Concession Contract Renegotiations: Some Efficiency vs. Equity Dilemmas

Transcription

1 Concesson Contract enegotatons: Some Effcency vs. Euty Dlemmas Antono Estace Lucía Quesaa Setember 00 Abstract e analyze te ossblty of trae-offs between effcency an euty an of strbutonal conflcts n te context of nfrastructure contract renegotaton n evelong countres. To o so we resent a moel of contract awar troug an aucton of te rgt to oerate an nfrastructure servce to a rvate monooly an conser te ossblty of renegotaton. To entfy te otental sources of trae -offs we track own te ossble outcomes of varous renegotaton strateges for te monooly runnng te concesson an for te two grous of consumers rc an oor wo alternate n ower accorng to a majorty-votng rule. One of te most mortant olcy mlcatons of te moel s tat f te fear of avng frms-rven renegotatons s a major concern effcency soul not te only varable tat matters wen selectng an oerator. Inee consumers may want to awar te concesson to a less effcent frm n orer to reuce te robablty of renegotaton snce lower robabltes of frm rven renegotatons (ues to eman socks for nstance are assocate wt ger welfare levels for all users. Ts aer was starte wle te secon autor was vstng at Te orl Bank urng te summer of 000 an s an nut n a seres of courses organze for evelong countres olcymakers. e are enebte to Lus Guasc Jean-Jacues Laffont an Matas Dewatront for useful scussons comments an suggestons Te orl Bank Insttute. E-mal: aestace@worlbank.org GEMAQ Unversty of Toulouse. Manufacture es Tabacs Allée e Brenne 3000 Toulouse France an Centro e Estuos Económcos e la egulacón UADE Buenos Ares Argentna. E-mal: Luca.Quesaa@unv-tlse.fr.

2 . Introucton Durng te 990s nvestment n utlty rojects assocate wt some form of rvatzaton ae u to about US$70 bllon n evelong countres. Ts fgure reflects nvestment commtments mae troug almost 500 contracts between governments an rvate oerators n te context of major nfrastructure restructurng rograms. Te contracts most of tem concesson contracts were generally aware troug auctons to make te most of cometton for te market ex ante an mnmze te nee for scretonary regulatory ecsons ex ost (Crames an Estace (998. An ongong analyss of a samle of 000 utltes an transort contracts sgne urng te 990s (Guasc (00 suggests tat renegotaton aens n aroun 50% of te cases. Moreover t suggests tat te os of renegotatons are te gest wen te aucton crtera s rven by te esre to mnmze te average tarff to be a by users of te servces be out. More generally te ffculty s often relate to a oor ntal effort to assess te sources of eman fluctuatons often relate to a lmte ablty to ay of a goo sare of te oulaton. Tese stylze facts efne te olcy ssues ts aer s concerne wt. If awarng te contract to te ber romsng to cut te most te average tarff leas to renegotatons t seems far to be concerne wt te ossblty tat renegotaton wll en u ncreasng tarffs an enalzng te oorest. Te man urose of ts aer s to ceck for some of te contons uner wc tere s an effcency-euty trae-off n te context of contract renegotaton an to entfy te wnners an te losers of varous olcy otons n a cangng oltcal envronment. Secton glgts te man results of te lterature on contract ncomleteness an renegotaton an sows tat te euty menson as essentally been left out so far from te scusson. Secton 3 resents a moel of contract awar troug an aucton. In vew of te exerence of te last ecae te ossblty of renegotaton s exlctly recognze as art of te ntal contract rearaton. enegotaton can be rven eter by te frm n wc case t wll aen f te frm's rofts are negatve or by te government n wc case t wll occur f a new grou s n ower at te renegotaton stage. To also take nto account te ossble nfluence of strbutonal ssues n te context of renegotaton we assume tat te economy s comose by two grous of consumers rc an oor wo alternate n ower accorng to a majorty-votng rule. In Secton we solve te moel an we obtan te

3 caracterstcs of te wnnng frm an tose of te renegotate rce. Secton 5 analyzes te results n terms of effcency an euty trae-offs. Secton 6 conclues.. enegotaton an contract ncomleteness enegotaton an contract ncomleteness are ute relate n te lterature. Te sngle ossblty of renegotaton mles some ncomleteness because te artes are not able to commt not to renegotate. But even ts kn of ncomleteness s not enoug to exlan wy renegotaton occurs n te real worl. Inee te renegotaton-roof rncle states tat any outcome tat a rncal can get can be aceve troug a contract tat leaves no room for renegotaton. Unfortunately tere s no unue framework uner wc contract ncomleteness can be analyze. Trole (999 summarzes te man reasons tat coul generate ncomlete contracts. He mentons bascally tree reasons. Frst some contngences tat may arse urng te executon of te contract cannot be foreseen wen te contract s sgne. Secon even f te artes coul antcate all te ossble contngences tat soul be nclue n te contract tey mgt be so numerous tat te costs of escrbng tem all can be robtve. In tat case te artes ave to trae off te beneft of avng a more comreensve contract wt te cost of wrtng new clauses. Fnally te contract can only be contngent on varables tat coul be verfe by a tr arty usually an arbtraton commsson or a egulatory Agency n te context of nfrastructure rvatzaton contracts. If ts were not te case te contract woul not be enforceable. A semnal aer on renegotaton an ncomlete nformaton s Hart an Moore (988. In ter moel tey assume tat actons an future contngences are all ex ost observable but tey are not verfable by a tr arty. Tey sow tat n tat case te artes cannot aceve an ex ante otmal level of nvestment an renegotaton s use to aceve ex ost effcency. On te oter an Agon et al. (99 argues tat te assumton of unverfablty s not enoug to exlan uner-nvestment. Very often ts roblem can be overcome wt an arorate ex ante esgn of te renegotaton rocess tself. In ter moel layng wt te efault oton an te barganng ower at te renegotaton stage restores effcency even wen actons are not verfable. Closer to te moel resente ere are te aers of Jeon an Laffont (999 an Kartaceva an Quesaa (000. Jeon an Laffont (999 moele a government regulatng a frm uner some uncertanty contons on te eman face by te frm 3

