Optimal Two-stage Auctions with Costly Information Acquisition

Size: px
Start display at page:

Download "Optimal Two-stage Auctions with Costly Information Acquisition"

Transcription

1 Optmal Two-stage Auctons wth Costly Informaton Acquston Jngfeng Lu Lxn Ye December 2014 Abstract We consder an aucton envronment wth costly entry wheren the cost manly stems from nformaton acquston. Bdders are endowed wth orgnal estmates ( types ) about ther prvate values and can further learn ther true values of the object for sale by ncurrng an entry cost. We frst derve an ntegral form of the envelope formula as requred by ncentve compatble two-stage mechansms, based on whch we demonstrate that optmalty of the generalzed Myerson allocaton rule s robust to our settng wth costly nformaton acquston. Optmal entry s thus to admt the set of bdders that maxmzes expected vrtual surplus adjusted by both the second-stage sgnal and entry cost. We show that our optmal entry and allocaton rules are both IR and IC mplementable, and furthermore, n an mportant class of envronments, they can be mplemented va a two-stage aucton that s essentally a handcap aucton augmented wth an entry mechansm. Keywords: Two-stage auctons, entry, nformaton acquston, sequental screenng, handcap auctons, optmal mechansms. JEL Classfcaton: D44, D80, D82. 1 INTRODUCTION In hgh-valued asset sales, buyers often need to go through a due dlgence process before developng fnal bds. Due dlgence s usually a process to update or acqure nformaton about the value of the asset for sale or to prepare for the bddng process (e.g., to establsh qualfcatons to bd). Ths process s costly and s usually modeled as entry as t s closely montored by the auctoneer. For a sale of an asset worth bllons of dollars, the entry cost can run from tens of thousands to mllons of dollars. 1 We thank semnar partcpants at Unversty of Mchgan, the Conference n Honor of Paul Mlgrom s 65th Brthday, the Mdwest Economc Theory Conference, NSF/CEME Decentralzaton Conference, and n partcular, Drk Bergemann, Tlman Börgers, Yeon-Koo Che, Jeff Ely, L Hao, Preston McAfee, Davd Mller, Ilya Segal, Xanwen Sh and Juuso Tokka for very helpful comments and suggestons. All remanng errors are our own. Department of Economcs, Natonal Unversty of Sngapore, 10 Kent Rdge Crescent, Sngapore Tel: (65) , Emal: ecsljf@nus.edu.sg. Department of Economcs, The Oho State Unversty, 449A Arps Hall, 1945 North Hgh Street, Columbus, OH Tel.: (614) Emal: ye.45@osu.edu. 1 A more detaled descrpton of a typcal due dlgence process s provded n Secton 4. 1

2 Gven the substantal entry cost, t s unrealstc to assume that whoever s nterested would necessarly go through the costly entry process. The success of a sale thus very much reles on whether the most qualfed bdders would commt to the due dlgence process and partcpate n the fnal sale. Manly motvated by the need for entry screenng, varants of two-stage sellng mechansms have emerged n the real world. A leadng example of the two-stage aucton procedure s known as ndcatve bddng, whch s commonly used n sales of complcated busness assets wth very hgh values. It works as follows: the auctoneer actvely markets the assets to a large group of potentally nterested buyers. The potental buyers are then asked to submt non-bndng bds, based on whch a fnal set of bdders s shortlsted to advance to the second stage. The auctoneer then communcates only wth these fnal bdders, provdng them wth extensve access to nformaton about the assets, 2 and fnally runs the aucton (typcally usng bndng sealed bds). The use of ths two-stage aucton procedure s qute wdespread. For example, n response to the restructurng of the electrc power ndustry n the U.S. whch was desgned to separate power generaton from transmsson and dstrbuton bllons of dollars of electrcal generatng assets were dvested through ths two-stage aucton procedure over the last two decades. 3 Ths two-stage aucton procedure s also commonly used n prvatzaton, takeover, and merger and acquston contests. 4 Fnally, t s commonly used n the nsttutonal real estate market, whch has an annual sales volume n the order of $60 to $100 bllon. 5 Ye (2007) was the frst study of ndcatve bddng based on the assumpton of costly nformaton acquston. 6 Ye s analyss suggests that the current desgn of ndcatve bddng cannot relably select the most qualfed bdders for the fnal sale, as there does not exst a symmetrc, strctly ncreasng equlbrum bd functon n the ndcatve bddng stage. In a more recent paper, by restrctng ndcatve bds to a fnte dscrete doman, Qunt and Hendrcks (2013) show that a symmetrc equlbrum exsts n weakly-monotone strateges. But agan, the hghest-value bdders are not always selected, as bdder types pool over a fnte number of bds. Wthout safely selectng the most qualfed bdders for the fnal sale, the mechansm s hardly optmal n maxmzng expected revenue. What the optmal mechansm s n ths two-stage aucton envronment remans an open queston n the lterature, and ths paper seeks to provde an answer. We model the two-stage aucton envronment as follows. Before entry, each potental bdder s endowed wth a prvate sgnal, α, whch can be regarded as her pre-entry type. After entry (by ncurrng a common entry cost, c), each bdder fully observes her (prvate) value v, whch s postvely correlated to her pre-entry type. Gven costly entry, t s not feasble for all potental bdders to be ncluded n the fnal sale. As such, we consder the class of two-stage mechansms wth the frst stage allocatng entry 2 Data rooms, whch are descrbed n Secton 4, are typcally set up to facltate bdders due dlgence process. 3 A lst of ndustry examples usng ths two-stage aucton desgn can be found n Ye (2007). 4 Leadng examples nclude the prvatzaton of the Italan Ol and Energy Corporaton (ENI), the acquston of Ireland s largest cable televson provder Cablelnk Lmted, and the takeover contest for South Korea s second largest conglomerate Daewoo Motors. 5 See Foley (2003) for a detaled account. 6 Boone and Goeree (2009) provde an analyss of pre-qualfyng auctons, whch are smlar to ndcatve bddng. 2

3 rghts and the second stage allocatng the asset. 7 Despte the potental complcaton due to both sequental screenng and costly nformaton acquston, we are able to completely characterze the optmal revenue-maxmzng two-stage mechansms. Our analyss benefts greatly from recent developments n the lterature of sequental screenng (e.g., Courty and L, 2000; Esö and Szentes, 2007; Bergemann and Wambach, 2014; and Pavan, Segal, and Tokka, 2014). 8 In partcular, our analyss follows Esö and Szentes closely, and our techncal contrbuton s to extend ther analyss to dynamc auctons wth costly nformaton acquston. We frst derve an ntegral form of the envelope formula as a necessary condton for ncentve compatblty for our twostage mechansms, whch extends the valdty of the envelope theorem to dynamc auctons wth costly nformaton acquston. Based on ths derved envelope formula, we are able to show that the optmal allocaton rule of the asset n our second stage s the same as that dentfed by Esö and Szentes, whch requres that, among the shortlsted bdders, the asset be allocated to the bdder wth the hghest vrtual value adjusted by the second-stage sgnal. Our analyss thus suggests that the optmalty of the generalzed Myerson optmal allocaton rule (adjusted by second-round sgnals) s robust to the dynamc aucton settng wth costly entry. The frst-stage entry rght allocaton mechansm s new to the orgnal Esö-Szentes framework, and we show that the optmal entry rule s to admt the set of bdders that gves rse to the maxmum expected vrtual surplus (adjusted by both the second-stage sgnal and entry cost). Alternatvely, gven the regularty assumpton and that buyers are ex ante symmetrc n our model, the optmal entry rule s to admt the bdders n descendng order of ther pre-entry types, the hghest type frst, the second hghest type second, etc., provded that ther margnal contrbuton to the expected vrtual surplus s postve. Therefore, the optmal number of shortlsted bdders typcally depends on the reported type profle from the potental bdders, whch s endogenously determned. We then show that specfc payment rules can be constructed n each stage to mplement both optmal entry and allocaton rules truthfully. For an mportant settng where one s value s lnear n her frst sgnal, Esö and Szentes show that ther optmal mechansm can be mplemented over two rounds va a so-called handcap aucton: n the frst round (before observng the second-stage sgnal), each buyer selects a prce premum by payng a fee accordng to a pre-announced schedule. In the second round (after observng the second-stage sgnals), buyers compete n a second-prce or Englsh aucton, where the wnner obtans the object at a prce equal to the second-hghest bd plus the prce premum selected from the frst round. In our settng wth entry, the mplementaton s presumably more complcated, as optmal entry needs to be mplemented pror to the fnal aucton. Indeed, now the mplementaton requres that an (optmal) entry rule be augmented to the handcap aucton. So n our case the optmal mechansm s mplemented va a two-stage aucton, wth the frst stage beng an aucton for entry rghts (as well as the prce prema) and the second stage 7 The focus on two-stage mechansms should be regarded as a constrant, whch s dscussed n Secton 4. 8 Early work on dynamc contractng wth a sngle agent are due to Baron and Besanko (1984) and Rordan and Sappngton (1987). 3

4 beng a second-prce or Englsh aucton for the asset. Other than the connecton wth sequental screenng and dynamc auctons mentoned above, our paper s related to the lterature on nformaton acquston n auctons (see, for example, Persco, 2000; Compte and Jehel, 2001; and Rezende, 2013). These papers focus on bdders ncentves to acqure nformaton n dfferent aucton formats. Our paper dffers from thers n that we follow the normatve approach to dentfy optmal mechansms wth nformaton acquston. To the extent that nformaton acquston s modeled as entry, our paper s closely related to the growng lterature on auctons wth costly entry. 9 Ths lterature can be summarzed nto three branches. In the frst branch, bdders are assumed to possess no prvate nformaton before entry and they learn ther prvate values or sgnals only after entry (see, for example, McAfee and McMllan, 1987; Engelbrecht-Wggans, 1993; Tan, 1992; Levn and Smth, 1994; and Ye, 2004). In the second branch, t s assumed that bdders are endowed wth prvate nformaton about ther values but have to ncur entry costs to partcpate n an aucton (see, for example, Samuelson, 1985; Stegeman, 1996; Campbell, 1998; Menezes and Montero, 2000; Tan and Ylankaya, 2006; Cao and Tan, 2009; and Lu, 2009). Fnally, n the thrd branch, bdders are endowed wth some prvate nformaton before entry, and are able to acqure addtonal prvate nformaton after entry (Ye, 2007; Qunt and Hendrcks, 2013). The framework n ths current paper nests all the models mentoned above as specal cases. Our paper thus characterzes optmal mechansms for a very general framework n the lterature on auctons wth costly entry. Our research s also related to a small lterature on auctons of entry rghts. In a poneerng work, Fullerton and McAfee (1999) ntroduce auctons for entry rghts to shortlst contestants for a fnal tournament. Ye (2007) extends ther approach to the settng of two-stage auctons descrbed above. Our current approach dffers from thers n the way the set of fnalsts s determned: whle n ther approach the number of fnalsts to be selected s fxed and pre-announced, n our entry rght allocaton mechansm the selecton of shortlsted bdders s contngent on the reported bd profle, makng the number of fnalsts endogenously determned. For ths reason the entry rght allocaton mechansm examned n ths research s more general. 10 In another relevant paper, Lu and Ye (2013) explore optmal two-stage mechansms n an envronment where bdders are characterzed by heterogenous and prvate nformaton acquston costs before entry. In that settng the pre-entry type s the entry cost, whch s nether correlated to nor part of the value of the asset for sale. As such, there s no beneft to make the second-stage mechansm contngent on the reports of the pre-entry types, resultng n a much smpler characterzaton of optmal mechansms. The settng n ths current paper s dfferent, as the pre-entry type s correlated to the value of the asset, hence there are potental gans to make the second-stage mechansm contngent on frst-stage reports. Indeed, n our current settng, the optmal allocaton and payment rules n the second stage do depend on the frst-stage reports. Therefore the characterzaton of optmal mechansms s more demandng n 9 See Bergemann and Välmäk (2006) for a thoughtful survey of ths lterature. 10 In fact, t resemble mult-unt auctons wth endogenously determned supply (see, e.g., McAdams, 2007). 4

