Judicial Review as a Constraint on Tyranny of the Majority

Transcription

1 Judcal Revew as a Constrant on Tyranny of the Majorty Robert K. Fleck Department of Agrcultural Economcs and Economcs Montana State Unversty Bozeman, MT phone: (406) e-mal: rfleck@montana.edu F. Andrew Hanssen Department of Economcs Colby College Watervlle, ME phone: (207) e-mal: ahanssen@colby.edu Aprl 30, 2010 Abstract: We develop a theoretcal model to analyze the role of judcal revew. The model dentfes condtons under whch the optmal court wll not prevent tyrannes of the majorty and, ndeed, even the mnorty may prefer that the court not block such tyrannes. These results hnge on the tmng of two events: the lftng of the vel of gnorance wth respect to who gans and who loses from the polcy subject to judcal revew, and a random shock (.e., new nformaton) that affects the level of the payoffs generated by that polcy. We demonstrate how to nterpret the fndngs of our model wthn the context of three controversal rulngs (Serrano v. Prest, Kelo v. Cty of New London, and In re Marrage Cases). In so dong, we demonstrate how the mplcatons of our model allow us to dstngush scenaros n whch judcal constrants on majorty rule are socally benefcal from those n whch they are socally harmful. Through ths theoretcal exercse and dscusson of cases, our paper clarfes what condtons must be consdered n order to evaluate the proper role of judcal revew. For helpful comments, we thank Tony Cookson, Nck Parker, Randy Rucker, Chrs Stoddard, and semnar partcpants at Montana State Unversty.

2 If a majorty be unted by a common nterest, the rghts of the mnorty wll be nsecure. James Madson, Federalst 10 I. Introducton The potental for democratc majortes to tyrannze mnortes has long been recognzed. To guard aganst t, the ancent Greeks posted that even the wll of the people must be subject to 1 the rule of law. More than 2000 years later, Locke and Montesqueu emphaszed the mportance of checks and balances, and the Federalst Papers famously debated the best means for dealng wth 2 the threat posed by dvergent nterests. In the Unted States today, a wdely held vew among legal scholars s that the judcary s (or at least should be) a crtcal bulwark aganst tyrannzng 3 majortes. Recent theoretcal work by economsts ponts n the same drecton for example, Maskn and Trole (2004) model the tradeoff between accountablty and ndependence and conclude 1 Ths s evnced, for example, by Thucyddes account of Percles famous oraton: We are free and tolerant n our prvate lves; but n publc affars we keep to the law. Ths s because t commands our deep respect. We gve our obedence to those whom we put n postons of authorty, and we obey the laws themselves, especally those whch are for the protecton of the oppressed, and those unwrtten laws whch t s an acknowledged shame to break. See, e.g., Jowett (1881, ). 2 It s of great mportance... to guard one part of the socety aganst the njustce of the other part. Dfferent nterests necessarly exst n dfferent classes of ctzens (Madson, Federalst 10). The federal structure was promoted on the grounds that the dversty of nterests wthn t would, when combned wth varous checks and balances, reduce the potental problems arsng from tyranny by a majorty. Presumably, such a structure was supported by ctzens who saw the potental to cycle through beng n the tyrannzed group. 3 For example, Kurland (1969, 45) wrtes that the prmary functons of the U.S. Supreme Court are to protect the ndvdual aganst the Levathan of government and to protect mnortes aganst oppresson by majortes. See Bckel (1962) and Ely (1980) for classc works. A varety of concurrng opnons have been ssued by varous courts, the most famous of whch s Justce Stone s Carolene Products (1938) footnote, assertng that courts should step n when prejudce aganst dscrete and nsular mnortes... tends serously to curtal operaton of those poltcal processes ordnarly reled upon to protect mnortes (see, e.g., Fredman 2002). See Klarman (1996, 1-2) for smlar sentments expressed n other Supreme Court decsons. On the arguments aganst judcal revew, see, e.g., Waldron (2006). 1

3 that relyng on an ndependent court s preferable when the majorty s preferences are lkely to 4 nflct large negatve externaltes on the mnorty (p. 1050). Indeed, t s not clear how else to constran a majorty wthn a democratc settng how can a socety establsh a decson-makng system that reles on majorty rule except when the majorty s wrong? In short, whle scholars have questoned the ablty of courts to protect tyrannzed mnortes (e.g., Dahl 1957, 1989; Klarman 1996), or how to square such protecton wth democratc prncples (e.g., Bckel 1962; Ely 1980), the conventonal wsdom s that courts are a potentally mportant mechansm for lmtng majorty abuses of mnorty rghts. 5 In ths paper, we examne the condtons under whch a court that systematcally rejects tyrannous polces wll make potentally tyrannzed mnortes better off (or worse off) than would a court that merely followed the wll of the majorty. Our theoretcal analyss starts by characterzng two mportant roles of the judcal system, then asks what happens when the court serves those two roles jontly. One role s, as dscussed above, protectng mnortes from majortes. The other s 4 La Porta et al. (2004) attrbute ther emprcal fndng that economc freedoms are greater where courts are more ndependent to the fact that courts can check legslatve majortes that threaten to nfrnge upon mnorty rghts. Hanssen (2004) fnds evdence that more ndependent courts are establshed where competton between rval poltcal partes s tghter, consstent wth the hypothess that ndependent courts render polcy changes more costly. Our paper s also related to the noton that legal and poltcal nsttutons may serve as commtment devces (e.g., Landes and Posner 1975; North and Wengast 1989; Wengast 1995; Qan and Wengast 1997; Acemoglu and Robnson 2000, 2001; Falaschett and Mller 2001; Fleck and Hanssen 2006; Anderson and Parker 2008; Falaschett 2008). 5 Dahl (1957, 1989), McCloskey (1960), and many others have argued that there are few, f any, nstances n whch courts have stood frm n the face of ferce publc opnon. There s also a vast lterature debatng how courts should and to what extent they do adjust ther rulngs durng tmes of crss; Epsten et al. (2005) revew the lterature and provde the most careful emprcal analyss of court behavor durng tmes of crses. As Epsten et al. (2005, 9) conclude, The justces are, n fact, sgnfcantly more lkely to curtal rghts and lbertes durng tmes of war and other nternatonal threats, yet nterestngly whle the presence of war does affect cases unrelated to the war, there s no evdence that the presence of war affects cases drectly related to the war. 2

