Does Bcameralsm Matter? Mchael Cutrone Dept. of Poltcs Prnceton Unversty Nolan McCarty Woodrow Wlson School Prnceton Unversty
. Introducton Perhaps the most conspcuous varaton n modern legslatures concerns the practce of grantng legslatve authorty to two separate chambers wth dstnct membershps. Whle the majorty of natonal governments empower but a sngle chamber, at least a thrd of natonal legslatures practce some form of bcameralsm as do 49 of the 50 Amercan state governments. Scholars have made a number of arguments to explan the emergence of bcameral legslatures. One of the most common arguments for the emergence of bcameralsm n Brtan and ts Amercan colones s that t helped to preserve mxed governments, to ensure that upper class elements of socety were protected (Wood 969, Tsebels and Money 997). In such settngs, bcameralsm allowed the upper chamber, domnated by arstocrats, to have a veto on polcy. More generally, an explct role of some bcameral systems has been the protecton of some mnorty who s overrepresented n the upper chamber. A second ratonale for bcameralsm s the preservaton of federalsm. The Unted States, Germany, and other federal systems use a bcameral system n order to ensure the representaton of the nterests of ndvdual states and provnces, as well as the populaton Tsebels and Money (997, 5) defne bcameral legslatures as those whose delberatons nvolve two dstnct assembles. Ths defnton, however, masks consderable varaton n the roles of each chamber n polcymakng. Many upper chambers have legslatve prerogatves that are lmted n mportant ways; for nstance, the Brtsh House of Lords s unable to orgnate monetary legslaton and, at best, can delay blls for a year rather than permanently veto those they dsagree wth. For our purposes, we wsh to defne bcameralsm more narrowly. We defne bcameralsm as the requrement of concurrent majorty support from dstnct assembles for new legslaton. It s mportant to note that our defnton treats concurrent majortes as a necessary, but not suffcent, condton for enactng legslaton. Thus, t does not preclude other legslatve procedures or consttutonal requrements, such as the sgnature of the executve or supermajortaransm wthn one of the chambers. Our defnton does not map cleanly onto Ljphart s (984) dstncton between strong and weak bcameralsm. Hs dchotomy classfes systems where both chambers have smlar consttuent bases as weak even f they requred concurrent majortes.
of the country. Under federal bcameralsm, the lower house s typcally apportoned on the bass of populaton, whle the upper house s dvded amongst the regonal unts. Some countres, such as the Unted States, provde equal representaton for the states regardless of ther populaton or geographc sze, whle others, lke the Federal Republc of Germany, unequally apporton the upper chamber by provdng addtonal representaton to the larger unts. However, despte ts promnence, the role of bcameralsm n contemporary legslatures has not receved the scholarly attenton that other legslatve nsttutons have. In ths essay, we revew and analyze many of the arguments made on behalf of bcameralsm usng the tools of modern legslatve analyss -- the spatal model, multlateral barganng theory, and games of ncomplete nformaton. Importantly, ths analytcal approach allows us to dstngush the effects of bcameralsm from those of the nsttutonal features whch are often packaged wth t, such as supermajortaran requrements, dfferng terms of offce, and malapportonment. We also revew exstng evdence of bcameralsm s effect on polcymakng.. Spatal Models of Polcymakng The spatal model of polcymakng has become the workhorse model n the study of legslatve nsttutons. Its stark parsmony makes tractable the analyss of a number of nsttutonal arrangements. Such models have proven qute useful n studyng the consequences of bcameralsm and other mult-nsttutonal polcymakng settngs (.e., Tsebels and Money 997, Ferejohn and Shpan 990). We wll not revew the arguments about the role of bcameralsm n the formaton and duraton of parlamentary cabnets (Druckman and Thes 00, Dermeer, Eraslan and Merlo 003).
Before we draw out the mplcatons for bcameralsm, consder a baselne uncameral model. Assume that a uncameral legslature has an odd number of members, n, wth deal ponts, x, arrayed along a sngle deologcal dmenson represented by the real lne. We ndex the deal ponts from left to rght so that x s the leftmost member and x n s the rghtmost member. The legslature seeks to pass legslaton to change a polcy wth a status quo q, whch s also represented by a pont on the spectrum. We assume that each legslator has sngle-peaked, symmetrc preferences so that member weakly prefers y to z f and only f y x z x. The standard predcton, based on Black s theorem, s that the outcome of ths type of chamber would le at the deal pont of the medan legslator x m where n + m =. Any other polcy outcomes could be defeated by some other polcy proposal n a parwse majorty vote. Ths outcome s ndependent of the status quo locaton, q. Krehbel (998) and Brady and Volden (997) have extended the nsghts of the smple spatal model nto mult-nsttutonal settngs. Krehbel s formulaton, dubbed Pvotal Poltcs, s based on the nteracton of pvotal legslators. A legslator s pvotal f her support s necessary for the passage of new legslaton gvng the nsttutonal structure of the polcymakng process and the dstrbuton of preferences. Krehbel s model can easly be modfed to accommodate bcameralsm and concurrent majortes. To do so, consder a basc spatal model wth two legslatve chambers wth szes n and n, both odd. Under concurrent majortaransm, any revson to the status quo must receve majorty support n both chambers. Let 3