5 ths work, and the mplementaton of the optmal mechansm s also more sophstcated. The rest of the paper s organzed as follows. Secton 2 presents the model. Secton 3 characterzes the optmal mechansm and ts aucton mplementaton. Secton 4 dscusses some restrctons n our analyss. Secton 5 concludes. 2 THE MODEL The nformaton structure n our model s closest to that n Esö and Szentes (2007). The man dfferences are that n Esö and Szentes, the addtonal nformaton s controlled by the seller, and they focus on the seller s ncentve to dsclose (wthout observng) addtonal sgnals to the buyers. In our settng, however, t s costly for the bdders to acqure addtonal nformaton, and we focus on the bdders ncentve for nformaton acquston (entry). In addton, all buyers are ncluded n the fnal sale n Esö and Szentes, but due to costly entry n our settng, not all buyers wll be wllng to partcpate n the fnal aucton. As such, we wll addtonally consder entry mechansms whch s the major dfference from the analyss n Esö and Szentes. Formally, a sngle ndvsble asset s offered for sale to N potentally nterested buyers. The seller and bdders are assumed to be rsk neutral. The seller s own valuaton for the asset s normalzed to 0. Buyer s true valuaton for the asset s v. However, ntally she only observes a nosy sgnal of t, α, whch s her prvate nformaton and can be nterpreted as her orgnal type. After ncurrng a common nformaton acquston cost (or entry cost) of c, bdder fully observes her ex post value, v. The pars, v ) are assumed to be ndependent across. 11 Ex ante, α follows dstrbuton F( ) wth ts assocated densty f ( ) on support α,α]. 12 We assume that f s postve on the nterval α,α] and satsfes the monotone hazard rate condton; that s, f /(1 F) s weakly ncreasng. Gven α, the ex post value v follows dstrbuton H α H( α ) wth ts densty h α h( α ) over support v,v] R. 13 The values N and c and dstrbutons F and H α are all common knowledge. Followng the sgnal orthogonalzaton technque ntroduced by Esö and Szentes (2007), 14 there exst functons u and s, such that u, s ) v, where u s strctly ncreasng n both arguments, and s s ndependent of α. In partcular, s can be constructed as follows: s = H(v α ), 11 As n Esö and Szentes (2007) and Pavan, Segal, and Tokka (2014), ths assumpton rules out the possblty of full rent extracton (Crémer and McLean, 1988). 12 Esö and Szentes allow α s to be drawn from dfferent dstrbutons. Our procedure can be extended to accommodate asymmetrc dstrbutons for α s. For ease of characterzng our optmal entry rght allocaton rule, we assume that α s are drawn from a common dstrbuton, so that bdders are ex ante symmetrc. Note that wth dfferent realzatons of α s, bdder heterogenety before entry s stll captured n our model. 13 Followng the dynamc mechansm desgn lterature, we assume that the support of v s ndependent of the frst-stage sgnal α. 14 The use of ths technque becomes standard n the lterature (see, e.g., Pavan, Segal, and Tokka, 2014, and Bergemann and Wambach, 2014). 5

6 whch s the percentle of the value realzaton to bdder. 15 valuaton can be computed as v = H 1 α (s ) u, s ). Thus gven type α and sgnal s, the We wll denote the c.d.f. of s by G. 16 We mantan the followng assumptons that are adopted n Esö and Szentes (2007): Assumpton 1. ( H α (v)/ α)/h α (v) s ncreasng n v. Assumpton 2. ( H α (v)/ α)/h α (v) s ncreasng n α. Esö and Szentes show that Assumpton 1 s equvalent to u 12 0 and Assumpton 2 s equvalent to u 11 /u 1 u 12 /u 2. Assumpton 1 thus states that the margnal mpact of the new nformaton on buyer s value s decreasng n her type α. Assumpton 2 mples that an ncrease n α, holdng u, s ) constant, weakly decreases the margnal value of α. Assumptons 1 and 2 can thus be nterpreted as a knd of substtutablty n buyer s posteror valuaton between α and s. Example 1. (Ye, 2007): Each potental bdder s endowed wth a prvate value component α before entry; after entry, each buyer learns another prvate value component s, where s s ndependent of α. The ex post value u, s )=α + s. By the lnearty of u, s ), Assumptons 1 and 2 hold. Example 2. (Adapted from Esö and Szentes, 2007): v s drawn from a normal dstrbuton wth mean µ and precson (nverse varance) τ 0. The pre-entry type, α, s normally dstrbuted wth mean v and precson τ v. After entry, the buyer can observe her true value, v. It s clear that v and α are strctly afflated. The dstrbuton of α, whch s normal, satsfes the hazard rate condton. The cdf of v condtonal on α, H α, s normal wth mean (τ 0 µ+τ v α )/(τ 0 + τ v ) and precson τ 0 + τ v. Defne s = H α (v ) and let u, s )= H 1 α (s ) v. Obvously u s strctly ncreasng n s. It can be verfed that u 1, s )=τ v /(τ 0 + τ v ), whch s a constant. Therefore, u s lnear and strctly ncreasng n α. Hence Assumptons 1 and 2 hold. Snce nformaton acquston s modeled as entry n our settng, we consder a mechansm desgn framework n whch the seller exercses entry control. In addton, we restrct our analyss to two-stage mechansms: the frst stage s the entry rght allocaton mechansm, and the second stage s the prvate good provson mechansm. Note that n ths mechansm desgn framework, the second-stage mechansm can be made contngent on the frst-stage reports. Followng Myerson (1986) and Pavan, Segal, and Tokka (2014), we restrct to drect mechansms where agents report ther types truthfully at each stage on the equlbrum path. We assume that the frst-stage reported profle α s revealed to all admtted bdders so that the frst-stage entry allocaton 15 It s easly seen that s s unformly dstrbuted over 0,1], and s hence statstcally ndependent of the ntal nformaton α. 16 G could be assumed to be unform on 0,1]. More generally, all s s satsfyng u,s ) v are postve monotonc transformaton of each other (Lemma 1 n Esö and Szentes). 6

7 and payments are mmedately verfable. 17 Ths revelaton polcy turns out to be optmal, n the sense that no other revelaton polcy (e.g., not revealng or partally revealng α) can generate hgher expected revenue to the seller. For ths reason, our restrcton to fully reveal α s wthout loss of generalty n our search for optmal mechansms. A detaled dscusson s relegated to Secton 4. As n Pavan, Segal, and Tokka (2014), the revelaton polcy concerned n ths paper s about the frst-stage nformaton and outcome, ncludng the agents frst-stage reports, ther payments, and the agents beng shortlsted. In our paper, the prncpal has no control over the ways n whch the second-stage new nformaton s revealed to bdders. A shortlsted bdder wll be fully nformed about her true value v after ncurrng the entry cost. As such, we are not concerned about the dscrmnatory nformaton dsclosure ssue studed n L and Sh (2013). As n Esö and Szentes, we can focus attenton on equvalent drect mechansms that requre bdders to report s s, rather than v s. Note that reportng,v ) s equvalent to reportng, s = H α (v )). Let N=1,2,..., N} denote the set of all the potental buyers and 2 N denote the collecton of all the subsets (subgroups) of N, ncludng the empty set, φ. The frst-stage mechansm s characterzed by the shortlstng rule A g ) and payment rule x ), = 1,2,..., N. Gven the reported profle α, the shortlstng rule, A g : α,α ] N 0,1], assgns a probablty to each subgroup g 2 N, where g 2 N A g ) = 1. The payment rule x : α,α ] N R, specfes bdder s frst-stage payment gven the reported profle α. 18 Gven the frst-stage reported profle α, the second-stage mechansm s characterzed by p g,sg ), the probablty that the asset s allocated to buyer g, and t g,sg ), the payment to the seller made by buyer g, g 2 N. 3 THE ANALYSIS We start wth the second stage. Suppose group g s shortlsted, and the profle α reported n the frst stage s revealed as publc nformaton to the shortlsted bdders. Frst, suppose α s truthfully reported at the frst stage and group g s shortlsted. Assume that they follow the recommendaton and ncur the nformaton acquston cost c to dscover s g. 19 Gven the announced α and s, defne the nterm wnnng probablty and expected payment rule as P g, s )= E g s p g,sg ) and T g, s )=E g s t g,sg ), where s g = sg \s }, g and g 2 N. Then bdder s second-stage nterm expected payoff when she observes s but reports ŝ s as follows: π g ;s, ŝ )=E g s u, s )p g, ŝ,s g ) tg, ŝ,s g )]= u, s )P g, ŝ ) T g, ŝ ). 17 In Esö and Szentes, there s no such need for nterm verfcaton, as ther allocaton and payment rules are executed at the end of the mechansm. 18 Note that our shortlstng rule A g ) s more general than specfyng each bdder s probablty of beng shortlsted gven ther reported type profle α. 19 As wll be shown, the equlbrum expected proft from gong forward s postve for a buyer upon entry, so n equlbrum, a bdder does have an ncentve to follow the recommendaton to acqure (costly) nformaton and partcpate n the fnal aucton once admtted (as droppng out only results n zero proft). 7

8 The second-stage ncentve compatblty (IC) condtons requre that π g ;s, ŝ ) π g ;s, s ), g,α,s, ŝ. (1) Frst, the followng lemma s standard n the tradtonal screenng lterature: Lemma 1. Suppose α s truthfully revealed from the frst stage and P g, s ), g, s contnuous and weakly ncreasng n s where g denotes the group beng shortlsted, then the second-stage ncentve compatblty condton (1) holds f and only f s π g ;s, s )= π g ;ŝ, ŝ )+ ŝ u 2,τ)P g,τ)dτ, s >ŝ, g. (2) (2) s an ntegral form of the envelope formula. Next, we consder the case when ˆα nstead of α s reported by bdder whle others report ther types truthfully. As demonstrated n Esö and Szentes (2007), whenever a bdder had msrepresented her type n the frst stage, she would correct her le n the second stage. Formally n our settng, suppose α s truthfully revealed from the frst stage and the second-stage mechansm s ncentve-compatble gven a truthfully revealed α. Then buyer of type α who reported ˆα n the frst round wll report ŝ = σ, ˆα, s ) f she observes s n the second stage such that 20 u, s )= u( ˆα,σ, ˆα, s )). (3) Reportng ŝ after a le ˆα s equvalent to revealng v truthfully regardless of the frst-stage report. The optmalty of ths strategy has been establshed n general for the Markov envronments by Pavan, Segal and Tokka (2014). Our two-stage settng resembles the Markov envronment defned n Pavan, Segal, and Tokka snce the agents payoffs only depend on ther second-stage true types (v s) and the allocaton outcome, but not on ther frst-stage true types. For ths reason, an agent s reportng ncentve n the second stage depends only on her current type and her frst-stage report, but not on her frst-stage true type. Snce t s optmal for the agent to report her value truthfully when the past report has been truthful, t s also optmal for her to report her value truthfully even f she has led n the frst stage. Note that ŝ does not depend on α, g, or s g. Defne π g, ˆα ;s, ŝ ) = E g s u, s )p g, ˆα, ŝ,s g ) tg, ˆα, ŝ,s g )] = u, s )P g, ˆα, ŝ ) T g, ˆα, ŝ ); π g, ˆα ;α ) = E s π g, ˆα ;s, ŝ = σ, ˆα, s )). π g, ˆα ;α ) s the expected second-stage payoff for the type-α bdder f she reported ˆα n the frst 20 The exstence of σ (,, ) reles on the assumpton that the support of v does not depend on the frst-stage sgnal α. 8

