Anti-gravity Underground?

Since its introduction in 1946, the Gravity Model has proven to be a valuable aid in understanding and predicting forces at play in human mobility. Historically, the model has been applied primarily at the intercity and interstate level. In this paper we investigate whether the scope of the model extends to the intra-urban. Using a large dataset of trips made on the London rail network, we derive a number of proxies for mass and distance, and using these proxies we examine the extent to which the estimated passenger flows of the Gravity Model fit the observed data. We find that there is a good correlation between the estimates and observations, with a Pearson’s correlation coefficient (PCC) of 0.720. We then extend the investigation to examine how well the model fits when focusing on trips of differing purpose, and we also modify the model to take into account the attractiveness of a location. We find that the original model better fits leisure trips than commute trips, but that the modified model closes this gap, and indeed offers an improvement overall, increasing PCC up to 0.789. We close with a discussion of the validity of our results and the direction of future research.