m n + = and m n + =. It s easly seen that f m prefers the status quo q to some proposal y, a majorty of chamber wll also prefer q to y. Thus, m s support s necessary or pvotal to the passage of any revson to q. Smlarly, m s pvotal n chamber. Snce m and m must agree to any polcy change, the model predcts that any status quo n the nterval mn { xm, xm }, max { xm, xm } cannot be legslated upon. Any attempt to revse q would be vetoed by one of the chamber medans. However, when the status quo s outsde ths grdlock nterval, the two chambers wll prefer to enact new legslaton. Clearly, the new legslaton wll le n the { xm xm } { xm xm } mn,, max, nterval, because otherwse some other polcy proposal wll be strctly preferred by some concurrent majorty. The pvotal poltcs model does not gve specfc pont predctons about whch polces wll be adopted when q s outsde the grdlock nterval. Such predctons wll depend on specfc protocols for nter-branch barganng such as dfferental rghts of ntaton and procedures for the reconclaton of dfferences such as the conference commttee. 3 Whle qute smplstc, ths model demonstrates two of the arguments whch are forwarded n support of bcameral systems. The frst s that bcameralsm may lead to more stable polces. When the medans of the two chambers dverge such that { xm xm } { xm xm } mn,, max, s a non-empty set, a set of polces wll be stable n the presence of small electoral shocks whch shft the chamber medans. Rker (99) 3 See Tsebels and Money 997 for models of nter-chamber reconclaton procedures. 4
espouses ths stablty argument as a ratonale for bcameral nsttutons. 4 Also apparent from ths llustraton s that bcameralsm requres compromse agreements between the majortes of each chamber. Polcy outcomes wll le n the nterval between the two chamber medans. Ths compromse effect les at the heart of arguments about the presence of bcameralsm n federal and consocatonal democraces (Ljphart 984). Whle theoretcally compellng, the stablty and compromse ratonales depend on a degree of preference dvergence across branches. Wthout a systematc dfference between x m and x m, few compromses between the chambers are lkely to be stable. Thus, f stablty and compromse were the consttutonal desgner s objectve, one would expect to see bcameralsm operate n conjuncton wth dfferent electoral rules for each chamber. In many cases, the electoral bases and procedures dffer dramatcally across chambers as n the U.S. or Germany. However, many systems have chambers wth congruent preferences (Ljphart 984) such as those whch preval n the Amercan state legslatures, especally followng Baker v. Carr whch elmnated many malapportoned upper chambers. Even allowng for dosyncratc dfferences n chamber medans, uncameralsm and congruent bcameralsm should produce nearly dentcal results. Whle short term polces would fluctuate based on the preferences of these two medans, long run polces would locate at the expected medan the same outcome whch occurs n a uncameral legslature. One objecton to ths purely preference based model of bcameralsm s that t gnores the role that poltcal partes mght play n the polcymakng process (e.g. Cox 4 More specfcally, Rker argues that multcameralsm allows for smple majorty rule when an ssue s one-dmensonal and a medan-voter equlbrum exsts, but dscourages decsons for multdmensonal ssues where an equlbrum s unlkely to exst. 5
and McCubbns 003, and Chou and Rothenberg 003). Cox and McCubbns suggest a model of partsan gatekeepng whch produces a uncameral grdlock nterval between the deal pont of the medan member of the majorty party and the chamber medan. Therefore, f we let x J be the deal pont of the medan member of the majorty party of chamber, the grdlock nterval wthn chamber s mn { x, x }, max { x, x } J m J m. Snce polcy change requres that q not fall n ether grdlock nterval, the bcameral grdlock nterval s smply the unon of both chamber ntervals and the non-partsan grdlock nterval or mn { xj, xj, xm, xm }, max { xj, xj, xm, xm }. Clearly, ths partsan grdlock nterval wll be largest when the majorty party member of one chamber les to the rght of the medan whle the other les to the left. Generally, ths wll occur when the dfferent partes control the dfferent chambers. The spatal model can also ncorporate a number of features that are often assocated wth bcameralsm such as supermajorty requrements for one of the n + chambers. Now assume that chamber requres k > to pass new legslaton. Ths requrement now makes members k and n k pvotal for changes to the status quo. Therefore, the grdlock nterval s mn { xm, xn k}, max { xm, xk}. It s easy to see that supermajortaransm wll ncrease the grdlock nterval. However, n cases where the two chambers are reasonably congruent, supermajortaransm wll cause m to no longer be pvotal, makng one chamber, n some sense, redundant. Perhaps more mportantly, ths analyss shows that the key features of bcameralsm, stablty and compromse, can 6
be obtaned by uncameral legslatures wth sutably chosen supermajorty procedures. The spatal model provdes no ratonale for choosng one nsttuton over the other.. Spatal Models of Malapportonment Gven that our revew of spatal models suggests that bcameralsm should matter only when the chambers are apportoned dfferently, the obvous queston s: under what crcumstances would malapportonment be a reasonable consttutonal desgn. Such an answer s provded n a recent paper by Crémer and Palfrey (999). They consder a model where the ctzens of varous jursdctons must decde on a level of centralzaton and representaton at a consttutonal phase, takng nto full account what polcy outcomes wll result from the consttutonal choce. Centralzaton s modeled as the extent to whch the polcy outcome vares across jursdcton. Wth complete centralzaton, the polcy outcome n each jursdcton s dentcal whereas wth decentralzaton each jursdcton sets ts own polcy. Snce Crémer and Palfrey assume that there s ncomplete nformaton about voter preferences, rsk averse voters wll prefer some centralzaton n order to reduce varaton n polcy outcomes. When selectng the type of representaton, voters choose the weghts that fnal outcome wll place on the outcome of dstrct electons. In the case of populaton representaton, the weghts are proportonal to dstrct populaton. In unt representaton, each dstrct receves the same weght, regardless of populaton. Crémer and Palfrey show that: Voters wth extreme polcy preferences (relatve to the expected medan of the centralzed polty) prefer completely decentralzed polcymakng. The greater 7