9 stage (and everyone else reported truthfully). Parallel to Lemma 5 n Esö and Szentes, we can show the followng lemma: Lemma 2. Suppose α s truthfully revealed from the frst stage and the second-stage mechansm s ncentve-compatble gven a truthfully revealed α. If buyer of type α who reported ˆα n the frst stage s shortlsted n group g, her expected payoff from the second stage s gven by α π g, ˆα ;α )= π g ( ˆα, ˆα ;α )+ ˆα u 1 (y, s )P g ( ˆα,α,σ (y, ˆα, s ))d ydg (s ). (4) Throughout, g wll be used to denote the group ncludng bdder. (4) should agan be regarded as an ntegral form of the envelope formula: the wnnng probablty (P g ) s now obtaned when evaluatng at ŝ = σ (y, ˆα, s ) (whch s optmal gven the frst-round le ˆα ). We are now ready to consder the frst-stage IC mechansm. Let π, ˆα ) be the expected payoff (net of the entry cost) for a type-α bdder who reports ˆα n the frst stage. By (3), we have π, ˆα ) = E α g A g ( ˆα,α ) π g, ˆα ;α ) c] x ( ˆα,α ) } (5) = E α g A g ( ˆα,α ) E s ( u, s )P g, ˆα, ŝ ) T g, ˆα, ŝ ) ) c ]} x ( ˆα ), where ŝ = σ, ˆα, s ) and x ( ˆα )= E α x ( ˆα,α ). The followng lemma characterzes the bdder s expected payoff n an IC two-stage mechansm wth costly entry. Lemma 3. If the two-stage mechansm s ncentve compatble and E α A g,α )P g,α, s ) s contnuous n α then buyer s expected payoff (as a functon of her pre-entry type) can be expressed as π,α )=π,α)+ α α u 1 (y, s ) Eα A g (y,α )P g (y,α, s ) ] dg (s )d y. (6) g Proof. See Appendx. Note that g Eα A g (y,α )P g (y,α, s ) ] s buyer s equlbrum probablty of eventually wnnng the asset wth sgnals (y, s ) n our settng. Thus (6) s also an ntegral form of the envelope formula. Under a set of regular condtons, Pavan, Segal, and Tokka (2014) show that the envelope formula contnues to hold n the dynamc mechansm desgn settng. Lemma 3 can be regarded as an extenson of ther result to a dynamc mechansm desgn settng wth costly nformaton acquston. 9

10 3.1 The Optmal Two-stage Mechansms We are now ready to derve the seller s expected payoff from an IC two-stage mechansm. By Lemma 3, we have Eπ,α ) = π,α)+ α α α α u 1 (y, s ) Eα A g (y,α )P g (y,α, s ) ] dg (s )d ydf ) g u 1, s ) α 1 F ) = π,α)+ Eα A g,α )P g α f ),α, s ) ] dg (s )df ) g = π,α)+ E α A g 1 F ) )]},α ) u 1, s )P g g f ),α, s )dg (s. The second equalty above s due to Fubn s Theorem. Thus ]} N N Eπ,α )=,α)+ E α A =1 =1π g )E s p g,sg ) 1 F ) u 1, s ). g g f ) The total expected surplus from the two-stage mechansm s ]} TS=E α A g )E s p g,sg )u, s ) g c. g The seller s expected revenue s thus gven by g ER = TS g N Eπ,α ) =1 ( = E α A g )E s p g,sg ) u, s ) 1 F ) ]} ) N u 1, s ) g c π,α), (7) g f ) =1 where A g ) s the shortlstng rule and p g,sg ) s the second-stage allocaton rule. To maxmze ER subject to IC and IR (ndvdual ratonalty), the seller sets π,α)=0 for all = 1,2,..., N;.e., no rent should be gven to the buyer wth the lowest possble (pre-entry) type. Defne the vrtual value adjusted by the second-stage sgnal as follows: w, s )= u, s ) 1 F ) u 1, s ). (8) f ) From the expresson of the expected revenue, we can derve the optmal allocaton rules n both stages 10

11 as follows. At the second stage, gven the revealed α and the shortlsted group g, s g, 21,s g 1 f = argmax j g w j, s j )} and w, s ) 0 )= g, g. (9) 0 otherwse p g So as also dentfed by Esö and Szentes, the asset should be awarded to the bdder wth the hghest non-negatve vrtual value adjusted by the second-stage sgnal, whch s a generalzaton of the optmal allocaton rule n Myerson (1981). Our analyss shows that the generalzed Myerson allocaton rule s robust to settngs wth costly entry. By Lemma 3, a buyer s expected payoff does not depend on the entry cost, whch mples that the seller bears all the entry costs (ndrectly) n equlbrum. As such, costly entry wll affect the fnal allocaton only through ts effect on the entry rght allocaton rule. Defne the expected vrtual surplus (the vrtual value less the entry cost) as follows: w g )= E s g p g,s g )w, s ) g c ]. Then at the frst stage, contngent on the revealed α, the optmal shortlstng rule s as follows: 22 A g 1 f g= argmax g w g )} and w g ) 0 )= g. (10) 0 otherwse The optmal shortlstng rule admts the set of bdders that gves rse to the maxmal expected vrtual surplus. Alternatvely, the optmal shortlstng rule admts the bdders n descendng order of ther margnal contrbuton to the expected vrtual surplus the bdder wth the hghest contrbuton frst, the bdder wth the second-hghest contrbuton second, etc. provded that ther margnal contrbuton s postve. Smlarly to Esö and Szentes, followng Assumptons 1 and 2, we can establsh the followng propertes of the optmal second-stage allocaton rule: 23 Corollary 1. () p g P g,s g ) ncreases n both α and s, g, g, α, and s g, whch mples that,α, s ) ncreases n both α and s, g, α ; () If α > ˆα, s < ŝ and u, s )= u( ˆα, ŝ ), then p g,α, s,s g ) p g ( ˆα,α, ŝ,s g ), whch mples P g,α, s ) P g ( ˆα,α, ŝ ), g, α. Property () above suggests that whenever α > ˆα, s < ŝ and u, s )= u( ˆα, ŝ ), the optmal allocaton rule favors the truth-tellng par, s ). Gven α, let s ) be defned such that w, s ))=0. To dentfy propertes of the optmal shortlst- 21 Tes occur wth probablty zero and are hence gnored. 22 Agan tes occur wth probablty zero and are hence gnored. 23 Assumpton 2 s used to show property (). 11

12 ng rule, we defne a truncated random varable as follows: w + w, s ) f w, s ) 0 or equvalently s s ), s )=. 0 otherwse Note that condtonal on α, w + s are ndependent across g. Let S g ;α ) denote buyer s margnal contrbuton to the expected vrtual surplus, g, then S g ;α )= S g ) S g ), g, αg, where α g = αg \α } and The followng two propertes are obvous: S g )=E s g max g w+, s )}, g, α g. (1) S g ;α ) ncreases wth α, and decreases wth α j, j, g, g. (2) S g ;α ) S g ;α ), α, g, g g. The revenue-optmal shortlstng rule can be alternatvely descrbed as follows. For gven α, we can rank all α from the hghest to the lowest. The seller admts bdders one by one n descendng order of α s as long as the bdder s margnal contrbuton to the expected vrtual surplus s nonnegatve,.e. S g ;α ) c= S g ) S g ) c 0, where g denotes the group of bdders wth the hghest g types before entry. Two propertes follow mmedately from the optmal shortlstng rule A g : Corollary 2. () Gven α, f bdder wth α s shortlsted, then she would also be shortlsted wth a hgher type α (> α ); () Suppose s shortlsted gven α, then bdder would reman beng shortlsted as long as α s hgher than a threshold ˆα ). As α ncreases, the shortlsted group weakly shrnks. As α ncreases from ˆα ), the bdders n g )\} would be excluded one by one (wth the lowest type orgnally shortlsted beng excluded frst). We are now ready to show that the optmal fnal good allocaton and entry rght allocaton rules (9) and (10) are truthfully mplementable by some well constructed payment rules n both stages. Theorem 1. Under Assumptons 1 and 2, the optmal fnal good allocaton and entry rght allocaton rules (9) and (10) are IR and IC mplementable. Proof. u, s ) ncreases wth s and by Assumpton 1, u 1, s ) (weakly) decreases wth s. Ths mples that w, s ) ncreases wth s. By the fnal good allocaton rule (9), the wnnng probablty P g, s ) s weakly ncreasng n s. By Lemma 1, the second-stage mechansm s ncentve compatble (gven α and g). Thus, gven the truthfully revealed α and shortlsted group g, a second-stage payment rule, 12

13 p g say, t g,s g ), g, g, can be constructed to truthfully mplement the second-stage allocaton rule,s g ), g, g whle mantanng the second-stage IR constrants (to partcpate n the secondstage mechansm),.e. π g,α ;s, s ) 0 on equlbrum path. Ths resembles the Myerson (1981) settng wth asymmetrc bdders. We use π g, ˆα ;α ) to denote the second-stage expected payoff to buyer of type α f she announces ˆα and s shortlsted n group g, gven that everyone else announces α truthfully at the frst stage. π g, ˆα ;α ) s well defned gven Lemma 2. Therefore, when buyer of type α announces ˆα whle others reveal α truthfully, her frst-stage expected payoff can be wrtten as follows: π, ˆα )=E α g A g ( ˆα,α ) π g, ˆα ;α ) c] x ( ˆα,α ) }, where x s the frst-stage payment rule. Next, we wll show that the optmal shortlstng rule (10) s truthfully mplementable by a properly chosen frst-stage payment rule x, together wth the second-stage payment rules t g chosen above. Note that by (5), we have π,α )=E α g A g,α ) π g,α ;α ) c] x,α ) }. (11) Construct the frst-stage payment rule as follows: x ) = A g,α ) π g,α ;α ) c] g α u 1 (y, s ) Eα A g (y,α )P g (y,α, s ) ] dg (s )d y (12) α g Substtutng (12) nto (11), we can verfy that π,α )= α α u 1 (y, s ) Eα A g (y,α )P g (y,α, s ) ] dg (s )d y, g whch s precsely equaton (6) wth π,α)=0 (the optmalty requrement). Note that π,α ) 0, so IR s satsfed n the frst stage. Suppose that all buyers except report ther types α truthfully. Consder buyer wth α contemplatng to msreport to ˆα < α. The devaton payoff s =π, ˆα ) π,α )=π, ˆα ) π ( ˆα, ˆα )]+π ( ˆα, ˆα ) π,α )]. 13

