* I would like to thank Scott Davies, John Fox, and Jane Synge for their generous contribution to the manuscript. I would also like to thank the anonymous reviewers for their many helpful suggestions. As well, I am grateful to Statistics Canada for making the National Graduates Surveys available at their Research Data Centre at McMaster University.
1. The transition from the Fordist to the post-industrial era coincided with a shift to services which, in turn, brought about changes in the industrial mix of jobs. The transition began slowly in the 1950’s and 1960’s, and then more rapidly in the 1970’s. By the 1980’s, the service economy became a major source of new employment in the North American economy (Myles, Picot, and Wannell, 1993: 172).
2. The transition from Fordist to post-industrial era was also accompanied by a transition to a different ‘class-scheme’ in North America (See Esping-Anderson, 1993: 12). The transition is also believed to be responsible for changes in political arrangements in Canada (See Jenson, 1989).
3. The latter, less optimistic, images of the new economy emerged when it became clear that the post-war boom ended during the 1970’s (Myles et al., 1993: 172).
4. The terms “knowledge” and “postindustrial” are generally used synonymously to reflect the changing economy in the post-war era (Rinehart, 1987: 73).
5. Interestingly, however, it has been found that there are little, if any, differences between liberal and vocational university graduates in their perceived possession and acquisition of employability skills (Lin et al., 2000).
6. One study by Ross Finnie (2000b) does make comparisons over time using the 1982, 1986, and 1990 NGS. However, the results are descriptive. Much of the analysis involves the use of either frequencies or cross tabulations. As well, this study does not include the most recent NGS cohort, those who graduated in 1995.
7. A common problem when using a series of cross-sectional surveys is that not all of the questions are asked in every survey. In addition, the wording of questions can often change over time, as can the response options. This is a particular limitation of the National Graduates Surveys, given that Statistics Canada and Human Resources Development Canada made certain changes in the content of the questions asked of the four cohorts. Therefore, a great deal of care and attention was devoted to recoding the variables so that meaningful comparisons could be made across surveys. Only variables that are compatible, or could be made compatible, across the four surveys were used in the analyses.
8. The selection criteria are also based, in part, on the variables available in all of the surveys. The proportions of graduates selected are relatively similar across the four cohorts (13.5% were excluded in 1982, 14.4% in 1986, 12% in 1990, and 13.9% in 1995).
9. Statistics Canada has also grouped graduates with a bachelor of education degree (i.e. B.ed’s) in this category because they are not considered to be professional programs of the same level as those included in category number five.
10. This category was coded by Statistics Canada. It also includes graduates of other highly specialized programs (i.e. advanced teaching and nursing programs). Due to the high degree of heterogeneity, the results obtained for graduates of this category should be interpreted with some caution.
11. Unfortunately, the field of study and level of schooling variables in the NGS do not distinguish between trades graduates of highly skilled programs (i.e. plumbers and electricians) and those of vocational programs (i.e. hairdressers and travel agents).
12. The number of cases censored was small (177 cases). However, as a precautionary measure, the regression models for earnings were re-estimated using the Tobit model, which is appropriate for censored data (Tobin, 1958; see also Breen, 1996). Since the estimates obtained from the Tobit model were identical to those of the OLS model, the estimates from the more parsimonious (OLS) model in Table 1.
13. Each model includes the sociodemographic variables sex, age, marital status, region, language of interview, number of children, mother’s education, father’s education, and NGS graduation cohort (1982, 1986, 1990, and 1995). The coefficients for these variables are available upon request from the author.
14. Incidentally, an additional model including two three-way interactions was also estimated. The first three-way interaction was among gender, NGS cohort, and level of schooling, and second was among gender, NGS cohort, and field of study. While the effects of these interactions were statistically significant (p<.001), as would be expected with such a large sample, the general earnings and employment patterns for men and women were very similar. Therefore, these results are not provided here. They are available from the author by request.
15. Since the effects of the interactions between NGS cohort and the level of schooling, and NGS cohort and field of study are statistically significant, the parameter estimates from Model 1 are not interpreted.
16. When not otherwise stated, the relationships between the dependent variable and each independent variable are to be interpreted as controlling for all of the other variables included in the respective models.
17. While one might expect that some of the wage premium to graduates with higher level credentials could be explained by the fact that they may have had prior participation in the labour market, part of this effect would likely be captured by the age variable included in the analysis.
18. The earnings decline during both periods is statistically significant (p<.001).
19. By comparison, the average earnings in the population were $22,490 in 1984, $23,946 in 1988, $23.667 in 1992, and $23.130 in 1997 (Statistics Canada, 2001b).
20. The sharp earnings decline of health graduates during both periods is statistically significant (p<.001).
21. Respondents employed full time represent the reference group.
22. The coefficients for the sociodemographic variables, as well as the extended set of coefficients for the interaction terms, are available upon request from the author.
23. Since the effects of the interaction terms between NGS cohort and each of level of schooling and field of study are statistically significant in Model 2, the parameter estimates from Model 1 are not discussed.
24. All predicted probabilities are calculated holding the other variables constant at their means.
25. Except for the 1986 cohort, where their respective probabilities of being employed full time are the same.
26. An interesting exception is professional graduates of the 1995 cohort. Their probability of being employed full-time in 1997 is surprisingly low, .78.
27. Although, the probability of being employed part time is nearly identical for college and trades graduates of the 1995 cohort. Their respective probabilities are (.125) for trades, and (.124) for college. This difference is not statistically significant. Incidentally, the corresponding part-time employment rates in Canada were 16.8 in 1984, 16.8 in 1988, 18.5 in 1992, and 19.0 in 1997 (Statistics Canada, 2001c: 5).
28. In comparison with these figures, the overall unemployment levels for the population in 1984, 1988, 1992, and 1997 are 11.3, 7.8, 11.2, and 11.2, respectively (Statistics Canada, 2001d).
29. Graduates of professional programs of the 1995 cohort are an exception. They have the highest probability of being unemployed (.16).
30. For example, from 1982 to 1995 the number of college graduates increased by 8%, whereas graduates of bachelors and professional programs increased by 47%; the number of graduates with master’s degrees increased by 72%, and the number of graduates with Ph.D.’s increased by 114% (see Statistics Canada, 1995).
31. Such an interpretation is consistent with the credentialist perspective, which asserts that an oversupply of graduates leads to the devaluation of credentials (Collins, 1979; Livingstone, 1998).
32. The mother’s and father’s education variables only represent rough proxies for parental SES, and could not be used for more elaborate analysis involving parental background.
33. In fact, the amount of student debt has increased substantially. For example, college and university undergraduates alone owed 65% more (at the time of graduation) in 1990 than they did in 1982 (Finnie and Garneau, 1996; see also Statistics Canada, 2001a: 28).
34. “Gate-keeping” might also be used to ensure that the salaries of professionals are not undermined by allowing the supply of recently trained professionals in a field (for example, law or social work) to outstrip the demand for these professionals.
35. Rinehart’s (1987) argument is that while the new economy may create higher level “professional” jobs, it may also be responsible for a disproportional shift into part-time work, even among the more highly educated workforce.