The Advantage of Second Chambers in Republican Legislatures: An Informational Theory. James R. Rogers, Ph.D, J.D.

Transcription

1 The dvantage of Seond Chambers in Repblian Legislatres: n Informational Theory James R. Rogers, Ph.D, J.D. Department of Politial Siene Texas &M University College Station, TX ( rogers@polisi.tam.ed bstrat: iameralism initially arose in systems of mixed government in whih the two legislative hambers were ontrolled by different soial estates with differing sbstantive interests. The rise of repblianism in the modern era raises a theoretial problem for biameralism: whether seond legislative hambers are neessary when both hambers are ongrent, representing the same set of eletors and interests. game theoreti model of biameral deision-making is developed that identifies an informational welfare advantage for ongrent biameralism: Two-hamber legislatres are more apt to adopt the best stattory means relative to a given end than are one-hamber legislatres. In short, biameralism prodes better legislation than niameralism. Importantly, this advantage is demonstrated when the two hambers have ongrent preferenes. The theory ths identifies a jstifiation for repblian biameralism. Grant spport for this projet is grateflly aknowledged from the TMU Sholarly Enhanement Program, the Nebraska Center for Pbli Poliy, and the merian Politis Program of the sh Shool of Government and Pbli Servie. n earlier version of this paper was presented at the 998 nnal Meeting of the merian Politial Siene ssoiation.

2 Sine the period of the earliest U.S. state onstittions, the jstifiation for seond legislative hambers in the repblian governments of the states has proven problematial. To be sre, the olonial governments had a form of biameralism and in the shift from olony to state eah of the U.S. states maintained or soon adopted biameral instittions. The otward niformity of the new state legislatres with the old olonial legislatres, however, masks the fndamental shift in onstittional theory that needed to or dring this era to jstify the ontined existene of seond hambers. The old olonial governments had attempted to mirror, however dimly, the mixed government of the ritish onstittion in whih the three soial estates of the realm the king, the nobility, and the people provided heks and balanes on the power of eah other. This was refleted in parliamentary biameralism with one hamber devoted to the people and the other to the nobility. This was also refleted in merian olonial government with a poplar assembly for the people and a onsel whih was spposed to represent the king s interests and to be peopled by nasent merian aristorats. With the Revoltion, however, merian politial instittions beame thoroghly repblianized. Why then ontine to have two hambers when both simply represent the same set of eletors and, hene, were expeted to experiene a high degree of ongrene over legislative otomes? If biameralism was to ontine, it needed a new jstifiation. While this need for a distintly repblian jstifiation for two legislative hambers has been mh stdied and domented by modern historians of the Revoltionary era (Wood 969, 97-55, Krman 997, -54, f., Greene 994, it has been all bt ignored in modern attempts by politial sientists to aont for the vale of seond legislative hambers. Instead, modern biameral stdies identify ontined jstifiation for seond hambers only when the sbstantive poliy preferenes of the two hambers diverge. While this argment jstifies biameralism in mixed regimes, it is largely irrelevant to repblian biameralism and to the legal and politial evoltion of highly ongrent legislative hambers in U.S. states. This paper seeks to fill this gap in biameral theory by identifying a rationale for biameralism in repblian politial systems that is, to provide a jstifiation for seond hambers that does not rely on Georgia, Pennsylvania, and Vermont (ntil 86 initially had niameral legislatres (arnett 95, Morey 89, Senning 97.

3 there being any sbstantial divergene between the two hambers in primitive poliy preferenes. The paper is divided into for setions. Part I dissses the rise of the biameral problemati in repblian governments with a fos on the experiene of U.S. states. Part II then sets ot a simple ommon vale/information model of biameral deision-making. Part III examines the behavior between biameral hambers that the model implies. Part IV derives the welfare impliations of ongrent biameralism from the eilibrim behavior identified in Part III. Part V onldes the paper. Part I. Politial and Sholarly Interest in iameralism iameralism evolved historially as an instittional refletion of fndamental politial and lass leavages (Lijphart 984, 95, Longley and Oleszek 989, 4, Tsebelis and Money 997, -. In its most elebrated historial form the mixed government of Great ritain nobles were alloated one hamber and ommoners were alloated the other. ease the two hambers represented different estates, eah needed the protetion of a legislative veto against measres that wold advane one estate s interests at the expense of the other. In extolling the ritish example Montesie ths wrote that nobles mst form a body that has a right to hek the lientiosness of the people bt that the people mst also have a hamber to oppose any enroahment by the aristoray (748/949, 55. The theory of mixed government had ome to merian shores at the hands of the ritish olonists (Greene 994, f., Wood 969 and Krman 997. Upon separation from ritain, however, the traditional, mixed government rationale for biameralism faed a major theoretial problem in appliation to the new U.S. states: the nderlying soial estates neessary for a mixed regime did not exist. The new states did not have two soial estates, let alone three. Rather, the homogeneity of orders in the United States (Wood 969, 7 implied two homogenos branhes in state biameral legislatres shold both hambers share the same set of eletors (ibid., Wood onldes, The people in the new states seemed to be eleting the same kinds of persons to both hoses of the legislatres, ths reating a homogeneity of interest between the two branhes and destroying the prpose for institting a mixed polity (ibid., p. 6. Jefferson (784/987, for example, had objeted to the strtre Virginia senate bease, being hose by the same eletors it was too homogenos with the hose of delegates when

