Introduction Theory Experimental Design Results E¢ ciency in Multi-Object Auctions with (Failed) Resale Marco Pagnozzi Krista Jabs Saral June 2014 Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Auctions are often followed by a resale market, where winners can resell the objects acquired Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Auctions are often followed by a resale market, where winners can resell the objects acquired – Spectrum licenses Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Auctions are often followed by a resale market, where winners can resell the objects acquired – Spectrum licenses – Treasury bills Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Auctions are often followed by a resale market, where winners can resell the objects acquired – Spectrum licenses – Treasury bills – Emission permits ... Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Auctions are often followed by a resale market, where winners can resell the objects acquired – Spectrum licenses – Treasury bills – Emission permits ... How does resale a¤ect bidders’strategies, e¢ ciency and the seller’s revenue? Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Auctions are often followed by a resale market, where winners can resell the objects acquired – Spectrum licenses – Treasury bills – Emission permits ... How does resale a¤ect bidders’strategies, e¢ ciency and the seller’s revenue? Should resale be allowed? Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Auctions are often followed by a resale market, where winners can resell the objects acquired – Spectrum licenses – Treasury bills – Emission permits ... How does resale a¤ect bidders’strategies, e¢ ciency and the seller’s revenue? Should resale be allowed? What if resale is uncertain? Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Why does resale happen? Bidders do not participate in the auction (Milgrom, 1987; Bikhchandani & Huang, 1989) Bidders’valuations change after the auction (Haile, 2000, 2003) Value uncertainty (in 1st - price auctions) (Gupta & Lebrun, 1999; Hafalir & Krishna, 2007) Auction price a¤ects bargaining in resale market (Pagnozzi, 2007) Strategic behavior: demand reduction and speculation (Garratt & Tröger 2006; Pagnozzi, 2009, 2010) Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results In multi-object auctions, bidders often bid less than value for marginal units to keep the auction price low (Demand Reduction – Wilson, 1979; Ausubel & Cramton, 98) (e.g., FCC auctions, German GSM auction, electricity) Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results In multi-object auctions, bidders often bid less than value for marginal units to keep the auction price low (Demand Reduction – Wilson, 1979; Ausubel & Cramton, 98) (e.g., FCC auctions, German GSM auction, electricity) Demand reduction reduces the seller’s revenue and yields an ine¢ cient allocation, making bidders willing to trade Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results In multi-object auctions, bidders often bid less than value for marginal units to keep the auction price low (Demand Reduction – Wilson, 1979; Ausubel & Cramton, 98) (e.g., FCC auctions, German GSM auction, electricity) Demand reduction reduces the seller’s revenue and yields an ine¢ cient allocation, making bidders willing to trade Resale induces weak (low-value) bidders to speculate: bid aggressively to win and sell to strong (high-value) bidders Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results In multi-object auctions, bidders often bid less than value for marginal units to keep the auction price low (Demand Reduction – Wilson, 1979; Ausubel & Cramton, 98) (e.g., FCC auctions, German GSM auction, electricity) Demand reduction reduces the seller’s revenue and yields an ine¢ cient allocation, making bidders willing to trade Resale induces weak (low-value) bidders to speculate: bid aggressively to win and sell to strong (high-value) bidders Resale increases strong bidders’incentive to reduce demand, because they can purchase after the auction the units lost Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results UK 3.4GHz Auction (June 2003) Simultaneous Ascending Auction for 15 licenses for broadband wireless services Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results UK 3.4GHz Auction (June 2003) Simultaneous Ascending Auction for 15 licenses for broadband wireless services PCCW (a Hong-Kong telecom company) was the