Job-Related Training Activity by Immigrants to Canada

Transcription

1 Job-Related Training Activity by Immigrants to Canada Job-Related Training Activity by Immigrants to Canada 469 DEREK HUM WAYNE SIMPSON Department of Economics University of Manitoba Winnipeg Manitoba L Enquête de 1998 sur l Education et la Formation des Adultes identifie pour la première fois les immigrants et est utilisée pour comparer la formation des immigrants avec celle des Canadiens de naissance. Des recherches antérieures indiquent que les immigrants n acquièrent généralement pas, après leur arrivée, le même degré de formation que les Canadiens. En outre, si le capital humain étranger a une valeur moindre sur le marché du travail du pays d accueil, la formation sera limitée pour les immigrants les plus âgés. Nous constatons que la formation se réduit environ d un an par année de retard de la migration, tant pour les hommes que pour les femmes, à partir d échantillons d immigrants, considérés en groupes ou séparément, et de Canadiens de naissance. Les immigrants qui arrivent au Canada déjà adultes reçoivent moins de formation que ceux qui arrivent très jeunes, alors que les immigrants qui arrivent encore enfants font à peu près aussi bien que les Canadiens de souche. Il se peut que le désavantage en matière de formation s explique en partie par les contraintes financières, mais nous rejetons d autres explications telles que la langue qui sont communément avancées. The 1998 Adult Education and Training Survey (AETS) identifies immigrants for the first time and is used to compare the training experiences of immigrants and native-born Canadians. Previous Canadian research indicates that immigrants generally acquire less human capital after arrival than the native-born. Further, if foreign human capital has reduced value in the host labour market, training will be limited for older migrants. We find that training is reduced by about one year for each year that migration is delayed for both men and women in both pooled and separate samples of immigrants and the native-born. Immigrants who arrive in Canada as adults train less than those who arrive as children, while immigrants who arrive as children do about as well as the native-born. Financial constraints may explain some of the training disadvantage, but other common explanations, such as language, are rejected. INTRODUCTION Aperennial policy issue concerns the economic outcomes of immigrants to Canada. What problems do they face in the job market? Do they undertake job training to acquire Canadian credentials? If not, what are some of the barriers, and what policies might Canada adopt to upgrade the job skills of immigrants? Are the labour market experiences of current immigrants similar to those who have come in the past? The ease with which immigrants integrate into the Canadian labour market is especially important in deciding the number of immigrants to admit, since the economy s capacity to welcome more newcomers depends on the ease and speed with which immigrants achieve success.

2 470 Derek Hum and Wayne Simpson Immigrants to Canada experience a disruption to their job history upon arrival. Canadian studies employing cross-section data invariably find that immigrants face an initial earnings disadvantage at entry relative to native-born workers. Many studies also suggest that immigrants eventually catch up to the native-born (Abbott and Beach 1993; Baker and Benjamin 1994 and 1997; Bloom, Grenier and Gunderson 1995; Grant 1999; Hum and Simpson 1999; Li 2001; McDonald and Worswick 1998; Meng 1987; Miller 1992). This appears to be the common view, and no doubt influences policy stance affecting immigration. Although such an integration scenario implies faster earnings growth for immigrants than native-born workers over some period after arrival, Hum and Simpson (2000) find no evidence of faster growth in recent panel data from the Survey of Labour and Income Dynamics (SLID). They conclude that economic assimilation may now be slower than past studies using cross-sectional evidence imply. Instead, they find that human capital accumulation (additional years of schooling and work experience) has a significant impact on wage growth. Job-related training is known to affect wage progression, raising the question whether recent immigrants, whatever their level of human capital upon arrival, undertake as much training as Canadians. If they do not, can we determine why they do not? There is, however, little research on differences in training activity between immigrants and the native-born, particularly job-related training beyond formal schooling. This paper focuses on the training activity of immigrant and native-born Canadians beyond formal schooling, which we call postschool training. Although anecdotal evidence abounds that immigrants are denied training opportunities in the labour market, there has been no formal analysis of the issue. This omission is significant since a clear link has been established between work-related training, earnings growth, productivity growth, and job performance in numerous studies (e.g., Barron, Black and Loewenstein 1989; Bartel 1992; Lynch 1992). The Adult Education and Training Survey (AETS) contains information on training, employment, and individual characteristics and circumstances. Furthermore, the AETS distinguishes immigrants for the first time in 1998, permitting us to compare the postschool training activity of immigrants and nonimmigrants. The next section sets out an economic model of training activity for immigrants and reviews the sparse international evidence on this question. We then describe the AETS data and develop an empirical approach in the context of the larger literature explaining training activity. The penultimate section presents our empirical results on the training activity of immigrants. The final section reports evidence on perceived barriers to training for immigrants. POSTSCHOOL TRAINING AND IMMIGRATION The human capital model explains postschool training activity as one of a series of investment decisions made by an individual over the life cycle (e.g., Blinder and Weiss 1976; Weiss 1986; Polachek and Siebert 1993). This model can provide a basis for examining immigrant training activity. For a given level of formal schooling, an individual will subsequently invest in additional training until the marginal benefits no longer exceed marginal costs, including foregone earnings. The marginal benefits of training include the discounted incremental earnings stream from the skills acquired, which declines with age. At some point near anticipated retirement, the future earnings stream will be sufficiently short that virtually no training will be justified, yielding a characteristic pattern of declining training activity with age. Those with greater ability would be expected to invest more, while those with higher borrowing costs would be expected to invest less. Mincer and Ofek (1982, 19) note that, if selective migration were to yield more able workers, one might expect greater investment in human capital among immigrants than among the native-born, other factors (e.g., age and borrowing costs) being constant.