4 the whole prpose of biameralism was to introde the inflene of different interests or different priniples (65. Historian Mar W. Krman similarly onldes: Many revoltionaries dispensed altogether with the idea of mixed government and envisioned a senate simply as an instittional restraint on the power of the representatives. These men based their views on the assmption that, in a repbli, there old be only one interest that of the people. To organize government arond onfliting estates wold enorage and even bless politial disagreement.... Those biameralists, as well as men who spported a single assembly, envisioned an ndivided, harmonios polity. Having no rank abot that of freemen, the essayist Sals Popli delared, she has bt one interest to onslt, and that interest... is the tre and only interest of men as members of soiety. (997, 45 Three argments for the ontination of biameralism sbseently developed among merian onstittionalists. First, that two hambers, even if sharing the same set of eletors, serve as a hek pon the other and ths rede the probability of legislative tyranny (Wood 998, Seondly, in the ontext of the U.S. national government, a dissimilarity in... genis between Hose and Senate, as Madison ommended, wold promote stable poliy otomes (The Federalist No. 6, 79. Finally, almost alone in ommending the vale of ongrent hambers, James Wilson arged that biameral deision-making proded an informational advantage for the legislative proess relative to niameral deision-making. We briefly onsider rationale in trn. First, while legislative tyranny was an ndobted fear for early merians, the modern ase for biameralism annot really trn on it. U.S. states now have a long tradition of repblian politis and enjoy a nmber of instittional safegards that mitigate modern fears of legislative tyranny: the Forteenth mendment provides the Federal jdiiary a negative on nreasonable state legislation, exetives wield the veto in greater proportion than originally antiipated, and the national government retains the ommission to garantee to eah state a Repblian Form of Government (U.S. Const. art IV, se 4. Frther, as an empirial matter, Nebraska has had over sixty years of experiene with niameralism and has yet to manifest any nie penhant for legislative tyranny relative to the 49 biameral U.S. states. Ths, it does not appear that the fear of legislative tyranny is a ompelling reason to maintain biameral legislative systems in modern repblian regimes. That biameralism promotes stability is the jstifiation most often advaned in the modern 4

5 literatre on biameralism for the instittion. Relying on soial hoie models of biameral deisionmaking, one set of modern sholars has arged that biameralism shold be maintained even in modern repblian polities bease it promotes poliy stability (Hammond and Miller 987, 989, Levmore 99, Riker 99a 99b, Hammond, Miller, and Kile 996, Tsebelis and Money 997, f., hanan and Tllok 96. Stability in this literatre is nderstood both in its traditional Madisonian sense and in the more speialized sense of reding the domain of voting yles. There are several weaknesses, however, with this jstifiation for biameralism. First, and most importantly, stability does not have nambigos welfare impliations. The literatre itself often reognizes this. For example, Tsebelis and Money observe: [G]reater stability (in the sense of preserving the stats o and fos of onflit along one dimension are the main featres (both advantages and disadvantages of biameralism. If an instittional designer desires poliy stability... biameralism delivers.... On the other hand, if rapid hange is desired, a onstittional designer wold be better off avoiding biameral... instittions (7. In these models, preferenes over ameral hoie are simply proxies for preferenes over stats o poliies. Conservatives prefer biameral proesses bease it inreases the diffilty of adopting new legislation; liberals prefer niameral proesses bease it inreases the ease of adopting new legislation. Hammond and Miller note the same problem (987, 70, as does Madison (The Federalist No. 6, 78. Indeed, in the ratifiation debates srronding the U.S. Constittion, biameral opponents agreed that biameralism promoted stability, bt arged that was a reason to oppose the instittion. The stability-inding ality of biameralism was also a main argment Progressive-era niameralists asserted to indit the instittion. seond problem with the soial hoie ase for biameralism is that the analyses have largely ignored the impliations of reoniliation mehanisms for the stability For example, the psedonymos Repblis arged in 788: re [seond hambers] not as likely to hek a good bill as a bid one? and has it not in fat often happened? (788/987, 7. Orfield (98 arged that the present system reslts in too mh hek and balane. For every poor measre that may be defeated nder the biameral system, it is likely that two or more good measres fail. In times of depression, this may reslt in lak of ation at ritial moments with something approahing anarhy ensing (-. Poplist U.S. Senator George Norris arged that speial interests, orporations, and monopolies se biameralism to prevent legislation (Norris 95a, 54, ibid. 95b, 65. In the 9 th entry, Jeremy entham arged similarly that biameral stability was one of the instittion s disadvantages (Rokow 98. 5

6 onlsion. 4 Representatives from both hambers gather in the niameral body of the onferene ommittee; an agenda setter in one hamber an ditate the agenda in the other hamber throgh motioning. Ths it is not lear jst how mh biameralism redes the size of the domain over whih majority yling may or (Levmore 99, 48-49, f., Tsebelis and Money 997, 6. One ommon onlsion in these attempts to identify stability as the modern rationale for biameralism is that the instittion generates a legislative advantage only if the hambers differ signifiantly from one another. Hammond and Miller find that The stability-inding properties of biameralism are... dependent on the existene of distintly different viewpoints in the two hambers (989, 9, f., ibid. 987, 60. hanan and Tllok onlde similarly that, nless the bases for representation are signifiantly different in the two hoses, there wold seem to be little exse for the two-hose system (96, 6. Differenes in hamber preferenes are ritial to generating Tsebelis and Money s stability reslts (997, 89, f., 9, 69, ease two ongrent hambers wold ostensibly not signifiantly affet poliy otomes, Lijphart desribed biameral systems with ongrent hambers as weak forms of biameralism (984, nd Deler (977 smmarized the onventional wisdom when he wrote: The ase for niameralism beomes sbstantially stronger if one an find little empirial distintion between the two hoses. If, in fat, little inter-hose variation exists, why maintain two hoses that merely delay poliy making? If sbstantial differenes are fond, a neessary (albeit not a sffiient ondition for maintaining biameral legislatres is met (775. Of orse, modern sholars are not alone in making this onlsion. s noted earlier, both Madison and Jefferson expliitly regarded dissimilar hambers as a neessary ondition for realizing a biameral advantage. James Wilson, however, artilated a deidedly minority position among early onstittional ommentators in identifying an informational welfare rationale for the ongrent hambers of repblian 4 While Tsebelis and Money devote a sbstantial amont of their analysis to reoniliation mehanisms, their soial hoie model of biameral ation in a mlti-dimensional setting abstrats away from the reoniliation mehanisms that they later analyze, as well as the onditions nder whih biameral hambers will hoose to employ a speifi reoniliation mehanism or hoose not to reonile at all. 5 Deler (977 smmarized the onventional wisdom when he wrote: The ase for niameralism beomes sbstantially stronger if one an find little empirial distintion between the two hoses. If, in fat, little inter-hose variation exists, why maintain two hoses that 6

