Â鶹´«Ã½AV

Paper presented at the annual meeting of the American Association for Public Opinion Research, May 12-15, 2005, Miami, Fla.

Every four years, a bright spotlight shines on the polling industry as the media and other political observers clamor for the latest information on who is winning the election and why. Pollsters are faced with the challenge of developing methods to accurately measure the preferences of a sometimes fickle electorate. The 2004 election proved to be a fairly good one for pollsters, as most generally predicted a close Bush victory. Of course, there is always a desire as well as a need to refine and improve pre-election polling methods. To that end, Â鶹´«Ã½AV has at times re-interviewed respondents who participated in its final pre-election polls, to see if they fulfilled their pre-election voting intentions, and to get a better sense of why they voted the way they did.

This paper reviews the major findings of Â鶹´«Ã½AV's 2004 post-election panel survey, such as why voters chose their preferred candidate and when they did so. The paper also discusses practical applications for election polling, such as the validity of the likely voter model and the growing trend toward early voting. While the two waves of interviewing were conducted in a relatively short timespan, they do give glimpses into the relative stability of certain basic political attitudes. Where possible, comparisons are made to Â鶹´«Ã½AV's 1996 election panel survey.

Voters were more likely to decide early and less likely to shift their voting perceptions in 2004 than in 1996. And reported voting participation, as would be expected, was much higher in 2004. While voting preferences were relatively stable in 2004, there was change evident in partisanship and opinions of Democratic presidential nominee John Kerry between the two waves of interviewing. Additionally, the 2004 study showed that Bush and Kerry voters had very different reasons for supporting their chosen candidate.

Methodology

The CNN/USA Today/Â鶹´«Ã½AV final pre-election poll -- a random sample of 2,014 U.S. adults -- was conducted Oct. 29-31, 2004. Each respondent was asked whether they would agree to be re-contacted by Â鶹´«Ã½AV and 1,567 agreed to do so. The group of respondents who agreed to be re-contacted were balanced politically -- 37% were Democrats, 35% Republicans, and 27% independents (using weighted percentages). Forty-seven percent were Bush supporters, 48% were Kerry supporters, and 5% were undecided or supporting minor party candidates (also weighted). Both proportions were similar to the full sample composition of the Oct. 29-31 poll.

The 1,567 adults who agreed to be re-interviewed were then re-contacted by Â鶹´«Ã½AV between Nov. 3 and Dec. 12, and interviews were successfully completed with 1,148. This sample consisted of 38% Republicans, 35% Democrats, and 27% independents (weighted), based on their pre-election interview responses. This suggests that the final sample was slightly more Republican, slightly less Democratic, and proportionately independent compared with the starting sample.

I. Voting Participation

The 2004 campaign was marked by high voter enthusiasm and predictions of higher-than-normal turnout were realized when 60.7% of the voting age public voted. One of the objectives of the panel survey is to see how many respondents actually voted, and how that compares with Â鶹´«Ã½AV's likely voter estimates.

Generally, speaking, survey respondents greatly overstate their voting intentions. In Â鶹´«Ã½AV's 2004 polling, at least 9 in 10 respondents consistently indicated they planned to vote on Nov. 2, about 50% higher than the actual turnout. That is the main reason likely voter models or screens are needed in pre-election polling. Â鶹´«Ã½AV's likely voter model asks respondents a series of questions on current voting intention and past voting behavior. The responses are scaled on a 0 to 7 scale, and Â鶹´«Ã½AV takes the highest scorers equivalent to the projected turnout (usually 50%, it was estimated at 60% in 2004) to comprise its likely voter sample.

Eighty-eight percent of the full post-election sample reported voting, while just 12% said they were unable to vote. It is difficult to judge the accuracy of these data compared with the real world because of the problem of over-reporting, but also because it is possible that survey respondents may be activated to vote after being polled about their voting intentions prior to the election. Â鶹´«Ã½AV had previously attempted, with a limited degree of success, to go back to voters' precincts to verify if they voted after the election. But in the absence of confirming evidence of voter participation, self-report data is the only alternative.

How do the voting reports compare with Â鶹´«Ã½AV's pre-election predictions? Seventy-eight percent of the final pre-election poll panel sample (as well as the pool of panel respondents) were considered likely voters by Â鶹´«Ã½AV based on its likely voter formula*. Ninety-seven percent of those identified as likely voters in the pre-election poll said that they voted.

