170 likes | 274 Views
Reexamining Education Fairness: An Experimental Study of College Admission Policies in China. Yang Zhang May 7, 2013. College Admission Policies in China. College Entrance Exam ( Gaokao ) A nation-wide comprehensive exam in every June
E N D
Reexamining Education Fairness: An Experimental Study of College Admission Policies in China Yang Zhang May 7, 2013
College Admission Policies in China • College Entrance Exam (Gaokao) • A nation-wide comprehensive exam in every June • Universities only consider Gaokao scores in recruiting new students. • College Admissions • Competition for college education is limited within each province because of the quota system. • Local variations in college admission policies • Immediate acceptance (most provinces) vs. deterred acceptance (e.g. Shanghai). • Exam content • Timing of application
Timing of Application Gaokao Exam Result Distribution • Low information • Not knowing your own and others’ Gaokaoscores. • Private/asymmetric information • Estimating your own score based on the official answers, but not knowing others’ scores. • Complete information • Knowing your rank among all students.
Research Questions • How does timing of application affect students’ strategies in college application? • How does timing of application affect the distribution of admission outcomes?
Three Games 1. Mimic Exams 2. College Entrance Exam • Randomly pick one mimic exam score.
Three Games Cont’d 3. College Application • Choose one university from Harvard and the UI. • Get 2 dollars from Harvard and 1 dollar from the UI. • Three games only differ in the timing of application. 4. College Admission • Each university only admits one student.
Defining Admission Fairness • Expected Rank Distribution • Ideal Outcome Distribution
Available Information in College Choice (e.g Student A) • Complete Information • Expected rank distribution (unnecessary) • A, B, and C’s Gaokao scores • Private/asymmetric information • Expected rank distribution • A’s Gaokao score • Low information • Expected rank distribution
Experiment • Subjects • 10 groups and 30 subjects • 21 Chinese (15 experienced), 6 Americans, 1 Taiwanese, and 2 Koreans • 8 groups for each treatment (complete, private, and low information) • Post-game discussion • Which games are fair and which are not fair? • What was your expectation of other players’ choices (belief)? • Does the experiment well reflect the reality? • Other thoughts?
Complete Information Game: Evaluating the Expected Strategies • Nash Equilibrium(a) • Observed Strategy Distribution
Complete Information Game: Evaluating the Expected Admission Outcomes • Expected outcome distribution • Luck in getting a higher score • Observed outcome distribution • Adjusted outcome distribution
Private Information Game: Evaluating the Expected Strategies
Private Information Game: Evaluating the Expected Admission Outcomes • Expected outcome distribution • Adjusted outcome distribution
Low Information Game: Evaluating the Expected Strategies • Bayesian Equilibrium • Observed Strategy Distribution
Low Information Game: Evaluating the Expected Admission Outcomes • Expected outcome distribution • Adjusted outcome distribution
Conclusions • Observed Strategies • Complete information: the experimental results perfectly fit the Nash equilibrium that A chooses Harvard, B chooses the UI, and C is indifferent between the two. • Private information: A is more likely to choose Harvard, and B is more likely to choose the UI no matter which score they pick. C chooses the UI most of the time. A bigger sample is needed to further evaluate the PBEs. • Low information: As the Bayesian equilibrium suggests, A always chooses Harvard and C always chooses the UI. B is risk-seeking that she applies for Harvard most of the time. • Adjusted Outcome Distributions • Complete information yields ideal outcome distribution. • Private information greatly favors C and slightly favors A at the cost of B. • Low information is biased against both A and B, but in favor of C, which contradicts the Bayesian equilibrium.
Future Research Directions • Under deterred acceptance algorithms, how will different levels of information affect college choices? • Two-sided matching game • Statistically evaluate the theoretical predictions with agent quantal response equilibrium. • A problem of chi2 test. • Policy innovation and diffusion in an authoritarian country. • “The adoption of a design is at least partly a political process. (Roth 2002: 1345)”