7 biameralism. He arged in his Letres on Law: [Many] reasons... may be assigned, why all the advantages, to be expeted from two branhes of a legislatre, may be gained and preserved, thogh those two branhes derive their athority from preisely the same sore.... [These inlde a] doble sore of information, preision, and sagaity in planning, digesting, omposing, omparing, and finishing the laws, both in form and sbstane (79/987, 78, emphasis added; f., Story 8/987, 80. The analysis below piks p on this minority insight and demonstrates that both the old and new literatre on biameralism has erred in onlding that divergent hambers are a neessary ondition for a biameral advantage. It shows this nder the ondition that both traditional and modern sholars have onsidered to be the weakest form of biameralism: when both hambers are ongrent. The argment over the existene of a benefit to a seond hamber in repblian politial systems is more than of aademi interest both in the U.S. and world-wide. While Nebraska is the only U.S. state rrently with a niameral legislatre, 6 the 990s have witnessed a level of interest in the reform rarely seen sine the 90s. 7 t least forteen states have offiially onsidered niameral proposals in reent years. In 997 legislative sessions, state legislators sponsored niameral proposals in laska, 8 Connetit, 9 Hawaii, 0 Iowa, Minnesota, New York, Soth Dakota, 4 and Wisonsin. 5 Florida s onstittional revision ommission narrowly defeated a 997 measre to eliminate one legislative hamber. 6 Dring 995 sessions, legislators introded measres proposing niameralism in Maine, 7 Massahsetts, 8 Pennsylvania, 9 and Vermont, 0 and legislation proposing the reform was introded in merely delay poliy making? If sbstantial differenes are fond, a neessary (albeit not a sffiient ondition for maintaining biameral legislatres is met ( Georgia, Pennsylvania, and Vermont (ntil 86 initially had niameral legislatres (arnett 95, Morey 89, Senning Nebraska s niameral legislatre held its first session in 96, after adopting the reform in a 94 initiative measre. Twenty-one state legislatres sbseently onsidered niameral proposals in their 97 legislative sessions (Johnson 98, H.J.R., 0 th Leg., st Sess. (K H.J.R. 5, 997 Reg. Sess. (CT 997; S.J.R. 6, 997 Reg. Sess. (CT S.. 60, 9 th State Leg. (HI 997. S.J.R. 7, 77 th Gen. ssembly, st Sess. (I 997. S.. 4 d, 998 st Spe. Sess. (MN 998; H.. 57, 80 th Reg. Sess. (MN 997; S.. 59, 80 th Reg. Sess. (MN 997. S.. 89, 0 th nnal Leg. Sess. (NY H.J.R. 006, 7d Leg. ssembly, 997 Reg. Sess. (SD J.R. 46, 9d Reg. Sess. (WI Proposal No. 77, CRC -0-pr, Fla. Const. Rev. Comm ( The plan was defeated 9-4 (Elmore S.. 658, 7 th Leg., d Reg. Sess. (ME 995. The proposal was defeated in the Hose 66- and (Higgins H , 80 th Gen. Ct., 996 Reg. Sess. (M H.. 8, 79 th Gen. ssembly, Reg. Sess. (P PR 05, Leg. Sess. (VT

8 the California legislatre in 99. California s Constittion Revision Commission initially adopted a niameral reommendation in 995 (Stall 995, bt failed to send it to the legislatre after several ommissioners hanged their votes (San Franiso Chronile 996. Interests extends beyond the borders of the U.S. as well. While the biameral form predominates among U.S. legislatres, the opposite is the ase world-wide, with 6 niameral and 56 biameral national legislatres (Tsebelis and Money 997, 45. There was a modest national trend toward niameralism after WWII, with New Zealand adopting the reform in 950, Denmark in 95, Kenya in 966, and Sweden in 970 (Longley and Olson 99. This trend reversed arond the time of the Soviet ollapse. Poland (P 989, Kyrgyzstan (Nash 994, elars (in 997, and the breakaway Dnestr Moldovan Repbli (in 995 replaed ommnist-era niameralism with biameral instittions. lbania s rling Soialist Party also annoned on Jly 0, 998 that both niameralism and biameralism wold be onsidered for the parliament s new strtre (C 998. eyond the former Soviet alliane, Haitian voters hose biameralism over niameralism in a 987 referendm, as did Moroan and lgerian voters in 997. Chinese offiials have also onsidered adding a seond legislative hamber, and Canada is onsidering reinvigorating its torpid pper hamber (Longley and Olson 99,. Ths, aonting for the existene and natre of a biameral advantage is a matter of pratial as well as sholarly interest. Part II. Desription of the Model stin-smith and anks (996 reently arged that information theoreti models potentially imply misleading reslts nless modeled within a game theoreti ontext. The information theoreti model of biameralism developed below ths sets the informational hoies of the ators in a game theoreti model. Ths, while the information theoreti aspets of the model developed below are oneptally very simple, given the nmber of possible deision periods and the nmber of possible deisions in eah period, the game theoreti model is nfortnately notation dense. We now trn to the model..c.. 4, Reg. Sess. (C 99; S.C.. 8, Reg. Sess. (C 99. 8

