Predicting Urban Growth with Machine Learning

Abstract

Using machine learning (ML) models, we predict the population growth for the next two, five, and ten years for American and Chinese urban areas. To this end, we construct a rich city-level data set encompassing information on transportation, output, amenities, and human capital. The ML models choose the main urban growth predictors through cross-validation. We find that human capital, real estate investment, amenities, and geographical lo-cation are strong future urban growth predictors for both countries. In the US, intercity transit is a stronger urban growth driver than intracity transit, whereas the opposite holds for China. Our models predict that coastal urban areas in the South Atlantic and the West South Central Division in the US, and southeastern coastal cities in China, will grow the most.

Keywords: City growth

JEL Classification: R1, R12, R15, R4