We revisit the question of why some participants in online auctions place their bids right before the time of closing. Using e-Bay data for two product categories, antiques and iPods, we propose count-data models to look at both the presence of the late-bidding phenomenon and its intensity. Our results reveal a notable difference when we compare extremely late bidders or snipers to moderately late bidders. The methodological implication is that the conclusions depend on the choice the late-bidding time window. Further, whereas the literature considers all bids placed during this time window as late bids, we only retain those bids made by participants who did not bid before reaching this time window. We believe that our approach is more consistent with the theoretical arguments provided to explain late-bidding behavior.