Chap 4

2 downloads 0 Views 79KB Size Report
related training using national survey data for Australia collected in 1989 which indicate that NESB immigrant workers were indeed much less likely than workers ...

Published in:

Ethnic and Racial Studies 20(4), October 1997, pp. 830-848. [http://dx.doi.org/10.1080/01419870.1997.9993991]

Participation of non-English-speakingbackground immigrants in work-related training Audrey VandenHeuvel and Mark Wooden

Acknowledgement: This research was supported by funding from the (Australian) Bureau of Immigration, Multicultural and Population Research.

Audrey VandenHeuvel is Senior Research Fellow, National Institute of Labour Studies, Flinders University of South Australia. Mark Wooden is Associate Professor and Deputy Director, National Institute of Labour Studies, Flinders University of South Australia.

ADDRESS: National Institute of Labour Studies, Flinders University of South Australia, GPO Box 2100, Adelaide, SA 5001, Australia.

Participation of non-English-speakingbackground immigrants in work-related training

Abstract

This article uses representative data on a sample of Australian workers to examine whether non-English-speaking-background (NESB) immigrants are disadvantaged in terms of access to, and participation in, employmentrelated training. Results from the estimation of logit models explaining participation in various types of training indicate that NESB immigrant workers have much lower probabilities of receiving training, even after holding other individual and employment characteristics constant. Further, this differential was generally found to be most pronounced for those immigrants with English-language difficulties. Decomposition of the training differential, however, suggests that part of the explanation for the lesser incidence of training among NESB immigrants may well lie in employer-based discrimination.

Keywords:

Immigrants; training; employment; non-English-speaking background

1

Participation of non-English-speakingbackground immigrants in work-related training

The expansion of education and training has become a key platform of government economic and social policy in Australia over the last decade. In particular, Australian governments have made a concerted attempt to raise training effort through such vehicles as the National Training Reform Agenda (NTRA) — a complex web of regulatory arrangements including national competency standards, a national framework for the recognition and accreditation of training courses and providers, and a new national entrylevel training system. 1 A key element of the rhetoric surrounding the NTRA has been a commitment to improving training opportunities for persons identified as disadvantaged in the labour market, such as non-Englishspeaking-background (NESB) immigrants. Certainly it is true that NESB immigrants have higher rates of unemployment compared with other groups in the labour market, and those that do secure employment tend to be heavily concentrated in low-skill occupations and in industries, such as manufacturing, which have been particularly vulnerable to structural change.

On the other hand, other developments, and especially the move away from a highly centralised system of wage determination towards a more decentralised American-style enterprise-based bargaining system, may work against these objectives. 2 Numerous commentators (e.g., Alcorso and Hage 1994; Bertone 1994; Federal Race Discrimination Commissioner 1994) have expressed concerned about the potential for enterprise bargaining to

2

reinforce the disadvantaged position of NESB immigrants. The combination of poor English skills, cultural factors and the lack of participation of NESB immigrants in trade union structures may result in enterprise agreements which do not adequately reflect the needs of immigrants, including their training needs.

While the extent and nature of NESB immigrant disadvantage in the Australian labour market has been the subject of substantial research (see Wooden 1994 for a review), few researchers have considered the possibility that one source of such disadvantage may lie in lower rates of access to, and participation in, employment-related training. A relative lack of training opportunities, for example, may cause NESB immigrants to be denied access to the same career paths and promotion opportunities available to other workers, effectively locking them into low-skill employment.

To date, the only significant work concerned with the relationship between participation in training and place of birth is that of Baker and Wooden (1991) and Foster, Marshall and Williams (1991). Both of these studies investigated the relationship between birthplace and employmentrelated training using national survey data for Australia collected in 1989 which indicate that NESB immigrant workers were indeed much less likely than workers from an English-speaking background (including persons born both inside and outside Australia) to receive training. Moreover, the analysis reported by Baker and Wooden (1991) also revealed that such differences remained statistically significant after taking into account other worker characteristics (such as age, experience, workplace size and occupation) which were also related to training incidence. Unfortunately, Baker and Wooden were unable to control for English-language ability, a factor which

3

is generally considered to be a very important influence on labour market outcomes for NESB immigrant workers.

In this article, we re-visit the issue of NESB immigrant participation in training, but make use of a more recent and richer data set which permits the construction of a control for English-language ability. Logit regression models of the incidence of different types of training are estimated, and the results are used to obtain estimates of the size of any training differential between

NESB

immigrant

workers

and

other

workers.