If panel respondents' vote reports are taken at face value, it suggests that the likely voter model does a reasonably good job of discerning voters from non-voters. Ninety-nine percent of those with the highest score on the likely voter scale reported voting. Roughly 9 in 10 at the next highest scores did so, but less than half at the bottom portion of the distribution did.

Post-Election Reported Voting Behavior Compared With Pre-Election Likely Voter Status,
2004 Election

Â鶹´«Ã½AV's likely voter model is probably one of the tighter likely voter screens in the industry, and is probably more likely to exclude actual voters than others, but also probably less likely to include non-voters. This sort of trade-off is one all pollsters need to make -- from being too inclusive to being too restrictive in likely voter screening. Â鶹´«Ã½AV's likely voter model has proven to be very accurate in projecting the final national popular vote total in each election.

Voting participation was much higher in 2004 than in 1996, and Â鶹´«Ã½AV's polling picked that up. In Â鶹´«Ã½AV's 1996 post-election panel survey, 78% reported voting (compared with 88% in 2004) in a year when 49% of Americans turned out. Additionally, many more respondents in the final panel sample qualified as likely voters in 2004 (83%) than in 1996 (67%). In 1996, 98% of those scoring highest on the likely voter scale reported voting, as did 94% at the second highest scale. One noticeable difference is that those with lowest scores on the likely voter scale were more likely to report voting in 2004 (44%) than in 1996 (28%), suggesting the possibility that more actual voters may have been excluded from the likely voter sample this past year.

Post-Election Reported Voting Behavior Compared With Pre-Election Likely Voter Status,
1996 Election

II. How and When Did People Vote?

One possible reason for the increase in voter turnout may have been efforts by the states to make it easier for people to vote. In the past few years, a growing number of states have changed their election laws to give voters the opportunity to vote in person before Election Day, and to ease the restrictions on voting by absentee ballot. In the 2004 election, about 30 states had no restrictions on voting early or voting via absentee ballots, including most states west of the Missouri River. There are an additional 10 states where voters are allowed to cast their ballots earlier than Election Day with only minimal limitations. Voters in several of these states were even able to begin casting ballots prior to the first presidential debate last year. Another state, Oregon, does not even operate polling centers to handle votes; Oregon voters are only able to vote through the mail.

Â鶹´«Ã½AV's 2004 post-election survey shows that 68% of panel respondents said they voted on Election Day itself. An additional 20% voted prior to Election Day, either in-person (8%) or by absentee ballot or mail (12%). Twelve percent did not vote.

Early voting is not distributed equally across demographic groups. For example, half of those living the western part of the United States said they voted either by absentee ballot or by mail (40%), or voted in person before Election Day through early voting opportunities in their state (10%). This compares with 4% of Easterners, 11% of Midwesterners, and 27% of Southerners who voted prior to Election Day by any of those means.

Additionally, older Americans were significantly more likely to vote early than were other Americans. Twenty-three percent of 18- to 29-year-old panel respondents who voted did so before Election Day, compared with 17% of 30- to 49-year-olds, 27% of 50- to 64-year-olds, and 31% of people aged 65 and older. Â鶹´«Ã½AV found similar age and regional differences in its standard pre-election polling in 2004.

Early voting certainly has implications for pre-election polls. Undoubtedly, the incidence of early voting and absentee voting will continue to rise as more states change their laws and more voters take advantage of the opportunities afforded them. This presents a significant challenge to the exit polls, but also to those using telephone pre-election polls. For instance:

• How should those who say they have already voted but do not qualify as likely voters be classified in the likely voter sample? Last year, CNN, USA Today, and Â鶹´«Ã½AV automatically counted early voters as likely voters.
  
  
• Secondly, how should vote preference questions be worded when -- especially in polls conducted close to Election Day -- many people would have already voted? CNN, USA Today, and Â鶹´«Ã½AV asked those who indicated (in a prior question) that they already voted whom they voted for, and those who had not the standard question of whom they would vote for if the election were held today**.
  
  
• Thirdly, developments in alternative voting opportunities -- especially the Oregon all-mail election -- may force survey organizations to modify their likely voter questions to accommodate these changes, especially when polling in those states.

III. Vote Choice

The CNN/USA Today/Â鶹´«Ã½AV final pre-election estimates were quite accurate when compared with the final outcome, with Bush at 49% support (he got 51%) and Kerry at 47% (he got 48%) among likely voters. The initial panel sample tilted slightly more in Bush's direction based on their responses in the pre-election poll -- 51% were Bush supporters, 45% Kerry supporters, 2% preferred Ralph Nader or other minor-party candidates, and 2% were undecided. The final post-election panel numbers also were slightly more favorable to Bush -- 53% of those who say they voted supported Bush, 45% Kerry, and 2% refused.

