Presidential Vetoes in the Early Republic: Changing Constitutional Norms or Electoral Reform? Nolan McCarty Woodrow Wilson School Princeton University

Transcription

1 Presidential Vetoes in the Early Reubli: Changing Constitutional Norms or Eletoral Reform? Nolan MCarty Woodrow Wilson Shool Prineton University Revised August 2002

2 1. Introdution Today it is generally reognized that the residential veto lays an imortant role in the legislative roess. The threat, either imliit or exliit, that a resident will refuse to affix his signature to legislation is believed to influene oliy outomes (MCarty and Poole 1995, Krehbiel 1998, Cameron 2000). Beyond its diret effet on oliy, veto olitis is also believed to lay an imortant role in defining artisan oliy onflits for the eletorate (Gilmour 1995, 2001; Groselose and MCarty, 2000). However, the work of many historians and olitial sientists suggests that the veto develoed these modern funtions relatively late. The author of a reent study of the veto makes the following haraterization: As with virtually every other ower enumerated in the onstitution, the veto ower evolved over time as exerimentation, irumstane, and umulative reedent ombined to give the ower its atual shae, eseially as to its frequeny, and other onditions of use (Sitzer, 1988). The basis of suh laims is that the veto was used saringly, if at all, during muh of the 19th entury. Perhas the most ommon exlanation of the infrequeny of vetoes fouses on norms surrounding the onstitutionally legitimate exerise of exeutive ower. Many sholars have argued that early residents and legislators viewed the veto rerogative very narrowly (Binkley 1947, Blak 1976, Remini 1967, Sitzer 1988, Skowronek 1993, Watson 1987, White 1956). Under the norms of the day, the veto was not onsidered a legislative ower of the resident. Rather, the dotrine of searation of owers held that the veto was rimarily an exeutive or judiial funtion. Its exeutive role was two-fold. First, it roteted the resident from enroahments of the legislature. The seond was to give the resident the oortunity to rejet bills so oorly or hastily 2

3 drafted that they ould not be effetively exeuted. Alternatively, the veto's judiial dimensions were suh that it was to rovide an oortunity for the resident to revent the enatment of unonstitutional laws. Aordingly, the veto ould only legitimately be alied to legislation that was learly unonstitutional, enroahed on exeutive ower, or was badly drafted. The modern onetion of the veto -- a tool to defeat or modify legislation that the resident finds objetionable on oliy grounds -- was onsidered to be antithetial to the searation of owers, reublian government, and legislative suremay. Thomas Jefferson s advie to President Washington over a bill hartering the Bank of the United States seems to suggest suh a restrited view: unless the President's mind on a view of everything whih is urged for and against this bill, is tolerably lear that it is unauthorized by the Constitution; if the ro and on hang so even as to balane judgment, a just reset for the wisdom of the legislature would naturally deide the balane in favor of their oinion. (Quoted in Bass 1972 and Watson 1987). Adherents of this view suggest that these norms did not breakdown until the administration of Andrew Jakson. Two of his vetoes, the bill reating the Maysville Road and legislation to reharter of the Bank of the United States, are laimed to reresent the first two serious violations of the onstitutional rosritions. In the words of Jakson s biograher Robert Remini: [Jakson] strethed the veto ower and laimed the right to blok legislation for reasons of oliy or exedieny rather than onstitutionality. Thereafter, Congress arefully onsidered the residential will in all legislation in order to avoid a ossible veto. Next, he broadened the olitial ower base of the resideny by taking the Bank issue to the eole and winning an overwhelming vitory in the residential eletion of Thereafter, Jakson did not hesitate to laim an augmentation of exeutive authority by virtue of his vitory at the olls...jakson widened the resident's resonsibility to inlude all the eole, a neessary aknowledgement if he was to draw olitial strength from their suort when he tangled with Congress. Moreover, he advaned the onet that the President is 3

4 the diret (and sole) reresentative of the eole, a revolutionary onet for its time. ( ) Thus, Jakson not only bloked legislation that he oosed as a matter of oliy, but he also asserted an absolute right to do so on the basis of his reresentation of the eole. Reently, Stehen Skowronek has written that these ations made a mokery of the remier oerating rinial of the Jeffersonian regime -- exeutive deferene to the legislature ( ). This onstrained view of the exeutive veto has ontinued to lay a role in modern jurisrudene of the searation of owers. In his artial dissent in Bukley v. Valeo, Justie White wrote that the veto s aim was not another hek against oor legislation but to shore u the Exeutive Branh against... the overweening ower of legislators. 1 Nevertheless, other sholars have questioned the saliene of these onstitutional norms in rosribing the aggressive use of the veto ( Bass 1972, Fisher 1985, Jakson 1967, and Moe 1987). They rejet the notion that vetoes grounded in oliy disutes were ontrary to the intent of the framers nor inimial to the true views of 19th entury residents and legislators. Vetoes were rare due to a number of other fators suh as olitially-weak residents, the availability of other methods of residential influene, and a simle lak of legislative ativity. Bass suggests...if the veto was used saringly, other fators than oneiving the veto as a limited ower had their influene. With fewer demands for legislation, bills were drafted with greater are and onsideration than revailed in later years, giving early residents less ause to use the negative. While the quality of bills was higher, quantity was lower, lessening the need to emloy the veto. (. 89) 1 Bukley v. Valeo, 424 U.S (1976). 4

