In this paper we propose an alternative to traditional hedonics for estimating new multiunit housing inflation, adjusting for quality changes. By relying on the within-site variation we control in a very general way for unobserved housing characteristics using site-specific effects. Precise location, transport, traffic, closeness to services, or construction quality are some of the unobserved but typically relevant housing characteristics that may bias estimated house price inflation, even when using hedonic methods. We also estimate standard hedonic equations and compare the results to those obtained with the alternative hedonic equations with site dummies. Our dataset is fairly rich in observable housing characteristics but, nevertheless, the quality-adjusted house price evolution is quite different in some cases. The data cover the construction of new housing in some of the large Spanish cities and in the smaller towns on the outskirts of the capital during part of the 1990s.