to workers age 25 and older and is broken down into older workers age 50 and older (as defined by the OECD 2006) and workers 25-49. It is also restricted to for-profit organizations because they were the only ones that had information on whether there was competition and, if so, whether it was local, regional or global. The WES is an ideal dataset for our analysis because it is a worker-firm matched dataset and hence has information on both workers and firms, with one of its primary focuses being on-the-job training. It has an extensive set of 54 indicators about older worker training, including the following: whether the worker received training in the past year; type and duration of training; instruction for on-the-job training; nature of the classroom courses; instruction for classroom training; whether training was offered but refused; and reasons for refusing training, including being too old or too late in one’s career. The sample size is substantial: about 4,000 older workers over the age of 50 and 12,000 between the ages of 25 and 49.

Table 1 presents the means of the 54 training indicators for older workers (50+) and younger workers (25-49). For the various categories, these are the percentages of workers who received that type of training. The differences are raw and unadjusted, being not controlled for factors other than age. They are the dependent variables used in our subsequent regression analyses to control for other factors influencing those outcomes.

Table 1

Profile of Training Indicators for Older and Younger Workers, WES 2003 (% responding yes for categorical measures; mean magnitude for continuous measures)

Profile of Training Indicators for Older and Younger Workers, WES 2003 (% responding yes for categorical measures; mean magnitude for continuous measures)

-> Voir la liste des tableaux

As indicated by the negative differences in column 3 of Panel 1, training was shorter-lasting and less prevalent among older workers than among younger workers for almost all of the dimensions of training. This is a common result found in the literature across different countries.[8]

Regression analyses

The gross difference between older and younger workers across a wide range of training indicators may reflect differences in their training-associated characteristics and differences in their propensity to take different types of training, after controlling for or netting out the other factors that influence the training indicators. The net effect can be considered a pure older worker effect, since it controls for the other determinants of the training indicators. It is estimated here as simply the coefficient on an older versus younger worker dummy variable, which is based on separate regressions for each of the 54 training indicators, after controlling them for a wide array of other variables that can affect those training indicators.[9] As indicated at the bottom of Table 2, the variables include the following: gender; visible minority status; Aboriginal status; immigrant status; marital status; education; presence of dependent children; full-time vs. part-time status; regular permanent vs. non-standard work; presence of a collective agreement; use of a computer at work; use of technology at work; number of employees at the firm;% part-time,% temporary; new goods or processes being introduced at work; whether there was no competition or local, regional or global competition; existence of individual or group incentive plans; whether overtime was worked; whether there was downsizing; occupation; industry and region. These full regressions are estimated separately for each of the 54 different training indicators.

Table 2

Effect of Being an Older Worker vs. Being a Younger Worker on Various Training Indicators After Controlling for Other Determinants of Training Indicators (Coefficient from an older worker dummy variable in OLS regression)

Effect of Being an Older Worker vs. Being a Younger Worker on Various Training Indicators After Controlling for Other Determinants of Training Indicators (Coefficient from an older worker dummy variable in OLS regression)

Significance is denoted by *** at the 0.01 level, ** at the 0.05 level and * at the 0.10 level

-> Voir la liste des tableaux

Control variables include: gender; visible minority status; Indigenous status; immigrant status; marital status; education; presence of dependent children; full-time vs. part-time status; regular permanent vs. non-standard work; presence of a collective agreement; use of a computer at work; use of technology at work; number of employees at the firm;% part-time,% temporary; new goods or processes being introduced at work; whether there was no competition or local, regional or global competition; existence of individual or group incentive plans; whether overtime was worked; whether there was downsizing; occupation; industry and region.

Table 2 presents the pure or net older worker effect (i.e., the coefficient on the dummy variable coded 1 if the worker was 50 or older and zero if 25-49).[10] A negative older worker coefficient means that older workers were less likely than younger workers to receive that training outcome, to use that type of training or instruction or to refuse training for that reason. A positive coefficient means the opposite. Please note that in the absence of causal estimation procedures the relationships here are associations and not causal.

As indicated in the first row of the top panel of Table 2, after we controlled for other determinants of training, older workers were significantly less likely than younger workers, by 5.2 percentage points, to receive some form of training. This probability was 10% lower than the mean of 53.1%.

The coefficient for the days of training received also indicates that older workers received fewer days of training than did younger workers. Specifically, after we controlled for the other determinants of training, older workers received significantly less training than did younger workers, namely 2.3 fewer days. While the difference seems small, it was half the 4.6 days of training that all workers received on average.

Older workers were thus less likely to be trained and received fewer days of training on average, after we controlled for other factors that affect training. The reason may be the reduced benefit of most types of training for older workers (given their shorter time horizons for amortizing the costs, and the lower productivity gains after training) and the higher costs (given their higher wages and hence higher opportunity cost, as well as possible psychological costs).

Overall, considering the negative effects of all of the prevalence and duration indicators, we see that older workers generally had a lower probability of being trained and received fewer days of training on average even after we controlled for other factors that influence the prevalence and duration of training. These pure older worker effects are generally substantial in size when compared with the means of column 1. The effects are small and insignificant only for two indicators: having received only on-the-job training and having received only classroom training.

As seen in Panel 2 of Table 2, the negative effects of different types of on-the-job training show that older workers were less likely to be trained across most types of on-the-job training, especially for managerial, supervisory and professional training. In many cases, however, the differences between older and younger workers were small and statistically insignificant. The notable exception is that older workers were much more likely to receive on-the-job training in computer software. This is understandable because they likely did not have such training in their earlier education or as part of their lifestyle. Because they more likely received training in computer software, they must have needed it for their work and could absorb it readily enough—otherwise such training would not likely continue.

