Statistics Canada - Government of Canada
Accessibility: General informationSkip all menus and go to content.Home - Statistics Canada logo Skip main menu and go to secondary menu. Français 1 of 5 Contact Us 2 of 5 Help 3 of 5 Search the website 4 of 5 Canada Site 5 of 5
Skip secondary menu and go to the module menu. The Daily 1 of 7
Census 2 of 7
Canadian Statistics 3 of 7 Community Profiles 4 of 7 Our Products and Services 5 of 7 Home 6 of 7
Other Links 7 of 7

Warning View the most recent version.

Archived Content

Information identified as archived is provided for reference, research or recordkeeping purposes. It is not subject to the Government of Canada Web Standards and has not been altered or updated since it was archived. Please "contact us" to request a format other than those available.

Skip module menu and go to content.
Analytical Studies Branch Research Paper Series - Logo

Analytical Studies Branch Research Paper Series


Volume 2007
Number 292

The Role of University Characteristics in Determining Post-graduation Outcomes: Panel Evidence from Three Recent Canadian Cohorts

Main page

More information

PDF version
of the study

Previous issues
in this series

The Role of University Characteristics in Determining Post-graduation Outcomes: Panel Evidence from Three Recent Canadian Cohorts

by Julian Betts, Christopher Ferrall and Ross Finnie

Executive summary

A major role of universities is to prepare their students for success in the labour market after they graduate. Surprisingly, we know very little about how universities' educational policies influence the success of their students. From a policy perspective, the role of field of study and university characteristics in determining labour market success of graduates is a compelling issue. Education feeds indirectly into both public and private budgets through productivity gains, earnings power and the tax base. How should scarce funds be spent to foster successful post-graduation outcomes?

In this paper, we use labour-market data from the 1982, 1986 and 1990 waves of the National Graduates Survey (NGS) to examine how graduates of Bachelor's programs in Canadian universities have fared in the labour market. Our specific goal is to test whether given types of educational spending in Canadian universities are helpful in increasing students' earnings five years after graduation. In addition we test for a link between the probability of employment five years after graduation and university characteristics.

Our data set includes measures of university traits at three separate periods (specific to each cohort of graduates) based on data gathered between 1978 and 1990. Such a model fully controls for all unobserved traits of each university that are fixed over time. Using university fixed effects, we identify the impact of certain university traits, such as professor-pupil ratio, on students' wages five years after graduation by (conditionally) correlating changes in these wages across cohorts with changes over time within each university in these traits.

We regress log annual earnings five years after graduation of men and women who graduated in 1982, 1986 or 1990 on a vector of personal and family traits. These include age and its square at the time of the survey, marital status, parental education, presence of children, province of residence, education completed prior to the Bachelor's degree, field of study, and the months of work experience prior to graduating from the Bachelor's program. The models condition on a vector of university characteristics, including measures of professor/student ratios, measures of the composition of the teaching staff, and median professor salaries.

To measure student success in the labour market, we model men's and women's annual earnings reported in the fifth year follow-up.

In order to measure university resources devoted to teaching, we merge the NGS data with data available from the federal government on the characteristics of each Canadian university: the ratio of full-time teaching personnel to undergraduate students, the proportion of teachers by rank (full, associate, assistant professors), median salaries of faculty by rank, and the proportion of graduate students in the total student population. The last of these variables, graduate share, serves as a proxy for the degree to which the university is oriented toward research.

We also examined the relation between average fees (including non-tuition fees) and subsequent earnings of graduates.

In addition, we obtained data on the proportion of the freshman with high school averages of 75% or higher, as well as the average high school grades of freshmen. The data, provided by Maclean's, are for 1994. Although these data are not matched to each cohort, and post-date the year of graduation of each student by 2 to 12 years, we used these variables in selected models to corroborate American evidence that the degree of selectivity in admissions at a university is positively associated with students' subsequent earnings.

Our regression sample consists of males and females who obtained a Bachelor's degree in 1982, 1986 or 1990, and who had valid data for age, province of residence, work experience prior to graduation, language spoken, and earnings. Earnings were top-coded at $150,000 per year.

As a starting point, we ran a random effects model that conditioned log wages on our two 1994 measures from Maclean's of the high school grades of incoming students. The coefficients on the two measures of high school achievement of freshman are positive whether they are entered together or alone, thus weakly corroborating evidence from the American literature that the selectivity of universities' undergraduate programs is positively linked to subsequent earnings.