Vote Preferences, Pre- and Post-Election Polls, 2004 Â鶹´«Ã½AV Election Panel

One of the lessons of the inaccurate 1948 pre-election polls was that polling should continue as close to Election Day as possible to pick up any late shifts in voter sentiment. The above data suggest there was relatively little change in respondents' voting preferences at the aggregate level. However, since the final pre-election poll was conducted between Friday evening (Oct. 29) and Sunday afternoon (Oct. 31), it is possible some voters changed their minds between the time they were interviewed and Election Day.

In reality, the post-election panel showed very little change in voters' stated intentions before the election and their post-election reports of vote preference. Ninety-four percent of voters reported voting for the candidate they preferred in the pre-election poll. That number increases to 98% when looking just at voters who had expressed a candidate preference prior to the election (as opposed to being undecided, for example)***.

Stability in Vote Choice, Pre- and Post-Election Surveys, 2004

Bush (98%) and Kerry (97%) intenders were equally likely to follow through on their intentions. More of the relatively small number of undecideds moved in Bush's direction than Kerry's, which runs counter to the idea that undecideds move disproportionately to the challenger. In fact, no national pre-election poll had Bush at 50%, so in theory Bush got at decent share of the undecided vote.

It is probably unlikely that voters would change their minds in such a short timespan, for example a panel conducted over several months should show more evidence of change. However, it should be noted that those who were surveyed later in the post-election interviewing phase were no more likely to show evidence of possible change in attitudes than those interviewed earlier.

The 1996 Â鶹´«Ã½AV Election Panel data show slightly more evidence of vote change, although this is a necessary outcome given the larger pool of undecided voters in the final pre-election poll that year.

At the aggregate level, voter preferences in the 1996 election panel survey were very similar to the final estimates of the 1996 CNN/USA Today/Â鶹´«Ã½AV pre-election poll. Clinton's vote share was 48% in the final pre-election poll and 46% in the panel, Dole's was 40% in both, and Perot's was 6% in the pre-election poll and 9% in the post-election poll.

Vote Preferences, Pre- and Post-Election Polls, 1996 Â鶹´«Ã½AV Election Panel

At the individual level, 88% maintained a consistent preference from pre- to post-election, compared with the 94% figure from 2004. Among those with a candidate preference before and after the election, 95% were consistent in their preference, and 5% were not in 1996, compared with 98% consistency in 2004****.

Stability in Vote Choice, Pre- and Post-Election Surveys, 1996

There were differences in candidate consistency among the various supporters in 1996, with Dole voters the most consistent and Perot voters the least:

• 95% of Dole voters stayed consistent, while 4% reported voting for another candidate
  
  
• 92% of Clinton voters stayed consistent, and 6% reported voting for another candidate
  
  
• 80% of Perot voters stayed consistent, while 19% reported voting for someone else

IV. Stability in Other Attitudes

Not all attitudes exhibited the same level of stability as did voter preference in the 2004 election. Some change could be expected in the election aftermath since the victor is usually showered with positive media attention, which can (and did) result in a post-election "bounce" in support -- Bush's job approval rating increased from 48% in the final pre-election poll to 55% in mid-November.

Â鶹´«Ã½AV asked four key political attitudes in both waves of the panel -- vote preference, Bush job approval, candidate favorable ratings, and party identification. Overall, very little change is evident in the aggregate -- at most opinion only shifted three percentage points -- on Bush's job approval rating and on Kerry's favorable rating.

Political Attitudes, Pre- and Post-Election, 2004 Â鶹´«Ã½AV Election Panel Respondents

As is typically the case with panel data, the aggregate numbers obscure some change occurring at the individual level. Although the vast majority of respondents gave consistent responses in both waves of the panel study, on some attitudes respondents were more likely to show change.

Political Attitudes, Pre- and Post-Election, 2004 Â鶹´«Ã½AV Election Panel Respondents

The greatest amount of movement on attitudes occurred on party identification and opinions of John Kerry. The case of the former is notable since many believe it is an essentially fixed characteristic almost like age, education, or race. That is probably true for the vast majority of Americans, but a substantial number have only weak orientations to the parties, and thus are more likely to shift their "allegiances" when asked about it in a poll.