highest-value bidder and was expected to win 15 licenses Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results UK 3.4GHz Auction (June 2003) Simultaneous Ascending Auction for 15 licenses for broadband wireless services PCCW (a Hong-Kong telecom company) was the highest-value bidder and was expected to win 15 licenses Red Spectrum and Public Hub were companies created for the auction and chose to be eligible for only 1 license Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results UK 3.4GHz Auction (June 2003) Simultaneous Ascending Auction for 15 licenses for broadband wireless services PCCW (a Hong-Kong telecom company) was the highest-value bidder and was expected to win 15 licenses Red Spectrum and Public Hub were companies created for the auction and chose to be eligible for only 1 license As soon as PCCW, RS and PH were the only bidders left, PCCW reduced demand to 13 licenses to end the auction Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results UK 3.4GHz Auction (June 2003) Simultaneous Ascending Auction for 15 licenses for broadband wireless services PCCW (a Hong-Kong telecom company) was the highest-value bidder and was expected to win 15 licenses Red Spectrum and Public Hub were companies created for the auction and chose to be eligible for only 1 license As soon as PCCW, RS and PH were the only bidders left, PCCW reduced demand to 13 licenses to end the auction PCCW’s failure to win all licenses was described as “a surprise, ... a ga¤e,” “a costly mistake that may cost the chance of o¤ering a nationwide service” Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results UK 3.4GHz Auction (June 2003) Simultaneous Ascending Auction for 15 licenses for broadband wireless services PCCW (a Hong-Kong telecom company) was the highest-value bidder and was expected to win 15 licenses Red Spectrum and Public Hub were companies created for the auction and chose to be eligible for only 1 license As soon as PCCW, RS and PH were the only bidders left, PCCW reduced demand to 13 licenses to end the auction PCCW’s failure to win all licenses was described as “a surprise, ... a ga¤e,” “a costly mistake that may cost the chance of o¤ering a nationwide service” ... but PCCW took over RS and PH, obtaining all licenses Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Experimental literature on auctions with resale Lange et al. (2004), Georganas (2007), Saral (2008), Georganas & Kagel (2011), Jog & Kosmopoulou (2014) consider single-unit auctions with resale either "automatic" or through auction Pagnozzi & Saral (2013), Filiz-Ozbay et al. (2013) consider multi-object auctions with di¤erent resale forms and valuation structures First study of auctions when resale may exogenously fail ... Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Overview In multi-object uniform-price auctions with asymmetric bidders and resale through bargaining: 1 Demand reduction and speculation emerge when the auction winner can resell Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Overview In multi-object uniform-price auctions with asymmetric bidders and resale through bargaining: 1 Demand reduction and speculation emerge when the auction winner can resell 2 E¤ects of resale on e¢ ciency and seller’s revenue Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Overview In multi-object uniform-price auctions with asymmetric bidders and resale through bargaining: 1 Demand reduction and speculation emerge when the auction winner can resell 2 E¤ects of resale on e¢ ciency and seller’s revenue 3 E¤ects of changing the probability of resale Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Overview In multi-object uniform-price auctions with asymmetric bidders and resale through bargaining: 1 Demand reduction and speculation emerge when the auction winner can resell 2 E¤ects of resale on e¢ ciency and seller’s revenue 3 E¤ects of changing the probability of resale 4 Bargaining in resale market Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Model Bidding Strategies THEORETICAL BACKGROUND Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Model Bidding Strategies Model 2 units of an identical good for sale Uniform-price auction: the 2 highest bids win, and winner(s) pay the 3rd -highest bid for each unit Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Model Bidding Strategies Model 2 units of an identical good for sale Uniform-price auction: the 2 highest bids win, and winner(s) pay