4 an ts fxe cost. Tey assume tat bot te government an te frm are myoc an are not able to antcate future renegotatons. Ex ost f te roft of te frm s negatve te regulatory contract s renegotate. In ts context renegotaton occurs more often wen te frm s neffcent wc s not always te case n our moel. Fnally Kartaceva an Quesaa (000 assume a government auctonng te concesson of some ublc utlty. Frms antcate tat renegotaton wll occur f eman s low an take ts fact nto account at te aucton stage. Frms know tat f rofts are negatve te government wll refer to renegotate rater tan to sto roucton an tey wll face no cometton at tat ont. As a conseuence te announce b wll be base ownwar gven tat a frm exects renegotaton f t makes losses. Furtermore fferent frms can ave fferent egrees of barganng ower at te renegotaton stage. Bot effects gve ncentves for frms to announce a b lower tan te one tat woul ave been announce wtout te ossblty of renegotaton. Ts mles also tat te ex ost robablty of renegotaton s ger. Ts aer ffers from te revous ones n tat we allow ere to ave renegotaton bot rven by te frm an by te government. Of course te outcome of renegotaton wll be fferent n one case or n te oter. Moreover we allow n our moel to ave two fferent tyes of consumers an terefore we are able to analyze ow renegotaton affects fferently eac of tese grous. 3. Te moel Te government wants to concesson a ublc utlty. To coose te frm to oerate te concesson t runs an aucton an cooses te frm tat bs te lowest rce. Te cosen frm wll rouce for two eros. e are not lookng for te otmal aucton from te government's ersectve an conser te mecansm as gven. Inee we wll assume tat bot te government an te frm are naïve n game-teoretcal sense meanng tat future stages are not taken nto account wen a ecson or acton s unertaken. In ts artcular case t means tat te government oes not conser or oes not care about te ossblty of renegotaton wen esgnng te aucton an tat neter o te frms wen ecng on te b. Ts lmte ratonalty on te agents beavor s wat allows renegotaton to aen n eulbrum. Inee wtout suc naïve agents renegotaton woul not be an ssue. Ts beavor may be exlane n several ways. It may be te case tat te agents overestmate ter commtment ower. More realstcally from te vew ont of te government wat may matter more s tat te oltcal cycles are muc sorter tan te tycal

5 concesson contract uraton an more often tan not te oltcal turnover s suc tat te rvatzng team of a government s no longer aroun wen renegotaton takes lace an ence as lttle ncentve to be too concerne about t. From te vewont of te oerator beng naïve may smly reflect overconfence n ter ablty to get anytng tey want from a weak government or regulator an s ence erceve as a no rsk atttue towars te market. On te suly se eac frm s caracterze by two arameters. Te frst arameter s ts margnal cost of roucton [ ] an s frm s rvate nformaton. Te secon arameter s te barganng ower te frm woul ave wenever renegotaton occurs [ 0] an s ex ante unknown to anyone but woul become common knowlege at te renegotaton stage. To smlfy we assume tat all are neenent an entcally strbute wt a cumulatve strbuton functon F( an robablty ensty functon f(. Te roft of frm s ( ( I were s te rce of te ublc utlty ( s te eman functon an I s a fxe cost. e conser an ncreasng return to scale roucton tecnology wt a constant margnal cost. Te assumton of a constant margnal cost s mae just to smlfy comutatons an all our results wll stll be val wt a more general cost functon of te form C(( wt C/ < 0. At te aucton stage te only fference across frms s te margnal cost. Terefore a frm wt a low value of (a low margnal cost s unambguously more effcent tan anoter frm wt a larger. On te oter an te barganng ower s statstcally neenent of te margnal cost of roucton an s neenently rawn from a cumulatve strbuton functon G( wt robablty ensty functon g(. Alternatvely one coul assume tat eac frm knows ts own arameter but ts woul comlcate te eulbrum of te aucton roceure wtout ang muc to te analyss. On te eman se we assume tat tere are two grous of consumers. Te frst grou (nexe by te rc consumers arorates all te rents of te frm wereas te secon grou (nexe by te oor consumers oes not sare te rofts of te frm. Te roorton of rc consumers s β [0]. Bot grous of consumers alternate n government eenng on weter β s larger or smaller tan /. Tat s f β > / ten rc consumers are a majorty an tey are n ower. On te oter an f β < / oor consumers are majorty. e wll call S( te gross consumer utlty an u( S( te net utlty. Te eman functon s ten 5

