Market Competition, Institutions, and Contracting. Draft: Not for Citation or Circulation

Transcription

1 Market Competton, Insttutons, and Contractng Prelmnary Draft Prepared for Presentaton, Department of Agrcultural and Appled Economcs, Unversty of Wsconsn September 18, 009 Draft: Not for Ctaton or Crculaton James MacDonald Unted States Department of Agrculture-Economc Researc Servce Room N M Street NW Wasngton, DC macdonal@ers.usda.gov Steven Y. Wu Department of Agrcultural Economcs Purdue Unversty 403 West State Street, Krannert Bldg. West Lafayette, IN sywu@purdue.edu September 009 Copyrgt 009 by Steven Y. Wu. All rgts reserved. Readers may make verbatm copes of ts document for non-commercal purposes by any means, provded tat ts copyrgt notce appears of all suc copes.

2 Ts paper nvestgates te mpact of market competton on contractng outcomes, wt a partcular focus on effcency and dstrbuton. Wle textbook mcroeconomc models clearly predct welfare losses wen frms ave market power, t s not clear weter suc effcency losses apply to stuatons nvolvng contractng. Contractual relatonsps are typcally desgned to solve agency problems wereby prncpals contract wt agents wo may ave obectves tat dffer from te prncpal s obectves. Of paramount mportance s te ablty of te prncpal to desgn effectve ncentves tat mtgate msalgnment of obectves. Standard teory offers few predctons on weter market concentraton decreases or enances a prncpal s ablty resolve agency problems, altoug ntutvely, t mgt seem plausble tat wen te prncpal as more power, te prncpal as more lattude to structure contracts tat mtgate agency ssues. Indeed, f market power facltates te ablty of prncpals to manage agency conflcts, ten te effcency losses due to decreased quantty under tradtonal market power models mgt be partally offset makng t more dffcult to dentfy aggregate welfare losses n nput markets caracterzed by bot monopsony and contractng. We frst develop a teory under two dfferent nsttutonal envronments tat vary by te degree to wc contractual oblgatons can be enforced by a trd party. We subsequent test our teory usng expermental economcs. We abstract from quantty concerns, wc s te focus on tradtonal market power models, because focusng exclusvely on te agency ssue allows us to cleanly dentfy te mpact of market competton on neffcences caused by agency concerns. An obvous next step s to combne two models, wc we reserve for a future paper. Te motvaton for ts work comes from recent concerns expressed by some polcy makers and farm advocates tat ncreasng consoldaton of large agrbusnesses as eroded te barganng power and proftablty of many farmers tat are under contract wt processors. Among te ssues tat ave receved attenton nclude alleged opportunstc beavor by processors and te dffculty of verfyng weter te oblgatons of contractng partes ave been met. Some examples of allegatons made by farm advocates are tat contractors often create complex and dffcult to understand compensaton formulas n contracts, provde naccurate cas flow proectons to growers to nduce contract acceptance, and offer contracts tat provde te contractor wt consderable dscretonary flexblty (Moeller 003). As a 1

3 consequence, some growers ave complaned tat ter ncomes fall below expectatons based on nformaton provded by contractors (Scrader and Wlson 001). We develop a teoretcal model under te assumpton tat people ave te opton of engagng n one-sot contractng or formng relatonal agreements troug repeat tradng. We beleve tese assumptons are farly general and cover a wde array of contractng stuatons n practce. We assume people structure contracts to enforce promses made by all contractng partes. Tese oblgatons can be eter trd-party enforceable or can be self-enforcng va a relatonal contract based on a reputaton based, perfect Bayesan Nas equlbrum tat mposes sequental ratonalty on te partes. Assumng tat te partes are suffcently patent and are reasonably optmstc about te exstence of cooperatve types n te populaton, even purely self-nterested partes wll onor te contract on te equlbrum pat. We derve te optmal contracts under two enforcement envronments. We sow tat wen contract enforcement s perfect, ten te optmal contract tat maxmzes ont surplus s a complete contract tat enforces all relevant oblgatons. However, wen enforcement s mperfect, ncomplete contracts are unavodable. Noneteless, we fnd tat tere s an optmal degree of ncompleteness tat leaves only enoug gaps n te contract to balance dscretonary flexblty across partes. However, te prncpal always prefers a contract tat provdes er wt maxmum dscretonary flexblty. In oter words, se prefers a contract tat s more ncomplete (n a drecton tat s favorable to te prncpal) tan wat s socally optmal. Intutvely, se can use er flexblty to eter to reward or puns te agent, or more negatvely, to extract rents from te agent. Wen concentraton n favor of te prncpal s added, te prncpal as more leverage to mplement contracts tat exceed te optmal degree of ncompleteness. Our model may explan wy many agrcultural contracts are slent on mportant oblgatons, suc as wen tecnologcal upgrades are to be made or wen pay rates can be adusted, leavng tese oblgatons largely to te dscreton of te processor. We test our predctons usng expermental economcs. Te expermental desgn nvolves tradng between subects tat are randomly assgned to be buyers (processors) and sellers (growers) of an abstract good. Buyers and sellers trade across many dentcal tradng perods were n eac tradng perod, buyers can endogenously offer a range of possble contracts n order to trade a unt of good tat vares n qualty. A typcal contract ncludes oblgatons suc

4 as payment terms, wc ncludes a fxed prce P and possbly ncentve bonuses or deducts to be made by a buyer, and qualty, Q, to be produced by te seller. Tus, a contract s a prcebonus/deduct-qualty trplcate, (P, B/D,Q) of mutual oblgatons wc a seller can accept or reect. We allow buyers to endogenously coose weter to nclude some or all of te oblgatons n a contract and to determne weter oblgatons are bndng. A bndng oblgaton (enforced by te computer) s analogous to a legally enforceable oblgaton, wc means te party responsble for te oblgaton as no lattude to devate from te oblgaton ex post. If an oblgaton s ncluded but not bndng, ten tere s lattude to devate. We nclude four treatments. In treatment RS1, Q can be made bndng and tere s a balanced number of buyers and sellers. Ts mmcs an envronment were contracts can be perfectly enforced and tere s no concentraton. Treatment RS1B s te same as RS1 except tere are fewer buyers and sellers so tat tere s concentraton n favor of buyers. Buyers ave more barganng power as some sellers wll be unemployed. Next, treatment RS s te same as RS1 except Q can no longer be made bndng. Tat s, Q cannot be enforced by a trd-party so contract enforcement s mperfect. Fnally, treatment RSB s te same as RS except buyers ave more barganng power. In nterpretng our results, we dstngus between ex ante effcency and ex post effcency. Ex ante effcency refers to te ablty of contractng partes to take advantage of tradng opportuntes tat yeld postve gans from trade. Wen barganng breaks down or wen contracts are frequently reected, ten ex ante effcency decreases as fewer transactons are executed resultng n mssed opportuntes to create surplus. Ex post effcency refers to te ablty of te prncpal to motvate te agent to undertaken surplus enancng actvtes condtonal on contract acceptance. For example, f te agent exerts effort tat enances qualty, ts ncreases te sze of te pe of a gven contractng relatonsp. Our key teoretcal results and prelmnary expermental results suggest tat, wen mportant performance oblgatons cannot be enforced by a trd-party, contracts tat provde buyers wt dscreton tend to be more ex post effcent tan contracts tat pn down te buyer s oblgatons up-front. Intutvely, buyer-flexble contracts allow te buyer to better matc payments to seller performance, wc enances nformal ncentves. Moreover, even n envronments were self-enforcement breaks down, buyers can use ter dscreton to rent-seek; 3