4 actng as an agent of the publc when uncoverng and evaluatng nformaton (whch courts regularly 6 do n actvtes rangng from crmnal trals to overseeng the enforcement of economc regulaton). Our analyss focuses on nformaton and the tmng of two events: When are voters out from behnd a vel of gnorance wth respect to beng among the benefcares or among the losers from a prospectve polcy, and when are shocks that nfluence the net socal benefts of that prospectve polcy observed? The order n whch those two nformaton-revealng events occur s crtcal when courts play a dual role (protectng mnortes, uncoverng and evaluatng nformaton). In some cases, the optmal decson rule for the court s smple: overrule majorty votes that generate more costs than benefts for socety. Yet n other cases, followng such a decson rule wll be a mstake: When the ntal majorty rule decsons take place after the vel s lfted, but before the polcy s net benefts are revealed, t can be socally desrable for a court not to block majorty-supported polces that are revealed ex post to be socally harmful. That s, t may be better for the mnorty as well as the majorty to have a court that makes decsons solely on the bass of benefts to the majorty, or to have no judcal revew at all. Our analyss has practcal mplcatons for understandng why the court s role n protectng mnorty rghts wll be both complcated and controversal. As the model shows, even under relatvely deal condtons a court that can commt credbly to future rulngs, shocks that (once observed) ndcate the exact benefts of polcy, and a publc that comples wth court rulngs judcal efforts to protect mnortes from majorty-supported neffcent polces may turn out to be 6 Researchers have suggested that when courts can uncover (or react to) relevant nformaton more rapdly than can the publc, court oversght can be welfare-enhancng (e.g., Maskn and Trole 2004; Fleck and Hanssen forthcomng). 3

5 counter-productve. Although tyranny of the majorty and judcal revew have both generated vast lteratures, our paper makes a unque contrbuton. We are the frst to model how the judcary s optmal role n protectng mnortes relates to ts role n uncoverng and evaluatng nformaton about polcy, and we model ths n a way that llumnates how the court s optmal decson rule depends upon when the publc observes nformaton. The paper s theoretcal fndngs thus contrbute to the branch of the endogenous nsttutons lterature that seeks to dentfy what types of judcal nsttutons are most valuable n what types of crcumstances (e.g., Glaeser, Johnson, and Shlefer 2001; Aghon, Alesna, and Trebb 2004; Maskn and Trole 2004). Moreover, as we dscuss n Secton III, the model s mplcatons can be used to understand key aspects of some famously controversal judcal rulngs (e.g., Serrano v. Prest, Kelo v. Cty of New London, and In re Marrage Cases). II. Theoretcal Model Assumptons The model consders events over four tme perods. The publc makes decsons usng majorty rule votng, and, except for the court, the publc has no mechansm to overcome tme nconsstency problems. In each perod, when ndvdual members of the publc vote, they act n a forward-lookng manner and seek polces that maxmze ther ndvdual expected benefts. Thus, the publc wll vote to change a polcy f and only f the change would brng more than half of the ndvduals an ncrease n expected benefts. Table 1 presents the order of events over the four tme perods. In perod 1, the publc decdes whether to pass or reject a gven polcy, denoted as polcy, and provdes nstructons to a 4

6 court that wll have the power to overturn some publc decsons wth respect to polcy. To focus our analyss on the optmal role of the court, we assume that the court wll act as the publc s perfect agent n the sense that t wll rule n the manner desred by the perod 1 majorty of the publc. If the publc passes polcy n perod 1, then n perod 2 the publc (whch n some cases wll have more nformaton than t dd n perod 1) makes a majorty rule decson to retan or revoke polcy. In perod 3, the court can overrule the publc s perod 2 choce: If the publc n perod 2 retaned polcy, the court may let polcy stand or overturn polcy, whereas f the publc n perod 2 revoked polcy, the court may let the revocaton stand or renstate polcy. 7 It s worth notng what t means to assume, as we have, that the court wll act as the publc s agent. Gven a court that follows the nstructons t receves from the publc n perod 1, one may thnk of those nstructons as akn to a consttuton, from whch the court wll not devate. And ths, of course, allows the court to serve as a commtment devce. 8 Table 1 also defnes a set of scenaros (Cases A, B, C, D, and E) wth respect to when the publc and the court observe nformaton. In all cases, the values of all parameters (defned below) are known from the begnnng (.e., perod 1). What dffers between cases s () when the vel of gnorance s lfted, meanng when members of the publc learn ther type, and () when a shock wth respect to the benefts of polcy s observed by the publc and the court. In Cases A through D, the 7 Note that our model does not nclude the case when the publc rejects polcy n perod 1, then passes polcy n perod 2. As the reader wll see shortly, ths case could easly ft the model (e.g., suppose the vel s lfted n perod 2 and the majorty then seeks to pass a tyrannzng polcy ). Addng ths case would add lttle to our analyss because we already consder (n Cases A and B) condtons where reversng a perod 1 polcy decson (passng polcy ) would be favored by a majorty of voters n perod 2. Also note that we do not allow the court to enact polcy f the voters have never adopted t; n other words, the court can block or renstate legslaton, but t cannot pass new polces. 8 Our approach thus abstracts from real-world devatons, drven by such thngs as judcal deology, or unclear or napplcable consttutonal gudelnes. 5

