Distance is the spice, but not the whole enchilada: Country-pair psychic distance stimuli and country fixed effects in a deep learning implementation of the trade flow model

Deep learning is used to analyze temporal trade flow data from 62 countries from 2017 to 2021. The model incorporates 63 explanatory country fixed effects and country-pair psychic distance stimuli. This advanced computer-age statistical approach goes beyond the limitations of traditional OLS regression. The model demonstrates that country fixed effects contribute at least as much to the variations in trade flows as do the distance-related factors. The study also shows that distance stimuli related to democracy, education, and religion do not negatively influence trade flows. Remarkably, the deep learning model can effectively train itself solely on country fixed effects. This prompts a reevaluation of the classic trade flow gravity model, which typically places heavy reliance on distance-related variables.