Choice of Ontario high schools and its impact on university applications

Abstract

The extent to which increasing students' ability to choose between schools can impact their educational outcomes continues to generate significant research interest. I take advantage of the unique context in the province of Ontario, where two publicly funded school systems operate in parallel. I find a small positive impact of school choice on student applications to university. However, most of the impact is in terms of ‘cross-effects’; the most robust finding is that the more Catholic high schools accessible from a neighbourhood, the better the public high schools perform. This is suggestive that one mechanism through which choice affects school outcomes is through competition between public and Catholic school boards.

Keywords:

• school choice
• competition
• economics of education

Acknowledgements

I am grateful to my thesis committee of Abigail Payne, Stephen Jones and Martin Dooley for their ongoing help and support. I thank Rajashri Chakrabarti for helpful comments at the 2011 meetings of the Association for Education Finance and Policy. I also thank the staff at McMaster's PEDAL laboratory for making available the data for this project. I acknowledge financial support from the Canadian Labour Market and Skills Researcher Network in the form of a PhD fellowship. I also acknowledge financial support from the Social Sciences and Humanities Research Council through my supervisor, Abigail Payne.

Notes

1 In fact, there are four, but I exclude the French Catholic and French Public Boards, since they are not in the choice set of most of the students in my sample.

2 However, students of all faiths attending Catholic school may be required to take courses in Catholicism. Whether a non-Catholic can avoid taking these courses (and the degree of difficulty in doing so) varies by board and sometimes by school.

3 Students are allowed to pay extra to apply to more than three universities, and roughly half of students in my sample take up this option.

4 Since virtually all students in the OUAC data apply to exactly three universities, using alternative definitions of having applied makes little difference. The possibility of paying for additional applications beyond three does exist, but few students take this option up.

5 In fact, in my sample years, the percentage of all Ontario students who are in grade 12 or OAC varies from a low of 29% in 2004 (grade 12 students only) to a high of 34% in 2001 (grade 12 and OAC students combined).

6 It is recognized that, for various reasons including school capacity constraints, a student may not actually be able to attend all of the schools which I define as ‘accessible’. That said, a count of schools to which at least some of one's neighbours are attending should provide a reasonable proxy of the actual extent of choice for a given student.

7 To be more precise, I check whether the centroids of any of the postal codes within the DA fall within the school travel zone.

8 I exclude French high schools and high schools with less than 25 students, as these likely are not in the choice set of most high school students.

9 I estimate all equations using OLS. Alternative estimates for Tables 6 and 8 using Papke and Woolridge's (1996) generalized linear model (GLM) strategies for proportional variables are reported in the appendix and give qualitatively similar results. I have also experimented with Tobit regressions, also with quite similar results (likely since only 11% of observations are at the minimum 0%). For all regressions using the difference in matched proportions, ΔUAR, the resulting variables are in fact, quite normally distributed around a mean of close to zero, no observations approach the theoretical boundaries of −1 or 1 and the use of OLS is therefore justified.

10 In fact, looking ahead to Table 5, students on the Toronto side (the side with more choice) of the matched DA pairs score slightly worse than students on the Durham, Peel or York sides (which have less choice).

11 I do not use the boundaries of Kawartha DSB or Upper Grand DSB because: (a) their borders do not necessarily follow the Catholic school board boundaries and (b) there are very few matched DA observations.

12 Matched DAs along these three boundaries account for roughly 90% of all matches in the sample. Matches along other boundaries (i.e. Halton–Hamilton) have too few observations for meaningful comparative tables.

13 Control variables for all regressions are: DA's average household income, Percentage of DA population holding a bachelors degree, Percentage of DA immigrants, Percentage of DA recent immigrants, Percentage of DA speaking English at home, Percentage of DA of Southwest Asian origin, Percentage of DA of East Asian origin, average dwelling value in DA, population density of the DA's fsa, percentage of DA of Catholic religion and year dummy variables.

14 I do, of course, control for the obvious factors such as population density, parental education levels and income.