Abstract
To what extent can offering more choice in schooling lead to ‘cream skimming,’ or the sorting of students by ability? I study whether increased choice leads to student sorting by ability into high schools in the Greater Toronto Area. On average, 41% of students ‘opt out’ of the high school to which they would normally be assigned based on their residence. Students are more likely to opt out in areas where accessibility to other schools is greatest due to population density and explicit ‘open enrolment’ policies. While students of higher ability are generally more likely to opt out, an interaction term between school choice and ability is insignificant, suggesting that increased choice does not have differential impacts by student ability. Findings are robust to changes in assumptions about instrument exogeneity.
Acknowledgements
I wish to thank my thesis committee of Abigail Payne, Martin Dooley and Stephen Jones for their ongoing attention, support and suggestions. Thanks also to the Education Quality and Accountability Office for making their data available through the Public Economics Data Analysis Laboratory (PEDAL) at McMaster University.
Notes
1 In fact, students may also have the choice of a French secular or French Catholic school board. However, as the number of students attending French schools is quite small, and since most students do not meet the French-speaking requirements of the French schools, I focus only on English schools in my analysis. Likewise, I ignore private schools and students, who account for less than 5% of enrolments in Ontario.
2 See Belfield and Levin (Citation2002) for a thorough US review of the impacts of increased school choice on student outcomes. There continues to be a large amount of interest in this topic including Hoxby (Citation2003), Greene and Kang (Citation2004), Bettinger (Citation2005), Figlio and Rouse (Citation2006), Imberman (Citation2011), Marlow (Citation2010) and Figlio and Hart (Citation2014). Card et al. (Citation2010) and Leonard (Citation2015) study the issue for elementary schools and secondary schools, respectively, in Ontario, Canada. Lefebvre et al. (Citation2011) study competition between private and public schools in Quebec, Canada.
3 In addition, there are 12 French school boards which have a relatively small attendance. I exclude these boards in my analysis since most Toronto-area students (i.e., those without French-speaking parents) are not able to attend.
4 However, Catholic Boards vary in terms of what is acceptable proof of Catholicism.
5 Students with no Grade 9 match have a mean Grade 6 math score of 2.36, which is approximately one-half of an SD below the mean score of 2.77 for the full sample. However, the test scores of these students are not significantly different from those students who write the Grade 6 test, but do not write the Grade 9 test (but for whom I can match the Grade 9 school).
6 The first wedge extends due north from each school and subsequent wedges are defined in a clockwise direction. The counts of travel zones are fairly robust to starting in alternative directions.
7 These students account for 104 observations on the math test and 483 observations on the reading test. Since there are so few students with scores of 0, excluding or including them makes no difference to the results.
8 Appendix 1 shows that within 5 km from a school board boundary, students and neighbourhoods are otherwise quite similar at different distances from the board boundary.
9 I have also estimated both logit and probit models with no changes to the signs or significance of coefficients.
10 Appendix 2 shows similar regressions to where each test score (i.e. scores of 0 through 4) is entered as separate dummy variables rather than a single linear score. Clearly, students who score a 2 (below provincial standard) opt out less than students meeting the provincial standard (a score of 3) and students scoring a 4 (above provincial standard) opt our more. Due to small sample sizes, not much can be said about the students scoring 0 or 1 on the test.