14 Snce (6) s satsfed by the constructon of x ), we have π ( ˆα, ˆα ) π,α )= α ˆα u 1 (y, s ) Eα A g (y,α )P g (y,α, s ) ] dg (s )d y. g Recall the defntons of π, ˆα ) above, we have from Lemma 2 that π, ˆα ) π ( ˆα, ˆα )= Therefore, we have = α + ˆα α α ˆα E α A g (y,α ) g ˆα E α u 1 (y, s ) Eα A g ( ˆα,α )P g ( ˆα,α,σ (y, ˆα, s )) ] dg (s )d y. g u 1 (y, s )P g g A g ( ˆα,α ) A g (y,α )] ( ˆα,α,σ (y, ˆα, s )) P g (y,α, s )]dg (s )d y u 1 (y, s )P g ( ˆα,α,σ (y, ˆα, s ))dg (s )d y. (13) From Corollary 1 (), we have P g ( ˆα,α,σ (y, ˆα, s )) P g (y,α, s ) 0, whch mples that the frst term n s negatve. We now consder the second term n when y > ˆα. By Corollary 2, the optmal shortlstng rule mples that gven α, when buyer s admtted wth a hgher α, she must be admtted to a group wth a weakly smaller sze. If y and ˆα are admtted n the same group, then A g ( ˆα,α )= A g (y,α ) and ths term n s zero. We now turn to the case where g ( ˆα,α ) g (y,α ) }. Note that A g (,α ) s 1 for the shortlsted group, and 0 for all other groups. Therefore, A g ( ˆα,α ) A g (y,α )]u 1 (y, s )P g g = u 1 (y, s )P g ( ˆα,α ) 0, ( ˆα,α,σ (y, ˆα, s )) ( ˆα,α,σ (y, ˆα, s )) P g (y,α ) ( ˆα,α,σ (y, ˆα, s ))] whch mples that the second term n s negatve. Snce g ( ˆα,α ) g (y,α ) }, we must have P g ( ˆα,α ) ( ˆα,α,σ (y, ˆα, s )) P g (y,α ) ( ˆα,α,σ (y, ˆα, s )),.e. entrant wns wth a smaller probablty f a strctly bgger group s shortlsted. A smlar argument can be used to rule out devaton to ˆα > α. It s worth notng that Assumptons 1 and 2 are suffcent but not necessary for the optmal entry rule to be truthfully mplementable: the necessary and suffcent condton s that defned n (13) s non-postve, whch s also the ntegral monotoncty condton characterzed by Pavan, Segal, and Tokka (2014). Example 3. Assumptons 1 and 2 are not necessary for Theorem 1 to hold. One can verfy that for the 14

15 case wth H α (v)= v α, where v 0,1], Assumpton 1 fals, but Corollary 1 holds. In ths case, v=s 1/α u, s). Thus ] ] w, s)= s 1/α 1 F) 1 1 F) 1 1+ f ) α logs1/α = u, s) 1+ logu, s). f ) α When there s only one potental bdder, the second-stage allocaton rule can be mplemented va a take-t-or-leave t offer wth a prce P) = s) 1/α, where s) s defned by w, s)) = 0. Note that } s)=exp α 2 f ), whch decreases wth α. Thus P ) < 0. Defne w) = E s max0,w, s)}. The 1 F) optmal shortlstng rule s gven by A)=1 f and only f α α, where w )=0. We next derve the frst-stage payment rule. Note that 1 ] π, ˆα)= A( ˆα) max0, s 1/α P( ˆα)}ds c x( ˆα), α, ˆα α. 0 The FOC requres that A)=1. We thus have π, ˆα) ˆα ˆα=α = 0. Takng ˆα>α α, we have s( ˆα)<s) s ). Note A( ˆα)= π, ˆα) π,α)= s) s( ˆα) Snce s) s( ˆα) s1/α P( ˆα)]ds/ ˆα ˆα=α = 0, we have 1 s 1/α P( ˆα)]ds+ P) P( ˆα)]ds+x) x( ˆα)]. s) π, ˆα) ˆα ˆα=α= P )(1 s)) x )=0. Thus x )= P )(1 s)), together wth boundary condton x )= 1 0 max0, s1/α P )}ds c, jontly determnes the payment functon x) for the frst stage. 3.2 Implementaton of Optmal Mechansms When u, s ) s lnear n α,.e., when u, s )= u 1 α + r(s ) for some constant u 1 and functon r, we wll demonstrate that the optmal mechansm can be mplemented va a two-stage aucton, wth the frst stage beng an aucton for both entry rghts and prce prema and the second stage beng a second-prce or Englsh aucton for the fnal good. Ths two-stage aucton can be regarded as a handcap aucton ntroduced n Esö and Szentes, augmented by an addtonal aucton at the entry stage. 24 More specfcally, our two-stage aucton works as follows. The frst stage s an all-pay aucton, where bdders need to pay what they bd, regardless of beng awarded entry rghts or not. Suppose buyer, knowng her type α, bds an amount b, = 1,2,..., N. After all the frst-stage bds are collected, underlyng types wll be recovered based on a recovery functon, x 1, such that buyer s perceved type α s x 1 (b ), = 1,2,..., N. Gven the recovered type profle α } N =1, the entry rghts are mplemented 24 The assumpton that u 1 s constant s satsfed n both Examples 1 and 2. 15

16 accordng to the optmal entry rule (10), and a prce premum s determned for each shortlsted buyer accordng to the followng premum schedule: p( α ) = u 1 (1 F( α ))/f ( α ). Both the recovery functon x 1 and the premum determnaton rule p are made publc at the outset of the game, whch reman common knowledge throughout the aucton process. Upon beng admtted, each entrant bdder wll ncur the nformaton acquston cost and partcpate n the second-round bddng. The second stage s a tradtonal second-prce or Englsh aucton wth a zero reserve prce, but the wnner s requred to pay her premum over the prce. 25 Ths mechansm s referred to as the handcap aucton n Esö and Szentes, snce the buyers compete under unequal condtons: a bdder wth a smaller premum has an advantage. In our settng, the handcap aucton s modfed so that the optmal entry rule s also mplemented after the frst-round bddng. In Esö and Szentes, buyers pay fees regardless of wnnng the fnal good or not; n our settng, buyers pay b s regardless of beng admtted to the fnal sale or not. For ths reason, the frst-stage aucton s a varant of the all-pay aucton. In the second-stage aucton, t s a (weakly) domnant strategy for entrant bdder wth a prce premum p (determned from the frst stage) to bd u, s ) p. Assumng that all the entrant buyers follow ths weakly domnant strategy n the second stage, the mechansm can be represented by a par of functons, p : α,α] R + and x : α,α] R for = 1,2,..., N, where p ) s the prce premum for a buyer who bds an amount of b = x ). Theorem 2. If u 1 s constant then the optmal mechansm of Theorem 1 can be mplemented va a twostage aucton descrbed above wth the recovery functon x 1 and prce premum functon p defned as follows: p ) = 1 F ) u 1, (14) f ) x ) = E α, s ),0} c ] A g,α )E s g maxw, s ) w g g α E α A g,α )E s g α g u 1 (y, s )1 } w(y,s )>w g,s g ) where w, s )= u, s ) p ) and w g, s )=max j, j g wj, s j ),0 }. Proof. See Appendx. ] d y, (15) The mplementaton s establshed by showng that x ( ) as defned n (15) consttutes a symmetrc (strctly) ncreasng equlbrum bd functon n the (reduced) all-pay aucton game, wth the second stage beng replaced by ts assocated equlbrum payoffs. A major (and tedous) step n the proof of Theorem 2 s to establsh that x ) as defned n (15) s strctly ncreasng for α α,α], where α,α) s the mnmum type that could possbly be allocated wth an entry rght n equlbrum. 26 Thus the recovery 25 Should there be only one entrant, the prce premum for ths sole entrant becomes the effectve reserve prce. 26 α s defned such that } E s max u,s ) 1 F ) f u 1,0 = c. ) 16

17 functon x 1 ( ) s well defned over α,α] and a (truncated) profle of pre-entry types can be recovered from ther bds. 27 Optmal entry can then be mplemented based on the recovered type profle accordng to (10). Upon beng selected n a group, say, g, everyone wll follow the (weakly) domnant strateges n the second round bddng (to bd ther value less the prce premum), so buyer g wth pre-entry type α wll wn the asset f and only f u, s ) 1 F ) u 1 max 0, max u j, s j ) 1 F j) u 1 ]}. f ) j g, j f j ) Hence the optmal allocaton rule (9) can ndeed be mplemented, provded that bddng accordng to x ( ) consttutes a symmetrc equlbrum n the (reduced) frst-stage aucton game, whch s establshed n the second step of the proof. When u 1 s not a constant, n partcular, f u 1 s a functon of s, then one s optmal prce premum also depends on her second-stage sgnal. As such, one s second-stage bd also affects the (total) prce to pay should she wn the object even under a second-prce aucton. A drect consequence s that t s no longer a (weakly) domnant strategy for bdder to bd w, s ). To avod such an nconvenence, as n Esö and Szentes, we also focus on the case n whch u 1 s a constant for aucton mplementaton. 28 Snce x ( ) s ncreasng whle p( ) s decreasng, the prce premum s decreasng n the frst-stage bds. Thus a buyer wth a hgher pre-entry type bds hgher n the frst round, whch results n a hgher probablty to be admtted and a lower prce premum. The equlbrum (entry) fee x has an ntutve nterpretaton. As can be seen from (15), one s entry fee equals her expected proft from entry less her nformatonal rent due to her prvate nformaton about her type α. So the addtonal nformaton from the second-stage sgnals does not contrbute to buyers rents: the seller approprates all rents from entry by chargng each buyer an upfront entry fee equal to her value of entry (or equvalently, the value of addtonal nformaton). 29 Fxng the aucton rules n the second round, t s also clear by the envelope theorem that payoff e- quvalence holds among all the entry mechansms n whch (1) the same entry rule (10) s mplemented, and (2) the buyer wth type α makes zero expected proft. In lght of ths equvalence result, based on x ) we can derve the canddate equlbrum bd functon under any alternatve entry mechansm. If the canddate equlbrum bd functon so derved s strctly ncreasng, then the nverse of the equlbrum bd functon can serve as the recovery functon for the mplementaton of optmal entry. For example, we can consder a dscrmnatory-prce aucton, where only bdders who are awarded entry rghts need to pay, and they pay what they bd. Usng the payoff equvalence, the canddate equlbrum bd functons That s, α s the mnmal type that one can possbly be shortlsted (as a sole entrant). We assume α,α). 27 We assume that buyers wth types below α wll stay away from bddng. But that should not affect the mplementaton of optmal entry as those buyers should not be admtted anyway. 28 The aucton mplementaton n the envronment where u 1 depends on s s an open queston. 29 Ths seems to be a robust predcton n optmal dynamc mechansm desgn (see, for example, Courty and L, 2000 and Pavan, Segal, and Tokka, 2014). As ponted out n Esö and Szentes, the value of addtonal nformaton s not well defned, however, as t depends on the specfc rules n the second-round aucton. 17

