Association for Public Policy Analysis & Management

To address the college undermatch problem among low-income students, there have been growing demand-side efforts to improve students’ decision making in college choice and application, including informational interventions and nudges (e.g., Bettinger et al., 2012; Hoxby & Avery, 2013; Page & Scott-Clayton, 2016). There is still a dearth of evidence of the potential impacts of institutional factors from the supply side. To fill this gap, we apply matching theory and a national sample of 30 million high school graduates in China to examine how admission matching mechanisms generate/reduce inequality in college access and choices between low-income students and their high-income counterparts.

In the past fifteen years, school choice mechanism has been one of the most innovative and important education policies to better allocate K-12 students in U.S. public schools (Abdulkadiroğlu & Sönmez, 2003; Pathak & Sönmez, 2013). In higher education, decentralized assignments have been increasingly replaced by centralized ones worldwide (Machado & Szerman, 2016). Centralized college admission system operates in quite a similar way as U.S. K-12 school choice, which also calls for the improvement of matching mechanism. College application and admission process requires students to strategize their choices. Owing to a lack of access to information and guidance, this institutional feature disproportionally leads disadvantaged students to make uninformed choices than their advantaged peers.

China has the largest centralized matching market in the world. College admissions in China operates through the national College Entrance Exam, the scores of which are the sole criteria used to rank students. This process helps researchers identify the impacts of matching mechanisms on student choices and admissions by controlling for the only confounding factor (academic achievement), which is not possible in decentralized systems. More importantly, there are large temporal and regional variations in the Chinese province-centralized systems. We take the advantage of these exogenous variations to evaluate the causal policy impacts of different matching mechanisms (e.g., Boston mechanism, the deferred acceptance mechanism).

We use a unique administrative dataset on every student who entered college in China from 2005 to 2012. The dataset contains basic information on student background, college entrance exam scores, and college-major admissions results. We merge this student-level dataset with extensive county-level and provincial level covariates over time, including the full set of college applications and admissions rules. Our baseline specification is the difference-in-differences-in-differences (DDD) model that accounts for the heterogeneous impacts of matching mechanisms on students in different income and achievement distributions. We show supplemental IV-DDD results by exploiting a national policy nudge by the central government in 2008.

Existing theoretical and lab experimental studies primarily examine the differences between sequential and parallel mechanisms among all students (Chen & Kristen, 2017; Chen et al., 2015), this paper provides novel empirical evidence of a wide range of heterogeneity in matching mechanisms and of the differential impacts on low-income students, which has not been thoroughly examined. This paper will significantly contribute to the literature of school choice mechanism design and the on-going policy efforts of improving college choice for disadvantaged students.