7 publc makes ts perod 1 decsons (whether to pass polcy, how to nstruct the court) behnd the vel; naturally, ndvduals behnd the vel are homogeneous. In Case E, however, the vel s lfted n perod 1. In Case A, the shock s observed n perod 1, whereas n Case B, the shock s not observed untl Perod 4, whch s after all the publc and court decsons are made. For Cases C through E, the shock s observed n perod 3 (.e., n tme for the court to make use of the nformaton, but too late to nfluence the publc s majorty rule decsons). Defne For polcy, let v represent the net value (per capta) of polcy, where v = mx + (1-m )y m sze of majorty (.5<m <1) 1-m sze of mnorty x per capta net gans to majorty y per capta net gans to mnorty and x and y have fxed components (x, y ) and a random component (e ): x = x + e y = y + e where e (the shock) s drawn from unform dstrbuton over -e to e, wth e 0. Implcatons To begn, consder some basc ponts about what the perod 1 publc would want the court to do. Recall that each member of the publc seeks to maxmze hs or her own expected benefts. Thus, f the publc s behnd the vel n perod 1 (Cases A through D), the entre perod 1 publc wll desre a court that promotes effcency (.e., polcy that maxmzes Ev ). By contrast, a publc out from behnd the vel may be dvded n terms of what t wants, and members of the majorty wll want a court that behaves n a manner that maxmzes expected benefts to the majorty (Ex ). Hence, 6

8 n Case E (when the vel s lfted n perod 1), the perod 1 majorty wll nstruct the court to act as an agent of the majorty. We wll now consder each of the cases n turn. Cases A and B: Basc tyranny Cases A and B focus on the basc tyranny of the majorty problem. Regardless of whether the shock s observed before the court rulng (Case A) or after the court rulng (Case B), the only valuable role of the court s preventng tyrannes. Because the publc s behnd the vel n perod 1, the publc wll unanmously favor the followng: () passng polcy n perod 1 f and only f the known net benefts (Case A) or expected net benefts (Case B) are postve and () nstructng the court to enforce the perod 1 polcy decson (.e., block any perod 2 polcy changes). In short, there s no way for the publc as a whole to do better than to follow a smple rule: pass (and retan) a polcy f and only f t yelds v > 0 (n Case A) or Ev > 0 (n Case B). 9 Illustratons To see why the court s valuable, consder the followng examples, whch we wll refer to as polcy 1 and polcy 2: Example Polcy 1 m 1 =.6 x 1 = -1 y = 2 1 Example Polcy 2 m 2 =.6 x 2 = -1 y 2 = 2 e = The mathematcal logc s as follows. For Case A, pass polcy f and only f mx + (1-m )y > 0; for Case B, pass polcy f and only f mx + (1-m )y > 0. 7

9 10 Note that polcy 1 s exactly what polcy 2 would be wth an always zero shock. We wll use polcy 1 to llustrate Case A (where the shock s known n perod 1) and polcy 2 to llustrate Case B (where the shock remans unknown untl perod 4). The benefts of polcy 1 n Case A are postve (v 1 =.2), as are the expected benefts of polcy 2 n Case B (Ev 2 =.2). Hence, n ether set of crcumstances, behnd-the-vel voters all favor passng the polcy. Yet a majorty (60%) of the publc wll favor revokng the polcy once the vel s lfted. Thus, the behnd-the-vel publc wll nstruct the court to renstate the polcy n perod 3 f the publc revokes the polcy n perod 2. Knowng ths, the perod 2 publc wll not bother to revoke the polcy. The value of the court s thus as a commtment devce for the behnd-the-vel publc. Case C: Court responds to shocks Case C focuses on the second role of the court: Respondng to new nformaton that takes the form of shocks (e ). There wll be no tyranny of the majorty n Case C, because the publc remans behnd the vel for both perods n whch t makes majorty rule decsons. Thus, the entre perod 1 publc (and perod 2 publc) wll want the court to overturn polcy f and only f the shock s suffcently unfavorable that the socal benefts are negatve (v <0). Because the court can overturn polces n such cases, the publc wll pass polcy n perod 1 f there s any postve probablty that v wll be postve. Of course, the majorty of the publc n perod 4 may dslke the court s rulng: By basng ts decson on socal benefts (v ) rather than benefts to the majorty (x ), the court rulng may overturn polcy when x >0 or allow polcy to stand when x <0. A smple but mportant pont s that there are two categores of polces for whch the court wll beneft the publc. Frst, a gven polcy may usually, but not always, have postve benefts n 10 In other words, an equvalent way to defne polcy 1 s: m 1 =.6, x 1 = -1, y 1 = 2, e 1 = 0. 8

10 the absence of a court. The publc would pass such a polcy even f the court dd not exst, yet have a hgher expected payoff wth a court that prevents outcomes wth negatve socal benefts (.e., 11 prevents v <0). Second, a gven polcy may usually, but not always, have negatve expected benefts n the absence of a court. The publc would not pass such polces n the absence of a court, 12 but wll n the presence of a court, because the court renders the expected benefts postve. Illustraton Consder the followng: Example Polcy 3 m 3 =.6 x 3 = -1 y 3 = 1 e = 2 3 Polcy 3 dffers from polcy 2 only n that t has a lower expected payoff to the mnorty suffcently lower to generate Ev 3 = -.2 for passng polcy 3 n the absence of a court. Yet the expected return of passng polcy 3 s.405 f the court always overturns polcy 3 when v 2< Ths s easy to show formally. If 0 < mx + (1-m )y < e and there s no court, then Ev = mx + (1- m )y > 0, and 0 < P(v <0) < P(v >0). In other words, polcy has postve expected benefts (and hence wll be passed n perod 1), though n some cases the realzed benefts turn out to be negatve. By contrast, n the presence of a court that overrules polcy when v <0, the publc s assured of v 0. Because such a court would turn what would otherwse be outcomes wth v <0 nto outcomes wth v =0, yet would not change the dstrbuton of outcomes wth v >0, the court clearly ncreases expected benefts. Polcy 2 llustrates ths case: Wthout the court, the publc s expected return s Ev 2, whch equals.2; wth the court, the publc s expected return s P(v >0)E(v v >0), whch equals Agan, ths s easy to show formally. If e < mx + (1-m )y < 0 and there s no court, then Ev = mx + (1-m )y < 0, and 0 < P(v >0) < P(v <0). In other words, polcy has negatve expected benefts (and hence wll not be passed n perod 1), though n some cases the realzed benefts would, because of favorable shocks, turn out to be postve. By contrast, n the presence of a court that overrules polcy when v <0, the publc s assured of v Wth the court, the publc s expected return s P(v 3>0)E(v 3 v 3>0). Ths yelds: (1.8/4)(.9) =

