We explore the estimation of origin-destination (OD), city-pair, air passengers, in order to explicitly take into account spatial autocorrelation. To our knowledge, we are the first to test the presence of spatial autocorrelation and apply spatial econometric OD flow models to air transport. Drawing on a world sample of 279 cities, over 2010-2012, we find significant evidence of spatial autocorrelation in air passenger flows. Thus, contrary to common practice, we need to incorporate the spatial structure present in the data, when estimating OD air passengers. Importantly, failure to do it, may lead to inefficient estimated coefficients and prediction bias.