3 Job-Related Training Activity by Immigrants to Canada 471 A second feature is that current training cost depends on past investments. Those who invest more in formal schooling receive more postschool training at any age because it is cheaper, which provides an explanation for the steeper earnings growth experienced by more educated workers. Thus, a pattern of higher training activity for workers possessing a greater stock of human capital is a second crucial prediction of the model. We argue that this feature may be particularly relevant in the case of immigrants, who arrive at all ages. Any immigrant will face some degree of disruption associated with differences in language, culture, and labour market networks. For those who immigrate at a young age, these disruptions are easily overcome. Immigrant children, for example, face some adjustments during the period of formal schooling but are likely to leave the education system with skills comparable to the native-born. For older immigrants, however, adjustment may be more difficult. For example, there are well-documented problems associated with the recognition of foreign experience and adjustment to a foreign work culture for those who immigrate in mid-career. If foreign schooling and training has reduced value in the host country job market, this increases the marginal cost of training at any age because these costs depend on the perceived human capital stock. The higher marginal training costs for immigrants will reduce the amount of training activity, leading us to predict a negative correlation between age at migration and training for immigrants compared to native-born workers. Mincer and Ofek (1982, 19) suggest that the interruption associated with migration may cause immigrants to train less insofar as intermittent workers are lesser lifetime investors [in human capital] than continuous workers, although they do not test this hypothesis directly. Rather, they point to the apparently rapid integration of immigrants as evidence that other factors are involved. There is little direct evidence on the training activity of immigrants. Although a large literature exists on adult training or on immigrant outcomes, the two topics have been combined, to our knowledge, only in two Australian studies. Miller (1994) finds that immigrants from non-english-speaking countries train less than either natives or immigrants from Englishspeaking countries and that foreign qualifications tend to reduce the incidence of training relative to comparable domestic qualifications, particularly for males. Moreover, immigrants from non-english-speaking countries get far less in-house training, even though in-house training appears to be unrelated to the source of qualifications. These results are consistent with the idea that foreign qualifications are discounted in the host country, but also with the idea that other disruptions related to language fluency may reduce training activity among those who immigrate as adults. A paper by Kennedy et al. (1994) finds corroborative evidence that Australian workplaces having a higher proportion of workers from a non-english-speaking background engage in less in-house training. Other studies focus on post-migration investment by immigrants in formal education, which is available in far more data sets. Borjas (1982) found that Cubans assimilate faster than other Hispanics. His explanation is that Cubans invest more in education in the United States soon after arrival, but his data does not identify foreign students separately from other migrants and does not directly measure years of post-migration education. Chiswick and Miller (1994), using Australian data that records post-migration education, find that age at migration has a strong negative effect. Recent cohorts are also much less likely to have obtained qualifications, which Chiswick and Miller interpret as a positive effect related to years since migration. Moreover, Chiswick and Miller find that both pre-migration qualifications and high occupational status increase the probability of obtaining qualifications in Australia. This suggests that foreign human capital complements domestic human capital, counter to Borjas finding. In summary, there is some evidence suggesting that immigrants may experience problems acquiring

