CENTER FOR AMERICAN POLITICS AND PUBLIC POLICY OCCASIONAL PAPER SERIES (92-1) Deficit Politics and the 1990 Elections Gary C. Jacobson February, 1992 Center for American Politics and Public Policy University of California, Los Angeles 310 GSLIS Building, 405 Hilgard Avenue Los Angeles, California 90024 (310) 206-3109 *** 1300 19th Street, NW, Suite 300 Washington, D.C. 20036 (202) 296-8226 Deficit Politics and the 1990 Elections Gary C. Jacobson Abstract Shortly before election day, 1990, Congress passed a landmark deficit- reduction bill that raised taxes and fees and cut entitlements for broad segments of the public. Despite the bill's manifest unpopularity, all but a handful of incumbents were reelected. Although both outcomes appear to contradict the logic of the "electoral connection," closer inspection shows that congressional action on the deficit was permeated by electoral considerations and had a major impact on voters' decisions. Member of both houses displayed acute sensitivity to the electoral risks they faced in trying to reduce the deficit, which is one reason so many of them escaped defeat. Another, equally important, reason is that very few of them faced opponents capable of exploiting the anti-incumbent sentiment the issue had aroused. A great majority of incumbents were thus in a position to absorb the anticipated political damage without risking their careers, which is why Congress was able to pass the bill. Analysis of Congress' actions and the electorate's response in 1990 provides rich insight into the complexities of the electoral connection under divided partisan control of the federal government. In October 1990, less than two weeks before election day, Congress passed and President Bush signed a landmark deficit reduction package that raised taxes, imposed new user fees, and cut entitlements. The action was as conspicuous as it was unpopular. Both the contentious and confusing process that produced the bill and the bill itself left voters manifestly angry with the Congress and the president. Yet only a handful of incumbents--fifteen representatives, a single senator--were denied reelection, and only a fraction of the losers could plausibly blame their defeat on voters' anger over budget politics (Jacobson 1991). Superficially, both the decision to raise taxes and cut entitlements and the absence of widespread retaliation by irate voters would appear to violate the logic of the "electoral connection" (Mayhew 1974). Most members of Congress put themselves on record only days before an election in support of legislation that promised immediate and palpable pain for large segments of the public in order to achieve the more distant and uncertain benefits of lower deficits. The electorate, despite hostility to the legislation and overwhelming disapproval of Congress's performance, meted out precious little punishment. We might be tempted to conclude that majorities in the House and Senate did not worry about the electoral consequences of supporting the deficit reduction bill (rightly, as it turned out), and voters did not bother to express their anger with it at the polls. On a closer inspection, of course, neither conclusion holds up. In fact, congressional action to rein in the deficit was permeated by electoral considerations, and budget politics had a major impact on voting in the election. Members displayed acute sensitivity to the electoral implications of what their actions, which helps to explain why the electorate's wrath, though clearly expressed at the polls, cost only a handful members their seats. Indeed, the bill passed only because few members faced opponents who posed a serious electoral threat; most were in a position to absorb the anticipated political damage without risking their careers. The events of October and November 1990 offer unusual insight into the complex linkages between electoral and legislative politics. The research reported in this article is designed to take advantage of the unique opportunity these events present for enriching our understanding of the electoral connection. The Electoral Connection In the domain of policy, discussions of the electoral connection commonly focus on how electoral considerations shape the policy choices of individual members of Congress and how, in turn, the policy choices of individual members affect their electoral support. This, as R. Douglas Arnold (1990) points out, by no means exhausts the possible linkages. Arnold distinguishes four decision rules that congressional voters might use in responding to policy issues, depending on the basis of choice (policy positions or policy effects) and the object of choice (parties or candidates). Voters may decide on the basis of (1) the parties' positions on policy issues, (2) the candidates' positions on policy issues, (3) the governing party's (that is, the president's party's) performance in office, or (4) the incumbent's performance in office. As we shall see, the events of 1990 require that we add a fifth potential decision rule to Arnold's list: Voters may also use an institutional performance rule, voting for or (more to the point for 1990) against individual incumbents on the basis of their evaluation of Congress's performance. Members of Congress may find reason to adjust their behavior on account of any or all of these potential decision rules when making policy choices. They are, of course, most sensitive to choices that might affect how voters rate their own performance, not only because their personal standing with constituents is crucial to survival in a predominantly candidate-centered electoral process, but also because it is the only part of the action fully under their control. But because voters' decisions may also be influenced by the issue positions associated with parties and by assessments of the president's performance on policy issues, members of Congress have reason to consider how their actions might affect these things as well. Other things equal, they prefer to do things that make their party and their leaders look good and to avoid doing things that have the opposite effect. Similarly, it may serve their electoral interests to embarrass the other party and its leaders. Naturally, members are happiest when they can adopt positions on issues that enhance their own and their party's standing simultaneously, especially if the other party is thereby put at a disadvantage. The tough choices come when alternative electoral considerations pull in opposite directions. For reasons inherent in the politics of budgets and deficits during the Reagan-Bush years, the various collective components of the electoral connection assumed a far larger role in budget politics in 1990 than is probably typical. This is one reason why the case rewards close scrutiny. The Electoral Origins of the Deficit Issue The first step in understanding how electoral politics shaped the battle to tame the deficit monster 1990 is to recognize that monster is itself a creature of electoral politics. During the 1980 campaign, Republicans discovered the political magic of the supply-side economics. The magic lay in its claim that the government could reduce taxes without reducing spending or adding to deficits, because lower taxes would stimulate so much economic growth that revenues would not actually decline. Republicans could thus promise to reduce taxes without having to concede that programs would also have to be slashed or red ink would flow. Upon taking office, the Reagan administration duly proposed large supply-side tax cuts. Congressional Democrats quickly surrendered to the political appeal of reducing taxes and even sought to outbid the administration by offering additional tax breaks. The administration also attracted enough Democratic votes to impose cuts in domestic social programs to go along with the tax cuts in 1981, but these reductions were more than offset by increases in defense spending. When a sharp recession hit in 1982 and the supply-side dividend fell far short of optimistic projections, large budget deficits began to accumulate. Faced with growing economic distress, Congress resisted further reductions in domestic spending. The recession also emboldened Democrats to attack the equity and efficacy of Republican economic policies, and this became the main theme of their 1982 congressional campaigns. The 1982 election enlarged the Democratic House majority by 26 seats, stiffening congressional opposition to further domestic cutbacks. The experience of the early 1980s taught Republicans that opposition to taxes was good politics, but attacking popular domestic programs was not. Democrats learned that most large domestic programs were indeed popular, but people did not like paying for them. In ensuing campaigns, voters were thus offered the opportunity to declare themselves simultaneously for both low taxes and generous domestic spending. They could--and did--elect Republican presidents on a promise of "no new taxes" and Democratic majorities in Congress committed to maintaining or increasing their favorite domestic programs (Jacobson 1990). Neither side was candid about the price. Republicans insisted that economic growth would