Bid More, Pay Less Overbidding and the Bidder s Curse in Teleshopping Auctions LH1: Auctions Stony Brook Center for Game Theory, NY, July 17 th, 2017 E-Mail: Fabian.Ocker@kit.edu KIT Karlsruhe Institute of Technology www.kit.edu
Bid more, pay less A German thing? 3
Outline 1 Introduction: 1-2-3.tv Auctions 2 Related Literature and Data Set 3 Theoretic Analysis 4 Hypotheses and Results 5 Summary and Outlook 4
Introduction: General Information Interactive auction-teleshopping channel. Start of broadcast: October 1 st, 2004. Current broadcasting time: 365 days per year à 20 hours per day. Broadcast is split into eight product categories, e.g. Jewlery, Beauty & Wellness, Household & Kitchen. Feature: Two sales channels. Customers bid in the auction or purchase for a fixed online shop price. Consequently, an (objective) market price is available. Distribution channels for auction broadcast: Offline: TV & telephone, Online: Website or App. 5
Introduction: Applied Auction Mechanism Auction form: Dutch (sell) Auction Bidding rule: Auction price declines continuously until a bidder signalizes his willingness to pay at the current price. Implementation: Current auction price is presented on an auction watch. The auction terminates once a bidder signalizes his willingness to pay. Scoring rule: The signalizing bidder wins the auction. Price rule: The final price equals the price shown on the auction watch when the winning bidder signalized his willingness to pay. Feature: Multi-unit auction Uniform price auction (lowest accepted bid, LAB). Homogeneous goods (complements possible). 6
Related Literature (on Single-unit Auctions) Amyx and Luehlfing (2006) with a data set of 416 online auctions. First evidence for overbidding when simultaneously a fixed price is available. 9% of overbidding, 14% mean percentage of overbidding. Malmendier and Lee (2011) with two data sets of 2,200 online auctions. Name the overbidding phenomenon as Bidder s Curse. 42%/48% of Bidder s Curse, 2%/10% mean percentage of overbidding. Best explanation approach for Bidder s Curse is limited attention. Schneider (2016) with a data set of 552 online auctions. Limited attention is a premature explanation approach. Search costs for price information need to be considered. 23% of Bidder s Curse. Freeman, Kimbrough and Reiss (2017) conduct a laboratory experiment. Overbidding increases when search costs for price information are high. 7
Data Set Auction outcomes from January 19 th to March 23 rd, 2016 (65 days). Crawler (in C#) wrote data from www.1-2-3.tv in MS Excel. Each submitted bid is reported (around 700,00 bids). à Note: reported bid = sold good Date, product, distribution channel, uniform price, online-shop price, etc. Several contributions to existing literature: Substantially greater data set. Systematic analysis across different product categories. Extension of the analysis to multi-unit auctions. Consideration of teleshopping auctions. 8
Theoretic Analysis IPV-model of the 1-2-3.tv auction: Valuation of bidder depends solely on own signal. Individual signal is known by each bidder before the auction. Valuations of other bidders are unknown (distribution common knowledge). Assumption: No transaction costs. Analysis of bidding strategy (following Malmendier and Lee, 2011): Extension of the multi-unit Dutch auction to a two-stage game: First stage: Multi-unit auction with uniform pricing (HRB/LAB). Second stage: Purchase of goods for a fixed price. Furthermore: Distinction of single-unit and multi-unit demand. Main result: Rational bidders do not overbid the online shop fixed price. 9
Hypotheses and Results (1/2) Do bidders in the 1.2.3.tv auctions behave according to theory? Hypothesis 1 (Overbidding in 1-2-3.tv auctions): Bidders do not submit bids higher than the simultaneously available online shop price for the same good. Hypothesis 2 (Bidder s Curse in 1-2-3.tv auctions): None of the final uniform auction prices exceed the simultaneously available online shop prices. Findings Finding 1 (Overbidding in 1-2-3.tv auctions): In 25.55% of all auctions, bids are higher than the simultaneously available online shop price for the same good. Finding 2 (Bidder s Curse in 1-2-3.tv auctions): In 5.18% of all auctions, the final uniform auctions price is higher than the simultaneously available online shop price. 11
Hypotheses and Results (2/2) What are influencing factors for overbidding and the Bidder s Curse? Finding 3 (Search cost in 1-2-3.tv auctions - overbidding): Offline-bidders overbid greater and more often than online-bidders: Relative frequency of overbidding of 26.77% (19.88%) for offline (online) bidders, Average percentage of overbidding of 9.67% (8.98%) for offline (online) bidders. Finding 4 (Search cost in 1-2-3.tv auctions Bidder s Curse): Offline-bidders do not experience the Bidder s Curse more often than online-bidders: Relative frequency of Bidder s Curse of 5.31% (4.66%) for offline (online) bidders, Average percentage of Bidder s Curse of 5.84% (5.61%) for offline (online) bidders. Finding 5 (Learning effect in 1-2-3.tv auctions): The Top 10 most frequent customers do not experience a learning effect, but overbid greater and more often than the average 1-2-3.tv customer. The most frequent customer submitted 488 bids with total expenses of 37,625. 12
Summary and Outlook We find that overbidding and the Bidder s Curse are also present in multi-unit teleshopping auctions. However, the frequency of the Bidder s Curse is (far) lower than in studies on single-unit auctions. We argue that this is due to multi-unit auctions with uniform pricing. Here, overbidding does not mandatorily result in the Bidder s Curse. In other words, overbidding is less risky. We find that offline-bidders overbid greater and more often than online-bidders, and reason in line with recent scientific work that this is linked to different search costs of these two types of bidders. Further research could focus on empirical investigation of other formats of teleshopping auctions. other shops that offer two sales channels. 13
Thank you for your attention! M.Sc. Karlsruhe Institute for Technology (KIT) Fabian.Ocker@kit.edu http://games.econ.kit.edu 14 16.07.17