This paper proposes a novel approach for the empirical analysis of multi-unit auctions, to which participants submit supply or demand functions observable by the researcher. The approach allows for the evaluation of firm-level market power in a private information setting, and avoids having to model the market mechanism. It relies on econometric methods that treat the observed bid functions as function-valued random elements. Notably, a functional instrumental variable estimator is developed. The method is applied to the New York electricity
market using rich data on firm-level bids and marginal costs for 2013-2015. In this market, daily bids are disclosed three months later in order to limit strategic behaviors. I estimate firm-level market power and compare actual bidding behavior to profit-maximizing behavior under private information. I find consistent evidence of optimal bidding, suggesting that firms are well aware of their own market power and behave accordingly. Therefore, the late disclosure of bids is not sufficient to preclude firms from acting strategically, most likely due to the
repeated nature of those auctions.