9 There are three possible ators in the model: hamber, hamber (whih always at, and a onferene ommittee (, whih may or may not be alled. I = {,, } is the set of instittional ators. The hambers onsider two versions of a bill on the same sbjet matter, l in one of two states (s of the world withot onfsion named and, s L = {, }. Eah bill is best S = {, }. For simpliity, either state of the world ors a priori with eal probability, pr(s = = pr(s = = 0.5. The hambers and the onferene ommittee share ommon preferenes over the otomes, desribed by (. P ( l s ( l s ( 0 ( l s ( l s P where P R +. dopting law in state is better than adopting no bill, whih is better than adopting law in state. The payoff strtre noted in relationship (. expliitly assmes a ommon vale held between the hambers. While this makes the analysis easier to develop relative to the alternatives, it is important to stress that the ommon vales assmption is sbstantively motivated preisely bease it aptres the key problemati of repblian biameralism that of homogenos preferenes aross the hambers. The theoretial problem posed by repblianism for biameral instittions is preisely that of whether a seond hamber is sefl when hamber preferenes are ongrent. The hambers may se different reoniliation mehanisms, sh as motioning and onferening. There is a ost to employing a reoniliation mehanism. Motioning ( a bill bak to the originating hamber osts eah hamber R + and onferening a disagreement osts R +, with < (Longley and Oleszek 989, 68. These are opportnity and transation osts and may vary depending on the point the hambers are at in the legislative session (ibid.. The ost of eah reoniliation mehanism old be assmed to differ between the hambers, bt no sbstantive reslt depends ritially on assming eal osts aross the hambers and it does provide some eonomy in deriving reslts. The hambers and the onferene (if held observe a signal that is private to the instittion and a pbli signal ommniated by the ations of instittions in previos periods. eliefs abot the tre state of the world are derived from these signals. The natre of the private signals reeived by eah hamber This assmption an be relaxed withot signifiantly altering the reslts derived below, bt it provides a great deal of eonomy. 9

10 and the onferene (if held is disssed first. Derivation of the eilibrim impliations of the model draws on several different probability measres (f., Hirshleifer and Riley 99, egin with: (. i t m s, = pr(m s, m i M = {a,b}, s {,}, i {,, }, t {,,, 4}. This expresses the onditional probability that instittion i reeives private message m given state s in period t. Eah instittion that holds a session observes a private signal, { a, b}, i {,, }, that is i, t i, t a b orrelated with the tre state of the world, 0.5. Similar reslts to those proven below old i, t be derived withot reiring that a i, t b m i, bt the eality does permit more eonomial proofs and i, t i, t a b permits the onditional probabilities to be abbreviated by i =. In general, i j for i j, i, j {,, }. The vale of i for i {,, } is assigned in period zero. While the ontent of the message is private knowledge, the orrelation vale of eah instittion s signal with the tre state of the world is ommon knowledge. Other sefl probability measres are: (. (.4 (.5 (.6 i t m, = the nonditional probability of i {,,} reeiving message m in period t. i t s, = the nonditional (prior probability of state s for instittion i in period t. i t m s j,, = the joint probability of state s and message m for instittion i in period t. i t m s, = the onditional (posterior probability of state s, given message m in period t for instittion i. s noted, hambers reeive signals from the other hamber (in the form of a bill or from the onferene ommittee (in the form of a bill reommendation. Instittions ating in periods, and 4 observe both their private signal and the ations taken by instittions in earlier periods. The ombination i, t i, t of private and pbli signals the instittions reeive is represented by. These signals are detailed below for eah stage of the game. The expansion of the message set implies a straightforward translation of relationships (. throgh (.6. For example, i t, i, t wold be the nonditional probability of i {,, } reeiving message i,t in period t; i, t s i t, wold be the onditional probability of state s, 0

11 given message i,t in period t for instittion i, et. ease the indexing spersripts on i,t only repeat the indexing spersripts on the probabilities, they are sppressed. Ths, for example, i, t i, t i, t i, t i, t, i, t s, et. s Let i t v, be the poliy ation (or vote deision hosen by instittion i in period t, with i t V, = {,} denoting the set of feasible poliy deisions for i {,, } in period t {,,, 4}. In addition to voting in period, hamber also selets a reoniliation option, = {,, }, from a set inlding a onferene ommittee (, motioning (, or voting on the proposal withot a reoniliation option (. (It wold be tehnially more orret to permit hamber to invite a onferene in period pon reeiving hamber s amended bill. Doing so, however, only adds an additional step in the analysis withot providing any additional insights. So hamber is permitted to all the onferene. Let t h be the ommonly observed history of the game throgh period t. The set of possible t histories at the end of period t is denoted H. The hambers and onferene have private information and beliefs. These are aptred by the hambers and onferene s posterior distribtion over S, whih are dependent on the history of the game. These are desribed in greater detail below for eah period of the i,t game. strategy for instittion i in period t is denoted. strategy for the entire game is denoted i, i, i, i,4 i ={,,, }, with i,t = if instittion i does not have a feasible ation in period t. The seene of the game is as follows (see Figre. Period 0. The vale of i for i {,, } is assigned. Period. Chamber s ation set in period is, V = {, }. The message reeived in period (the period with no preeding ations by hamber is M { a, b}. The prior distribtion on S is, fter observing its signal, hamber s posterior distribtion over S in period is jst s,, s m s. strategy for hamber in period is a fntion that takes its posterior into the set of,,

