We examine bidding behavior in a clock auction in which price is set by the lowest-accepted bid and provisional winners are reported each round (the LABpw auction). This format was used in the India 3G spectrum auction. In the standard theory, the auction performs poorly. In particular it yields lower revenues and is less efficient than the more standard clock auction with exit bids and highest-rejected-bid pricing (the HRB auction). However, the LABpw auction performs well in the lab, achieving higher revenues than the HRB auction.