provide sufficient revenues to maintain popular government programs and shrink the deficit without additional taxes. And, taking Walter Mondale's fate in 1984 as an object lesson, Democrats continued to champion popular domestic programs and initiatives without mentioning that the money to finance them would have to be borrowed or raised through higher taxes. Budget action faithful to these electoral stances guaranteed continued deficits. Republican administrations accepted large deficits rather than raise taxes to pay for government programs that Democratic congresses refused to cut; and Democratic congresses accepted large deficits rather than reduce government programs when Republican presidents refused to initiate tax increases to pay for them. There is nothing mysterious about the continuing budget deficits; they were the product of a compelling political logic and were the equilibrium outcome of conflicting party preferences under divided partisan control of the federal government (McCubbins 1991). But the deficits also became a major political headache, in part because the mythical analogy to the family budget remains strong in the popular imagination, and in part because most informed observers regarded the growing mountain of government debt as a serious threat to the economy's long-term vitality. Persistent deficits soon became "exhibit A" in every indictment of national politics and its practitioners. The principal attempt to cope with the deficit problem politically during the 1980s was embodied in the Gramm-Rudman-Hollings scheme, enacted in 1985, which required Congress to achieve a balanced budget by meeting a series of progressively lower yearly deficit targets. If Congress and the president did not agree on where to make the required cuts, they were to occur automatically across the board on all programs not explicitly exempted. In effect, Congress tried to force itself and the president to chose between automatic cuts certain to be painful to important constituencies and making hard, but more considered choices on programs and revenues themselves. The political logic of Gramm-Rudman-Hollings is transparent: Instead of having to go on record as voting for cuts in specific programs, members of Congress could support deficit reduction in general while taking credit for fighting to prevent cuts that would otherwise occur automatically. The initial deadline for reaching a balanced budget was 1991. When Gramm-Rudman-Hollings was revised in 1987, the deadline was extended to 1993. Each year it became more difficult to reach the deficit targets, as the easy moves--changes in accounting procedures and other tricks commonly dismissed as "blue smoke and mirrors"--were exhausted. Both the Democratic Congress and the Republican White House remained strongly disposed to resist spending cuts or tax increases that would impose immediate and palpable costs on their core constituencies to achieve the more indirect and distant benefits of a smaller deficit. As long as both sides saw smaller political damage in deficits than in what they would have to give up to cut a deal, no solution was possible, and most of the maneuvering went into avoiding partisan blame for failure to reach agreement. This last observation underlines the central point of the narration thus far: both sides displayed acute concern for their party's image--in Arnold's terms, voters' perceptions of their party's position--on budget issues. The reasons are obvious. Republicans had made opposition to taxes the centerpiece of their strategy for becoming the majority party. Democrats sought to avoid the "tax and spend" brand while burnishing their image as defenders of education, the environment, old people, poor people, farmers, working people, and so on. But some members on both sides also wanted their party to assume the mantle of fiscal responsibility. And Republicans had an additional concern: perceptions of their party's position and performance would be shaped largely by what the Republican president was able to accomplish. To be sure, members still had to look out for their own reputations on budget issues--hence, for instance, the overwhelming support for Gramm-Rudman-Hollings once the bill became a test of fiscal responsibility. But they also recognized an important individual stake in their party's collective image on budget issues. This put all participants in a serious bind when leaders on both sides finally decided that large, permanent deficits were too destructive economically (and, eventually, politically) to be tolerated further. The only feasible outcome was a compromise that included both spending cuts and tax increases. But such a compromise required that both parties betray campaign promises made to their core supporters, abandoning positions that were central to the images they sought to project. It also meant that many individual members of Congress would be pressed to betray their own personal campaign promises. If either side balked, however, it could be branded as irresponsible and unfit to govern. And failure to do something about the deficit after it had been become the dominant domestic political issue could damage everyone, especially if the draconian automatic cuts mandated by Gramm-Rudman-Hollings were allowed to take effect. There is more than a touch of poetic justice in this turn of events. Republican presidents and congressional Democrats prospered politically during the 1980s by pandering to their respective constituencies, promising something for nothing, benefits and programs without the taxes to pay for them. When leaders of both parties finally felt compelled to take on the deficit, they found themselves menaced by the political wrath of an indignant public they had left quite unprepared to face fiscal reality. Budget Action in 1990 In May of 1990, President Bush sought to break the budget stalemate by inviting Congressional leaders to a budget summit with "no preconditions," implying that tax increases were on the table. Democratic negotiators, however, feared a trap, and little progress was made until the president explicitly agreed that "tax revenue increases" would be part the package. After much additional haggling, the negotiators cut a deal to reduce the projected deficit by nearly $500 billion over the next five years through a combination of tax and fee increases and caps on benefits. Both the unedifying process (which included months of unproductive closed-door negotiation, numerous leaks, trial balloons, and false starts, and a brief shutdown of some federal agencies when a deadline was not met) and the product (an agreement that raised taxes and cut back on some popular social programs) angered large segments of the public. The politically painful compromise was not reached easily. The initial package, negotiated in secret between the White House and Democratic congressional leaders, was rejected by majorities on both sides of the aisle. Many members were angry at having been left out of the process, but what really killed the bill was its content. The attempt to find a middle way out of the impasse alienated majorities on both ends of the ideological spectrum. Conservative Republicans refused to join Bush in breaking the talismanic pledge of "no new taxes." Liberal Democrats, who objected to its regressive taxes on alcohol, gasoline, and tobacco and its cuts in Medicare, felt free to vote down the package once Republican leaders had failed to deliver their promised majority. (Part of the deal was that leaders on both sides would deliver majorities for the bill; blame for the pain would be bipartisan.) Opposition to the bill was solidified by its manifest unpopularity. President Bush "went public" with a televised appeal for citizens to lobby their representatives to support the package, but this only stimulated more opposition; the more people learned about the bill, the less they liked it. Members who faced an appreciable electoral risk in 1990 were especially sensitive to the package's liabilities. A moment of introspection told them how their challengers would handle the issue. The advantage to be taken from attacking a package that satisfied no one was too obvious to ignore; every challenger could claim to favor a "better" combination of elements, with "better" defined in whatever terms would sell locally. As Vic Fazio, a House Democrat, put it, "most challengers will simply bash the institution. If they're conservative they'll talk about no new taxes and if they're liberals they'll talk about no cuts in popular programs."