18 under the dscrmnatory-prce aucton s gven by x D )= x )/Pr α g,α ) }, where g,α ) s the set of shortlsted bdders determned by the optmal entry rule (10), gven the reported type profle,α ). A dscrmnatory-prce aucton can mplement optmal entry f and only f x D ) so derved s strctly ncreasng. 3.3 Applcatons Our optmal mechansm analyss s general enough to encompass many exstng models n the lterature on auctons wth costly entry. Below we demonstrate how we can apply our general optmal mechansm to specal models prevously studed. 1. Bdders do not have pre-entry types and only learn about ther values after entry (e.g., McAfee and McMllan, 1987; Tan, 1992; and Levn and Smth, 1994). In ths case, u, s ) = s. Hence w, s )= s, whch mples that the optmal aucton s ex post effcent, and the optmal entry s to select a set of bdders that results n the maxmal expected socal surplus. Snce bdders are dentcal before entry, optmal entry s entrely characterzed by n, the optmal number of bdders to be selected. The mplementaton s somewhat smple: the second round s a standard aucton (frst-prce, second-prce, or Englsh aucton no prce premum s nvolved). The frst round (entry stage) s to select exactly n bdders, and whomever selected s requred to pay an upfront entry fee e, whch s set so that no rent s left for the entrants ex ante. 2. Bdders know ther values before entry, and entry s merely a bd preparaton process (wthout value updatng) (e.g. Samuelson,1985; Stegeman, 1996; Campbell, 1998; Menezes and Montero, 2000; Tan and Ylankaya, 2006; Cao and Tan, 2009; and Lu, 2009). In ths settng, u, s )=α, and hence w, s ) = α (1 F ))/f ). It s easly verfed that accordng to Theorem 1, the optmal allocaton rules can be descrbed as follows: the bdder wth the hghest type ) s admtted as the sole entrant to wn the tem, as long as her contrbuton to the vrtual surplus w, s ) c s postve. The optmal mechansm can be mplemented as follows: each buyer pays what she bds n the frst stage (regardless of beng admtted or not), and the only entrant wns the tem at a prce equal to her prce premum determned from her frst-round bd. For an llustraton, below we derve the equlbrum frst-stage bd functon x. Consder a bdder wth type α > α, where α (1 F ))/f ) = c. Suppose, n the (reduced) frst-stage drect game, bdder reports ˆα, a suffcently small devaton from α. Her expected payoff s then gven by π, ˆα ) = α 1 F( ˆα ] ( ) ) c Pr α f ( ˆα ) (1) < ˆα x ( ˆα ) 18

19 = α 1 F( ˆα ] ) c F f ( ˆα ) α ( ˆα ) x ( ˆα ), (16) (1) where α s the hghest type among all the buyers other than. (1) Incentve compatblty mples dπ,α ) dα = F α ). (1) Thus we have π,α )= α 1 F ] ) α c F f ) α ) x )= F (1) α (τ) dτ. (17) α (1) Substtutng ˆα = α nto (16) to obtan the equlbrum expected payoff, and then equatng that wth (17), gves the (symmetrc) equlbrum frst-stage bd functon x )= α 1 F ] ) α c F f ) α ) F (1) α (τ) dτ. (18) α (1) It s easly verfed that x ( ) s strctly ncreasng so types can be recovered from frst-round bds n equlbrum. If a dscrmnatory-prce aucton s conducted nstead, by payoff equvalence the canddate equlbrum bd functon n the frst round s gven by x D )= x ) F α (1) ) = α 1 F ) f ) ] c 1 F α ) (1) α α F α (1) (τ) dτ. (19) It s also easly verfed that x D ( ) s strctly ncreasng. Thus the dscrmnatory-prce aucton works to mplement optmal entry n ths context as well. 3. Each bdder s endowed wth pre-entry type α, and learns an addtonal prvate value component s (e.g., Ye, 2007; Qunt and Hendrcks, 2013). The total value s gven by u, s )=α + s. Hence w, s )=α +s (1 F ))/f ). The optmal second-stage allocaton rule thus requres that the asset be allocated to the entrant bdder wth the hghest vrtual value w, s ) provded that t s nonnegatve. The optmal entry rule requres that bdders be admtted n descendng order of ther pre-entry types, as long as ther contrbuton to the expected vrtual surplus s nonnegatve. To further llustrate the optmal entry rule, we assume that α s dstrbuted unformly over 0,1] and s follows a Bernoull dstrbuton, takng value 1 ( Hgh ) wth probablty q and 0 ( Low ) wth probablty 1 q. Then w, s )=2α + s 1. If only one buyer (the one wth the hghest type α (1) ) s admtted, the expected vrtual surplus s gven by w 1 = E ( 2α (1) + s 1 1 ) c= 2α (1) + q 1 c. So the optmal number of entrants n 1 f 2α (1) + q 1 c 0. For ease of computaton we assume 19

20 that α (1) α (2).5 (so that the vrtual values from the top two bdders are guaranteed to be nonnegatve). If two top buyers are admtted, the expected vrtual surplus s gven by w 2 = E max 2α (1) + s 1 1,2α (2) + s 2 1 }] 2c = E max 2α (1) + s 1,2α (2) + s 2 }] 1 2c = Pr(s 1 = 1) (2α (1) +1 ) +Pr(s 1 = s 2 = 0) 2α (1) +Pr(s 1 = 0, s 2 = 1) (2α (2) +1 ) 1 2c = q (2α (1) +1 ) +(1 q) 2 2α (1) +(1 q)q (2α (2) +1 ) 1 2c So the optmal number of entrants n 2 f the ncremental expected vrtual surplus w=w 2 w 1 = q(1 q) 1 2 ( α (1) α (2) )] c 0. Contnung ths procedure of calculaton, 30 t can be verfed that n n f q(1 q) n ( α (1) α (n) )] c 0, 31 or α (1) α (n) c q(1 q) n 1 ]. (20) Ths condton s ntutve: the admsson of the n-th hghest buyer s more lkely to be justfed f (1) the probablty that she wll turn out to be the wnner n the second round s suffcently hgh; (2) the entry cost s suffcently low; or (3) her type s suffcently close to the hghest type. It s thus clear that the optmal number of entrants, n, s determned by the followng condtons: α (1) α (n ) c q(1 q) n 1 ], α (1) α (n +1)> c q(1 q) n ]. 4 DISCUSSION In our precedng analyss, we have focused on the revelaton polcy so that the frst-stage reports are fully revealed to the shortlsted bdders. Due to ths partcular revelaton polcy, one concern s that there mght be some loss of generalty n dentfyng optmal mechansms. To address ths concern, we next dentfy an upper bound for the expected revenue that can be acheved by examnng a relaxed settng by droppng the IC and IR constrants for the shortlsted n the second stage so that all shortlsted bdders must ncur entry costs to learn ther second-stage sgnals as n our orgnal setup, and regardless of ther second-stage sgnals, they must partcpate n the second-stage sellng mechansm and report truthfully ther second stage sgnals. As a result, regardless of the dsclosure polcy of the frst-stage reports, the hghest possble expected revenue achevable n ths relaxed settng should mpose an upper bound for the expected revenue that can be obtaned n our orgnal setup, where the bdders second-stage IC and IR must both be satsfed. A useful observaton s that n the relaxed settng, bdders can only msreport ther frst-stage sgnals, and the shortlsted buyers belefs on buyers frst-stage type profles have no 30 We contnue to consder the case α (1) > α (2)... α (n).5 so that the vrtual value from these buyers wll be postve. 31 The addton of the nth hghest buyer only contrbutes to the expected vrtual surplus when she turns out to be the only one havng a good shot n the second stage (.e., s n = 1, whle s 1 =...= s n 1 = 0). 20

21 mpact on ther second-stage decsons (as shortlsted must enter and truthfully report ther second-stage sgnals). Ths observaton mples that the revelaton polcy of the frst-stage reports s not relevant to the mechansm desgn n the relaxed settng. Consequently, the hghest expected revenue attanable n ths relaxed settng does not depend on the prevalng dsclosure polcy of the frst-stage sgnals. We next proceed to dentfy ths bound. In the relaxed settng, the mechansms are specfed exactly the same as n Secton 2. All potental bdders report ther types α, gvng rse to a reported type profle α. The mechansm specfes the frststage shortlstng rule A g ) and payment rule x,α ). Every shortlsted bdder j ncurs cost c to dscover her second-stage sgnal s j. The second-stage sellng mechansm specfes the wnnng probablty p g,sg ) and payment rule t g,sg ), g, g 2 N. Recall that P g, s ) = E g s p g,sg ) and T g, s ) = E g s t g,sg ). For shortlsted bdder g wth type α, her nterm expected payoff when she reports ˆα and others report truthfully s gven by π, ˆα )= E α g A g ( ˆα,α )E s ((u, s )P g ( ˆα,α, s ) T g ( ˆα,α, s )) c] x ( ˆα,α )}. (21) Applyng the envelope theorem, the IC condton π,α ) π, ˆα ) leads to the followng necessary condton: Therefore, we have dπ,α ) = π, ˆα ) ˆα =α dα α = E α A g,α )E s u 1, s )P g,α, s )]}. g π,α ) α = π,α)+ E α A g (y,α )E s u 1 (y, s )P g (y,α, s )]d y α g α = π,α)+ E α u 1 (y, s ) A g (y,α )P g (y,α, s )dg (s )d y. α g Note that the above expresson s exactly the same as (6), whch mples that the seller s expected revenue must be the same as n (7); n other words, the expected revenue upper bound n the relaxed settng s acheved n our orgnal settng. In ths sense, there s no loss of generalty to derve optmal mechansms by only consderng mechansms that fully reveal the buyers frst-stage reports to all admtted bdders. Another mportant aspect n our analyss s that we model nformaton acquston as entry. An mplcaton s that nformaton acquston s mandatory, n the sense that a bdder s not allowed to bd wthout gong through the due dlgence process. Ths assumpton s due to the specfc nsttutonal setup we are tryng to model. For example, data rooms are usually provded by the sellng party to dsclose a large amount of confdental data to bdders durng the due dlgence process. A typcal data room s a contnually montored space that the bdders and ther advsers wll vst n order to nspect and report on the varous documents and data made avalable. Often only one bdder at a tme wll 21

Optimal Two-stage Auctions with Costly Information Acquisition

Optimal Two-stage Auctions with Costly Information Acquisition Optmal Two-stage Auctons wth Costly Informaton Acquston Jngfeng Lu Lxn Ye May 2014 Abstract We consder an aucton envronment wth costly entry wheren the cost manly stems from nformaton acquston. Bdders

More information

Optimal Two-stage Auctions with Costly Information Acquisition

Optimal Two-stage Auctions with Costly Information Acquisition Optmal Two-stage Auctons wth Costly Informaton Acquston Jngfeng Lu Lxn Ye Ths verson: March 2017 Abstract We study optmal two-stage mechansms n an aucton envronment where bdders are endowed wth orgnal

More information

Module 17: Mechanism Design & Optimal Auctions

Module 17: Mechanism Design & Optimal Auctions Module 7: Mechansm Desgn & Optmal Auctons Informaton Economcs (Ec 55) George Georgads Examples: Auctons Blateral trade Producton and dstrbuton n socety General Setup N agents Each agent has prvate nformaton

More information

Welfare Properties of General Equilibrium. What can be said about optimality properties of resource allocation implied by general equilibrium?

Welfare Properties of General Equilibrium. What can be said about optimality properties of resource allocation implied by general equilibrium? APPLIED WELFARE ECONOMICS AND POLICY ANALYSIS Welfare Propertes of General Equlbrum What can be sad about optmalty propertes of resource allocaton mpled by general equlbrum? Any crteron used to compare

More information

CS286r Assign One. Answer Key

CS286r Assign One. Answer Key CS286r Assgn One Answer Key 1 Game theory 1.1 1.1.1 Let off-equlbrum strateges also be that people contnue to play n Nash equlbrum. Devatng from any Nash equlbrum s a weakly domnated strategy. That s,

More information

Vickrey Auction VCG Combinatorial Auctions. Mechanism Design. Algorithms and Data Structures. Winter 2016

Vickrey Auction VCG Combinatorial Auctions. Mechanism Design. Algorithms and Data Structures. Winter 2016 Mechansm Desgn Algorthms and Data Structures Wnter 2016 1 / 39 Vckrey Aucton Vckrey-Clarke-Groves Mechansms Sngle-Mnded Combnatoral Auctons 2 / 39 Mechansm Desgn (wth Money) Set A of outcomes to choose

More information

2E Pattern Recognition Solutions to Introduction to Pattern Recognition, Chapter 2: Bayesian pattern classification

2E Pattern Recognition Solutions to Introduction to Pattern Recognition, Chapter 2: Bayesian pattern classification E395 - Pattern Recognton Solutons to Introducton to Pattern Recognton, Chapter : Bayesan pattern classfcaton Preface Ths document s a soluton manual for selected exercses from Introducton to Pattern Recognton

More information

Assortment Optimization under MNL

Assortment Optimization under MNL Assortment Optmzaton under MNL Haotan Song Aprl 30, 2017 1 Introducton The assortment optmzaton problem ams to fnd the revenue-maxmzng assortment of products to offer when the prces of products are fxed.

