Earnings and employment outcomes for male and female postsecondary graduates of coop and non‐coop programmes

Abstract

New postsecondary graduates seeking entry into the Canadian workforce may experience more favourable labour market outcomes if they augment their education with job‐related training. Employers, looking for new ways of differentiating among equally credentialed prospects, may prefer job candidates with practical training. These new realities have led to a steady increase in the availability and popularity of cooperative education programmes in Canadian colleges and universities. Few existing studies, however, have sought to examine whether or not earnings premiums or employment advantages exist for coop graduates. This paper compares the earnings and employment outcomes of postsecondary graduates with coop credentials to those with traditional, non‐coop credentials. Even once controlling for a number of factors, coop programmes provide the greatest advantage at the university level, particularly among male graduates. In terms of employment status, college males and university females experienced the greatest advantages to attending coop programmes.

Keywords:

• cooperative education
• earnings
• employment status
• Canada
• university
• college

Acknowledgement

We are grateful to the Social Sciences and Humanities Council for helping to fund this project. While the research and analysis are based on data from Statistics Canada, the opinions expressed do not represent the views of Statistics Canada.

Notes

1. Unfortunately, there is very little population‐based research available in North America to evaluate the economic benefits of cooperative programmes at the postsecondary level, particularly relating to the labour market outcomes of youth who have recently made their school‐to‐work transitions.

2. At the time these graduates accounted for 93% of postsecondary graduates with cooperative credentials (see Darch 1995).

3. Unfortunately, coop/non‐coop findings of this study were not thoroughly discussed because the coop/non‐coop distinction was used as control variable for a broader statistical analysis comparing the labour market outcomes of graduates of various fields of study.

4. All figures were reported in 1997 CDN dollars.

5. Technically, many professional degree programmes (i.e. law, pharmacy, and medicine) are undergraduate programmes. However, they are generally classified separately because they are advanced degree programmes that have rigorous admission requirements based on undergraduate grades as well as standardised testing.

6. Moreover, it was not possible with the NGS to distinguish coop programmes from similar programmes offered by professional programmes such as medical residencies or legal internships (e.g. articling).

7. Logistic regression is the most common generalised linear model, where the binomial distribution is used in the likelihood function and the logit link, is used to map the mean of the response variable to a set of linear predictors (see McCullagh and Nelder 1989). Thus, the resulting linear equation is represent the regression parameters, the X’s are the explanatory variables, and π in this analysis is the probability of being employed full‐time.

8. The regression equation for the generalised linear model is conveniently expressed in matrix form as g(µ) = η = Xβ, where g(µ) = η is the invertible link function connecting the linear predictor, η = Xβ, to the mean of the response variable y. The log link, g(µ) = log _e µ, is used to relate the mean, µ, to expectation of the response variable. The right hand side of the equation consists of β, a vector of coefficients, and X, the design or model matrix of predictors that includes categorical, quantitative, and polynomial regressors.

9. The estimates in Table 1 exclude graduates of trades programmes so that direct comparisons of coop and non‐coop programmes can be made at the college and university levels. Contrasts between graduates of coop and non‐coop programmes for all postsecondary graduates (including trades graduates) in this study are provided in Appendix A.

10. Due to the large sample size even relatively small differences are statistically significant. The estimates are weighted to adjust the sample according to population characteristics.

11. Indicator (0–1 dummy) coding is used to provide contrasts for the categorical variables.

12. The estimates are converted into meaningful quantities (predicted probabilities) by taking the inverse of the link function, g ⁻¹(Xβ), before plotting the fitted values in the corresponding figures.

13. When not otherwise stated the results for models with more than one explanatory variable are reported as controlling for the other variables in the model. Wald tests are used for individual parameter estimates; whereas likelihood ratio tests are used for categorical variables with multiple parameter estimates.

14. To save space the estimates and corresponding tests of statistical significance for the control variables are not discussed; however, they are provided in Table 1.

15. Means are used for quantitative variables and proportions are used for categorical variables.

16. Preliminary diagnostics reveal a curvilinear relationship between age and earnings. Thus, a quadratic regressor for age was added to the model. Polynomial contrasts were used for age and age squared to adjust for multicollinearity.

17. When the five‐year follow‐up to the 2000 NGS becomes available at Statistics Canada’s research data centres, social and policy researchers will be able to determine whether recent graduates of cooperative programmes are able to maintain their relative labour market advantages over the early stages of their careers.