The 1996 House elections were unique in a number of ways. The Democratic presidential candidate, despite consistent leads of 15 to 20 percent, refused to ask the electorate for a Democratic Congress. Labor, under new leadership, spent more time and money trying to influence election results than it had in more than two decades. The Speaker of the House of Representatives became a focal point in congressional campaigns from Connecticut to California, belying the claim that congressional elections are purely local events. The most important aspect of the 1996 elections was that for only the third time since 1930 the Republican Party had an opportunity to retain control of the House of Representatives. From the Democratic takeover of the House of Representatives in 1930, two years before Franklin Roosevelt and the New Deal, until 1996, the Republicans had won only two majorities in the House-1946 and 1952-in 1948 and 1954 they lost this majority status.
Some commentators compared 1996 with 1948 in that the 1948 elections featured an incumbent Democratic president who, it was thought, after the loss of the 1946 elections would be a sure loser. Yet in a miracle comeback featuring President Harry Truman' s attacks on the "do nothing" Republican Congress, Truman was reelected and the Democrats were the majority party in the House. In 1952, Dwight Eisenhower led the Republicans to their first presidential win in twenty years and brought with him a Republican House. The victorious Republicans promptly began to fight over the extent to which, if any, the New Deal should be rolled back, and after November 1954 the Republicans were again the minority party. In the forty years that followed, the Republicans in the House were the minority and nothing seemed likely to change this status. Republican presidents won overwhelming victories--Nixon in 1972, Reagan in 1980 and 1984-and the House remained Democratic. Changing leaders from Joseph Martin to Charlie Halleck, Halleck to Gerald Ford, Ford to Robert Michael seemed irrelevant. Then in 1994 the Republicans won control of the House by using President Bill Clinton as a foil. Analyses showed that voting with Clinton in moderate-to-conservative districts cost thirty-four Democratic incumbents their jobs. In addition to the thirty-four incumbent seats, the Republicans picked up twenty-two of thirty-one open Democratic seats, giving them a fifty-six-seat swing.1
Could this anti-Clinton victory be sustained in 1996, given that Clinton seemed to have a sizable lead over Senator Dole?
The 104th Congress, the first Republican House in forty years, proceeded to push for and vote consistently for a conservative agenda. They voted to eliminate federal mandates on states, for a balanced budget amendment, and, more important, for cuts in expenditures designed to balance the budget by 2002. The shutdown of the government twice in late 1995 reversed the poll results for President Clinton; by the time of the party conventions in the summer of 1996 he was leading Senator Robert Dole by as much as twenty points in some polls. The analogies to 1948 were obvious, with many pundits predicting a Democratic Congress in 1997, and there seemed reason to believe it possible. President Clinton was clearly leading Senator Dole; the Democratic Party worked hard to recruit and fund good challengers to vulnerable Republicans; and labor spent on unprecedented amount of time and money attempting to defeat Republicans. The campaign themes were that these Republicans had gone too far and were too radical; Speaker Newt Gingrich became the Democrat' s foil in 1996 as Clinton had been for Republicans in 1994. Given that the Democrats needed to take only twenty seats to capture a majority in 1997, a repeat of 1948 and 1954 seemed possible. The actual election resulted in a nine- seat swing against the Republicans, and thus the Republicans retained the House in consecutive elections for the first time in more than sixty years. This essay analyzes why the Republicans kept their majority status and, as such, features analyses of the incumbency advantage, especially among freshmen; the effect of voting on reelection rates-did they go too far?-vulnerability in 1994; and an overview of the nature of the Republican majority in the 105th Congress. We conclude with an analysis of Republican prospects to the year 2000.
National Swing and Regional Results
Throughout the campaign major polls showed President Clinton leading Senator Bob Dole by double digits (sometimes as much as twenty-two points), which, combined with the size of the Republican win in 1994, meant that the electoral swing would go against Republicans. On election day President Clinton' s lead over Senator Dole fell to eight points, and the swing against Republican candidates for the House was 3.4 percent. That is, on average Republican congressional candidates received 3.4 percent fewer votes than they had in 1994. Of the twenty-six post-World War II elections, sixteen had swings lower than the swing in 1996; of those only four were against the Republicans. The five swings in the 3 percent range had an average seat loss of almost sixteen; however, the comparable swings against Republicans in 1970, 1982, and 1986 resulted in Republican seat losses of seven, twenty-six, and three, respectively. Thus the nine-seat loss for a 3.4 percent swing is lower than the Republican average of about twelve seats.
