Sandy Maisel's PARTIES AND ELECTIONS IN AMERICA

National or Local Concerns: The Race for the House

A couple of weeks ago I wrote about national political operatives trying to influence local congressional campaigns, because the stakes are so high for the national parties. This midterm election is all about control of the two houses of Congress. As I wrote last week, control of the Senate is very much up for grabs. On October 25th, the untimely death of Senator Paul Wellstone [MN], who was defending one of the most vulnerable seats in the Senate for the Democrats, only increased the uncertainty. With the exception of the Senate race in Minnesota, a great deal is known about each of the hotly contested races, though winners are not always predicable. What about the House?

Few of us claim to be expert on all 435 House races; certainly I do not. But I—and any careful political observer—can make some comments on how one should look at these 435 races. Look quickly back at Table 5.3 in the textbook. Note the incredible percentage of incumbents seeking re-election who do so successfully. Note also the percentage who win with more than 60% of the vote. The starting place for analyzing House elections, even more than Senate or gubernatorial elections, is that most incumbents win and win big. A corollary to that is that the outcomes of races involving most incumbents can easily be predicted well in advance of election day.

So we start any analysis of House races looking at the sure wins for one party or the other, mostly those seats in which incumbents are seeking re-election. The current breakdown of seats in the House of Representatives is 222 Republicans, 209 Democrats, one Independent, and three vacancies. The Democrats would have to make a net gain of seven seats (taking into account which party controlled the vacant seats) to win control.

A number of different analysts “rate” the two parties’ chances in all House seats. As I have written earlier, I tend to rely on the Cook Political Report and the Rothenberg Political Report. Their latest forecasts say that about 12-15 seats are all that should currently be considered toss-ups; they rate fewer than 30 other seats as not assured for one party or the other. These and other analysts all claim that the Republicans have an advantage, somewhere in the range of ten seats, among those races in which the victor seems to be assured.

Taking both analysts’ lists into account, I rate 39 seats as either only leaning one way or the other or true toss-ups. My count has the Republicans ahead in the safe seats three, 196-194. I also have the Republicans ahead in the seats in which they are leading, but not assured of victory. If one combines these two categories, the Republicans have 214 seats and the Democrats 204, with 16 as true toss-ups. Three of those sixteen seats are currently held by Democrats; eight by Republicans. The other five are new seats, created because of reapportionment and redistricting; in two of those (CT-5, and PA 17), two incumbents have been pitted against each other. To gain control, the Democrats would have to win 13 of these 16 races, assuming that Independent Bernie Sanders (VT) votes with the Democrats. That is a tall task. And that is why most analysts think that the Republicans are likely to retain control of the House. Analyzing on a district-by-district basis, I reach the same conclusion.

But political scientists for some time have attempted to predict the cumulative outcome of House elections using aggregate statistics. The best-known and most accurate of these models is one developed by Professor Gary Jacobson and the University of California, San Diego. Jacobson’s model incorporates polling data on the citizens’ evaluation of the President, on the change in real per capita income in the year prior to the election, and seat exposure to predict gain and loss by the President’s party. (The President’s party has lost seats in all but two elections in the last three-quarters of a century; the district-by-district predictions imply that the Republicans might well gain seats in the 2002 election.) While this model is theoretical compelling, it has not done terribly well at predicting results, explaining only about 70% of the variance in outcomes of those elections to which it has been applied (not a bad result, but also not useful if one wants to know which party will control the House.)

Alan Abramowitz, a political scientist at Emory University, has recently proposed a much simpler model, based on the so-called “generic vote” question that the Gallup Poll has been asking for more than half a century. Using only the results of that question, a measure of the partisan divide between the parties, and a variable noting the president’s party affiliation. The generic vote question captures long-term party affiliation and short-term responses to a campaign, the House partisanship captures incumbent advantage, and the president’s party affiliation captures any negative feelings toward the president’s party. Using these three simple variables, Abramowitz’s model explains 96% of the variance in the elections since 1950, with a standard error of only 5.4 seats. Very impressive results. (While this paper has not been published, Abramowitz has run quite rigorous tests on his model. I am sure he would share them with those who are interested.)

The Gallup Poll conducted earlier this month (October 3-6) showed the Democrats with a generic vote lead of one-percent over the Republicans, with five percent undecided. This number has held pretty constant for some weeks. The three-variable model Abramowitz has developed predicts a sizeable Democratic gain of 14 seats if the generic vote question results remain the same. Given the margin of error in his most rigorous tests, Abramowitz predicts (assuming the generic vote results do not change) that the Democrats will win between 214 and 238 seats, controlling the House for all results except those at the lowest end of the margin of error.

You might wonder why these results seem so at odds with the district-by-district analysis. Which should we believe? The sort answer is that these results stress national trends. It is difficult to see how the district-by-district analysts, who rely on constant reference back to knowledgeable informants in the districts, can capture these trends. Thus, the Abramowitz model would have picked up the Republican movement at the end of the 1994 election better than the political analysts did. This model also allows for changing predictions as the generic vote results change, with the approach of the election. (According to the model, the Republicans would have to lead in the generic vote test in order to control the House, though their control would be well within the margin of error were the generic vote tied.)

The national picture for this election is painted against a backdrop of a war on terrorism, a potential war with Iraq, burgeoning corporate scandals, a stock market that was in free fall for a number of weeks, questions about the state of the economy, and concerns about a number of specific issues like education, payment for health care, and social security. Whether these play an important role in determining the outcome of the election, and if so, to whose advantage, remains to be seen. So too does the question of whether local or national factors will have more of an impact as the toss up races and those in which one party or the other is leading but not assured of victory are decided. You should focus on these races and these factors as you observe the results on Election Night.