4 472 Derek Hum and Wayne Simpson work-related training, particularly those with language problems in the host country. These problems may in turn account for some of the difficulties immigrants experience in achieving earnings parity over time with their native-born counterparts. We also argue that these problems will depend on the age at which migration occurs, particularly whether migration occurs in childhood, prior to entry into the labour market, or as adults. We investigate these issues below. OUR EMPIRICAL APPROACH The Adult Education and Training Survey (AETS) is a comprehensive cross-sectional survey on adult participation in formal education and training in Canada. The 1998 AETS was conducted as a supplement to the Labour Force Survey of January 1998 to identify education and training during It covers learning obtained through programs as well as courses, learning which is job-related as well as taken for personal interest, learning undertaken fulltime as well as part-time, and learning that is employer-sponsored as well as non-employer sponsored. The survey captures training in universities and colleges, private and commercial institutions, and on-site at the workplace. It also captures distance learning, such as Internet courses. We classify training as work related if it is either employersponsored or judged by the respondent to be taken for job or career reasons. For those who received any training, the survey records the total duration of all training taken during the reference year, which we adopt as our primary measure of training activity. In this study we exclude full-time and part-time students to focus on work-related training. Further, our analysis is limited to those who were employed in 1997, which automatically excludes the most recent immigrants; that is, any AETS respondent who might have immigrated in January 1998 or during 1997 but who would not yet have begun employment. 1 Associated with the details on training activity is a variety of useful information on personal and job characteristics, which we discuss in some detail below. In particular, the 1998 AETS asks questions on immigrant status for the first time and can be used to assess whether training experiences differ between the native and foreign born. Limited information is also gathered about such matters as situational and institutional barriers to participation, the subject matter of courses taken, and respondents perceptions about the course. Table 1 indicates the extent of training by immigrant status for men and women employed in We identify those who immigrate as adults (18 years of age or older) as a group that may have more limited training opportunities. For employed men, the results generally indicate that men and women who are native-born participated in postschool training at a higher rate than immigrants and that the gap was greater for those who immigrated as adults. Consider first the fourth row of Table 1, which estimates participation in training among workers. While 32.7 percent of native-born men and 39.6 percent of native-born women participated in training in 1997, only 29.4 percent of immigrant men and 32.7 percent of immigrant women participated in training. Participation declines further to 23 percent for men and 28.5 percent for women when only those migrating as adults are considered. The pattern for work-related training in the sixth row is similar: 27.8 percent of native-born men and 30.8 percent of native-born women participated in workrelated training in 1997, compared to 23.2 percent of immigrant men and 26.3 percent of immigrant women and, for those who immigrated as adults, only 18.5 percent of men and 22.8 percent of women. Clearly, most training reported by employed respondents is work-related and the general patterns by immigrant status are similar. It is more interesting to consider training duration, represented by average hours of any training. (See the eighth row of Table 1.) Training duration encompasses both participation rates and the intensity of training activity among those participating. For men, the pattern is comparable to the one for

5 Job-Related Training Activity by Immigrants to Canada 473 TABLE 1 Training Activity Among Employed Men and Women, by Immigrant Status MEN WOMEN Immigrants Immigrated Native- Immigrants Immigrated Nativeas Adults 2 Born as Adults 2 Born 1 Pop. est (,000) 1 1,809 1,246 7,470 1,908 1,361 7,737 2 Employed (,000) 1, ,676 1, ,612 3 Any training (,000) , Any training (%) Work-related training (,000) , ,420 6 Work-related training (%) Hours any training (,000) 3 34,440 18, ,800 51,870 24, ,100 8 Average hours any training Hours work-related training (,000) 28,710 14, ,200 46,040 22, , Average hours work-related training Notes: 1 Sample results weighted to reflect Canadian adult population excluding full-time and part-time students. 2 Defined as immigrants 18 years of age or older. 3 Sample size is slightly smaller than listed because of a small number of missing values for training duration Source: 1998 Adult Education and Training Survey Master (Internal) File. Calculations by the authors. participation. The average duration of training is 44.1 hours per year for native-born men, but only 27.5 hours for immigrant men and 22.9 hours for men who immigrate as adults. For work-related training (see the last row of Table 1), the average duration was 36.7 hours for native-born men compared to 23 hours for immigrants and 17.8 hours for men who immigrate as adults. In other words, age at immigration appears to matter a great deal: men who immigrate as adults receive only about half as much training, work-related or otherwise, as those born in Canada. For women, the gap in training between the nativeand foreign-born is less apparent. Native-born women trained 43.6 hours on average in 1997 compared to 51.1 hours for immigrant women and 37.6 hours for women who immigrated as adults. For work-related training, native-born women trained 35.3 hours as opposed to 45.6 hours for all immigrant women and 34.1 hours for women who immigrated as adults. Thus, there is evidence of fewer training opportunities for women who immigrate as adults, but it is clearly weaker than for men in these simple aggregate comparisons. But do these figures stand up to more careful scrutiny? Can we account for these differences in training activity in other ways? We now develop a multivariate model of training duration to assess more carefully the relative performance of immigrants. As noted previously, a pattern of training activity that declines with age, while generally increasing with schooling and previous training, has been

