What case should you bring to court to alter precedent in your favour - a strong or a weak case?

Transcription

1 What cas should you bring to court to altr prcdnt in your favour - a strong or a wak cas? Hnrik Borchgrvink Dpartmnt of conomics, Univrsity of Oslo, PB 95 Blindrn, 37 Oslo, Norway Sptmbr, 9 Abstract Taking a cas to court in an unsttld lgal fild yilds nw prcdnt which affcts both subsqunt court cass and out of court sttlmnts. A long-run playr will sk to bring to court th cass that giv him th bst possibl prcdnt changs. This papr builds a modl to study th optimal cas to bring to court. Th modl shows that th optimal cas is compltly diffrnt undr two diffrnt, plausibl lgal rgims. JEL classification: K4 Kywords: Prcdnt; Slction for trial Introduction Evolution of prcdnt is affctd by th slction of cass prsntd to th court (s.g. Hadfild (99)). My papr studis what cas a disputant should bring to court to affct prcdnt in his favor. I show that th optimal cas to bring to court is compltly diffrnt undr two diffrnt, plausibl rgims. What cas a disputant would prfr bringing to court to affct prcdnt is so far not thoroughly studid in th litratur. Thanks to Ton Ogndal, Halvor Mhlum and Endr Stavang h.h.c.borchgrvink@con.uio.no, tl: , fax: This is surprising givn th larg litratur

2 on fficincy in volution of prcdnt in common law, s.g. Gnnaioli and Shlifr (7). My papr argus that a long-run playr can affct prcdnt volution, also in th long run, by slcting which cass to bring to court. Th argumnt rquirs an agnt who can control slction of court cass and who hav long-run intrst in affcting lgal prcdnt. Thr ar at last two typs who fulfill ths rquirmnts, stat authoritis who can choos in which cass whr thy will prss chargs, and firms who ar financially strong nough to bar unwantd cass through sttlmnt offrs. Th modl is prsntd through th cas of a firm. Th firm wants to minimiz th xpctd damag paymnts in th futur. Th firm is dfndant and choos whthr to sttl or tak th cas to court. Plaintiffs ar assumd on-shot playrs unintrstd in prcdnt, typically an individual. Prcdnt volution is modld as a procss whr nw prcdnt covrs a prviously uncovrd shar of th population of possibl futur cass. A cas covrd by prcdnt is commonly known to gt a crtain judgmnt (liabl or not liabl) with crtainty if it is takn to court. Th modl shows that of th cass not covrd by prcdnt, th optimal cas for th dfndant to bring to court to altr prcdnt in his favor is a strong cas (i.. wak vidnc against dfndant) in a lgal rgim whr judgs constantly adjust thir prsonal dcision standard for th cass uncovrd by prcdnt as prcdnt volvs. Th optimal cas is, on th contrary, a wak cas (strong vidnc against dfndant) in a lgal rgim whr judgs do not adjust thir idiosyncratic dcision standard as prcdnt volvs. Th rason for th optimal cas bing compltly diffrnt undr th two rgims is: Whn judgs do not adjust thir dcision standard thr is only th dirct ffct of th nw prcdnt covring mor cass. Whn judgs adjust thir dcision standard, thr is in addition an updating of th population of dcision standards for cass uncovrd by prcdnt. Th rsult holds for a broad rang of possibl distributions of th dcision standards of judgs. First, I build a modl and study what cas dfndant will bring in th vry first cas in court in a lgal fild without prcdnt. Thn, I us th rsults of th modl to discuss incntivs in th long run whn prcdnt can b changd mor than onc.

3 Th modl. St-up Th firm facs an xognous stram of cass against him,.g. suits for damags. For simplicity, assum thr ar on cas ach priod. Th disputd amount is J in all possibl cass. A cas has an vidnc lvl [, ]. Th cass ar indpndntly, uniformly distributd ovr this intrval. Highr mans a wors cas for dfndant (bttr for plaintiff). Th vidnc lvl is obsrvabl and vrifiabl... Prcdnt tchnology Th court judgmnt in th first cas brought to court,, binds all cass if was found not liabl, i.. dfndant won th cas, or binds all cass if was found liabl. This prcdnt tchnology is vry simpl, but capturs th natur of a cas law rgim with th star dcisis principl: Th law is nothing but th st of judgmnts mad in th past. For cass not covrd by prcdnt thr is a commonly known probability distribution dscribing th probability of a liabl judgmnt in court for ach vidnc lvl. Figur illustrats for th cas whr was found not liabl. Th cas whn is found liabl is paralll. Figur : Illustration of prcdnt and rmaining lgal uncrtainty. Two diffrnt lgal rgims I considr two diffrnt, plausibl lgal rgims. Thy diffr in how a nw prcdnt ffcts th liability probability distribution ovr th rmaining intrval whr thr is no 3

