Abstract: We study how collocation — geographic clustering of business establishments owned by the same parent company — influences the productivity of establishments over time. Using Census micro data on the full population of U.S. hotels from 1987-2007, we find that doubling the intensity of collocation is associated with a productivity increase of about 2% in pre-existing establishments. We use changes in collocation to tease out how much of the effect is persistent versus contemporaneous, and find that a significant component of the productivity gain persists after collocation ceases. These and other results are consistent with the idea that old establishments learn from new establishments. We replicate our main findings on Census micro data on the full population of U.S. restaurants from 1987-2007, suggesting that the collocation effects we document in the hotel industry may generalize to other industries.