6 474 Derek Hum and Wayne Simpson widely used to explain observed life-cycle patterns of hours worked and earnings. This represents an important starting point for any study of training activity. The AETS does not capture previous training activity, but it does capture the highest level of schooling completed which, insofar as education affects the subsequent level of postschool training activity, serves as a useful proxy for all past education and training. In addition, Blinder and Weiss (1976) develop a model that explains both hours worked and human capital accumulation over the life cycle. Thus, any model seeking to explain differences in postschool training activity needs first to account for differences across individuals in formal schooling completed, age, and hours worked (or the extent of labour force attachment). A fourth factor is job tenure. Much training contains both job-specific and general components. Therefore, time elapsed within the current job may be an important determinant of training activity, since the incentives to invest are greater near the beginning of a job for both the worker and the firm (Borjas 1975, 1981). One might expect training to decline throughout the tenure of the current job, but Borjas argues that firms may defer offering training until a worker s permanence with the firm can be assessed. Thus, the pattern of actual training activity with lengthening job tenure may be complex and non-linear. In our view, schooling (and past training, if it were available), age, hours worked, and job tenure constitute a set of core economic determinants of training decisions at any point in time. We are surprised to find how few studies of adult training include this set of variables. In particular, we would note that simple correlations of training activity and individual characteristics are misleading because the human capital approach, stripped to its bare essentials, establishes the case for a multivariate analysis of the factors accounting for training. Even among multivariate analyses of training decisions in which age and education are typically included, factors such as job tenure are often ignored. Moreover, many studies of training include other explanatory variables that appear to influence the training decision. Exclusion of these variables could bias estimates and perhaps provide misleading comparisons of training opportunities for native and foreign-born workers. Given the heterogeneity of individuals and jobs in a microdata set like the AETS, this would not be surprising. To guard against potential specification bias, we adopt an encompassing regression approach in which we admit both our core set of economic determinants and a wide variety of other factors related to personal or job heterogeneity that have been cited in the literature. This allows us to determine whether there is a robust relationship between immigration status, age at immigration, and training activity. Accordingly, we adopt the duration of any training activity as a proxy for the proportion of time spent training during the year. If we let y represent the observed duration of training with a limiting value of zero (no training during the reference year), then our model of adult training accumulation may be interpreted as the unobserved propensity for training y * for a particular individual in the AETS sample. We then have a standard latent variable model of training of the form y= y * if y * > 0 * 0 if y 0 [1] where the propensity for training depends on the core economic variables and, insofar as training decisions are shared between workers and employers, on a variety of personal and job characteristics that we take from our assessment below of the literature on adult training. The model may be estimated directly using the maximum likelihood (tobit) technique where, as in this case, there is no good reason to believe that the factors determining participation and the duration of training for participants differ. In terms of personal characteristics that affect adult training, many studies find that the determinants of

7 Job-Related Training Activity by Immigrants to Canada 475 training differ between men and women (Green 1991, 1993; Miller 1994; Wooden and VandenHeuval 1997; Green and Zanchi 1997). Greenhalgh and Stewart (1987) find that women receive less training than men early in their careers, and that married women continue to receive less training throughout their careers, implying a marital status effect as well. Holtmann and Idson (1991), on the other hand, find no significant marital status effect. Greenhalgh and Stewart also find that the presence of young children reduces training participation for both men and women, but Miller (1994) finds that very young children (0-2 years) reduce training only for women. Duncan and Hoffman (1979) find that the expectation of having children reduces the chances of training for women. In our analysis, we estimate separate models for men and women that take account of marital status and the presence of young children. Many studies examine the role of occupation in training activity. Holtmann and Idson (1991) find that white collar workers are more likely to receive training, while Green (1993) and Miller (1994) argue that it is occupational status that matters. Altonji and Spletzer (1991) find that training incidence increases with the verbal, math, and clerical skill requirements of an occupation and decreases with manual skill requirements. As in many other studies, however, no allowance is made for such crucial variables as age, previous education and training, or job tenure. We include various measures of occupational status available in the AETS in our analysis. 2 With respect to job characteristics, studies have repeatedly found that larger firms train more (Barron, Black and Loewenstein, 1987; Booth 1991; Holtmann and Idson 1991; Green 1993; Miller 1994; Simpson 1984). Holtmann and Idson argue that larger firms have a greater incentive to train because they can pool training risks. Mincer (1981, 1984) finds that non-union firms train less in the US but argues that compressed wage schedules and the dominance of seniority rules for wage progression in unionized firms may reduce incentives for general training. Canadian evidence, on the other hand, has found no significant training effect based on union status (Betcherman, Leckie and McMullen 1997; Simpson 1984). Kennedy et al. (1994) find that unionization per se does not affect training in Australia, but that a unique variable representing union bargaining activity or voice appears to be positively correlated with training. Our analysis includes measures to capture firm size and union status. Industrial and location factors may also affect training, and we capture these elements in our study. Lillard and Tan (1986) find that workers in industries undergoing rapid technological change receive more training, particularly more educated workers, but Booth (1991) finds no significant effect for industrial categories for women. Lynch and Black (1998) find that workers in manufacturing firms with high capital-labour ratios, and workers in nonmanufacturing firms with high research and development activity, conduct more training. A number of studies have found regional factors to be significant, including Betcherman, Leckie and McMullen (1997), Booth (1991), Lillard and Tan (1986), and Simpson (1984). DO IMMIGRANTS HAVE COMPARABLE TRAINING OPPORTUNITIES? Table 2 provides descriptive statistics for men and women by immigrant status for the variables used in our analysis. Our dependent variable is the duration of work-related training, as a measure of the proportion of time spent in training activity during the reference period. As reported in Table 1, nativeborn men acquire an average of 13.7 hours, or 60 percent, more training than foreign-born men. The gap is reversed for women, where foreign-born women receive 10.3 hours, or 29 percent, more work-related training. What factors might explain these differences between immigrant and nativeborn men and women?