Further,

decomposition analysis, akin to that used in studies of wage discrimination between men and women, is used to examine the extent to which the observed training differential is attributable to demographic and employment characteristics rather than to employer discrimination.

Theoretical perspectives

Why should NESB immigrant workers be less likely to access training? Two very different theoretical models — the human capital model and the labour market segmentation model — assist in providing insights into why NESB immigrant workers might be less likely to participate in work-related training.

The key assumption underlying the human capital model (see Becker 1964; Mincer 1974) is that investment in training on the part of both the employer and the employee is based on cost-benefit considerations. For employers, the benefits from training accrue in the form of increased productivity. Provision of training, however, involves costs, such as course fees, instructors, additional supervision time and most importantly, foregone

4

output during training. For workers, the main benefits from training come in the form of higher earnings and increased promotion and career opportunities. Participation in training, even when the employer bears the burden of the up-front cost, also involves costs for the employee, most notably in the form of foregone (or lower) earnings while training. 3

As a result of the costs involved in training provision, employers do not randomly provide training opportunities to their workers. Instead, training will be offered to workers when the costs of that training are lower than the long-term benefits the firm expects to derive from providing the training. Expected benefits, in turn, are dependent on perceptions of the effectiveness of the training and likely length of job tenure. In the case of NESB workers, it is usually argued that, as a result of English-language difficulties (and cultural differences), training is less effective and, in turn, more costly. Thus, NESB immigrants are considered to be less attractive training propositions for employers.

With regards to whether or not an individual worker will accept an offer of training, the human capital model suggests that NESB immigrants should be less likely to take up offers of training due to their assessment of the net benefits of such training. Most obviously, the returns to training will be relatively low for workers with language difficulties, which in turn will act as a disincentive to voluntary participation in training.

In contrast to the human capital model with its emphasis on the choices of employers and workers, the second theory — the labour market segmentation model (Doeringer and Piore 1971) — emphasises the characteristics of the job and the job market. In its simplest form, labour

5

market segmentation models distinguish between jobs which fall into the primary labour market and those which fall into the secondary labour market. Primary labour market jobs offer high wages, good working conditions, employment stability and good training opportunities. Secondary labour market jobs, in contrast, tend to offer low wages, few fringe benefits, poor working conditions and few training opportunities. Thus in this theory, the training needs of the job are technologically determined by the job itself — that is, a specific amount of training is considered to be intrinsic to each job (Duncan and Hoffman 1979).

According to this theory, then, in order to gain access to training, one must secure a job in the primary sector. However, workers tend to be assigned to the two different labour market sectors according to the average productivity of the group with which they share common observable characteristics. Thus particular groups will tend to be concentrated in the secondary labour market, and hence presented with few training opportunities, while other groups will be concentrated in the primary labour market where training opportunities are in greater abundance. It is generally agreed that the characteristics of NESB immigrant workers, due to communication problems inhibiting skill formation, have led to their concentration in the secondary labour market. In summary, this theory predicts that NESB immigrant workers will tend to obtain employment in industries, occupations and jobs where training requirements and opportunities are relatively modest.

There are also at least two other reasons why immigrants might receive less training (Baker and Wooden 1991). First, the immigrant selection process may favour immigrants with skills obviating the need for them to

6

receive training. Second, lower levels of training among NESB immigrant workers may result if employers deliberately discriminate against workers from this group. This might occur in the form of being deliberately excluded from the training process, or through reduced promotion prospects which, in turn, will discourage NESB immigrant workers from devoting time and energy to training.

Finally, it is worth noting that it is possible that there are other influences working in the opposite direction to those discussed above and thus that NESB immigrants could receive more, rather than less, training. Most importantly, the self-selection aspect of immigration may mean that immigrants are a relatively highly motivated group who may therefore have a relatively strong desire to enhance existing, and acquire new, skills.

Data and method

Definitions

‘Immigrant’ and ‘training’ are key concepts in this paper, and thus require clarification. We make use of the conventional definition of ‘immigrant’ — persons resident in Australia who were born in an overseas country. Generally, immigrants are classified into two groups — those from nonEnglish-speaking

backgrounds

and

those

from

English-speaking

backgrounds (ESB). Due to a lack of more appropriate data, previous researchers have relied on country of origin data to make this classification, with those born in the main English-speaking countries (i.e., United Kingdom, Ireland, the United States of America, Canada, New Zealand and the Republic of South Africa) classified as coming from an ESB, and

7

immigrants born in all other countries classified as from a NESB. This classification system, however, will tend to misclassify a number of persons. Many persons born in Hong Kong, for example, have English as a first language, while many persons born in Quebec (Canada) do not have English as a first language.