5 .e. extract rents from sellers. As a consequence, sellers are more reluctant to accept gly dscretonary contracts tat provde buyers wt dscreton. Ts reduces ex ante effcency. However, addng market power ncreases sellers wllngness to accept all classes of contracts, wc ncreases ex ante effcency. Overall effcency appears to ncrease wt concentraton as ex post surplus appears to be largely unaffected wle ex ante acceptance rates ncreases. However, a dark sde of te story s tat tere appears to be substantal ncrease n rent-seekng under concentraton. Our prelmnary expermental results sow tat a sgnfcant fracton of sellers earned profts tat are so low, tey would ave been better off not contractng n te frst place. However, standard economc teory treats market concentraton drven rent-seekng as a dstrbutonal ssue rater tan a socal welfare loss. Noneteless, n practce, suc rent-seekng creates perverse dstrbutve effects wc can create socal, poltcal, and legal costs. Tese costs can create poltcal pressures to regulate wc can ndrectly reduce effcency. Tese results, owever, sould be nterpreted wt cauton as tey are prelmnary. Follow-up experments wll be conducted to determne weter te results reman robust. A Smple Model of Contractng Ts secton develops a smple, two perod contractng model between a contractor (buyer/processor) and contractee (seller/suppler). Suppose tat a buyer and seller can potentally trade one unt of a good wt a qualty ndex q [ q, q], were q s observable but may or may not be trd-party enforceable. Wen a contractng nsttuton exsts to verfy and enforce q, ten complete contractng s possble. However, f te contractng nsttuton s mperfect or mssng, ten neter q nor payments tat are contngent on q can be made legally bndng n a contract. Te lack of trd-party enforceablty s a realstc assumpton n agrcultural contractng as growers often complan about te lack of trd-party verfcaton of performance outcomes. If trade occurs at some prce, p, along wt a bonus, b, ten te payoffs to te buyer and seller are π = R( q) p b and U = p+ b c( q). Te revenue functon, R( q ), obeys Rq ( ) = 0, R ( q) > 0 and R ( q) 0. Te cost of producng a good of qualty q s cq ( ), were cq ( ) = 0, c ( q) > 0 and c ( q) 0. Hence, te buyer and seller s profts from excange are functons of q. If no trade occurs, ten te buyer earns π and te seller earns a reservaton payoff of u. Tnk 4

6 of tese payoffs as expected profts from fndng a dfferent tradng partner. Socal surplus s ten gven by S = R( q) c( q) u π. Assume Sq ( ) > 0 = Sq ( ) and R ( q) c ( q), q [ q, q], so tat trade at q = q s socally effcent. Te tmng of a one-sot tradng (stage) game s as follows. At tme 0, te buyer can make a take-t-or-leave-t offer to te seller. Te contract may specfy a base prce, P, a bonus B, and/or qualty, Q. At tme 1, te seller decdes weter to accept or reect te contract. If te contract s reected, te partes fnd oter tradng partners and earn expected payoffs π and u. If t s accepted, te partes move to tme were te seller cooses actual qualty q, wc may not equal Q, snce qualty s trd-party enforceable (.e. bndng). At tme 3, after q s observed, te buyer cooses actual prce, p wc may dffer from P f prce s not trd-party enforceable. If a bonus B was specfed, ten t s pad f q Q wen B s bndng. Oterwse, te buyer may coose to wtold a bonus or coose some b B. For smplcty and to economze on notaton, we wll ust assume tat a seller onors te contract by delverng q = Q as opposed to q Q. Perfect Contractng Insttutons: Q Trd-Party Enforceable Wen contractng nsttutons exst to verfy Q and make t publc nformaton so tat t can be enforced by a trd-party, ten te optmal contract can be derved from te followng problem: (1) max ( R( Q) P B) QPB,, s.t. P+ B c( Q) u In problem (1), te prncpal cooses contract terms Q, P, and B, n order to maxmze profts subect to te constrant tat te agent s wllng to accept te contract. To solve te problem, note tat a proft maxmzng prncpal would never leave te agent wt rents so tat one can assume tat te partcpaton constrant s bndng and substtute t nto te prncpal s obectve functon to get: () max ( R( Q) u c( Q) ) Q wc gves te frst-order condton: (3) R ( Q) c ( Q) = 0. By assumpton, R ( Q) c ( Q) Q [ q, q] so te buyer s requested Q= q= q. Wt Q n and, t s easy to recover P= p= c( q) + u to nduce te seller s partcpaton. Note tat B=0 as ts 5

