The objective of the research in this study is to evaluate the pricing and market conduct of milk manufacturers and retailers. Using data from a U.S. Midwestern state, we estimate a random coefficient logit demand model (RCL) to empirically investigate a range of possible scenarios in the milk supply chain. These include vertical leader-follower model with underlying Bertrand-Nash pricing, models allowing for nonlinear pricing contracts, and collusion scenarios at various levels in the supply chain. This study contributes to the literature in the following ways. First, it generalizes the RCL demand model via Box-Cox power transformation. While previous studies rely on ad hoc specified linear indirect utility, this procedure allows data to determine the functional form of utility. Power transformation parameters cannot be obtained analytically with product-level data, given that consumer choices are unobserved. We propose an algorithm to estimate the transformation and consumer heterogeneous taste parameters sequentially. The model is identified using annual variation in consumer demographics along with cross-sectional and time series variation in milk consumption. Finally, the milk choice set is allowed to vary across markets. It should be mentioned that jointly estimating the manufacturing sector, the vertical channel, and the retail sector will more likely yield reliable estimates of structural parameters vis-à-vis studies investigating food supply chain sectors in isolation. Several key results are obtained from the research. First, the estimate of demand “superelasticity”suggests that retailers have incentives to adjust retail markups to enhance their market power. Next, supply selection bias associated with imposing restriction on the demand-side framework is shown to have formidable policy implications. Namely, empirical results from the general demand show that retailers are more powerful than they would appear otherwise. In the face of high concentration and a small presence of Wall-Mart in these markets this seems a plausible scenario.