(1) In other words, challengers were free to pander as usual, and they did. A survey taken by Congressional Quarterly found every challenger save one opposed the initial deficit-reduction legislation (Hook 1990a). Voting on the First Budget Proposal Only House members voted on the first deficit-reduction proposal; Senators were excused from taking a stand when it failed to pass in the House. Representatives plainly concluded that supporting the bill was politically dangerous and acted accordingly; other things equal, the greater electoral risk they faced, the less likely they were to vote for it. This is shown by the probit equation reported in Table 1. The measure of electoral risk is taken from Congressional Quarterly's classification, published in October, and is entered as a series of dummy variables. CQ's "tossup" and "leaning" categories are combined because of small N's--28 for the combined category. I also separated unopposed incumbents from those also considered "safe" but who did have a major-party challenger. The omitted category is "running for higher office." The measure of ideology, the American Conservative Union's ratings for 1989, is entered in quadratic form because the first budget package was attacked from both ideological extremes; the rating takes values from 0 (most liberal) to 100 (most conservative). [Table 1 here] Clearly, electoral risk had a major effect on the probability that a member would support either proposal. Other things equal, those House members most at risk--those seeking higher office or in the tightest House races--were least likely to vote for the bill, while those facing no electoral risk at all (because they were retiring) were most likely to vote for it.(2) Probabilities across the other categories arrange themselves as we would expect. The differences are remarkably large, with values ranging from .14 to .94 across the risk categories. The vote on the first budget package was also influenced by ideology, but not, independently, by partisanship. Support was heaviest among moderates and thinnest among the ideological extremists in each party. This is apparent from Figure 1, which shows how the expected probability of voting for deficit reduction varies over the ideological spectrum (with other variables set at their means). Insofar as ideological stances embody explicit or implicit promises to voters, ideological voting is itself a manifestation of the electoral connection. [Figure 1] The bill's failure sent the negotiators back to work (not without a brief shutdown of the government after Bush vetoed a continuing resolution meant to keep things going while negotiations continued). There followed a period of confusion, fed by the president's indecision about what tax increases he would accept as part of a revised package--by one count, Bush took six positions in four days.(3) The debate allowed Democrats to resurrect the "fairness" issue by charging that Bush and his party were not opposed to new taxes, but only to new taxes on the very rich. Boxed into this losing position, the Republican administration capitulated. The final outcome was a deficit reduction package far more congenial to Democrats than the initial proposal, and it represented a clear political defeat for the president (Hager 1990). Voting on the Second Budget Proposal The Democrat's victory was reflected in congressional voting on the legislation. The House vote on the second budget package was partisan. But again, level of electoral risk was strongly related to support for the bill by members of both parties. The evidence is in Table 2. Opposition to the second package also increased with conservatism; the slight decline in support among the most liberal members reflect the Black Caucus's protest vote against the bill in retaliation for Bush's veto of the civil rights bill (Hager 1990); see Figure 1. [Table 2 here] Electoral considerations were even more important in shaping senators' votes on the second package. Senators up for reelection in 1990 were less likely to vote for the bill than those whose terms were not up, though the difference is not great (46.7 percent compared to 61.5 percent). Far more striking electoral effects appear within these two categories. Senators facing reelection divided sharply on the budget vote according to the magnitude of the threat represented by their opposition, as Table 3 demonstrates. The table lists these senators in ascending order of their opponent's campaign expenditures (in dollars per voting age individual in the state). Only one of the thirteen senators whose opponents spent less than 5.8 cents per voting-age citizen voted against the bill; only two of the remaining nineteen senators whose opponents spent more than this amount voted for the bill. These two exceptions "prove" the rule. One of them, Rudy Boschwitz of Minnesota, was the only senator to lose in 1990. The other, Daniel Akaka of Hawaii, was free to vote for the bill because his opponent, Representative Pat Saika, had voted for it in the House eleven hours earlier. [Table 3 here] No consideration other than electoral risk appears to have entered their calculus; a quick review of the list confirms that neither ideology nor partisanship divided this set of senators; Jay Rockefeller and Strom Thurmond voted for the bill, Tom Harkin and Jesse Helms voted against it.(4) This is decidedly not the case for senators whose seat were not at stake in 1990. Ideology had a strong effect on their votes--more conservative members were more likely to oppose the bill--and, curiously, although Democrats were more inclined to vote for the bill than were Republicans (76.3 percent to 40.7 percent), once ideology is taken into account, Democrats were actually less likely to support the bill. More significant for our purposes, however, is evidence that electoral considerations influenced even those senators who did not face reelection in 1990. Respondents in the 1990 Senate Election Study (Miller et al. 1991b) were asked to rate senators not up for election on the 100-point "feeling thermometer" scale. The mean thermometer rating gives us a serviceable measure of each senator's current level of popularity in the state; the means ranged from 38 to 78, with an grand mean of 61. This measure was strongly related to the budget vote, as the probit equation in Table 4 demonstrates. Figure 2 illustrates the estimated effect of this variable on the probability of voting for the bill (with the other variables set at their means). [Table 4 and Figure 2 here] Using the thermometer rating as an independent variable raises the question of causation, because the rating was given after the election and hence after the senator voted on the budget. It might be argued that the budget vote affected the thermometer rating rather than the reverse. But the bill itself was highly unpopular, and majorities surveyed in both the 1990 Post-Election Study (Miller, et al., 1991a) and the Senate Election Study strongly opposed raising taxes to reduce the deficit (see Table 5).(5) So it is most unlikely that voting for the bill would have produced higher ratings for Senators. [Table 5 here] A much more plausible interpretation is that stronger support back home made it easier for senators to vote for an unpopular bill, that senators viewed more critically by their constituents were less inclined to risk casting an unpopular vote. Notice also that support for the bill was negatively related to the per capita tax burden borne by citizens of a state, indicating that senators representing taxpayers already facing large tax burdens were reluctant to go on record for increasing them. Here, then, is intriguing evidence that prospective electoral considerations may shape decisions even when election day is two or more years away. Nonetheless, the protection afforded two-thirds of the senators by the six-year term was essential to the bill's passage, for it was opposed by a majority of members seeking reelection in 1990. In sum, electoral expectations strongly shaped the budget votes of both senators and representatives. They clearly--and, as we shall see, correctly--recognized that raising taxes and cutting benefits would be unpopular, so action to reduce the deficit depended on the votes of those members who were sufficiently secure to absorb some political damage without risking their seats. The deficit reduction bill was passed with the votes of members in both houses who faced no serious, immediate electoral threat. The Public Responds Members of Congress were well advised to fear the public's reaction to having their taxes increased and benefits reduced. Polls taken at the time indicate that the events of October were politically damaging to all involved. Both the president and Congress immediately suffered a serious loss of public esteem. Bush's approval ratings, already sliding from the high level stimulated by his response to Iraq's invasion of Kuwait (76 percent approving in a late-September Gallup Poll), dropped sharply. The proportion of Americans approving his job performance fell from 66 percent to 54 percent between early and late October Gallup surveys, and other national polls showed nearly identical declines (Keene and Ladd 1990, p. 96). Reviews of Bush's budget performance were decisively negative; 64 percent of the respondents in the October 28-31 New York Times/CBS News Poll disapproved of his handling of the budget; only 27 percent approved (1990, p. 6). The president suffered further embarrassment when the press got hold of a memo written by Edward Rollins, co-chairman of the National Republican Congressional Committee, advising Republicans to run against the White House's position on the budget. Rollins justified his action by arguing that "my job today is to help House Republicans survive and help other Republican candidates. I have to give the best political advice I can and in this case, it is to run away from the budget package."