5 To make the ase that onstitutional norms layed but a small role, these sholars oint to early vetoes that were not justified on either grounds of onstitutionality or of legislative enroahment. They oint to Washington's rejetion of a military redution bill and Madison's rejetion of the national bank harter as vetoes justified by few if any onstitutional issues. 2 Furthermore, they argue that many of the onstitutional objetions aomanying other veto messages were little more than window dressing for underlying oliy objetions. Desite substantial sholarly attention, this debate is far from resolved either in terms of why the veto was rarely used or what imliations, if any, we might draw about exeutive-legislative relations in the early reubli. Emirial work to date has foused exlusively on analysis of residential veto messages and the statements, often selfserving, of residents and legislators. Neither the norms hyothesis nor any of its alternatives has been subjeted to rigorous tests. Rather than draw testable inferenes from the underlying hyotheses, the debate has foused solely on the signifiane of ereived deartures from the osited norms. In this aer, I add to this debate by roviding my own argument about why usage of the veto hanged over time. This argument is loosely based on my revious work with Tim Groselose on the role of an attentive eletorate on olitial bargaining. In that aer, we argue that the rimary ause of residential vetoes (at least on imortant legislation) is the inentive of legislators and residents to use veto bargaining to define issue ositions before the eletorate. I argue here that hanges in the eletoral environment of the resideny, in artiular inreasing oular artiiation and 2 However, those who argue the entrality of onstitutional norms have ountered that these vetoes were justified by exeutive onerns over how the oliies would be imlemented, thus falling well within the 5

6 delining influene of olitial elites in residential eletions, enhaned the inentives to engage in blame game olitis. As a result, the use of the veto inreased, and it beame inreasingly tied to eletoral olitis and artisan oliy onflits. Thus, while agreeing that the nature of veto usage hanges in the 1820 s, I argue that these hanges resulted from a demoratization of the residential offie, not the breakdown of onstitutional norms. However, I also argue that the hanges that ourred were limited only to the frequeny of veto usage, not to the role of the veto in shaing legislative outomes. Below I rovide both qualitative and quantitative evidene that re-jaksonian residents had about as muh imat on legislative outomes as subsequent hief exeutives. The aer roeeds as follows. In the next setion, I rovide a brief history of the veto from the olonial era to the ratifiation of the Constitution. This disussion is designed both to rovide bakground and to assess laims that the Constitution rosribed vetoes based on oliy disagreements. Setion 3 reviews the evidene about the justifiations rovided in veto messages. As will beome lear, this evidene is quite mixed, suggesting the need for more rigorous emirial tests. Setions 4 and 5 rovide a framework for roviding a ounterfatual of re-jaksonian veto usage under more ontemorary atterns of behavior. To this end, Setion 4 reviews reent theoretial work on the veto, inluding Groselose and MCarty s blame game model. In setion 5, the reditions of these models are used to seify an emirial model of veto usage from Three findings of this setion rovide key suort of my arguments. First, veto usage is systematially lower before 1829 even when ontrolling for the fators that have reviously been urorted to exlain the differene. Seond, the atterns of veto usage are systemially different rior to It is not simly fewer vetoes, but those range of legitimate uses. 6

7 that our are inonsistent with the reditions of the emirial model. Finally, and most imortantly, the emirial model illustrates the imortane of eletoral olitis in generating vetoes in the modern eriod. After addressing the issues surrounding the usage of the veto, I turn to an analysis of the onsequenes for residential influene on legislation. A reasonable interretation of the onstitutional norms hyothesis is that residents had less imat on legislative outomes rior to Jakson's vetoes than after. I test this hyothesis in two ways. First, I losely examine the assage of the Missouri Comromise over Monroe s resumed oosition, an examle sholars have often used to demonstrate exeutive weakness eretuated by the onstraining norms. Seond, I develo an emirial model of the robability that legislation oosed by the resident asses the House of Reresentatives. If the norms hyothesis is orret, this robability should have been higher before It was not, suggesting that only the frequeny of veto usage hanged, not its effet on oliy outomes. 2. Exeutive Power and the Constitution The exeutive veto was not a oular institution during the olonial era. As Gerhard Caser (1997) has written, the most notable feature of revolutionary state onstitutions was the deendene of the exeutive on the legislature. In most states, the exeutive was hosen by the legislature for very short terms in offie and given authority narrowly onfined to administrative matters. A number of hyotheses have been ut forward as to why these onstitutions so severely onstrained the exeutive. First, Amerian olonists were long frustrated by the ereived abuses of royal governors in 7