Results for men (when one measure of university traits at a time is added to a basic model including personal demographic traits and past educational experiences) suggest that male graduates' earnings are positively linked to professor-to-student ratios. An increase of 0.01 in the professor-to-pupil ratio is predicted to raise students' earnings five years after graduation by about 0.25%. This represents a modest increase in university staffing, given that for the average student in our sample the professor-to-pupil ratio was 0.089, with a standard deviation of 0.084.

In contrast, a number of other university traits, such as median professor salary and the share of graduate students in the overall student population, are not significantly linked with undergraduates' earnings five years after graduation.

Higher fees, as measured by the total fees paid by arts undergraduates are positively associated with men's later earnings, providing indirect evidence that students benefit, to some extent, from fee hikes and consequent increases in expenditures on undergraduate education by universities during the time under study.

Another regressor of note is that the coefficients on the dummies indicating the field in which the graduate specialized are quite large. Apparently, a man's major influences his earnings in a significant way. Overall, there is a gap in predicted earnings between those in the most highly paid field "Other Health" and those in the lowest paying major, Fine Arts/Humanities, of about 60%.

For both men and women, an increase in enrollment of 1,000, or an increase of about 8% at a typical university, is associated with a 1% drop in earnings five years after graduation. But for women, in the final model that incorporates all of the university traits at once, enrollment remains the only significant university variable, in contrast to the results for men.

After dropping the controls for majors, for men, professor/student ratio, undergraduate enrollment, and fees remain significant, but the coefficients rise by a third to a half. The implication is that expansions in the professor/student ratio, increases in fees, and reductions in overall enrollment allow students a greater opportunity to enroll in majors that have high payoffs in the labour market.

When the controls for major are dropped for women, each university characteristic except for professor salary becomes statistically significant. In addition, the coefficients grow in absolute size by about half. Thus, changes in many of a university's characteristics may induce changes in a woman's major that, in turn, affect wages. Factors encouraging women to enroll in more highly paying majors include an increase in the professor/student ratio, a drop in overall enrollment, an increase in the share of graduate students in the overall student body, and an increase in fees.

To check for non-linearities, we reran the basic models after adding squared terms for each university characteristic. For men, we found a concave relation between log wages and tuition and fees, and a positive relation between log wages and tuition and fees up to about $5,900. Still, these results suggest that benefits from increases in tuition and fees arise primarily in the universities with the lowest tuition and the biggest gap between desired expenditures and actual budget.

For women, there is a positive concave relation between log wages and two university characteristics: tuition and fees and median professor salary. The quadratic relationship of log wages with tuition and fees is similar to that for men, with a peak at $5,800, but this relation is only weakly significant. More strongly significant is the quadratic in median professor salary, where a positive but diminishing link with log wages emerges, up to $58,800, beyond which the relationship becomes negative.

For men, none of the university characteristics is significantly related to the probability of employment. For women, with one important exception, none of the university characteristics is significant, and the signs of the coefficients match the results for men. As for men, the variable that is most highly significant is median professor salary, but unlike the model for men, it becomes marginally significant (t=1.91). A $1,000 increase in median professor salary is predicted to increase the probability of employment for female graduates by 0.36%.

This paper presents the first analysis of the link between university resources and earnings of Canadian undergraduates after they graduate. It also represents one of the first times that university fixed effects have been used to control for unobserved and fixed traits of the university. We find strong evidence that there are fixed and unobserved wage effects associated with attendance at different universities. After incorporating fixed effects we find that university traits, in particular the professor/student ratio, enrollment, and fees can explain some of the observed inter-university differences in earnings.

What do our findings imply for the earlier literature on American university quality, which has used purely cross-sectional variation to identify the impact of university spending? First, our consistent and strong rejection of the hypothesis that undergraduates' earnings are identical across campuses after controlling for the standard set of measures of university quality, for personal background, and for selectivity in admissions raises questions about the interpretation of earlier findings. Second, our fixed-effect approach offers more direct policy guidance. Our analysis cannot tell undergraduates which university is the best to attend, but we can predict the likely outcome if resources change over time.

The more important finding is how small such effects are. Variations in university spending may have a much smaller impact on graduates' earnings than do variations in unobserved traits across universities and the large observed variations in earnings across university majors.

View the publication The Role of University Characteristics in Determining Post-graduation Outcomes: Panel Evidence from Three Recent Canadian Cohorts in PDF format.

You need to use the free Adobe Reader to view PDF documents. To view (open) these files, simply click on the link. To download (save) them, right-click on the link. Note that if you are using Internet Explorer or AOL, PDF documents sometimes do not open properly. See Troubleshooting PDFs. PDF documents may not be accessible by some devices. For more information, visit the Adobe website or contact us for assistance.

Home | Search | Contact Us | Français Top of page
Date modified: 2007-02-26 Important Notices