Measuring Remoteness Using a Data-Driven Approach

Datasets of schools or hospitals often include an urban-rural divide drawn by government. Such partition is typically determined by subjective thresholds for a few variables, such as access to transportation and local population size, leaving aside relevant factors despite datavailability. We propose to measure eremoteness f by mapping a comprehensive set of covariates onto a scalar, and define an objective score of remoteness using a standard selection model. We apply the proposed method to data from Taiwanese public elementary schools. Our method replaces 35% and 47% respectively of the current official list of "remote" and "extra-remote" campuses, shifting the remoteness designation to those furthest from train stations, having the highest teacher vacancy percentages, and located in the least populous areas with the least well-educated populations. The campus- and district-level variables used are publicly available and periodically updated in most advanced economies, and the statistical model can be easily implemented.