A more accurate method of classification considers whether the first language spoken by the immigrants was English, with those whose first language was not English classified as a NESB immigrant, and others as an ESB immigrant. In this paper, we are able to use this classification system, since our data source, the 1993 Survey of Training and Education (SOTE), provides us with information on first language spoken. Thus, our definitions of NESB and ESB take into account not just birthplace, but also Englishlanguage background (even though we often simply use the term ‘birthplace’ groups in this article).

At the most basic level, ‘training’ can be thought of as any activity which assists individuals to develop, learn and maintain skills related to job performance and competency. This definition thus includes a wide variety of activities ranging from basic schooling through to learning on-the-job. Consequently, most training statistics attempt to distinguish between different types of training. In this study, three major types of training are identified — structured on-the-job (or in-house) training, structured off-thejob (or external) training, and unstructured on-the-job training.

In the SOTE data, ‘structured on-the-job training’ refers to formal training courses which were organised by an employer or business primarily for their own staff, and made use of their own staff or training consultants to

8

provide the training. The types of activities specifically considered were lectures, seminars, tutorials, workshops, conferences, group sessions, audiovisual presentations, demonstration sessions and self-paced courses.

‘Structured off-the-job training’, on the other hand, relates to formal courses organised and conducted by training or educational establishments, agencies or consultants other than the respondent’s employer. Study for an educational qualification was specifically excluded. Note that only training undertaken by individuals while in employment is considered.

Finally, ‘unstructured on-the-job training’ covers informal activities undertaken by a worker to improve job skills. In the SOTE, such training was defined specifically to include being shown how to do the job, watching others work, asking questions of co-workers and teaching self.

Data

The 1993 Survey of Training and Education is the second household survey undertaken by the Australian Bureau of Statistics (ABS) with the primary aim of collecting information on a range of topics relating to the training and education experiences of the Australian work force. 4 As described by the ABS (1994, p. 49), the SOTE was conducted over a six-week period during April and May 1993, and involved a stratified sample of 12 600 dwellings.

The scope of the survey was restricted to persons aged 15 to 64 years (inclusive) who had worked as a wage or salary earner in the previous 12 months, as well as those who, at the time of the survey, were employers, self-employed, unemployed or marginally attached to the labour force. 5

9

However, residents of an overseas country, persons aged 15 to 20 years who were still at school, and members of the permanent defence forces were excluded. Data were collected by personal interview with all persons who fell within the scope of the survey.

Overall, the SOTE provides data on the education and training experiences of 20 889 persons. However, given our interest in work-related training, we only consider those persons who worked in a wage or salary job in the previous 12 months, reducing the sample to 15 644 persons.

Data on training events relate to any training that took place in the 12 months prior to interview. As summarised in Table 1, these data indicate quite clearly that the incidence of all types of training is much lower among NESB immigrant workers than among both Australia-born workers and ESB immigrant workers. Moreover, simple non-parametric tests indicate that these differences across birthplace groups are highly significant.

— INSERT TABLE 1 ABOUT HERE —

However, and as shown in previous research (e.g., Duncan and Hoffman 1979; Greenhalgh and Stewart 1987; Baker & Wooden 1991, 1992; Booth 1991, 1993; Lillard and Tan 1992; Lynch 1992; Green 1993; Miller 1994), a variety of other factors, both on the part of workers (e.g., age, education, family status) and on the part of firms (e.g., sector, industry, workplace size), are also likely to be associated with participation in training. This means that if workers from the three birthplace groups differ with regard to these factors, these differences alone may explain at least some of the variation in participation in training. Thus, it is necessary to take existing

10

differences in other factors into account in order to gain a fuller understanding of the factors that explain why NESB immigrant workers have relatively low rates of participation in training. This is the aim of the remainder of this article.