11 Case D: Court blocks potental tyrannes and responds to shocks In Case D, the court plays jontly the two roles llustrated separately by Cases A, B, and C. The sole characterstc that dstngushes Case D from Cases A and B s the perod n whch the shock s observed: perod 3 n Case D, rather than perod 1 n Case A or perod 4 n Case B. And the sole dfference between Case D and Case C s when the vel s lfted: perod 2 n Case D nstead of perod 4 n Case C. Identfyng the optmal role of the court s straghtforward because the tmng of events allows the frst-best outcome to be reached. The court observes the shock n perod 3 and, therefore, can base ts rulng on a known value of v. Because the vel s not lfted untl perod 2, the perod 1 publc unanmously desres effcent polcy enforced by court rulngs. The perod 1 publc therefore nstructs the court to approve (or renstate) polcy f v >0 and overturn polcy f v <0. And, of course, the perod 1 publc wll pass polcy f t has any chance of beng valuable that s, f P(v >0) > 0. In short, for any gven polcy, the outcome of Case D wll be () the same as Case C, because n Case D the court blocks potental tyrannes, and () the same as Case A, because n Case D the court weghs shocks n the same manner that the behnd-the-vel publc would n Case A. Case E: Court gnores potental tyrannes yet responds to shocks Case E has the same set-up as Case D, except that the vel s lfted n perod 1. Ths change has major mplcatons for the role of the court. Because the nstructons for the court wll be determned by the already-dentfed perod 1 majorty, the court wll rule on the bass of benefts to the majorty. The outcome s smlar to that of Case C, where the court s only role s to montor shocks, except that wth Case E, the court wll overturn polcy f x <0 (.e., f e <-x ). A tyranny of the majorty wll occur anytme x <0<v or v <0< x. 10

12 Illustratons Agan consder polces 2 and 3. The majorty would nstruct the court to overturn polcy 2 f e 2<1 and to overturn polcy 3 f e 3<1, even though polcy 2 would be effcent f e 2>-.2 and polcy 3 would be effcent f e 3>.2. Thus, -.2<e 2<1 would lead to a tyranny of the majorty, as would.2<e 3<1. Key Implcatons: The Optmal Court The polcy-relevant mplcatons of the model come from a comparson of the cases just 14 dscussed and, n partcular, how they relate to the optmal role of the court. When dscussng Cases A through E, we followed our model s assumpton that a forward-lookng perod 1 publc makes a majorty rule decson when nstructng the court, whch then follows those nstructons. In four of those fve cases (A, B, C, and D), the optmal court behavor s obvous from the analyss we have already presented, because the tmng of the events makes t effcent for the court to rule on the bass of socal benefts (v or Ev ). The most nterestng and less obvous mplcatons relate to the optmal court n Case E. The key queston s ths: When facng the possblty of a Case E tyranny, would t be desrable to have a court that, nstead of followng the majorty s nstructons, was credbly commtted to overturnng neffcent polces (.e., polces that yeld v < 0)? To address ths ssue, 15 we wll retan the assumpton that the court can (and must) commt credbly to polcy. We wll also 14 Recall that our defnton of optmal court behavor s the role that the behnd-the-vel publc would assgn to the court (assumng the court can credbly commt to future behavor). 15 Ths elmnates the possblty that court can rule based on v >0 after successfully convncng the perod 1 majorty that the court wll rule based on x >0. If the court could mslead the publc n ths manner, the frst-best would be attanable. 11

13 assume that drawng the most favorable shock would lead to postve socal benefts (v >0). 16 The optmal court n Case E To dentfy a general soluton for optmal court behavor wth respect to a gven polcy, t fb c fb s useful to defne three cutoffs (e, e, and e *) such that: e >e ndcates the range of e where v >0 c (the cutoff for the frst-best); e >e ndcates the range of e where the majorty wll have a suffcent ncentve to pass polcy (the cutoff mposed by the ncentve constrant); e >e * ndcates the range of e n whch a court followng optmal nstructons (gven the ncentve constrant) wll approve (or renstate) polcy. One can dentfy e drectly from the composton of v : e = -[m x + (1-m )y ]. c fb fb To solve for e, one must determne a lower-end cutoff for the dstrbuton of e such that a unform dstrbuton of e over the range from e to e wll yeld Ex =0. Ths shows: e = -2x - e. The c c soluton for e * s now smple: e * = max (e, e ). Equvalently, ths can be expressed drectly n fb c terms of the parameters: e * = max (-[mx + (1-m )y ], -2x - e ). 17 Illustratons To llustrate the mportance of consderng whether the ncentve constrant places a bndng restrcton on e *, once agan consder our example polces. In some crcumstances (e.g., polcy 3), the sgn of the socal benefts (v ) wll determne the optmal court crteron, and the frst-best 16 Mathematcally: [mx + (1-m )y ] + e > 0. Otherwse, the soluton s obvous: It would be optmal for the court always to block the polcy. 17 fb c c fb fb To see why e * = max (e, e ), note the followng. If e <e, then e *=e, because that wll lead fb c c c to the frst-best outcome. If e <e, then e *=e, because any cutoff below e would lead to zero benefts (because the majorty would not pass polcy ), whle any hgher cutoff would block polcy for shocks n c fb c c some range above e, even though e <e guarantees that all shocks above e generate postve socal benefts. 12

