11/ We revisit the city-size premium using our estimated location effects and find an elasticity of 0.007. This is 2-3 times lower than estimates not controlling for establishment composition.

9/ MAIN RESULTS: location effects account for only 2-5% of spatial wage differences. Worker and establishment composition for 30% and 17%, respectively. Remaining half stems from the co-location of high-ability workers & high-paying firms, particularly in high-paying locations.

8/ We provide a series of tests that confirm the validity of our design, including quartile by quartile event study graphs, placebo tests and a survey of entrepreneurs.

7/ Identification: “double-movers” design, leveraging both workers who move across establishments and locations, and establishments that relocate across CZs.

6/ We estimate a three-way fixed effects model that accounts for (a) worker ability, (b) establishment productivity / pay policy, and (c) location-specific factors—such as infrastructure and agglomeration forces.

4/ Establishments which relocate to another commuting zone replace ~50% of their workforce. Establishments relocating to a 1% higher paying location increase pay by ~0.15%.

3/ Establishments which relocate maintain the same activity, organization, and overall input composition.

2/ In both France and the US, 4% of establishments relocate every year. 25% of those moves involve a change in commuting zone (CZ).