opportunty to get the polcy they want from ther dstrct outweghs the varance reducton afforded by centralzaton. Moderate voters (those close to expected medan) are unanmously opposed to populaton representaton for any level of centralzaton. Ths s because the unt rule unquely mnmzes the varaton n centralzed polcy. Gven hgh levels of centralzaton, extreme voters from large dstrcts wll prefer some populaton representaton. Ths ncreases ther votng weght n the centralzed legslature whch move polcy back towards ther deal pont. Conversely, as long as the level of centralzaton s suffcently low, moderate voters from small dstrcts wll prefer populaton representaton because ths reduces the nfluence of extreme small dstrcts on the centralzed component of polcy. Ths s the functonal equvalent of cedng addtonal soveregnty to the center. Gven these preferences, they examne the votng equlbra n the consttutonal stage. If a majorty rule equlbrum exsts 5, they show that t nvolves representaton based solely on the unt prncple. However, f the level of centralzaton s fxed and representaton s voted on separately, the condtonal votng equlbrum generates representaton whch s a mx between the populaton and unt prncples. The ratonale s that f centralzaton s fxed, extreme voters from large dstrcts and moderate voters from small dstrcts wll vote n favor of postve levels of populaton representaton. 5 In ths context a majorty rule equlbrum s a combnaton of centralzaton and representaton for whch no other combnaton s preferred by a majorty. 8
The Crémer-Palfrey model provdes a reasonable mcro-foundaton for endogenously malapportoned legslatures. However, t falls short of a ratonale for bcameralsm because t black boxes the legslatve nsttutons that make centralzed polcy. Thus, t s plausble that a malapportoned uncameral legslature could provde the representatonal foundaton for greater centralzaton just as well as two chambers based on dfferent representatonal prncples.. Evdence The key predcton of the undmensonal spatal models of polcymakng s that bcameralsm matters only when preferences of each chamber are dssmlar. In the context of Amercan poltcs, there have been mportant perods n whch preferences dverged dramatcally across chambers. For example, the antebellum balance rule parng the admsson of slave and free states contrbuted to the Senate beng sgnfcantly more pro-slavery than the House (Wengast 998). After the war, the admsson of lowpopulaton Republcan states n the West gave the Republcans a sgnfcantly larger advantage n the Senate (Wengast and Stewart 99). 6 Generally, however, the chambers have been qute congruent n ther preferences. Fgure shows the percentage of seats n both chambers held by Democrats snce the restoraton of the two-party system at the end of Reconstructon. 6 McCarty, Poole, and Rosenthal (000) fnd, however, that the substantve effect of the Republcan rotten boroughs s small and short-lved. 9
Fgure Democratc Seat Share: 879-004 0.9 0.8 0.7 Percent 0.6 0.5 0.4 0.3 0. 879 885 89 897 903 909 95 9 97 933 939 House 945 Year 95 957 963 Senate 969 975 98 987 993 999 There are only two perods n whch the seat shares dverge sgnfcantly for an extended perod of tme. The frst s the aforementoned post-reconstructon Republcan bas n the Senate. The second s that caused by Republcan control of the Senate followng Ronald Reagan s electon n 980. However, congruency was almost completely restored after the 986 electon when most of the Republcan freshman class was defeated. If we look at measures of the medan preferences of both chambers, we fnd almost the same pattern. Fgure plots the st dmenson, common space adjusted DW- NOMINATE score for each chamber s medan. 7 7 For a dscusson of common space-adjusted DW-NOMINATE scores, see McCarty, Poole, and Rosenthal 997 and Poole 998. The frst dmenson captures each legslator s poston on a lberal-conservatve scale whch runs roughly from - (very lberal) to (very conservatve). 0
Fgure Chamber Medans 0.4 0.3 0. 0. Medan 0-0. -0. -0.3-0.4 879 883 887 89 895 899 903 907 9 95 99 93 97 93 935 939 Senate 943 Year 947 95 955 959 963 967 97 975 979 983 987 99 995 999 Based on preference measures, there are three addtonal perods of ncongruence caused by more conservatve Houses durng the 90s, 950-960s and followng the 994 ascendancy of the Republcans to the majorty. The conservatve bas of the House n the 90s s accounted for by the number of progressve Republcans n the Senate. Durng the 950s and 960s, southern Democrats caused the House to be more conservatve than the Senate. Smlarly, the large number of conservatves who entered the House n 994 explans the contemporary dfference. House Gven Krehbel s arguments about the U.S., perhaps the relevant effect of bcameralsm should be measured n term of the contrbuton of the House to the grdlock nterval. The House wll only have a postve contrbuton to grdlock so long as ts
overrde pvots are more extreme than those of the Senate. In Fgure 3, the sold lne plot the grdlock nterval from 937 to 00 usng common space-adjusted DW-NOMINATE scores. The dotted-lne shows what the wdth of the grdlock nterval would have been n the absence of the House. Thus, clearly the House generally makes a small contrbuton to grdlock except n the 980s. However, most of the varaton n the grdlock nterval s due to the Senate preferences. Fgure 3 Grdlock Intervals 0.7 0.6 0.5 0.4 Wdth 0.3 0. 0. 0 937 939 94 943 945 947 949 95 953 955 957 959 Total Grdlock Interval 96 963 965 967 969 Year 97 973 975 977 Grdlock Interval- Senate Only 979 98 983 985 987 989 99 993 995 997 999 00 The two seres are correlated at the.9 level. If we computed the grdlock nterval n the absence of the Senate, the House grdlock nterval correlates wth the actual nterval at only the.5 level. Thus, the pvotal poltcs model predcts that the effects of bcameralsm at the natonal level should be small. Now we consder the mplcatons for U.S. state legslatures. Not surprsngly, state legslatures n the U.S. are qute congruent as well. In fgure 3, we plot a measure of partsan ncongruence for each regon n each year snce 954. Ths measure s smply
PI % DemsLower House = % DemsUpper House Thus, t takes on a value of 0 when the partsan composton s dentcal across chambers. Before Baker v. Carr, Republcans were often overrepresented n upper houses outsde the South. However, after the mplementaton of one person-one vote, there s no systematc tendency for partsan ncongruence. Party Incongruence of State Legslatures 954-004 0.4 0.35 0.3 Incongruence 0.5 0. 0.5 0. 0.05 0 954 956 958 960 96 964 966 968 970 97 Northeast South Mdwest West 974 976 978 980 98 984 986 988 990 99 994 996 998 000 00 004 Unfortunately, we lack measures of preference estmates at the replcate our analyss of the natonal level, but t seems lkely that the mplcatons for bcameralsm wll be the same. Whle nter-chamber dfferences are small, ths does not preclude the possblty that varaton n these dfferences has mportant consequences for polcy. However, few studes have looked at the effect of nter-chamber dfferences. Bnder (999, 003) fnds that the dfferences n chamber preferences are negatvely correlated wth her measure of legslatve producton n the post-war II perod. In her 999 paper, she measures 3