12 ,, ations, ( V. The set of histories at the end of period is given by, H V. Period. Chamber s ation set is omprised of a deision to vote for or, V {, }, and, a reoniliation hoie = {,, }. Chamber s ation set is represented by,, D V, with an ation in period being the dple d, D,. The message reeived by hamber in period is,,, V M {, } { a, b}. For hamber in period the prior distribtion on S is a fntion of and onjetres abot hamber s eilibrim ation. The posterior probability for hamber in period is,,, s ( v, m s =. strategy for hamber in period is a fntion,,, (,, h D s ( v, m for all h H. The ommonly known history at the end of period is H H D,. Period. Chamber s posterior probability on S in period given h H is,, s =. s ( m, d, If a onferene is alled, its posterior probability over S given h H is,, s ( v s =,,. Depending, d, m on the period ation of hamber, hamber or the onferene ommittee might have feasible ations or neither may have feasible ations. Ths, the ation sets for hamber and the onferene ommittee in period mst be restrited to depend on hamber s period ation. In period, the onferene ommittee s ation set is,, {, } if d {(,,(, } V. otherwise The message reeived by the onferene ommittee in period is [ V D M] { }.,,,,,,, s ( v, d, m strategy for the onferene ommittee is (, h V., The period ation set for hamber is,,, {, } if d {(,,(, } V. otherwise

13 Figre. Otline of the iameral Deision-Making Proess Period 0 Period Chamber pdates beliefs given private signal Correlation vales assigned to hambers and onferene Chamber adopts version Chamber or adopts of version bill or of bill Period Chamber pdates beliefs given private signal and hamber s bill Chamber aepts Chamber s bill withot amendment Game ends Chamber amends hamber s bill and proposes onferene reoniliation Chamber amends hamber s bill and motions it bak for reonsideration Chamber rejets hamber s bill withot proposing reoniliation Game ends Period Chamber pdates belief given hamber s ation; or onferene observes private signal and earlier ations Conferene Committee reommends version or nder a losed rle Chamber onrs in or rejets hamber s amendment Game ends Period 4 Chambers and pdate beliefs given onferene reommendation Chambers and aept or rejet onferene reommendation Game ends

14 The message reeived by hamber in period is [ M D ] { }. strategy for,,,,,, V s ( m, d hamber in period is (, h., The ommonly known history at the end of period is H H V, V,. Period 4. In this period, the hambers onsider the onferene proposal nder a losed rle, so hamber i s ation set, i {,}, is given by V i,4 {ept, Rejet} if if v v,, {, }. The message reeived by hamber in period 4 is [ M D V ] { }. The message,4,4,,,4,4 reeived by hamber in period 4 is [ V M V ] { }. strategy for hamber in,,,4,4,4,, V s ( m, d, v period 4 is a fntion (, h. strategy for hamber in period 4 is a fntion,4 (,4 s ( v,,, h V, m, v,4. The legislatre selets a poliy with the hambers voting in periods,, and/or 4, depending on the history of the game. The legislative seletion rle is therefore defined by: ( v,, v, if if if v v v, t, t, t v v v,,, for t for t {,}or if v {,}or if v for t {,} and i,, and v and v,4,4 {,}sh that v i,4 v v,4,4 rejet. aept aept The soltion onept is perfet ayesian eilibrim. tors do not play weakly dominated strategies (whih prodes a nie eilibrim for eah ondition. Informational voting ors when i t v, = (respetively, if the private message reeived by the ator is i m = a (. Several aspets of the model merit some disssion. First, the onferene ommittee is modeled here as an endogenosly hosen message servie. Chambers may vote again on a proposal after reeiving an additional signal regarding the state of the world. The onferene literatre reognizes that onferenes perform this fntion in the legislative proess. For example, Longely and Oleszek write: 4

15 nother [onferene ommittee] negotiating strategy inldes onferees efforts to obtain information both poliy and politial.... Providers are typially other ongressmen, the president, exetive branh offiials, representatives of interest grops, ommittee and personal staff, and the media. t times, these third parties and their information an be instrmental in the resoltion of onferene disptes.... [C]onferenes are sbjet to ageny reestimates, the availability of new information, and a host of other fators... that may move onferenes... toward one hamber s figres bease they reflet hanging realities and nderstandings better (989, 05,, f., ibid., at 50, 07, 84. Similarly, Krehbiel finds some empirial spport for the hypothesis that onferene ommittees serve an informational role in the legislative proess. He onldes that ills that go to onferene will tend to be those abot whose onseenes legislators are nertain and ths in need of the servies of expert agents from whom the parent hamber stands to benefit (99,. Seondly, note that an exogenos voting seene is imposed on the hambers in that hamber votes first, even if hamber has better information (i.e.,. Given that the hambers are always at least as well off if the best informed hamber originates legislation, the assmption might initially appear implasible. There are two reasons that this is an inoffensive assmption, one tehnial and one methodologial. First, there is a straight-forward tehnial fix for this: instead of a point orrelation level being assigned in period 0 and being ommon knowledge, the hambers an be assigned a distribtion of orrelation levels with a known mean and variane. Natre wold not assign a partilar vale to an instittion ntil the beginning of the period in whih it ats. The hamber with the best distribtion (in the sense of stohasti dominane wold then vote first and the hamber with the seond best distribtion wold vote seond. t given that eah hamber s probabilisti assessment of the world is drawn from a distribtion, the originating hamber will sometimes be less informed than the seond voting hamber. That otome is all that is neessary to derive reslts similar to those below. There is also a sbstantive reason to permit the less informed hamber to originate legislation. Endogenos timing models in eonomis (Dagherty and Reinganm 994 and politial siene (Rogers 998 in whih ators are asymmetrially informed and hoose the seene in whih they order their deisions onsistently derive mltiple perfet ayesian eilibria. One eilibrim is the antiipated 5