Most of the change that occurred on party identification involved respondents shifting from one party to independent status (7% of panel respondents did this). The second most common movement involved a shift from independent status to party status (4% did so). The least common was movement from one party to the other, only 1% moved from identifying as Republican to identifying as a Democrat, or vice versa.

The idea that independents are more likely to shift is supported by the fact that there was roughly half the amount of movement -- just 7% -- in 'leaned' party identification (in which independents are asked to which party they lean). Also, as the table below shows, just 84% of independents were consistent in their reported party affiliation, compared with 90% of both Republicans and Democrats.

Consistency of Party Identification, Based on Party Identification in Pre-Election Poll

Opinions of Kerry also showed a non-trivial amount of change from one panel wave to the next, as 12% of respondents gave different views on Kerry between the two waves. That includes 5% who went from an unfavorable to a favorable view, 2% from favorable to unfavorable, 3% who had no opinion before the election but did so after, and 2% who had an opinion before the election but did not after.

Panel respondents were less likely to change their views on Bush -- in fact the three attitudes most closely tied to Bush (favorable, job approval, and vote) generally showed the least amount of change of the attitudes measured before and after the election.

• Seven percent of panel respondents' basic views of Bush changed, with 4% moving from an unfavorable to a favorable view. [1% went from favorable to unfavorable, 1% from no opinion to an opinion, and 1% from an opinion to no opinion].
  
  
• Similarly, 8% changed their view on the way Bush was handling his job as president, with 3% moving from disapproval to approval and 2% from approval to disapproval. [1% from an opinion to no opinion; 3% from no opinion to an opinion (more approval than disapproval)]
  
  
• As discussed in the previous section, only 6% of all voters (and 2% of those with a candidate preference) changed their voting preference from pre- to post-election.

A comparison to 1996 is not possible on these items since vote choice was the only question included in both the final pre-election poll and the post-election panel survey.

V. When the Final Decision Was Made

One reason for the relatively limited evidence of voter preference change is that the vast majority had long made up their mind about whom they were supporting. This was evident throughout the year in the relatively small number (by historical standards) of "swing voters," whose choices are usually though to be a critical determinant of who wins the election.

The notion of longstanding vote decisions is supported by the 2004 panel data. The vast majority of voters, 67%, said they made up their minds about whom to vote for before the conventions. Ten percent did so after the conventions, 10% during the debates, and 4% after the debates. Only 8% say they decided in the last week before the election, including 2% who indicated they made up their minds on Election Day.

The data show some interesting differences between Bush supporters and Kerry supporters, with Bush voters making up their minds earlier and Kerry voters more likely to make up their minds closer to the election.

More than 7 in 10 Bush voters made up their minds before the conventions, and another 8% saying they did so after the conventions but before the debates. Eight percent of Bush voters made up their minds during the debates, leaving just 10% who decided after the debates (3%), in the last week (4%) or on Election Day (3%).

Perhaps because Kerry was the challenger, his supporters joined his camp later than the incumbent Bush's supporters. Still, the majority (60%) of Kerry voters say they made their minds up to vote for him before the conventions. Thirteen percent decided after the conventions but before the debates. Eleven percent chose Kerry during the debates, and 6% after the debates (Kerry was generally considered to have won all three debates). Eight percent of Kerry voters made up their minds in the last week (7%) or on Election Day (1%).

In 1996, Â鶹´«Ã½AV also asked voters when they decided. The question has a more limited timespan, only asking about the week before the election or earlier (this question was repeated in 2004, but then those who indicated they had decided before the last week were asked a follow-up to get a better idea of when they made their decision during the campaign). It is clear that voters in 2004 made up their minds earlier than those in 1996. In 1996, 81% of voters made up their minds before the last week of the election, compared with 92% in 2004.

Those differences could very well be due to the presence of a significant third-party option in 1996 -- Ross Perot. More than 8 in 10 respondents who voted for Clinton or Dole made up their minds before the last week of the campaigns. Perot voters, however, were much more likely to decide to vote for Perot closer to the election. Just 57% of Perot supporters made up their minds to vote for him before the last week of the election, while the remaining 43% did so in the last week, including one in eight who did so on Election Day itself.

In both the 1996 and 2004 elections, voters had long since decided whom they would be supporting before the frantic final days of the campaign. Of course, both elections involved incumbent presidents, so the choice facing voters in those elections may have been easier to make than in elections like 2000 when no incumbent ran. It is possible that an "open-seat" election like in 2000 or 2008 would show more evidence of late decision-making.

