Past studies in the literature of the law of one price (LOP) show statistically significant but economically subtle roles of geographical distance in regional price dispersions. In this paper, we challenge this empirical “death of distance” as a primary source of LOP violations investigating a unique daily data set of wholesale prices of agricultural products in Japan that enables us to identify source regions and observe product-delivery patterns to consuming regions. We build a simple structural model to explain the observed product-delivery patterns and argue that ignoring the underlying delivery choice results in a serious under-bias toward inferences on distance effects on regional price dispersions due to sample selection. Estimating a sample-selection model, on which theoretical restrictions of our structural model are imposed, with data of several agricultural products, we find quite large estimates of the distance elasticity of price differential compared with conventional estimates. This paper, hence, provides evidence that conventional estimates of the distance elasticity could be heavily biased downwards and spuriously underestimate the role transportation costs play in regional price dispersions and LOP violations.