More information

Econ107 Applied Econometrics Topic 3: Classical Model (Studenmund, Chapter 4)

Econ107 Applied Econometrics Topic 3: Classical Model (Studenmund, Chapter 4) I. Classcal Assumptons Econ7 Appled Econometrcs Topc 3: Classcal Model (Studenmund, Chapter 4) We have defned OLS and studed some algebrac propertes of OLS. In ths topc we wll study statstcal propertes

More information

Microeconomics: Auctions

Microeconomics: Auctions Mcroeconomcs: Auctons Frédérc Robert-coud ovember 8, Abstract We rst characterze the PBE n a smple rst prce and second prce sealed bd aucton wth prvate values. The key result s that the expected revenue

More information

(1 ) (1 ) 0 (1 ) (1 ) 0

(1 ) (1 ) 0 (1 ) (1 ) 0 Appendx A Appendx A contans proofs for resubmsson "Contractng Informaton Securty n the Presence of Double oral Hazard" Proof of Lemma 1: Assume that, to the contrary, BS efforts are achevable under a blateral

More information

Pricing Network Services by Jun Shu, Pravin Varaiya

Pricing Network Services by Jun Shu, Pravin Varaiya Prcng Network Servces by Jun Shu, Pravn Varaya Presented by Hayden So September 25, 2003 Introducton: Two Network Problems Engneerng: A game theoretcal sound congeston control mechansm that s ncentve compatble

More information

The Multiple Classical Linear Regression Model (CLRM): Specification and Assumptions. 1. Introduction

The Multiple Classical Linear Regression Model (CLRM): Specification and Assumptions. 1. Introduction ECONOMICS 5* -- NOTE (Summary) ECON 5* -- NOTE The Multple Classcal Lnear Regresson Model (CLRM): Specfcaton and Assumptons. Introducton CLRM stands for the Classcal Lnear Regresson Model. The CLRM s also

More information

Abstract Single Crossing and the Value Dimension

Abstract Single Crossing and the Value Dimension Abstract Sngle Crossng and the Value Dmenson Davd Rahman September 24, 2007 Abstract When auctonng an ndvsble good wthout consumpton externaltes, abstract sngle crossng s necessary and suffcent to mplement

More information

Implementation and Detection

Implementation and Detection 1 December 18 2014 Implementaton and Detecton Htosh Matsushma Department of Economcs Unversty of Tokyo 2 Ths paper consders mplementaton of scf: Mechansm Desgn wth Unqueness CP attempts to mplement scf

More information

Optimal Dynamic Mechanism Design and the Virtual Pivot Mechanism

Optimal Dynamic Mechanism Design and the Virtual Pivot Mechanism Optmal Dynamc Mechansm Desgn and the Vrtual Pvot Mechansm Sham M. Kakade Ilan Lobel Hamd Nazerzadeh March 25, 2011 Abstract We consder the problem of desgnng optmal mechansms for settngs where agents have

More information

A Generalized Vickrey Auction

A Generalized Vickrey Auction A Generalzed Vckrey Aucton Lawrence M. Ausubel* Unversty of Maryland September 1999 Abstract In aucton envronments where bdders have pure prvate values, the Vckrey aucton (Vckrey, 1961) provdes a smple

More information

Equilibrium with Complete Markets. Instructor: Dmytro Hryshko

Equilibrium with Complete Markets. Instructor: Dmytro Hryshko Equlbrum wth Complete Markets Instructor: Dmytro Hryshko 1 / 33 Readngs Ljungqvst and Sargent. Recursve Macroeconomc Theory. MIT Press. Chapter 8. 2 / 33 Equlbrum n pure exchange, nfnte horzon economes,

More information

Introduction. 1. The Model

Introduction. 1. The Model H23, Q5 Introducton In the feld of polluton regulaton the problems stemmng from the asymmetry of nformaton between the regulator and the pollutng frms have been thoroughly studed. The semnal works by Wetzman

More information

Perfect Competition and the Nash Bargaining Solution

Perfect Competition and the Nash Bargaining Solution Perfect Competton and the Nash Barganng Soluton Renhard John Department of Economcs Unversty of Bonn Adenauerallee 24-42 53113 Bonn, Germany emal: rohn@un-bonn.de May 2005 Abstract For a lnear exchange

More information

Market structure and Innovation

Market structure and Innovation Market structure and Innovaton Ths presentaton s based on the paper Market structure and Innovaton authored by Glenn C. Loury, publshed n The Quarterly Journal of Economcs, Vol. 93, No.3 ( Aug 1979) I.

More information

k t+1 + c t A t k t, t=0

k t+1 + c t A t k t, t=0 Macro II (UC3M, MA/PhD Econ) Professor: Matthas Kredler Fnal Exam 6 May 208 You have 50 mnutes to complete the exam There are 80 ponts n total The exam has 4 pages If somethng n the queston s unclear,

More information

Games of Threats. Elon Kohlberg Abraham Neyman. Working Paper

Games of Threats. Elon Kohlberg Abraham Neyman. Working Paper Games of Threats Elon Kohlberg Abraham Neyman Workng Paper 18-023 Games of Threats Elon Kohlberg Harvard Busness School Abraham Neyman The Hebrew Unversty of Jerusalem Workng Paper 18-023 Copyrght 2017

More information

Economics 101. Lecture 4 - Equilibrium and Efficiency

Economics 101. Lecture 4 - Equilibrium and Efficiency Economcs 0 Lecture 4 - Equlbrum and Effcency Intro As dscussed n the prevous lecture, we wll now move from an envronment where we looed at consumers mang decsons n solaton to analyzng economes full of

More information

Composite Hypotheses testing

Composite Hypotheses testing Composte ypotheses testng In many hypothess testng problems there are many possble dstrbutons that can occur under each of the hypotheses. The output of the source s a set of parameters (ponts n a parameter

More information

Module 9. Lecture 6. Duality in Assignment Problems

Module 9. Lecture 6. Duality in Assignment Problems Module 9 1 Lecture 6 Dualty n Assgnment Problems In ths lecture we attempt to answer few other mportant questons posed n earler lecture for (AP) and see how some of them can be explaned through the concept

More information

Endogenous timing in a mixed oligopoly consisting of a single public firm and foreign competitors. Abstract

Endogenous timing in a mixed oligopoly consisting of a single public firm and foreign competitors. Abstract Endogenous tmng n a mxed olgopoly consstng o a sngle publc rm and oregn compettors Yuanzhu Lu Chna Economcs and Management Academy, Central Unversty o Fnance and Economcs Abstract We nvestgate endogenous

More information

Online Appendix. t=1 (p t w)q t. Then the first order condition shows that

Online Appendix. t=1 (p t w)q t. Then the first order condition shows that Artcle forthcomng to ; manuscrpt no (Please, provde the manuscrpt number!) 1 Onlne Appendx Appendx E: Proofs Proof of Proposton 1 Frst we derve the equlbrum when the manufacturer does not vertcally ntegrate

More information

Credit Card Pricing and Impact of Adverse Selection

Credit Card Pricing and Impact of Adverse Selection Credt Card Prcng and Impact of Adverse Selecton Bo Huang and Lyn C. Thomas Unversty of Southampton Contents Background Aucton model of credt card solctaton - Errors n probablty of beng Good - Errors n

More information

Problem Set 2 Solutions

Problem Set 2 Solutions FDPE Mcroeconomcs 4: Informaton Economcs Sprng 07 Juuso Välmäk TA: Chrstan Krestel Problem Set Solutons Problem Prove the followng clam: Let g,h : [0, ) R be contnuous and dfferentable such that ) g (0)

More information

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

Module 3 LOSSY IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur Module 3 LOSSY IMAGE COMPRESSION SYSTEMS Verson ECE IIT, Kharagpur Lesson 6 Theory of Quantzaton Verson ECE IIT, Kharagpur Instructonal Objectves At the end of ths lesson, the students should be able to:

More information

3.1 Expectation of Functions of Several Random Variables. )' be a k-dimensional discrete or continuous random vector, with joint PMF p (, E X E X1 E X

3.1 Expectation of Functions of Several Random Variables. )' be a k-dimensional discrete or continuous random vector, with joint PMF p (, E X E X1 E X Statstcs 1: Probablty Theory II 37 3 EPECTATION OF SEVERAL RANDOM VARIABLES As n Probablty Theory I, the nterest n most stuatons les not on the actual dstrbuton of a random vector, but rather on a number

More information

Lecture 14: Bandits with Budget Constraints

Lecture 14: Bandits with Budget Constraints IEOR 8100-001: Learnng and Optmzaton for Sequental Decson Makng 03/07/16 Lecture 14: andts wth udget Constrants Instructor: Shpra Agrawal Scrbed by: Zhpeng Lu 1 Problem defnton In the regular Mult-armed

More information

Maximizing the number of nonnegative subsets

Maximizing the number of nonnegative subsets Maxmzng the number of nonnegatve subsets Noga Alon Hao Huang December 1, 213 Abstract Gven a set of n real numbers, f the sum of elements of every subset of sze larger than k s negatve, what s the maxmum

More information

Kernel Methods and SVMs Extension

Kernel Methods and SVMs Extension Kernel Methods and SVMs Extenson The purpose of ths document s to revew materal covered n Machne Learnng 1 Supervsed Learnng regardng support vector machnes (SVMs). Ths document also provdes a general

More information

Incentive Compatible Transfers in Linear Environments

Incentive Compatible Transfers in Linear Environments Incentve Compatble Transfers n Lnear Envronments Nenad Kos Dept. of Economcs, IGIER Boccon Unversty Matthas Messner Dept. of Economcs, IGIER Boccon Unversty August 5, 2010 Abstract We study the mechansm

More information

Problem Set 9 Solutions

Problem Set 9 Solutions Desgn and Analyss of Algorthms May 4, 2015 Massachusetts Insttute of Technology 6.046J/18.410J Profs. Erk Demane, Srn Devadas, and Nancy Lynch Problem Set 9 Solutons Problem Set 9 Solutons Ths problem

More information

Lecture 3: Probability Distributions

Lecture 3: Probability Distributions Lecture 3: Probablty Dstrbutons Random Varables Let us begn by defnng a sample space as a set of outcomes from an experment. We denote ths by S. A random varable s a functon whch maps outcomes nto the

More information

Online Appendix: Reciprocity with Many Goods

Online Appendix: Reciprocity with Many Goods T D T A : O A Kyle Bagwell Stanford Unversty and NBER Robert W. Stager Dartmouth College and NBER March 2016 Abstract Ths onlne Appendx extends to a many-good settng the man features of recprocty emphaszed

More information

CHAPTER 17 Amortized Analysis

CHAPTER 17 Amortized Analysis CHAPTER 7 Amortzed Analyss In an amortzed analyss, the tme requred to perform a sequence of data structure operatons s averaged over all the operatons performed. It can be used to show that the average

More information

Salmon: Lectures on partial differential equations. Consider the general linear, second-order PDE in the form. ,x 2

Salmon: Lectures on partial differential equations. Consider the general linear, second-order PDE in the form. ,x 2 Salmon: Lectures on partal dfferental equatons 5. Classfcaton of second-order equatons There are general methods for classfyng hgher-order partal dfferental equatons. One s very general (applyng even to

More information

Lecture 4. Instructor: Haipeng Luo

Lecture 4. Instructor: Haipeng Luo Lecture 4 Instructor: Hapeng Luo In the followng lectures, we focus on the expert problem and study more adaptve algorthms. Although Hedge s proven to be worst-case optmal, one may wonder how well t would

More information

Ex post implementation in environments with private goods

Ex post implementation in environments with private goods Theoretcal Economcs 1 (2006), 369 393 1555-7561/20060369 Ex post mplementaton n envronments wth prvate goods SUSHIL BIKHCHANDANI Anderson School of Management, Unversty of Calforna, Los Angeles We prove

More information

University of California, Davis Date: June 22, 2009 Department of Agricultural and Resource Economics. PRELIMINARY EXAMINATION FOR THE Ph.D.