VI. Reasons for Candidate Support in the 2004 Election

Â鶹´«Ã½AV's post-election panel survey shows that Iraq, terrorism, and President George W. Bush's job performance were key reasons why voters voted the way they did last year. But people voting for Bush and John Kerry differed in the priority they assigned to these issues. Bush voters most often mentioned moral values, terrorism, and satisfaction with Bush's job performance. Kerry voters cited Iraq, the economy, and dissatisfaction with Bush.

Additionally, more than 8 in 10 voters rated the candidates' policies on terrorism and on Iraq as at least very important considerations in their votes. More than 4 in 10 Kerry voters said the fact that he "was not Bush" was an "extremely important" reason for their choices.

The post-election panel survey first asked respondents in an open-ended manner to say what were the most important one or two reasons why they voted for the candidate they chose. Twenty-two percent of respondents volunteered something about the situation in Iraq and 12% mentioned terrorism. Broadly speaking, national security seemed to be the top concern for voters.

A substantial proportion mentioned something about Bush's job performance (21%), either in positive or negative terms, which would be expected in an election involving an incumbent candidate.

Fifteen percent -- mostly Bush voters -- cited morals, values, or religious beliefs, and an additional 9% mentioned honesty, integrity, or ethics.

Economic issues, mentioned by 16%, were another common theme, mostly expressed by Kerry voters. Thirteen percent of voters talked about their dislike of the opponent, 11% about their own candidate's platforms or agendas, and 10% said their candidate's leadership qualities were central to their vote choices.

Most Frequent Issues Mentioned as Reason for Presidential Vote

As suggested above, there was considerable variation in the most common responses given by Bush and Kerry voters. Bush voters were most likely to cite moral values, terrorism, satisfaction with Bush's job performance, leadership qualities, Iraq, dislike of Kerry, and honesty and ethics as the reasons for their votes.

Kerry voters were most likely to cite Iraq, economic issues, dissatisfaction with Bush's job performance, overall dislike of Bush, and favoring Kerry's agendas as the reasons for their choices.

Most Frequent Issues Mentioned as Reason for Presidential Vote
by Candidate Choice

While popular interpretations of the election outcome centered on the moral values issue, these data suggest it was not the overriding concern of voters, placing behind national security concerns such as Iraq and terrorism and evaluations of Bush's job performance. But moral values were definitely a key consideration for Bush voters, especially.

Rating Specific Issues and Dimensions

Â鶹´«Ã½AV also asked respondents to rate the importance of 13 specific reasons for their vote choice. More than 8 in 10 voters said the candidates' policies for dealing with terrorism (84%) and Iraq (83%) and the candidates' leadership characteristics (82%) were extremely or very important to their votes. The candidates' intelligence (78%), economic policies (77%), domestic policies (75%), and policies on moral values issues (73%) were rated as extremely or very important by more than 7 in 10 voters. Relatively few said that the candidates' party affiliations, performances in the debates, convention speeches, or vice presidential running mates were at least very important.

Rated Importance of Reasons for Vote

Again, there were different patterns of results depending on which candidate the voters preferred.

Bush voters apparently were attracted most by the president's terrorism policies and his leadership characteristics -- 9 in 10 say these were at least very important to their votes, including more than 4 in 10 who said they both were extremely important. Bush's policies on moral values issues and Iraq also figured prominently in the ratings among his supporters.

Rated Importance of Reasons for Vote
Bush Voters

Kerry voters were most likely to cite his policies on domestic issues, his intelligence, and his policies on Iraq and the economy as extremely or very important reasons for voting for Kerry. Terrorism policies ranked lower down the list, as did policies on moral values issues.

Seventy-five percent say the fact that Kerry "was not Bush" was at least very important to their votes, including 41% who said it was an extremely important reason for supporting Kerry. No other item was rated "extremely important" by as many Kerry voters.

Rated Importance of Reasons for Vote
Kerry Voters

While not ranked near the top of issues, Kerry's convention speech and especially his performance in the debates were more likely to be rated as important by Kerry voters than Bush's convention speech and debate performance by Bush voters. Sixty-five percent of Kerry voters rated their chosen candidate's debate performance as at least a very important reason for their votes, while only 25% of Bush voters did so. Forty-four percent of Kerry voters rated his convention speech as extremely or very important to their votes, while only 33% of Bush voters did the same for their candidate's speech. This is not surprising given that the incumbent can be evaluated more on the basis of his four years in office while the lesser-known challenger's most prominent campaign moments are the convention and the debates.