4 prcdnt. Th two mchanisms considrd rprsnt, arguably, th main two basically diffrnt mchanisms to considr in this stting. On rsmbls truncation and th othr rsmbls cnsoring. Th following is common for both rgims: Th judg who will dcid a cas takn to court is drawn from th population of judgs. If th cas is not covrd by prcdnt, th judg dcids th cas liabl if th vidnc lvl is largr than his idiosyncratic dcision standard. For a givn prcdnt ach judg has a givn idiosyncratic dcision standard. Dfin g( ) as th pdf of judgs idiosyncratic dcision standards bfor any prcdnt. Th support of g( ) is [, ]... Rlativists rgim: Rlativ dcision standards Each judg is ithr laissz-fair, and want all cass to b not liabl, or strict and want all cass to b liabl. Howvr, th judgs hav diffrnt idiosyncratic (political) cost of confronting th prcdnt by giving a cas rlativly clos to a liabl judgmnt or a cas rlativly clos to a not liabl judgmnt. Aftr th first prcdnt is mad, or is rplacd with th first cas. That is, th support of th distribution of judgs dcision standards is updatd to th updatd vidnc intrval without prcdnt,, if th first cas in court is found not liabl, and, if it is found liabl. Aftr th first prcdnt, th probability of a liabl judgmnt in court for any nw cas is thn as xprssd blow and illustratd in Figur and Figur 3. was found not liabl: F (;, ) = G(;, ) = g( )d < G(;, ) = g( )d < was found liabl: F (;, ) =.. Absolutists rgim: Absolut dcision standards In this rgim, judgs dcision standards do not chang with th nw prcdnt. In othr words, th support of th distribution of judgs dcision standards is [, ] irrspctiv of prcdnt. A judg acts in accordanc to his idiosyncratic dcision standard whn not 4

5 Rlativists rgim F(;,) Figur : Probability of a liabl judgmnt undr Rlativists rgim aftr a not liabl first cas. Rlativists rgim F(;,) Figur 3: Probability of a liabl judgmnt undr Rlativists rgim aftr a liabl first cas. 5

6 bound by prcdnt. A judg is bound by prcdnt whn prcdnt forcs him to giv a cas th opposit judgmnt of that of his own dcision standard. It follows that for a cas uncovrd by prcdnt, th probability of gtting a judg who finds th cas liabl, or not liabl, is th sam as bfor th first prcdnt. Aftr th first prcdnt, th probability of a liabl judgmnt in court for any nw cas is thn as xprssd blow and illustratd in Figur 4 and Figur 5. was found not liabl: F (;, ) = G(E) = g( )d < Absolutists rgim F(;,) Figur 4: Probability of a liabl judgmnt undr Absolutists rgim aftr a not liabl first cas. G(E) = g( )d < was found liabl: F (;, i ) = 6

7 Absolutists rgim F(;,) Figur 5: Probability of a liabl judgmnt undr Absolutists rgim aftr a liabl first cas...3 Symmtry For simplicity, I assum that g( ) is symmtric, so G(;, ) is symmtric undr both rgims. Undr Rlativists rgim G(.) is thn symmtric also aftr th first prcdnt. For th Rlativists rgim, symmtry implis that thr ar as many laissz-fair judgs as strict judgs and th cost distribution within ach group is idntical. For th Absolutists rgim, symmtry implis that th idiosyncratic dcision standards ar symmtrically distributd around th mid-point =...4 Comparison of th ffct of nw prcdnt undr th two rgims Undr Absolutists rgim thr is only th dirct prcdnt ffct from th cass that bcam covrd by th nw prcdnt. Undr Rlativists rgim, howvr, thr is an additional ffct ovr th cass still not covrd, du to th updating of th distribution of th judgs dcision standards. Th two ffcts undr Rlativists rgim work in th sam dirction: If is found not liabl, th dirct ffct is th mony dfndant savs for all cass blow and qual to. In addition, dfndant will pay lss for all cass with vidnc This is not a vry rstrictiv assumption as on is fr to dsign th scal of th support of g(.), [, ]. 7