Multivariate analysis

The approach used in our analyses closely follows that of the earlier Australian studies into training participation undertaken by Baker and Wooden (1991, 1992) and subsequently Miller (1994). In particular, these studies utilised probit or logit models (both of which are specifically designed for the estimation of models where the dependent variable can assume only discrete, rather than continuous, values) to estimate reducedform equations explaining the incidence of various types of training. In this study, binary logit methods are used to estimate the incidence of both types of on-the-job training. In the case of off-the-job training, however, multinomial logit methods are used since we allow for three possible outcomes: no participation, participation with employer support (for at least some of the training) and participation without employer support. Unlike the other two forms of training considered, participation in structured off-thejob training is not necessarily contingent on employer support (since workers can attend these courses without any funding from their employer or leave from work). The distinction between whether or not the training was supported by an employer may be of large importance in determining whether immigrants are disadvantaged in the training process. Thus, all three possible outcomes are considered.

11

Previous research has suggested that the effects of the independent variables on the incidence of training differ for men and women (Miller 1994). A test of our data provides support for this conclusion. 6 Thus, all analyses are undertaken separately for men and women.

The explanatory variables included in the analyses also closely resemble those included in the work of both Baker and Wooden (1991, 1992) and Miller (1994). However, compared with the data used in these earlier studies, the SOTE gathered data on a number of important additional variables. In particular, we are able to include measures of union membership and the extent of English-language difficulties.

The variables considered in our analysis can be clustered into six major groups, as follows: 7 i)

factors associated with the expected length of the employment relationship — age and employment status (i.e., permanent versus casual employee);

ii)

skills — proxied by education and work experience;

iii)

training requirements of the job — proxied by occupation and whether the worker began his/her job during the survey reference period;

iv)

firm characteristics — firm size, sector, and industry;

v)

union membership; and

vi)

other demographic characteristics — marital status, age and presence of dependent children, birthplace and language background, extent of any English-language difficulties, and location of residence.

12

A summary of the data and the chief variables considered in the analyses is provided in Table 2. Note that with the exceptions of years of tenure, years of occupational experience and weeks of employment, all variables are discrete rather than continuous, and hence enter the analysis in binary form. Thus, for example, when controlling for ‘sector’ we simply included a variable which took the value ‘1’ if the worker’s job was in the public sector and ‘0’ if it was in the private sector.

Given their binary nature, the construction of most variables is relatively straightforward and requires little explanation. Nevertheless, three variables require some comment. First, the measure of English-language difficulties was constructed from information provided by the interviewer. Specifically, if the interviewer indicated that the interview was completed in English but with difficulty, conducted in another language or conducted by proxy for language reasons, the worker was coded as having English-language difficulties.

Second, we do not classify all Australia-born persons into one group in our analyses. Instead, we distinguish between those Australia-born workers whose first language was English and those whose first language was not English. 8 By making this distinction, we can more clearly identify whether immigrant status per se makes a difference in training participation or whether a non-English-speaking background is the more critical characteristic.

The third variable requiring further explanation is occupational experience. This variable is a measure of a person’s total years of experience

13

in the occupation of their job. It does not refer to years in that occupation with any one specific employer.

It is also worth noting that we tested for interaction effects between groups of key variables. As a result, the final model specification includes variables capturing interaction effects between birthplace and Englishlanguage difficulties, between educational attainment and where the highest level of education was completed, and between employment status (permanent versus casual employee) and hours of work variables. In all of the models tested, inclusion of these interaction effects improved the overall fit.

One problem which arises from the nature of the SOTE data for those who had more than one job in the previous year concerns the linking of information on participation in training (which relates to all jobs) with job characteristics, which only relate to the job in which the worker spent most time during the year (referred to as their main job). We therefore excluded from our analyses those observations where it was unclear whether the training received related to the individual’s main job. 9

Results

To keep technical detail to a minimum, we have chosen not to report the full logit results from the estimation of our preferred model for participation in each of the different types of training, and instead concentrate on describing the implications of our results. 10 Overall, the model performed extremely well and, more importantly, the estimated signs on the coefficients are generally in line with a priori expectations and previous research. For

14

example, in the case of structured on-the-job training, we found significant relationships with age for men (an inverted U-shaped relationship is found whereby training is lowest at both younger and older ages) but not for women. The absence of any obvious relationship between training incidence and age for women probably reflects their greater likelihood of career interruptions. Other variables found to vary significantly with the incidence of structured on-the-job training include educational attainment (a positive relationship), occupation (with training incidence lowest for workers in unskilled and semi-skilled blue-collar occupations, and highest among managers and administrators and professionals), employment status (casual workers, especially part-time casuals, participated in less training), firm size (workers in large firms are more likely to receive training), union membership (union members participate in more training, though the relationship is only significant for women), weeks of employment during the reference year (positive) and place of residence (metropolitan residents have a lower incidence of structured on-the-job training). This pattern of results is replicated, to varying degrees, in the models explaining the other types of training, with perhaps the most important difference being that in the case of unstructured on-the-job training, the effects of tenure and experience are of much greater importance.

