If Republicans increase the number of seats they now hold in the United States Senate by seven, they become the majority party for the first time in eight years. They just might do that.
An equation developed by two political scientists who have studied the Presidential and Congressional elections held since World War II predicts a Democratic loss of between four and seven seats.
There are now 56 Democrats and 44 Republicans in the Senate. Senators will be elected to 21 seats now held by Democrats and 13 now held by Republicans between now and January. (The successor to David Boren, Democrat of Oklahoma, will be picked early in 1995.) A shift of seven seats would therefore give the Republicans a majority.
The equation is one of several developed by Michael S. Lewis-Beck and Tom W. Rice and published in their 1992 book Forecasting Elections. Political scientists have long shied away from forecasting the outcome of individual U. S. Senate races and certainly from predicting overall changes. But Lewis-Beck and Rice have tackled the job successfully.
Their equation considers three factors: the President’s popularity in the June Gallup Poll, the number of seats of the Presidential party that are being voted on, and the party of the Presidency. These three factors account for 83 percent of the shifts in Senate seats since 1948. The average error in the formula since 1948 has been 1.68.
The calculation (done by me, not the developers of the equation) comes out to -5.58. (The minus sign means a loss to the Presidential party). Applying the average error, the calculation works out to a loss of between 3.9 and 7.26 seats. The calculations assume that the relative importance of the factors remain the same as their 1948-to-date average.
The data that lie behind the equation show some interesting relationships, not all of them what you might think:
The performance of the economy, a strong factor in Presidential elections and congressional elections in Presidential years, is not so important in mid-term elections. In fact, the formula only considers the economy in Presidential years.
Eight Presidential popularity points in the June Gallup poll are worth about one Senate seat.
It doesn’t matter very much which party won the Senate seat six years before.
Democratic seats are somewhat more secure than Republican seats.
The equation for forecasting Presidential elections has correctly predicted all but one of the elections since World War II. The one incorrect prediction was the 1960 Kennedy-Nixon election. The Presidential model contains two more variables than the Senate model: party strength (the number of House seats the incumbent party lost in the last midterm election), and candidate appeal (how well did the candidate do in the primaries?)
In this equation, the only polling number used is the popularity number for the July before the election. That number here is more important than at mid-term. One Gallup point is worth .86 percent of the electoral college, or 4.6 electoral votes. A one percent improvement in the economy in the last six months is worth 7.76 percent, or 41.5 electoral votes.
Again, the data show an unexpected result: the July popularity poll (four months before the election) is a more accurate predictor of the electoral vote than the poll one month before the election. The model developed by the authors is more accurate than Gallup estimates based on day before election data.
The statistical techniques Lewis-Beck and Rice have developed are very impressive in analyzing past elections. I can’t wait to see how accurate they will prove in predicting the changes in the 34 U. S. Senate that will be decided by next year.
It will be even more fun in 1996 to analyze the Presidential model.
