Typical plant-level data sets do not report quantities. This paper shows that estimating mark-ups (price-cost ratio) in product-differentiated industries using deflated sales to proxy quantity is not appropriate due to unobserved price heterogeneity. This paper presents an econometric model for estimating mark-ups that controls for unobserved prices. The model shows that ignoring price heterogeneity results in mark-up estimates that converge to one, whatever the value of the true price-cost ratio. Estimates obtained using real data are consistent with this result, as they reveal that ignoring price heterogeneity leads to spurious evidence of firms with little or no market power.