bcameral dfferences as the dfferences n chamber medans usng W-NOMINATE scores. However, these scores are not generally comparable across chambers. Her measure correlates only weakly wth those derved from scores that facltate nterchamber comparsons. In her 003 book, Bnder uses a new measure of bcameral dfferences based on agreement scores on conference reports. However, there s some crcularty n the argument that legslatve producton s hgher when the House and Senate vote smlarly on conference reports. In addton conference reports are a hghly selected sample snce a mere twenty percent of blls go to conference (Longley and Oleszek 989). Chou and Rothenberg (003) also test several grdlock models fully ncorporatng bcameralsm and partsan effects. They fnd very lttle evdence of effects attrbutable solely to bcameralsm. Lookng at the state level, Rogers (003) looks at the effects on legslatve productvty of movng from a bcameral legslature to a uncameral one, and vce versa. Unfortunately, he s able to examne only four cases and fnds mxed evdence across the cases. Thus, the queston of the mportance of nterchamber dfferences n U.S. polcymakng remans somewhat contested. 3. Multdmensonal Spatal Models Some authors have stressed the role of bcameralsm n ameloratng the ntranstvty of majorty rule (McKelvey 976, Schofeld 978). These authors have focused on the queston of whether or not t can produce a core or reduce the sze of the 4
uncovered set n the absence of a Condorcet wnner n a uncameral legslature. 8 Cox and McKelvey (984) demonstrate that the necessary condton for the exstence of a core n a multcameral legslature s the ncongruence of the medan preferences across chambers. In a undmensonal settng, the core wll exst and wll be the nterval connectng the medans of the two chambers the same as the grdlock nterval we extracted from the Pvotal Poltcs model. Hammond and Mller (987) extended ths lne of analyss and deduced the necessary condtons for the core n two-dmensons. They defne a bcameral medan as a lne whch dvdes the legslature such that a majorty of the members of both chambers le on ether sde of the lne (ncludng the lne tself). They show that the core exsts f only one bcameral medan lne exsts. Typcally, ths condton wll not be met f there s preference congruence between the chambers. Furthermore, Tsebels (993) proves the generc non-exstence of the core when the polcy space s larger than two dmensons. Because the condtons for bcameralsm to generate a core are so demandng, Tsebels focuses on the weaker soluton concept of the uncovered set. Unlke the bcameral core, the bcameral uncovered set s guaranteed to exst. He shows that the bcameral uncovered set s always at least as bg as the uncameral uncovered set. 3. Evdence Heller (997) uses the multdmensonal spatal model to explore dfferences n fscal polcy between uncameral and bcameral systems. He argues that the grdlock 8 The core s the set of alternatves that cannot by some other alternatve gven the votng rule. The uncovered set s based on the coverng relaton: some alternatve y covers x f y defeats x accordng to the votng rule and f the set of ponts that defeat y s strctly smaller than the set of ponts that defeat x. The uncovered set s the set of ponts are not covered by some other alternatve. 5
caused by bcameralsm s often overcome by hgher levels of government spendng. He tests ths clam on 7 parlamentary systems from 965 to 990 and fnds that bcameralsm s assocated wth hgher annual government defcts budget defcts. However, hs study has two mportant lmtatons. Frst, he does not dsaggregate annual defcts nto ther expendture and revenue components. Thus, hs clam that bcameralsm causes excessve spendng growth cannot be tested aganst the alternatve theores that bcameral grdlock constrans revenue. Second, the mplcaton of hs model that the effect of bcameralsm should be greater, the less congruent the chambers s not tested. In hs 00 artcle, Heller supplements the multdmensonal model wth partsan ncentves for logrollng. He concludes that bcameral systems where the chambers have smlar partsan compostons should generate hgh levels of logrollng and therefore spend more and produce larger defcts. He tests ths predcton on nne bcameral parlamentary governments and fnds that defcts and expendtures are negatvely correlated wth a number of measures of partsan dfferences across chambers. Gven the problems of testng multdmensonal models on observatonal data, Bottom et al (000) use laboratory experments test the exstence of a bcameral core. Ther results provde support the stablty-nducng propertes of bcameralsm, but the external valdty of such experments s hard to substantate. 9 4. Bcameralsm and Dstrbutve Poltcs 9 There are two major dffcultes n testng socal choce models n the laboratory. The frst s that, whle socal choce models are nsttuton free, experments must have protocols for proposal makng and votng. The second problem s that t s dffcult to know whether or not laboratory condtons (such as congruence) match real world condtons. 6
