Archived — Working Paper Number 5: Steppin' Out: An Analysis of Recent University Graduates into the Labour Market

Prepared by Ross Finnie, School of Public Administration, Carleton University and Statistics Canada, May 1995


Summary

This study provides a descriptive analysis of a sample of bachelor-level university graduates derived from "The Follow-Up of 1982 Graduates" database, with an emphasis on comparisons between Natural Science and Engineering, and Non-NSE graduates, and men versus women. The unique nature of the data and the mix of cross-tabulations and regression analysis covering many different aspects of the education program and early labour market experiences gives a perspective on the school-to-work transition which did not previously exist. This is especially useful for the evaluation of the Canada Scholarships Program which encourages university graduates to enrol in engineering and the sciences.

Results of the Cross-Tabulations

  • Most graduates are working or back in school five years after leaving, university, although some pass through an initial period of joblessness before finding -employment. Activity rates vary considerably by field of study and sex, with Engineering (ENG) and Mathematics/Physical Sciences (MATHSCI) graduates having higher rates of full-time employment than others, and women more likely to be found in part-time jobs than men.
  • ENG and MATHSCI graduates appear to be more concerned with developing -specialized knowledge and job skills and improving their career prospects when choosing their education program. Non-NSE Graduates put greater weight on the acquisition of general communication, social and reasoning skills, while Agricultural and Biological Science (AGBIOSC) graduates resemble the Non-NSE group more than the other science graduates. Women claimed to have been generally more concerned with all the criteria than men, but it is not clear if this reflects different choices, more careful decision making. or simply the manner in which they respond to the questions.
  • Satisfaction with the different aspects of the programs corresponds to the pre-programme priorities just cited. ENG and MATHSCI graduates are happier with the narrower career aspects of their programs; Non-NSE men and women express greater satisfaction with the more general developmental aspects; while the AGBIOSC group is less happy than the other NSE groups in terms of the job-specific aspects of the program — below the Non-NSE group in terms of general developments and, generally, the least satisfied with their programs. The groups expressed similar opinions in terms of the importance of the learning satisfaction aspect of the program, and all were more or less equally satisfied on this count.
  • The job-education match is closest for ENG and MATHSCI graduates, followed by the Non-NSE group, AGBIOSC, and the SOCSCI graduates having the weakest job-education match of all. There was a general movement over time into jobs more closely related to the program of study for all groups, which is further evidence of the gradual or step-wise nature of the integration into the labour market for many of these graduates. Match patterns were similar for men and women.
  • Overall, these graduates generally express high levels of satisfaction with their jobs, but are less content with their earnings. The AGBIOSC graduates are the least satisfied in this regard, the MATHSCI group the happiest, and the Non-NSE and ENG men and women lay between. There are no clear gender patterns in these outcomes.
  • The overall evaluation of the program — would it be chosen all over again if given the chance? — roughly follows the job evaluation patterns, with the ENG and MATHSCI graduates most likely to respond in the affirmative, followed by the general Non-NSE group, then the AGBIOSC graduates and the SOCSCI men and women. Patterns are generally similar for men and women. While approval rates are around three quarters at the highest, a full 40 percent of the least-satisfied groups say they would have preferred another program, although no one seems to regret the general decision to have gone to university. Approval ratings are clearly correlated with having a full-time job or being back in school, which suggests that there is perhaps a role for the simple policy of helping students identify fields where they are more likely to find good employment opportunities (although the issue is obviously more complicated than this).
  • The ENG and MATHSCI graduates are clustered in a couple of occupations and industries, while the other groups are more widely distributed. Mean earnings and the rate of part-time work vary significantly by occupation and industry. Women are more likely to be in part-time jobs and their mean earnings are lower than men's almost everywhere – sometimes much lower, meaning that there are significant gender gaps even after controlling for field of education and the industry and occupation of employment.
  • ENG and MATHSCI men and women earned significantly more than their non-science counterparts in 1984, and AGBIOSC men and women made considerably less. But by 1987 – just three years later – the ENG and MATHSCI men had lower mean earnings than the Non-NSE group, while the women in these fields actually had a slightly increased advantage relative to the Non-NSE comparison group.
  • The gender earnings gap was relatively uniform across all educational groups in 1984 around 10 percent when part-time workers are included. The gap increases everywhere by 1987, but by much less among the ENG and MATHSCI graduates than others. As a result, the advantage of the ENG and MATHSCI women must be seen in terms of their not falling as far behind the men in their field as occurs elsewhere. Five years after graduation, the gender earnings gap was 20 to 25 percent for the Non-NSE and AGBIOSC graduates, and just over 10 percent for the ENG and MATHSCI men and women.
  • It is interesting to contrast these gender earnings gaps with the similar levels of satisfaction regarding remuneration expressed by men and women. It could be that women are happy to be in the jobs they have, and are indeed fairly paid; alternatively, they might not like their jobs, but feel the pay is fair under the circumstances; or it could be that they are resigned to making less than men and they express their satisfaction within the context of a general resignation to pay inequity.
  • The gender earnings gap is clearly related to family responsibilities: it is greater among men and women who were married or who had children.

Regression Analysis Summary

The general possibilities and limits of regression analysis were established, and the work reported here should be thought of as descriptive. The analyses were put in the context of always choosing between:

  • wanting to add explanatory variables to the regressions which can rightfully account for male-female differences in earnings; and
  • concern that such "controls" might themselves be the outcomes of discrimination processes, thus leading to overstatements of the portion of the gap. which can be "explained" (and thus underestimating the share which might be due to discrimination).

The procedure adopted was to start with very simple models to establish an initial overview of the gender earnings gap, and then to add variables in order to provide a decomposition of these differences.