6 etermne by utlty maxmzaton wc mles tat f s te rce of te goo ( s suc tat S ((. Te utlty functon s stocastc n suc a way tat were U [ ] s a eman sock an s te mean eman functon. e also assume tat once te eman functon s realze t remans te same n bot eros. Tat s at te en of te frst ero wen renegotaton can aen te eman for te secon ero s erfectly known. A rc majorty wll want to maxmze ts own surlus. Tey receve a roorton β of te total net utlty an all te frm s rofts terefore β [S(( (] β S(( ( β ( ( I. On te oter an a oor majorty wll maxmze ( β [S(( (] because tey get only a roorton of te total net utlty eual to ( β. Any grou as to take nto account te constrant tat te rofts of te frm are non-negatve for te frm to be wllng to rouce. Fnally we assume tat after roucton te rofts of te frm are erfectly observable by te government an terefore renegotaton occurs uner comlete nformaton. Bot te roft an te eman are observe by te government an terefore te margnal cost can be nferre. Te tmng of te game s as follows:. Eac frm learns ts rvate nformaton.. A frst-rce seale-b aucton s run an eac frm cooses a rce to sell te goo. Te frm wt te lowest b wns te aucton an s selecte to rouce te goo for eros. 3. Te actual eman s realze an s observe by te government an te frm.. Te frst ero of roucton ens an te frm an te government observe te frst ero roft of te frm. 5. If te roft s negatve te frm asks te government to renegotate te contract. If tere s a new majorty an te rce tat can be obtane wen renegotatng s smaller tan te actual rce te government asks for renegotaton. e assume tat at ts stage bot te government an te frm are locke n te relatons. e moel renegotaton as a Nas barganng game wt barganng owers for te frm an for te government. 6

7 6. If an agreement s reace te frm runs te concesson n te secon ero wt a new rce. Bot te government an te frm wll get 0 f tere s no agreement. 7. If tere s no renegotaton te orgnal contract revals. Note tat we assume tat te frm only asks for renegotatons wen rofts are negatve. Te reason s te followng. Suose tat te frm asks for renegotatons wen t s makng ostve rofts because t exects to ncrease te rce. If te government rejects te offer te frm can eter leave te market an get 0 (ts outse oton or stay an get some ostve roft. So te government knows tat te frm wll always stay an terefore wll reject any renegotaton tat ncreases te rce. On te oter an f te majorty oes not cange n te secon ero tere s no renegotaton rven by te government. So we are assumng tat tere s some lmte commtment on te government se. A gven grou can commt not to renegotate as long as t s n ower an te frm s not makng losses but s not able neter to force a new grou not to renegotate nor to oblge a frm to stay n te market wen rofts are negatve. So n terms of commtment ower our moel s les n between te stylze cases resente n Laffont an Trole (993 namely fullcommtment one ero-commtment an commtment only wen no renegotaton s mutually benefcal.. Otmal bng an renegotaton In ts secton we wll look for te soluton of te aucton game an ts mlcatons on te renegotaton rocess. Because we assume tat frms are myoc n te sense tat tey o not conser te ossblty of renegotaton at te aucton stage we start by solvng te eulbrum of te frst-rce seale-b aucton conserng tat a frm f t wns wll rouce for two eros wt te same rce.. Frst stage: soluton of te frst-rce seale-b aucton e look for a symmetrc Bayesan Nas eulbrum of te aucton n wc eac frm bs accorng to a functon ( ncreasng n. So te objectve functon of frm can be wrtten as 7

8 max γ ( E [( ( ( I] were γ( s te robablty of wnnng te aucton. Terefore γ( Pr ( < j j. Because we want to fn te Nas eulbrum we assume tat all te frms but are layng te eulbrum strategy an we look for te best resonse of frm. After tat we wll ceck tat nee tere s a symmetrc eulbrum an te functon ( s ncreasng. Because all te oter frms are bng accorng to te functon (j j we ave tat γ( Pr ( < (j j. Now because (j s ncreasng t can be nverte so tat j (j. So frm wll wn te aucton f j j > (. Tus γ( Pr ( ( < j j. Usng te assumton of neenence of j γ( [ F( (] N wt N beng te number of frms artcatng n te aucton. Tus frm s objectve functon wrtes: N max [ F( ( ] E [( ( ( I] ( Prooston Te wnnng frm s te most effcent one n te ool. Te execte roft of te wnnng frm s strctly ostve at te symmetrc eulbrum for any fnte N an for N te wnnng b s lower tan te monooly rce. Proof. See aenx. Because te frm oes not take nto account tat an ex ost renegotaton wt te government coul ncrease te rce n te secon ero te frst-rce seale-b aucton succees n selectng te most effcent frm to rouce te goo. Te ntuton s tat gven te rce of te goo a more effcent frm as always ger rofts an terefore s wllng to reuce a lttle bt ts b n orer to ncrease te robablty of wnnng te aucton. Ts wll always be true as long as frms are symmetrc n te sense tat all margnal costs are rawn from te same strbuton functon. More generally f te frms margnal costs are rawn from fferent strbuton functons or f tey were correlate wt a known I te concesson may be aware to a frm wose margnal cost s not te lowest one n te samle. But not necessarly te most effcent n an absolute sense. 8

9 On te oter an cometton at ts stage mles tat te rce of te goo s M smaller tan te monooly rce ( but larger tan te amsey rce ρ ( ρ ρ efne suc tat ( ( ( ( I comute at 0. emember tat our assumton of ncreasng returns to scale mles tat cometton n te market s not esrable. Moreover te fact tat total costs are not observable ex-ante by te government makes regulaton by average cost rcng (or amsey rcng mossble. Terefore te best way to avo ex ante monooly rcng s to ntrouce cometton for te market. Te benefts are llustrate n Prooston : In orer to ncrease te robablty of wnnng te aucton frms refer to carge a rce lower tan te monooly rce. Furtermore as te number of frms artcatng n te aucton ncreases te rce converges to te amsey rce.. enegotaton Tere are two cases n wc renegotaton can aen n ts moel. In te frst case te frm trggers renegotaton f te eman sock s suc tat te roft becomes negatve. In te secon case te government may trgger renegotaton. A new government majorty wll ask for renegotatons f tese can reuce te rce... Fgurng out te canges n rces from renegotaton. Snce te robablty of renegotaton s nfluence by te resultng rce we frst nee to see ow te renegotate rce eens on te barganng ower of te frm. Of course ts renegotate rce wll een on te majorty at te renegotaton stage. Te aenx sows ow te roblem s solve analytcally. In a nutsell te otmal renegotaton for eac majorty case wll result from te maxmzaton by te new rulng majorty of ts own welfare subject to a ostve constrant on rofts an takng nto account te strengt of ts barganng ower. Te frst orer contons of ts otmzaton roblem yel te followng result: Prooston (a If te rc wn te electon for te secon ero (.e. β > / te renegotate rce s gven by: 9