8 using vetoes to extrat onessions from olonial legislatures, inluding inreases in their ersonal salaries (Moe 1987; Watson 1987). In addition to those vetoes, olonial legislation was also subjet to a veto (atually reeal) by the Board of Trade and Plantations of the Privy Counil in London (Moe, 1987). This negative was used on almost 500 olonial ats between (Russell 1915). Given these frustrating exerienes, the olonists roved extremely reetive to arguments in favor of legislative dominane. Thus, an anti-exeutive sentiment manifested itself both in rovisions for weak or non-existent exeutives in the new state onstitutions and in the lak of a national exeutive under the Artiles of Confederation. However, this era of legislative dominane did not last very long. In 1776, South Carolina gave its governor an absolute veto, but those rovisions were reealed two years later (Thore 1909; Watson, 1987; MDonald 1994). In New York s 1777 onstitution, a qualified veto was granted to a ounil of revision onsisting of the governor and members of the state judiiary. In 1780, Massahusetts adoted the form that was later to revail at the Federal Convention, a qualified veto subjet to 2/3's override in both houses. While those rovisions generated a large amount of ontroversy during the ratifiation of that harter (MDonald 1994), they beome a bluerint for onstitutional revisions in other states. Aording to historian Gordon Wood, The Massahusetts onstitution of 1780 not only had a diret influene on the New Hamshire Constitution adoted in 1784 but it seemed to many in the 1780's to limax a seond wave of state onstitutional onstrution. In its struture at least, it ame to reresent muh of what reformers in other states desired for their own onstitutions -- a strengthening of the governor at the exense of the legislature, artiularly the lower house. The exeutive ower, as the New Hamshire onstitution of 1781 delared in defense of its roosed onstitution, had beome the ative rinial of all governments: it is the soul, and without it the body oliti is but a dead orse. (. 435) 8

9 Delegates to the onstitutional onvention were aware of the defiienies aused by weak exeutives, and they generally agreed that stes should be taken to reate a more owerful and indeendent national exeutive. Given this onsensus, the key debates about the exeutive ower were less about ends than about means. The key roblem they faed was how to reate an offie that would be seen as legitimate by an eletorate who viewed exeutive ower with some susiion, yet at the same time would be emowered to vigorously exeute the law. The veto rovisions were an imortant omonent of this balaning at. At one extreme, rovisions ould be made too strong and beome a soure of oosition at the ratifying onventions. Others also feared that rerogatives might be made so extensive that residents would be unwilling to use them for fear of ubli ensure. However, diluting the rovisions would transform the resideny into little more than a legislative ageny. At the onvention, debate on the veto foused rimarily on two key issues. The first foused on the ease with whih Congress ould override the veto while the seond onerned whether the resident should be able to at unilaterally or whether the rerogative should be given to a ounil of revision of whih the resident was but one member. The issue of the suermajority for legislative override was relatively straightforward, itting advoates of stronger exeutives against those more fearful of exeutive ower. While Alexander Hamilton argued for an absolute veto, the majority sentiment ranged from a 2/3s to 3/4s override deending on the urrent status of other rovisions for residential owers and tenure in offie. While the onvention was working under the assumtion of a shorter residential tenure, eletoral deendene on Congress, and an ative role of the Senate in exeutive matters, the 3/4s override was 9

10 more oular. When these rovisions were eventually abandoned in favor of longer terms, eligibility for reeletion, eletoral indeendene, and a smaller role for the Senate, a onsensus for a 2/3s override was emented. This redution in the override suermajority was intended to offset the dramati exansion of exeutive ower over the last few weeks of the onvention. The debate over a roosal to adot a New York-insired ounil of revision rovides some insight into the intentions of the founders. A roosal that the veto rerogative be shared with the Sureme Court was made at a time when the onvention had tentatively agreed to a qualified veto whih the resident ould invoke unilaterally. However, this was not an attemt to weaken the exeutive. This measure was suorted by two advoates of a strong exeutive, James Madison and James Wilson. Their advoay was based on the remise that veto was an imortant instrument of exeutive influene over legislation (Rakove 1997). Suorters of the roosal argued that the veto was not simly intended to resist legislative enroahments. Virginia s George Mason argued that it must be used to revent unjust and erniious laws. Governuer Morris added that the veto was neessary to make the resident the guardian of the eole, even of the lower lasses, against legislative tyranny, against the Great & the wealthy who in the ourse of things will neessarily omose the legislative body (both quoted in Rakove 1997). Their suort for the ounil was due to their doubts that a single individual would have the olitial wherewithal to oose legislative majorities. A ounil, on the other hand, would be better ositioned to generate olitial suort for its hallenges to the legislature. This would make the veto more redible as a ounterbalane to legislative dominane. 10

11 Nevertheless, the oonents of the ounil veto did not fous their ritiism on the breadth of the veto ower advoated by Madison, Wilson, and Morris, but they onentrated rimarily on whether judiial artiiation was onsistent with the searation of owers (Rakove 1997). They argued that artiiation in the drafting of laws omromised the judiial funtion of interreting the laws. Eventually, Madison and Wilson's motion fell 4 states to 3 with 2 divided. Given the tenor of the debate, it is reasonable to onlude that the rejetion was of the form of the veto rather than its soe. Imortantly, no effort was made at the onvention to exliitly narrow the soe of the exeutive veto. Furthermore, the ratifiation debates also do not suggest a onsensus for a limited soe. Contrary to the eretion that exeutive ower was softedaled during ratifiation, Hamilton asserted boldly that the veto...not only serves as a shield to the Exeutive, but it furnishes an additional seurity against the enation of imroer laws. It establishes a salutary hek uon the legislative body, alulated to guard the ommunity against the effets of fation, reiitany, or of any imulse unfriendly to the ubli good, whih may haen to influene a majority of that body. Unless one asserts that Hamilton maintained a one-to-one orresondene between imroer and unonstitutional, this statement seems to oen the soe entirely to the resident s ersonal views of ubli oliy. While muh has been made of Hamilton s argument that the veto would be used as saringly as had the royal rerogative of the British king, he also asserted that Instead of an absolute negative, it is roosed to give the Exeutive the qualified negative... This is a ower whih would be more readily exerised than the other. A man who might be afraid to defeat a law by a single VETO, might not srule to return it for reonsideration; subjet to being finally rejeted only in the event of more than one-third of eah house onurring in the suffiieny of his objetions... In roortion as it is less likely to offend, it would be more at to be exerised; and for this reason in may in ratie be found more effetual. 11