(6) They could run, but they could not hide. Republican pollsters reported the "bottom dropping out" in late October.(7) Questioned in August on their intended House vote, New York Times/CBS News Poll respondents favored Republican candidates over Democrats 41 percent to 37 percent; by the late October poll, Republicans had fallen behind, 38 percent to 45 percent (1990, p. 8). Though most congressional Republicans opposed the budget package, opposition could not insulate them from the political heat it generated. The president did not get all the blame, to be sure. The public's view of Congress was even more negative. A Wall Street Journal/NBC poll reported that 71 percent of respondents disapproved of Congress' handling of the budget (Hook 1990b), and two separate polls taken in late October found only 23 percent approving Congress's overall performance (Keene and Ladd 1991, p. 84). As usual, individual members were rated much higher than the Congress, but even their level of approval was, by some accounts, abnormally low. The New York Times/CBS News survey found only 51 percent approving of their own representative's performance, with 30 percent disapproving; two months earlier, the respective figures were 60 percent and 21 percent, with 40 percent also approving of Congress' performance (1990, p. 6). The parallel decline in popular approval of the institution and its members is consistent with Richard Born's recent (1990) demonstration that evaluations of Congress strongly affect evaluations of its members (though the latter maintain a much higher average level of approval). Election Day The electorate's wrath at elected officials (especially Republicans) was also clearly visible in the aggregate vote cast on election day. The average vote for House incumbents was 3.9 percentage points lower than it had been in 1988, reaching its lowest level--64.5 percent--since 1974. Almost three quarters of the House incumbents who were not freshmen won a smaller share of the votes in 1990 than they had in 1988. This is the highest proportion of such incumbents suffering a decline in support in any postwar election and is more than three standard deviations beyond the mean for the period. Most striking, though, was that for the first time in any postwar election the mean vote for incumbents of both parties fell. Normally, the aggregate House vote reflects a national partisan tide; the average vote for one party's incumbents increases, the average vote for the other party's incumbents decreases. Before 1990, this pattern was broken only twice in postwar elections--1968 and 1988--when incumbents of both parties increased their average vote. Uniquely in 1990, the swing against incumbents exceeded the partisan swing, so both parties' incumbents typically lost support. An anti-Republican tide was also evident in the aggregate House vote. The average Republican incumbent's vote fell more (-3.7 points) than that of the average Democratic incumbent (-2.0 points),(8) but the clearest sign of the anti-Republican trend showed up in open seats, where Republicans lost 6 of the 17 seats they were defending, while taking none of the 12 seats held by Democrats. Republican officials tried to claim a victory of sorts by pointing out that their party's loss of seats in the House was well below the midterm average for the president's party (26 in elections since 1946); but Republicans went into the election with the smallest share of seats ever held by a president's party at the midterm. In fact, Republicans entered the 102nd Congress in 1991 with precisely the same number of House seats they had held after losing 26 seats in 1982 (167 seats, 38.4 percent of the total, after Phil Gramm had switched parties) and with their smallest Senate contingent (44) in a decade. The anti-incumbent and anti-Republican trends are more difficult to discern in the 1990 Senate elections, yet traces of both appear in Senate voting as well. The net outcome--one incumbent defeated, a single seat changing party control--represents the smallest change since senators have been elected directly by voters. This result is all the more remarkable because Senate incumbents have been, on average, three times as vulnerable as House incumbents over the past half-century and because both parties are now competitive in virtually every state.(9) Moreover, in contrast to other recent Senate elections, 1990 produced relatively few close contests; only 4 seats were won with less than 53 percent of the two-party vote, and 22 of the 35 winners received more than 60 percent. Some signs of the national trends identified in the House elections nonetheless surfaced. Minnesota Senator Rudy Boschwitz's defeat at the hands of Paul Wellstone, a political novice who spent only $1.3 million against Boschwitz's $6.2 million, was a surprise. So was New Jersey Senator Bill Bradley's narrow victory despite an ever greater spending advantage ($9.5 million to $801,660). In both cases, many voters appeared to be venting their disgust with other politicians and with politics in general on the most available target, a complacent incumbent (Benenson 1990). The trend against Republicans is also evident--in what the party did not achieve. Republicans had hoped to make inroads into the Democratic majority, positioning themselves to retake the Senate in 1992, when the Democrats must defend 20 seats. Five Republican House members gave up safe seats to run against Senate Democrats thought to be vulnerable; all of them lost, and three of their House seats were taken over by Democrats. As members of the House, they were, of course, unable to derive much benefit from whatever anti-incumbent sentiment prevailed in their states. The average Republican incumbent's vote fell by 6.2 percentage points compared to 1984, while the average Democrat's vote increased by 3.7 percentage points; but keep in mind that 1984 had been a rather good Republican year. Considering the aggregate results, it appears that voters angered by the budget battles and their resolution did desert incumbents in general and Republicans in particular, especially in House elections. On the other hand, the average vote loss was not, by itself, large enough to defeat more than a few incumbents. The incidence of victory or defeat depended on how the vote loss was distributed, leading to the next question: Did voters discriminate? Did a vote for the deficit reduction package cost additional votes on election day? Were members in danger able to protect themselves by voting against the unpopular measure? These questions are not easily answered, because voting on deficit reduction legislation was so strongly conditioned by electoral expectations. Nonetheless, the available evidence suggests that some voters did discriminate, although only among Democratic House incumbents and only under certain conditions--just those conditions under which these members were least likely to take the unpopular stance. Republicans suffered the same negative consequences regardless of how they voted. Voting in House Elections Did voting for the deficit reduction bill reduce a representative's vote on election day? The regression equations in Table 6 suggest that it did, but only among Democrats and only to the degree that the challenger had the resources to exploit the issue. According to the first equation, voting for the second bill cost a representative about 1 percent of the vote, taking level of electoral threat (represented by the lagged vote and challenger's spending) into account, but the estimated coefficient is only slightly larger than its standard error and hence not statistically significant. The effect is larger (though still short of statistical significance) when the deficit reduction vote is interacted with the challenger's spending. The coefficients indicate that if the challenger spent less than about $10,000, the incumbent did better if he or she voted for the bill. At higher levels of spending by the challenger, voting for the bill reduced the incumbent's vote; at $100,000, the expected reduction is 1.7 percentage points; at $250,000, 2.4 percentage points; and at $500,000, 2.9 percentage points. [Table 6 here] Further investigation shows that this interactive effect is largely confined to elections involving Democratic incumbents; the coefficients for Republican incumbents are substantively small and statistically insignificant. Because equations with interaction terms are not so easy to interpret, the lower section of the table displays the estimated electoral effects of the budget vote at several levels of spending by challengers for incumbents of each party; the estimates are computed with the incumbent's previous vote set at 60 percent. These results suggest that voting for the bill cost Democratic incumbents support, but only to the extent that the Republican challenger had the money to exploit the issue. Another way of putting it is that the impact of the Republican challenger's expenditures more than doubled if the Democrat created an exploitable issue by voting for the