10 max ( ( β ( ( η( ( ( ( ρ { ( ( } for 0 ( 0] ( were ( s suc tat ( ( ( ( β η( an η( > s te eman elastcty. 0 (b If te oor wn te electon (.e. β < / te renegotate rce s gven by: η ρ ( ( for 0 Proof. See aenx. ( ( β ( ( ( ( 0] Prooston s ute ntutve an reasonable. It means tat te renegotate rce s a wegte average of te government s most referre rce an te frm s most referre rce an te wegts are gven by eac arty s barganng ower. Te extreme solutons of ts Nas barganng game are actually ute nterestng as well. If te frm as all te barganng ower ( te renegotate rce s eual to te monooly rce. Ts s true watever te grou n ower. On te oter extreme wen te government as all te barganng ower ( 0 te renegotate rce wll be eual to te amsey rce (eual to te average cost. Tese extremes efne te uer an lower boun for te renegotate rce revalng uner ntermeate values of. (3 If tere s a majorty of rc consumers an β s close to / t coul be te case tat te renegotate rce for 0 s larger tan te amsey rce f te amsey rce s smaller tan. But of course t wll never be lower tan te amsey rce. 0

11 Te ervatve of te frst orer contons to te otmzaton roblem yelng ( an (3 see aenx for a formal ervaton roves useful atonal nformaton on ow te effects of te canges n eac one of te olcy arameters. Te most nterestng results are summarze n: Prooston 3. Te comaratve statc results suggest tat: s ncreasng n I an for meanng tat ceters arbus te renegotate rce wll be ger for ger levels of costs (bot fxe an margnal an for ger levels of barganng ower for te frm; keeng fxe te cost of te frm te larger te barganng ower te closer te rce to te corresonng monooly rce. s constant n β meanng tat weter β s slgtly lower tan / or te wole socety belongs to te oor category te resultng rce wll be te same for gven values of I an. So tngs won t get any better for te oor after tey renegotate snce uner ermanent rulng by te oor rces are lkely to be as low as ossble from te begnnng. s ecreasng n β meanng tat f tere s a rc majorty n te secon ero te renegotate rce wll be smaller f te wole socety s rc tan f β s slgtly ger tan /. For gven values of te arameters an for any β wt eualty wen meanng tat cangng majorty from te rc to te oor wll result n lower tarff. Proof. See aenx. Fgure summarzes te man results an llustrates ow te renegotate rce canges wt te roorton of rc consumers.

12 0 / Fgure : enegotate rce as a functon of β β.. Fgurng out wo wll ntate te renegotaton an wen Usng te results of Prooston we comute te robablty of renegotaton as a functon of te arameters of te moel. Defne ϕ Pr(β > / / β < / Pr(β < / / β > / te robablty of avng a new majorty n ero. For smlcty we assume tat ts robablty s te same weter n te frst ero tere s a rc majorty or a oor majorty. Te total robablty of renegotaton f tere s ntally a majorty k (Pr( / k s ten te sum of te robablty of renegotaton rven by te frm (Pr(f an te robablty of renegotaton rven by te government gven tat n te frst ero grou k s n ower (Pr(g / k : Pr( / k Pr(f Pr(g / k Conser now searately te cases n wc te frm an te government ntate te renegotaton. As exlane before te frm wll ask for renegotatons wen ts rofts are negatve wc means:

13 Pr ( Pr f ( < 0 Pr( ( ( ( ( ( < I Pr < I ( ( ( ( Defne (( [ 0 suc tat ( ( I f ( ( < I I ( ( ( f ( ( ( So gves te mnmum eman sock tat wll allow te frm to make ostve rofts an terefore te frm wll not ntate renegotatons f. ( t ts efnton ( ( Pr Pr < because we f ave assume a unform strbuton for te eman sock. Te rce b n te frst ero s ger tan te amsey rce for 0 so < 0 an Pr(f < /. Terefore only negatve eman socks can trgger renegotaton by te frm. Te government wll ask for renegotatons f tere s a new majorty k wc occurs wt robablty ϕ an only wen te rce can be ecrease wc means bot tat te roft of te frm s ostve so te frm can accet to rouce wt a lower rce an tat ts barganng ower s small enoug because te renegotate rce ncreases wt. Tus Pr ( / k Pr( Majorty Pr( 0 Pr ö ( g ( ( ( ϕ G( ö ( ( (5 were ö ( ( s efne suc tat ( ö ( ( ( So te renegotate rce wll be lower tan te ntal rce f an only f < ö ( (. 3