the 3rd -highest bid for each unit 2 asymmetric (risk-neutral) bidders: – S (strong) demands 2 units and has high value vs – W (weak) demands 1 unit and has low value vw Marco Pagnozzi, Krista Jabs Saral U [30; 50] U [10; 30] E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Model Bidding Strategies Model 2 units of an identical good for sale Uniform-price auction: the 2 highest bids win, and winner(s) pay the 3rd -highest bid for each unit 2 asymmetric (risk-neutral) bidders: – S (strong) demands 2 units and has high value vs – W (weak) demands 1 unit and has low value vw U [30; 50] U [10; 30] ! Either S wins both units or S and W win one unit each Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Model Bidding Strategies Resale Market After the auction, if W wins a unit, he may resell it to S Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Model Bidding Strategies Resale Market After the auction, if W wins a unit, he may resell it to S Resale takes place through bargaining Gains from trade are vS vW S obtains a share α of the gains from trade W obtains a share (1 α) of the gains from trade (results are robust to many alternative models of resale market) Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Model Bidding Strategies Resale Market After the auction, if W wins a unit, he may resell it to S Resale takes place through bargaining Gains from trade are vS vW S obtains a share α of the gains from trade W obtains a share (1 α) of the gains from trade (results are robust to many alternative models of resale market) q = exogenous probability of a resale market (trading friction or measure of market e¢ ciency) Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Model Bidding Strategies W’s Bidding Strategy W bids up to bW vW + q (1 | Marco Pagnozzi, Krista Jabs Saral α) (E [vS j W wins] {z expected resale pro…t vW ) } E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Model Bidding Strategies W’s Bidding Strategy W bids up to bW vW + q (1 | α) (E [vS j W wins] {z expected resale pro…t vW ) } q = 0 (no resale): W bids vW (dominant strategy) Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Model Bidding Strategies W’s Bidding Strategy W bids up to bW vW + q (1 | α) (E [vS j W wins] {z expected resale pro…t vW ) } q = 0 (no resale): W bids vW (dominant strategy) q > 0: W speculates because of the option to resell and bids higher than vW Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Model Bidding Strategies S’s Bidding Strategy Since W bids bW , in the auction S can Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Model Bidding Strategies S’s Bidding Strategy Since W bids bW , in the auction S can outbid W and win 2 units, obtaining 2 (vS Marco Pagnozzi, Krista Jabs Saral E [bW ]) E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Model Bidding Strategies S’s Bidding Strategy Since W bids bW , in the auction S can outbid W and win 2 units, obtaining 2 (vS E [bW ]) 2nd unit (reduce demand), win 1 unit bid 0 for the and then try to buy from W , obtaining vs 0 | {z } auction pro…t Marco Pagnozzi, Krista Jabs Saral + qα (vS E [vW ]) {z } | expected resale pro…t E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Model Bidding Strategies S’s Bidding Strategy Since W bids bW , in the auction S can outbid W and win 2 units, obtaining 2 (vS E [bW ]) 2nd unit (reduce demand), win 1 unit bid 0 for the and then try to buy from W , obtaining vs 0 | {z } + qα (vS E [vW ]) {z } | vS (1 + qα) qαE [vW ] = auction pro…t expected resale pro…t Indi¤erent type vS is such that: 2 (vS E [vW ] q (1 Marco Pagnozzi, Krista Jabs Saral α) (E [vS j vS < vS ] E [vW ])) E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Model Bidding Strategies S’s Bidding Strategy Since W bids bW , in the auction S can outbid W and win 2 units, obtaining 2 (vS E [bW ]) 2nd unit (reduce demand), win 1 unit bid 0 for the and then try to buy from W , obtaining vs 0 | {z } + qα (vS E [vW ]) {z } | vS (1 + qα) qαE [vW ] = auction pro…t expected resale pro…t Indi¤erent type vS is such that: 2 (vS E [vW ] , q (1 vS Marco Pagnozzi, Krista Jabs Saral α) (E [vS j vS < vS ] 40 E [vW ])) 10q (1 + α) 1 q E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Model Bidding Strategies S’s Bidding Strategy ) S reduces demand i¤ vS < vS = Marco Pagnozzi, Krista Jabs Saral 40 10q (1 + α) 1 q E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Model Bidding Strategies S’s Bidding Strategy ) S reduces demand