14 fb 18 outcome wll be obtaned wth e*=e. Yet n other cases (e.g., polcy 2), the frst-best cannot be obtaned, and the optmal court crteron wll be determned by the locaton of the ncentve c constrant: e*=e. In these cases, a court that sought always to prevent tyrannes of the majorty fb 19 (.e., usng e as ts cutoff) would not be dong what the publc would want behnd the vel. Moreover, even the mnorty may be better off wth a court that acts based on the nterests of the majorty (decdng to overturn polcy f and only f x <0) rather than on the frst-best crteron (decdng to overturn polcy f and only f v <0). Wth polcy 2, for example, not only would a behnd-the-vel publc unanmously prefer a court that ruled based on x 2 to a court that ruled based 20 on v 2, but lftng the vel would not change that unanmous preference. Put another way, the expected gans to the mnorty, as well as the majorty, can be hgher when the court gnores the mnorty s benefts. Can judcal revew based on a frst-best crteron be worse than no judcal revew? We wll now dscuss a smple extenson of the model. In the precedng analyss, we showed why t may be desrable (from the perspectve of the mnorty and majorty) to have a court that 18 For polcy 3, recall that e3.2 s the cutoff for effcency. When faced wth a court that uses e 3<.2 as the crteron for overturnng polces, passng polcy 3 wll yeld expected gans of.045 for members of majorty. In these crcumstances, the perod 1 majorty wll pass polcy 3. (To fnd the expected return: P(v 3>0)E(x 3 v 3>0) = (1.8/4)(.1) =.045.) Wth polcy 3 passed and the court ensurng v3 0, the frst-best polcy s obtaned as voters behnd the vel would want. 19 Wth polcy 2, a court that uses a frst-best cutoff for overturnng polcy wll yeld an expected outcome worse than what s attanable. The reason s that such a court would overturn polces wth e 2<-.2 (and not otherwse), and ths would yeld an expected return of for members of the majorty. (To fnd the expected return: P(v 2>0)E(x 2 v 2>0) = (2.2/4)(-.1) = ) Hence, n the presence of such a court, the majorty would not pass polcy 2, leavng both majorty and mnorty wth benefts of If the court rules based on the majorty s preferences, the expected return to members of the mnorty s P(x 2>0)E(y 2 x 2>0) = (.25)(3.5) =.875. If the court rules based on socal benefts beng postve (v >0), the return to members of the mnorty wll be zero (because the majorty wll not pass polcy 2). 2 13

15 montors shocks n a manner that allows some majorty-supported neffcent outcomes. Note that those fndngs requred a court that responds to shocks. An addtonal queston to consder s whether the absence of judcal revew mght be better than a court that blocks neffcent polces fb usng a frst-best crteron (.e., e*=e ). The answer s yes, f we relax one of our model s assumptons. Recall that our basc model has shocks affectng the mnorty and majorty n the same manner. In realty, of course, some shocks wll be good news for one group and bad news for another. To account for heterogeneous effects of shocks, we can redefne y as follows: y = y + ëe, where ë s an exogenous parameter. Returnng to Case E, wth ë <0 t s easy to see the potental value of gnorng shocks. The logc s smlar to what we showed earler: A court that approves (or renstates) polcy on the bass of v may cause the majorty to refran from passng polcy. When a gven shock s effect on the majorty s n the opposte drecton of ts effect on the mnorty (.e., when ë <0), court decsons based on the sgn of v may tend to approve polcy when the majorty wants t overturned, and overturn polcy when the majorty wants t approved. And ths can be the case even f a polcy would, n the absence of a court altogether, generate postve expected benefts 21 for both the mnorty and the majorty. Thus, there are condtons n whch the majorty and mnorty, when facng a court that weghs the benefts of polcy to the entre publc, would expect to gan from elmnatng the judcal revew process. 21 To see why, consder Case E wth polcy 4 defned as follows: m 4=.6, x 4=.4, y 4=9.4, e 4=6, ë 4=-4. fb fb fb Ths yelds v 4 = 4 - e 4 and, therefore, e 4 =4, wth e 4 now defned such that polcy s overturned for e 4 <e 4. Ths would lead to polcy 4 beng approved over the range -6<e 4<4, whch s equvalent to -5.6<x 4<4.4. fb Hence, the majorty would not pass polcy 4 n the presence of a court that ruled based on e 4. If nstead there were no court, the majorty would pass polcy 4, because dong so would yeld postve expected benefts for the majorty (.e., Ex 4=.4). And the absence of the court would also yeld postve expected benefts for the mnorty (.e., Ey =9.4). 4 14

16 III. Dscusson Ths secton llustrates how the model apples to real polcy ssues. As a startng pont, we wll dscuss how Cases A and B correspond to basc tyranny of the majorty problems encountered by courts that s, when the court s role s protectng mnortes from majortes. We wll also dscuss brefly how Case C corresponds to common types of montorng roles played by the court that s, when the court s role s to serve as the publc s agent for uncoverng and evaluatng nformaton. Then we wll turn to the most nnovatve parts of our model, n whch courts play a dual role (protectng mnortes, montorng), to show why understandng the dfference between Case D and Case E can llumnate dlemmas faced by courts. Judcal Protecton from Basc Tyrannes (Cases A and B) The quntessental example of tyranny of the majorty s the targetng of easly dentfable mnortes for harsh treatment. As an llustraton, suppose that polcy s a prohbton on enslavng or mprsonng people because of a characterstc such as race or relgon. In ths case, t seems qute safe to assert that polcy would be desred behnd the vel. Hence, a socety may seek to employ some non-majortaran mechansm, such as a judcally-enforced consttutonal provson, to enforce polcy. Yet the adopton of such provsons wll depend on the degree to whch the laws are wrtten behnd the vel. Consder, for example, the U.S. Consttuton. In ts modern form, t has broad provsons to protect ctzens from slavery and arbtrary mprsonment. Arguably these provde, along the lnes of Cases A and B, some measure of protecton for all ctzens each of whom belongs, for example, to a protected mnorty group n terms of relgous afflaton (or nonafflaton). Of course, the orgnal U.S. Consttuton protected a narrower class of people and, most flagrantly, allowed slavery. The ndvduals who desgned the Consttuton were not behnd the vel 15