The vote swing against the Republicans varied by region. In the Northeast and Midwest the vote swing against Republican congressional candidates from 1994 to 1996 was on average about 5 percent. In terms of lost seats the swing translated into a six-seat loss in the Northeast and a four-seat loss in the Midwest. In the southern and border states the swing against Republicans was only 1 percent, and in this region Republicans picked up four seats. In the West the swing was 3 percentage points against the Republicans, and they lost three seats. Overall, the Democrats picked up thirteen seats in nonsouthern regions, while in the southern and border states, Republicans gained four seats for a net loss of nine seats.
When the 103d House began, Democrats were the majority party in every region, whereas with the opening of the 104th House, Democrats were the majority in only the Northeast. When the 105th House meets the Republicans will be the majority party in the Midwest (55 of 105), the southern and border states (86 of 148), and the West (51 of 93). Republican maintenance of their majority occurred despite the 3.4 percent swing against them nationally. The traditional explanations for why parties maintain congressional majorities despite their party' s foiled presidential candidates is incumbency. In 1972, 1980, and 1984 there were swings against Democratic members of Congress, and although they lost seats they nevertheless maintained their majority. In 1996, the Republican congressional candidates experienced a similar phenomena-their party' s presidential candidate loses, swing goes against them, they lose seats but retain the majority.
The swing in popular vote does vary by incumbency status. Previous research has shown that incumbents have a personal vote that is independent of the partisanship in the district. This personal vote is measured in two ways: (1) when incumbents retire their party' s vote share declines (retirement slump), and (2) in incumbents' first run for reelection, their vote total increases (sophomore surge). Over the last twenty years, retirement slump has averaged about 9 to 10 percent, whereas sophomore surge has been about 7 or 8 percent. These figures have been down in the last two elections-1992, 1994, and 1996 followed this short-run tradition. In 1996, the slump was around 7 percent and sophomore surge was about 3.7 percent. The incumbency advantage was strong overall as more than 95 percent of all running incumbents won. Even though Republicans lost nine seats overall, about 92 percent of Republican incumbents were reelected. To lose the House in 1996, the Republicans would have had to lose many of the freshmen from the 104th House because Republicans were supposed to gain seats from Democrats in open southern and border seats. We now turn to in-depth analyses of open seats, voting records and their election impact, the freshmen, and the effectiveness of labor' s targeting Republican incumbents.
In 1994 Republicans captured twenty-two of thirty-one open Democratic seats, which was crucial for the Republican takeover. The open seats captured by Republicans in 1994 were those where President Clinton had not done well in 1992 (i.e., conservative-to- moderately conservative districts). In the 1996 election, there were fifty-three open seats (thirty Democrat and twenty-three Republican) and, overall, the Republicans picked up six seats. By region, in the Democratic Northeast there were seven open seats (four Republican) and of these there were no party switches. In the Midwest there were fourteen open seats (nine Republican, five Democrat); six seats stayed Republican, three stayed Democrat, two switched from Democrat to Republican, and three switched the other way, giving Democrats a one-open-seat pickup in the Midwest. In the southern and border states there were twenty-four open seats, of which four stayed Republican, twelve stayed Democrat. Of those southern seats that switched, seven went from Democrat to Republican and only one went the other way, giving the Republicans a gain of six. In the West, there were eight open seats (five Republican and three Democrat). All five Republican seats remained Republican, and one Democratic seat went to the Republicans.
The Republican gains in open seats followed the 1994 pattern in that districts that switched to Republican tended to be conservative as measured by President Clinton' s 1992 vote total.
Open districts that did not switch parties were conservative or liberal-in Republican (both 1994 and 1996) districts, Clinton averaged 36 percent of the vote; in Democratic (both 1994 and 1996) districts, Clinton averaged 52 percent. In districts going Republican the Clinton vote total was less than 40 percent; in districts going Democratic it was more than 40 percent. The swing figures show that, in districts that stayed consistent, there was only a slight swing against Republicans, whereas the switched districts show major swings. Therefore, Republicans continue to gain seats from Democrats in districts where the average voter is conservative. The southern and border states are the most conservative districts on average, which is where Republicans continue to gain seats. Moreover, these are the areas where, over the next two elections, we expect to see more Republicans gain open seats.