12 In site of the role of legislative dominane in the dotrine of the oonents of the onstitution, there was little if any diret ritiism of its veto rovisions (Moe 1987). While the Anti-Federalists found many faults with the roosed onstitution, the rovision that allowed an exeutive to hallenge the judgments of as many as 2/3 of the legislature was not one of them. 3. Vetoes in the Early Reubli Although onstitutional rosritions on veto usage did not arise from the framing and ratifiation of the Constitution, it is not yet ossible to dismiss the norms hyothesis omletely. An alternative exlanation is that exeutive restraint develoed as a onstitutional rinile as reedents were established during the first residential administrations. It was well understood by President Washington and others that his behavior in offie was likely to determine how future residents used their rerogatives. Even suh trivialities as the use of titles and the roer rotools for soializing were heavily srutinized for the ways in whih they would effet future residents. The use of the veto was no different. There were numerous debates within Washington s abinet about the roer soe of the veto. Most members wanted him to use the veto aggressively to establish a reedent. Jefferson, erhas the least symatheti to a strong resideny, went so far as to suggest that he seek out bills to veto. In the end, Washington only used the veto twie and the first one oming three years into his term (see Table 1). The first bill onerned 12

13 aortionment of the House of Reresentatives. In Washington s view the lan was learly unonstitutional as it was so heavily biased towards the northeast that it onstituted a de fato reeal of the 3/5 s Comromise. That the first veto ame only after three years and was generally agreed to be a flagrant violation of the onstitutional bargain may have heled to establish the rinile of a veto ower limited in soe. However, Washington s seond veto message ontained little onstitutional analysis. When Congress voted to redue the army by mustering out two omanies of light dragoons, Washington objeted on several grounds. Some of these may be safely ategorized as relatively minor administrative details suh as legality of aying the troos between their legal and atual disharge dates. However, Washington ointedly argues that these omanies were needed to seure the frontier against Native Amerians, a lear oliy disagreement. As Table 1 oints out, the other vetoes of the era were also justified by a mix of objetions. Some learly dealt in onstitutional issues suh as the vetoes onerning hurh-state relations and the veto of a lan to allow Sureme Court justies to try ases in distrit ourt. Yet others were based rimarily on oliy objetions. Madison vetoed the harter for the Seond Bank of the United States beause the mehanisms for olitial ontrol were deemed insuffiient and a naturalization law whih rovided too many inentives for fraud. Categorizing other vetoes is even more diffiult, beause it is hard to disern whether the stated onstitutional objetions are sinere or whether they essentially reakaged oliy differenes. The two internal imrovements vetoes raise onstitutional objetions that were inonsistent with settled dotrine. Madison vetoed an internal 13

14 imrovements bill on his last day of offie with the objetion that Congress laked the ower to romote suh rojets. However, Jefferson, with Madison in the abinet, had earlier signed the bill authorizing the building of the Cumberland Road and Madison as resident signed aroriations for building extensions of the road (Bass 1972). When Monroe later vetoed legislation for federally olleted tolls on the Cumberland Road, his veto message hinged on a tortured distintion between Congress s right to aroriate money for roads and its right to administer internal imrovements. That Madison and Monroe felt omelled to use onstitutional language might suggest to some the imortane of the restritive norms. On the other hand, it may simly show how easy it was to gussy u oliy disagreements with the language of onstitutionalism. With standards for onstitutional analysis so low, the norms hardly seem onstraining. Insert Table 1 Here More evidene bearing on the norms hyothesis an be obtained by looking at ongressional resonses to the vetoes. For muh of the 19th entury, Congress felt obliged to take reorded votes on motions to override on every regular veto. If legislators sought to maintain a norm of legislative suremay, we ought to see this refleted in greater suort for the motion to override the veto than for the original bill. 3 From the last olumn of Table 1, we find sant evidene that legislators may have sought to unish the resident by overriding his veto. In fat, there is but one ase in whih the motion to override get even 50% suort, suggesting that many members who suorted legislation voted against the override motion. Only one override vote seems roughly onsistent with 14

15 the norms hyothesis -- Washington's veto of the Army redution bill. However, while there was no reorded final assage, an amendment to restore the dragoon omanies failed Thus, Washington's 40% suort on the override was almost double the 21% suort for restoring the dragoons rior to his veto. So it seems unlikely that Congress tried to use its override authority to enfore a norm against oliy vetoes. 4 A final iee of evidene also suggests the absene of the restritive norms. If the ideologial ommitment to legislative suremay that underlay the norms was strong, one would exet to see it refleted in the onstitutions of new states and in the revisions to existing state harters. To the extent that itizens wanted to onstrain the legislative influene of their governors, they should have withheld the veto or restrited its soe. The evidene from the onstitutions adoted during the early 19th entury omletely undermine this view. Figure 1 rovides the number of states in whih the governor had some form of veto ower and the number of states in whih he did not. 5 Insert Figure 1 about here Note that the modal attern shifted dramatially towards the veto over this eriod. In fat, no state droed its veto rovisions. From 1800 to 1850, 34 state onstitutions were adoted, inluding those of new states and revisions of old harters. Only two states, Illinois and Ohio, entered the union with a onstitution laking an exeutive veto, while four states (Maryland, Delaware, Rhode Island, and Virginia) adoted new 3 Krehbiel (1999) also uses omarisons of final assage and override votes to measure residential influene in the ost-ww II era. 4 The ublished House debates on the motion to override foused solely on the merits of Washington's objetions, not his right to veto legislation (Annals of Congress, 4th Congress, 2nd Sess ) 5 This data is olleted from Thore (1909). 15