i¤ vS < vS = 40 10q (1 + α) 1 q When q = 0, S reduces demand i¤ vS < 40 S ’s incentive to reduce demand giving up 1 unit is lower when he has a higher value Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Model Bidding Strategies S’s Bidding Strategy ) S reduces demand i¤ vS < vS = 40 10q (1 + α) 1 q When q = 0, S reduces demand i¤ vS < 40 S ’s incentive to reduce demand giving up 1 unit is lower when he has a higher value q" ) less costly to lose auction ) Marco Pagnozzi, Krista Jabs Saral vS " E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Model Bidding Strategies S’s Bidding Strategy ) S reduces demand i¤ vS < vS = 40 10q (1 + α) 1 q When q = 0, S reduces demand i¤ vS < 40 S ’s incentive to reduce demand giving up 1 unit is lower when he has a higher value q" ) less costly to lose auction ) α" ) bW # vS " ) cheaper to outbid W Marco Pagnozzi, Krista Jabs Saral ) vS # E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Model Bidding Strategies S’s Bidding Strategy ) S reduces demand i¤ vS < vS = 40 10q (1 + α) 1 q When q = 0, S reduces demand i¤ vS < 40 S ’s incentive to reduce demand giving up 1 unit is lower when he has a higher value q" ) less costly to lose auction ) α" ) bW # vS " ) cheaper to outbid W ) vS # When α = 12 , S always reduce demand if q > 0.3 S wins 1 unit at price 0 and then buys from W with high probability (rather than pay both units) Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Model Bidding Strategies Summing up: 1 Without resale (q = 0), W bids vW and S reduces demand if and only if vS < 40 Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Model Bidding Strategies Summing up: 1 Without resale (q = 0), W bids vW and S reduces demand if and only if vS < 40 2 With resale (q 0), W bids above vW and S always reduces demand Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Model Bidding Strategies Summing up: 1 Without resale (q = 0), W bids vW and S reduces demand if and only if vS < 40 2 With resale (q 0), W bids above vW and S always reduces demand Due to uncertainty of resale, risk-averse bidders speculate and reduce demand less (with resale) Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results EXPERIMENTAL DESIGN Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results vW U [10; 30]; vS U [30; 50] (same values for all treatments) Subjects are randomly assigned to S or W and do not change role Each subject participates in only 1 treatment Bidders are randomly rematched each period Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Uniform-Price Ascending Clock Auction Bidders choose when to drop out of the auction as the price increases When one bidder drops out, the auction is over (# of units on sale = # of units demanded) Winner(s) pay the dropout price for each unit Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Uniform-Price Ascending Clock Auction Bidders choose when to drop out of the auction as the price increases When one bidder drops out, the auction is over (# of units on sale = # of units demanded) Winner(s) pay the dropout price for each unit Resale Market Unstructured bargaining: multiple o¤ers, chat and exit option Exogenous probability determines if bidders enter resale when W wins a unit Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Treatments - between subjects design 1. No Resale: after the auction, there is no resale market 2. 30% Resale: if W wins the auction, bidders participate in resale with 30% probability 3. 50% Resale: if W wins the auction, bidders participate in resale with 50% probability 4. 100% Resale: if W wins the auction, bidders participate in resale with certainty Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Session Information 3 sessions of 16 subjects per treatment (48 subjects per treatment) All sessions had at least 20 periods, maximum 30 (# periods di¤er because of bargaining & 2 hour limit) All 12 sessions were run in the xs/fs laboratory at FSU Spring 2011-Fall 2012; Fall 2013-Spring 2014 Mostly undergraduate subjects Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Weak Bidding Strong Bidding E¢ ciency and Revenue Resale Market RESULTS Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Weak Bidding Strong Bidding E¢ ciency and Revenue Resale Market Weak Bidding Hypotheses No Resale: W bids value Resale (q > 0): W bids above value Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Without resale, W tends to bid value 0 10 20 30 40 No Resale 10 15 20 25 30 Does W speculate more with resale?