8 lvl in th intrval,, bcaus ach judg s dcision standard is movd closr to as th vidnc intrval uncovrd by prcdnt is shortnd from blow. Similarly if is found liabl. 3 Tak to court or sttl? Dfndant has two objctivs whn a cas mrgs and h has to choos btwn taking th cas to court or sttling it: Minimiz paymnts to plaintiff in th particular cas and improv futur prcdnt. To focus on th prcdnt objctiv, th amount paid in sttlmnts is assumd qual to th xpctd damag pay if th cas was takn to court: F (;, ) J. F (.) is assumd common knowldg. Court costs ar ignord for simplicity. For th objctiv of minimizing damag paymnts to plaintiff in th particular cas in hand, th risk nutral dfndant is thus indiffrnt btwn court cas and sttlmnt. Plaintiff is assumd to accpt th sttlmnt amount, or at last ɛ mor: Plaintiff is by assumption not abl to xtract any of th valu of a potntial prcdnt chang from th dfndant. 3 Consquntly, dfndant controls th slction of cass to th court. For th objctiv of improving prcdnt th choic btwn sttlmnt and court cas mattrs. Futur damag paymnts in both sttlmnts and court cass ar changd if th cas is takn to court, confr th prsntation of th two rgims. Consquntly, dfndant can affct prcdnt through th slction of cass to court. If dfndant could not control whthr a cas gos to court or not, i. if som random cas wr takn to court to mak prcdnt, th prcdnt chang would also b random and thus x ant in ithr litigants favor bcaus th distribution of judgs is assumd symmtric. 3. Two priods Assum thr ar only two priods: Bfor and aftr th vry first cas. This simplification nabls m to dscrib problm of th dfndant without assuming particular forms of th judg population function g( ). Dfndant will sk to minimiz xpctd damag Avoiding futur court costs is clarly anothr incntiv for taking cass to court to obtain a prcdnt undr which sttlmnts can b mad in futur, s.g. Robson and Skaprdas (8). This incntiv is not th focus of my papr. 3 In othr words, dfndant has all bargaining powr. 8

9 paymnt in th scond priod by optimally choosing sttlmnt or court in th cas in th first priod. Th xpctd damag pay this priod and nxt priod undr th two options ar xprssd blow. Dfndant will choos th lowst of thm to minimiz his paymnts. For ach possibl cas nxt priod,, dfndant will pay F ( ;, ) J if is sttld outsid court and F ( ;, ) J or F ( ;, ) J if is takn to court. Th possibl cass nxt priod ar continuously and uniformly distributd ovr [, ], so th wight to b put on ach cas is th sam. Consquntly, whn comparing th two options sttl and tak to court, it is sufficint to hav βj F (.)d as xpctd damag pay for th cas which appars nxt priod. Sttl: S = F ( ;, ) J + βj F (;, )d () Tak to court: C = F ( ;, )[J + βj F (;, )d] + ( F ( ;, ))[ + βj F (;, )d] () This minimization problm can b simplifid in th following way. First, paymnt today, F ( ;, ) J, which is qual undr th two options, can b subtractd from both. Thn th xprssions only contain xpctd damag pay nxt priod. Th minimization problm is thn: min[ F (;, )d, F ( ;, )[ F (;, )d] + ( F ( ;, ))[ F (;, )d]] (3) In othr words, th option, sttl or tak to court, which minimizs damag paymnts is th on whr th xpctd siz of th ara undr F (.) is smallst. 3. Rlativists rgim Considr th cas and in Figur 6. If dfndant sttls th cas, th xpctd damag pay nxt priod is th sam as undr prvailing prcdnt bfor th cas matrializd. This xpctd damag pay is rprsntd by th ara undr th graph from to. If 9