7 ncluson would be redundant snce t would play no ncentve role. Ts s because Q can be drectly specfed nto a contract and enforced by a trd-party suc as a court. Assume tat any devaton from Q by a buyer would trgger a court mandated punsment tat s suffcently severe to deter srkng. Note tat snce ts contract nduces acceptance and Q= q= q, t s fully effcent. No oter contract would yeld ger surplus. One can of course, construct an equally effcent complete contract by not specfyng Q drectly, but ncludng a bonus tat satsfed te agent s ncentve compatblty constrant; tat s, B cq ( ) cq ( ). In ts case, te court would enforce B rater tan Q but eter contract leads to te same result. Fnally, te role of repetton from te repeated game matters lttle wt perfect enforcement as repeat tradng s not necessary to provde self-enforcement of oblgatons. Proposton 1: Wen q can be perfectly and costlessly enforced by a trd-party, ten te contract tat drectly specfes Q= q= q and P= p= c( q) + u s optmal. Te prncpal s proft s π = R( q) c( q) u and te agent earns s reservaton proft u. Proposton 1 predcts tat wt perfect contractng nsttutons, we sould observe te effcent outcome and te agent sould earn no rents. Imperfect Contractng Insttutons: Q not Trd-Party Enforceable Wen Q s not trd-party enforceable, ten a full reputaton perfect Bayesan equlbrum can be used to provde self-enforcement of eter Q, or B (te bonus s contngent on qualty). Te reputaton equlbrum assumes tat tere are two types of players: (1) tose tat are cooperatve and terefore onor ter contract agreements, and () strctly self-nterested players wo wll srk on te bonus and prce unless t s n ter self-nterest not to do so. Assume tat te strategy space of te cooperatve type s restrcted so tat tey eter onor ter oblgatons wen te oter party as not srked, and srk f te oter party as srked. Usng te logc of Kreps, et. al. (198) and Healy (007), a repeated game can provde ncentves to selfs buyers to make ter payments n all but te fnal perod of a fntely repeated game. In essence, selfs players mmc cooperatve players n all but te last perod to preserve ter reputaton. To keep tngs smple, we restrct te analyss to a two perod repeated game, toug results generalze easly to te T perod case. Moreover, t s assumed tat te buyer can endogenously coose te contract to ncorporate greater or lesser degrees of 6

8 dscreton by coosng to eter make P bndng by makng t legally enforceable, or by relyng on an nformal andsake agreement about P. Under a andsake agreement, te buyer as te dscreton of gong back on s word and devate on promsed payments. Note tat even f contractng nsttutons suc as effcent and unbased courts do not exst, one can easly mmc perfect enforcement of P troug an upfront payment of P. Note tat t s not possble to make te terms Q and B bndng gven te unenforceablty of Q. Te economcs lterature tends to empasze tree maor types of ncomplete or relatonal contracts tat provde varyng degrees of dscreton to te contractng partes, ncludng (1) a repeat purcase mecansm (RPM) wc guarantees a prce P across all contngences and specfes no bonus (Klen and Leffler 1981); () a dscretonary bonus contract wc guarantees a base prce of P and an nformal, non-enforceable bonus; and (3) an llusory promse (Boot, et. al. 1993), wc s a andsake agreement tat only nformally promses a payment n excange for a qualty level Q. A unque contrbuton of our study s tat we do not mpose te specfc form of contract on our subects and nstead assume tat te buyer can endogenously coose bot te contractual parameters as well as te form of te contract ncludng te tree mentoned above. It sould be noted tat te dscretonary bonus contract, llusory promse and RPM do not exaust te set of all possble contracts but tey do nest maor features of te set of possble contracts. Tus, studyng tese tree types of contracts can generate nsgts tat can be generalzed to oter contracts avalable to te contractng partes. We wll begn by analyzng a dscretonary bonus contract and ten dscuss ow results are affected under te oter two contracts. Te two-perod repeated game s analyzed by backward nducton. Startng n te second and fnal perod, note tat te stage-game of ts repeated game s a sequental move game wt te buyer beng te last mover. Backward nductng wtn te stage game, note tat a selfs buyer wll renege on any promsed bonus, B, even f te seller delvers q = Q, were te subscrpt denotes te perod. Backward nductng to te seller s move, note tat te seller wll be uncertan as to weter se s dealng wt a selfs or cooperatve buyer. We wll assume toug tat te seller olds a pror belef tat a buyer s of a cooperatve type. Denote ts belef as a probablty [ 0,1]. In ts secton, we wll focus on te case were s omogeneous across all contractng partcpants. We do ts as t provdes te most parsmonous analyss of te 7

9 contractng outcomes and ntuton. Ts assumpton wll ave to be relaxed later to generate addtonal results tat are useful for nterpretng te expermental data. Te seller wll onor te contract by delverng q = Q only f er expected payoff from dong so exceeds er payoff from breacng te contract, (4) [ ] [ ] P+ B cq ( ) + (1 ) P cq ( ) P cq ( ) cq ( ) cq ( ) were te subscrpt refers to te second perod, solvng for B yelds B Ts mples tat te seller wll meet contractual oblgatons f te bonus s suffcently large. Backward nductng furter to te buyer s contract offer stage, note tat for a gven belef,, te buyer s optmal bonus offer ust satsfes te ncentve compatblty cq ( ) cq ( ) condton B. Tus, te optmal bonus s determned smply by lettng ts nequalty old wt equalty, wc yelds, cq ( ) cq ( ) (5) B = In order to nduce te seller to accept te contract, te buyer s coce of P must satsfy te seller s partcpaton constrant, (6) [ ( )] (1 )[ ( )] P+ B cq + P cq u wc, combned wt (5), yelds, (7) P = u + c( q) were equalty s assumed because te buyer would not pay more tan necessary to nduce partcpaton. Equatons (5) and (7) embed te agent s second perod ncentve compatblty and partcpaton constrants and can be substtuted drectly nto te buyer s obectve functon π = R( Q) P B to get (8) cq ( ) cq ( ) max R( Q ) u c( q) Q c ( Q ) wc yelds te frst dervatve R ( Q ). Ts may be greater tan, equal to, or less tan zero dependng on te sze of. Clearly, ts condton dffers from te frst best condton of 8

10 R ( Q) c ( Q) 0, Q [ q, q], wc mples Q = q. Tus, we can expect Q < q to preval f sellers are suffcently pessmstc. Ts soluton would also old for any one-sot outcome. One last constrant tat needs to be consdered s te buyer s ncentve to offer a contract tat requests Q > q. Ts s mportant because f s suffcently low, ten te buyer as to pay a large bonus. At some pont, t may erode te buyer s profts f s suffcently low. Hence, te followng constrant must old for cooperatve buyers, R( Q ) P B R( q) p( q) (9) were p( q) = u + c( q), a prce tat s ust g enoug to cover te seller s reservaton utlty f te seller produces mnmal qualty. It s stragtforward to sow tat selfs buyers always contract for greater tan mnmal qualty snce tey wll srk on te bonus. Terefore, sellers belefs do not matter. Substtutng (5) and (7) nto (9) and ten solvng for yelds, cq ( ) cq ( ) (10) = b RQ ( ) Rq ( ) Hence, f sellers belefs are suffcently optmstc so tat tey do not ave to be offered an excessvely large bonus, ten t s wortwle for te buyer to contract for supra-mnmal qualty. Wt repetton and suffcently optmstc belefs, t s possble to sustan ger qualty n earler perods. Backward nductng to perod 1, note tat a buyer s contract desgn problem s, (11) max ( R( Q) P B ) Q1, P1, B1 s.t. 1 1 P + B c( Q ) u () () [ ] [ ] () RQ ( ) P B+ δ [ RQ ( ) P] RQ ( ) P+ δπ P + B c( Q ) + P + B c( Q ) + (1 ) P c( Q ) P c( q) + u (v) b were δ te buyer s dscount factor and δ < 1. If te buyer does not dscount te future, t s stragtforward to sow tat e wll always srk on te bonus. For smplcty, assume tat te seller does not dscount. Constrant (9) s a partcpaton constrant and constrants (9 & 9) are dynamc ncentve constrants or self-enforcement constrants. If tese constrants are satsfed, ten t would be sequentally ratonal for even selfs buyers and sellers to onor ter 9