Gven ts hstorcal ratonale as an nsttuton to provde benefts for specfed classes and groups, a natural queston to ask s whether bcameralsm s an effectve way of engneerng partcular dstrbutve outcomes. A number of dstrbutve mplcatons for bcameralsm and related nsttutonal arrangements can be derved from the legslatve dvde-the-dollar barganng games poneered by Baron and Ferejohn (989). Before dscussng specfc models addressng bcameralsm, we revew the basc framework. Assume that a legslature wth N (an odd number) members must allocate one unt of resources (.e., a dollar). Baron and Ferejohn consder barganng protocol wth a random recognton rule under whch at the begnnng of each perod one of the players s selected to make a proposal. Let p be the probablty that legslator s selected to make the proposal, and we assume that ths probablty of recognton s constant over tme. We wll focus on the closed rule verson of the model where the proposer makes a taket-leave-t offer for the current legslatve sesson. The proposer n each perod makes an offer ( ) x, x,..., xn such that x s the share for player and we requre N = x =. If a N + smple majorty, n =, vote for the proposal t passes, the benefts are allocated, and the game ends. If ths proposal s rejected, a new proposer s chosen at the begnnng of the next sesson. All players dscount payoffs secured n future sessons by a factor δ. Ths game has lots of subgame perfect equlbra. In fact, for suffcently large N and δ, there s a subgame perfect equlbrum that can support any dvson of the dollar. Thus, Baron and Ferejohn and subsequent authors generally lmt ther analyses to statonary equlbra. A statonary equlbrum to ths game s one n whch: 7
. A proposer proposes the same dvson every tme she s recognzed regardless of the hstory of the game.. Members vote only on the bass of the current proposal and expectatons about future proposals. Because of assumpton, future proposals wll have the same dstrbuton of outcomes n each perod. Let v ( h ) be the expected utlty for player for the barganng subgame begnnng n t tme t gven some hstory of play, h t. Ths s known as legslator s contnuaton value. Gven the assumpton of statonarty, contnuaton values are ndependent of the hstory of play so that v( ht) = v for all h t, ncludng the ntal node h 0. Therefore, the contnuaton value of each player s exactly the expected utlty of the game. Fnally, we wll focus only on equlbra n whch voters do not choose weakly domnated strateges n the votng stage. Therefore, a voter wll accept any proposal that provdes at least as much as the dscounted contnuaton value. Therefore, any voter who receves a share x δ v wll vote n favor of the proposal whle any voter recevng less than v δ wll vote aganst. 0 Thus, the proposer wll allocate δ v to the n members wth the lowest contnuaton values. As a benchmark for comparson wth the models that we dscuss below, consder the case where all members have the same recognton probablty so that p = for all N 0 The requrement that legslators vote n favor of the proposal when ndfferent s a requrement of subgame perfecton n ths model. 8
. In ths case, the unque expected payoffs from the statonary subgame perfect equlbrum are v =. Thus, the dollar s splt evenly n expectaton. N 4. Concurrent and Supermajortaransm McCarty (000) consders an extenson of the Baron-Ferejohn model to study the dstrbutonal effects of concurrent majortes and a number of other features assocated wth bcameralsm. He assumes that there are two chambers, and, wth szes m+ m = N. A proposal must receve at least k n chambers =, to pass. In each perod, a proposer s selected at random where each member of chamber s selected wth probablty p. Thus, the proposal probabltes are constant wthn chambers, but may vary across chambers. Ths may reflect consttutonal provsons that gve certan chambers an advantage n ntatng certan types of legslaton. McCarty derves the rato of the expected payoffs for a member of chamber to a member chamber v, r =, as a functon of the key parameters k, m, and the chamber-specfc tme v dscount factor δ. Hs model predcts that r = k p δ m k p δ m From ths equaton a number of mplcatons about bcameralsm can be derved. For our purposes, the most mportant s that the sze of the chamber, m, does not have an k k effect ndependent of the chamber s majorty requrement k. If = (as would be m m 9
the case f both chambers were majortaran), the relatve payoffs depend only on the allocaton of proposal power and the dscount factors. If both chambers are co-equal n ther ablty to ntate legslaton and dscount the future equally, the requrement of concurrent majortes does not have dstrbutve mplcatons. Therefore, the fact that upper chambers are generally smaller does not make t more powerful. Ths predcton stands n drect contrast to power ndces such as those of Shapley and Shubk (954). Such ndces are based on the assumpton that all wnnng coaltons are equally lkely, therefore members of the smaller chamber are more lkely to be ncluded. In the McCarty model, legslatve proposers choose majortes n each chamber to mnmze the total costs. Thus, competton to be ncluded n the majorty coalton for the chamber elmnates any small chamber advantage. Whle hs model predcts that concurrent majortaransm does not have dstrbutve consequences, a chamber s use of supermajorty requrements such as the U.S. Senate s cloture provson benefts ts members. In ths sense, the model s results are very smlar to those of Dermeer and Myerson (999), who argue that, n a bcameral system, each chamber would lke to ntroduce at least as many nternal veto ponts as the other chamber. It s also consstent wth our argument we derved from from the pvotal poltcs model that supermajortaransm s more consequental than concurrent majortes. Secondly, note that the relatve payoffs of chamber to chamber are ncreasng n p and decreasng n p. Thus, consttutonal procedures that gve dfferent chambers dfferental rghts to ntate legslaton have dstrbutonal consequences. 0
Fnally, consder the mplcatons of tme dscountng. Ceters parbus, the chamber whose members have the hghest dscount factors get more of the benefts. Snce one would naturally assume a correlaton between the dscount factor and term length, a chamber whose members are elected for longer terms should get more of the dollar. A lmtaton of McCarty s model s that t mplctly assumes that legslators represent dsjont consttuences whereas n actual bcameral systems voters are typcally represented on both levels. Ansolabehere, Snyder, and Tng (003) (whch we dscuss n more detal n a later secton) develop a dstrbutve model of bcameralsm whch ncorporates dual representaton. Consstent wth McCarty, they fnd that, absent malapportonment or supermajortaransm, per capta benefts are equal for all voters. 4. Bcameral Pork Sequental choce models can also be used to make predctons about the extent to whch bcameral legslatures wll be more or less fscally prudent than uncameral legslatures. In ths secton, we extend the models of Baron and Ferejohn (989) and McCarty (000) to determne whch system s most lkely to pass legslaton whose total costs exceed total benefts. Consder a set of possble spendng proposals wth varyng levels of aggregate beneft B and total cost T. Followng the same closed rule descrbed n the prevous secton, a proposer s selected n each perod to propose an allocaton of B under closed rule. If the proposal passes, the benefts are allocated accordng to the proposal and each legslator pays the same per capta tax, T t. If the proposal fals, no benefts are N