* 40 30 20 10 0 0 10 20 30 40 50 30% Resale 50 No Resale bid 10 15 20 25 30 10 20 25 30 25 30 40 30 20 10 0 0 10 20 30 40 50 100% Resale 50 50% Resale 15 10 15 20 25 30 10 v alue * Fewer observations with resale because of W winning more and fewer periods 15 20 Bidding by W – Random E¤ects Tobit (unobserved bids censored at the auction price) W’s Bid Constant vw 30% Resale 50% Resale 100% Resale vw 30% Resale vw 50% Resale vw 100% Resale Period Coe¢ cient 1.20 0.99*** 3.89** 4.77** 6.20** –0.07 –0.16** –0.23** –0.03 *** and ** indicate statistical signi…cance at 1% and 5%, bootstrapped standard errors Bidding by W – Random E¤ects Tobit (unobserved bids censored at the auction price) W’s Bid Constant vw 30% Resale 50% Resale 100% Resale vw 30% Resale vw 50% Resale vw 100% Resale Period Coe¢ cient 1.20 0.99*** 3.89** 4.77** 6.20** - Bids are higher with resale - E¤ect weaker with more uncertainty –0.07 –0.16** –0.23** –0.03 *** and ** indicate statistical signi…cance at 1% and 5%, bootstrapped standard errors Introduction Theory Experimental Design Results Weak Bidding Strong Bidding E¢ ciency and Revenue Resale Market Strong Bidding Hypotheses No resale: S reduces demand (bidding zero) if vS < 40, S does not reduce demand if vS > 40 Resale (q 0): S always reduces demand Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Weak Bidding Strong Bidding E¢ ciency and Revenue Resale Market Demand Reduction by S: % S’s bids 2 v s < 40 v s > 40 No Resale 30% Resale 50% Resale 100% Resale 29.9 43.2 44.8 48.1 12.4 34.8 37.5 49.5 % W Wins v s < 40 v s > 40 51.9 74.2 73.7 74.0 25.1 55.7 61.8 70.5 % of all auctions where vs < 40 or vs > 40 for both bids and wins Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Without resale, S reduces demand more when v s < 40 0 10 20 30 40 50 No Resale 30 35 40 45 50 With resale, S reduces demand more, for all values but uncertainty reduces demand reduction 40 30 20 10 0 0 10 20 30 40 50 30% Resale 50 No Resale bid 30 35 40 45 50 30 40 45 50 45 50 40 30 20 10 0 0 10 20 30 40 50 100% Resale 50 50% Resale 35 30 35 40 45 50 v alue 30 35 40 Bidding by S – Random E¤ects Tobit (unobserved bids censored at the auction price) Strong Bid Constant vs vs >40 30% Resale 50% Resale 100% Resale vs >40 30% Resale vs >40 50% Resale vs >40 100% Resale Period Coe¢ cient 5.26 0.58*** 4.99* –6.71** –7.31** –10.34*** –4.59 –5.75* –9.71*** –0.41*** *** and ** indicate statistical signi…cance at 1% and 5% Bidding by S – Random E¤ects Tobit (unobserved bids censored at the auction price) Strong Bid Constant vs vs >40 30% Resale 50% Resale 100% Resale vs >40 30% Resale vs >40 50% Resale vs >40 100% Resale Period Coe¢ cient 5.26 0.58*** 4.99* –6.71** –7.31** –10.34*** –4.59 –5.75* –9.71*** –0.41*** *** and ** indicate statistical signi…cance at 1% and 5% - Without resale, S bids higher when he has a high value Bidding by S – Random E¤ects Tobit (unobserved bids censored at the auction price) Strong Bid Constant vs vs >40 30% Resale 50% Resale 100% Resale vs >40 30% Resale vs >40 50% Resale vs >40 100% Resale Period Coe¢ cient 5.26 0.58*** 4.99* –6.71** –7.31** –10.34*** –4.59 –5.75* –9.71*** –0.41*** *** and ** indicate statistical signi…cance at 1% and 5% - Bids are lower with resale especially with less uncertainty Bidding by S – Random E¤ects Tobit (unobserved bids censored at the auction price) Strong Bid Constant vs vs >40 30% Resale 50% Resale 100% Resale vs >40 30% Resale vs >40 50% Resale vs >40 100% Resale Period Coe¢ cient 5.26 0.58*** 4.99* –6.71** –7.31** –10.34*** –4.59 –5.75* –9.71*** –0.41*** *** and ** indicate statistical signi…cance at 1% and 5% - Lower bids at high values when resale is less uncertain Bidding by S – Random E¤ects Tobit (unobserved bids censored at the auction price) Strong Bid Constant vs vs >40 30% Resale 50% Resale 100% Resale vs >40 30% Resale vs >40 50% Resale vs >40 100% Resale Period Coe¢ cient 5.26 0.58*** 4.99* –6.71** –7.31** –10.34*** –4.59 –5.75* –9.71*** –0.41*** *** and ** indicate statistical signi…cance at 1% and 5% - Signi…cant learning by S Introduction Theory Experimental Design Results Weak Bidding Strong Bidding E¢ ciency and Revenue Resale Market E¢ ciency and Revenue Hypotheses Auction e¢ ciency is lower with resale than without (because of demand reduction) Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Weak Bidding Strong Bidding E¢ ciency and Revenue Resale Market E¢ ciency and Revenue