University of California, Davis Date: June 22, 2009 Department of Agricultural and Resource Economics. PRELIMINARY EXAMINATION FOR THE Ph.D. Unversty of Calforna, Davs Date: June 22, 29 Department of Agrcultural and Resource Economcs Department of Economcs Tme: 5 hours Mcroeconomcs Readng Tme: 2 mnutes PRELIMIARY EXAMIATIO FOR THE Ph.D. DEGREE

More information

ECE559VV Project Report

ECE559VV Project Report ECE559VV Project Report (Supplementary Notes Loc Xuan Bu I. MAX SUM-RATE SCHEDULING: THE UPLINK CASE We have seen (n the presentaton that, for downlnk (broadcast channels, the strategy maxmzng the sum-rate

More information

Information Structures in Optimal Auctions

Information Structures in Optimal Auctions Informaton Structures n Optmal Auctons Dr Bergemann y Martn Pesendorfer z January 2007 Abstract A seller wshes to sell an obect to one of multple bdders. The valuatons of the bdders are prvately nown.

More information

Vickrey Auctions with Reserve Pricing

Vickrey Auctions with Reserve Pricing Vckrey Auctons wth Reserve Prcng Lawrence M. Ausubel and Peter Cramton Unversty of Maryland 28 June 1999 Prelmnary and Incomplete Abstract We generalze the Vckrey aucton to allow for reserve prcng n a

More information

Notes on Frequency Estimation in Data Streams

Notes on Frequency Estimation in Data Streams Notes on Frequency Estmaton n Data Streams In (one of) the data streamng model(s), the data s a sequence of arrvals a 1, a 2,..., a m of the form a j = (, v) where s the dentty of the tem and belongs to

More information

princeton univ. F 17 cos 521: Advanced Algorithm Design Lecture 7: LP Duality Lecturer: Matt Weinberg

princeton univ. F 17 cos 521: Advanced Algorithm Design Lecture 7: LP Duality Lecturer: Matt Weinberg prnceton unv. F 17 cos 521: Advanced Algorthm Desgn Lecture 7: LP Dualty Lecturer: Matt Wenberg Scrbe: LP Dualty s an extremely useful tool for analyzng structural propertes of lnear programs. Whle there

More information

Tit-For-Tat Equilibria in Discounted Repeated Games with. Private Monitoring

Tit-For-Tat Equilibria in Discounted Repeated Games with. Private Monitoring 1 Tt-For-Tat Equlbra n Dscounted Repeated Games wth Prvate Montorng Htosh Matsushma 1 Department of Economcs, Unversty of Tokyo 2 Aprl 24, 2007 Abstract We nvestgate nfntely repeated games wth mperfect

More information

ON THE EQUIVALENCE OF ORDINAL BAYESIAN INCENTIVE COMPATIBILITY AND DOMINANT STRATEGY INCENTIVE COMPATIBILITY FOR RANDOM RULES

ON THE EQUIVALENCE OF ORDINAL BAYESIAN INCENTIVE COMPATIBILITY AND DOMINANT STRATEGY INCENTIVE COMPATIBILITY FOR RANDOM RULES ON THE EQUIVALENCE OF ORDINAL BAYESIAN INCENTIVE COMPATIBILITY AND DOMINANT STRATEGY INCENTIVE COMPATIBILITY FOR RANDOM RULES Madhuparna Karmokar 1 and Souvk Roy 1 1 Economc Research Unt, Indan Statstcal

More information

A new construction of 3-separable matrices via an improved decoding of Macula s construction

A new construction of 3-separable matrices via an improved decoding of Macula s construction Dscrete Optmzaton 5 008 700 704 Contents lsts avalable at ScenceDrect Dscrete Optmzaton journal homepage: wwwelsevercom/locate/dsopt A new constructon of 3-separable matrces va an mproved decodng of Macula

More information

Hila Etzion. Min-Seok Pang

Hila Etzion. Min-Seok Pang RESERCH RTICLE COPLEENTRY ONLINE SERVICES IN COPETITIVE RKETS: INTINING PROFITILITY IN THE PRESENCE OF NETWORK EFFECTS Hla Etzon Department of Technology and Operatons, Stephen. Ross School of usness,

More information

Estimation: Part 2. Chapter GREG estimation

Estimation: Part 2. Chapter GREG estimation Chapter 9 Estmaton: Part 2 9. GREG estmaton In Chapter 8, we have seen that the regresson estmator s an effcent estmator when there s a lnear relatonshp between y and x. In ths chapter, we generalzed the

More information

P R. Lecture 4. Theory and Applications of Pattern Recognition. Dept. of Electrical and Computer Engineering /

P R. Lecture 4. Theory and Applications of Pattern Recognition. Dept. of Electrical and Computer Engineering / Theory and Applcatons of Pattern Recognton 003, Rob Polkar, Rowan Unversty, Glassboro, NJ Lecture 4 Bayes Classfcaton Rule Dept. of Electrcal and Computer Engneerng 0909.40.0 / 0909.504.04 Theory & Applcatons

More information

More metrics on cartesian products

More metrics on cartesian products More metrcs on cartesan products If (X, d ) are metrc spaces for 1 n, then n Secton II4 of the lecture notes we defned three metrcs on X whose underlyng topologes are the product topology The purpose of

More information

x i1 =1 for all i (the constant ).

x i1 =1 for all i (the constant ). Chapter 5 The Multple Regresson Model Consder an economc model where the dependent varable s a functon of K explanatory varables. The economc model has the form: y = f ( x,x,..., ) xk Approxmate ths by

More information

4 Analysis of Variance (ANOVA) 5 ANOVA. 5.1 Introduction. 5.2 Fixed Effects ANOVA

4 Analysis of Variance (ANOVA) 5 ANOVA. 5.1 Introduction. 5.2 Fixed Effects ANOVA 4 Analyss of Varance (ANOVA) 5 ANOVA 51 Introducton ANOVA ANOVA s a way to estmate and test the means of multple populatons We wll start wth one-way ANOVA If the populatons ncluded n the study are selected

More information

Mechanisms with Evidence: Commitment and Robustness 1

Mechanisms with Evidence: Commitment and Robustness 1 Mechansms wth Evdence: Commtment and Robustness 1 Elchanan Ben-Porath 2 Edde Dekel 3 Barton L. Lpman 4 Frst Draft January 2017 Current Draft July 2018 1 We thank numerous semnar audences and Joel Sobel

More information

The Order Relation and Trace Inequalities for. Hermitian Operators

The Order Relation and Trace Inequalities for. Hermitian Operators Internatonal Mathematcal Forum, Vol 3, 08, no, 507-57 HIKARI Ltd, wwwm-hkarcom https://doorg/0988/mf088055 The Order Relaton and Trace Inequaltes for Hermtan Operators Y Huang School of Informaton Scence

More information

Edge Isoperimetric Inequalities

Edge Isoperimetric Inequalities November 7, 2005 Ross M. Rchardson Edge Isopermetrc Inequaltes 1 Four Questons Recall that n the last lecture we looked at the problem of sopermetrc nequaltes n the hypercube, Q n. Our noton of boundary

More information

Robustly Optimal Auctions with Unknown Resale Opportunities

Robustly Optimal Auctions with Unknown Resale Opportunities Robustly Optmal Auctons wth Unknown Resale Opportuntes Gabrel Carroll Ilya Segal Department of Economcs, Stanford Unversty, Stanford, CA 94305 August 22, 2016 Abstract We study robust revenue maxmzaton

More information

Lecture Notes on Linear Regression

Lecture Notes on Linear Regression Lecture Notes on Lnear Regresson Feng L fl@sdueducn Shandong Unversty, Chna Lnear Regresson Problem In regresson problem, we am at predct a contnuous target value gven an nput feature vector We assume

More information

Mechanisms with Evidence: Commitment and Robustness 1

Mechanisms with Evidence: Commitment and Robustness 1 Mechansms wth Evdence: Commtment and Robustness 1 Elchanan Ben-Porath 2 Edde Dekel 3 Barton L. Lpman 4 Frst Draft January 2017 1 We thank the Natonal Scence Foundaton, grant SES 0820333 (Dekel), and the

More information

Winter 2008 CS567 Stochastic Linear/Integer Programming Guest Lecturer: Xu, Huan

Winter 2008 CS567 Stochastic Linear/Integer Programming Guest Lecturer: Xu, Huan Wnter 2008 CS567 Stochastc Lnear/Integer Programmng Guest Lecturer: Xu, Huan Class 2: More Modelng Examples 1 Capacty Expanson Capacty expanson models optmal choces of the tmng and levels of nvestments

More information

The Second Anti-Mathima on Game Theory

The Second Anti-Mathima on Game Theory The Second Ant-Mathma on Game Theory Ath. Kehagas December 1 2006 1 Introducton In ths note we wll examne the noton of game equlbrum for three types of games 1. 2-player 2-acton zero-sum games 2. 2-player

More information

Smooth Games, Price of Anarchy and Composability of Auctions - a Quick Tutorial

Smooth Games, Price of Anarchy and Composability of Auctions - a Quick Tutorial Smooth Games, Prce of Anarchy and Composablty of Auctons - a Quck Tutoral Abhshek Snha Laboratory for Informaton and Decson Systems, Massachusetts Insttute of Technology, Cambrdge, MA 02139 Emal: snhaa@mt.edu

More information

Chapter 13: Multiple Regression

Chapter 13: Multiple Regression Chapter 13: Multple Regresson 13.1 Developng the multple-regresson Model The general model can be descrbed as: It smplfes for two ndependent varables: The sample ft parameter b 0, b 1, and b are used to

More information

CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE

CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE Analytcal soluton s usually not possble when exctaton vares arbtrarly wth tme or f the system s nonlnear. Such problems can be solved by numercal tmesteppng

More information

Linear Approximation with Regularization and Moving Least Squares

Linear Approximation with Regularization and Moving Least Squares Lnear Approxmaton wth Regularzaton and Movng Least Squares Igor Grešovn May 007 Revson 4.6 (Revson : March 004). 5 4 3 0.5 3 3.5 4 Contents: Lnear Fttng...4. Weghted Least Squares n Functon Approxmaton...