16 eilibrim in whih the more informed ator moves first and the less informed ator moves seond. t these stdies also onsistently derive a seond perfet ayesian eilibrim in whih the less informed ator moves first and the most informed ator moves seond. s above, that this eilibrim otome ors sometimes is all that is neessary to derive reslts similar to those disssed below. There is, however, a more general faial hallenge to the need for biameralism to help legislatres aggregate information. Given the large poplation reslts of the Condoret jry theorem (Feddersen and Pesendorfer 998, a niameral legislatre of, say, 50 members old hoose orret deisions with a high aggregate probability. Ths, even if biameralism inreases the probability of a orret deision relative to niameralism, the marginal inrease in that probability might be so low that it wold be faially implasible that the benefit of a seond hamber wold be worth the ost. There are at least two reasons that biameral deision-making might nonetheless be benefiial. First, the Condoret theorem depends on the absene of orrelated information that might arise, say, from legislative deliberation (f., sten-smith and anks 996, Lahda 99. t one hallmark of biameralism is that deliberation in the two hambers is aostially separated. This separation an be nderstood preisely as a mehanism to inde norrelated deliberation and, hene, to reate sperior otomes relative to those available shold the same set of legislators deliberate in a niameral body. Seondly, hamber deisions are often prodts of information gathered by a handfl of informed ommittee members (Gilligan and Krehbiel 987, Krehbiel 99. ggregating two aostially separated, small-poplation ommittee deisions aross hambers wold permit signifiant inreases in the probability of seleting a right poliy relative to a niameral legislatre deferring to one small poplation ommittee deision. We now trn to the reslts of the model. Part III. Eilibrim Reoniliation ehavior This part aonts for the eilibrim behavior predited in the interation between ongrent biameral hambers. This jstifies the behavior assmed in deriving the welfare reslts in the next Part. Montesie identified aosti separation as a neessary aspet of biameralism. He arged that legislative power shold be divided between two grops of legislators, eah having their assemblies and deliberations apart (748/949, 55. 6

17 Part. of the lemma states the eilibrim onditions nder whih a bill will die withot reoniliation. Part. states the onditions nder whih a bill is adopted withot amendment. Part.C states the onditions nder whih a bill will be amended by the seond hamber and motioned bak to the originating hamber. Part.D states the onditions nder whih a onferene ommittee will be alled. The relevant behavior ors when hamber messages are different, so that is the fos of the lemma. Lemma. Let (m, m = (a, (or (m, m = (b, a, then there is a nie eilibrim in eah of the following partitions: [] If <,, P( < 0, and P ( < 0, then both hambers vote informatively in periods and. Chamber s bill dies in hamber withot reoniliation. [] If >, and P ( 0, then hamber proposes informatively and hamber votes ninformatively for hamber s bill in period. The bill is adopted by hamber withot amendment. [C] If <,,,, P( 0 and P ( ( ( < 0, then hamber will amend hamber s bill and motion it bak to hamber for reonsideration. Chamber will onr in hamber s amendment. In eilibrim, both hambers vote informatively in their period and period votes. Chamber votes ninformatively for the amended bill in period. [D] [i] If >, and P ( 0, then hamber will propose a onferene ommittee in period and the onferene reommendation will be adopted in period 4. In periods,, and, hamber, hamber, and the onferene ommittee, respetively, all vote informatively. In period 4, the hamber that agrees with the onferene reommendation votes informatively, the hamber whose private message disagrees with the onferene reommendation votes ninformatively. [ii] If <,, P( 0 and P ( 0 then hamber will propose a onferene ommittee in period and its reommendation will be adopted in period 4. In periods,, and, hamber, hamber, and the onferene ommittee, respetively, all vote informatively. In period 4, the hamber that agrees with the onferene reommendation votes informatively, the hamber whose private message disagrees with the onferene reommendation will vote ninformatively. [iii] If <,,, P( 0 bt P (, ( ( > 0, then hamber will propose a onferene ommittee in period and its reommendation will be adopted in period 4. In periods,, and, hamber, hamber, and the onferene ommittee, respetively, all vote informatively. In period 4, the hamber that agrees with the onferene reommendation votes informatively, the hamber whose private message disagrees with the onferene reommendation will vote ninformatively. Proof: See ppendix. The hoies whether to disagree with a proposal from the other hamber, as well as hoies to 7

18 reonile the disagreement and of seleting a reoniliation mehanism, are nie in eah partition stated in the lemma and are derived endogenosly from the model. While the lemma may be visally dense, only three sets of variables with a total of six elements aont for all of the reslts: the payoff of the bill (P; the expertise of the instittions (,, and ; and the ost of the reoniliation mehanisms, ( and. The period prior probability,, and the period posterior probability,, ( a, (reall that, =,, are ompted from,, and. s noted above, the eilibrim reslt is that hamber always votes informatively in period. Therefore, for heristi prposes, the message that hamber, observes,, is stated in terms of the private messages that eah hamber reeives (e.g., (m, m = (a, instead of (V,, m = (,. We now trn to the welfare impliations of ongrent biameralism. Part IV: The iameral dvantage In his speeh opening the 94 Nebraska niameral initiative ampaign, Progressive U.S. Senator George Norris smmarized the basi argment against biameralism this way: The alifiations of members of both branhes of or State legislatre are exatly the same. They represent exatly the same idea. The offiial dties they are to perform are of exatly the same natre. Why shold we then have two bodies instead of one, and brden or taxpayers with the neessarily inreased expense, to attain the objet that an be flly attained by one hose instead of two? (77 For Norris, seond hambers in repblian systems were ongrent hambers and, ths, were wastefl redndanies. (Reall that Progressives regarded biameral stability as a disadvantage of the instittion. This part develops the argment that biameral deision-making even when the hambers are ongrent prodes better legislative otomes than obtain with niameral deision-making. The analytial argment is similar to the one developed for the onferene ommittee in Part III, exept here we move bak one step in the legislative proess and aont for the vale of the information provided by the seond hamber (and by the onomitant reoniliation mehanisms. The relevant onstittional hoie then weighs the expeted vale of the information that a seond hamber provides to the legislative proess against the ost of that hamber. First, the expeted soial gain of niameral legislation is ompted. Redefine the payoff P 8

