In the Senate data, the point for 2002 is also well above the regression line while points for several other elections including 19 are also well above or below the line. In the House data, the only point that appears to be conspicuously far removed from the line is the one for 2002 - the midterm election that occurred shortly after the 9/11 attacks at a time when the incumbent president, George W. The closer fit of the data points to the prediction line in the House data is very clear from these figures. The line in each graph is the regression line generated by the equations in Table 2. The size of the seat exposure effect is especially noteworthy considering that the number of Senate seats in play is only about one-twelfth of the number of House seats in play.įigure 1 displays a scatterplot of the relationship between the forecasts generated by the generic ballot model and the actual seat swing in House and Senate elections between 19. Based on the Senate results, a one-point increase or decrease in the generic ballot margin for the president’s party would be expected to produce a swing of about 0.2 seats and every additional seat defended by the president’s party would be expected to produce a net loss of more than 0.8 seats. Based on the House results, a one-point increase or decrease in the generic ballot margin for the president’s party would be expected to produce a swing of about 1.7 seats and every additional seat defended by the president’s party would be expected to produce a net loss of more than 0.6 seats. The results in Table 2 show that the generic ballot and the number of seats defended by the president’s party have strong and statistically significant effects in both House and Senate elections. Table 2: Results of regression analyses of House and Senate seat change in midterm elections, 1946-2018 This is not surprising given the much smaller number of Senate seats contested in each election and the larger potential impact of local factors such as candidate quality and fundraising. The Senate model is not quite as accurate, explaining about 60% of the variance in seat swing. The House model is highly accurate, explaining over 80% of the variance in seat swing. Table 2 displays the results of regression analyses of House and Senate seat change in midterm elections between 19. The larger the presidential party’s deficit on the generic ballot and the more seats it is defending, the more seats it tends to lose. The generic ballot provides a measure of the national political environment at the time of the midterm election while the number of seats defended by the president’s party provides a measure of exposure to risk. Two predictors largely explain the variation in seat swing in midterm elections: the generic ballot and the number of House and Senate seats defended by the president’s party. Table 1: Change in seats for president’s party in midterm elections, 1946-2018 House results have ranged from a loss of 64 seats to a gain of eight seats while Senate results have ranged from a loss of 13 seats to a gain of four seats. Beyond this general tendency, however, the magnitude of these losses has varied considerably. The average seat loss has been almost 27 seats in the House and between three and four seats in the Senate. Table 1 shows that the president’s party has lost House seats in 17 of 19 midterm elections since World War II and Senate seats in 13 of 19. The tendency of the president’s party to lose seats in Congress in midterm elections is one of the best-known regularities in American politics. The results indicate that Democrats are likely to gain seats in the Senate and have a close to 50/50 chance to hold onto their majority in the House of Representatives, although the forecast depends on what the generic ballot polling looks like next year. I use the estimates from these models to make conditional forecasts of the results of the 2022 House and Senate elections. The generic ballot model uses two predictors - the generic ballot along with the number of seats defended by the president’s party - to generate forecasts of seat swing in midterm elections. In this article, I use generic ballot polls to construct a model for forecasting seat change in midterm elections. A recent article by Seth Moskowitz in the Crystal Ball showed that polls on the generic ballot question, while not perfect, generally give a good indication of the national popular vote in recent U.S. The generic ballot - a question in which survey respondents are asked which party they prefer for Congress without providing names of individual candidates - has proven to be a useful tool for explaining the national outcomes of House and Senate elections. Sign up: Free daily newsletter Sign up! Projecting 2022 through the House generic ballot
0 Comments
Leave a Reply. |