Hypotheses Auction e¢ ciency is lower with resale than without (because of demand reduction) Final e¢ ciency depends on the amount of demand reduction and the probability of resale Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Weak Bidding Strong Bidding E¢ ciency and Revenue Resale Market E¢ ciency and Revenue Hypotheses Auction e¢ ciency is lower with resale than without (because of demand reduction) Final e¢ ciency depends on the amount of demand reduction and the probability of resale Seller’s revenue is lower with resale than without (because of demand reduction) Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Auction E¢ ciency: Frequency of S winning 2 units ! resale reduces e¢ ciency because of demand reduction and speculation 30 35 40 45 50 30 35 40 45 50 v alue .8 1 100% Resale 0 .2 .4 .6 .8 .6 .2 0 0 .2 .4 .6 .8 1 50% Resale 1 30% Resale .4 .6 .4 .2 0 relative frequency .8 1 No Resale 30 35 40 45 50 30 35 40 45 50 Final E¢ ciency: Resale increases e¢ ciency after the auction ! ambiguous e¤ect on …nal e¢ ciency 30 35 40 45 50 30 35 40 45 50 v alue .8 1 100% Resale 0 .2 .4 .6 .8 .6 .2 0 0 .2 .4 .6 .8 1 50% Resale 1 30% Resale .4 .6 .4 .2 0 relative frequency .8 1 No Resale 30 35 40 45 50 30 35 40 45 50 Introduction Theory Experimental Design Results Weak Bidding Strong Bidding E¢ ciency and Revenue Resale Market Auction E¢ ciency: winner’s value/S’s value Resale Prob. 0% 30% 50% 100% Auction E¢ ciency .82 .69 .68 .65 .82 .75 .82 .95 .75 .75 .75 .75 (average) Final E¢ ciency (average) Random E¢ ciency Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Weak Bidding Strong Bidding E¢ ciency and Revenue Resale Market FInal E¢ ciency: …nal owner’s value/S’s value Resale Prob. 0% 30% 50% 100% Auction E¢ ciency .82 .69 .68 .65 .82 .75 .82 .95 .75 .75 .75 .75 (average) Final E¢ ciency (average) Random E¢ ciency 100% Resale 0% Resale Marco Pagnozzi, Krista Jabs Saral 50% Resale 30%Resale E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Weak Bidding Strong Bidding E¢ ciency and Revenue Resale Market Seller’s Revenue: – Resale reduces revenue because it induces S to reduce demand (when S wins revenue is higher because of speculation by W ) Resale Prob. 0% 30% 50% 100% Average Revenue Revenue (W wins) Revenue (S wins) 14.6 8.0 18.8 11.2 6.6 17.2 10.0 6.0 18.6 8.4 5.2 17.2 Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Weak Bidding Strong Bidding E¢ ciency and Revenue Resale Market E¤ects of changing the probability of resale on success of resale: 30% Resale 50% Resale 100% Resale Resale Possible (W won) Successful Resale Exogenous Failure Endogenous Failure 65% 68% 73% 19% 42% 79% 75% 47% 0% 6% 11% 21% Subjects agree to trade more often when resale is less likely Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Weak Bidding Strong Bidding E¢ ciency and Revenue Resale Market Average O¤er, Resale Price, Resale Earnings First O¤er Last O¤er Weak / Strong Weak / Strong 30% Resale 34 / 21 28 / 25 50% Resale 34 / 20 28 / 25 100% Resale 34 / 19 29 / 25 Resale Price Earnings (Auction Price) Weak / Strong 26 7 / 12 27 7 / 12 27 8 / 12 (11) (10) (8) No signi…cant di¤erences across treatments S earns more than W in resale Both S and W improve o¤ers through bargaining Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Weak Bidding Strong Bidding E¢ ciency and Revenue Resale Market Total Earnings: Auction + Resale Pro…ts Average Total Earnings No Resale 30% Resale 50% Resale 100% Resale 15.83 Weak Bidder 4.95 9.86 12.02 (std. dev) (8.838) (11.171) (12.134) (12.620) Strong Bidder 38.64 37.73 40.45 44.62 (std. dev) (16.911) (16.122) (16.260) (17.147) W always obtains higher pro…ts with resale S obtains higher pro…ts with resale when it’s more certain Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale Introduction Theory Experimental Design Results Weak Bidding Strong Bidding E¢ ciency and Revenue Resale Market Conclusions Experiments of multi-object auction with resale and asymmetric bidders uncertain resale through bargaining Without resale, strong bidders with low values reduce demand less With resale (even when uncertain), weak bidders speculate and strong bidders reduce demand more frequently Resale reduces auction e¢ ciency and seller’s revenue Very uncertain resale decreases …nal e¢ ciency (compared to an auction without resale) Strong bidders earn more than bidders in resale Marco Pagnozzi, Krista Jabs Saral E¢ ciency in Multi-Object Auctions with (Failed) Resale