The Republican majority pursued an aggressive legislative agenda that included reforms of House procedures, balancing the federal budget, reducing the size of appropriations, and reforming regulatory and litigation decision processes. Voting behavior, as measured by the Americans for Democratic Action (ADA) scores, provides a useful index of just how conservative this agenda was. The ADA, which has been publishing its voting scores since 1947, measures the proportion of times a member of Congress has supported the liberal position on a major vote. The higher the score, the more liberal the member's voting record. The ADA chooses the votes that it considers major and classifies the vote as liberal or conservative. An ADA high vote score is based on votes against the Republican balanced budget package, product liability and regulatory overhead, election spending increases, and a prohibition on abortion procedures for federal workers. During the first session of the 104th Congress, more than half of all Republicans had 0 percent ADA vote scores and about 90 percent had ADA vote scores less than or equal to 20 percent. Their average ADA vote score was only 6 percent.
Figure 1 compares the voting record of the Republicans in the 104th Congress to that of Republicans in each Congress since 1946. Although a comparison of ADA vote scores across Congresses is fraught with difficulties, the chart is nevertheless instructive.
2 Throughout the post-World War II era, House Republicans averaged a 15 percent ADA score. The prior House Republican record for conservative voting, an ADA score of 11 percent, was set in 1947. The 6 percent average ADA vote score among Republicans in the 104th Congress makes it the most conservative on record.
In an earlier paper on the 1994 congressional elections, we found that the voting behavior of incumbent Democrats contributed to their losing control of the House. Specifically, we found that, among Democratic incumbents from moderate-to-conservative districts, the greater their voting support for the president's policies, the higher their probability of losing.
Democratic incumbents in districts in which Bill Clinton received between 40 and 50 percent of the 1992 presidential vote and who supported the president less than 60 percent of the time fared well in their reelection bids. Incumbents from similar districts who supported the president at least three-fourths of the time, however, suffered large reelection losses. The results are more pronounced in districts in which Bill Clinton received less than 40 percent of the 1992 presidential vote.
In this essay, we employ similar tests on Republican incumbents in 1996. Following the approach in our earlier paper, we classify districts as moderate to liberal if the district' s 1992 popular vote for Bill Clinton exceeds 40 percent of the total votes cast.3 Mr. Clinton won 80 percent of these districts.
Republican incumbents seeking reelection are classified into three groups based on the degree of conservatism of their voting records. The first category, labeled "perfect conservative," consists of members with an ADA score of 0 percent. These are members who always voted for the conservative position on major pieces of legislation in 1995. The second category, labeled "strongly conservative," contains members with ADA scores of above 0 percent but less than or equal to 20 percent. These members voted at least once, but less than or equal to 20 percent of the time, for the liberal position on major legislation in 1995. The third category, labeled "moderately conservative," consists of members whose ADA scores in 1995 were greater than 20 percent, that is, members who voted more than 20 percent of the time for the liberal position on major legislation.
In districts where Bill Clinton received between 40 and 50 percent of the 1992 presidential vote, there is no systematic difference in reelection rates. An equal percentage of those with perfect conservative voting records and with moderately conservative voting records were reelected. There are too few incumbent Republicans in districts where Bill Clinton received more than 50 percent of the 1992 presidential vote to conclude much of anything. The results for all incumbents from moderate-to-liberal districts reported in the last two column' s resemble those of the moderate districts. There is no evidence that voting behavior harmed the reelection prospects of Republican incumbents from moderate-to-liberal districts.
The criteria for classifying congressional districts as moderate to liberal are arbitrary. To ensure that our results are not driven by choosing 40 percent of the 1992 presidential vote as the threshold for classifying districts as moderate to liberal, we also computed reelection rates using a higher threshold. Under this threshold, a district is classified as moderate to liberal if the 1992 presidential vote for Bill Clinton exceeded the national average of 44 percent; thus, twenty-three districts with a Republican incumbent who stood for reelection are classified as moderate to liberal. In 1992, President Clinton won more than 90 percent of these districts, but there is no evidence that Republican incumbents were harmed by conservative voting. Republican incumbents representing six of these districts had perfect conservative voting records. All six were reelected. Eleven incumbents had strongly conservative voting records, and seven were victorious. An additional six incumbents had moderately liberal voting records. Five of them were returned to office in 1996.
