Late bidding is one of the key phenomena in online consumer auctions. Existing studies suggest that there is no strategic reason to bid late in soft-close auctions, and that late bidding should be much less prevalent in such auctions than in hard-close ones. This dissertation shows otherwise, both theoretically and empirically. In the first essay, I extend the existing hard-close sniping theories with an assumption such that each bidder has a positive probability to be absent at the end of an auction. I show that, in all the major existing theoretical models, strategic late bidding arises in equilibrium in soft-close auctions. In the second essay, I use a unique, year-long natural experiment consisting of jewelry auctions offered by a same seller on eBay and Overstock Auctions to investigate the effects of auction closing rules on late bidding. I show that late bidding is not only prevalent in the soft-close auctions but significantly more so during certain stages of an auction than in the hard-close ones. In the third essay, I use a separate large-scale natural experiment made possible by private listings on eBay to test the expert theory of sniping but find no consistent evidence.