Japan college-access model shows how moving barriers can widen regional gaps

Students living near a wide selection of colleges can pursue higher education without leaving home. Those elsewhere may face relocation expenses and the emotional burden of moving away. A preprint by Se-il Mun, Shun Fujita and Lei Qin investigates how these differences could shape college enrollment across Japan.

Their central finding is conditional but striking: eliminating moving costs in a simulated college market substantially narrows regional enrollment gaps. Yet competition for limited places also changes, leaving some metropolitan students worse off. The results are model-based counterfactuals, not observations of an implemented policy.

How geography enters the admissions model

The researchers build an equilibrium model covering Japan’s 47 prefectures. Students choose whether to attend a four-year college and, if so, which one. Their modeled choices balance educational quality against tuition, moving costs and the effort required to reach an entrance-exam threshold. An additional random preference component allows otherwise similar students to choose differently.

Moving costs increase with distance. They represent generalized burdens, including financial expenses and nonfinancial difficulties such as separation from familiar support networks—not simply transport bills.

The model distinguishes assumed underlying ability from exam performance: students with lower ability can reach a given qualifying score by investing more effort. Ability is not measured in the data. Instead, the researchers impose the same truncated normal distribution, described using an IQ scale, in every prefecture.

Colleges have fixed tuition and enrollment capacity. Their qualifying scores adjust with demand, acting as a rationing mechanism for scarce seats. Higher thresholds increase preparation costs, but also raise modeled educational quality through an assumed peer effect. Educational resources provide another quality component, represented by an index combining research grants and student–teacher ratios.

For computation, the authors aggregate institutions into 107 groups by location, public or private status, and selected quality distinctions. They estimate parameters by repeatedly solving the admissions equilibrium and minimizing discrepancies between predicted and observed prefecture-to-prefecture enrollment flows. Inputs combine several years: income and research-funding measures are from 2019, while capacity data primarily come from 2021, with 2020 substitutes where needed.

A narrower gap, with winners and losers

The model reproduces some observed patterns more closely than others. Predicted and observed enrollment flows have an overall correlation of 0.908, but the correlation falls to 0.668 for movement between prefectures, compared with 0.987 for enrollment within a prefecture. The authors also report that the model overestimates inflows into several major-city prefectures, potentially reflecting omitted deterrents such as higher living costs.

The main experiment sets moving costs to zero and recalculates admissions thresholds. Across prefectures, simulated enrollment rates range from 16.2% to 71.7% in the baseline, compared with 33.5% to 66.3% without moving costs. These are simulated ranges, not a before-and-after comparison in real enrollment data.

For example, Iwate’s modeled enrollment rate rises from 27.2% to 51.4%, while Tokyo’s falls from 71.7% to 66.3%. Removing geographic barriers allows more students from poorly served regions to compete for attractive institutions. Because capacity remains fixed, access is redistributed rather than expanded through new seats.

The authors value changes using compensating variation: a monetary measure of the difference in expected utility between the two scenarios. They report a total gain of ¥21.25 billion, or about ¥19,500 per prospective student on a population-weighted basis. This is a modeled welfare equivalent, not an observed increase in earnings or household income.

Regional outcomes differ sharply. Average gains reach approximately ¥590,000 in Shimane, whereas Tokyo records an average loss of about ¥330,000. In the model, stronger competition raises preparation costs for students who previously enjoyed privileged proximity to high-quality colleges.

Why the policy interpretation needs caution

The paper separately reports an unweighted average welfare gain across prefectures of ¥170,000. That is not the national per-student effect: giving each prefecture equal weight produces a different answer from weighting by its student population.

A hypothetical subsidy offsetting the modeled distance-related burden would total about ¥2.424 trillion. The authors regard this as impractically large. However, eliminating psychological and other nonfinancial burdens through payments is itself a strong assumption; the exercise does not establish the effects of a realistic subsidy program.

Other limitations matter. Aggregate data cannot reveal detailed household differences, and students are assumed to obtain the score needed for admission without uncertainty. Peer effects permit multiple equilibria, so the numerical procedure starts from observed qualifying scores and allows relaxed capacity tolerances. Reported tiny standard errors are described as local numerical diagnostics, not robust confidence intervals.

The preprint therefore offers a framework for examining geographic access—not a definitive policy forecast. Its clearest message is that helping students move can change who gets existing college places, while the scale and distribution of benefits depend on admissions competition and the model’s assumptions.