In many markets, buyers make discrete choices between differentiated products and negotiate prices that are specific to the choice. We develop for estimation a model for this class of markets which is consistent with non-cooperative models of bargaining between a buyer and competing sellers. We show that when the buyer’s utility has GEV disturbances, the model has a tractable likelihood function which can be used with transaction-level data giving the selected product and its price. We estimate the model using data from the UK brick industry and use it to measure market power and analyze mergers. We analyze how spatial differentiation and ownership concentration affect the distribution of market power across transactions. In counterfactuals we find that switching from individually-negotiated to uniform pricing causes markups, and merger price effects, to increase on average but to decrease for a minority of transactions.