Abstract
China’s National College Entrance Examination (NCEE, also known as Gaokao) system has transitioned to a "major-oriented" admission model under recent reforms. However, persistent information asymmetry—particularly among socioeconomically disadvantaged students—undermines rational decision-making. Misconceptions about career prospects, limited access to labor market data, and inadequate college application strategy guidance exacerbate mismatched higher education choices, perpetuating educational inequities. Behavioral economics suggests that nudge interventions can improve decision-making without restricting autonomy. Prior studies demonstrate that targeted information provision (e.g., earnings statistics, career trajectories) significantly influences educational choices, especially for disadvantaged groups. This RCT would evaluate the efficacy of personalized information interventions in optimizing NCEE application and improving long-term academic/professional outcomes.
This RCT is conducted in partnership with public senior high schools in Yunnan province. Approximately 2,000 graduating students (2025 NCEE cohort) will be randomized into treatment and control groups. The intervention would be consisted with two main components. The first component is the baseline information intervention, which would be embedded in the questionnaire survey before application procedure begins. The treatment group would receive the labor market statistics (e.g., discipline-specific earnings from China College Graduate Employment Reports or China Labor Statistical Yearbooks), while control group would receive nothing. The second component is the personalized strategy guidance. All survey participants are randomly assigned to two groups and requested to submit draft application forms on Days 1, 2, and 3 during the application period. For each participant in the treatment group, they would receive a machine learning (ML) algorithm generated tailored feedback reports (e.g., risk-adjusted major recommendations) based on their Day-1 draft on Day 2. To ensure equity, control group participants receive identical feedback reports on Day 3, based on their Day-2 drafts.
This study employs a multi-phase data collection framework to capture both immediate and longitudinal outcomes. Primary outcomes focus on observable behavioral changes during the higher education application process, including revisions between application form drafts and finalized application forms, as well as ultimate admission results (institutions and majors secured). Secondary outcomes extend to post-enrollment experiences, measured through surveys assessing academic satisfaction, perceived alignment between chosen majors and personal preferences, and career preparedness. To elucidate underlying mechanisms, pre- and post-intervention questionnaires will systematically track shifts in participants’ beliefs about academic and career prospects, their utilization of provided information, and decision-making strategies during the application process.
Causal effects of the intervention will be estimated through intention-to-treat (ITT) analysis, which compares application strategy modifications and admission outcomes between the treatment and control groups while preserving randomization integrity. Additionally, a machine learning-driven, cost-benefit framework simulates counterfactual scenarios using provincial admission datasets. This approach quantifies long-term labor market disparities attributable to the intervention by modeling earnings trajectories under different major-institution pairing scenarios. Sensitivity analyses will further validate robustness across heterogeneous subgroups.
To mitigate equity concerns, all participants—including controls—receive full intervention materials after trial completion (before completing final college application). Personally identifiable information (e.g., names, national ID numbers) is rigorously excluded from datasets, with encryption protocols adhering to China’s Statistics Law. Collaborative oversight by partner schools ensures ethical compliance, guaranteeing that experimental procedures neither disrupt standard application workflows nor disadvantage any participant group.
By synergizing behavioral nudges with machine learning, this study represents a novel methodological advance in China’s education policy research. Its findings hold transformative potential for designing scalable, low-cost interventions that address systemic inequities in higher education access.