bill. Republicans incumbents suffered generally (note the coefficients on "Democratic incumbent" in the first two equations in Table 6) no matter how they voted, with the extent of their problems varying strongly with the Democratic challengers' campaign expenditures either way. Despite running away from Bush on the issue--three--quarters of them voted against the second package--House Republicans evidently could not escape guilt by association when the president's betrayed his "no new taxes" pledge. Survey data reinforce both of these conclusions. The 1990 Post- Election Study asked respondents if they approved or disapproved of the deficit reduction package (see the Table 5 for the wording of the question). We can thus determine whether the respondent agreed or disagreed with the incumbent's vote. Analysis of the data shows that this variable was significantly related to the individual vote only in cases where the incumbent voted for the bill and the respondent could spontaneously recall the name of the challenger, and only a tiny proportion of respondents (3.7 percent of the self-reported voters) fall into this category. The evidence is in the upper section of Table 7. [Table 7 here] The number of cases is far to small to separate Democratic and Republican incumbents, but the larger sample found in the Senate Election Study permits such a distinction. This survey requires the use of a more indirect measure of agreement: Whether the respondent approved or disapproved of raising taxes to reduce the deficit and whether the House incumbent voted for the deficit reduction bill. Nonetheless, the results in the lower half of Table 8 replicate those of the upper half quite satisfactorily. Furthermore, when Democrats and Republicans are treated separately (in cases where the incumbent voted for the bill and the respondent recalled the challenger's name), agreement or disagreement has a very strong effect on voting for the Democratic incumbent (91.4 percent if in agreement, 26.2 percent if not), but not on voting for the Republican (31.5 percent if in agreement, 59.2 percent if not, a difference in the "wrong" direction). Thus we find that only Democratic incumbents who voted for the deficit reduction bill were rewarded or punished for their vote, and only by voters who were able to recall the challenger's name (and thus could be expected to have comparatively high levels of information on other aspects of the House campaign). But, as we would surmise from our earlier analysis of electoral risk, where the campaign was vigorous enough to etch the Republican challenger's name in voters' memories, the Democrat was much less likely to have voted for the bill. This is confirmed by the data in Table 8. In the 23 Democratic districts where the Republican challenger spent more than $300,000, more than 40 percent of the respondents recalled the challenger's name; in the other 224, fewer than 10 percent recalled the challenger's name. Support for the either bill was much lower among Democrats with well-financed opponents. [Table 8 here] When faced by an opponent capable of exploiting the budget issue in the election, then, Democratic incumbents were much more reluctant to leave themselves vulnerable. For this reason, although the deficit reduction vote cost individual House Democrats some votes, it rarely cost them their seats. If we rerun the equation in Table 7 using probit, with winning or losing as the dependent variable, there is no relationship whatever between a Democrat's (or, for that matter, a Republican's) vote on the bill and the probability of winning reelection, either as a main effect or interactively with the challenger's level of spending. The individual vote for House Republicans was not related to their vote on the budget. It was, however, related to the respondent's approval of George Bush's performance as president. The Post-Election Survey asked whether voters approved or disapproved of Bush's handling of the economy, foreign policy, the environment, and the situation in the Persian Gulf as well as his overall performance in office. The president was rated lowest on his handling of the economy (40 percent approving, 53 percent disapproving), and this rating was most strongly related to the vote in Republican districts (its coefficient was the only one of the ratings coefficients exceeding conventional levels of statistical significance). A probit model of the vote incorporating this variable appears in Table 9. [Table 9 here] The second equation in Table 9 reveals that ratings of Bush's economic performance were strongly conditioned by budget politics as well as by respondents' economic experiences and perceptions. Opinions on the deficit reduction deal had a significant effect on assessments of Bush's handling of the economy. Moreover, those respondents who considered the deficit or taxes to be the most important issue in the election (14 percent of the sample) were much less likely to approve of the president's handling of the economy. Thus the damage the president suffered in his performance ratings, especially on the economy, cost Republican incumbents votes, and it cost them votes whether they ran from the president's package (as Rollins had recommended) or not. Evaluations of Bush's handling of the economy had no significant effect on the vote for House Democrats. Senate Elections The budget vote in the Senate was so completely determined by electoral conditions (recall Table 3) that it is futile try to measure its effect on the aggregate vote with available data. Survey data provide some evidence that voting for the bill cost a senator votes on election day, but the results are highly sensitive to the circumstance that Rudy Boschwitz didn't realize what kind of trouble he was in and voted for the bill. The first probit equation in Table 10 indicates that, taking the challenger's level of spending into account, voters had a significantly lower probability of voting for a Senate incumbent who voted for the deficit reduction package. But when Minnesota respondents are removed from the data set, the coefficient on the budget vote shrinks to insignificance. The most plausible interpretation of these results is that, with the exception of Boschwitz, any senator with sufficiently active opposition to suffer any appreciable damage from voting for the bill did not vote for it. [Table 10 here] The second equation adds some variables testing whether the issue had an impact on the Senate vote independent of the behavior of individual senators. Respondents' views on "which party worked hardest to produce a fair solution to reducing the deficit" were strongly related to the vote (even with party identification controlled); and respondents who mentioned the deficit or taxes as the most important national issue were less inclined to vote for the incumbent. In an equivalent equation for House voters, the party performance variable was significantly related to the House vote, but identifying the budget as the most important national issue was not. Some voters in both kinds of contests were evidently guided by Arnold's "party performance" criterion on the deficit issue. Both the aggregate and survey evidence indicate that Republican Senate incumbents were at a general disadvantage in 1990 (observe the coefficients on "Democratic incumbent" in Table 10). Again, it appears that Republicans were hurt by the events of October regardless of their personal responsibility for them. Constituents' views of senators not up for election may also have been influenced by the vote on the second deficit reduction package. Because, as we have seen, the more popular senators tended to vote for the bill, the mean thermometer rating for senators not facing reelection was significantly higher (by about 5 degrees) among those who voted for the bill. But in rating senators who voted for the budget package, those respondents who strongly favored raising taxes to reduce the deficit rated the senator about 5 degrees higher than those who strongly opposed doing so. In rating senators who voted against the package, the order was reversed; respondents who strongly favored raising taxes to reduce the deficit rated the senator about 5 degrees lower than those who strongly opposed doing so. In both cases, the differences are statistically significant.