17 wth respect to ther own race (and other crcumstances) and, hence, wrote laws that protected themselves whle allowng an already dentfed subset of the populaton to reman enslaved. 22 Another mportant example arses when majorty rule decsons have the potental to make polcy reforms tme-nconsstent. Suppose polcy s an effcency-enhancng measure that would generate large socal gans concentrated among an ex ante undentfed mnorty of the populaton perhaps the new entrepreneurs who succeed as a result of the deregulaton or prvatzaton of ndustres. In ths case, the majorty of the publc may favor reversng the polcy once the wnners and losers are dentfed. And that may, n turn, leave voters unable to commt credbly to mantanng, and thus unable to reap the benefts from, effcency-promotng reforms (e.g., Fernandez and Rodrk 1991; Fleck 2000; Falaschett 2008). Under such crcumstances, f a court can block a majorty rule decson to revoke polcy (as n Cases A and B), t can provde a way around the commtment problem arsng from tme-nconsstent majorty rule decsons (e.g., Hanssen 2004). Judcal Montorng n the Absence of Potental Tyrannes (Case C) In contrast to the cases just dscussed, at tmes the entre (or nearly entre) publc has algned objectves but cannot easly observe the way a law s mplemented. Perhaps the most obvous example s that of a potentally valuable, but potentally harmful, power delegated to poltcans. To see how ths fts Case C, let polcy represent the delegaton of, say, spendng authorty and polce powers to elected poltcans. When poltcans use publc funds and/or polce powers for personal 22 Of course, the ssue of protectng mnortes dd not dsappear wth the end of slavery, and Consttutonal protectons of mnorty rghts have not been unformly respected by the courts. One partcularly notorous example s the nternment of Japanese Amercans followng Pearl Harbor. A vast lterature examnes the role of courts durng tmes of crss, ncludng recent cases related to the War on Terror; see, e.g., Epsten et al. (2005). 16

18 gan (analogous to a negatve shock from the general publc s perspectve), vrtually all of the publc would prefer those powers to be revoked or narrowed n scope (perhaps just temporarly). Thus, f courts can uncover (and/or react to) relevant nformaton more rapdly than can the publc, court oversght can be welfare-enhancng whle creatng few, f any, losers among the publc. 23 Shocks and Tyrannes: A Dual Role for Courts (Cases D and E) The man pont here s that the optmal decson rules for real world courts depend on the degree to whch the vel has been lfted when voters (or ther elected representatves) set polcy. That s, do the crcumstances correspond more closely to Case D, n whch the court revews polces for whch the ntal decsons (.e., those made by the publc n perod 1) were made when voters were stll behnd the vel, or do the crcumstances correspond more closely to Case E, n whch the ntal decsons were made after the lkely wnners and losers were dentfed. The real world of judcal revew s, of course, suffcently complex that one cannot classfy rulngs cleanly nto those two categores. Nevertheless, one can gan substantal nsght by applyng the basc logc of Cases D and E to actual court rulngs. To demonstrate ths, we wll consder examples related to equalty n the provson of educaton, the takng of property, and same-sex marrage. Equalty n publc educaton: Serrano v. Prest In Serrano v. Prest, the Calforna Supreme Court mandated an equalzaton of per capta 24 school expendtures across dstrcts n the state. In arrvng at ts decson, the Court emphaszed 23 On the way that the court s ablty to react quckly to new nformaton helps determne ts optmal role, see Maskn and Trole (2004) and Fleck and Hanssen (forthcomng). 24 The Serrano decson was delvered n three parts (n 1971, 1976, and 1977). The 1976 decson (Serrano v. Prest, 135 Cal. Rptr. 345, 1976) had the greatest effect, specfyng that no dstrct could dffer from another by more than $100 per student. See, e.g., Fschel (1989, 1996). Although the Serrano decson s the most famous of ts knd, courts n other states have ssued smlar rulngs. As Husted and Kenny (2002, 17

19 th equal protecton under the 14 Amendment of the Unted States Consttuton, whch t took to be 25 substantally equvalent to clauses n the Calforna state consttuton. In order to frame ths rulng n the context of our model, the crtcal queston s what polces were at stake; that s, what polces dd the Court strke down, and what new polces were mplemented as a result? Our model suggests two possble nterpretatons, and comparng them helps to explan why the Serrano decson has been so controversal. One possble nterpretaton s that the Court enforced a behnd-the-vel law (.e., equal protecton) that unequal school spendng volated, and that by dong so t moved polcy toward what a behnd-the-vel publc would want. To see ths, suppose that the behnd-the-vel allocaton of educaton servces gven the shock that made the returns to educaton large n the modern era would be hgh qualty and egaltaran, yet the ex post majorty would dvde the budget unequally, 26 yeldng hgh qualty educaton for the majorty, and low qualty for the mnorty. To llustrate ths 567) report, A thrd of the states educaton fnance systems now have been ruled unconsttutonal. 25 Although t s not central to our argument, one mght ask why the Serrano decson came when t dd, rather than earler (especally gven that the Equal Protecton Clause was not new). One contrbutng factor s that the value of homes skyrocketed n a large number of neghborhoods, ncreasng dfferences n levels of fundng across dstrcts. Fschel (1996, 625) wrtes, Calforna housng prces exploded there s no better word for t durng the 1970s.... No manland state had such an enormous ncrease n sngle famly home values durng the 1970s. Fschel states that the value of sngle famly homes n Calforna rose from 27 percent and 35 percent hgher than n the rest of the U.S. n 1960 and 1970, to 70 percent hgher n In the context of our model, the court may have been, n essence, reactng to a shock (to home prces) that ncreased the behnd-the-vel value of mandatng the redstrbuton of property tax revenue between dstrcts for the purpose of equalzng school fundng. 26 Note that the shock related to the returns to educaton pertans to new nformaton revealed subsequent to the ncorporaton of equal protecton nto the Consttuton. Ths matters for applyng our model because, f the tmng of events corresponded to Case A (.e., the value of educaton n the future s known when a consttuton s wrtten), the publc would smply wrte a behnd-the-vel law specfyng unformly hgh educaton qualty, and nstruct the court to overturn any devaton from that provson. What the rsng mportance of educaton as a publc polcy ssue means for modern courts s that they must nterpret a consttuton that was wrtten before publc school budgets became a major polcy ssue large scale publc th fundng of educaton dd not come untl the 19 century. On the reasons for the rse of publc fundng n the 18