8 476 Derek Hum and Wayne Simpson TABLE 2 Descriptive Statistics for Workers in 1998 AETS by Sex and Immigrant Status MEN WOMEN Variable All Immigrants Native-Born All Immigrants Native-Born (Values are percentages unless indicated) Duration of work-related training (hrs.) Immigrant Immigrant after age Education No high school Some high school High school degree Postsecondary diploma University Bachelor s degree Graduate degree Age Under 20 years years years years years years years and over Mean age (yrs.) Hours worked per week Job tenure Under 6 months months 1 year years years years Over 20 years Mean job tenure (mos.) Region Atlantic Quebec Ontario Prairies Alberta British Columbia continued

9 Job-Related Training Activity by Immigrants to Canada 477 TABLE 2 (Continued) Urban French* Ethnic origin Black Aboriginal Disabled Married or common law Married previously Preschool children Full-time job Permanent job Changed jobs during year Self-employed Union coverage Firm size 500 or more employees employees employees Under 20 employees Supervisor Professional/managerial Blue collar Public sector Goods sector Industry Primary Manufacturing Construction Transportation Trade Services Sample size 8, ,699 8, ,335 MEN WOMEN Variable All Immigrants Native-Born All Immigrants Native-Born (Values are percentages unless indicated) *Based on preferred language of response to questionnaire (English or French). Notes: 1. Results are weighted to reflect Canadian population estimates. 2. Full-time and part-time students are excluded from the sample. 3. Excludes cases where valid hours worked (zero or positive) are not reported. Source: 1998 Adult Education and Training Survey Master (Internal) File. Calculations by the authors.

10 478 Derek Hum and Wayne Simpson Table 2 illustrates other differences between immigrants and native-born persons. Among the core economic variables we identified, immigrant men and women are slightly older, less educated, and work more hours per week. Immigrant men have shorter average job tenure, but immigrant women have a slightly longer average tenure. Immigrants are regionally concentrated in Ontario and British Columbia, more likely to live in urban areas, less likely to be French-speaking, more likely to have a full-time job, less likely to be self-employed or covered by a collective agreement, and less likely to work in the public sector. Immigrant men are less likely to be found in large firms (200 or more employees) but immigrant women are more likely to be found in the largest firms (500 or more employees). These differences in the characteristics of immigrant and native-born workers might account for some of the observed differences in training activity between these groups. We use multivariate techniques (tobit regression) to see if observed aggregate differences in workrelated training remain after accounting for these other factors. In particular, we examine immigrant status, and the age at which immigration occurred, both of which are provided for the first time in the 1998 AETS. We use a dummy variable to identify those who immigrate as adults (18 years of age or older) and, alternatively, we introduce age at migration directly into our regression specification. First, we pool the immigrant and native-born samples and estimate four regression specifications for men and women separately: (i) a completely general specification which includes immigrant status variables, our core economic variables (age, education, hours worked, and job tenure), a full set of personal characteristics, and a full set of job characteristics (that is, the full set of regressors listed in Table 2); (ii) a specification which removes the job characteristics; (iii) a specification which removes both the job and personal characteristics; i.e., including only the immigrant status and core economic variables; and (iv) a specification which includes only immigrant status variables. By gradually removing factors other than immigrant status, we assess their influence and determine whether our results are robust to specification differences. Table 3 summarizes the results for the immigrant status variables for these different specifications. Table 4 presents detailed results for the full set of regressors. We discuss these results in turn. From Table 3, it is fairly easy to generalize the differences in training activity for immigrants and the foreign born, controlling for a wide variety of other factors. The effect of introducing a dummy variable to differentiate immigrants from the nativeborn in the first column of results indicates that immigrant men train significantly less than nativeborn men, but there is no significant difference for women. The coefficient estimate is 48.6 to 75.1 hours per year for men, depending on the specification. This represents the estimated gap in training received by immigrant men relative to native-born men, other factors (as specified) considered. It also implies that employed immigrant men as a group receive between 13 and 20 fewer hours of training per year than employed native-born men, depending on the specification. We next estimate the effect due to the age at immigration. We initially introduce age at immigration and its square, but the squared term is never significant; accordingly, the third column of results in Table 3 reports a significant linear reduction in training opportunities for older immigrants for both men and women. The estimates are remarkably similar for men and women and imply, for all but the barest specification, that training declines by about a year for each year that migration is delayed. For women who immigrate as children, the results suggest that they train more than those who are native-born, since the coefficient on the immigrant dummy variable is positive and significant for all specifications. This apparent advantage turns to a disadvantage between 16 and 19 years of age, depending on the specification. For men, there is no significant advantage for those who immigrate as children, except in the barest