14 From Prooston we can fn an exresson for te maxmum barganng ower of te frm tat wll nuce a new majorty to ask for renegotatons. Inee relacng an by ( n an 3 an solvng for we obtan: ö ( η( ( β ( ( η( ( β ( ( η( ( (6 ö ( β ( ( β ( ( η( ( (7 were (. So ö s ncreasng n an ecreasng n. Because te renegotate rce s always lower wen tere s a oor majorty t can be sown tat gven (( ö ö < meanng tat ceters arbus Pr(g/ > Pr(g/ tat s te robablty of renegotaton s larger f n te frst ero te rc consumers were n ower as llustrate by Fgure. Intutvely f te grou of rc consumers s n ower n ero ten te government can ask for renegotatons only f n ero te grou of oor consumers becomes a majorty. An watever te barganng ower an te margnal cost of te frm barganng wt te grou of oor consumers wll aceve a lower rce tan barganng wt a grou of rc consumers as state n Prooston 3. Terefore te renegotate rce wll be lower tan te ol rce for a larger range of values of wen te new majorty s a oor one. Analytcally usng ( an (5 te robablty of renegotaton wen grou k s n ower n ero s eual to: Pr( / k ( ( ( ( ϕ G( ö ( (..3 Fgurng out ow te ntal effcency level rves renegotaton os Snce regulators ten to be manly concerne about effcency t seems mortant to be able to assess te mortance of te ntal effcency levels obtane troug te aucton for te robablty of avng a renegotaton uner any tye of majorty. It also

15 5 els n assessng te mact of regulatory ecson amng at cangng tese effcency levels. Te formula for Pr( / k allows us to aress ts concern an to test for te mact of effcency canges for bot sources of renegotaton. Inee: g f k k / Pr Pr / Pr { ( ( ( ( ö ö ö ö ö ö / Pr Pr ϕ ϕ g f G g G g k Notce tat te effcency arameter of te frm affects bot te robablty of renegotaton rven by te frm wc s etermne by an te robablty of renegotaton rven by te government etermne bot by an ö. Startng wt te robablty of avng te frms reuestng a renegotaton te result allows us to nvestgate n artcular weter a more effcent frm (wc s te frm tat s gong to be selecte by te aucton roceure s always referre by te two grous of consumers. It suggests tat canges n te effcency of te frm as two ooste effects on t e robablty of renegotaton rven by te frm: ( te os of renegotaton ecreases wt te frm's effcency (.e. ncreases wt because a more effcent frm as lower costs an ts reuces te range of socks for wc te roft s negatve; ( te os of renegotaton ncreases wt ts effcency because a more effcent frm bs a lower rce an ts ncreases te range of socks for wc te roft s negatve. So a ror we cannot know weter te robablty of renegotaton rven by te frm ncreases or ecreases wt.

16 Te same s true for te robablty of renegotaton ntate by te government. On te one an ts robablty ecreases wt te frm s effcency because for a more effcent frm:. te renegotate rce s smaller an terefore te maxmum barganng ower tat trggers renegotaton ecreases;. te rce n te frst ero s smaller an terefore te range of eman socks for wc te frm s rofts are ostve ecreases. On te oter an te robablty of renegotaton rven by te government ncreases wt te frm s effcency because for a more effcent frm:. te maxmum barganng ower tat trggers renegotaton ncreases;. costs are lower an terefore te range of eman socks for wc te frm s rofts are ostve ncreases. 5. Effects of renegotaton on te effcency-euty trae-off In ts secton we scuss ow te moel eveloe so far answers uestons suc as: Is tere a conflct between effcency an euty? How oes renegotaton affect rc an oor consumers? Is tere any conflct between te two grous of consumers? at s te role laye by te barganng ower of te frm n ts context? 5. y users may refer to start wt an neffcent frm One source of trae-off between effcency an euty arses wen users ave a collectve ncentve not to refer to ave te most effcent frm wnnng te aucton knowng tat a renegotaton s lkely. If renegotaton were not ossble n oter wors f bot artes coul erfectly commt to te concesson contract sgne at te aucton stage tere woul be no ossble conflct between effcency an euty. Inee te aucton allows te government to coose te most effcent frm wtn te canates an ts frm carges te lowest ossble rce wc beneft te oor consumers n a larger roorton. An gven tat te effcency level s ex ante unknown tere s no better way to select te frm. en te ossblty of renegotaton s oene a conflct between effcency an euty may aear. For nstance we know tat te welfare of te oor consumers ecreases wen te frm ntates renegotaton because te rce ncreases. On te oter an we also know tat ts welfare ncreases wen te government ntates 6

17 renegotaton because n ts case te rce ecreases. In te revous Secton we sowe tat te robablty of renegotaton mgt ncrease or ecrease wt te effcency arameter of te frm. e want to know ow te oor consumers welfare Pr( f canges wt te effcency arameter of te frm. So suose tat < 0 an Pr( g / k > 0 tat s an mrovement n effcency (a ecrease n mroves te cances of a renegotaton by te frm wc te oor slke an reuces tose reueste by te government wc te oor lke. Ts suggests tat oor consumers may refer to ave a less effcent frm wnnng te aucton to rove oortuntes for government rven renegotatons wc wll lea to realstc tarff reuctons reuests. But of course tey ave to take nto account tat a less effcent frm wll oerate wt a ger rce n te frst ero an also n te secon ero f tere s no renegotaton. To see ow te two effects work togeter conser te followng very smle (but very artcular too examle. Assume frst tat ϕ 0 so te government never asks for renegotatons an secon tat te robablty of renegotaton rven by te frm ecreases wt 3 tat s < 0 (8 Ts mles tat a more effcent frm wll more often ave negatve rofts n te frst ero an terefore wll ask for renegotatons more often. But remember also tat wen te frm as negatve rofts te renegotaton wll always ncrease te rce. Now te magntue of te ncrease n rce wll een on te barganng ower of te frm at te renegotaton stage wc s ex ante unknown. So conser for nstance two frms caracterze by an > 0 so frm s more effcent tan frm an because of (8 te robablty of renegotaton s larger f frm s selecte. Te ex ante execte welfare for te two grous of consumers as a functon of s te sum of te excete welfare n ero gven te rce level carge ntally by 3 If te only source of renegotatons s negatve rofts ten f te robablty of renegotaton rven by te frm ncreases wt all consumers wll always refer te most effcent frm. Inee a more effcent frm wll carge a smaller rce n te frst ero wc s goo an wll trgger renegotaton less often wc s also goo. 7

