As Election Day approaches, many political commentators are asking whether the 2010 midterm elections could be a reprise of 1994, when Republicans picked up eight seats in the Senate and 52 seats in the House of Representatives to take control of both chambers for the first time in 40 years. There is almost universal agreement that Republicans are poised to make major gains in both the House and the Senate. And while the GOP's chances of gaining the 10 seats needed to take control of the upper chamber appear remote, results from the generic ballot forecasting model indicate that the 39 seats required to take back the House of Representatives are well within reach.

There are some striking similarities between the mood of the American public today and the mood of the country 16 years ago. The most important similarity is that President Obama, like President Clinton in 1994, has seen his approval ratings fall below 50% into what is generally considered the danger zone for an incumbent president and his party (Mellman Reference Mellman2010). The Democratic-controlled 111th Congress, like the Democratic-controlled 103rd Congress, is very unpopular, with a June Gallup poll showing an approval rating of 21% (Jones Reference Jones2010). And, according to the same poll, only 24% of Americans are satisfied with the way things are going in the country. Given these results, it is not surprising that Republicans have been running even with or ahead of Democrats when registered voters are asked which party they want to control the next Congress. This Republican edge also existed in the summer of 1994, and it represents a dramatic change from the 2006 and 2008 election cycles, during which the Democrats enjoyed large advantages on this generic ballot question.

THE GENERIC BALLOT MODEL

According to a statistical model that uses the generic ballot and presidential approval to measure the national political climate, Republicans have an approximately 50% chance of regaining control of the House of Representatives in November. The model uses four independent variables to predict seat change in congressional elections.Footnote ¹ Two variables measure the national political climate in the election year: the president's net approval rating as registered in an early September Gallup poll and the results of the generic ballot question in the same poll. The other two variables reflect structural features that are fixed for a given election: a dummy variable for midterm elections that is positive in Democratic midterm years and negative in Republican midterm years, and the number of seats held by Republicans before the election, which measures the relative exposure of both parties to risk—the more seats a party holds going into an election, the more seats it can expect to lose or the fewer seats it can expect to gain.

The generic ballot model does an excellent job of predicting the results of recent wave elections, especially for the House of Representatives. For 1994, using data from elections between 1946 and 1992, the model predicts Republican gains of 46 House seats and eight Senate seats, which comes very close to the actual gains of 52 House seats and eight Senate seats. For 2006, using data from elections between 1946 and 2004, the model predicts Democratic gains of 36 House seats and three Senate seats, also close to the actual gains of 30 House seats and six Senate seats.

Table 1 displays the results of a regression analysis of the outcomes of congressional elections between 1946 and 2008 using these four predictors. The dependent variable in this analysis is the change in Republican House or Senate seats. All of the estimated coefficients are in the expected direction and statistically significant. As expected, the more seats the Republican Party holds before the election, the fewer seats it tends to gain or the more seats it tends to lose; similarly, the larger the Republican lead or the smaller the Republican deficit on the generic ballot, the more seats Republicans can expect to win, and the more popular the Republican president or the less popular the Democratic president, the more seats the Republicans can expect to win. In addition, regardless of the president's popularity or the results of the generic ballot, the president's party tends to lose seats in midterm elections. This presidential midterm variable has a very large and highly significant negative coefficient, which results in the president's party almost always losing seats in midterm elections, especially in the House.

Table 1

Estimates for House and Senate Forecasting Models

Not surprisingly, given the relatively small number of Senate seats contested in each election and the larger proportion of closely contested races, the Senate model is not as accurate as the House model. Otherwise, however, the results of the two models are fairly similar once the different number of seats at stake in each chamber is taken into account. The main difference between the Senate and House results is that the seat exposure variable has a much larger impact on Senate elections than on House elections. In fact, the single strongest predictor of Republican seat change in Senate elections is the number of Republican seats at stake in the election. The larger impact of this variable in the Senate model probably reflects the greater competitiveness of Senate contests and therefore the greater risk associated with seat exposure in Senate elections.