11 Job-Related Training Activity by Immigrants to Canada 479 TABLE 3 A Summary of the Estimated Effects of Immigrant Status on Training for Men and Women for Various Regression Specifications 1 Regression Specification Immigrant Age at Immigration Adult Immigrant Effect 2 Effect 3 Effect 4 Imm Imm Age Imm Adult Men Full set of regressors 69.4 (4.6) 2.5 (0.1) 3.7 (3.2) 12.5 (0.6) (3.8) No job characteristics 75.1 (5.0) 7.8 (0.3) 4.2 (3.6) 12.2 (0.6) (4.2) No job or personal characteristics 48.6 (3.4) 33.8 (1.3) 4.2 (3.6) 16.8 (0.8) (4.3) Only immigrant status regressors 57.9 (3.9) 56.5 (2.0) 5.6 (4.7) 24.4 (1.1) (5.1) Women Full set of regressors 13.3 (0.9) 61.5 (2.4) 3.8 (3.4) 35.2 (1.7) 82.0 (3.3) No job characteristics 20.9 (1.5) 79.3 (3.1) 5.0 (4.5) 40.3 (2.0) (4.2) No job or personal characteristics 6.1 (0.5) 91.4 (3.6) 4.8 (4.3) 50.6 (2.6) 93.5 (3.8) Only immigrant status regressors 12.9 (0.9) (3.7) 5.5 (4.8) 60.1 (2.9) (4.6) Notes: 1 Based on Tobit regression estimates of equation [1] where y * = Xβ+ξ and X contains (i) a full set of regressors, i.e., immigrant status regressors, economic (life cycle) controls, and controls for personal and job characteristics (observed heterogeneity); (ii) no job characteristics, i.e., only immigrant status regressors, economic controls and controls for personal characteristics; (iii) no job or personal characteristics, i.e., only immigrant status and economic variables as regressors; and (iv) only immigrant status regressors. For detailed results for the full set of regressors, see Table 4. Parentheses contain t-values. Significant coefficients are in bold. 2 Only immigrant status regressor is a dummy variable =1 if immigrant and =0 if native-born (Imm); other control variables as stipulated in rows of table. 3 Immigrant status regressors are immigrant dummy variable (Imm) and age at immigration (Age). 4 Immigrant status regressors are immigrant dummy variable (Imm) and dummy variable to indicate adult immigrant, i.e., immigrant after age 17 (Adult). specification, where males who immigrate before the age of ten appear to have some advantage. In the final column of Table 3, we report results that replace age at migration with a dummy variable to indicate whether or not immigration occurred in adulthood, defined as after 17 years of age. Our results indicate that men who immigrate in childhood receive the same amount of training as their native-born counterparts; however, men who immigrate as adults receive 27.4 fewer hours of training for the full set of regressors, a result that is statistically significant. The results for women are similar. For the full set of regressors, there is no significant difference in training activity between women who immigrate as children and the native-born, but for adult immigrants, there is a significant reduction of 24.6 hours of training per year. Table 4 presents complete results for the full set of regressors in order to assess factors other than immigrant status on training duration. In particular,

12 480 Derek Hum and Wayne Simpson TABLE 4 Estimates of the Determinants of Work-Related Training Duration for Employed Men and Women for the Full Set of Regressors Regressors MEN WOMEN Immigrant 2.5 (2.3) 61.5 (2.4) Age at immigration 3.7 (3.2) 3.8 (3.4) Education Some high school (6.8) (10.0) High school degree 78.1 (4.8) (7.6) Postsecondary diploma 26.1 (1.4) 99.3 (5.0) Bachelor s degree 19.3 (1.0) 83.4 (4.3) Graduate degree 6.7 (0.2) 89.6 (2.8) Age years (5.2) 89.0 (2.0) years (9.4) 92.9 (2.2) years (10.0) (2.7) years (10.4) (2.8) years (11.7) (4.9) 65 years and over (9.6) (3.5) Hours worked (per week) 0.8 (1.2) 0.6 (0.8) Job tenure year (4.1) 95.3 (4.2) 1 5 years (7.8) (9.4) 5 10 years (6.5) (6.9) years (6.0) (7.0) Over 20 years (5.1) (6.3) Region Atlantic 3.8 (0.2) 9.5 (0.5) Quebec 33.8 (1.1) 13.5 (0.5) Prairies 31.6 (1.5) 37.2 (1.8) Alberta 19.8 (1.1) 31.6 (1.8) British Columbia 6.2 (0.4) 61.2 (3.9) Urban 1.3 (0.1) 2.1 (0.1) French* (3.7) 77.2 (2.8) Ethnic origin (2.3) 34.8 (0.7) Black Aboriginal 9.0 (0.2) 28.4 (0.7) Disabled 52.5 (1.9) 60.9 (2.6) Married or common law 31.1 (1.3) 9.9 (0.6)... continued