The second measure of voting behavior used in this essay is the incumbent' s support for the Contract with America as measured by member votes on a dozen bills in 1995 containing Contract with America provisions, including the constitutional balanced budget amendment, regulatory reform, term limits, the balanced budget package of tax and spending cuts, Clean Water Act revisions, and welfare reform.
These results are essentially the same as those using ADA scores and are decidedly different from our earlier results for Democrats in 1994. Although Republicans with only moderate support for the Contract with America have marginally higher reelection rates than those with perfect support, the difference is not statistically significant.
Our sophisticated empirical analysis confirms these conclusions. The detailed results of this analysis are presented in the appendix. Briefly, using a maximum likelihood statistical method similar to regression analysis, we estimated the impact of voting behavior on the
probability of Republican incumbents losing reelection. The estimated impact of voting behavior controls for the district' s 1992 presidential vote for Bill Clinton, whether or not the incumbent was targeted by the AFL-CIO, and the prior electoral experience of the Democratic challenger. Estimates were made separately for all Republican incumbents and for freshmen Republican incumbents. None of the estimates of voting behavior were even close to approaching statistical significance at the 5 or 10 percent confidence level.
Organized Labor's Campaign
In January 1996, John Sweeney, the newly elected president of the AFL-CIO, announced that organized labor would mount an extraordinary political campaign against incumbent Republicans. The AFL-CIO promised to spend $35 million to unseat Republicans who had voted against the union' s position on key issues. This amount represents the largest independent political campaign expenditure by a single organization in American politics. The 1996 elections were characterized by independent political groups, such as the Sierra Club and other interest groups, choosing to spend their resources themselves rather than through the official campaigns of candidates for office. The $35 million labor expenditure, nevertheless, stands out as an extraordinary sum.
The initial announcement called for the campaign to target seventy-five Republican incumbents, forty of whom were freshmen. (The Democrats had to capture only nineteen spots in order to recapture the House.) The main thrust of the campaign was television and radio ads attacking Republicans for their votes on Medicare, education, and the minimum wage. The AFL-CIO' s campaign, however, also included hiring state field directors and staff for get-out-the-vote efforts, mass mailings, voter guides, and other campaign activities.
Information on how much the AFL-CIO spent, in which districts, and how it allocated the money among TV ads, radio ads, and other campaign activities is scant. Since the AFL-CIO's campaign was independent, it is not required to make the details of its expenditures public. Two measures of the AFL-CIO's campaign at the congressional-district level, however, are publicly available, both of which were compiled by the Congressional Quarterly. The first is whether or not the AFL-CIO ran any TV or radio ads in the district. The second is whether or not the AFL-CIO ran electronic voter education guides in the district. The voter education guides are a better indicator that the union' s effort in the district was relatively intense. Therefore, we have used them to measure the impact of the AFL-CIO' s campaign.
Under this measure, 35 districts were heavily targeted by the campaign and 178 districts were not. The reelection rate of incumbent Republicans in targeted districts (71 percent) was far lower than it was in districts that were not targeted (96 percent). The lower reelection rate, however, does not necessarily reflect the impact of the AFL-CIO' s campaign, for the campaign targeted Republican incumbents that were particularly vulnerable. Similarly, it avoided targeting safe Republican districts.
For targeted Republican incumbents, the average margin was 8 percentage points in 1994. In contrast, the average 1994 margin among incumbents whom the AFL-CIO chose not to target was 32 percentage points. The columns show the proportion of targeted and nontargeted districts that are considered vulnerable districts versus safe districts. (Vulnerable districts are defined as those in which the Republican incumbent won his 1994 election with less than 55 percent of the popular vote; safe districts are those in which the Republican incumbent won his 1994 election with greater than 60 percent of the popular vote.) Among those districts targeted by the AFL-CIO campaign, 37 percent were held by vulnerable Republicans. In contrast, only 15 percent of the districts the AFL-CIO avoided were vulnerable. Only 9 percent of the targeted seats were safe seats, whereas 75 percent of seats avoided by the AFL-CIO were safe seats. In view of the fact that the AFL-CIO campaign disproportionately targeted Republican incumbents in vulnerable districts, determining the precise impact of the campaign requires the use of more than simple cross-tabulations.
