News Release

Political scientists' models predict Democratic takeover of House of Representatives

In forecasts made as long as six months ago, scholars predict likely Democratic gain of 22-29 seats in the House, and 2-3 seats in the Senate.

Peer-Reviewed Publication

American Political Science Association

Washington, DC--Election forecasting models completed by political scientists months before recent events predict significant Democratic gains in the 2006 midterm elections, including a likely 22 seats in the U.S. House of Representatives and 2-3 seats in the U.S. Senate.

The predictions appear in two articles coauthored by Carl Klarner and Stan Buchanan (both of Indiana State University) and a third article authored by Alan Abramowitz (Emory University). The articles appear in the October 2006 issue of PS: Political Science and Politics, a journal of the American Political Science Association (APSA), and are available online at:

Klarner and Buchanan employ the same framework to forecast the House and Senate elections. Traditionally, election forecasting models rely primarily either on district- and state-level analyses or aggregate methods measuring national trends. The authors develop a new predictive model that combines both approaches and uses the strengths of each while minimizing their analytical weaknesses. As a result, "both national and district conditions, as well as how those conditions combine to influence election outcomes, may be examined" observe the authors.

For their House forecast, made in late April 2006, Klarner and Buchanan consider three factors in examining past elections from 1972 and on: district partisan composition, candidate attributes, and national partisan tides. The weighting of these factors is based on a range of historical and empirical data--including most recent House vote for Democrats in a district; results of the most recent presidential vote; incumbency; prior House experience in candidates; national vote intentions reported in surveys; presidential approval; performance of the economy; and the "midterm penalty" for the president's party. The model's House predictions include the following items of note:

  • Democrats will receive 224 seats in the House--6 more than needed for control and a gain of 22 seats overall

  • There is a 94.9% probability that Democrats will win the House

  • Even when very pro-Republican assumptions were made about who would win primaries, the model predicts Democrats have a 69.2% chance of winning the House and will likely emerge with 220 seats

For their Senate forecast, made on June 30, 2006, the authors employ a similar model and data in examining past elections from 1972 and on: a state's partisan composition, candidate attributes, and national partisan tides. They weight these factors based on an analysis of Senate vote results for Democrats in the two most recent elections; average results of the two most recent presidential votes in the state; incumbency; voting records; office-holding experience; national vote intentions; presidential approval; performance of the economy; and the "midterm penalty" for the president's party. The model's Senate predictions include the following items of note:

  • Democrats will have 48 Senate seats after the election--a net gain of 3 seats, but 3 short of a majority even when Bernie Sanders and Jim Jeffords (I-VT) are counted as Democrats

  • There is a 4.7% chance the Democrats will win a majority in the Senate

  • The most competitive Senate races in 2006 are those in Pennsylvania (51% chance of a Democratic win), Minnesota (57%), Ohio (39%), Nevada (36%), and Rhode Island (61%)

In his analysis, completed in June and finalized in September, Alan Abramowitz employs a model assessing national political conditions and candidate behavior and which reinforces the Klarner and Buchanan forecasts. The Abramowitz model draws upon U.S. House election data from 1946-2004 to predict the 2006 outcome; a simpler Senate version based only on national conditions covers elections from the same period to predict the 2006 outcome. The House model considers six factors: the percentage of GOP seats in the previous Congress; the midterm penalty for the president's party; net presidential approval; the difference between the GOP and Democratic percentage of the generic vote in early September; the difference between GOP and Democratic open seats; and the difference in office-holding experience between GOP and Democratic challengers. The model's predictions include the following items of note:

  • A Democratic gain of 29 seats in the House of Representatives

  • A Democratic gain of 2.5 seats in the Senate, limited by a relatively small GOP exposure of 15 seats in this election cycle

  • An advantage of 10 points in the generic vote produces a swing of about 2 Senate seats with all else equal

By integrating long-term data analysis with current local and national political factors, these forecasting models reflect ongoing efforts by political scientists to analyze election dynamics in the US. Notably, well before recent events such as the Foley scandal, Woodward book, and Iraq NIE have made predictions of a Democratic House takeover commonplace, these scholars have used existing academic literature and their own research to assess the likely outcome of the 2006 midterm elections.


The American Political Science Association (est. 1903) is the leading professional organization for the study of politics and has over 14,000 members in 80 countries. For more news and information about political science research visit the APSA media website,

Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.