The 2006 midterm elections were largely a referendum on the Bush administration and the war in Iraq.

Throughout the nation, in state after state, Fox News Election Day telephone surveys show voters basing their vote for House and Senate more on national issues than on state and local issues.

Those who support the president and the war in Iraq largely voted for the Republicans in their district. Those who oppose the war or who have an unfavorable view of the president typically voted for the Democrat.

Surveys also showed that the president’s job approval rating was well below 50 percent — with the Iraq war highly unpopular, as well. Furthermore, many voters felt a disconnect between the Iraq mission and the overall War on Terror. The result was a clear gain in Democratic seats in both the House and Senate.

In Tennessee, the one close state in which a majority of voters did not oppose the president or the war in Iraq, the Republican Senate candidate proved successful.

Throughout the nation, many voters considered the economy and the war in Iraq in deciding their vote. In many instances, voters who placed high importance on the economy or on the war in Iraq were likely to vote Democratic.

Another national trend was the importance of political affiliation. In many key districts, 40 percent or more of all voters say the political party of the candidates was more important this year than in the past. In most cases, these voters supported the Democrats in their districts.

Finally, several Republican candidates who were perceived as having agreed too much with the president were punished by voters by wide margins.

Note: All polls are in PDF format.

• Senate
• Congressional District 2
• Congressional District 4
• Congressional District 5

• Senate
• Congressional District 6

• Senate

• Senate
• Congressional District 18

• Senate
• Congressional District 6
• Congressional District 7
• Congressional District 8

• Senate

• Senate

FOX News/Opinion Dynamics Election Day Poll Methodology

The FOX News/Opinion Dynamics Election Day poll was carried out on November 6 and 7, 2006 using standard statistical methods to select a representative sample of likely voters and voters who had already cast a ballot. Interviews were conducted by telephone Monday night prior to the election and throughout Election Day.

The Election Day poll was conducted in seven states with competitive senate and house races. A total of 900 interviews were completed statewide in each of the seven states. Additional interviews were completed in eleven Congressional Districts within the seven states, to ensure 500 completed interviews in each district.


Samples for Election Day polls in each state were random digit samples of telephone numbers selected using the “probability proportionate to size” method, which means numbers from across the state were selected in proportion to the number of voters in each area of the state.
A computer selected the first eight digits of an actual working number and then appended a two-digit random number to produce a random-digit dial (RDD) sample. An RDD sample allows for contacting not only listed and unlisted numbers, but also households with new numbers.
Each respondent was screened to establish him/her as a registered voter. Respondents were asked if they have already cast a ballot. If not, interviewers screened for “likely voters” through a additional questions about voting intention and knowledge of voting location.


The RDD selected phone numbers were sent to the interviewers through computer-assisted telephone interviewing (CATI) software. Both the software and human supervisors monitored each step of the interviewing process. While calls were automatically dialed, the system did not use predictive dialing so prospective respondents always found a live interviewer when they answered their phone.


For a sample of about 900 interviews, used in the statewide polls, the error due to sampling is plus or minus three percentage points. For example, when the survey says “47% of voters...” then chances are very strong that no less than 44% and no more than 50% of all voters would have responded the same way. For a sample of about 500 interviews, used in the district polls, the error due to sampling is plus or minus 4.4 percentage points.
In addition to sampling error, question wording and question order can influence poll results.


Generally, Fox News poll results are not weighted. The probability sample, if conducted properly, should accurately reflect attitudes within each area. However, particularly because the survey was conducted over a short period (limiting the opportunity for callbacks), some demographic deviation is possible. For this reason weighting was used to bring the sample into conformity with other samples within each state and district.