11 contractual oblgatons. In partcular, even selfs buyers wll mmc cooperatve buyers n te frst perod to preserve ter reputaton. If a selfs buyer srks on te bonus, ten te seller would update er belef n perod to =0 and know wt certanty tat te buyer s selfs and would eter not deal wt te buyer agan n te second perod, or only sgn a contract tat guarantees reservaton profts. Te soluton to (11) can be obtaned by frst substtutng second perod optmal prce and bonus (5) and (7) nto constrants (9) and (9). Ten assumng bot constrants bnd wt equalty, one can obtan te solutons, (1) B1 = cq ( 1) cq ( ) (13) P1 = u + c( q) Furtermore, (7), (1), and (13) can be substtuted nto constrant (9) and ten solvng for δ yelds (assume constrant bnds), cq ( ) cq ( ) (14) δb = RQ ( ) cq ( ) u π Expresson (14) gves us te lower bound on te dscount factor for wc a selfs buyer wll onor ts bonus n te frst perod. Tat s, selfs buyers mmc f δ δ. Fnally, substtutng te optmal values for P 1 and B 1 nto te buyer s obectve functon gves us te problem, (15) max ( R( Q ) u c( Q )) Q1 1 1 Wc yelds te frst order Kun-Tucker condton R ( Q1) c ( Q1) 0. By assumpton, R ( q) c ( q), q [ q, q]. Hence, te optmal soluton s a corner soluton at Q1 = q. Tus, te dscretonary bonus contract s ex post effcent as te buyer requests te gest level of qualty. We now turn our attenton to llusory promses or andsake agreements. Suppose tat te buyer offers an llusory promse wc ncludes no enforceable terms. Te prmary dfference between te llusory promse and te dscretonary bonus presented above s tat te fxed prce P s also made dscretonary so tat te buyer can renege (refuse to pay). If a purely self-nterested buyer reneges, assume tat e makes te most proftable devaton, wc s p = 0. Real world examples of suc contracts nclude cases were payment takes place after a ob s complete or after delvery. Ts exposes te seller to sgnfcant counter-party rsk. 10 b

12 One can proceed wt te analyss of llusory promses usng smlar steps as before toug addtonal scenaros must be consdered gven tat sellers are facng sgnfcantly more counter-party rsk. Startng wt te last mover n te last perod, note tat a selfs buyer wll renege on bot te bonus and P. Backward nductng to te seller s move, we can modfy (4) to determne weter te seller wll delver te promsed qualty. Inequalty (4) s modfed to, (16) [ ] [ ] P+ B cq ( ) + (1 ) cq ( ) cq ( ) Ts nequalty can be used to solve for te ncentve compatble bonus, wc turns out to be dentcal to (5). Te seller wll accept te contract f, (17) [ ( )] (1 )[ ( )] P+ B cq + cq u Note, owever, tat only (17) can be bndng snce by assumpton, u 0 c( q). Solvng for (17) yelds, u+ c( Q ) (18) P + B = Equatons (18) can be substtuted drectly nto te buyer s obectve functon to get (19) u + c( Q ) max RQ ( ) Q c ( Q ) wc yelds te frst order condton R ( Q ) wc may be greater tan, equal to or less tan zero dependng on te sze of. Ts s dentcal to te dscretonary bonus case and mples Q requests Q < q f sellers are suffcently pessmstc. Anoter constrant to be aware of s te ncentve for te buyer to offer a contract tat > q. For cooperatve buyers, ts s gven by (0) RQ ( ) P B max { Rq ( ) P( q), π} Constrant (0) states tat a buyer wll only offer a contract for Q > q f t s more proftable tan te oter two optons of eter offerng mnmal qualty contract or offerng no contract at all to te current seller. Note tat n te dscretonary bonus case, offerng a mnmal qualty contract s always at least as proftable as no contract. But n te llusory promse case, offerng no contract may be more proftable tan a mnmum qualty f belefs are pessmstc ( small). 11

13 Ts s because sellers know tat buyer can renege on even P ( ) q and leave sellers wt a loss of cq ( ) u. As a consequence, sellers wll only accept a mnmal qualty contract f (1) P( q) c( q) + (1 ) c( q) u Ts mples tat u+ c( q) () P ( q) =. Tat s, as seller pessmsm ncreases ( drops), P ( ) q as to ncrease to compensate te seller for counter-party rsk. Wen s small enoug, t s more proftable for te cooperatve buyer, wo wll onor er contract but must also bear te cost of negatve externaltes generated by selfs buyers, to offer no contract. Pluggng n te optmal contract value (18) and () nto (0) yelds, (3) cq ( ) cq ( ) u + c( Q ) max, = I RQ ( ) Rq ( ) RQ ( ) π Hence, belefs must satsfy (3) n order for te buyer to offer a contract tat requests Q > q. It s also mportant to note tat f <, te buyer may not even offer a contract requestng q. It I would only do so f Rq ( ) P ( q) π, wc olds f, u + c( q) (4) = RQ ( ) π Iq. Wt repetton and suffcently optmstc belefs, t s possble to sustan ger qualty n earler perods. Movng back to perod 1, te contract desgn problem s, (5) max ( R( Q ) P B ) Q1, P1, B1 s.t P + B c( Q ) u () () [ ] [ ] P + B c( Q ) + P + B c( Q ) + (1 ) c( Q ) c( q) + u () δ [ ] RQ ( ) P B+ RQ ( ) RQ ( ) + δπ (v) Solvng for (5) gves us, I 1