10 dfndant taks th cas to th court and loss, th nw prcdnt is,. Th corrsponding nw damag pay is th ara undr th graph from to. This nw prcdnt implis an incras in xpctd pay nxt priod. In othr words, dfndant loss th ara L compard to th prvailing prcdnt. Similarly, if dfndant wins th cas, th nw prcdnt is, and dfndant wins th ara W. Th probability-wightd sum of L and W is what dfndant compars to no chang in prcdnt whn choosing sttl or tak to court. Rlativists rgim, Expctd damag paymnt as intgrals of F(.) J F(.),Damag paymnt L W Figur 6: Rlativists rgim. Th thr possibl prcdnts nxt priod. 3.3 Absolutists rgim Considr th cas in Figur 7 for illustration undr absolutists rgim. If dfndant loss th cas, th nw prcdnt is, and dfndant loss th ara L, bcaus only th judgs with idiosyncratic dcision standard btwn and will b unbound by prcdnt. Similarly, if dfndant wins th cas, h wins th ara W. Dfndant compars th probability-wightd sum of L and W to no prcdnt chang whn choosing sttl or tak to court.

11 Absolutists rgim, Expctd damag paymnt as intgrals of F(.) J F(.),Damag paymnt L W Figur 7: Absolutists rgim. Th thr possibl prcdnts nxt priod. 3.4 Dfndant s choic I will dscrib for which intrvals dfndant will want to tak th cas to court and not. A cas, < <, for which xpctd chang in damag pay nxt priod is zro, is calld ẽ. In th following I considr only symmtric distributions g(.) with no or on local maximum point, i.. g (.) for and g (.) for Rlativists rgim Undr Rlativists rgim it is asy to driv a gnral xprssion for th xpctd damag pay nxt priod. Dfndant pays J for cass if is found not liabl and J for cass if is found liabl. For cass in th intrval uncovrd by prcdnt, dfndant pays on avrag J, bcaus G(.) is symmtric. Thrfor xpctd damag pay nxt priod can b xprssd: is found not liabl F (;, )d = + ( ) = ( ) (4)

12 is found liabl F (;, )d = ( ) + ( ) (5) = Expctd chang in xpctd damag pay nxt priod from taking to court can thrfor b xprssd: F ( ;, ) F (;, )d + ( F ( ;, )) F (;, )d F (;, )d = F ( ;, )( ) + ( F ( ;, )) ( ) = (F ( ;, ) ) (6) Proposition Undr Rlativists rgim, i) Expctd chang in damag pay is zro for =, = and = ẽ =, ii) For uniform judg distribution g( ) xpctd chang in damag pay is zro for all, iii) For symmtric judg distribution g( ) with a mod, xpctd chang in damag pay is in dfndant s favor for cass < ẽ and in dfndant s disfavor for > ẽ. Proof. i) F (;, ) = and F (;, ) = mak Equation 6 zro. Bcaus F (;, ) is symmtric by assumption, F ( ;, ) =, so Equation 6 is zro for = ẽ =. ii) For uniform F (;, ), F (;, ) = for all, so Equation 6 is zro for all. iii) A symmtric F (;, ) with a mod yilds F (;, ) < and Equation 6< for < <, and F (;, ) > and Equation 6> for < <. Th rsult for symmtric judg distribution with a mod has th following intuition: If is clos to, losing th cas implis a vry bad prcdnt for dfndant, but th probability of losing is vry small. Th probability of winning th cas is larg and th rsulting prcdnt improvmnt is not only stablishing zro damag pay in th small intrval blow, but also to pay a littl lss in xpctation for all cass abov. Thrfor, dfndant has incntivs to bring to court cass blow ẽ =. In othr