allocated, dscountng occurs wth a common factorδ, and a new proposer s selected n the next perod. As above assume that there are two chambers wth membershps m and m where m+ m = N. Further, we assume that k and k votes are requred n each chamber for passage. To keep notaton smpler, q k = be the requred proporton of votes for m passage n each chamber. Agan we focus on symmetrc statonary equlbra and elmnate weakly domnated strateges. Therefore, a member of chamber wll vote n favor of any proposal for whch the net benefts must exceed the dscounted contnuaton value.e. x t δ v. As a benchmark for comparson, consder a uncameral legslature requrng k + k of N votes. Ths s smply a fuson of the two chambers and ther votng rules. A drect applcaton of Baron and Ferejohn (989) shows that any project such that ( δ )( k + k) δ ( + ) B > T N k k wll be enacted n a uncameral chamber. Note that ths crtcal beneft-cost rato s less than snce N > k + k. Thus, the uncameral legslature enact many neffcent programs where B < T. Now we consder whether bcameralsm ncreases or decreases the tendency to enact neffcent projects. For a proposal to pass, a proposer from chamber must obtan k other votes from chamber (she wll vote for her own proposal) and k votes from chamber. A proposer from chamber has to buld the analogous coalton. Snce each vote costsδ v + t, the net benefts of proposng are
( )( δ ) ( δ ) z = B k v + t k v + t t () ( δ ) ( )( δ ) z = B k v + t k v + t t () for proposers from chambers and respectvely. Now we can compute the contnuaton values for members of each chamber. Assumng that proposers randomze across members when ndfferent, we can show that φ v = ( z + φδ v) t N N φ v = ( z + φδ v) t N N (3) (4) where m φ = k + k and φ = k + m k are the probabltes that each member of m m chamber () s selected. Note that smple algebra reveals that φ = qn. and From these equatons, t can be verfed that [ δ ] [ δ ] = ( ) q v q v q q t mv + mv = B T The key for determnng whether or not a proposal wll pass s to verfy that t s ratonal for the proposer to make a proposal. Ths ratonalty condton s z δ v for =,. Otherwse, the proposer would do better by not makng a proposal. Consder the case where both chambers use the same votng rule so that q = q. B T Then the only soluton to these equatons () to (4) s v = v = whch mples that N z δ v f and only f 3
( δ )( k + k) δ ( + ) B > T N k k Ths s exactly the same threshold as the uncameral case. Thus, f both chambers have the same recognton probabltes and votng rules, there s no dfference between bcameralsm and uncameralsm when t comes to the pork barrel. Ths contradcts the predctons that Heller (997) derved from the mult-dmensonal spatal model. A full analyss of the bcameral pork game s beyond the scope of ths chapter. However, ths sketch suggests that, as we saw n the purely dstrbutve game, any effect of bcameralsm must depend on votng rules that vary across chambers, asymmetrc recognton probabltes, or as we dscuss n the next secton, malapportonment. 4.3 Malapportonment Dstrbutve legslatve models also speak drectly to the effects of malapportonment. As we dscussed above, the unque statonary subgame perfect equlbrum of the Baron-Ferejohn model predcts that f all legslators have the same proposal power, ther ex ante payoffs wll be dentcal. Snce legslatve payoffs are equal, the per capta payoffs to consttuents wll be much hgher n dstrcts wth fewer voters. Thus, malapportonment wll lead to skewed dstrbutons of benefts. Whle the malapportonment result from the Baron-Ferejohn model s somewhat mechancal, a recent model of Ansolabehere, Snyder, and Tng (003) produces a rcher set of mplcatons of malapportonment. In ther basc model, 4
The lower chamber (House) represents dstrcts wth equal populaton and the upper chamber (Senate) represents states contanng dfferent numbers of dstrcts. Each dstrct has one representatve as does each state. Publc expendtures are allocated to the dstrct level and legslators are responsve to ther medan voters. Thus, House members seek to maxmze the benefts gong to her dstrct whle a senator s assumed to maxmze the benefts gong to the medan dstrct n her state. Both chambers vote by majorty rule wth all proposals emanatng from the House. Each perod begns wth a House member selected at random to propose a dvson of the dollar whch s voted on by the House and Senate. If the proposal obtans majorty support n both chambers, t passes and the game ends. If not, the game moves to the next perod and a new proposer s selected from the House. The authors show that all symmetrc, statonary, subgame perfect Nash equlbra the expected payoffs to all House members are dentcal, regardless of the sze of ther state. Snce all dstrcts are equal populaton, per capta beneft levels are constant despte malapportonment n the Senate. However, f the game s modfed so that senators may make proposals, there s a small state advantage attrbutable to malapportoned proposal rghts. 4.4 Evdence Symmetry here mples that all house members from states of the same sze are treated symmetrcally. The authors ndcate that there are other payoff dstrbutons sustanable when ths assumpton s dropped. 5