20 more formally, let polcy j represent unequal shares of a fxed budget, so that a court rejecton of polcy j would lead to an equal dvson of that fxed budget. If ths descrbes what was at stake n Serrano, then the optmal court would have based ts rulng on a frst-best crteron (as n Case D) and, hence, dsallowed unequal shares essentally what the court dd. But now consder that, n realty, the educaton budget s not fxed and, furthermore, that equal protecton does not allow a court to set specfc levels of spendng. These factors pont to another possble nterpretaton of the Serrano decson. Let polcy k represent a hgh level of spendng for majorty students and a low level of spendng for mnorty students. If ths descrbes what s at stake n Serrano-style applcatons of equal protecton, then rulngs based on frst-best crtera may be sub-optmal (as n Case E). The key pont s that, wthout assumng a fxed budget, t s not clear that rejectng polcy k (even though t reflects a tyranny) wll mprove the outcome for the mnorty. Put another way, even f courts can rely on equal protecton (whch can be vewed reasonably as a behnd-the-vel law) to guarantee relatvely equal budget shares, voters wll set (or choose poltcans who set) the level of the budget after the vel s lfted. Thus, f the majorty knows that anythng wth unequal fundng (such as polcy k) wll be overturned as a volaton of equal protecton, t may pass nstead a polcy wth a small, equally dvded budget. In vew of these two contrastng nterpretatons of Serrano, what does the evdence say? The economcs lterature suggests that Serrano dd ndeed harm the group that the court sought to help (chldren from poorer famles), because t caused an overall declne n educaton fundng (e.g., Fschel 1989, 1996, 2008; Downes 1992; Slva and Sonstele 1995). The argument fts our Case E: Because the majorty of the electorate responded ratonally to ts knowledge of how the court would th 19 century, see Stoddard (2009). 19

21 rule, the educaton budget fell relatve to what t otherwse would have been, and so dd the overall (mnorty as well as majorty) qualty of K-12 educaton n Calforna. 27 Judcal constrants on takngs: Kelo v. Cty of New London The takng of a mnorty s property s a standard example of tyranny of the majorty, and t s thus easy to see why a behnd-the-vel publc would seek to constran such actons. Ths s, of course, an mportant reason why the U.S. Consttuton allows the use of emnent doman only when 28 property s taken for a publc use and just compensaton s pad. Yet despte these 29 consttutonal lmtatons, there has always been controversy over the use of emnent doman. The debate (and wdespread furor) over the most recent major rulng Kelo v. Cty of New London 27 Hamlton (1975) shows that, gven Tebout sortng (Tebout 1956), local property taxes and local zonng ensure the provson of the optmal amount of congestable government-provded goods, such as schools. In other words, property taxes are equvalent to a fee for servces. By strctly lmtng the ablty of dstrcts to use local property taxes to provde local publc goods, Serrano undercut support for property taxes (and thus spendng on schools). A number or scholars Wllam Fschel most promnently have suggested that Serrano, by breakng the lnk between local property taxes and spendng on local schools, nspred Calforna s 1978 Proposton 13, whch greatly reduced spendng. Fschel attrbutes the popularty of Proposton 13 to the second Serrano opnon (Fschel 1989, 1996, 2004a, 2008). For a counter to Fschel s argument, see Stark and Zasloff (2003), and see Fschel (2004a) for a response. Slva and Sonstele (1995) conclude that Serrano was responsble for about 50 percent of Calforna s relatve (to the rest of the country) drop n per student spendng. Fnally, t s worth notng that the overall declne n school qualty does not appear to be just the prce pad for desrable mprovements n low-end student performance: Downes (1992) fnds that sxth grade achevement test scores were as unequally dstrbuted n as they had been n th The 5 Amendment to the U.S. Consttuton contans the phrase nor shall prvate property be taken th for publc use, wthout just compensaton, and the 14 Amendment states that U.S. ctzens cannot be deprved of lfe, lberty, or property wthout due process of law, whch effectvely extends the 5 th th Amendment s protectons to actons by ndvdual states (the 5 Amendment was ntally nterpreted as applyng only to the federal government). Most state consttutons contan smlar wordng, wth some statespecfc elaboratons. 29 On the use of, and controverses over, emnent doman n the early Unted States and the colonal era, see Hart (1995, 1996a, 1996b). Fschel (1995, 2004b) and Somn (2004) examne more recent controverses. Fleck and Hanssen (forthcomng) provde an overvew of the recurrng controverses over, and the changng judcal nterpretatons of, emnent doman laws. Also see Epsten (1985) and Hermaln (2005). 20

22 provdes a good llustraton. In Kelo, the U.S. Supreme Court ruled (5 to 4) for the cty of New London, thus allowng the use of emnent doman to take non-blghted homes n order to provde 30 land for prvate development. Our model suggests three possble nterpretatons of the rulng: two that ft Case D and one that fts Case E. The frst nterpretaton s that even those harmed through the loss of ther homes mght have (n prncple) supported the redevelopment plan behnd the vel. Ths scenaro fts Case D of the model, wth the Kelo decson beng the effcent one. In other words, the socal benefts from usng emnent doman may have been postve. A few homeowners were harmed and, hence, complaned but only after the vel was lfted. The second nterpretaton s that the use of emnent doman n the Kelo case corresponds to a basc tyranny, also as llustrated n Case D. Suppose polcy allows the use of emnent doman n the manner employed by New London, and the shock observed for New London n partcular 31 rendered the overall benefts (v ) negatve. From ths perspectve, t s at least arguable that the Court made a mstake by falng to protect a tyrannzed mnorty from neffcent polcy. And, ndeed, Justce O Connor s Kelo dssent emphaszed her concern about the manner n whch the poltcal system can be expected to wegh the gans and losses from redevelopment projects: The benefcares are lkely to be those ctzens wth dsproportonate nfluence and power n the poltcal process, ncludng large corporatons and development frms. As for the vctms, the government now has lcense to transfer property from those wth fewer resources to those wth more. 30 Susette Kelo, et al., v. Cty of New London, Connectcut, et al., 545 U.S (2005). 31 Of course, n the real world, nether the mnorty nor the majorty would be homogeneous n terms of recevng benefts, but those who stood to lose would be few (those forced to sell ther homes at below ther reservaton prces), and the potental benefcares, at least accordng to the promoters of the development project, would be many (e.g., f the project generated tax revenue for local use). 21

