Using a new dataset on household purchases of personal computers (PCs), we document positive correlations between buyers' incomes and the prices they pay for seemingly identical PCs. These results suggest that firms may be successful at separating the market and charging different prices to consumers with different levels of willingness to pay. We consider the implications of this kind of market separation for price and quality measurement via a theoretical model based on Mussa and Rosen (1978). The model suggests that, in markets like these, standard methods that do not account for this heterogeneity can understate inflation in a cost-of-living context. Consistent with the model, our empirical work shows that controlling for income yields indexes that show slower price declines than seen in standard indexes. This understatement of the cost-of-living measure likely mitigates the unrelated upward biases found in recent studies by Bils (2009), Erickson and Pakes (2010), Broda and Weinstein (2010).