(12) The Outcome Despite the abundant evidence of anti-incumbent sentiment, few incumbents were actually defeated. Part of the reason is that, as we have seen, most of the vulnerable incumbents--those with formidable opponents-- voted against the bill, so it could not be used against them personally. Equally important, however, was the fact that so few members faced opponents who were capable of exploiting the opportunity deficit-reduction politics had, by raising public ire at incumbents as a class, handed them. In general, incumbents lose only with the confluence of (1) an issue giving voters a good reason to desert the incumbent, (2) an acceptable alternative, and (3) enough money to acquaint voters with both. Each component is necessary; none is sufficient by itself. Without exploitable issues, even experienced, well-financed challengers can expect to make little headway. And, more to the point for 1990, even incumbents vulnerable on the issues are returned to office if they avoid a challenger with the skill, appeal, and resources to exploit their vulnerability (Jacobson 1992). Deficit reduction politics had produced issues that offered considerable leverage against incumbents of both parties, but remarkably few incumbents faced challengers capable of exploiting them. Eighty-four House incumbents had no major-party opponent at all; five senators were also unopposed in the general election. The number of uncontested House seats has not been so high since the 1950s, before partisan competition had spread to the South. Those House incumbents who were opposed faced the weakest collection of challengers in any postwar election. A simple index of a challenger's quality is prior political experience; those challengers who have previously held elective office are classified as high quality challengers, those who have not, are not. Though obviously crude, this measure of quality is strongly related to various indices of electoral success (Jacobson 1990). By this standard, both parties fielded an extraordinarily small number of high-quality challengers in 1990. Only 10 percent of incumbents of either party faced opponents who had ever won any elective office, a proportion falling more than two standard deviations below the postwar mean (compare, for example, 1982, when 40 percent of the Democratic challengers had held elective office). Not surprisingly, little money flowed into the coffers of such an unpromising lot: they raised and spent less in real terms than any class of challengers since 1974 (Jacobson, forthcoming). More than a few post-election analysts argued that the 1990 election underlined the decisive power of incumbency because so few incumbents were defeated even though the public's disgust with government was at a peak. Measured against the weakness of their opposition, however, the overall performance of House incumbents was unimpressive. Parameters from a model regressing the challenger's vote on the national swing, the previous vote in the district, campaign spending by the incumbent and the challenger, the quality of the challenger, and a trend term, estimated for elections from 1972 through 1988 and applied to 1990, predict the average challenger to win 30.6 percent of the vote, much lower than the 35.5 percent the average challenger actually won.(13) In other words, incumbents did much worse on election day than the quality and vigor of their opposition warranted. A more formidable set of challengers might well have sent a lot more incumbents packing. Why Were Challenges So Feeble? What accounts for the extraordinary weakness of both parties' challengers in the 1990 House elections? A good part of the explanation lies in the timing of emerging trends. The events that left voters angry with the Congress and the president took place late in the election year. National conditions early in the election year, when final decisions about candidacy must be taken, were far less promising to challengers of either party (Jacobson 1991). The prospect of redistricting after 1990 also probably deterred experienced potential challengers of both parties. Why endure the hard work, expense, and risk to one's political future by mounting a full-scale campaign to capture a district which may change drastically, or even disappear, before 1992? It made more sense to wait a couple of years until the new boundaries are in place and the new districts, along with anticipated retirements, produce a bumper crop of open seats. This kind of thinking would be reinforced by the overblown depictions of the incumbency advantage that now permeate media coverage of congressional elections (and help to make the notion that incumbents are untouchable a self-fulfilling prophecy). Thus when the public mood turned hostile to officeholders as a class and incumbency itself became a potential liability, relatively few challengers of either party were poised to exploit their windfall, and so while most House incumbents saw their vote drop, often sharply, only a small number were turned out of office. Most of the House incumbents who did suffer defeat were identified as vulnerable well before the events of the fall and so attracted formidable opposition quite apart from anyone's reading of national trends. Four could be tied directly to the savings- and-loan debacle; three were beset by ethics scandals; several others held marginal seats or had shown other individual signs of vulnerability. Only a couple--Doug Walgren of Pennsylvania and Robert Kastenmeier of Wisconsin--can be considered surprise victims of anti-incumbent sentiment, though a score of others, most notably Newt Gingrich, the Minority Whip, had close calls (Jacobson, 1991). Conclusions Electoral politics had a powerful effect on budget politics in 1990, and budget politics had a powerful effect on voting in the 1990 election, though the ultimate impact on election outcomes was muted by the defensive behavior of those members most at risk and by the absence of effective opposition in so many districts and states. Members of Congress were acutely conscious of the political risks posed by enacting a highly salient and unpopular piece of legislation just before the election. But they were able to do so--and, by and large, get away with it--by engaging in a kind of division of labor. A large majority of members who faced an appreciable electoral threat opposed the legislation and thus could duck personal responsibility for it. Most of those who could hazard the politically dangerous work involved in "making good public policy" did so. Indeed, because the action imposed short-run costs designed to produce long-term benefits, anyone not at risk in 1990 was better off making the hard, unpopular choices now rather than later. Members who faced feeble opposition or did not have to run 1990 had ample incentive to get the issue behind them before their next campaign, when they might not be so fortunate. Moreover, even those members at electoral risk who opposed the bill may still have benefitted politically from its enactment, because all of the alternative outcomes promised to make voters even angrier. The automatic sequestration required by Gramm-Rudman would have devastated important government agencies and programs (Wildavsky 1992). Failure to do something about the deficit after staking so much on the issue would have opened all sides to charges of cowardice, irresponsibility, and incompetence. The deficit issue had evolved to a stage where every outcome promised to be politically painful. Incumbents who voted against the bill could at least deny personal responsibility for the outcome that did occur. The deficit reduction package, unpopular as it was, may still have minimized the damage for all concerned. The bill passed because enough senators and representatives felt safe enough to risk voting for it. In the Senate, this was an effect of constitutional design: the winning votes came from senators who were not up for election in 1990. In the House, it was an effect of unusually weak challengers, which reflected, in part, the timing of budget action. It is axiomatic that raising taxes during an election year is bad electoral politics. But a corollary suggested by 1990 is that if you must raise taxes during an election year, be sure to do it so late that your opposition has no time to mobilize and exploit the issue effectively. The president was in the same strategic position as members of Congress who faced no serious electoral threat 1990; it made political sense to take the heat by cutting a deal now in hopes of making life easier for himself in 1992. But left congressional Republicans to absorb more than their share of the immediate political damage. House Republicans had little opportunity to influence the outcome, and a large majority of them opposed the bill; but voters nonetheless took their anger at the legislation and the Republican president out on them. Indeed, Republicans suffered doubly, when Bush betrayed his "no new taxes" pledge, diminishing his popularity and staining his party's image, and when Democrats were able to resurrect the "fairness" issue. The electorate expressed its anger at deficit reduction politics in several ways. There was some electoral penalty for members who voted for the bill--but it was limited largely to House Democrats who faced well- financed opponents, and most of them had taken care to vote against the bill. With a few exceptions--four of the six House Democrats who lost had voted for the second package, as had Senator Boschwitz--incumbents ducked individual responsibility for inflicting pain in any circumstance where it might have cost them reelection. They were less successful in escaping collective responsibility for Congress' actions. The general decline in support for incumbents of both parties is indicative. Divided government diffuses responsibility for unpopular policies, leaving voters bereft of partisan cues for assigning blame. The simplest alternative is to turn on incumbents as a class. The popular enthusiasm for term limits (Jacobson 1991) is another expression of this response. Republicans also suffered collectively through guilt by association with the Republican president even though most of them had refused to join his retreat on "no new taxes." Voters plainly did enforce collective responsibility in 1990, though the impact of their reaction was muted by the dearth of challengers capable of building on voters' anti- incumbent sentiments. Members of Congress should not count on being so fortunate in the future. Notes 1. Robin Toner, "In Budget Chaos, Democrats See Chance of "Normal" Off-Year Victory," New York Times, 14 October 1990, p. El. 2. Other plausible measures of electoral risk--the incumbent's vote margin in 1988, the amount of money raised and spent by the challenger-- support the same substantive conclusions. 