14 (6) P1+ B1 c( Q1) c( q) And solvng (5) yelds (7) P1+ B1 u + c( Q1) Note tat (6) and (7) cannot bot be bndng. In fact, (7) s tgter and must be bndng. Intutvely, f te seller srks, e never gans because te buyer can wtold total payments so n some sense, te seller s locked n as n a old-up stuaton. Ts gves te seller very strong ncentves to onor te agreement. Hence, te optmal contract s, (8) P1+ B1 = u + c( Q1) Equaton (8) can be substtuted nto (5) and ten solvng for δ yelds (assume constrant bnds), cq ( 1) + u (9) δ I = RQ ( ) π 1 Expresson (9) gves te lower bound on te dscount factor for wc even a selfs buyer wll onor te contract n te frst perod. It s also stragtforward to sow tat, so long as (5v) s not breaced, ten te buyer wll request te gest level of qualty, Q = q. Tus, te llusory promse s ex post effcent altoug t wll be sown tat t s not as ex ante effcent as te dscretonary bonus contract. Fnally, te RPM optmal soluton can be derved by backward nducton. Note tat under te RPM, te buyer s commttng to pay a fxed prce P under all contngences so tat te buyer as no ex post dscreton. However, snce Q s not enforceable, te seller as te dscreton to coose q<q. Tus, after te contract s sgned, te seller s te last mover. Assume tat selfs sellers wll coose q = q regardless of te level of P and Q specfed n te contract wereas cooperatve sellers wll always onor te contract after acceptng te contract. Assume tat te buyer s belefs about a seller beng of a cooperatve type s also snce we ave assumed omogeneous belefs for te entre populaton of traders. Ten n te second perod, te buyer solves te followng contract desgn problem: (30) max ( R( Q) P) (1 ) ( R( q) P) Q, P + s.t. () P c( Q) u 13

15 () ( ) ( ) R( Q ) P + (1 ) R( q) P R( q) P( q) Tat s, te buyer maxmzes er expected profts subect to cooperatve sellers acceptng te contract. Constrant (14) caracterzes te buyer s ncentve to offer a contract tat requests more tan mnmal qualty, were P( q) = u + c( q) and P = u + c( Q) for Q > q. Solvng ts constrant for, one can see tat a buyer would only offer a contract for Q a commtment to pay P ( ) > P q, f te buyer s suffcently optmstc, > q, wc requres (31) cq ( ) cq ( ) = RQ ( ) Rq ( ) RPM. If (31) s breaced, ten te buyer wll only offer a contract tat pays P ( q) = u + c( q) and requests q. However, f (31) s satsfed, ten we can determne te level of Q tat te buyer requests by substtutng te seller s partcpaton constrant (30) nto te obectve functon, (31) max ( RQ ( ) u cq ( ) ) + (1 ) ( Rq ( ) u cq ( ) ) Q wc yelds te frst order condton, c ( Q ) (3) R ( Q ) = 0 Ts s dentcal to dscretonary bonus and llusory promse outcomes. Hence, buyers request te same fnal perod qualty under all tree contracts. Te optmal prce can be determned va te bndng partcpaton constrant to get, (33) P = u + c( Q) Note from (3) tat as pessmsm ncreases, so tat gets smaller, te buyer reduces ts qualty request n order to lower ts payment (33). Note tat te selfs seller wll always earn rents under ts contract snce te selfs seller wll only delver q and stll receve (33). Ts neffcency causes te buyer to edge er rsk of encounterng selfs sellers by reducng te level of qualty requested to below q. Anoter way to tnk about ts s tat selfs buyers always earn rents snce tey are pad a prce tat would satsfy ter partcpaton constrant f tey onored te contract, but wll only produce mnmal qualty. As suc, te cost of payng rents to selfs buyers rases te cost of contractng for g qualty. Movng back to perod 1, te contract desgn problem s, 14

16 (34) max ( R( Q ) P) (1 ) ( R( Q ) P) Q1, P1 + s.t () P1 c( Q1) u () () (v) P c( Q ) + P c( q) P c( q) + u 1 ( ) ( ) RQ ( 1) P1+ δ RQ ( ) P (1 ) Rq ( ) P + Rq ( ) Pq ( ) + δπ RPM were te second constrant s te selfs seller s dynamc ncentve constrant. Wen ts constrant s satsfed, ten te selfs seller wll mmc te cooperatve seller n te frst perod and onor te contract. Te trd constrant s te buyer s ncentves to offer a contract tat requests Q1 > q. Te last constrant s smply (31) wc states tat te buyer as to be suffcently optmstc so tat se wll not offer only a mnmal qualty contract n perod. To solve for te optmal frst perod contract, substtute (33) nto (34) and after cancelng and rearrangng terms, te constrant reduces to cq ( ) cq ( 1), wc mples, (35) Q Q1 Tat s, te expected second perod output (and ence P ) must be suffcently g to nduce te selfs seller to delay er srkng decson untl te second perod. If se srks n te frst perod, ten te buyer wll know wt certanty tat se s a selfs type and would update er belef to = 0 and would not offer er P n te second perod. Consequently, for a suffcently g P, even selfs sellers prefer to onor te contracts n te frst perod. Assumng tat (35) bnds, te optmal frst perod contract s dentcal to te second perod contract and yelds te same qualty request;.e. Q = Q = Q1. An mportant pont to note s tat ts contract s not RPM ex post effcent n eter perod. Te qualty level requested, Q RPM, wll be dentcal to second perod qualty requested under dscretonary bonus contracts and llusory promses but wll fall sort of frst perod requests under tese oter contracts. Constrant (34) can be solved to get te dscount factor cutoff pont for te buyer to offer a contract tat requests Q1 soluton s, > q. Te 15

17 (36) Rq ( ) RQ ( RPM ) + cq ( RPM ) cq ( ) δ = δ R( QRPM ) + (1 ) R( q) c( QRPM ) u π It s mportant to note tat te numerator of (36) s negatve gven te modelng assumptons on te revenue and cost functons. Ts mples tat f te denomnator s postve, ten (36) olds for any postve dscount factor. On te oter and, f te denomnator s also negatve, ten δ RPM s greater tan one, wc mples tat te buyer wll offer a mnmal qualty contract regardless of te dscount factor. Snce δ [ 0,1], one can essentally treat (36) as bnary were δ RPM s effectvely eter equal to 1 or 0. Note tat weter te denomnator s postve or negatve depends on. In fact, one can easly sow tat wen (31) s satsfed, ten te denomnator s non-negatve wereas wen (31) s volated, ten te denomnator s negatve. Hence, we ave, 1 f < RPM (37) δ RPM = 0 f RPM Te followng proposton summarzes te maor results derved above. Proposton : Wen q cannot be trd-party enforced, belefs are omogeneous, and contracts are self-enforcng, ten 1. te dscretonary bonus contract, llusory promse, and RPM all specfy te same second c ( Q ) perod qualty request Q wc satsfy R ( Q ) = 0;. te dscretonary bonus and llusory promse specfes frst perod qualty request Q1 wle te RPM specfes Q1 = Q = QRPM ; RPM = q, 3. te optmal contract payments are: a) Dscretonary bonus - P = u + c( q), B1 = cq ( ) cq ( ); B cq ( ) cq ( ) =, P1 = u + c( q), and u+ c( Q ) b) Illusory promse - P + B = and P1+ B1 = u + c( q) ; c) RPM - P = u + c( Q ) = P1. RPM 16

