This paper provides estimates of the net depreciation rate for rental housing using a unique confidential data set from the Bureau of Labor Statistics that covers over 30,000 rental units from 1998 to 2009. Our data and econometric approach allow us to add to the literature in three main ways. First, we can control for unobserved quality (including cohort effects) by allowing for unit-specific fixed effects. Our results suggest that estimates of the depreciation rate for rental housing that ignore unobserved heterogeneity suffer from omitted-variable bias and potentially from selection bias, and that these biases can be large. Second, we use a dummy-variable approach to estimate aging effects, thereby avoiding ad hoc assumptions about functional form. We find that rent for a typical housing unit at first falls rapidly with age, flattens, and then begins to rise with age for older units. This nonmonotonic pattern is a feature of many other studies of both age–rent and age–price profiles for housing, and it seems likely that the upward-sloped portion of such profiles is the result of unobserved improvements and changes in style. We show that the upward slopes of our estimated profiles are only partially eliminated when we attempt to control for major improvements by excluding housing units that see very large jumps in rent. Third, we present estimates of age–rent profiles separately for different types of structures and for different regions of the country.