Our revew of the dstrbutve models suggests that the effects of bcameralsm should be prmarly assocated wth supermajortaransm and malapportonment. Whle there s lttle emprcal work on the dstrbutve effects of supermajortaransm, there s a rch emprcal lterature on malapportonment. Before the Supreme Court ruled that malapportoned state legslatures were unconsttutonal, scholars (.e., Adran 960, Dye 966, Jewell 96) were extremely nterested n the mplcatons of malapportonment. Ths research emphaszed ts mplcatons for levels of party competton, nequtable dstrbuton of state fundng, and the falure to adopt certan socal polces (Lee and Oppenhemer 999, 4). There contnues to be a debate about the consequences of elmnatng malapportonment n the states, though recent research has found large effects on the allocaton of state spendng (Ansolabehere, Gerber, and Snyder 00). Work on the effects of malapportonment on dstrbutve polcy has focused on the small-state bas created by the representaton of states n the Senate. Lee and Oppenhemer (999, 6) fnd that, based on the 990 s apportonment, 3 states are overrepresented due to equal representaton of the states, whle 4 states were underrepresented and 5 receved an approxmately proportonal amount. Atlas, et al. (995) fnd a sgnfcant postve relatonshp between a state s US Senate representaton per capta and the state s net recepts from federal expendture. Lee and Oppenhemer (999, 74) consder the mpact of Senate malapportonment on both dscretonary and non-dscretonary fund allocatons. They fnd that states who are overrepresented receve dsproportonate allocatons of both 6
dscretonary fundng and non-dscretonary fundng. These relatonshps are consstent across polcy areas. Thes (998), n a comparatve case study of agrcultural spendng, fnds compellng evdence that rural overrepresentaton n the U.S. Senate blunted retrenchment n agrcultural spendng compared to Japan where both chambers became controlled by urban nterests smultaneously n the 970s when the Lberal Democratc Party became a predomnately urban party. 5. Informatonal Explanatons Recognzng that exstng theoretcal arguments on behalf of bcameralsm lacked much bte when the chambers have smlar dstrbutons of preferences, Rogers (00) attempts to provde an nformatonal ratonale for bcameralsm. Hs model s based on the nteracton of three actors: chambers and and a conference commttee (C). Ths legslature must choose between polces A and B, or not to act (e.g. polcy φ). All actors share the same state contngent preferences such that they all prefer polcy A (B) n state A (B) to the null polcy whch s preferred to polcy A (B) n state B (A). Each chamber receves a sgnal s { A, B} about the state of the world. The sgnal s correct wth probablty q > for {,, C}. We wll refer to q as player s sgnal qualty. Assume that each player receves for the correct polcy, - for the ncorrect polcy, and 0 from the default outcome. The sequence of the game s as follows. In state, chamber proposes one of the polces A or B. In response, chamber may ether accept chamber s choce, amend t and send t back (.e. propose the other polcy), amend the bll and propose the formaton 7
of a conference commttee to reconcle the dfferences, or reject the bll outrght leadng to polcy φ. Rogers shows that each of the outcomes may be acheved as part of a perfect Bayesan equlbrum. As a benchmark, note that a uncameral legslature wth sgnal qualty q u receves a payoff of qu. Rogers compares ths outcome to the outcome when chamber wth sgnal qualty q and the possble use of a conference commttee wth qualty q c. Not surprsngly, the aggregate utlty must be weakly ncreasng. After all, the lowest qualty chamber has a ncentve to at least defer to the hgher qualty chamber. And generally, the sgnals can be pooled snce both chambers have an ncentve to reveal ther nformaton truthfully. The only case where the payoffs of bcameralsm and uncameralsm are the same s when the second chamber has a much lower qualty sgnal and the cost of usng the conference procedure are large. Whle suggestve, Rogers model s somewhat restrctve and t s unclear how well t would generalze. Frst of all, each chamber s modeled as a untary actor wth a fxed sgnal qualty. Thus, t does not address whether bcameralsm s preferred to reforms wthn the uncameral legslature that enhance ts sgnal qualty. For example, ts not clear that benefcal effects of the conference procedure could be replcated wth the uncameral legslature. Secondly, t does not address whether t would be more sensble to smply ncrease the sze of a uncameral legslature rather than add a second body. To move towards askng the queston n these ways, we sketch some mplcatons from models of votng under ncomplete nformaton and common values (the so-called Condorcet jury problem ). These models seem to suggest a much more crcumscrbed nformatonal beneft of bcameralsm. 8
5. Non-Strategc Jury Theorem Consder n legslators who must decde whether to choose polcy 0 or polcy. They all have a common preference for choosng the correct polcy. They each get an ndependent sgnal s { 0,} about whch polcy s the common preference. We assume that each player s sgnal s correct wth probablty π. The probablty of makng a correct decson under majorty rule s therefore n j (, π) = π ( π) P n j= n+ n j Note that P( n, π ) s ncreasng n both of ts arguments. Now consder mplementng bcameralsm by dvdng the n legslators nto two chambers wth m and m members where m+ m = n+. 3 We assume that each chamber votes va majorty rule. We wll desgnate polcy 0 as the default polcy whch s to be adopted f the chambers do not agree. Then the probablty of a correct choce s n j { } ( ) ( ) { } ( ) ( )( ( )) Pr state = P m, π P m, π + Pr state = 0 P m, π P m, π Snce > (, π) > (, π), P( n, π) P( m, π) P( m, π) P n P m >, the bcameral system does worse n state. In state 0, bcameralsm s more lkely produce the correct decson snce deadlock produces the favorable result. Thus, superorty of bcameralsm would Allowng the sgnal qualty to vary across ndvduals s unlkely to change our analyss so long as sgnal qualty does not vary systematcally across chambers n the bcameral case. 3 The fundamental methodologcal problem n the comparatve study of bcameralsm s that t s mpossble to dvde an odd-numbered legslature nto two odd-numbered chambers. Perhaps ths suggests the mportance of trcameralsm. 9
