This paper examines the effects of incentivizing industrial users to reduce their electricity consumption using demand response auctions, in which rewards for curtailment depend on auction outcomes. Because true baseline consumption is unobserved, firms can strategically adjust both bids and consumption, leading to upward‐biased estimates of program effectiveness. Using data on bids, auction outcomes, and hourly electricity consumption from steel producers in Taiwan, this paper employs a regression discontinuity design to show that not accounting for firms' strategic bidding behavior can lead to an overestimation of electricity reductions by at least 50%.