Study Forecasts Mid-Term Election Results

Analysis predicts 32-seat gain for Democrats

A new study, "Forecasting House Seats from Generic Congressional Polls," authored by Assistant Professor of Government Joseph Bafumi, predicts that based on current ballot polling data, the Democratic party can expect to gain 32 seats in the U.S. House of Representatives in the November 7 midterm Congressional elections. Such a gain would guarantee the Democrats a House majority and exceed many current forecasts for Democratic gains in the House.

The study, coauthored by Robert Erikson of Columbia University and Christopher Wlezien of Temple University, translates the results of generic Congressional polls conducted by CNN, USA Today/Gallup, ABC/Washington Post, Fox/Opinion Dynamics and Newsweek into the partisan midterm swing. The polling data they used asked respondents which party they would choose in the next election. According to that analysis, Democratic Congressional candidates are predicted to receive 55 percent of the votes cast, plus or minus a few percentage points.

The study then goes on to analyze whether that majority of votes is likely to translate into a majority of House seats. Combining the partisan swing, estimated from the generic Congressional polls, with historical data at the Congressional district level, the authors predict which party will win each district race. They found that if Democrats receive even 53 percent of votes cast, the probability that they will win enough seats to take over the House increases to more than 90 percent.

Although Bafumi and his colleagues are far from the first prognosticators to analyze poll data as a predictor of a party's political fortunes, their analysis differs from that of most pollsters and pundits in its rigorous and nonpartisan evaluation of the true statistical value of polling. According to Bafumi, "This sophisticated statistical model suggests that the Democrats should out-perform the expectations of many in 2006."

By GENEVIEVE HAAS

Questions or comments about this article? We welcome your feedback.