This article tests whether the correlation between wages and concentration of employment can be explained by unobserved worker productivity. Residential location is used as a proxy for unobserved productivity, and average commute time to workplace is used to test whether location-based productivity differences are compensated away by longer commutes. Analyses using confidential data from the 2000 Decennial Census find that estimates of agglomeration wage premia within metropolitan areas are robust to comparisons within residential location and that estimates do not persist after controlling for commuting costs, suggesting that the productivity differences across locations are due to location, not individual unobservables.