16 onstitutions before 1850 without adding a veto. Not a single state onstitution restrits the soe of the veto to onstitutional or administrative objetions. While the reeding evidene suggests that onstitutional norms were robably not very imortant, it is also lear that early residents did not use the veto very often. There were only ten vetoes rior to 1829, and half of these were by a single resident. Nevertheless, it would be remature to onlude that residential or ongressional behavior differs substantially from urrent atterns. There ould be many fators whih aount for the auity of vetoes. First, legislative outut of Congress was very low, roviding few oortunities for the veto. Presidents tyially had artisan majorities in Congress during this eriod imlying that there would be few disagreements. Therefore, the only way to establish that atterns of veto olitis hanged is to onstrut a ounterfatual of what veto usage would have been if residents and legislators had layed the modern veto game. 4. The Modern Veto Before omaring re-jaksonian veto usage to that redited by ontemorary behavioral atterns, it is imortant to seify learly what those atterns are. Unfortunately, there is little sholarly onsensus on the auses of veto ativity. However, the develoment of formal models of residential-legislative bargaining has been a fertile area for researh. Many of these models make quite exliit reditions about veto usage. In the analysis that follows, these theories will lay dual roles. First, they rovide guidane on how to seify a reditive model whih an be used to test to see if re- Jaksonian veto behavior is anomalous. More imortantly, they also rovide a set of 16

17 otential arguments as to why suh hanges may have taken lae. Below I review a number of theoretial arguments about veto usage in order to reate measures for the reditive model and to rovide otential hyotheses for why usage atterns may have hanged. 4.1 Models of the Exeutive Veto In this setion, I review a variety of models of the residential veto in order to derive reditions about veto usage and residential influene in the legislative roess. These reditions will hel us to sort out ometing laims about the develoment of the exeutive veto. I have attemted to kee tehnial disussions to a minimum (and lae them aroriately in the aendix) Preliminaries To kee the models relatively simle, I abstrat from biameralism and from other features of the internal legislative roess to model the legislature as a unitary ator C. 6 These assumtions allow me to model the veto as bilateral bargaining between C and the resident, P. Also, all of the models resented in this hater will fous on olitial bargaining over a one dimensional oliy sae. 7 For simliity of exosition, I assume that C and P evaluate oliy alternatives solely on their roximity to their most referred oliies 6 I refer to think of C as the median member of the hamber. Adherents of other theories of legislative deision making an interret C aording to their own onsiene (arty leader, majority arty median, et.) Suh elesiastial disutes need not detain us here. 7 For models relaxing the assumtions of unidimensionality, see MCarty, Miller and Hammond, Tsebelis. The assumtion of unidimensionality is not all that onsequential for my uroses. Most of the reditions derived would hold for a multidimensional model so long as the legislature is treated as a unitary ator. 17

18 whih I denote and, resetively. Thus, the oliy utility funtions for C and P are x and x given any alternative x. In some of the models disussed in this hater, an additional layer will be relevant for determining whether or not the resident s veto will be overridden. Following Krehbiel (1999), O is the override ivot and has an ideal oint o. This ivot is defined as the legislator losest to the resident for whom exatly 1/3 of the legislature has ideal oints either lower or higher than hers. 8 I assume that O s utility funtion has the same form as C and P. Finally, all of the models in this hater are single shot affairs in whih the legislature makes a single roosal and the resident makes an u or down deision about aeting it or vetoing it. Formally, C makes a roosal b to hange some status quo or reversion oliy q. If P aets the offer, b is the final oliy and the game ends. If b is vetoed, a vote on a motion to override ours. If O suorts the motion, the bill is suessful and b is the new oliy. If O does not suort the motion, it fails and q remains the oliy The Comlete Information Model A tyial oint of dearture for analyzing the effets of exeutive veto ower is the assumtion that all ators are erfetly informed about the referenes and ations of all other layers. Under these assumtions, there is no unertainty about how the resident or override ivots will resond to a legislative roosal. Therefore, C an hoose b otimally given its orret exetations about the future. 18