13 words, undr Rlativists rgim, dfndant will bring to court only if it is mor likly that h wins th cas than loss it. For uniform distribution, th rsult that dfndant is indiffrnt btwn sttling and bringing to court any cas, is a consqunc of th xtrm symmtry of th uniform distribution. Figur 8 dpicts th situation of uniform distribution for an <. Subscript d in Figur 8 rfrs to th part of th chang which is du to dirct prcdnt covrag and u th part that is du to updating of th distribution of judgs. Expctd chang in damag paymnt is qual to zro du to symmtry: F ( )(L d + L u ) ( F ( ))(W d + W u ) = ( ) ( ) = (7) Rlativists rgim, Expctd damag paymnt as intgrals of F(.) J Ld F(.) F( ) Lu Wd Wu Figur 8: Rlativists rgim. Uniform judg distribution. From Figur 8 and Equation 7 it is asy to s what happns whn judg distribution is changd from uniform to a symmtric with mod: Th straight lins in Figur 8 bcom S-shapd. Not that L = L d + L u and W = W d + W u will b unchangd bcaus also th distribution with mod is symmtric. Howvr, F ( ) will dcras and thus ( F ( )) incrass. Consquntly, th vrsion of Equation 7 for distribution with mod will b ngativ, which mans that xpctd damag pay from taking to court is rducd. Figur 9 xmplifis using th triangl distribution. 3

14 Rlativists rgim, Triangl distribution J Rlativists rgim, Triangl distribution F(;,) Expctd chang in damag pay (a) (b) Figur 9: Rlativists rgim with triangl judg distribution: (a) Th cdf, (b) Expctd chang in damag pay [CHECK th triang with mov in not from /-8. Insrt mtring on y-axis on all thr] From Figur 8 and Equation 7 it is also asy to anticipat th rsult undr Absolutists rgim which is tratd in nxt subsction. Undr Absolutists rgim thr is no updating ffct, so L u = W u =. For Equation 7, that implis subtracting F ( )L u and adding ( F ( ))W u. Bcaus L u = W u, having sam ground lin and hight, whil F ( ) < ( F ( )), th modifid vrsion of Equation 7 will bcom positiv. That is, undr Absolutists rgim and uniform judg distribution xpctd chang in damag pay from taking < to court is positiv. In othr words, th opposit rsult of Rlativists rgim and symmtric judg distribution with mod Absolutists rgim Undr Absolutists rgim, an xprssion for th xpctd damag pay can b drivd in th following way. Bcaus th distribution of judgs g( ) is symmtric and [, ], th ara undr F (;, ) is qual to ( ) =. Aftr an has mad prcdnt, th rmaining ara is subtractd th ara cut off by th prcdnt. If was found liabl, th full pay intrval must b addd: is found not liabl F (;, )d (8) 4

15 is found liabl F (;, )d + (9) Givn ths xprssions for xpctd damag pay, th xpctd chang in damag pay from taking cas to court can b writtn: F ( ;, ) ( F (;, ))d ( F ( ;, )) F (;, )d () Proposition Undr Absolutists rgim, i) Expctd chang in damag pay is zro for =, = and = ẽ =, ii) For uniform g(.), xpctd chang in damag pay is in dfndant s favor for cass > ẽ and in dfndant s disfavor for < ẽ. iii) For symmtric judg distributions with mod, xpctd prcdnt chang is in dfndant s disfavor for cass clos to,, and in dfndant s favor for cass clos to,. Proof. i) F (;, ) = and F (;, ) = mak Equation zro. Bcaus F (;, ) is symmtric by assumption, F ( ;, ) =, so Equation is zro for = ẽ =. ii) For uniform distribution, Equation is: ( ) ( ) = ( 3 + ) which is > for < < and < for < <. iii) First, considr a vry small. Lt F ( ) = ɛ. Lt F (;, )d = δ ɛ. Thn Equation can b writtn ɛ( ( δ)) ( ɛ)δ = ɛ( + δ δ ) which is > as also δ is smallr than ɛ. ɛ Scond, considr th paralll cas for an qually clos to. Thn Equation can b writtn ( ɛ)δ ɛ( ( δ)) = ɛ( + δ + δ ) which is < for th sam rason. ɛ Th rsult for symmtric judg distribution with a mod has th following intuition: If is clos to, although th probability of winning th cas is vry high, th prcdnt gain from winning is vry small, sinc is clos to. Th vry larg prcdnt loss from losing th cas is dominating vn though probability of losing is vry low. For an clos to it is th othr way around. This rsult is valid for all judg distributions with a mod. Figur xmplifis using th triangl distribution. Howvr, xpctd chang in damag paymnt for around, has no gnral answr valid for all symmtric judg distributions with a mod. Of cours, distributions sufficintly clos to th uniform distribution will hav th sam rsult as uniform distribution, 5