19 from above to identify soial gain rather than (only legislative payoffs. Let m be the signal observed by a niameral hamber and let m = a. Then the expeted payoff if the niameral legislatre were to adopt bill is: P ( P = ( P > 0 for > 0.5. Next, we want to add one hamber to the niameral legislatre and ompte the gain that wold be aired if the biameral legislatre were to onsider the same bill. To do so, hamber is arbitrarily designated to be the existing niameral hamber. The other hamber is arbitrarily designated hamber. These are arbitrary designations now bease they no longer designate voting seene. Either hamber may vote first or seond on a bill. s before, either hamber may be the more or the less informed hamber, sometimes > and sometimes >. Consider the following theorem. Theorem: Given (m, m = (a, (resp. (b, a and the eilibrim behavior identified in the lemma, the soial vale added by a seond hamber in the onsideration of one bill relative to its onsideration by a niameral body is: [] for < and, [i] the bill does not go to a onferene ommittee, P ( > 0; or [ii] the bill goes to a onferene ommittee, P ( > 0; [] for > and, [i] the bill goes to a onferene ommittee, P ( > 0. 4 [ii] the bill does not go to a onferene, the vale added by a seond hamber is zero. Proof: See ppendix. Importantly, note that the theorem identifies a positive informational gain from a seond hamber both when the seond hamber is more informed than the first/niameral hamber, >, and when the first/niameral hamber is the most informed hamber, >. The existene of a onferene ommittee is rial for the biameral advantage in the latter ase. Nonetheless, to the extent that a seond hamber will sometimes have higher ality information than the first hamber, then biameral systems will be informationally advantageos even if they do not employ onferene ommittees as reoniliation mehanisms (and many biameral systems do not, Tsebelis and Money 99, Three nmerial examples are provided to give a sense of the different otomes identified in the 0. 4, Reall from the lemma that a onferene is alled only if s, whih garantees that P ( > 9

20 theorem. The examples are, of orse, ompletely stylized. Example 4... = 0.6, = 0.7, and P = $00,000,000. There is no onferene ommittee. The benefit of a seond hamber relative to a niameral deision is: P ( = $00,000,000( = $0,000,000. Example 4... = 0.6, = 0.7, = 0.65 and P = $00,000,000. onferene is held. The benefit of a seond hamber relative to a niameral deision is: P ( = $00,000,000 ( 0.65(0.7 (0.6((0.65( = $,800,000. Example 4..C. = 0.7, = 0.6, = 0.65 and P = $00,000,000. onferene is held. The benefit of a seond hamber relative to a niameral deision is: P ( = $00,000,000 ( 0.65(0.6 (0.7((0.65( = $,800,000. s noted, the theorem expresses the vale of a seond hamber for one bill. The total soial prodt of biameralism wold be the vale of the seond hamber smmed aross all legislation that it affets. Call this total soial prodt of a seond hamber. The informational servie provided by a seond hamber is always weakly prodtive: deisions are always at least as good, and sometimes better, with two hambers than with one hamber. Therefore > 0. t simply bease > 0 does not mean that seond hambers always represent net soial gains. The soial prodt of biameralism mst be reded by the ost of a seond hamber,. Therefore the onstittional deision regarding a seond hamber depends on whether the expetation is that > 0 or < 0. 5 Consider several things abot. First, there is a pbli goods aspet to legislative otpt. On the benefits side, everyone within a politial jrisdition lives nder the same set of laws. s with any pbli good, then, the vale of legislative otpt is smmed aross the poplation while the ost of legislation remains onstant (it osts no more to legislate for 5 million people than for 50 million. This permits an intitive predition regarding when biameralism might and might not be ost effetive. It is more likely they are now ignored. 5 The internal legislative osts of onferening or motioning a bill are negligible relative to any gain smmed over an entire polity, so 0