19 Given that there is no unertainty, the game an be solved bakwards by onsidering the deisions of the override ivots on an override motion. Clearly, eah ivot will vote to override if she refers b to q. Thus, we an define a set of bills B ( q ) that O refers to q so that an override motion would always be suessful. This set is given in figure 2. Given the assumtions about referene symmetry, this set is just [ q,2o q] or [ 2 o q, q] roosal in B ( ) O deending on whether or not o> q. As long as C makes a q, the veto will be overridden and the roosal beomes the new oliy. Insert Figure 2 about here O Having determined whih bills an override a veto, I an now omute whih bills will be aeted by the resident. First of all, it is reasonable assume that the resident will aet any bill whih would have been overridden. 9 Thus, the bills in BO ( ) q will not be vetoed. Nor will he veto any bills that he refers to the status quo. Formally, let BP ( q ) be the set of bills for whih b q. Under my assumtions, BP ( q ) is given by either [ q,2 q] or [ 2 qq, ] or not > q as in figure 3. Insert Figure 3 about here deending on whether 8 If the resident has a low ideal oint, O is the 33 rd erentile legislator. Alternatively, if the resident has a relatively high ideal oint, O is the 67 th erentile legislator. 9 There are a number of rationales for this assumtion running from the ratial to the very tehnial. The easiest is to note that with any infinitesimal ost of being overridden, the resident would stritly refer to sign any bill destined to ass over his veto. The tehnial reason is that any other assumtion about the resident s behavior on suh a roosal is inonsistent with a subgame erfet Nash equilibrium. 19

20 Sine C is erfetly informed about the other layers referenes, she knows for ertain that any bill in either BO ( q ) or BP ( ) q will be suessful. She need only ass her favorite bill from these sets. If q is her favorite bill in these sets, she should not legislate at all. 10 Thus, it is straightforward to omute C s otimal behavior. In the aendix of this hater, I formally resent Proosition 1 whih fully desribes the roosal making and veto behavior in the subgame erfet Nash equilibrium to this veto game. This roosition generates a number of seifi reditions that will be useful in understanding the develoment of the veto. The first result deals with the usage of the veto. Sine the resident never vetoes any bill that an be overridden and the legislature never makes any roosal that will be vetoed, all we should observe in equilibrium are assed bills or inative legislatures. Predition 1: If all ators are erfetly informed about the referenes of all other ators, vetoes should not our. While seemingly simle, redition 1 has some very owerful imliations. Most imortantly it demonstrates that it is imossible to infer anything about the soe of the veto ower from the frequeny of its use. In this very simly model, the veto moves oliy away from that referred by C yet we never see it used. Thus, one annot draw the inferene that if the veto is never used, it is imotent. It is artiularly damning to any inferene about norms in the early reubli whih is based solely on the infrequeny of veto usage. 10 Again small any small osts of legislating will lead her to stritly referene not a assing a bill destined 20

21 The seond redition, whih I resent grahially, is that the exeutive veto has oliy onsequenes even if it not used. In Figures 4 and 5, I resent the equilibrium oliy outomes for all status quo oints and a oule of different referene onfigurations. These oliy outomes with the veto an be omared with the result that would our without an exeutive veto, * b =. Note that for status quo oints around and o, the veto moves oliy away from towards the ositions favored by the resident and the override ivot. Insert Figures 4 and 5 Predition 2: Under the exeutive veto, oliy may be resonsive to the referenes of the resident or the override ivot. The roosition also leads to some imortant reditions about residential suort for legislation. Sine in the absene of the veto oliy will be determined solely by C s referenes, often legislation will ass that is oosed by the resident, i.e. legislation he would veto if he ould. Of ourse, whether or not the resident will get rolled in this way deends on the osition of the status quo, so Figure 5 indiates the set of q for whih oliy will move ontrary to the resident s referenes. However, if the resident an utilize the veto, the irumstanes under whih he gets rolled are redued dramatially. In fat, if the resident s ideal oint lies between that of C and O (as in Figure 2), he never gets rolled. If the resident s referenes are more extreme than the override ivot s (as in Figure 3), he may get rolled when veto-roof legislation is to be vetoed. 21

22 assed. However, this will our far less often than when P laks the right to veto legislation. 11 Insert Figure 5 Here Predition 3: The robability that legislation oosed by the resident asses is lower when he has a veto. 12 While this too is a fairly obvious redition, it also omes in handy in assesses ometing exlanations for the develoment of the residential veto ower. If as others have argued, the resident was onstrained from using the veto on oliy grounds, we would exet to see muh higher residential roll rates rior to the breakdown of this norm than afterwards. We would also exet to see roll rates resonding to variations in the override ivot only after the establishment of the oliy veto. I test this imliation of the norms hyothesis in setion 5.2. However, while these reditions will turn out to be quite useful, obviously, a model that redits that vetoes will not our will not take us very far. I now turn to some models in whih vetoes do our and examine their imliations for ossible lues about exeutive-legislative relations in the 19 th entury The Inomlete Information Veto Model 11 The aendix ontains exat onditions for residential rolls for both models. 12 By stating the redition in terms of the robability that the resident will be rolled rather than the set of status quos under whih a roll ours, I am imliitly assuming that the distribution of status quo oints does not vary aross the models. As Krehbiel oints out, this assumtion an be roblemati sine different models of olletive hoie redit different dynami evolution of status quos. 22