3. Christopher Matthews, "Bush, GOP in Vulnerable Position for Midterm Vote," San Francisco Examiner, 14 October 1990, p. 1. 4. On the other hand, ideology appears to have determined the votes of the three retiring senators, all of whom were conservative Republicans, and all of whom voted against the bill. 5. Absolute majorities opposed raising taxes to reduce the deficit in 48 of the 50 states surveyed in the Senate Election Study; a plurality opposed raising taxes in one of the remaining two states; only in one state (Vermont) did a plurality (50 percent) of the respondents favor such a policy. 6. Andrew Rosenthal, "Bush Mounts Effort to Quell G.O.P. Rebellion Over Taxes," New York Times, 26 October 1990, p. A11. 7. R. W. Apple, Jr., "Government Itself Is the Big Casualty," New York Times, 25 October 1990, p. A15. 8. Based on incumbents who ran and faced major party opposition in both elections. Comparing averages for all opposed incumbents in both elections, Democrats were down -3.1 percentage points, Republicans, down -5.6 percentage points. 9. Thirty-two states have chosen senators from both parties in elections since 1978, and all but one of the remaining states--Hawaii--have elected governors of the party opposite their senators' during this time. 10. I have not included the incumbent's spending as a variable because, in this model, it's coefficient is insignificant and perversely signed. 11. We cannot, however, determine whether the respondent actually knew how the incumbent voted, because that question was not asked. 12. I derived these estimates by regressing the thermometer rating of senator in each category on party identification, how well the respondent thought the senator kept in touch with constituents, and the respondent's views on raising taxes to reduce the deficit. 13. The model for 1972-1988 is Challenger's vote- b s.e. Constant 9.57 .75 Challenger's party's vote, previous election .33 .01 Mean swing to challenger's party .68 .03 Challenger's spending 3.36 .09 Incumbent's spending .16 .18 High quality challenger 1.34 .27 Trend -.30 .05 N = 2667 R2 = .69 Spending is entered as the log of spending in thousands of dollars, adjusted for inflation (1990-1.00), with all candidates assumed to have spent at least $5,000 (the log of zero is undefined; candidates spending less than $5,000 do not have to report their expenditures). A high quality challenger is one who has previously held elective public office. The trend term is 1972=1, 1974=2, ....1988=9. Appendix: Description of Variables The survey data analyzed in this article are from the American National Election Study, 1990: Post-Election Survey (Miller, et al., 1991a) and the American National Election Study, 1990: Senate Election Study (Miller, et al, 1991b). The variables emplyed in the probit equations in tables 9 and 10 were constructed from the following questions in the post-election survey (the number is parentheses): House vote: 1 if for the incumbent, 0 if for the challenger (612) Party ID: 1 if strong, weak, or leaning Democrat, 0 if pure independent, -1 if strong, weak, or leaning Republican (643) Bush's rating on the economy: 1 if respondent approves of George Bush's handling of the economy, -1 if respondent disapproves, 0 otherwise (203) Approve deficit reduction deal: 1 if respondent approves of "recent agreement that increases taxes and reduces government services," -1 if disapproves, 0 otherwise (735) Deficit/taxes most important national issue: 1 if respondent mentions the deficit or taxes as the most important issue (numbers 415 and 416 in the Campaign Issues Code) in any of the three responses recorded to an open ended question, 0 if not (702, 703, 704) Family finances better/worse: 1 if respondent reported family better off financially than a year ago, -1 if worse off, 0 otherwise (901) Senate vote: 1 if for the incumbent, 0 if for the challenger (617) Party same as incumbent: 1 if respondent's party identification matched incumbent's party, -1 if it matched the challenger's party, 0 if the respondent was an independent leaning toward neither party (643) Democratic incumbent: 1 if the Senate incumbent was a Democrat, 0 if Republican (102) Incumbent's party better on deficit: 1 if respondent thought Senate incumbent's party "worked hardest to produce a fair solution to reducing the deficit," -1 if the challenger's party, 0 otherwise (736) References Arnold, R. Douglas. 1990. The Logic of Congressional Action. New Haven: Yale University Press. Benenson, Bob. 1990. "Republicans' Net Loss: One Seat and Many Expectations." Congressional Quarterly Weekly Report 48 (10 November):3824-3929. Born, Richard. 1990. "The Shared Fortunes of Congress and Congressmen: Members May Run, But They Can't Hide." Journal of Politics 52:1223-1241. Federal Election Commission. 1991. "1990 Congressional Election Spending Drops to Low Point," news release, February 22. Gibbs, Nancy. 1990. "Keep the Burns In." Time 136 (19 November):32-42. Hager, George. 1990. "Deficit Deal Ever So Fragile As Hours Dwindle Away," Congressional Quarterly Weekly Report 48 (27 October):3574-3582. Hook, Janet. 1990a. "Budget Ordeal Poses Question: Why Can't Congress Be Led? Congressional Quarterly Weekly Report 48 (20 October):3471-3473. _______. 1990b. "Public's Patience Wears Thin As Congress Winds Down." Congressional Quarterly Weekly Report 48 (27 October):3569. Jacobson, Gary C. 1990. The Electoral Origins of Divided Government. Boulder, Colorado: Westview Press. _______. 1991. "Divided Government, Strategic Politics, and the 1990 Congressional Elections," paper presented at the Annual Meeting of the Midwest Political Science Association, Chicago, Illinois, April 18-20. _______. 1992. The Politics of Congressional Elections. 3rd ed. New York: HarperCollins. _______. Forthcoming. "When Opportunity Knocks, No One's Home: The Misallocation of Resources in House Campaigns." In Congress Reconsidered, 5th ed., ed Lawrence C. Dodd and Bruce I. Oppenheimer. Washington, D.C.: Congressional Quarterly Press. Jacobson, Gary C., and Kernell, Samuel. 1983. Strategy and Choice in Congressional Elections. 2nd ed. New Haven: Yale University Press. Keene, Karlyn, and Everett Carl Ladd, eds. 1991. "Public Opinion and Demographic Report," The American Enterprise 2 (January/February):83-101. Mayhew, David R. 1974. Congress: The Electoral Connection. New Haven: Yale University Press. McCubbins, Mathew. 1991. "Goverment on Law-Away: Federal Spending and Deficits under Divided Government." In The Politics of Divided Government, ed. Samuel Kernell and Gary W. Cox. Boulder, CO: Westview Press. Niller, Warren E., Donald R. Kinder, Stever J. Rosenstone, and the National Election Studies. 1991a. American National Election Study. 1990: Post-Election Survey ICPS Early Release Versionl [computer file]. Ann Arbor, MI: University of Michigan, Center for Political Studies [producer], 1991. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor]. Miller, Warren E., Donald R. Kinder, Stever J. Rosenstone, and the National Election Studies. 1991b. American National Election Study. 1990: Senate Election Study [computer file]. Ann Arbor, MI: University of Michigan, Center for Political Studies [producer], 1991. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor]. New York Times/CBS News Poll. 1990. Second October Survey, October 28-31. Wildavsky, Aaron. 