In our more sophisticated empirical work, we estimate the partial impact of AFL-CIO targeting controlling for two measures of vulnerability: the degree of liberalism in the district, as measured by the percentage of the district' s 1992 presidential vote for Bill Clinton, and the incumbent' s vote margin in the preceding House election. The details of our statistical work are reported in the appendix.
In our statistical work, we find that, after controlling for the incumbent's vulnerability, Republican incumbents in AFL-CIO targeted districts had a higher probability of losing their reelection bids. The effect is large and statistically significant. Being on the receiving end of the AFL-CIO' s campaign decreased the likelihood that a Republican incumbent would win reelection by 56 percent. Perhaps surprisingly, the AFL-CIO campaign' s impact was felt only among nonfreshmen incumbents. We find no evidence that the targeted freshmen were adversely affected by the campaign.
The finding of a statistically significant impact for nonfreshmen has to be placed in context. There were only five nonfreshmen targeted by the AFL-CIO campaign: Gary Franks (Conn.), Peter Torkildsen (Mass.), Bob Franks (N.J.), James Walsh (N.Y.), and Martin Hoke (Ohio). Three of these, Gary Franks, Hoke, and Torkildsen, were defeated.
In many respects, the fate of the Republican control of the House hung on the freshman class. Seventy-four Republicans were elected for the first time to the House in 1994, the largest freshman class in the House in years. In 1996, seventy-one of them chose to stand for reelection. With only a nineteen-seat advantage, significant losses among the freshmen would end Republican control over the House. Freshmen are traditionally more vulnerable to defeat than more- senior incumbents, especially in recent years. First, the proportion of freshmen who lost in the early 1990s exceeds the proportion who lost in any of the earlier three decades. Second, from the 1960s until the 1980s, freshmen could count on between a 6 and 8 percent increase in their popular vote as they realized the benefits of incumbency. In the early 1990s, this increase has averaged only about half that amount.
While in office, the Republican freshmen acted with an extraordinary degree of philosophical cohesion, quickly becoming an influential voting block on the key issues that dominated the 104th Congress. From the seven-year timetable for a balanced budget to a terrorism bill that expanded the role of the Federal Bureau of Investigation, the freshmen were instrumental in preventing the Republican caucus from veering from its conservative path. And, on occasion, they displayed a degree of independence, refusing to follow the age-old advice given to all freshmen: "To get along, you have to go along." For example, when senior Republicans voted to expand defense appropriations without offsetting reductions in other programs, the freshmen bolted and joined with Democrats to defeat the defense appropriations conference report.
Many Republican freshmen adopted the attitude of classmate Steve Largent (Okla.), who summarized how he would like his legislative career to be remembered by saying "brilliant but brief." Their voting record, as measured by their ADA voting scores, reflected this attitude. Thirty-seven of the seventy freshmen who stood for reelection had a perfect conservative voting record (a 0 ADA score) during their first year in office. Another seventeen freshmen maintained a ADA score of less than 10 percent. Only three freshmen had an ADA score of greater than 20 percent.
The aggressive pursuit of a decidedly conservative legislative agenda and a consistent voting record in support of this agenda caused many observers to opine that the freshmen had put their election prospects at high risk. If true, they jeopardized Republican control over the House.
Voters, however, returned fifty-nine of the seventy Republican freshmen who stood for reelection in 1996 to office. By historical standards, this 84 percent reelection rate is only slightly lower than normal for freshmen. For example, the large Democratic freshman classes of 1972, 1974, 1976, and 1978, each of which numbered about fifty members, had retention rates of 81, 96, 96, and 100 percent, respectively. In contrast, the Democratic freshman class of 1964, out of which thirty-five members stood for reelection, had a reelection rate of 81 percent.
A deeper understanding of how the freshmen fared can be gained by examining the swing (the average change) in the popular vote between 1994 and 1996. n average, Republican freshmen who stood for reelection in 1996 received the same share of the popular vote in their districts as they received in 1994. Senior Republican incumbents, in contrast, received about 5 percentage points less, for a net swing for Republican freshmen of a positive 5 percentage points. The Democratic freshmen in 1996 had a 2.2 percent advantage over senior Democrats. The sophomore surge to Republicans is close to the post-1964 average of 6 to 8 percent, while the sophomore surge to Democrats is less than half the Republican number. Part of these differences can be explained by looking at the characteristics of the districts represented by each group.