16 Absolutists rgim, Triangl distribution J Absolutists rgim, Triangl distribution F(;,) Expctd chang in damag pay (a) (b) Figur : Absolutists rgim with triangl judg distribution: (a) Th cdf, (b) Expctd chang in damag pay i.. xpctd chang in damag pay changs sign only for =. Figur shows that th triangl distribution is sufficintly similar to th uniform distribution in this rspct. Howvr, th othr xtrm, whn th cumulativ dnsity function is vry stp for =, xpctd chang in damag pay will also incras for. In Figur (a), such a distribution is for simplicity xmplifid with a thr-pics-linar cdf. Th corrsponding xpctd chang in damag pay has two mor points whr xpctd chang in damag pay changs sign, s Figur (b). Not that also for this xtrm distribution for th bulk of th intrval [, ] it is still so that dfndant will want to bring cass h xpcts to los. Absolutists rgim Absolutists rgim F(;,) Expctd chang in damag pay (a) (b) Figur : Illustration of a judg distribution with stp cdf: (a) Th cdf, (b) Expctd chang in damag pay 6

17 4 Long run Thr ar implications from th two-priod modl for th outcom undr infinitly many priods. First, rdfin prcdnt to opn for continud prcdnt changs. Lt prcdnt b dfind by a pair l, h whr l is th so far highst vidnc lvl that has bn found not liabl in court and h is th so far lowst vidnc lvl that has bn found liabl in court. 4. Rlativists rgim Th probability of a liabl judgmnt in court for a givn cas i and prcdnt l, h undr this rgim is thn as xprssd blow and illustratd in Figur. i l i F ( i ; l, h) = G( i ; l, h) = g( l h l )d l < E < h l h i F(;l,h), Rlativists rgim J F(;l,h) l h Figur : Probability of a liabl judgmnt undr rlativists rgim Bcaus th population of judgs dcision standards ar constantly updatd to b symmtrically distributd ovr th vidnc intrval not covrd by prcdnt, l, h, dfndant will always hav incntivs to tak som cass to court, confr Proposition. Consquntly, prcdnt will in th long run convrg to covr all cass, unlss dfndant 7

18 has som costs of bringing cass to court that at som point will prvnt him from bringing mor cass bcaus th gain from nw prcdnt dcrass whn th vidnc intrval not covrd by prcdnt dcrass. Howvr, dfndant will not tak any cas with i < (l + h) to court. Dfndant will hav incntivs to wait for cass clos to l. H will wait longr, th mor wight h puts on futur damag paymnt, i.. whn β is high. Again this is du to th constant updating of th dcision standards. Th risk of losing and paying mor in th futur is lowr whn prcdnt is volving with smallr stps. To illustrat, F (l + δ; l, h) > F (l + δ ; l, h) + F (l + δ; l + δ, h), whr th vidnc intrval δ is covrd in on stp on th lft hand sid, and in two qual stps on th right hand sid. Givn that dfndant wins in both options in th inquality abov, h pays mor for th scond cas undr th right hand sid option, but if dfndant has sufficintly high β this is dominatd by lowr xpctd pay in futur bcaus of th lowr risk of losing th intrval (l + δ, h) for all futur. 4.. Absolutists rgim Th probability of a liabl judgmnt in court for a givn cas i and prcdnt l, h undr Absolutists rgim is as xprssd blow and illustratd in Figur 3. i l i F ( i ; l, h) = G( i ) = g( )d l < i < h h i If dfndant loss a cas i that h taks to court, h is lowrd to h = i. Consquntly, thr is thn lss to los in subsqunt cass brought to court. Whn thr is lss to los from taking a cas to th court, incntivs for taking cass to court incrass, so dfndant will dfinitly kp on taking som cass to court. Similarly, if dfndant wins a cas takn to court, thr will b lss to win in th futur, so incntivs to bring cass to court dclins. Nvrthlss, thr will always b som cass clos to h in th uncovrd lgal intrval which lowrs xpctd damag pay in futur. Similarly, cass clos to l will always hav xpctd incras in damag pay if takn to court. This follows of part iii) in Proposition from th two-priod modl. This part is valid for any prcdnt l, h: Equation for a cas i l, h bcoms: 8