Turning to the variables at the centre of interest in this study — those relating to birthplace and immigrant status — the results suggest that, in the main, NESB immigrant workers are significantly less likely to participate in training, even after controlling for a wide range of worker and job characteristics. Further, we found that the English-language skills of these immigrants is very important in determining the likelihood of training participation. In particular, we found that English-language skills interact

15

with background such that persons from a NESB who also have poor English-language skills are especially disadvantaged in terms of access to training.

In the case of structured on-the-job training, NESB immigrant men with English-language difficulties are estimated to have a probability of participation which is 50 per cent that of an otherwise comparable Australiaborn man whose first language was English. For NESB immigrant men without English-language difficulties, the estimated coefficient also implies a negative effect, but was not large enough to enable us to reject the hypothesis of no relationship. In contrast, among women, NESB immigrants both with and without English-language difficulties fare much worse than Australia-born women. Those with English-language difficulties are estimated to have a probability of attending structured on-the-job training which is 45 per cent that of Australia-born women, while for those without English-language difficulties, the estimated ratio is 63 per cent.

A similar pattern of results is also evident in the case of unstructured onthe-job training, with the slight difference that it is only those NESB immigrants

with

English-language difficulties

who appear to be

disadvantaged. The estimated coefficients imply that NESB immigrants who had problems with the English language have a probability of receiving unstructured training which is between 58 and 63 per cent (for men and women, respectively) that of Australia-born workers with similar characteristics.

In the case of structured off-the-job training, the results are less straightforward. The estimated coefficients imply again that NESB

16

immigrants with English-language difficulties have an extremely small probability of attending off-the-job training with employer support. However, as a direct result of the very low incidence of such training in this group, extremely high standard errors are attached to these estimates, and hence these results are not statistically significant. What we can be confident about, however, is that all other things equal, NESB immigrants as a group are far less likely to attend structured off-the-job training courses. Indeed, re-estimation of our model after removing the English-language distinction indicates that NESB immigrant men have a probability of attending structured off-the-job training with employer support which is 46 per cent of Australia-born men and a probability of attending structured off-the-job training without such support which is 53 per cent of Australia-born men. For NESB immigrant women, the respective probabilities are 49 and 48 per cent.

As noted earlier, an innovative feature of our analyses is the ability to identify persons born in Australia whose first language is not English (presumably comprising some Aborigines born in remote areas as well as some second-generation immigrants). The regression results indicate that this group of workers has rates of participation in the various types of training which are almost as low as those of NESB immigrants with English-language difficulties. Only in the case of men attending structured off-the-job training courses (either supported or non-supported) was this not the case.

In contrast to the disadvantaged position of both NESB immigrants and Australia-born workers whose first language was not English, ESB immigrant workers fared much better. In line with previous research, ESB

17

immigrants generally appear no less likely than Australia-born workers to participate in training. Indeed, in the case of structured on-the-job training, male ESB immigrants are found to be significantly more likely (24% more likely) than Australia-born workers to have participated.

Another interesting finding that has implications for the extent of immigrant disadvantage in training relates to the education variables. Previous studies have typically found that the incidence of training rises with educational attainment. Our results, however, indicate that this is only true of education completed in Australia. Education which is completed overseas appears to have very little augmenting effect on the likelihood of participating in any of the three types of employment-based training. In the case of structured on-the-job training, for example, the probability that a man with a university degree from an overseas institution attended a structured on-the-job training course is estimated to be 1.5 times greater than that of a man who attended school in Australia but left before the age of 16. In contrast, the enhanced probability associated with a degree obtained in Australia is 2.5. Further, while trade qualifications, other post-school qualifications and completion of secondary school are all associated with much enhanced probabilities of attending training courses if those qualifications were obtained in Australia, the same is not true of similar levels of qualifications obtained overseas.