Republican freshmen appear to have solidified voter support in conservative-leaning congressional districts, maintained their winning advantage in moderate districts and, as expected, suffered their largest losses in liberal-leaning districts. There were thirteen Republican freshmen who were from congressional districts in which President Clinton received less than 35 percent of the popular vote in 1992, compared with his national total of 44 percent. All thirteen freshmen from these conservative-leaning districts were reelected. In the process, they increased their share of the popular vote in their districts by nearly 3 percentage points. Forty-eight Republican freshmen came from districts in which President Clinton received between 35 and 45 percent of the popular vote in 1992. Forty-one were returned to office, and on average they maintained their winning popular vote advantage. Nine Republican freshmen came from districts in which President Clinton received more than 44 percent of the popular vote in 1992. In these liberal-leaning districts, four Republican freshmen were defeated; on average their share of the popular vote declined by a modest 2 percentage points.
The 1996 congressional House elections represent trends that have been under way for decades. Since 1964, Republican congressional candidates have continued to run well in southern and border states. They now have a solid majority and are likely to continue to pick up additional seats in this region. From Ronald Reagan's victories in 1980 through 1996, Republicans have gained seats from Democrats by defeating Democratic incumbents at a higher rate than Democrats have defeated Republican incumbents. In addition, since 1980 Republicans have captured more Democratic open seats than vice versa. This represents a decided change from the 1964-1979 period, when exactly the reverse occurred.
A third trend has been that the members of the minority party have retired at a faster rate than the members of the majority party. In general, this results from the fact that being in the minority means having little or no influence over the legislative agenda, process, or policy outcome. Thus, over the next four years, we expect Democrats to retire at higher rates than Republicans, especially given that the vast majority of sitting Democrats were present during the years in which they were the majority.
A final trend benefiting Republicans is that, in every election in every off-year (nonpresidential) congressional election since the Civil War, except one, the president's party has lost seats in the House. With House Democrats cognizant of this fact, numerous retirements among senior Democrats are likely.
To obtain a more precise estimate of the importance of each of the aforementioned factors, we turn to statistical methods that permit us to isolate the separate effects of each factor and thus to estimate the impact of each factor on the likelihood that a Republican incumbent would be beaten by a Democratic challenger. This procedure is a maximum likelihood technique known as probit analysis.
We used two samples for our analysis. The first consists of all 209 Republican incumbents who sought reelection in 1996. The second consists of the seventy Republican freshmen who sought reelection. In each sample, our statistical analysis seeks to explain the probability that a Republican incumbent would be defeated. To explain this outcome, we have included four key variables. The first is the percentage of the popular vote for Bill Clinton in the district' s 1992 presidential election. This variable is a crude index of the district' s degree of liberalism. (Ideally, we would also have the district' s 1996 popular vote for President Clinton, but at the time of writing, the 1996 presidential vote by district was unavailable.) The second variable is the incumbent' s voting record. Two measures of incumbents' voting records are the ADA' s voting score in 1995 and a vote score on Contract with America votes in 1995. The third variable, compiled by the Congressional Quarterly, indicates whether the Republican incumbent was targeted by the AFL-CIO political campaign. It assumes a value of one if the AFL-CIO showed its voter education guide in the member' s district and zero otherwise.
(We also experimented with a measure of each Democratic challenger' s electoral experience. This variable is assigned a value of one if the challenger had held any elective office before the 1996 elections and zero otherwise. This measure could not be obtained for all challenges and has been dropped from the results reported below.)
1 The Democrats won four open Republican seats in 1994, reducing the Republican gain to fifty-two seats.
2 The principal difficulty arises from the fact that the legislation on which votes are cast differs from one Congress to the next. One Congress may produce legislation that is more liberal than the next. Identical ADA vote scores in the two Congresses would not necessarily reflect the same degree of liberal voting. Grossclose and Snyder (unpublished manuscript, Massachusetts Institute of Technology, 1995) devised a technique to adjust these votes for this problem. Even with this adjustment the 1994 ADA scores are lower than the earlier scores.
3 Recall that Bill Clinton received 44 percent of the nationwide popular vote in 1992.