19 F(;l,h), Absolutists rgim J F(;l,h) l h Figur 3: Probability of a liabl judgmnt undr absolutists rgim h F ( i ; l, h) i ( F (; l, h))d ( F ( i ; l, h)) i l F (; l, h)d () Equation is > for th cas of ( i = l + ɛ; l, h) and < for ( i = h ɛ; l, h) bcaus, in both cass, on of th intgrals bcoms infinitsimally small for infinitsimally small ɛ. In othr words, as long as thr ar cass not covrd by prcdnt, thr xists cass which dfndant will want to tak to th court. Bcaus thr is no updating of th population of judgs, th probability of losing dos nithr dclin nor incras by ltting prcdnt volv in small stps rathr than larg stps - contrary to undr Rlativists rgim. Howvr, also undr Absolutists rgim, dfndant has incntivs to not tak to court any cas that in xpctation improvs prcdnt. Dfndant will hav incntivs to wait for cass clos to h: In cass clos to h, dfndant gains most if h is lucky and wins, and h dos not los much if h loss. If dfndant in stad brings to court a cas furthr away from h (but still with xpctd rduction in damag pay), h incrass th probability of winning th cas. Howvr, if h loss it, h has for all futur lost th intrval btwn this cas and th formr h. It follows that th mor patint dfndant is and th longr th horizon, it is mor likly that dfndant only will bring to court cass clos to h, i.. waiting for a surprisingly 9

20 good prcdnt. 5 Discussion 5. Spd of volution Undr both rgims, thr will for any prcdnt l, h b cass that dfndant wants to bring to court. Thrfor, in th long run, both rgims will convrg to full prcdnt covrag of th vidnc intrval. Howvr, as th uncrtain lgal intrval shrinks with nw prcdnt, both th valu of nw prcdnt and th probability of gtting a cas insid th uncrtain lgal intrval shrink. In othr words, prcdnt volution will b slowr ovr tim. In particular, if a cas clos to h is won undr Absolutists rgim, thr will b a vry small intrval lft to win in futur prcdnt improvmnts and th intrval whr dfndant has incntiv to bring cass will b vn smallr, so furthr prcdnt volution will b vry slow. Not that if a cost of taking cass to court rlativ to sttlmnt is addd to th modl, thr could b uncrtain lgal intrval lft in th long run. 5. Th paralll cas of an authority Th rsults of th modl ar translatabl to th cas of an authority,.g. th antitrust authority, contmplating which cas to bring to court in an unsttld lgal fild. Bcaus such an authority acts as plaintiff whil th cas of a firm, as prsntd in th modl, acts as dfndant, th main translation will b to chang signs. In addition, th concivably mor complx objctiv function of a stat authority will mak th pictur mor complicatd. Not that in th cas of an authority suing a privat lgal ntity, thr is no nd to assum asymmtry btwn th two litigants with rspct to intrst in prcdnt. Evn if also th privat lgal ntity is intrstd in prcdnt, it is th authority who both chooss which cass to prss chargs in and whthr thy will accpt a sttlmnt offr. For th sam rason, it is nithr ncssary to assum that ithr party has th ntir bargaining powr in sttlmnts.

21 6 Conclusion I hav built a modl to answr whthr a dfndant with long trm intrst in prcdnt has incntivs to bring a wak or a strong cas to court whn altring prcdnt in his favor is th objctiv. Th main rsult of th modl is that th optimal kind of cas to bring to court hings on th lgal systm, in particular whthr or not judgs adjust thir idiosyncratic dcision standard as prcdnt volvs. Th analysis shows that a dfndant will want to bring cass h xpcts to win whn judgs adjust thir dcision standard, whras if judgs do not adjust thir idiosyncratic dcision standard th dfndant will bring cass h xpcts to los in a gambl for a surprisingly good prcdnt. In th vry long run, thr will b no lgal uncrtainty lft undr both rgims, but th long-run prcdnt will b path dpndnt and thus affctd by th slction of cass that has bn prsntd to th court. Rfrncs Gnnaioli, N., Shlifr, A., 7. Th volution of common law. Journal of Political Economy 5 (), Hadfild, G., 99. Bias in th volution of lgal ruls. Gorgtown Law Journal 8, Robson, A., Skaprdas, S., 8. Costly nforcmnt of proprty rights and th coas thorm. Economic Thory 36 (), 9 8.