Such findings are very much in line with research into other labour market outcomes which point to low returns to education received overseas among the NESB immigrant population (see Wooden 1994). The explanation for a positive correlation between education and training is usually said to lie either in a greater interest in training on the part of more

18

educated workers or in a greater ability to learn, ‘implying both a willingness on the part of the employee to undergo training and lower costs associated with the training of those persons’ (Baker and Wooden 1992, p. 33). The results discussed here suggest that it is this relationship between education and the cost of training which is probably most relevant. That is, there is not the degree of complementarity between an overseas education and post-school training that there is between an education acquired in Australia and post-school training.

Overall, it would seem clear that NESB immigrant workers are at a distinct disadvantage in terms of access to employment-based training. Further, a major source of this disadvantage seems to lies in their different labour market endowments, especially the greater incidence of Englishlanguage difficulties, but also the lack of pre-requisite qualifications obtained in Australia, and the somewhat greater concentration of NESB immigrants in low training occupations and industries. Nevertheless, we still find, at least in the case of structured training, that NESB immigrants are less likely to participate in training (all other things constant). This unexplained differential may reflect a greater propensity on the part of employers to train Australia-born workers in preference to NESB immigrants; i.e., they may discriminate against NESB immigrant workers purely on the grounds of ethnicity.

In order to provide estimates of the relative magnitude of these ‘endowment’ and ‘discrimination’ effects, the regression analyses were repeated separately for the Australia-born (including those whose first language was not English) and NESB immigrant groups, and the results from these analyses were used to calculate average predicted probabilities of

19

training for the two groups. The differences in these probabilities were then decomposed into two parts, with the first part indicating differences in the mean endowments of Australia-born workers and NESB immigrant workers (i.e., demographic and job-related differences). The second part is the residual component which reflects ‘unjustified’ differences in training incidence among NESB immigrant and Australia-born workers, which in turn may reflect employer discrimination. 11 Results of the decomposition analyses for both men and women, and for each type of training, are reported in Table 3. 12

Table 3 shows that, for both men and women, it is structured on-the-job training where the absolute size of the training differential between NESB immigrant and Australia-born workers is greatest. The difference is especially large for women, where we see a 21.9 percentage point difference in the probability in favour of the Australia-born workers. More importantly, for both men and women, the major source of differences is the unjustified or discriminatory component, accounting for 74.8 and 66.2 per cent of the differential for men and women, respectively. Such results suggest that employers favour Australia-born workers when offering training places. This is not to say that differences in worker and job characteristics are not unimportant, especially English-language ability, but they are dwarfed by the size of the unjustified, and possibly discriminatory, component. A similar conclusion arises from an examination of the results of decomposing the structured off-the-job training differential. Indeed, in the case of employer-supported off-the-job training, differences in mean endowments account for just 16 per cent of the differential for men and less than 10 per cent of the differential for women.

20

In contrast, all of the estimated unstructured on-the-job training differential is the result of differences in endowments. Indeed, among women, the results suggest that NESB immigrants receive substantially more training than their labour market characteristics would warrant.

The foregoing would thus seem to provide strong evidence of discrimination in the provision by employers of structured training. Unfortunately, we cannot totally discount other possible explanations. Most importantly, the unjustified component may not be due to employer discrimination but instead result from unmeasured differences in other variables, and especially English-language ability. The measure we have of English-language ability is a measure of an individual’s ability to respond to interviewer questions. It is unclear how well this will be correlated with an individual’s ability to deal with structured learning environments.

Conclusion

Part of the explanation for the lower incidence of training among NESB immigrant workers lies in the fact that they are more likely to have characteristics that are associated with lower probabilities of training. For example, NESB immigrants were found to be much more likely than other employees to be employed in jobs as labourers or as plant and machine operators, and these occupations were found to be consistently and strongly related to a lower likelihood of training for all of the training types. Further, compared with Australia-born employees, NESB immigrant employees tend to be older, and the multivariate results indicated that older workers were less likely to participate in unstructured on-the-job training. As well, NESB immigrants were more likely than other employees to have gained their

21

education overseas. This was found to be especially disadvantageous, suggesting a lesser complementarity between overseas education and postschool training. Possible reasons for this include lower quality education and country specificity of overseas schooling.

Nevertheless, even with factors such as those mentioned above taken into account, the results indicate that there is still a significant effect of birthplace status on participation in training. That is, with all other factors held constant, NESB immigrants were less likely than English-speakingbackground employees to have participated in training. In most cases, however, this result was found to be critically linked to English-language proficiency. That is, it was among those NESB immigrant workers with English-language difficulties that the training deficit was greatest. Indeed, with respect to some types of training (especially unstructured on-the-job training, but also structured on-the-job training among men), those NESB immigrant workers with good English-language skills were not any less likely than Australia-born workers to have participated.