18 8 te frm an of te execte welfare resultng from te rce n te secon ero wc s etermne by te exstence or absence of renegotaton: 0 g E 0 g E were s an ncreasng functon of an ( s an ncreasng functon of. Te frst term s te execte welfare n te frst ero gven tat te frm carges a rce eual to (. Te secon term s te execte welfare n te secon ero f tere s no renegotaton so te rce s stll eual to (. Fnally te tr term s te execte welfare n te secon ero f tere s renegotaton n wc case te rce wll be eual to an s ex ante unknown. e are ntereste n ow te execte welfare woul cange f tere s a small ecrease n te effcency level (a small ncrease n te margnal cost. { } 0 0 g g E

19 E ( ( ( ( 0 ( g( { } g 0 Four effects of on te execte welfare can be stnguse. Te frst tree are te classcal effects: an ncrease n reuces welfare n eac ero weter tere s renegotaton or not bot because te frm s costs are ger an because te rce of te goo n any case s also ger. Tese effects are reresente by te frst 3 terms wc are negatve mlyng tat wt te same robablty of renegotaton bot grous of consumers woul ave referre frm over frm. Te fourt effect s gven by te fact tat wen ncreases te robablty of renegotaton gven te assumton mae n (8 ecreases ecreasng te range of eman socks for wc te rce s g (eual to te renegotate rce an terefore welfare ncreases meanng tat bot grous of consumers refer frm over frm. 5 Terefore te net effect of over te execte welfare can be eter negatve n wc case effcency s always better or ostve n wc case a less effcent frm mgt be referable f te fourt effect s large enoug. Te fourt effect wll be large an ence a less effcent frm wll be referre wen te barganng ower strbuton functon g( s base towars large values In te case of te rc consumers t s not always true tat an ncrease n rce wll reuce welfare. Inee f te rce s lower tan wc s te rce tat maxmzes rc consumers' welfare a small ncrease n rce wll ncrease welfare. Inee f β s close to 0 s close to te monooly rce an te rce wtout renegotaton s smaller tan. e wll assume tat n tose cases te rect ecrease n welfare ue to an ncrease n comletely offsets te ncrease ue to an ncrease n te rce so te net effect s negatve. 5 Of course f te robablty of renegotaton ecreases wt te effcency of te frm ( > 0 ts fourt effect woul go n te same recton as te frst tree an more effcency woul be always better for consumers. 9

20 of because more wegt s ut on te cases for wc te ncrease n te rce s te largest. In ractce ts bas exsts wen cometton s lmte on te suly se (as s te case n te water sector an to a lesser extent n te ower ort an ralways sectors for nstance were te number of global layers s suc tat cometton for te market selom results n no more tan -3 bers. 6 In aton to te euty effcency trae-off just escrbe tere s an unerlyng strbutonal conflct between rc an oor. For β small enoug (.e. wen te oor rule sgnfcantly 7 we ave tat > an terefore tere are cases n wc te oor consumers refer frm over frm wle te reverse s true for te rc consumers. Te effcency-euty trae-off contnues to ol as well snce te aucton roceure wll always select frm. Te bottom lne s tat we can conclue tat te effcency level s not te unue relevant arameter rvng te selecton of a frm to oerate a concesson euty ssues must be taken nto account as well an may n fact ave to reval over effcency conseraton. en renegotaton rven by te frm s an ssue (wen te uncertanty about market contons s g a more effcent frm can en-u oeratng wt a ute g rce after renegotatons. 5. en rc an oor consumers on t see eye to eye Te strbutonal conflct just elue suggests tat t s also mortant to unerstan more generally te fferences n effcency outcomes snce tey may result n conflcts between te two grous of consumers. In artcular we soul be ntereste n tryng to classfy te stuatons n wc rc consumers an oor consumers wll agree or sagree about te stuaton tey refer te most. Te ossble stuatons are as follows:. en at te aucton stage tere s a rc majorty (β > / an at te en of ero : 6 Any tme we can exect a close relatons between te oerators an te government (or a oltcally rven regulator colluson s a rsk; te fewer te alternatves te stronger te frm's barganng ower an ence te stronger te rsks of colluson. 7 η Te exact conton s tat β <. In artcular te conton s satsfe wen oor consumers are majorty. 0

21 (a Te wnnng frm as negatve rofts so renegotaton occurs watever te majorty n ero an te rce of te goo ncreases. (b Te frm as ostve rofts an n ero tere s agan a rc majorty so no renegotaton can aen watever te value of. (c Te frm as ostve rofts but n ero tere s a oor majorty so renegotaton occurs f te frm s barganng ower s low enoug ( <ö an te rce of te goo ecreases.. en at te aucton stage tere s a oor majorty (β < / an at te en of ero : (a Te frm as negatve rofts so renegotaton occurs watever te majorty n ero an te rce of te goo ncreases. (b Te frm as ostve rofts but n ero tere s a rc majorty so renegotaton occurs f te frm s barganng ower s low enoug ( < ö an te rce of te goo ecreases. (c Te frm as ostve rofts an n ero tere s agan a oor majorty so no renegotaton can aen watever te value of. Cases a an a sow tat f te frm as negatve rofts n te frst ero bot grous of consumers are better off wen oor consumers wn te electon an rule urng te secon ero. Ts s because te magntue of te rce ncrease an te rce wll ncrease after renegotatons uner any tye of majorty s always lower wen te oor are n carge n te secon ero. Inee accorng to Prooston 3 te renegotate rce wll be smaller f β < / (see Fgure. Furtermore te robablty of beng n one of tese two cases s te same because te rce carge n te frst ero s neenent of β. Cases b an c sow tat f te frm as ostve rofts an tere s no cange n majorty watever te majorty no renegotaton aens watever te barganng ower of te frm. Moreover gven our assumton tat te robablty of reservng te majorty s neenent of β bot cases also arse wt te same robablty. Te most nterestng cases are c an b wen te frm as ostve rofts but tere s a cange n majorty. In tese two cases renegotaton occurs f te barganng