23 Under the condtons O Connor descrbes, one can easly see how, n the absence of a judcal constrant, the proponents of neffcent redevelopment polces (v <0) that harm mnortes (y <0) would preval. 32 The thrd nterpretaton whch fts Case E n our model s that, even f the undesrablty of the redevelopment polcy under revew n the Kelo case were obvous, a judge could ratonally oppose overturnng that polcy. If the use of emnent doman n the Kelo case generated negatve socal benefts for reasons that could not have been perfectly foreseen (.e., because of a shock), but smlar uses of emnent doman would (n the presence of more favorable shocks) be benefcal, then t may be undesrable for a court to decde each case usng a frst-best crteron. In other words, consstently dsallowng emnent doman when observng negatve socal benefts (v <0) could lead to lower socal benefts n the long run (by dscouragng valuable, as well as harmful, applcatons of emnent doman powers). Does the basc logc of ths thrd nterpretaton (Case E) have practcal relevance to the Kelo case? Qute plausbly yes. The key pont here s that even some supporters of the Kelo decson beleve that the New London s use of emnent doman was a mstake. Most notably, Justce Stevens, who wrote the majorty opnon n Kelo, subsequently expressed the vew that the Kelo 33 redevelopment plans were undesrable. Nevertheless, he voted wth the majorty. Thus, Stevens 32 The key ssue here s that a mnorty bears the costs. For the mnorty to be tyrannzed, t need not be that a majorty actually obtan apprecable benefts. The promoters of redevelopment do, of course, generally clam to delver wdespread gans, but regardless of how exaggerated those clams are, an obvous concern would be a poltcally nfluental group of expected wnners (along wth an ndfferent majorty and a harmed mnorty). 33 As Greenhouse (2005) reports: Hs own vew, Justce Stevens told the Clark County Bar Assocaton, was that the free play of market forces s more lkely to produce acceptable results n the long run than the best-ntentoned plans of publc offcals. But he sad that the planned development ft the defnton of publc use that, n hs vew, the Consttuton permtted for the exercse of emnent doman. 22

24 opposed the Court s makng an attempt to enforce what was accordng to hs own statements the frst-best outcome. Although Case E of our model s not the only reason to oppose the use of a frst-best crteron (another s respect for precedent), t does provde a practcal justfcaton: When the behavor of a ratonal, forward-lookng publc makes the frst-best nfeasble, the optmal court may allow the majorty of the publc to harm the mnorty through socally undesrable polces. Ths s a reason for courts to follow an nter-temporally consstent nterpretaton of publc use (as Justce Stevens argues the court should), even f t requres courts to allow emnent doman to be employed on occasons when t generates negatve socal benefts. That sad, an essental pont to consder s whch members of the publc reman behnd the vel. Recall that the Cty of New London fought the case all the way to the Supreme Court, suggestng that the majorty of New London voters supported (or at least dd not strongly oppose) the proposed redevelopment. The reason, of course, s that voters n New London were already out from behnd the vel wth respect to whose homes were to be taken and only a small mnorty stood to lose ther homes. By contrast, voters n the rest of the country were generally behnd the vel wth respect to smlar (possble) plans n ther own communtes. How dd the behnd the vel voters 34 react? In an overwhelmngly negatve fashon. The result was substantal pressure to lmt emnent 35 doman powers. In terms of our model, the publc responded to Kelo by curtalng polces that 34 As Cole (2006, 1) puts t, n June 2005 the U.S. Supreme Court decded the case of Kelo v. New London Development Corporaton and all hell broke loose. He then dscusses (and catalogues) the publc reacton. On the publc and legslatve reacton to Kelo, also see, e.g., Bell (2006) and Somn (2009). 35 In short, even mddle class homes n pleasant neghborhoods were not safe from governmentsponsored redevelopment, and nether majorty rule nor the courts post-kelo could be counted on to do what the majorty would choose whle behnd the vel. Thus, current and prospectve homeowners who reman behnd the vel (n practce, the vast majorty of the publc) favor revokng cty governments power to nterpret publc use expansvely when takng prvate homes. For a dscusson of the way Kelo fts nto the long debate over the proper role of emnent doman, see Fleck and Hanssen (forthcomng). 23

25 would be benefcal only n the presence of a court that prevents tyrannes of the majorty. In other words, the publc has reacted as f the Court faled to prevent a Case D tyranny. Equal protecton and same-sex marrage: In re Marrage Cases To see how the ssue of same-sex marrage fts our model, consder frst that one can plausbly vew prohbtons on same-sex marrage as an example of tyranny of the majorty. More specfcally, one could make the followng argument: A behnd-the-vel publc that s, before ndvduals know whether ther desred marrage partners wll be of the same or dfferent sex would favor laws that allowed marrage for same-sex couples as well as for opposte-sex couples. But after the vel s lfted, a majorty (composed prncpally of heterosexuals) may seek to block marrages for same-sex couples. To understand our argument here, t s essental to recognze that our purpose s not to evaluate the merts of same-sex marrage, but to llustrate how Cases D and E of our model can be appled to the real world. For the llustraton to be useful, one need not accept the vew that a behnd-the-vel publc would favor gay rghts, but only that such a vew could motvate a court to conclude that restrctons on marrage volate equal protecton. The key queston s thus whether a court that vews prohbtons on same-sex marrage as a tyranny of the majorty would be helpng the cause of same-sex marrage f t overturned prohbtons on such marrages. Our model shows why the answer s not obvous. Consder the hghlghts of how the ssue has played out n Calforna, the most promnent battleground. In 2000, the majorty (61%) of Calforna voters passed Proposton 22, whch banned same-sex marrage. Subsequently, n ts 4-3 In re Marrage Cases (2008) decson, the Calforna Supreme Court found Proposton 22 unconsttutonal. If ths were the end of the story, the Court would arguably have appled a behndthe-vel law (equal protecton) to prevent a mnorty (supporters of same-sex marrage) from beng 24