18 Anoter factor to consder s te ablty of contractng partes to sustan self-enforcng relatonsps under te varous contracts. For buyers n partcular, even selfs buyers are more wllng to onor frst perod promses f ter tme preferences are suffcently strong. However, te mnmal tresold levels for δ to sustan self-enforcement dffers across contracts as eac contract leads to dfferent payoffs to buyers from eter onorng or srkng on te contract. We can make ts comparson by comparng te mnmum cut-off dscount factors derved earler for eac type of contract. Proposton 3: Buyers are more lkely to break promses and engage n rent-seekng under te contract tat provdes te buyer wt more dscreton. Consequently, tere s a smaller set of dscount factors tat wll dscplne selfs buyers under llusory promses. Under llusory promses, selfs buyers are less lkely to mmc cooperatve buyers and are more lkely to engage n rent seekng. Ts s because buyers earn te gest profts from srkng under llusory promses. Hence, tere s also te greatest pressure to break promses under llusory promses. Proposton 3 predcts tat we ougt to observe fewer relatonal agreements under contracts tat provde buyers wt more dscreton. Heterogeneous Belefs So far, te analyss as assumed tat all contractng partes sare te same belef. As suc, t s relatvely stragtforward to derve te contractng outcomes specfed n Proposton. However, n order to generate sarper predctons tat are consstent wt stylzed observatons n te real world, t s mportant to ntroduce addtonal frctons suc as wen contractng partes ave eterogeneous preferences. Assume tat tere are m buyers and n sellers and eac buyer as a belef [0,1] and eac seller as a belef [0,1] were =1,,n and =1,,m. A belef represents an ndvdual s subectve assessment about te probablty tat er counter-party s a cooperatve type wo wll onor contracts. In oter words, t s an ndvdual s assessment about te proporton of cooperatve types n te populaton of traders. An mportant pont to note s tat a buyer may not sare te same belef as te seller wt wom se s contractng wt. In ts case, t s possble tat te buyer desgns a contract under a set of belefs tat may not be consstent 17

19 wt te seller s belefs makng t more dffcult to satsfy te self-enforcement and partcpaton constrants. We begn te analyss by revstng te contract desgn problem for te dscretonary bonus contract. Recall tat n te dscretonary bonus contract desgn problem (11), te buyer maxmzes er own profts wt respect to contract varables subect to four constrants. Frst, consder wat appens to te seller s self-enforcement constrant (11) wt eterogeneous belefs. Te buyer desgns te optmal contract under te assumpton tat te probablty of tat an ndvdual s of a selfs type s. Tus, accordng to (5), (7), (1), and (13) te buyer wll optmally offer a contract P = u + c( q), B cq ( ) cq ( ) =, P1 = u + c( q), and B1 = cq ( 1) cq ( ). However, a selfs seller, wo as belef as a true self-enforcement constrant gven by, (38) [ ] [ ] P + B c( Q ) + P + B c( Q ) + (1 ) P c( Q ) P c( q) + u Substtutng te optmal contract nto (38) and smplfyng yelds, (39) Ub = u + u 1 c( Q ) c( q) u + u wc s satsfed f and only f. In oter words, f a buyer s more optmstc tan a seller, ten te buyer offers a bonus payment tat s too low to ncentvze a selfs seller to onor te contract. Turnng to te partcpaton constrant (11), we would lke to know weter te seller s wllng to accept te contract n te presence of eterogeneous belefs. To answer ts queston, we must consder two scenaros. In scenaro one, buyers ave suffcently strong tme preferences so tat (11) wll be satsfed and buyers wll pay te bonus B 1 n perod 1 f te seller delvers q1 Q1. Ts s te case were buyers care about ter reputaton because dscounted second perod payoffs are suffcently g to dscplne even selfs buyers n te frst perod. If te seller knows tat buyers care about ter reputaton, ten te seller assumes tat te bonus wll be pad f te seller onors te contract. In ts case, te seller s partcpaton constrant s, 18

20 (40) P + B c( q) u f 1 1 P c( q) u f < 1 Te partcpaton constrant depends on belefs because only n te case would te seller s self-enforcement constrant be satsfed n wc case e would delver q1 Q1. Wen, ten te seller wll not delver g qualty because te bonus s too small to ncentvze te seller. As a result, te buyer wll also wtold te bonus snce te buyer observes qualty before decdng on te bonus. It s stragtforward to sow by pluggng te optmal contract nto (40) tat te buyer s partcpaton constrant wll always be satsfed wt equalty. Hence, n reputaton based envronments, sellers are predcted to always accept a dscretonary bonus contract. For te oter scenaro were (11) s breaced ( δ < δ ), selfs buyers do not ave suffcently strong tme preferences to onor ter contracts. In ts case, te frst perod s dentcal to te second perod as contractng becomes essentally one-sot. Snce a seller beleves tat e wll encounter a cooperatve buyer wt probablty, te partcpaton constrant becomes, (41) [ ( )] (1 )[ ( )] P + B c Q + P c Q u t t t t t were te subscrpt t replaces 1 because ts scenaro can apply to eter perod 1 or perod. Note tat te buyer wt belef wll offer b cq ( t ) cq ( ) Bt = and Pt = u + c( q), were Q t satsfes c ( Qt ) R ( Qt ) = 0, (41) reduces to, (4) Ub, δ< δ u 1 c( Q) c( q) u b = Constrant (4) s satsfed f and only f. Oterwse, te seller would reect a dscretonary bonus contract. Next, we can analyze te mpact of eterogeneous belefs on llusory promses. Note u+ c( Q ) from (18) and (8) tat a buyer wt belefs offers an optmal contract P + B = 19