23 If one wants to exlain vetoes one must disense with at least one of the assumtions underlying the legislative agenda ontrol model. While the model resented in the last setion has a number of outrageously restritive assumtions, it turns out that very few of them are onsequential in the redition that vetoes should not our. One exetion is the assumtion that C has omlete information about the referenes of P and O. When there is suh unertainty, vetoes may our beause the legislature overestimates its ability to extrat onessions from the resident or the override ivot. Relaxing the assumtion of omlete information has been the starting oint for most of the reent work on veto bargaining (Matthews 1989, MCarty 1996, and Cameron 2000). To resent the basi flavor of these models, I onsider only a model without an override ossibility so that q remains the oliy in the event of a veto. To ature the unertainty that C faes about the resident s referenes, I assume that she believes that the resident is one of two referene tyes, a moderate with ideal oint m or an extremists with ideal oint e. I assume throughout that e< m<. Following usual ratie, I assume that C has beliefs about the resident s tye whih are ommon knowledge. 13 Let π be the robability that P is the extreme tye. The main imliation of the unertainty about referenes is that C no longer knows whih bills the resident will aet and whih he will veto. To see this, onsider Figure 6 where I assume that q < e. Here the set of bills that e is willing to aet over the status quo is a subset of those m is willing to aet. Thus, C an extrat a better bill from m than from e. C s hoie therefore is to deide whether to roose a bill that both tyes will aet like b e, or be more aggressive and roose b m whih is muh loser to 13 P knows these beliefs, C knows that P knows, et. 23

24 her ideal oint but might be vetoed. Clearly, this hoie deends on C s beliefs about P. If π is high, C will likely be deterred from the aggressive roosal and will make the safer one. On the other had if π is low, b m is a more attrative roosal, but on the off hane that the resident is the extreme tye it will be vetoed. In the aendix to this hater, I alulate the neessary onditions for a veto. For the referene onfiguration in anel a, I show that C will make the risky roosal (ossibly generating a veto) if and only if: m e π < m q (1.1) However, neessary onditions hange as moves loser to m, as in anel b. Here C s best risky roosal is her ideal oint. This fat alters the neessary ondition somewhat to: + q 2e π < q (1.2) It an easily be shown that (1.2) is more strit than (1.1) imlying that a veto is less likely to our. This is rimarily due to the fat that the oliy onsequenes of the hoie between roosals is muh lower without reduing the risk of a veto. Finally, note that in the extreme ase of anel where C s ideal oint is aetable to both tyes, no veto will our. These results lead us to redition number 4: Predition 4: Vetoes will be more likely when the exeted differene between the ideal oints of P and C is larger 24

25 While I will not undertake it, an extended model with a veto override rodues a arallel result. Predition 5: Vetoes will be more likely when the exeted differene between the ideal oints of O and C is larger. Insert Figure 6 about here The Blame Game Veto Model A more reent model argues that vetoes are less a rodut of legislative unertainty than they are of blame game eletoral olitis. In my work with Tim Groselose (2001), the legislator agenda setter an use its roosal ower to signal that the resident has oliy views whih are out of ste with the voters. Vetoes are generated when the agenda setter gets a larger ayoff from signaling that the resident has extreme referenes than she reeives from enating a new oliy. Thus, it is the eletorate s unertainty about the resident that is ruial, not that of the legislators. To illustrate a simle version of this model, onsider a new ator V, the voter. I assume that V also has linear referenes and an ideal oint v. Following the notation of the last setion, V believes the resident is tye e with robability π and tye m otherwise. I will fous on the ase where e< m< v. I assume that the voter evaluates the resident based on the exeted distane of the resident to her ideal oint. Therefore, the voter s evaluation is just ( 1 π) v πe m (1.3) 25

26 An imortant feature of this model is that P and C are how muh exeted utility V gets from the resident s osition. Clearly, the resident stands to benefit from a higher voter evaluation and would therefore like to engage in strategies designed to minimize π. In artiular, he may be willing to trade-off oliy gains for olitial oints. To ature these trade-offs, I assume that the resident weights oliy by P and the voter evaluations by 1. Therefore, the resident s ayoffs may be written as P ( 1 ) ( 1 ) x v πe π m (1.4) P The model an allow variation in terms of whether C benefits or loses from favorable evaluations of the resident. Let δ { 1,1} be an indiator of how the voter s evaluation effets C. If δ = 1, C benefits from a high evaluation. This might be the situation if C and P belong to the same olitial arty. Alternatively, if δ = 1, C refers lower evaluations of the resident and will take ations to attemt to raise π. Assuming that C is Congress s oliy weight, its utility funtion may be written as ( 1 ) ( 1 ) x δ v πe π m (1.5) An imortant assumtion of this model is that while V is uninformed about P s referenes, C is fully informed. Therefore, C may be able to redibly ommuniate its information through its hoie of bill. Similarly, the resident s deision of whether to veto or aet artiular roosals may also rovide information to voters about his referenes. Rather than a full analysis of the model, my uroses here are satisfied simly by 26

27 seifying onditions under whih vetoes our in equilibrium. Therefore, I will onentrate on the neessary onditions for the existene of a ure strategy equilibrium where C indues a veto from tye e but rooses aetable legislation to tye m. This is the only tye of searating equilibrium that rodues vetoes. 14 The first ondition, whih is stated and roven as Lemma 2 in the aendix, is that vetoes do not our in equilibrium unless Congress refers lower evaluations of the resident, i.e. δ = 1. Sine this situation is most likely to be reresented by divided ontrol of the branhes, this result leads to the following substantive redition: Predition 6: Vetoes are more likely during divided ontrol of the resideny and Congress. In addition to δ = 1, Proosition 3 in the aendix shows an equilibrium in whih vetoes our exists if and only if the following two onditions hold: ( 1 π)( m e) > ( 2e 2q) (1.6) 2 > π (1.7) These onditions rodue a number of reditions about veto ourrene. 15 First, note 14 There are some semi-ooling equilibria in whih C indues a veto with some robability. These turn out to be very hard to haraterize so I ignore them here. Thus, the interretation of the onditions disussed below is that they are the ones for whih a veto is most likely. 15 These are onditions are neessary for the ase of 1 2 ( ) > m q+ m e. Different ositions of result in slightly modified but qualitatively similar onditions. See the aendix for the other onditions. 27

