Causal Effect of the Shinkansen on Population Change in Japan: An Application of PSM-DID

The Shinkansen is a high-speed railway in Japan; it has been in operation since 1964 and has been used for approximately 60 years. Shinkansen has improved population mobility and communication between different regions in Japan, making it easier for people to reach different cities and areas, thereby increasing population balance and diversity. This study uses the Japanese Shinkansen as an example and empirically investigates the relationship between high-speed rail and population change. An empirical analysis using an econometric approach was conducted on 19 newly opened Shinkansen stations in Japan between 1995 and 2015, using panel data from 1995, 2000, 2005, 2010, and 2015. Specifically, we used third-level grid data from the national census and the propensity score matching and difference-in-difference (PSM-DID) method to quantitatively calculate the causal effect of the opening of the Shinkansen on the population increase or decrease in the areas surrounding the Shinkansen stations. We used propensity score matching (PSM) to mitigate self-selection bias caused by observable variables and improve comparability between the treatment and control groups. We then combined the DID (difference-in-difference) method with panel data to eliminate the endogeneity problem caused by unobservable variables and improve the robustness of the estimation results. We introduced a two-way fixed effects model (TWFE) in the DID method to control for time-fixed effects and individual-fixed effects. We controlled for individual-fixed effects by dividing them into three levels—region, prefecture, and city—thus obtaining analysis results from macroscopic to microscopic perspectives. The impact of Shinkansen stations on population change at different levels was analyzed in detail. The event study method was introduced to test the parallel trend hypothesis, the core hypothesis of DID, and the dynamic effect of the Shinkansen opening on population change was further analyzed along with the parallel trend test.