21 and P1+ B1 = u+ c Q1 constrant gven by, ( ). However, a selfs seller wt belefs as a true self-enforcement (43) [ ] [ ] P + B c( Q ) + P + B c( Q ) + (1 ) c( Q ) c( q) + u Substtutng te optmal contract nto (43) and smplfyng yelds, (44) UI = u + [ u + c( Q) ] c( Q) c( q) + u Constrant (44) s satsfed f and only f cq ( ) cq ( ). Note te rgt-and-sde of te u + c( Q ) nequalty s less tan 1 so tat can be smaller tan and stll satsfy te self-enforcement constrant. Ts s because te buyer as te ablty to mplement a severe punsment on te seller for not onorng s contract and terefore te seller as extra motvaton not to volate te constrant. Movng to te partcpaton constrant (5), we now consder bot te scenaro were buyers care about ter reputaton ( δ δ I ) and wen tey don t ( δ < δ I ). In te frst scenaro, even selfs buyers make promsed payments P1+ B1 so long as te seller delvers promsed qualty. In ts case, te seller s partcpaton constrant s, cq ( ) cq ( ) P1+ B1 c( q) u f (45) u + c( Q ) cq ( ) < u oterwse We can see mmedately tat f cq ( ) cq ( ) u + c( Q ) s not satsfed, ten te seller would reect te contract snce e antcpates earnng cq ( ), wc s non-postve and below te reservaton payoff. For te case were cq ( ) cq ( ) u + c( Q ), P1+ B1 = u+ c Q1 ( ) can be substtuted nto te partcpaton constrant wc smplfes to u u. In ts case, te seller would accept te contract. 0

22 For te scenaro were buyers ave weak tme preferences ( δ < δ ) so tat reputaton s nsuffcent to nduce selfs buyers to onor ter contracts, te partcpaton constrant n bot perods becomes, (46) [ ( )] (1 )[ ( )] P + B c Q + c Q u t t t t Substtutng P + B = u + c( Q ) nto (46) and smplfyng yelds, t t t UI, δ< δ = u+ c( Q ) ( ) b t c Qt u (47) [ ] I Note tat (47) s satsfed f and only f. Oterwse, te seller reects te contract. Next, we focus on te RPM contract. It sould be noted ere tat te last mover wtn te stage-game s te seller. Ts dffers from te dscretonary bonus and llusory promse case were te buyer s te contract desgner and te last mover. Terefore, te buyer s belef matters at te ex ante stage wle te contract s beng desgned. Tat s, te buyer s belef represents te buyers subectve probablty about te proporton of cooperatve types n te populaton and ten desgns te contract accordngly. But at te ex post stage; tat s, after contract acceptance, t s te seller wo as to antcpate te ex post move of te buyer, wo s te last mover. Hence, te seller s belefs also matter and t s te degree of congruence between buyer and seller belefs tat drve outcomes under te dscretonary bonus and llusory promse contracts. However, n te case of te RPM, te seller, beng te last mover, as notng to antcpate. Moreover, te seller does not desgn te contract so s belefs don t play any role watsoever. Instead, te buyer s belef matters bot durng te contract desgn stage and ex post wen te buyer as to antcpate te seller s move. Because only te buyer s belefs matter, tere s no congruency problem between te buyer and seller s belefs. As a consequence, te RPM results from te omogeneous belefs case carry over to te eterogeneous belefs scenaro. Te followng proposton summarzes te key results dscussed so far n ts secton. Proposton 4: Wen q cannot be trd-party enforced and belefs are eterogeneous, ten 1. te dscretonary bonus contract and llusory promse specfy frst perod qualty request Q1 = qbut te seller wll only delver ts qualty under te dscretonary bonus contract f and under te llusory promse f cq ( ) cq ( ) ; u + c( Q ) 1

23 . f δ δb, ten a dscretonary bonus contract wll satsfy sellers partcpaton constrants, and f δ δ and I cq ( ) cq ( ) u + c( Q ), ten an llusory promse wll satsfy sellers partcpaton constrants; 3. f δ < δb, δ < δ I, and, ten bot dscretonary bonus contracts and llusory promses wll satsfy sellers partcpaton constrants. One can also used te results from ts secton to generate te one-perod payoff functons of buyers and sellers evaluated at te optmal contracts. Tese value functons wll be useful for assessng ex ante effcency as we can assess te lkelood tat sellers are wllng to engage n trade under varous contracts. Moreover, we can assess te types of contracts tat buyers prefer to offer to sellers. Proposton 5: Wen belefs are eterogeneous, te frst perod payoffs under te optmal contracts are: 1. Under dscretonary bonus contracts, a) Sellers: Ub = u f δ δb Ub = u 1 c( Q ) c( q) f δ < δb. b) Buyers: Vb = R( q) c( q) u f δ δb and [ ] Vb = R( q) c( q) + (1 ) R( q) c( q) u f δ δb and < R( Q1) c( Q1) u cooperatve c ( Q1 ) Vb = f δ < δb and were R ( Q1 ) = 0. R( Q1 ) c( q) u selfs. Under llusory promses, a) Sellers: cq ( ) cq ( ) UI = u f δ δ I and u + c( Q )

24 u cooperatve U I = f δ δ I and c( q) selfs UI = [ u + c( Qt) ] c( Qt) f δ < δ I cq ( ) cq ( ) < u + c( Q ) b) Buyers: VI = R( q) c( q) u f δ δ I V I ( 1) ( 1) = R Q c Q u cooperatve R( Q ) 1 selfs f δ < δ I and 3. Under RPM, a) Sellers: URPM = u b) Buyers: VRPM = R( QRPM ) c( QRPM ) u f δ δ RPM VRPM = R( q) c( q) u f δ < δ RPM Te payoff functons derved n Proposton 5 lead to te followng corollary. Corollary 1: Te RPM s ex ante effcent as sellers ave no ncentves to reect any RPM contract tat results n gans from trade. Te llusory promse s te least ex ante effcent as cq ( ) cq ( ) sellers accept contracts only wen δ δ I and, δ δ I and te buyer s u + c( Q ) cooperatve, or f δ < δ I and. Te dscretonary bonus contract s moderately ex ante effcent as sellers always accept contracts wen δ δ. Sellers only reect contracts wen δ < δ b and < Corollary 1 s easy to sow usng te payoffs from Proposton 5. Essentally, te seller wll reect any contract tat doesn t offer an expected payoff tat s at least as g as reservaton payoff. Corollary 1 summarzes te condtons under wc sellers are wllng to accept a contract under eac of te contracts. Gven tat expected payoffs under te RPM are always as g as reservaton payoffs, t maxmzes te number of surplus creatng tradng opportuntes. Te dscretonary bonus contract also provdes sellers wt expected payoffs of at least 3 b