22 ower of te frm s small enoug. Accorng to Fgure ö ö < an terefore Pr( ö Pr( ö < < < an renegotaton occurs more often wen tere s a oor majorty n te secon ero. Furtermore even wen ö ö < < so renegotaton occurs uner any majorty te renegotate rce wll be lower n case c tan n case b because oor consumers are more concerne wt a ecrease n rce. tout ambguty oor consumers wll refer case c to case b. c consumers wll also refer case c as long as ( (. On te oter an f ö ö < < renegotaton wll occur n case c wle t woul not n case b. Te conflct between rc an oor consumers arses as follows. en te rc wn te electon for β small enoug ( s larger tan te renegotate rce obtane wt a oor majorty because as β goes to 0 ( aroaces te monooly rce. If ts were te case te grou of rc consumers refers not to renegotate an mantan a rce closer to te monooly rce. Ts beavor urts te oor wc woul ave bargane a muc lower rce f tey a won te electon an s tus te man source of conflct n te secon ero. A fnal source of strbutonal conflct s tat te loss ncurre by oor consumers wen te rce s ger tan te amsey rce s always larger tan te one ncurre by rc consumers because te latter grou receves te ncrease rofts of te frm. 6. Conclung remarks In ts aer we analyze te ossble sources of contract renegotaton an ter mact on te effcency-euty trae-off n non-comettve markets n wc regulaton by average cost rcng cannot be mlemente because tere s ncomlete nformaton about te frm s costs. To o ts we resente a moel of ncomlete contractng for concesson awar n wc neter te government nor te frm conser ex ante te ossblty of renegotaton yet renegotaton aens. Cometton for te contract troug an aucton ensures tat te consumers beneft by a rce tat s lower tan te monooly rce an allows te government to select te frm wt te lowest costs. e assume tat once te frm as starte to rouce te margnal cost of te frm can be nferre so at te renegotaton stage te government an te frm sare te same nformaton. Te government stll cannot always force te frm to carge a rce eual to te average cost because at ts stage te frm as acure some barganng ower an terefore te government s not able to extract all te rents from t.

23 Te moel roves ossble exlanatons for te varous tyes of renegotatons tat ave been observe n te last ecae an offers testable results. Frst te robablty of electon-rven renegotatons nucng reuctons n tarffs s larger f: ( n te frst ero te rc consumers (.e. te co-owners of te oeratng frms are n ower ( te oerators are makng rofts ( te oor take over after a cange n government an/or (v te government exects to ave a strong barganng ower. Secon watever te barganng ower an te margnal cost of te frm barganng wt te grou of oor consumers wll aceve a lower rce tan barganng wt a grou of rc consumers. Tr te smaller te grou of rc n te oulaton te ger te cances of avng g rces because tey receve te wole ncrease n te roft of te frm wle tey lose a small roorton of te reucton n te consumer surlus. Ts mles tat conflcts between rc an oor consumers are more lkely to occur te larger te relatve ower of te oor at te tme of renegotaton. Fourt te execte recton of te effects of canges n te frm s margnal cost on te robablty of renegotaton eens on a set of factors wt oosng sgn an s ence an emrcal matter reflectng te relatve strengt of tese factors. Fft te erverse effects of excessve barganng ower by te frm ten to enalze te oor relatvely more because tey ay a ger rce wtout beneftng by te ncrease n te frm s rofts. In terms of olcy avse te moel suggests tat f te fear of renegotaton s a major concern effcency s not te only varable tat matters wen selectng a frm to oerate a concesson. Inee consumers may want to awar te concesson to a less effcent frm n orer to reuce te robablty of renegotaton snce lower robabltes of frm rven renegotatons (ues to eman socks for nstance are assocate wt ger welfare levels. Secon te moel suggests also tat a benevolent welfare maxmzng government soul make every ossble effort to balance te barganng ower of ts regulators wt tat of te oerators to ensure a far treatment of all users. One of te actons to take an certanly not te least mortant s to mnmze te ossblty of colluson/corruton between te frm an te government unts n carge of te renegotatons an wo may not ave te same egree of benevolence. eucng barganng ower wll also reuce strategc bs an low-ballng by frms convnce tat tey can wn any renegotaton an get closer to monooly rces ex-ost. Tr any olcy nstrument tat reuces eman uncertanty wll be n te nterest of all users but n artcular te oor. Inee te mlct nsurance a by selectng a somewat neffcent frm at te aucton reuces 3

24 renegotatons ue to negatve eman socks wc n turn leaves te oor better off. In orer to refne some of te results atonal researc coul exlore te followng eas. Frst we ave consere te barganng ower of te frm as an exogenous arameter. In ractce t eens on several varables suc as te owners of te frm te oltcal nterference wt te coce of bers resultng from te blateral a te level of rofts or of sunk costs etc. A more careful stuy of te otental effects of eac one of tese factors seems arorate. Secon tere s a reutaton effect tat s not cature n te moel. In orer to avo te kn of strategc beavor aforementone te government can bul a reutaton of beng toug regarng renegotatons. Ts woul ave a ynamc ostve effect on welfare by avong unerbng an terefore ex ost non-esre ncreases n rce. A fnal atonal recton for mrovement coul be te moelng of te strbutonal ayoffs of nsurance mecansms to rotect from eman uncertanty.