21 that > 0 for larger jrisditions and it is more likely that < 0 for smaller jrisditions. Ths, large ontries and large states wold reeive the largest benefit from biameral instittions. For smaller jrisditions it beomes inreasingly likely that < 0. There seems to be some refletion of this in atal ameral hoies. Lijphart (984, for example, notes a general relationship between the size of a ontry s poplation and the presene of biameral legislatres. ll of the large ontries he stdied had biameral legislatres and most of the smaller ontries had niameral legislatres (9-94. It frther aonts for the predominane of biameralism among U.S. state governments and niameralism among U.S. loal governments. (Interestingly, the New York City Charter Revision Commission onsidered reommending a biameral legislatre for the ity in 989, Prdm 989. t what abot the reverse of the estion: If two hambers are good, might not three or more be better? The estion is not as farfethed as it might sond. Ygoslavia had a triameral system nder its ommnist regime (Corier-Jornal 99 and Soth fria, too, briefly had a triameral legislatre (Longley and Olson 99, 4. Estate-based triameral systems were well known dring the Middle ges (Tsebelis and Money 997, 9 and Sweden even had a form of adameralism (Sydow 99. Frther, as modeled above, onferene ommittees an be viewed as attempts to derive the benefits of triameralism on the heap by forming third hambers only when it is ost effetive to do so. (The literatre often identifies onferene ommittees with third legislative hambers, Vogler 97, Norris 95. t why not three or more permanent hambers? While a flly fleshed ot answer to this mehanism design estion wold take s too far afield, a ople of thoghts are offered. Mltiple private signals are sefl only when there is some likelihood that they will differ. This is why legislation adopted by two hambers is better fitted for its prpose than legislation adopted by only one hamber. t the information implied by these signals is averaged or aggregated by ayes rle, aonting for the ality of the signal as well as the information implied by the message. Eah additional signal, being aggregated with the previos signals, is ths less likely to be informationally pivotal. Under reasonable assmptions abot informational prodtivity, it beomes inreasingly nlikely that three or more permanent hambers

22 wold be ost effetive relative to two hambers that have the option of forming a third body only when they deem it ost effetive to do so. Finally, it shold be noted that the biameral advantage identified in the theorem is not derived from the prevention or delay of legislation. Rather, the advantage the theorem identifies is that biameralism prodes better laws relative to niameralism in the sense that it promotes a better fit between the prposes of legislation and the means adopted to ahieve those prposes. In fat, whether the total nmber of bills inreases or dereases nder biameralism with ongrent hambers is indeterminate. Note that this laim abot the total nmber of laws proded by a biameral legislatre shold not be onfsed with the fat that one hamber will often kill legislation passed by the other hamber. To be sre, with internal reoniliation osts, biameral onsideration reslts in the loss of some bills that wold have been adopted by a niameral legislatre. Yet biameralism also offers two hambers of legislators who may introde legislation. If, for example, the two hambers of a biameral legislatre were eah to send to the other hamber as many bills as a niameral hamber wold adopt, bt eah hamber kills less than half of the other hamber s bills, then the biameral legislatre wold in fat enat more laws than a niameral legislatre omposed of jst one of those two hambers. Nebraska s experiene may illstrate the argment. Nebraska s last biameral session held in 95 saw,056 bills introded and 9 bills adopted. The first niameral session held in 97 saw 58 bills introded and 0 bills adopted (Johnson 98, 4. While half as many bills were proposed, the final nmber of bills adopted in 97 was only slightly greater than in 95. This otome is onsistent with the informational theory proposed here. It is inonsistent with the stability theory of biameralism that identifies its primary effet to be the delay or prevention of legislation. Part V. Conlsion The model also has pratial impliations for states and nations onsidering ameral reform. The analysis developed a heretofore negleted informational rationale for seond hambers. This rationale does not reire that the two hambers have divergent preferenes over poliy otomes. It ths provides a jstifiation for biameralism even when hambers preferenes are ongrent, whih is the limiting ase

23 for biameral hambers in modern repblian systems. The informational advantage of biameralism does not stem from the instittion delaying or preventing the adoption of legislation. Rather, biameral proess is advantageos bease it prodes better legislation relative to a niameral proess. Frther, nlike previos stdies that soght to aont for a biameral advantage with models that abstrated away from the availability of interhamber reoniliation mehanisms, the analysis developed here expliitly inlded reoniliation options in its model of the biameral proess.

24 ibliography ssoiated Press Solidarity and Party Urging Restraint in Polish Vote, St Lois Post-Dispath (Jne 4:. stin-smith, David, and Jeffrey S. anks Information ggregation, Rationality, and the Condoret Jry Theorem, merian Politial Siene Review 90 (Marh: arnett, James D. 95. The iameral System in State Legislation, merian Politial Siene Review 9 (gst C Parliament to Deide Whether to Have Two Chambers, C Smmary of World roadasts (gst : EE/D95/. The Corier-Jornal. 99. Slovenes Go to Polls, The Corier-Jornal (Deember 7: 4. Deler, Egene Inter-Hose Differenes in merian State Legislatres. Jornal of Politis 9 (gst: Daghety, ndrew F., and Jennifer F. Reinganm symmetri Information isition and ehavior in Role Choie Models: n Endogenosly Generated Signaling Game, International Eonomi Review 5 (November: Elmore, Charles Jennings aks Hose-Senate Merger, The Palm eah Post (Deember : 4. Feddersen, Timothy, and Wolfgang Pesendorfer Conviting the Innoent: The Inferiority of Unanimos Jry Verdits nder Strategi Voting, merian Politial Siene Review 9 (Marh: -6. Gilligan, Thomas W., and Keith Krehbiel Colletive Deision-Making and Standing Committees: n Informational Rationale for Restritive mendment Proedres. Jornal of Law, Eonomis, and Organization (Fall: Green, Jak P Negotiated thorities: Essays in Colonial Politial and Constittional History. Charlottesville: University Press of Virginia. Hammond, Thomas H., and Gary J. Miller The Core of the Constittion. merian Politial Siene Review 8 (Deember: Higgins, Jay Uniameral Legislatre Trned Down, angor Daily News (Jne 0: Hirshleifer, Jak and John G. Riley. 99. The nalytis of Unertainty and Information. Cambridge: Cambridge University Press. Jefferson, Thomas. 784/987. Notes on the State of Virginia, in The Fonders Constittion. Volme One. Edited by P. Krland and R. Lerner. Chiago: University of Chiago Press. Johnson, lvin W. 98. The Uniameral Legislatre. Minneapolis: The University of Minnesota Press. 4