  • In 1984, ENG and MATHSCI men and women had substantially higher earnings than the Non-NSE graduates (16 and 11 percent respectively), while AGBIOSC graduates earned almost 10 percent less than the Non-NSE group.
  • These early earnings differences by field of study are very similar for men and women, and are partly related to differences in job attachment, as shown by the role of accumulated experience and part-time versus full-time work status in the earnings patterns.
  • The overall gender earnings gap was around 14 percent in 1984. A significant portion of this gap is associated with marriage and the presence of children: the initial results indicated that married men and those with children had substantially higher earnings than single and childless men, while for women the effects were much weaker. These effects account for about one half of the gender gap which remains after controlling for the different fields of study, and almost all of the gap which could be explained by the variables available in the data.
  • A good portion of the effects of marriage and children can, in turn, be related to differences in labour market attachment. In particular, marriage and children are associated with more experience and higher rates of full-time work for men. The remaining direct effects of the family status variables are small, but significant. Interpretations of causality must be made with caution.
  • The job-education match is an important determinant of earnings for all groups in 1984. Women in jobs directly related to their education fare particularly well, and there was no gap between the earnings of these women and men in similar situations.
  • While occupation and industry play a minor role in explaining the gender earnings gap, they were related to differences by field of study.
  • Adding a full set of interaction variables to allow for different relationships between the explanatory variables and earnings for men and women added to the explanatory power of the 1984 earnings model, but did not change the principal results of interest in any way.
  • By 1987, the sex-education patterns of earnings had changed substantially: while the ENG and MATHSCI men had lost most of the earnings premiums they enjoyed over Non-NSE men less than three years earlier, the advantages of the ENG and MATHSCI women relative to Non-NSE women actually increased (slightly) over this same period. The earnings of the AGBIOSC graduates lagged behind the Non-NSE group about as much as in the earlier period. Once again, these patterns are significantly related to differences in the accumulation of experience and the incidence of part-time work across fields.
  • The overall gender earnings gap rose from 14 percent in 1984 to 24 percent in 1987. The gap is smaller among the ENG and MATHSCI graduates due to the extra advantages of women in these fields, but they lagged behind all the same – just not as much as elsewhere.
  • About two fifths of the 1987 gender earnings gap is related to the marriage and children variables, suggesting that a major factor in these male-female earnings differences is the different impacts of family responsibilities on men's and women's earnings. A good portion of these effects are related to differences in job attachment (i.e., experience, part-time versus full-time, etc.).
  • As in 1984, the job-education match in 1987 is strongly related to earnings; unlike the earlier year, the gender gap is pretty similar across all categories of match.
  • The results are generally very robust across a variety of specifications. including, separate regressions by education, sex and even education-sex group. The only exception is that the marriage and children effects appear to vary by field of study, although some of the samples are small. There is no clear explanation of why this might be, and future research might pursue these observations further.
  • Fixed-effect models were implemented to control for certain unobservable individual characteristics which might bias the coefficient estimates, especially the marriage and children effects. The findings suggest that such bias is indeed quite strong. In particular, while the previous results suggest that men who were married and had children had higher earnings than others and women's earnings were more mixed, the fixed effect findings suggest that men's earnings are largely unaffected by marriage and parenthood while women's earnings fall significantly with marriage and parenthood.

These results are relevant to policy. In particular, while the analysis is limited in what it can say about actual beneficiaries of the existing Canada Scholarships Program, which encourages university students to enter the sciences and engineering, it certainly paints a picture of these fields which is perhaps at odds with the common presumptions underlying the programs. If there is such a demand for NSE graduates, why aren't their earnings higher? This is especially true for the agricultural and biological sciences, where earnings are uniformly lower than for the other NSE groups as well as in relation to the Non-NSE graduates. With 50 percent of the scholarships reserved for women, and the majority of NSE women in the AGBIOSC fields, are women being, encouraged to enter fields where they are likely to have disappointing careers? Furthermore, the disappointing results hold across almost the full array of measures employed, both subjective and objective, regarding evaluations of the educational experience and the record of labour market achievement.

The news is, however, by no means all bad. The ENG and MATHSCI men and women (four of the six NSE sex-education groups) have considerably higher earnings than Non-NSE two years after Graduation, and this must be considered as at least somewhat affirming of the Canada Scholarships Program. Further, the ENG and MATHSCI women's advantages hold as strongly a full five years after graduation, which would seem to validate, at least partly, the stated goal of encouraging women to enter the sciences. On the other hand, the ENG and MATHSCI men are characterized by only average or slightly above average earnings in the later year, while the AGBIOSC men and women have, as noted, consistently lower earnings. In summary, four of the six scholarship recipient groups do no better than other graduates in the longer term, and two of these have decidedly dismal performances.

This does not necessarily mean that the scholarship program is not working. In fact, the high-achieving students who obtain scholarships might do very well in all of these fields and better than they would have fared elsewhere. We simply cannot tell if this is true from these data; nor can we calculate the societal returns to the federal Government's investment in the Canada Scholarships Program, given that the market rates of return (i.e. earnings) may not reflect societal rates of return to investments in these areas of study. Also, the data employed here follow the graduates only five years after Graduation and cover only a single cohort, whereas results might be different over a longer term or for another cohort. What is required is data on the scholarship recipients themselves and, ideally, over a longer period of time.

Nevertheless, the findings presented here should cause one to pause and think. Perhaps more research or a fine tuning of the Canada Scholarships Program is required to ensure that the money is used to encourage students to enter into areas where they will be able to make a significant contribution to Canada's economic well-being and, at the same time, enjoy more successful and personally rewarding careers. It is hoped that this study has made a contribution to this review process. In the meantime, a dissemination of these findings might, by better informing students, allow them to make better education and career choices for themselves.

Date modified: