Down to the Wire

Wednesday, October 30, 2002

recent edition of The Economist magazine shows President Bush walking with his dog under a headline reading, “It’s the economy, Boss.” Suddenly, the onslaught of dismal economic news seems to be weighing heavily on the president and his party. The New York Times, the Washington Post, CNN, and the political chat shows have all been predicting that the November midterm congressional elections will turn out badly for the president and the Republicans. NPR recently cited an increase in the number of Americans who felt that the country was headed in the wrong direction as evidence that the Republicans were in trouble. This “new” conventional wisdom is a major shift from the earlier assumption that the 2002 elections would be characterized by little or no change.

Pundit opinion has shifted as a variety of factors seen as major liabilities for the Republicans converge: the precipitous decline in the stock market, the anemic state of the economy, and the rash of corporate accounting scandals. In short, the new received wisdom is that the Democrats will retake the House of Representatives and increase their majority in the Senate.

History also seems to agree with this new assessment. Since 1860, with only two exceptions (1934 and 1998), the incumbent president’s party has lost seats in midterm elections. In the post–World War II era, the average result for a president’s first midterm election is a loss of almost 26 seats. The three most recent Democratic presidents (Johnson, 1966; Carter, 1978; and Clinton, 1994) lost an average of 29 seats, whereas recent Republican presidents (Eisenhower, 1954; Nixon, 1970; Reagan, 1982; and Bush, 1990) lost about 17 seats on average. The smallest first-term losses were Kennedy in 1962 and Bush in 1990—both losing fewer than 10 seats. The only Republican president to go into a first midterm election with a congressional majority was Eisenhower in 1954; in that election the Republicans lost 19 House seats and control of Congress.

Given these factors—the economy and midterm history—it is no wonder that Democrats are counting on retaking the House of Representatives and increasing their narrow lead in the Senate. Current House Minority Leader Richard Gephardt has told senior Democrats that the party could pick up as many as 40 House seats if the unfolding corporate scandals remain on the political “radar screen” through November.

Our assessment of the 2002 congressional elections differs, in that we believe the elections will be exceedingly close. What basis do we have for this view, which goes against the “new” consensus? The answer lies in the way that we measure and assess electoral vulnerability and the probability of defeating an incumbent. In what follows we show, by using a consistent analysis based on a congressional district’s ideological tendencies, that the 2002 elections could be closer than the 1998 elections. Indeed, we may be entering a period of very close elections, such as those seen in the early 1950s.

Vulnerability and Trends

Pundits, columnists, and television hosts who fill the air with predictions about election outcomes use polls, generic ballot questions, seat-by-seat analysis, intuition, and the herd mentality to foretell results. Our strategy focuses on incumbency and open seats—and historical trends—as the way to understand congressional elections. The idea is simple, and the technique we use yields the same measure for all post–World War II elections and thus has some scientific credibility. Since the 1950s incumbent members of Congress have insulated themselves from national electoral trends by appealing to voters on an individual as well as a partisan basis. The result is that the average incumbent has built an 8–10 percent cushion against electoral trends that adversely affect her or his party. This “personal vote” helps explain how Eisenhower, Nixon, Reagan, and Clinton could win presidential elections by large margins and not bring congressional majorities with them. In short, incumbents vote and behave in ways that appeal to their constituents even when those positions diverge from their party’s positions.

The personal vote manifests the incumbent’s advantage: more money, greater name recognition, access to the media, free mail to constituents, permanent staff to assist constituents, the ability to bring political heavyweights and celebrities to the district for fund-raising, and so on. It is no wonder that in the postwar era members of Congress who seek reelection win more than 95 percent of the time. The hard part, then, is predicting which incumbents are vulnerable. And, to complicate matters, vulnerability depends on national trends. That is, a vulnerable incumbent is even more vulnerable when there is a national swing against his or her party and, of course, less vulnerable when there is a national trend toward her or his party.

Because very few incumbents get defeated and most members run again, it is in the open seats (no incumbent running) where most electoral change occurs. When a House seat opens, a number of events occur: Good candidates decide to run, both parties put money into the campaign, and, most important, the seat typically becomes more competitive. We use the term more competitive rather than simply competitive because some seats are safe for one party even when the seat is open. Thus, as in the case of incumbents, we need to ascertain vulnerability to capture by the other party.

The first step is to analyze presidential voting data for each congressional district, plus some demographic data. This allows us to ascertain the liberal (Democratic) or conservative (Republican) profile for each district. Then we estimate what each incumbent’s ideological record should be, given the profile for his or her district. Based on the difference between this predicted value and the actual ideological value for an incumbent (based on scores from Americans for Democratic Action), we can identify those incumbents who are “out of step” with their district and thus more vulnerable than other members. Not surprisingly, it turns out that the Democratic incumbents who lose are too liberal for their districts and that the Republicans who lose are too conservative for their districts.1 For open seats, we use roughly the same measure of the district’s tendency as we use in incumbent elections to estimate the probability that an open seat will go with one party or the other. Having measured the vulnerability of open and incumbent seats in this way, however, does not solve the problem of trends. A political trend toward one party will affect the election results by tipping close races in that direction.

House Projections

To solve this problem we chose two elections that represented trends—one Democrat and one Republican. Both elections featured an incumbent president’s first midterm election: 1978 for a Republican trend (Democrats lost 15 House seats) and 1982 for a Democratic trend (Republicans lost 26 House seats). Using our method, we determined that the number of vulnerable incumbents for the 2002 elections was 60, with 42 Democrats and 18 Republicans in the vulnerable category.2 (The technique simply plugs in the values based on the 1978 and 1982 elections and then estimates the 2002 election; see table 1.)

Republican Trend

Under the conditions of a Republican trend (1982), the vulnerable seats (42 Democrat, 18 Republican) yielded the Republicans a 4-seat gain, which gave the party a 203-to-190 advantage in base seats.3 The 42 open seats (24 Republican and 18 Democrat), combined with the probability that an open seat changes hands in a Republican year, yields a 24-to-18 Republican advantage in the open seats and a 227-to-208 Republican House.

For there to be a Republican trend the party would need as favorable a set of issues as they had in 1994 to move voters toward them. The only issues we see as capable of generating any Republican swing are those surrounding September 11 and homeland security. The president’s jump in popularity after September 11 has largely stayed in place, and Republicans are generally perceived as better on this issue than Democrats. The Democratic incumbents’ strategy has been to be very supportive of the president on all security issues, and the more vulnerable the incumbent, the more supportive she or he has been on national security issues. Thus we believe that the national security issue will not result in a Republican trend.

Democratic Trend

In the event of a Democratic trend year, we project that 7 Republicans will lose, giving the Democrats a base of 201 seats. Predicted gains for open seats give the Democrats 23 of the 42 open seats, for a Democratic House majority of 224 to 211.

For this scenario to occur the Democrats need a powerful issue. Among the issues they have tried to ride to victory are “what did the president know in regard to September 11 and when did he know it”; Enron, WorldCom, and corporate scandal generally; and economic issues such as the deficit, economic growth, privatizing Social Security, and so on. Issues about the president’s knowledge of September 11 came and went as it quickly became clear that President Bush and the Republicans were not vulnerable and that there existed a potential backlash against Democrats who used this issue. We do not think the corporate scandals on their own are the Democratic issue because House and Senate Republicans have voted with Democrats on tough legislation. Republicans might have been vulnerable, but they have been tough enough on the issue to make it hard for incumbents to be held accountable.

Another and perhaps more powerful reason we believe that the issue will not hurt Republicans is that the money trail from discredited executives shows a mixed pattern of giving. Kenneth Lay and Jeffrey Skilling of Enron gave mostly to Republicans but also gave to seven Democratic senators and to at least four Democratic House members. In contrast, Paul Allaire of Xerox and Samuel Waksal of ImClone gave exclusively to Democrats, including Hillary Clinton, HillPac, the Democratic National Committee, Tom Daschle, Dick Gephardt, Al Gore, Bill Bradley, Emily’s List, and more than a dozen Democratic senators and a handful of representatives. Bernard Ebbers of WorldCom gave about equally across parties, covering all relevant members of House and Senate committees affecting his business. John Sidgmore of WorldCom, in contrast, gave exclusively to Democrats, including the DNC. John Rigas’s (Adelphia) money went mainly to Republicans with some scattered Democrats benefiting from his contributions. In short, it does not seem clear that the corporate issue is one that the Democrats can make stick.

The major issue that pundits believe can hurt Republicans—and make it a Democratic trend year—is the state of the economy and the stock market. Journalists and pundits have focused on this as the Achilles’ heel of the Republicans, suggesting that the president and his party will be held responsible for these events and that voters will turn against Republican candidates. As Dan Balz and Dana Milbank noted in the Washington Post recently: “With the stock market sell-off, corporate accounting scandals and voter anxiety about shrunken retirement funds and the future of the economy, Bush knows from his father’s experience what inattention to economic matters can do to a presidency.”

The argument is that the president and his party will be held responsible for the economy and that voters will turn against Republican candidates. (The two trend years we chose—1978 and 1982—were both associated with significant downturns in the economy.) Our view is that the economy looks like it will grow in the 1.5–2 percent range in 2002, which generally signals a no-trend election. A slow recovery with a depressed stock market, however, could anger voters. But in that event, the Democrats will still have to convince voters that they—not the Republicans—can ensure growth in the economy and the market. This is likely to be a difficult task; it is not easy for challengers to convince voters that specific candidates are responsible for the overall state of the economy. In addition, Republican incumbents are separating themselves from policies that are not popular with their voters.

No Trend

In all likelihood, the final numbers in the 2002 election will be between the two trends shown in table 1 because there will be no partisan trend. In this case, the 194 Democratic and 199 Republican incumbents go relatively loss-free, with each party holding almost all of the seats where incumbents are now running.

In that case, the election will turn on the 42 open seats (no incumbent running), of which the Republicans currently hold 24. In these seats the number of vulnerable Republican seats is less than 30 percent, and, given redistricting, the Republicans might pick up three to five seats. Again, it is hard to see how there would be a major swing to the Democrats in these redistricted seats. Under any of these scenarios the congressional elections will be very close. In fact, using our model to measure all post–World
War II elections reveals that only the 1998 and 1954 midterm elections look to be as close. Thus, contrary to the received wisdom, this election will be exceedingly close, with incumbents faring very well.

Senate Projections

Elections in the Senate are not as predictable. First, the incumbent reelection rate is only around 66 percent. Second, the money difference between challengers and incumbents is not prohibitive. Third, Senate campaigns in most states are media events; thus it is easier for challengers to overcome incumbent advantages. And, finally, voters pay more attention to Senate elections than to House races. Given this, here we list the vulnerable Democratic and Republican seats, offer our opinion on the likely outcomes, and then simply add up the results.

There are 34 Senate seats up for election this year, with 30 incumbents up for reelection (16 Republicans, 14 Democrats). All four open seats (North Carolina, South Carolina, Tennessee, Texas) are Republican, meaning that, of the 34 seats up for election, 20 are currently in Republican hands. In each of the four open seats, the Republicans have the partisan advantage and have recruited and nominated good candidates. Thus we expect all four open seats to stay in the Republican camp.

Of the incumbent races, the early consensus is that the races in Arkansas, Colorado, Georgia, Iowa, Louisiana, Minnesota, Missouri, Montana, New Jersey, New Hampshire, Oregon, and South Dakota will determine who wins the Senate (see table 2). The seats Republicans now hold and could lose are Arkansas, New Hampshire, Oregon, and Colorado. We predict that the incumbent Republicans in Oregon and Colorado will be reelected. And while John E. Sununu’s victory in the September 10 New Hampshire primary boosts Republican chances, that race will still be a toss-up. Senator Tim Hutchinson in Arkansas is likewise in a very close race that will go down to the wire. Thus we see only two seats that Republicans stand a good chance of losing.

As for the Democratic incumbents, in Georgia, Montana, and Louisiana the Democrats are the likely winners even though President Bush carried each of those states. Representative Greg Ganske poses a challenge to Senator Tom Harkin of Iowa, but that seat is also likely to stay in the Democratic camp. This leaves Minnesota, Missouri, New Jersey, and South Dakota as the most likely seats to go Republican. Minnesota voted for Gore and is in the Progressive tradition of Hubert Humphrey; nevertheless, the Republicans in Minnesota have a good centrist candidate who is campaigning on the theme that Senator Paul Wellstone is too liberal to be effective. This, plus the Green Party running a Senate candidate, makes this election a very close call. In New Jersey, Senator Robert Torricelli’s recent admission of ethics violations has made this race close, although New Jersey is a Democratic state.* Both Minnesota and New Jersey are probably too close to call.

The two most likely Republican pickup states are Missouri and South Dakota, where majorities voted for Bush in 2000. In Missouri, incumbent Senator Jean Carnahan has never run a full campaign and her inexperience has given challenger Jim Talent a slight lead in the polls. In South Dakota, a strong Bush state, the president is pitted against South Dakota’s senior senator, Majority Leader Tom Daschle. Here, the Republican candidate, Congressman John Thune, has a lead over the incumbent senator, Tim Johnson. Thus, there are four races where Republicans could pick up seats: two are toss-ups (Minnesota and New Jersey) and two are states where the Republicans have a slight lead (South Dakota and Missouri).

Vast resources from both parties and many political interest groups are pouring into the 12 most competitive states. Any small trend toward either party could tip these races, which means that the Senate majority could be anywhere from 53 Democrats to 53 Republicans. The final outcome will probably hinge on one or two seats at most—with an evenly divided Senate or even a two-seat Republican majority strong possibilities. Just as with the House, this Senate election promises to be one of the closest midterm elections since 1954, when Senate control hung on a single seat.


1. For the purposes of this article we used the average of two estimations: one where “vulnerable” was defined as members more than 1.25 standard deviations in an appropriately liberal or conservative direction from their predicted score for their district, and one where the vulnerability standard was 1.5 standard deviations of difference.

2. This disparity in “at-risk” seats seems to favor Republicans on the surface, but, historically, “out-of-step” Democrats have managed to win reelection more often than their Republican counterparts.

3. We are defining “base” as the total number of incumbents running, after accounting for partisan losses or gains in those seats. Thus after the base is established, it is the open seats that matter for control of the chamber.

* This article was written on September 15. Two weeks later, on September 30, Senator Robert Torricelli withdrew from the New Jersey Senate race, confirming our prediction that his seat was in danger. Torricelli cited a desire to help Democrats retain control of the Senate and suggested that a substitute candidate not dogged by scandal would win. Torricelli's withdrawal and the subsequent attempts to place former senator Frank Lautenberg on the ballot in his place have thrown the race into chaos.