It cannot, however, be concluded that the acquisition of appropriate English-language skills will necessarily redress the problem of low levels of participation among NESB immigrant workers. Other evidence was uncovered which suggests that with respect to structured training (but not unstructured on-the-job training), the major source of disadvantage NESB immigrant workers face in accessing training lies in factors which cannot be justified within the framework of the regression models. This is often taken to indicate the presence of discriminatory forces. Unfortunately, this result might also reflect inadequacies in the data used to estimate our models. In particular, it is very possible that the results are affected by unmeasured

22

differences in English-language ability. Thus, while we cannot provide ‘hard’ evidence for employer-discrimination in training provision, we cannot discount it as a possibility either.

Finally, irrespective of the extent to which discrimination is responsible for the observed training differential, the results presented suggest that there are good reasons to be concerned about training opportunities for NESB immigrant workers in Australia. While the Federal government of Australia has committed itself to improving training opportunities for disadvantaged groups, including NESB immigrants, as part of its National Training Reform Agenda, how effective this Agenda will be in achieving this objective is far from certain. Furthermore, recent industrial reforms favouring a more decentralised enterprise-based bargaining system may make realising this goal even more difficult. Such conclusions, however, are highly speculative. At the time of the collection of the data analysed in this article, both the NTRA and enterprise-based bargaining were in their early stages of development and, in all likelihood, had affected only a small minority of Australian workplaces. Monitoring the impact of these reforms on access to training, and especially by immigrants, remains a research task for the future.

23

Notes

1

For a critical review of the National Training Reform Agenda, see the contributions to the policy forum published in The Australian Economic Review, No. 110, April–June, 1995. Other significant policy initiatives designed to lift the level of training provision include the Australian Traineeship System (a scheme designed to introduce apprentice-style training in non-trades areas), the now suspended Training Guarantee (a levy imposed on firms that did not meet minimum training expenditure levels), and incentives for unions and firms to incorporate training provisions into industrial agreements (and awards).

2

For overviews of the development and spread of enterprise bargaining in Australia, see Macklin, Goodwin and Docherty (1993), and Chapter 2 of the 1994 Annual Report on Enterprise Bargaining (Department of Industrial Relations 1995).

3

This does not mean that employee pay is reduced pay when attending training courses. Rather it assumes that employers take into account the probability of employees undertaking training when setting wage levels.

4

The first survey was conducted in 1989 and provided the data at the centre of the analysis by Baker and Wooden (1991).

5

The marginally attached are defined by the ABS as those persons not in the labour force but who wanted to work and who were either: (i) actively looking for work but did not meet the criteria for classification as unemployed; or (ii) were not actively looking for

24

work but were available to start work within four weeks (or could do so if child care was available). 6

The relevant chi-squared statistics for the test that the coefficients for men and women are the same were 145.2 for structured on-the-job training, 185.6 for structured off-the-job training, and 136.6 for unstructured on-the-job training. These are all highly significant, indicating the need to analyse the data for men and women separately.

7

In addition, since the likelihood of receiving training within the oneyear reference period will be closely related to the length of employment during that period, and since the sample includes persons who will not have been continuously employed throughout the year, we include a variable measuring the number of weeks employed during the survey reference period.

8

Just over three per cent of the population of Australia-born employees indicated that their first language was not English.

9

Specifically, with respect to structured on-the-job training, we excluded those employees for whom none of their involvement in such training occurred in their main job. This resulted in the removal of 182 observations (or 1.2 per cent of the sample). In the case of employersupported structured off-the-job training, those workers who participated in employer-supported training courses but without any support from the employer in their main job were excluded. In this instance, just 26 cases were lost. In the case of unstructured on-the-job training, no additional information was available on whether or not this form of training was undertaken in their main job but given the very high incidence of such training, we thought it appropriate to retain all observations.

25

10

For those interested, however, the results are available, on request, from the authors.

11

This unjustified component might also reflect other uncontrolled differences in worker preferences. However, as noted earlier, we might expect immigrants to be relatively highly motivated individuals and hence have stronger (rather than weaker) preferences for training than other workers. The only available evidence on this issue is consistent with this hypothesis (Stephens and Bertone 1995).

12

The decomposition method used is analogous to that proposed by Farber (1990) and used in Miller (1994). For a detailed explanation of the method, see Miller (1994, p. 553).