Class size limits are a common policy tool to regulate teacher workloads and improve learning. This paper examines their effects on classroom organization, focusing on the use of mixed-grade classrooms as a compliance strategy. I develop a model to study schools' organizational choices under different class size regimes and test its predictions drawing on rich administrative data and two policy changes that modified class size limits in British Columbia (BC). Empirical evidence from three quasi-experimental designs show that schools in BC primarily respond to classroom-level caps by forming mixed-grade classes rather than hiring additional teachers, with policy incentives unintentionally driving their large-scale adoption over time. These organizational responses undermine student achievement and create inequities in access to single-grade instruction as schools strategically manage enrollments. The results reveal a policy trade-off: classroom-level caps encourage the formation of mixed-grade classes, while average caps across classrooms mitigate this incentive by allowing greater flexibility. The findings underscore the importance of anticipating organizational responses in education policy design.

Public alternatives to traditional public schools are often viewed as a means to enhance school productivity, but also raise concerns that they may increase socioeconomic segregation and inequality. Studying these possible consequences of expanded school choice requires examining both the direct impact on students who choose these alternatives and the indirect impacts on those who do not. This paper examines these effects by studying the expansion of French Immersion programs in British Columbia. My analysis uses rich administrative data covering all public and private students in the province, linked to parental tax records and students' long-term outcomes. I characterize the direct and indirect effects of expanding choice through the lens of unordered choice models. Using the distance to new programs as a source of variation in exposure, I implement a heterogeneity-robust difference-in-differences estimator that leverages the quasi-random timing of program introductions. The introduction of new French Immersion programs draws students primarily from regular public schools and, to a lesser extent, from private schools, with enrollment shifts more likely among higher-income families and English-speaking households. Students in areas with large program-induced increases in French Immersion enrollment show significant gains in Grade 7 reading and math scores, as well as higher rates of full-time post-secondary attendance at age 19.

Misspecification is theoretically linked to failures in belief updating, but empirical evidence remains scarce. Using a field experiment in a university course, we show that misspecified beliefs are a major barrier to accurate updating. Students remain overconfident despite receiving informative test scores, and they substantially overestimate the noisiness in test scores.  A randomized controlled trial providing impersonal information about test score noisiness significantly improves students' predictions, closing up to one-third of the gap with a Bayesian benchmark. These results show that misspecification is an important constraint on belief updating but can be mitigated through information interventions.

This paper investigates why food insecurity has risen in recent years while poverty rates have remained stable. Interpreting this divergence is challenging because the two measures differ in both timing and conceptual scope: in the Canadian Income Survey (CIS), food insecurity is assessed over the past 12 months, while poverty is based on income from the previous calendar year, and a poverty indicator alone does not capture the broader distribution of financial vulnerability. Using newly linked administrative tax data and CIS food insecurity measures, we align income and reporting periods to assess the extent to which measurement differences explain the divergence. We then apply a Shapley decomposition to separate changes driven by income distribution from those arising from shifts in the conditional risk of food insecurity. We show that misalignment in the official CIS measures weakens the observed link between poverty and food insecurity, but correcting it does not explain the recent divergence between their trends. The results indicate that rising food insecurity primarily reflects growing vulnerability among non-poor households, particularly those with limited assets.

Special needs education plays an important role in  addressing the unique needs of students with disabilities. This paper investigates how funding models shape the provision of special education through two policy changes in British Columbia. The first reform eliminated targeted funding for specific special needs categories, creating variation that sheds light on the choice between a reporting-based model, where funding is tied to each identified student, and a census-based model, where funding is based on total district enrollment. I find that the removal of funding incentives led to lower identification rates and reduced services for affected students within the public sector. The second policy change increased funding for identified special needs students in independent schools to match the levels provided in public schools. I analyze how these increased funding incentives impact the sorting of students with disabilities between public and private schools, as well as their educational outcomes. This paper sheds light on the critical role that financial incentives play in the provision of education for students with special needs.

I investigate the impact of a large-scale class size reduction policy in British Columbia, triggered by a 2016 Supreme Court of Canada ruling. The ruling mandated school districts to reinstate historical class size limits, initially set decades prior and removed in 2002, and led British Columbia to allocate increased funding to facilitate these changes. The policy's implementation provides a natural experiment, where historical variation in class size rules across districts creates plausibly exogenous differences in treatment intensity among schools and districts. Using an event study design, I estimate the effects of this policy on student outcomes and examine how the effects differ across schools with varying baseline academic performance and concentrations of disadvantaged students.