25 reservaton payoffs except n te specal case were buyers do not care about ter reputaton and sellers are less optmstc tan buyers. Te llusory promse only results n expected profts tat meet reservaton payoffs n very specfc crcumstances. Te next corollary focus on te types of contracts tat buyers potental prefer. Tat s, buyers ougt to offer contracts tat leads to te gest expected payoffs. Corollary : Te llusory promse provdes te buyer wt a frst best level of expected profts f δ δ I. Te dscretonary bonus contracts results n a frst best level of expected profts only f δ δ b and. Te RPM never aceves a frst best level of expected profts for te buyer. An mplcaton of Corollary s tat buyers prefer llusory promses as tey are mostly lkely to provde buyers wt a level of expected profts tat matces te frst best level of profts aceved under te complete contract. Frst best level of expected profts s also possble under dscreton bonus contracts but te condtons are more restrctve as, n addton to requrng strong tme preferences, tey also requres seller belefs to be at least as optmstc as buyers. Te RPM s predcted to be least attractve to buyers as expected profts never reac frst best levels. So far, we ave establsed tat te llusory promse and dscretonary bonus contracts are ex post effcent as tey request qualty levels tat are at least as g as te RPM and can even request te frst best level, q. However, te RPM s te most ex ante effcent as t s lkely to maxmze te number of trades tat are executed. Tus, tere s a tenson tat exsts as buyers prefer to offer an llusory promse, but ts also may lead to te gest reecton rates and mssed tradng opportuntes. Te next proposton suggests tat reecton rates are lkely to be moderated by market power. Tat s, f tere s an excess supply of sellers so tat some sellers wll be unemployed, ten sellers reservaton payoffs wll decrease wc wll ncrease ter rate contract acceptance. As a consequence, ex ante effcency mgt ncrease as tere wll be fewer reectons and tus fewer mssed tradng opportuntes. Proposton 6: Wen buyer market power s ncreased, sellers wllngness to accept all contracts ncreases. m To sow ts, suppose tat tere are m buyers and n sellers were < 1 so tat tere s n concentraton n favor of buyers. Tat s, tere wll be sellers left wtout contracts. If te seller 4

26 reects a contract, ten tere s a (1 m/ n) probablty tat se wll not obtan anoter contract. Hence, er reservaton utlty must be modfed so tat uˆ = m u + (1 m/ n) u were u represents n te scrap value of unemployed producton assets or unemployment benefts, dependng on te context. Note tat u > uˆ so tat concentraton effectvely decreases te reservaton utlty makng t more lkely tat a seller wll accept a gven contract. It s mportant to empasze tat ts s a reduced form approac as prncpal-agent relatonsps are essentally partal equlbrum models were te outsde opton captures te barganng power. A partcularly nterestng aspect of Proposton 6 s tat concentraton can potentally ncrease ex ante effcency by ncreasng sellers wllngness to engage n trade. However, tere s a dark sde to ts story, wc s tat, gven sellers greater wllngness to accept contracts, buyers may offer more llusory promses so tat te net effcency effect of concentraton s unclear. Illusory promses can result n rentseekng, were wealt s transferred from sellers to buyers wtout creatng any gans n effcency. Accordng to Proposton 5, tere are some states of te world were te seller can earn negatve profts under llusory promses. In partcular, selfs buyers can default on payments leavng te seller wt negatve profts after accountng for te costs of producton. Wle we only provde an analytc model for tree types of contracts, our expermental desgn allows subects to endogenze a large set of contracts. For example, some contracts are allowed to ncorporate a deduct, D, rater tan a bonus. Oter contracts may omt Q altogeter so tat subects do not explctly communcate a qualty request. Explctly modelng tese possbltes s conceptually stragtforward, toug tedous, but would not add addtonal nsgts and terefore are omtted. Te man pont of te teoretcal model would contnue to old under tese addtonal contracts. Expermental Desgn Te basc expermental platform wll allow partes to coose te contractual form endogenously subect to exogenously mposed lmts to contract enforcement. Bot relatonal contracts and one-sot contracts are endogenously nested n te desgn. Te advantage of ts platform s tat t s consstent wt our teory and terefore allows us to test some mplcatons of te teory. We conduct four treatments. We conducted one treatment wtout market concentraton were 5

27 tere s a balanced number of buyers and sellers and one wt buyer market concentraton were tere are fewer buyers tan sellers. Eac market condton was splt nto two addtonal treatments one tat allowed for perfect enforcement of contracts and anoter were qualty cannot be enforced. Table 1 outlnes te set of treatments. In eac expermental sesson, subects were randomly assgned to be buyers (prncpals wo proposed and offered contracts) and sellers (agents wo can accept or reect contracts). Te laboratory ad room for twelve subects per-sesson and a sesson was conducted f a mnmum of sx subects sowed up. A typcal sesson s comprsed of 15 dentcal tradng perods to enable repeated game effects. Tus, te total number of possble trades per-sesson s 15 multpled by ½ te number of subects n te non-market power treatments (RS1 and RS). For example, f 1 subects partcpated, tere were sx buyer and sx sellers so tere were sx possble trades per round and 90 possble trades for te entre sesson. Snce not all sessons ncluded twelve subects, a total of 190 trades were possble across te 18 non-market power sessons. In te market power sessons (RS1B and RSB), tere were typcally fewer buyers tan sellers. Wen tere were an even number of subects, tere was typcally two more sellers tan buyers. For odd numbers of subects, tere was one more seller tan buyer. Tere were a total of 7 subects wo partcpated n te market power sessons resultng n a total of 1080 possble trades. Wtn eac of te 15 tradng perods n eac sesson, any buyer can trade wt any seller altoug eac buyer and seller was only allowed one trade per-round. Some sellers and buyers dd not trade eter because tey dd not reac contractual agreements or tey decded not to trade for tat perod. In eac perod, buyers and sellers trade one unt of a good wc vares n qualty. Hger qualty ncreases buyers revenues. But producng ger qualty also ncreases sellers costs. A buyer tus tres to nduce g qualty va a contract (a mutual promse, wc may be enforceable) wc specfes prce and qualty. Te key nsttutonal factor tat we vared s te degree to wc qualty s verfable and enforceable by a trd-party. If qualty s enforceable, ten te partes can wrte a contract tat stpulates a promsed level of qualty and to make t bndng. A bndng qualty offer s analogous to a perfectly legally enforceable term ts term was enforced by te computer. In oter words, f te partes agreed on terms and made tem bndng, tey ad no dscretonary lattude to devate from terms after agreeng to a 6