13 Job-Related Training Activity by Immigrants to Canada 481 TABLE 4 (Continued) Regressors MEN WOMEN Married previously 42.2 (1.5) 25.1 (1.2) Preschool children 5.5 (0.4) 32.7 (2.3) Full-time job 38.5 (1.3) 0.8 (0.0) Permanent job 18.9 (0.8) 73.8 (4.2) Changed jobs during year 86.8 (3.4) 77.3 (3.8) Self-employed 63.0 (3.1) 9.1 (0.4) Union coverage 35.7 (2.5) 36.8 (2.7) Firm Size 500 or more 85.1 (5.1) (9.3) employees 75.8 (3.3) (6.3) employees 68.4 (2.7) 96.3 (4.1) employees 14.2 (0.8) 64.0 (3.7) Supervisor 10.3 (0.8) 53.3 (4.4) Professional/managerial 84.4 (5.4) 44.7 (3.6) Blue collar 27.0 (1.6) 27.7 (1.2) Public sector 50.7 (2.7) 39.7 (2.7) Goods sector 71.9 (1.5) 38.1 (0.5) Industry Primary 0.0 (0.0) 0.2 (0.0) Construction 8.3 (0.3) 33.3 (0.6) Transportation 9.7 (0.2) 6.8 (0.1) Trade (2.0) (1.4) Services 55.2 (1.0) 62.2 (0.8) Constant (3.8) 62.2 (0.7) Sample size 8,681 8,161 Age at imm squared added 0.0 (0.5) 0.2 (0.6) Observed above limit 26.2% 28.9% Mean error Correlation of observed and expected 27.0% 29.6% *Based on preferred language of response to questionnaire (English or French). Notes: 1. Method of estimation is weighted Tobit regression, where weights reflect Canadian population; dependent variable is training duration, which is positive or zero. 2. Coefficient estimate b represents the effect of a one-unit change in the independent variable x on the latent index; the effect on the expected duration of training is F(xb).b where F(.) is the normal distribution function and equals the proportion of the sample above the limit (McDonald and Moffitt 1980, p. 319). 3. t-values in parentheses; bold indicates statistical significance at the 5 percent level. Source: 1998 Adult Education and Training Survey Master File.

14 482 Derek Hum and Wayne Simpson we would note that our core economic factors are generally important, and several of these aspects of human capital investment over the life cycle should be taken into account in any cross-sectional study of training. As expected, there is a highly significant pattern of declining training activity with age, although training declines more slowly with age for women, consistent with the delayed career paths experienced by some married women. There is also a significant pattern of training according to job tenure, but the pattern is more complex. For both men and women, workers receive more training in the first six months on the job as we would expect. There is then a decline in training activity up to five years on the job, after which training activity levels off and may even rise a bit. Also related to job tenure, men and women who changed jobs in the previous year (and hence must have low job tenure) report more training. Women with a university degree receive more training, but men do not; education appears to be more important in explaining the training activity of women than men. Hours worked is insignificant, as is the related indicator of full-time employment. Among personal characteristics, respondents with a disability train less, not unexpectedly. Black workers receive less training. The training deficit for black men is large and statistically significant, which may explain the general wage disadvantage of black men in Canada found in Hum and Simpson (1999). Marital status has no significant effect on training activity. The presence of preschool children, however, is associated with reduced training activity for women but not for men. This suggests that working women with young children postpone or reject training opportunities, which may slow job advancement. Workers with a higher occupational standing (supervisory and professional/managerial workers) and workers in the public sector train more, and selfemployed men train less, other factors taken in account. Among job characteristics, there is a strong correlation between firm size and training, as many studies have observed. Consistent with Mincer s results for the US, we find that unionized workers train less. Hence, the more limited access to unionized jobs for immigrants observed in Table 2 cannot explain the training gap between immigrants and the native-born. The type of production, represented by a series of industrial categories, variables indicating goods production, and blue collar jobs, is not significant. We also estimate the same set of regressions for separate male and female samples of immigrants and the native-born. The results for immigrants are summarized in Table 5 for the four specifications; the first column presents results for age at migration (the squared term is never significant and is therefore dropped), while the second column presents results for the adult immigrant indicator. The results are again similar for men and women in the first column indicating that, for those who receive training, there is a statistically significant reduction of four to five hours for each year that migration is delayed. This translates into a reduction of about one hour of training per year for each year migration is delayed, or about the same estimate that we obtained from the pooled data. The results in the second column are less consistent, however. Men who immigrate as adults receive less training; for the full set of regressors, they receive 17 hours less per year than their counterparts who migrated as children. For the full set of regressors, women who migrated as adults received about 13 hours less training, but the effect is not significant. For women, the only statistically significant effect occurs when all regressors except immigrant status are removed, which is probably the least reliable result. Thus, the results for the adult immigrant dummy are less robust and likely indicate that a simple immigrant-as-adult vs. immigrant-as-child distinction is less meaningful than age at migration per se. Table 6 provides the detailed results for men and women, immigrant and native-born, for the full set of regressors. While many of the results are similar to those reported for the pooled sample in Table 4,

