Penny Auctions

Abstract

The low setup cost of online auction platforms has facilitated the creation of new varieties of auction formats. A popular new auction variety invented by a German company ‘Swoopo’ in 2005 requires participants pay a small fee at each time of bidding (Platt et al., 2012). It was named Penny Auction, as each bid increased the current price by a fixed small increment, which was usually one cent. Swoopo’s success has brought hundreds of competitors running similar format pay-to-bid auctions around the world (Zimmerman, 2011). Swoopo remained one of the market leaders from its founding until it filed for bankruptcy in early 2011. At its peak, Swoopo operated internationally in 22 regions (including the United States and most EU countries), earned a profit of $28.3 million in 2008 and had 2.5 million users in 2009 (Oswald 2008; Stone 2009). Since then, more than 150 market entrants including BidCactus, BigDeal, and Quibids have tried to capture a piece of the market (Stone 2010), which swelled to 300,000 U.S. visitors daily in 2011.1 Table 1 shows the numbers of monthly unique visitors of a few of the largest penny auction websites, monitored by Compete.com, a web traffic monitoring company, showing penny auctions website even reached around 25% of eBay’s data at the end of 2010 (Wang Xu, 2016). Different penny auction websites may have slight differences in their auction settings, such as charging different amounts of bidding fees, with some general rules followed by most of the auctioneers. In general, each penny auction starts at a price of zero with a specified closing time displayed in a countdown clock. To place a bid, a bidder is required to pay a small bidding fee (usually between $0.60 and $1), which can either be charged immediately at the time of bidding, or deducted from his/her pre-purchased bidding credits. After a bid is placed, the current price increases by a fixed amount (usually between $0.01 and $0.15), substantially smaller than the bidding fee. After the original closing time is reached, each new bid can extend the auction by a set short amount of time (usually between 10~30 seconds). If no other bid is placed before the time expires, the last bidder wins and pays the current price. It differs from English auctions, where the winner of a penny auction pays not only the winning bid, but also the bidding fees incurred throughout the entire auction, while the auctioneer gains his revenue not only from the winning bid, but also from the aggregate bidding fees paid by all participating bidders. It also differs from ordinary online auction platforms such as eBay and Amazon, where a penny auction website hosts all its auctions and lists selected brand new items of limited variety, such as popular consumer electronics, video game consoles, gift cards and packs of bidding credits, which have a relatively welldefined market value. Penny auctions have drawn academic attention because of their similarity and differences to some well-known auction mechanisms such as Martin Shubik’s Dollar Auction, and War of Attrition and all-pay auctions (Hinnosaar, 2016). A variety of theoretical models based on complete information and risk neutral sellers and bidders predict zero expected profit, but data suggest otherwise. Empirical evidence has shown that bidding fees are primary revenue sources of penny auction operators who earn substantial and consistent profits, robust over time. Although some penny auctions ended with high profit margins, not all auctions were profitable, e.g. Swoopo only made positive profit in half of its auctions (Platt, Price et al. 2010; Byers, Mitzenmacher et al. 2010). A simple example would help us understand the profitability of the mechanism more intuitively. Suppose in some penny auction, bidders pay $1 to place a bid, which raises the current price by $0.02, then for every $1 increase in the winning bid, the seller collects $50 additional revenue from bidding fees. Thus, as long as the winning bid reaches 1.96% of the seller’s valuation, he would break even, and if the winning bid reaches 5% of his valuation, his profit margin would be 200%. Despite its profitability to the auctioneer, a penny auction may still look attractive to many bidders, especially the newcomers, who observe that the current price of an auction item is usually remarkably low and the winning bid of a closed auction is also usually low. Swoopo managed to deliver expensive consumer products at significantly low prices that beat all traditional retailers in most of its auctions, while retaining a high profit margin. Note that although Swoopo gained negative profits in around half of its auctions, the other half were successful enough to result in an overall profit margin of 50% from 166,000 auctions spanning from September 2005 to June 2009. In fact, the median winning bid of Swoopo auctions is only 10% of the retail price of the auction items (Augenblick, 2011). Despite these obvious pros, the tricky con is that costs spent on past bids are technically sunk and have no bearing on one’s likelihood of winning once being outbid, which is one of the reasons that newcomers often spend much more than they plan to. Note that placing a bid in a penny auction has a distinct interpretation from bidding in a traditional auction, in which a bid represents a bidder’s willingness to pay for the auctioned item and is not payable unless it turns out to be the winning bid, while money spent on bids in a penny auction is unrecoverable. Structures of most penny auction websites, including relatively small bidding fees and discounted packages of bidding credits, promote irrational