1992. The New Politics of the Budgetary Process, 2nd ed. New York: HarperCollins. Table 1. Probit Estimates of House Voting on the Fiscal 1991 Budget Resolution: Conference Report (H Con Res 310, 5 October 1990) Variable Coefficient t-statistic Constant -1.734*** -3.08 Democrat -.067 -.30 ACU Rating .035*** 4.14 ACU Rating, squared -.0004*** -4.52 Electoral risk: "Tossup" or "Leaning" -.071 -.11 "Favored" .775 1.36 "Safe" 1.203** 2.31 Unopposed 1.477** 2.77 Retiring 2.562*** 3.82 Log Likelihood (N=433) -260.43 Percent correctly predicted (null=58.7) 64.7 Estimated probablities for risk categories with other variables set at means: Electoral risk: Seeking higher office .16 "Tossup" or "Leaning" .14 "Favored" .41 "Safe" .58 Unopposed .69 Retiring .94 Note: The dependent variable is 1 if the member voted for the bill, 0 if the member voted against the bill; American Conservative Union (ACU) ratings are for 1989; the categories of electoral risk are from Congressional Quarterly Weekly Report, 13 October 1990, pp. 3359-3371; I separated unopposed incumbents from those designated "safe" but who had major-party opponents; the omitted category is "seeking higher office"; "Democrat" is scored 1 if the member is a Democrat, 0 if Republican. **p<.01, one-tailed test. ***p<.001, one-tailed test. Table 2. Probit Estimates of House Voting on the Fiscal 1991 Budget Reconciation Act/Conference Report (HR 5835, 27 October 1990) Variable Coefficient t-statistic Constant -.610 -1.22 Democrat .467* 2.07 ACU Rating .011 1.18 ACU Rating, squared -.0003** -2.87 Electoral risk: "Tossup" or "Leaning" -.028 -.05 "Favored" .238 .46 "Safe" .779* 1.70 Unopposed .881* 1.85 Retiring 1.867*** 3.02 Log Likelihood (N=433) -229.97 Percent correctly predicted (null=53.2) 74.7 Estimated probablities for risk categories with other variables set at means: Democrats Republicans Electoral risk: Seeking higher office .44 .27 "Tossup" or "Leaning" .42 .25 "Favored" .52 .34 "Safe" .72 .55 Unopposed .76 .59 Retiring .95 .89 Note: The dependent variable is 1 if the member voted for the bill, 0 if the member voted against the bill; American Conservative Union (ACU) ratings are from 1989; the categories of electoral risk are from Congressional Quarterly Weekly Report, October 13, 1990, pp. 3359-3371; I separate unopposed incumbents from those designated "safe" but who face major-party opponents; the omitted category is "seeking higher office;" "Democrat" is scored 1 if the member is a Democrat, 0 if Republican. *p<.05, one-tailed test. **p<.01, one-tailed test. ***p<.001, one-tailed test. Table 3. Challengers' Spending and Senate Voting on the Fiscal 1991 Budget Reconciliation Act/Conference Report (HR 5835, 27 October 1990) Challenger's Member Party Spending Vote on HR 5835 D. Pryor Dem. .000* for T. Cochran Rep. .000* for J. Warner Rep. .000* for S. Nunn Dem. .000* for B. Johnston Dem. .000* against T. Stevens Rep. .001 for A. Gore Dem. .002 for S. Thurmond Rep. .003 for N. Kassebaum Rep. .009 for J. Rockefeller Dem. .015 for A. Simpson Rep. .017 for P. Domenici Rep. .037 for D. Boren Dem. .058 for B. Bradley Dem. .138 against P. Gramm Rep. .141 against D. Coats Rep. .267 against S. Levin Dem. .365 against R. Boschwitz Rep. .433 for J. Biden Dem. .513 against P. Simon Dem. .580 against H. Hefflin Dem. .629 against M. Hatfield Rep. .734 (absent) M. McConnell Rep. 1.073 against J. Kerry Dem. 1.144 against J. Exon Dem. 1.234 against M. Baucus Dem. 1.257 against J. Helms Rep. 1.645 against W. Cohen Rep. 1.866 against T. Harkin Dem. 2.392 against L. Pressler Rep. 2.600 against C. Pell Dem. 2.743 against D. Akaka Dem. 3.068 for Note: The challenger's spending is in dollars per voting age individual in the state. Sources: Vote on HR5835: Congressional Quarterly Weekly Report 48 (November 3, 1990), p. 3769; Challenger's spending: Federal Election Commission (1991). *Unopposed in the general election. Table 4. Probit Estimates of Voting by Senators Not Up For Reelection on the Fiscal 1991 Budget Reconciliation Act/Conference Report (HR 5835, 27 October 1990) Variable Coefficient t-statistic Constant 6.111 1.57 Democrat -4.575** -2.80 ACU Rating -.115*** -3.50 Mean Thermometer Score .126** 2.65 Tax Burden Per Capita -.006** -2.47 Log Likelihood (N=65) -16.8 Percent correctly predicted (null=61.5) 87.7 Note: The dependent variable is 1 if the senator voted for the bill, 0 if the senator voted against the bill; "Mean Thermometer Score" and "Tax Burden Per Capita" are from the American National Election Study, 1990: The Senate Election Study (Miller, et al, 1991b), variables 72 and 5487; ACU are the American Conservative Union's ratings for 1989; "Democrat" is scored 1 if the senator is a Democrat, 0 if Republican. **p<.01, one-tailed test. ***p<.001, one-tailed test. Table 5. Public Opinion on Taxes and Deficits (Percentages) Now thinking about the problem of reducing the federal deficit, do you approve or disapprove of the recent agreement to increase taxes and reduce government services? Post-Election Survey Approve 22.0 Approve some parts, disapprove others 6.0 Disapprove 65.8 DK/NA 6.3 Do you favor or oppose increasing taxes in order to reduce the deficit? Senate Election Study Post-Election Survey (weighted) Favor strongly 9.1 16.4 Favor, not strongly 16.0 22.5 Depends on tax 5.8 2.7 Oppose, not strongly 15.4 13.4 Oppose, strongly 50.1 39.0 DK/NA 3.6 6.1 Sources: Miller, et al., 1991a (variables 735 and 1019) and 1991b (variable 499). Table 6. The Effects of the Deficit Reduction Vote on the Incumbent's Vote Share in the 1990 House Elections All Candidates Democrats Republicans Variable (N 277) (N 167) (N=ll0) Constant 59.82 56.15*** 44.16*** 75.41*** (1.11) (9.60) (5.15) (9.01) Incumbent's Vote, 1988 .40*** .40*** .50*** .23** (8.01) (8.04) (7.87) (2.84) Democratic incumbent 2.87*** 2.79*** (3.31) (3.22) Voted for HR 5835 -.99 6.86 11.81* 5.25 (-1.16) (1.36) (1.69) (.57) Challenger's spending -2.26*** -1.91*** -1.11* -2.67*** (-8.27) (-5.49) (-2.01) (-6.18) Challenger's spending X -.74 -1.24* -.53 deficit reduction vote (-1.58) (-1.92) (-.62) R2 .40 .51 .52 .47 Estimated vote for incumbents, assuming a 60 percent margin in 1988: Democrats Republicans For Against Difference For Against Difference Challenger's Spending $10,000 64.3 63.9 .4 64.9 64.6 .4 $100,000 58.9 61.3 -2.6 57.6 58.5 -.9 $250,000 56.8 60.4 -3.6 54.7 56.0 -1.3 $500,000 55.1 59.6 -4.5 52.5 54.2 -1.7 Difference between $10,000 and $500,00 -9.2 -3.3 -12.4 -10.1 Note: The dependent variable is the incumbent's percentage share of the two-party vote; "Democratic incumbent" takes the value of 1 for Democratic incumbents, 0 for Republican incumbents; challenger's spending is the log of reported expenditures (Federal Election Commission 1991); t-statistics are in parentheses. *p<.05, one-tailed test. **p<.01, one-tailed test. ***p<.001, one-tailed test. Table 7. The Deficit Reduction Vote and Individual Voting Decisions in the 1990 House Elections (Percent Voting for the Incumbent) Agreement with Incumbent's Vote: Agreed Disagreed Sig. N Post-Election Study Incumbent voted for the bill: Recalls challenger's name 85.7 30.0 p<.02 17 Does not recall challenger 87.3 86.9 ns 270 Incumbent voted against the bill: Recalls challenger's name 43.5 44.4 ns 32 Does not recall challenger 78.9 82.2 ns 140 Senate Election Study Incumbent voted for the bill Recalls challenger's name 82.3 35.8 p<.001 85 Democratic incumbent (91.4) (26.2) p<.001 53 Republican incumbent (31.5) (59.2) ns 32 Does not recall challenger 88.0 80.4 ns 557 Incumbent voted against the bill Recalls challenger's name 36.3 54.9 ns 167 Does not recall challenger 85.2 82.6 ns 583 Note: Entries from the Senate Election Study are weighted, but the N's and significance tests are from unweighted data. Sources: Miller, et al, 1991a (variables 213, 217, and 221) and 1991b (variables 39 and 43). Table 8. Challenger's Spending, Name Recall, and Deficit Reduction Votes by Democratic Incumbents (Percentages) Challenger Spent: Less than $300,000 $300,000 or more Sig. Respondent recalled challenger's name Post-Election Study 4.5 (331) 40.0 (30) p<.001 Senate Election Study 8.4 (762) 48.0 (74) p<.001 Democratic incumbents voting for H Con Res 310 43.8 (244) 17.4 (23) p<.001 Democratic incumbents voting for HR 5835 72.9 (224) 43.5 (23) p<.001 Note: Percentages for the Senate Election Study are weighted; their N's (in parenthewses) are not. Sources: Miller, et al., 1991a (variables 213, 217, and 221) and 1991b (variables 39 and 43); Federal Election Commission (1991). Table 9. Probit Estimates of Voting for Republican House Incumbents (Post-Election Survey) Coefficient t-statistic Voting for Republican incumbent: Constant 3.741*** 4.16 Party ID -.772*** 5.78 Bush's rating on the economy .277* 2.08 Log Challenger's Spending -.294*** -3.51 Log Likelihood (N-167) -66.92 Percent correctly predicted (null=73.7) 81.4 Bush's rating on the economy: Constant .460*** 4.51 Party ID -.371*** -6.81 Approve deficit reduction deal .122* 2.03 Deficit/taxes most important issue -.388** -2.52 National Economy Better/Worse .548*** 5.13 Family Finances Better/Worse .242*** 3.75 Threshold .187*** 6.47 Log Likelihood (N=606) -479.46 Percent correctly predicted (null=53.8) 62.4 Note: See the Appendix for a description of the variables. *p<.05, one-tailed test. **p<.01, one-tailed test. ***p<.001, one-tailed test. Table 10. The Effects of the Deficit Reduction Vote on Individual Voting for the Incumbent in the 1990 Senate Elections (Post-Election Survey) Variable 1 2 Constant .018 .085 (.11) (.51) Party Same as Incumbent .746*** .574*** (8.81) (5.89) Democratic incumbent .395** .386* (2.31) (2.17) Deficit reduction vote -.709** -.682* (-2.60) (-2.43) Challenger's spending -.334*** -.359*** (-4.95) (-5.11) Incumbent's party better on deficit .522*** (4.08) Defict/Budget most important issue -.429* (-1.91) Log likelihood (N-367) -167.93 -156.88 Percent correctly predicted(null=68.4) 75.0 79.0 Note: t-statistics are in parentheses; challenger's spending is log of dollars per voting-age resident (Federal Election Commission 1991); see the Apppendix for a description of the survey variables. *p<.05, one-tailed test. **p<01, one-tailed test. ***p<.001, one-tailed test.