15 Job-Related Training Activity by Immigrants to Canada 483 TABLE 5 A Summary of the Estimated Effects of Age at Immigration on Training for Immigrant Men and Women for Various Regression Specifications Regression Specification Age at Immigration Effect Adult Immigrant Effect Men Full set of regressors 3.6 (3.1) 73.8 (2.9) No job characteristics 4.6 (4.1) 93.6 (3.8) No job or personal characteristics 4.4 (3.9) 93.3 (3.8) Only immigrant status regressors 4.4 (3.7) (4.1) Women Full set of regressors 4.3 (2.4) 47.8 (1.2) No job characteristics 5.2 (3.0) 62.2 (1.6) No job or personal characteristics 4.8 (2.8) 51.8 (1.4) Only immigrant status regressors 6.5 (3.5) (3.2) Note: Based on Tobit regression estimates of equation [1] for immigrants only where y * = Xβ+ξ and X contains (i) a full set of regressors, i.e., age at immigration or a dummy to indicate immigration as an adult, economic (life cycle) controls, and controls for personal and job characteristics (observed heterogeneity); (ii) no job characteristics, i.e., only immigrant status regressors, economic controls and controls for personal characteristics; (iii) no job or personal characteristics, i.e., only immigrant status and economic variables as regressors; and (iv) only immigrant status regressors. For detailed results for the full set of regressors for the immigrant and native-born samples, see Table 6. Parentheses contain t-values. Significant coefficients are in bold. there are some noteworthy differences. For men, despite the inclusion of age at migration, there appears to be a much stronger effect of age per se on training for immigrants. This implies a double training disadvantage for immigrants who are older and who have migrated as adults. The training disadvantage for black men remains large and is actually larger for native-born blacks, consistent with results that the wage disadvantage for black men is independent of immigration status (Hum and Simpson 1999), although the results are no longer significant at the 5 percent level. The effect of firm size on training is stronger for immigrants than the native-born. While we can only speculate without further evidence, this might support Holtmann and Idson s (1991) argument that larger firms have a greater incentive to train because they can pool the risks, provided that the risks associated with training immigrants are higher than those of training the native-born. WHAT ARE THE BARRIERS TO TRAINING FOR IMMIGRANTS? What do immigrants themselves perceive to be barriers to achieving training? The AETS asks respondents about job- or career-related training that they had not taken and, if so, why not. In Table 7, we separate these responses into three groups of workers (i) all immigrants, (ii) those who immigrate as adults, and (iii) the native-born to see whether there are identifiable barriers to training for

16 484 Derek Hum and Wayne Simpson TABLE 6 Estimates of the Determinants of Work-Related Training Duration for Employed Immigrant and Native-Born Men and Women for the Full Set of Regressors MEN WOMEN Variable Immigrant Native-Born Immigrant Native-Born Age at immigration 3.6 (3.1) 4.3 (2.4) Immigrant after age (2.9) 47.8 (1.2) Education Some high school (2.4) (2.4) (6.5) (1.9) (1.7) (10.1) High school degree 67.1 (1.9) 63.0 (1.8) 67.8 (3.8) (2.2) 98.1 (2.1) (7.5) Postsecondary diploma 4.2 (0.1) 2.1 (0.1) 34.0 (1.7) 35.9 (0.5) 24.5 (0.3) (5.8) Bachelor s degree 3.3 (0.1) 7.6 (0.2) 26.2 (1.3) 41.5 (0.6) 33.6 (0.5) (5.1) Graduate degree 44.3 (0.6) 37.4 (0.5) 24.1 (0.6) 26.8 (0.2) 30.9 (0.2) (3.8) Age years (7.8) (7.6) (3.3) (1.6) (1.6) 75.7 (1.7) years (8.3) (8.1) (7.5) (1.4) (1.5) 69.7 (1.6) years (8.1) (8.0) (8.4) (1.1) (1.3) (2.3) years (8.4) (8.3) (8.4) (0.8) (1.1) (2.6) years (8.4) (8.3) 514.3(10.0) (1.3) (1.7) (4.5) 65 years and over (0.0) (0.0) (8.4) 46.9 (0.2) (0.6) (4.0) Hours worked (per week) 0.3 (0.2) 0.3 (0.2) 0.7 (1.0) 3.5 (1.3) 3.2 (1.2) 1.4 (1.7) Job tenure year (3.7) (3.6) 78.4 (2.8) (2.4) (2.4) 64.0 (2.8) 1 5 years (5.0) (5.0) (6.3) (4.0) (3.9) (8.0) 5 10 years (4.2) (4.1) (5.2) (3.1) (2.9) (6.0) years (4.3) (4.1) (4.5) (3.7) (3.3) (5.6) Over 20 years (3.2) (3.1) (4.0) (2.7) (2.3) (5.6) Region Atlantic 89.2 (1.0) 90.7 (1.0) 9.0 (0.4) (0.8) (0.8) 7.5 (0.4) Quebec 32.4 (0.6) 25.3 (0.5) 37.5 (1.1) (1.8) (1.7) 14.8 (0.5) Prairies 69.1 (1.1) 68.1 (1.1) 21.6 (1.0) 12.2 (0.1) 12.2 (0.1) 43.8 (2.2) Alberta 36.1 (0.8) 37.8 (0.9) 11.5 (0.6) 8.2 (0.1) 13.4 (0.2) 41.2 (2.4) British Columbia 25.2 (0.8) 26.7 (0.8) 8.2 (0.4) (2.2) 95.5 (2.0) 65.6 (3.9) Urban 64.6 (1.1) 64.4 (1.1) 6.1 (0.4) (1.6) (1.8) 9.0 (0.6) French* (1.3) (1.4) (3.0) (1.2) (1.1) 55.0 (1.9) Ethnic origin Black (1.7) (1.6) (1.8) 24.9 (0.3) 9.7 (0.1) (0.8) Aboriginal (0.0) (0.0) 17.1 (0.3) (0.7) (0.7) 16.8 (0.4)... continued