depend entrely on whch outcome s desgnated as the default. Of course, f they had ths nformaton ex ante, they wouldn t need to vote! Another problem for bcameralsm s that t wll never be the ex post best decson rule. Under uncameral majorty rule, all legslators would agree ex post that t s best to mplement the majorty s preference. However, under bcameralsm, all legslators would lke to reverse any decson that dsagreed wth the majorty of all votes cast. Thus, decson-theoretc votng models do not support the concluson that bcameralsm serves an nformatonal functon. 5. Strategc Jury Models Of course, a crtcal objecton to the analyss of the prevous secton s that the legslators were non-strategc n that they voted based on ther prvate nformaton rather votng n the way they would f they were pvotal. In the bcameral context, ths would mply that legslators should condton ther vote based on ther belefs about the state of the world when they would provde the te-breakng vote n ther chamber and ther expectatons about the other chamber s decson. A full analyss of strategc votng wth ncomplete nformaton under bcameralsm s beyond the scope of ths essay. However, there are strong reasons to beleve that bcameralsm would provde no advantages over uncameralsm. Austen- Smth and Banks (996) show that, for a gven legslature, prors, and sgnal qualty, there s an optmal q-rule whch fully aggregates all nformaton and makes the optmal decson gven ths nformaton. Thus, a uncameral legslature usng the optmal votng rule would do at least as well as any bcameral arrangement. 30
5.3 Endogenous Informaton Acquston Whle the arguments of the precedng secton suggest that the requrement of concurrent majortes s unlkely to aggregate nformaton better than smple majorty rule, t may stll be the case that bcameralsm affects the ncentves of legslators to acqure nformaton and develop legslatve expertse. If nformaton conveys a legslatve advantage, bcameralsm mght nduce nter-chamber competton n nformaton acquston. On the other hand, f nformaton s a publc good, bcameralsm mght nduce more free-rdng. The only work on ths queston s Rogers (998), who develops a game-theoretc model of the decson of each chamber to become nformed about the consequences of pendng legslaton. The game s three perods: one for the acquston of nformaton and the followng two for the proposal of legslaton. In Perod 0, each chamber, h and s, decdes whether or not to become nformed at costs f h and f s. In Perod, each chamber decdes whether to ntroduce legslaton. If a sngle chamber ntroduces a bll, the other chamber may update ther nformaton (f the frst mover s nformed, whle the follower s not). Followng ths, the second chamber consders the legslaton and payoffs are awarded. In the perfect Bayesan equlbrum of ths game, the chamber wth the lower nformaton costs generally obtans a frst mover advantage as the second chamber free rdes off ts nformaton. Ths effect s enhanced when the two chambers have smlar polcy preferences. To test these hypotheses, Rogers employs a dataset of all legslaton adopted by both chambers n 33 state legslatures. Argung that the lower chambers have lower nformaton costs because of ther larger szes, Rogers regresses the percentage of 3
legslaton ntated by the lower chamber on the explanatory varables on the relatve sze of the lower chamber and a dummy varable for cases where a sngle party controlled both chambers by a /3s vote. He fnds that both of these key varables are strongly correlated wth the percentage of lower chamber ntroduced legslaton. Whle these results are supportve of the dea that bcameralsm affects the nformaton envronment of the legslature, the model s not conducve to an explct comparson of bcameralsm and uncameralsm. Thus, t remans unclear whether ntercameral competton over the agenda wll domnate free-rdng suffcently to provde an nformatonal ratonal for bcameralsm. 6. Conclusons In ths essay, we have consdered a number of arguments n favor of bcameralsm as an organzng prncpal for modern legslatures. When vewed through the tools of contemporary legslatve analyss spatal, multlateral barganng, and nformatonal models the case for bcameralsm seems less than overwhelmng. Even n models where bcameralsm mght have an effect, we fnd that the necessary condtons for such an effect are emprcally rare. Further, much of the emprcal evdence of the polcy effects bcameralsm s ether weak or attrbutable to ether malapportonment or supermajortaransm, outcomes that could theoretcally be produced n uncameral legslatures. The role of bcameralsm n modern legslatures s of more than smple academc nterest. Durng the past year, many proposals for the post-hussen consttuton of Iraq have dentfed bcameralsm as an mportant ngredent n buldng a stable democracy 3
n a naton beset wth strong relgous and ethc cleavages. 4 Whle clearly some form of federalsm and perhaps over-representaton of ethc and relgous groups wll be a vtal ngredent n a democratc Iraq, our revew questons whether bcameralsm s a necessary or even desrable addton to the mx. 4 These proposals and the postons of dfferent Iraq groups s dscussed n Publc Internatonal Law and Polcy Group (003). 33
Bblography Adran, Charles. 960. State and Local Governments. New York: McGraw-Hll. Ansolabehere, Stephen, Alan Gerber, and James Snyder. 00. Equal Votes, Equal Money: Court Ordered Resdtrctng and Publc Expendture n the Amercan States, Amercan Poltcal Scence Revew 96(4):767-778. Ansolabehere, Stephen, James Snyder, and Mchael Tng. 003. Barganng n Bcameral Legslatures: When and Why does Malapportonment Matter? Amercan Poltcal Scence Revew 97(3): 47-48. Austen-Smth, Davd and Jeffrey S. Banks. 996. Informaton Aggregaton, Ratonalty, and the Condorcet Jury Theorem Amercan Poltcal Scence Revew, 90():34-45. Atlas, Cary M., Thomas W. Gllgan, Robert J. Hendershott, and Mark A. Zupan. 995. Slcng the Federal Government Net Spendng Pe: Who Wns, Who Loses, and Why. Amercan Economc Revew 85:64-69. Baron, Davd P., and John A. Ferejohn. 989. Barganng n Legslatures. Amercan Poltcal Scence Revew 89: 8-06. Bnder, Sarah A. 999. The Dynamcs of Legslatve Grdlock, 947-996 Amercan Poltcal Scence Revew 93(3):59-533. Bnder, Sarah A. 003. Stalemate. Washngton D.C: Brookngs Insttuton Press. Bottom, Wllam P., Cheryl L. Eavey; Gary J. Mller; Jennfer Ncoll Vctor. 000. The Insttutonal Effect on Majorty Rule Instablty: Bcameralsm n Spatal Polcy Decsons. Amercan Journal of Poltcal Scence 44(3):53-540. Brady, Davd W., and Crag Volden. 997. Revolvng Grdlock: Poltcs and Polcy from Carter to Clnton. Boulder, Colo.: Westvew Press. Chou, Fang-Y and Lawrence Rothenberg. 003. When Pvotal Poltcs Meets Party Poltcs. Amercan Journal of Poltcal Scence 47(3): 503-5. Cox, Gary W. and Mathew McCubbns. 003. Legslatve Levathan Revsted. Unpublshed book manuscrpt. Cox, Gary W. and Rchard McKelvey. 984. A Ham Sandwch Theory for General Measures. Socal Choce and Welfare :75-83. Crémer, Jacques and Thomas R. Palfrey. 999. Poltcal Confederaton. Amercan Poltcal Scence Revew 93():69-83. 34