Experimental Design Details
We are going to collect data from an online auction website, Catawiki.com. The website is specialized in buying and selling novelty items and collectibles such as stamps, arts, antique cars, gemstones, etc. We specifically collect data from mixed diamond auctions, as these items are frequently offered on Catawiki.com and can be categorized as common value items. Moreover, we focus on auctions that do not list a reservation or minimum price for which the item will be sold (so that our low bids are credible). All auctions in our sample run approximately one week starting from Friday at noon. For each auction, we record the item’s specifications, the minimum and maximum value of the item estimated by Catawiki’s experts, and all bids along with each bidder’s website identifier from the beginning to the end of the auction. Prior to the start of the auctions, we create a sample by selecting all relevant auctions, e.g. mixed diamond auctions without a minimum price, from the list of “auctions opening on a specific date”. Within each trio of matched auctions, we then randomly assign each auction to one of our three experimental treatments. As soon as the auctions begin, we place the first bid on the items based on the rules of the treatments they have been randomly assigned to. In treatment 1, our control condition, we do not intervene in the bidding. In treatment 2, we place a jump bid. In treatment 3, we place an automatic bid, which implies that Catawiki’s bidding system will place incremental bids up to our bid on our behalf. The amount we bid will be uniformly randomly drawn from the set {30, 35, 40, 45, 50}. After all auctions have ended, we request the platform for the bidding data. If we happen to win an auction despite our low bids, we buy the item for the price that we have bid, plus 9% buyer’s fee and the shipping costs. We will then attempt to resell it to a local diamond purchaser.