With respect to instruction for on-the-job training (Panel 3), older workers were much less likely to be taught by a supervisor or a fellow worker. This finding is logical because they likely had fewer supervisors, being older, and likely had younger fellow workers who were reluctant to train older workers. Mentoring and on-the-job training likely went in the other direction, from older workers to younger ones.

With respect to classroom training (13 indicators were about its nature (Panel 4) and 6 about instruction (Panel 5), the differences between older and younger workers were very small and statistically insignificant. The same applied to the probability of having refused training.

With respect to reasons for refusing training (Panel 6), older workers were much more likely to say the courses were unsuitable, rather than too difficult, i.e., a barrier. Indeed, course difficulty did not significantly predict refusal of training. This fact is informative because it shows that employers and training institutions should design and implement courses that are suitable to older workers’ needs and capabilities, as discussed previously. The concept of “one-size-fits all” obviously does not apply to the design and implementation of training courses for older workers.

Although older workers were obviously more likely to refuse training because they thought they were too old or too late in their career, this reason was of extremely minor importance. Few workers even mentioned it as a reason. Older workers were less likely than younger workers to refuse training because they were busy with job duties, presumably because they had already found a fit with their duties and could handle them. Family responsibilities and health factors were not significant reasons for older workers refusing training, even though health declines with age.

Decomposition analysis

The previous analysis of Table 1 indicates that the unadjusted or raw difference in the mean probability of being trained was 9.3 percentage points, with younger workers having a higher probability (i.e., 55.2% versus 45.9% for older workers). The training gap can be broken down into two components (Oaxaca, 1973). One component can be attributed to differences between older and younger workers in the means of the characteristics (explanatory variables) that affect training indicators. The other component can be attributed to differences between older and younger workers in the propensity to undertake or receive training (i.e., regression coefficients) for a given set of characteristics.

As shown in Table 3, the 9.3 percentage point difference includes 4.1 percentage points or 44% that can be attributed to differences between older and younger workers in the means of the explanatory variables or characteristics (i.e., personal characteristics, employment and workplace characteristics, human resource practices that come under incentive schemes, and occupation/industry/region). In other words, the lower probability of older workers being trained is almost half explained by characteristics that lower this probability for older and younger workers alike.

Table 3

Breakdown of 0.093 Training Gap, with Probability of Receiving Training Being Higher for Younger Workers (0.552) than for Older Workers (0.459)

Breakdown of 0.093 Training Gap, with Probability of Receiving Training Being Higher for Younger Workers (0.552) than for Older Workers (0.459)

-> Voir la liste des tableaux

The remaining 5.2 percentage points or 56% of the training gap can be explained by a lower propensity among older workers to be trained (i.e., differences in the regression coefficients including the constants in each equation). This lower propensity among older workers to be trained reflects the higher likely cost of training older workers (higher opportunity cost of lost wages during the time spent in training and possible psychological costs) and lower expected benefits due to the shortness of their remaining work-life and any lowering of productivity from taking training, as discussed previously.

Summary observations for individuals and employers

The evidence presented here is generally consistent with older workers and their employers making rational decisions on various aspects of training. Older workers had a lower probability of being trained—almost half as much—because they had more of the characteristics (personal, employment, workplace, human resource practices and occupation/industry/region) that lower this probability for both older and younger workers. The remaining half of the training gap can be explained by a lower propensity among older workers to be trained, likely due to higher expected costs and lower expected benefits.

Training of older workers was less prevalent across most types of on-the-job training likely because the benefits were lower for them and the costs higher. As discussed previously, the benefits were lower because older workers had less time remaining in the labour force and because their greater experience could substitute for training. The costs might be higher because older workers earn more money and suffer a higher opportunity cost, as well as a higher psychological cost and greater difficulties in absorbing training. Discrimination might also be a barrier.

On-the-job training was more prevalent among older workers in only two cases. One case was training in computer software. This exception is consistent with older workers not being exposed to computers in their much earlier formal education and not acquiring computer skills through their lifestyle, as is the case with younger workers. Evidently, such training was required in their work, and they were able to absorb it; otherwise, it would likely be uncommon.

The other exception was training for health and safety. This might be a rational employer response to the extent that older workers incur higher costs for lost time due to accidents because they generally earn higher wages and recover over a longer period.

Supervisors and fellow workers were a less common source of on-the-job training for older workers, probably because the latter had fewer supervisors (often being themselves older workers) and because younger workers were less likely to train older workers. In contrast, self-learning was more common for older workers, probably because of the importance they attached to learning at their own pace.

Older and younger workers refused training in identical proportions. Perhaps surprisingly, older workers did not refuse because of expectable barriers like course difficulty, health problems or family responsibilities. The most important reason, and to a greater degree than among younger workers, was that the courses were unsuitable. This finding highlights the importance for employers and training institutions to design and implement training courses with older workers’ needs and capabilities in mind.

Can older workers be trained? Yes, they can be. But the training should be redesigned to meet their needs and capabilities in several ways: by making instruction slower and self-paced; by assigning hands-on practical exercises; by providing modular training to be taken in stages; by familiarizing the trainees with new equipment; and by minimizing required reading and amount of material covered. The concept of “one-size-fits- all” does not apply to the design and implementation of training courses for older workers.