25 7. Aenx Proof of Prooston. e are lookng for a subgame erfect Nas eulbrum of te game. Eac frm bs accorng to te frst orer conton: f ( ( ( I ( ( ( ( N 0 (9 F ( ( Now we nee to verfy tat a soluton of euaton 9 satsfes > 0 at te symmetrc eulbrum wc means tat te lowest rce wll be offere by te frm wt te lowest margnal cost. At te symmetrc eulbrum ( an ( so conton 9 wrtes: f ( ( ( ( ( I ( N F( ( ( ( ( ( ( ( > 0 (0 were te neualty ols because te rce s lower tan te monooly rce wenever N >. If te rce were ger tan te monooly rce te frm coul ncrease bot te execte roft an te robablty of renegotaton by ecreasng te rce. Ts also mles tat te execte roft of te frm s strctly ostve for any so te rce s lower tan te amsey rce. If ts were not te case te execte roft of te frm woul be negatve an terefore t woul wn by bng te amsey rce. Proof of Prooston. a Suose frst tat β > / so at te renegotaton stage tere s a rc majorty. Because of Nas barganng we know tat te otmal renegotaton rce wll solve te followng roblem: 5

26 max[ ] [ ] subject to 0 So usng te efntons of an an rememberng tat S ( we [ ] [ ] obtan te frst orer conton ( FOC 0 : ( ( β ( earrangng terms euaton can be foun. Te constrant s bnng for 0 f ρ ( > ( an s not bnng for any ( 0]. b Now suose tat β < / so at te renegotaton stage tere s a oor majorty. Te otmal renegotaton rce wll solve te followng roblem: max[ ] [ ] subject to 0 So usng te efntons of an an rememberng tat S ( we [ ] [ ] obtan te frst orer conton ( FOC 0 : ( ( β 6

27 7 earrangng terms euaton 3 can be foun. Te constrant s always bnng for 0 an s not bnng for any ( ] 0. Proof of Prooston 3. For comaratve statc we comute te ervatve of te frst orer conton wt resect to te arameter tat canges. To smlfy notatons we wll call (. Tus a FOC a sgn sgn for any arameter a. [ ] 0 > FOC [ ] 0 > I FOC 0 > FOC β 0 < FOC β β [ ] 0 > FOC

28 8 [ ] 0 > I FOC 0 > FOC β 0 FOC β β c gves our result. To sow tat t suffces to sow tat te frst orer conton wen tere s a oor majorty evaluate at s negatve meanng tat ecreasng te rce wll ncrease te objectve functon. Tat s we wll sow tat FOC ( < 0. S FOC Now takng nto account tat by efnton of FOC ( 0 we can relace an we obtan: S FOC β Fnally usng te efnton of ( we ave ( S FOC So FOC ( < 0 meanng tat.

29 eferences [] Agon P. M. Dewatront an P. ey (99 enegotaton Desgn wt Unverfable Informaton Econometrca 6: [] Bajar P. an S. Taell (000 Incentves versus Transacton Costs: A Teory of Procurement Contracts Mmeo. [3] Cung P.Y. (99 Incomlete Contracts Secfc Investment an sk Sarng evew of Economc Stues 58: [] Crames C. an A. Estace (998 egulatory Traeoffs n te Desgn of Concesson Contracts Utltes Polcy 7: -3. [5] Elyakme B. J.J. Laffont P. Losel an Q. Vuong (99 Frst-Prce Seale- B Auctons t Secret eservaton Prces Annales Econome et e Statstue 3: 5-. [6] Elyakme B. J.J. Laffont P. Losel an Q. Vuong (997 Auctonng an Barganng: An Econometrc Stuy of Tmber Auctons t Secret eservaton Prces Journal of Busness an Economc Statstcs 5: [7] Estace A. (00 Prvatzaton an egulaton of Transort Infrastructure n te 990s Te orl Bank esearc Observer 6 (: [8] Fuenberg D. an J. Trole (99 Game Teory MIT Press Cambrge. [9] Green J. an J.J. Laffont (99 enegotaton an te Form of Effcent Contracts Annales Econome et e Statstue 5/6: [0] Guasc J.L. (00 A ecae of concesson contracts n nfrastructure: Lessons from te exerence mmeo Te orl Bank. [] Hart O. an J. Moore (988 Incomlete Contracts an enegotaton Econometrca 56: [] Jeon D.S. an J.J. Laffont (999 enegotaton of Concesson Contracts: A Teoretcal Framework Mmeo. 9

30 [3] Kartaceva A.V. an L. Quesaa (000 Contract enegotatons on Concessons n Latn Amerca an Carbbean egon: An Economc Analyss an Emrcal Imlcatons Mmeo. [] Laffont J.J. an J. Trole (993 A Teory of Incentves n Procurement an egulaton MIT Press Cambrge. [5] Myerson. (98 Otmal Aucton Desgn Matematcs of Oeratons esearc 6.: [6] Segal I. an M. nston (000 Te Mrrlees Aroac to Mecansm Desgn wt enegotaton Mmeo. [7] Trole J. (999 Incomlete Contracts: ere o we Stan? Econometrca 67: Antono Estace M:\tena\luca\oor-reneg-august.oc August :8 AM 30