More information

Game Theory Course: Jackson, Leyton-Brown & Shoham. Vickrey-Clarke-Groves Mechanisms: Definitions

Game Theory Course: Jackson, Leyton-Brown & Shoham. Vickrey-Clarke-Groves Mechanisms: Definitions Vckrey-Clarke-Groves Mechansms: Defntons Game Theory Course: Jackson, Leyton-Brown & Shoham A postve result Recall that n the quaslnear utlty settng, a drect mechansm conssts of a choce rule and a payment

More information

EFFICIENT DYNAMIC AUCTIONS. Dirk Bergemann and Juuso Välimäki. October 2006 COWLES FOUNDATION DISCUSSION PAPER NO. 1584

EFFICIENT DYNAMIC AUCTIONS. Dirk Bergemann and Juuso Välimäki. October 2006 COWLES FOUNDATION DISCUSSION PAPER NO. 1584 EFFICIENT DYNAMIC AUCTIONS By Drk Bergemann and Juuso Välmäk October 2006 COWLES FOUNDATION DISCUSSION PAPER NO. 1584 COWLES FOUNDATION FOR RESEARCH IN ECONOMICS YALE UNIVERSITY Box 208281 New Haven, Connectcut

More information

Informational Size and Efficient Auctions

Informational Size and Efficient Auctions Revew of Economc Studes (2004) 71, 809 827 0034-6527/04/00330809$02.00 c 2004 The Revew of Economc Studes Lmted Informatonal Sze and Effcent Auctons RICHARD MCLEAN Rutgers Unversty and ANDREW POSTLEWAITE

More information

Revenue Maximization in a Spectrum Auction for Dynamic Spectrum Access

Revenue Maximization in a Spectrum Auction for Dynamic Spectrum Access Revenue Maxmzaton n a Spectrum Aucton for Dynamc Spectrum Access Al Kakhbod, Ashutosh Nayyar and Demosthens Teneketzs Department of Electrcal Engneerng and Computer Scence Unversty of Mchgan, Ann Arbor,

More information

CS : Algorithms and Uncertainty Lecture 17 Date: October 26, 2016

CS : Algorithms and Uncertainty Lecture 17 Date: October 26, 2016 CS 29-128: Algorthms and Uncertanty Lecture 17 Date: October 26, 2016 Instructor: Nkhl Bansal Scrbe: Mchael Denns 1 Introducton In ths lecture we wll be lookng nto the secretary problem, and an nterestng

More information

Statistics for Economics & Business

Statistics for Economics & Business Statstcs for Economcs & Busness Smple Lnear Regresson Learnng Objectves In ths chapter, you learn: How to use regresson analyss to predct the value of a dependent varable based on an ndependent varable

More information

Pricing and Resource Allocation Game Theoretic Models

Pricing and Resource Allocation Game Theoretic Models Prcng and Resource Allocaton Game Theoretc Models Zhy Huang Changbn Lu Q Zhang Computer and Informaton Scence December 8, 2009 Z. Huang, C. Lu, and Q. Zhang (CIS) Game Theoretc Models December 8, 2009

More information

College of Computer & Information Science Fall 2009 Northeastern University 20 October 2009

College of Computer & Information Science Fall 2009 Northeastern University 20 October 2009 College of Computer & Informaton Scence Fall 2009 Northeastern Unversty 20 October 2009 CS7880: Algorthmc Power Tools Scrbe: Jan Wen and Laura Poplawsk Lecture Outlne: Prmal-dual schema Network Desgn:

More information

j) = 1 (note sigma notation) ii. Continuous random variable (e.g. Normal distribution) 1. density function: f ( x) 0 and f ( x) dx = 1

j) = 1 (note sigma notation) ii. Continuous random variable (e.g. Normal distribution) 1. density function: f ( x) 0 and f ( x) dx = 1 Random varables Measure of central tendences and varablty (means and varances) Jont densty functons and ndependence Measures of assocaton (covarance and correlaton) Interestng result Condtonal dstrbutons

More information

Statistical Inference. 2.3 Summary Statistics Measures of Center and Spread. parameters ( population characteristics )

Statistical Inference. 2.3 Summary Statistics Measures of Center and Spread. parameters ( population characteristics ) Ismor Fscher, 8//008 Stat 54 / -8.3 Summary Statstcs Measures of Center and Spread Dstrbuton of dscrete contnuous POPULATION Random Varable, numercal True center =??? True spread =???? parameters ( populaton

More information

Optimal Allocation with Costly Verification 1

Optimal Allocation with Costly Verification 1 Optmal Allocaton wth Costly Verfcaton 1 Elchanan Ben-Porath 2 Edde Dekel 3 Barton L. Lpman 4 Frst Draft August 2012 1 We thank Rcky Vohra and numerous semnar audences for helpful comments. We also thank

More information

BOUNDEDNESS OF THE RIESZ TRANSFORM WITH MATRIX A 2 WEIGHTS

BOUNDEDNESS OF THE RIESZ TRANSFORM WITH MATRIX A 2 WEIGHTS BOUNDEDNESS OF THE IESZ TANSFOM WITH MATIX A WEIGHTS Introducton Let L = L ( n, be the functon space wth norm (ˆ f L = f(x C dx d < For a d d matrx valued functon W : wth W (x postve sem-defnte for all

More information

Structure and Drive Paul A. Jensen Copyright July 20, 2003

Structure and Drive Paul A. Jensen Copyright July 20, 2003 Structure and Drve Paul A. Jensen Copyrght July 20, 2003 A system s made up of several operatons wth flow passng between them. The structure of the system descrbes the flow paths from nputs to outputs.

More information

Andreas C. Drichoutis Agriculural University of Athens. Abstract

Andreas C. Drichoutis Agriculural University of Athens. Abstract Heteroskedastcty, the sngle crossng property and ordered response models Andreas C. Drchouts Agrculural Unversty of Athens Panagots Lazards Agrculural Unversty of Athens Rodolfo M. Nayga, Jr. Texas AMUnversty

More information

Global Sensitivity. Tuesday 20 th February, 2018

Global Sensitivity. Tuesday 20 th February, 2018 Global Senstvty Tuesday 2 th February, 28 ) Local Senstvty Most senstvty analyses [] are based on local estmates of senstvty, typcally by expandng the response n a Taylor seres about some specfc values

More information

Optimal auctions with ambiguity

Optimal auctions with ambiguity Theoretcal Economcs 1 (2006), 411 438 1555-7561/20060411 Optmal auctons wth ambguty SUBIR BOSE Department of Economcs, Unversty of Illnos at Urbana-Champagn EMRE OZDENOREN Department of Economcs, Unversty

More information

Lossy Compression. Compromise accuracy of reconstruction for increased compression.

Lossy Compression. Compromise accuracy of reconstruction for increased compression. Lossy Compresson Compromse accuracy of reconstructon for ncreased compresson. The reconstructon s usually vsbly ndstngushable from the orgnal mage. Typcally, one can get up to 0:1 compresson wth almost

More information

ANSWERS. Problem 1. and the moment generating function (mgf) by. defined for any real t. Use this to show that E( U) var( U)

ANSWERS. Problem 1. and the moment generating function (mgf) by. defined for any real t. Use this to show that E( U) var( U) Econ 413 Exam 13 H ANSWERS Settet er nndelt 9 deloppgaver, A,B,C, som alle anbefales å telle lkt for å gøre det ltt lettere å stå. Svar er gtt . Unfortunately, there s a prntng error n the hnt of

More information

Psychology 282 Lecture #24 Outline Regression Diagnostics: Outliers

Psychology 282 Lecture #24 Outline Regression Diagnostics: Outliers Psychology 282 Lecture #24 Outlne Regresson Dagnostcs: Outlers In an earler lecture we studed the statstcal assumptons underlyng the regresson model, ncludng the followng ponts: Formal statement of assumptons.

More information

A Robust Method for Calculating the Correlation Coefficient

A Robust Method for Calculating the Correlation Coefficient A Robust Method for Calculatng the Correlaton Coeffcent E.B. Nven and C. V. Deutsch Relatonshps between prmary and secondary data are frequently quantfed usng the correlaton coeffcent; however, the tradtonal

More information

Commitment and Robustness in Mechanisms with Evidence 1

Commitment and Robustness in Mechanisms with Evidence 1 Commtment and Robustness n Mechansms wth Evdence 1 Elchanan Ben-Porath 2 Edde Dekel 3 Barton L. Lpman 4 Frst Draft June 2016 1 We thank the Natonal Scence Foundaton, grant SES 0820333 (Dekel), and the

More information

Basically, if you have a dummy dependent variable you will be estimating a probability.

Basically, if you have a dummy dependent variable you will be estimating a probability. ECON 497: Lecture Notes 13 Page 1 of 1 Metropoltan State Unversty ECON 497: Research and Forecastng Lecture Notes 13 Dummy Dependent Varable Technques Studenmund Chapter 13 Bascally, f you have a dummy

More information

Foundations of Arithmetic

Foundations of Arithmetic Foundatons of Arthmetc Notaton We shall denote the sum and product of numbers n the usual notaton as a 2 + a 2 + a 3 + + a = a, a 1 a 2 a 3 a = a The notaton a b means a dvdes b,.e. ac = b where c s an

More information

NP-Completeness : Proofs

NP-Completeness : Proofs NP-Completeness : Proofs Proof Methods A method to show a decson problem Π NP-complete s as follows. (1) Show Π NP. (2) Choose an NP-complete problem Π. (3) Show Π Π. A method to show an optmzaton problem

More information

Exploring Bundling Sale Strategy in Online Service Markets with Network Effects

Exploring Bundling Sale Strategy in Online Service Markets with Network Effects Explorng Bundlng Sale Strategy n Onlne Servce Markets wth Network Effects Weje Wu Rchard T. B. Ma and John C. S. Lu Shangha Jao Tong Unversty Natonal Unversty of Sngapore The Chnese Unversty of Hong Kong

More information

find (x): given element x, return the canonical element of the set containing x;

find (x): given element x, return the canonical element of the set containing x; COS 43 Sprng, 009 Dsjont Set Unon Problem: Mantan a collecton of dsjont sets. Two operatons: fnd the set contanng a gven element; unte two sets nto one (destructvely). Approach: Canoncal element method:

More information

2.3 Nilpotent endomorphisms

2.3 Nilpotent endomorphisms s a block dagonal matrx, wth A Mat dm U (C) In fact, we can assume that B = B 1 B k, wth B an ordered bass of U, and that A = [f U ] B, where f U : U U s the restrcton of f to U 40 23 Nlpotent endomorphsms

More information

Temperature. Chapter Heat Engine

Temperature. Chapter Heat Engine Chapter 3 Temperature In prevous chapters of these notes we ntroduced the Prncple of Maxmum ntropy as a technque for estmatng probablty dstrbutons consstent wth constrants. In Chapter 9 we dscussed the

More information

Optimal Allocation with Costly Verification 1

Optimal Allocation with Costly Verification 1 Optmal Allocaton wth Costly Verfcaton 1 Elchanan Ben-Porath 2 Edde Dekel 3 Barton L. Lpman 4 Prelmnary Draft January 2012 1 We thank Rcky Vohra and numerous semnar audences for helpful comments. We also

More information

See Book Chapter 11 2 nd Edition (Chapter 10 1 st Edition)

See Book Chapter 11 2 nd Edition (Chapter 10 1 st Edition) Count Data Models See Book Chapter 11 2 nd Edton (Chapter 10 1 st Edton) Count data consst of non-negatve nteger values Examples: number of drver route changes per week, the number of trp departure changes

More information