28 that (1.6) annot be satisfied if m = e or π = 1. Thus, voter unertainty about the resident s referenes is ruial. Without this unertainty, orhestrating a veto has no signaling value to C so she might as well make aetable roosals to both tyes. Predition 7: Voter unertainty about the resident s referenes is neessary for equilibrium vetoes. Next, note that both (1.6) and (1.7) are easier to satisfy when π is lower. Sine the ex ante evaluation of the resent is dereasing in π, the model suggests that vetoes will be likely when the resident s ubli standing is high. Intuitively, Congress is going to find the blame game strategy most rofitable when it has negative information about the resident that is inonsistent with voter beliefs. Predition 8: Vetoes are more likely to our when the resident s standing is high. The next three redition are based on C and P s willingness to trade off oliy gains for olitial gains. Figure 7 illustrates how eah of the onditions are effeted by the oliy weights P and C. The area under the higher solid line reresent those ombinations of P and C that satisfy (1.6). Alternatively, the area above the lower line are those satisfying (1.7). Blame Game equilibrium may exist in the intersetion of these regions. Insert Figure 7 about here 28

29 First, note that ondition (1.6) an be met only when > suggesting that the resident must ut relatively more weight on the oliy outome than does the ongress. If this were not the ase, C would rather exloit P s willingness to make oliy onessions to get more referred oliies rather than rovoke vetoes. However, ondition (1.7) uts an uer bound on the differene in oliy weights. In this ase, if P is too muh larger than C, C loses the ability to redibly signal through its roosals. Predition 9: Vetoes will our on issues that the resident ares relatively more about than does ongress. The remaining reditions of the blame game model onern the imliations of fators that hange the oliy referene weights of both Congress and the resident. As is lear from Figure 7, the effet of a ommon shift of referene weights deends on the referene oint. If the weights are loated in the lowest region (i.e. below ondition (1.7) ), a ommon inrease in the saliene of oliy will make vetoes more likely. However, if ondition (1.7) is already satisfied, a ommon inrease will redue the likelihood of vetoes. Thus, any reditions will deend on assumtions about whether ondition (1.7) holds generally. In what follows, I assume that it does. 16 If we aet this assumtion, a key redition about the eletoral yle emerges. Sine it seems reasonable to assume that the oliy weights are lower near eletions, the model redits that veto usage will be higher during eletion years. 29

30 Predition 10: Vetoes will be more likely during eletion years. This insight also has a diret bearing on hyotheses about the historial develoment of the veto. As I ointed out in hater xx (eletoral system), the eletoral environment of the 19 th entury hanged in ways that fored residents and legislators to be more ognizant of mass suort for the resident. This also suggests a downward shift on the ommon art of the oliy weight, and leads to a redition about historial atterns. Predition 11: Vetoes should be more ommon after the emergene of the mass eletorate in residential eletions. One final redition (whih we will return to ritially examine in a later hater) emerges from the fat that only extreme tyes ever veto in the blame game model. Sine only tye e vetoes (Lemma 3), every veto is followed by a redution on suort from ( 1 π) v πe m to v e. Predition 12: Vetoes lead to lower ubli suort for the resident. 4.2 Emirial Seifiation I now turn to seifying a reditive model of veto usage based on the reditions of the theoretial models. I also inlude a number of ontrol variables suggested by existing emirial work. I estimate two models of veto usage using different deendent 16 Sine (1.7) does not only when C and P lae very low weights on oliy, I think that the region above 30

31 variables. The first model uses only the number of regular ubli vetoes er ongressional term from , while the seond inludes all ubli vetoes inluding oket vetoes. Both models generate substantively similar onlusions, but the model of all vetoes erforms somewhat better statistially so I fous on those results. The indeendent variables are measured as follows: Legislative outut Any reasonable model would redit that veto ativity is related to the overall amount of legislation resented to the resident. To ontrol for this effet, I use the natural logarithm of the number of ubli laws assed in a given session. 17 Of artiular theoretial interest is the extent to whih veto usage inreases roortionately with the number of ubli laws. Bass (1972) argues that lower levels of legislative ativity imly better legislation, thus roortionally fewer vetoes. Sine this imlies that the number of vetoes should grow at a higher rate than legislative outut, the oeffiient on the natural logarithm of ubli bills should be greater than one. However, both the inomlete information and blame game models an be reasonably interreted to redit that the number of vetoes should grow at a rate lower than the number of ubli laws. In the inomlete information model, more interation should inrease the level of information about the resident's referenes leading to a lower roortion of vetoes. Alternatively, in the blame game model, the olitial signaling value of vetoes may be subjet to diminishing returns whih would redit dereasing roortions of vetoes as legislative ativity rises. it is more emirially relevant 17 The rationale for using the natural logarithm will be made exliit shortly. 31