bidding behaviours and exploit the sunk cost fallacy2; as bidders continue to participate in an auction, and they spend more money on bids leading them to experience a higher psychological cost from leaving the auction (Eyster, 2002; Augenblick, 2011). On the other hand, for a newcomer of some penny auction websites with a bidding fee of $1, it is likely that he may not realise he has already spent $100 after placing 100 bids. Most penny auction websites do not provide full bidding histories to their visitors. For instance, when viewing a live, or completed auction on Swoopo, only the ten most recent bids are visible, and most other auctioneers provide between five to ten recent bids in live auctions; although BidCactus displayed a full list of all bidders and their total number of bids in completed auctions, full bidding histories were not provided. Furthermore, the bidding history of a bidder is usually not available for him/her to track either, so it is not obvious to a bidder how much he/she has spent so far in a live auction. From sellers’ points of view, unlike traditional auctioneers who earn from the winning bids, bidding fees are the primary source of revenue for penny auction websites. Their mechanisms and auction settings (such as tiny bidding increments) promote the total number of bids placed in each auction. In a traditional online auction, most bids are placed at the beginning and within the last minutes before it ends, and because the bidding increment is customisable, simply placing a bid may not make a bidder the current leader, which may discourage some bidders from participating. However, in a penny auction, an operator would prefer to keep participants bidding steadily throughout the auction, so settings such as zero starting price, fixed tiny price increment per bid, and high valuation of auction items, which maintain the attractiveness of placing a bid throughout the auction. Similarly, a penny auction operator would also want to maximise the auction’s length. In traditional online auctions, bidders tend to wait until the last moment to place their bids (Roth and Ockenfels, 2002). The dynamic countdown clock in a large bold font is a key feature adopted by all penny auction websites. Together with super short extension periods (usually 10~30 seconds) being counted down in seconds, it builds a great atmosphere to attract new bidders to join in as well as encouraging current participants to place bids continuously, resulting in auto-extension periods that can last for hours, and even days. For instance, 120 bids that only increase the current price by $1.20 could take up to one hour in an auction with a 30-second countdown timer. However, it is not reasonable to expect ordinary bidders to remain active by placing bids manually in an online auction for hours or days, thus automatic bidding tools such as BidButler of Swoopo, are available to all bidders. For example, a bidder can set his BidButler to place a bid on a certain auction every time when the countdown clock reaches the last second, such that he can stay in the auction as long as he has bidding credits remaining (MacDonald, 2011). Besides this, large operators such as Swoopo and BidRivals run their auctions internationally, so that bidders living in different time zones across the globe can compete in every single auction, which contribute to extending the auction’s length. In my study, I collect my original dataset of over half a million auctions over a timeframe of four years from BidCactus.com, one of the market leaders after the shutdown of Swoopo. The data show wide variability in profitability across categories of items and overall high profitability. I first replicate the Maximum Likelihood methodology of Platt et al. (2013) to estimate risk parameters implied by bidding behaviour for my original dataset, and found evidence of bidders’ experiences affecting the profitability, which was consistent with evidence from earlier studies using other datasets that reveal differential bidding behaviour based on experience. I therefore extend the Platt et al. (2013) model to allow for multiple types of bidders with different prior experiences and estimate type-specific risk parameters via Maximum Likelihood when there are up to three types of bidders in an auction. Bidders in different experience groups in our dataset are shown to have significantly different risk attitudes. When bidders are separated into two bidding groups of inexperienced bidders with participation experience of less than 20 prior auctions with the rest being experienced bidders, I observe that the more experienced bidders were more risk-seeking and bidding more aggressively, which contributes more to the seller’s revenue, regardless of whether they competed with rivals of the same type, or otherwise. Another interesting fact that was observed, was that 0.2% of bidders participated in 1,000 or more auctions, and placed over 20% of all bids. If we separate this small proportion of bidders into a new group called the super-experienced bidders, and called the rest of the experienced bidders as ordinary-experienced bidders, we observed that the super-experienced bidders whose behaviours are usually affected by type composition of rivals. They play more aggressively than the other two types when playing only against rivals of their own type; while they play more conservatively compared to the ordinary-experienced bidders (those with 20~1000 prior auction experiences) when there exist rivals of other types in an auction. Implication of bidders’ experience types on seller’s profitability are discussed in more details in Chapter 6 and 7.