VII. Other Panel Survey Findings

Mobilizing the Faithful

The conventional wisdom is that "moral values" were key to George W. Bush's victory in the 2004 election. Indeed, Bush's strategists made it a point to mobilize conservative Christian voters, among whom a sizable proportion did not vote in 2000.

Just as labor unions are instrumental in rallying Democratic-leaning voters, churches are also believed to be instrumental in mobilizing Republican-leaning (and in the black community, Democratic-leaning) voters. Mobilization efforts can be subtle (a friend or family member encouraging someone to vote) or more direct (an organized worker driving someone to the polls), and public (candidate campaign rallies) or private (telephone calls). Â鶹´«Ã½AV's post-election panel survey attempted to find out to what extent clergy made public and direct appeals to their congregations in an attempt to activate voters in this year's election.

Sixty-eight percent of panel respondents reported attending church, synagogue, or other religious services in the prior six months, while 31% said they had not. Those who attended were asked whether the priest, minister, or rabbi spoke favorably or unfavorably about a specific presidential candidate at any of these services. Eleven percent of religious service attendees said their clergy members had spoken about a candidate, while the vast majority, 87%, said they had not. Thus, it appears as if presidential politics was not a common theme from the pulpit.

Among the 11% who indicated a clergy member had spoken about presidential candidates, more than half said that he or she stopped short of urging the congregation members to support a particular candidate. Of the small minority of churchgoers who say that the clergy member did urge members to vote for a candidate, they were much more likely to say they were urged to vote for Bush rather than Kerry.

Did the priest, minister or rabbi urge the members of the congregation to vote for a specific presidential candidate, or not? [Which one?]

So while churchgoers were more likely to get a Bush appeal than a Kerry appeal during a religious service, neither was apparently common in U.S. houses of worship.

Voting Problems

One of the background stories leading up to the election was the potential for fraud or other voting irregularities. Indeed, some believe that came to pass and cite the exit polls showing Kerry leading as evidence. Â鶹´«Ã½AV asked those who indicated that they had voted in person if they were aware of any problems with voting at the location in which they cast their ballot. Only 7% did, while 93% did not.

In fact, the most common problem that voters personally encountered were "long lines to vote" -- 15% reported that. Only 3% said they had encountered questions about their voting registration, and 2% said they had experienced "confusing ballots" -- the legacy of the 2000 Florida presidential vote.

VIII. Summary and Conclusions

The 2004 Â鶹´«Ã½AV election panel provided additional insight into why Bush won the election, when voters decided, and what their voting experiences were. Many Kerry voters chose him because he was the alternative to Bush, but there are clear differences in the motivation of Bush and Kerry voters, at a gross level the perceived important of international versus domestic policy issues was a clear dividing line.

The results show that most voters voted the way they intended to, in fact the way they had long intended to. That makes sense because the vast majority of voters will support the candidate of their preferred party, and once the nominees are known, the decision is essentially made. Also, in elections involving an incumbent, the choice of four more years or not may be easier for the voter to make than the choice of candidate A or candidate B.

In the relatively short time period between waves, attitudes were basically stable -- especially views about Bush. Party identification and views of challenger John Kerry showed the most change, with most of the change in party affiliation accounted for by people moving between independent and partisan status.

The panel drove home the point that more and more voters are casting ballots prior to Election Day, which may force pollsters to alter the way they ask about voting intentions and preference.

The 2004 election presented some challenges to pollsters, such as an expected surge in voting among new voters, the rising proportion of cell phone usage, and new and varied ways of voting. In the final analysis, the polls rose to met those challenges, but the 2008 election will likely present a new set of challenges to address.

IX. Topline Results

Results are based on telephone interviews with 1,148 adults, aged 18 and older, conducted Nov. 3-Dec. 12, 2004. For results based on the total sample of adults, one can say with 95% confidence that the margin of sampling error is ±3 percentage points.

Survey respondents were first interviewed as part of a random national adult samples by Â鶹´«Ã½AV in its final national pre-election poll, conducted Oct. 29-31, 2004, at which time they indicated said they were willing to be re-interviewed by Â鶹´«Ã½AV. Respondents' pre- and post-election answers are shown for those questions that were asked on both surveys. 1148 out of 1567 respondents were successfully re-contacted and completed interviews.

The original sample of 1,567 consisted of 729 who said they planned to vote for John Kerry (48% after weighting) and 761 who said they planned to vote for George W. Bush (47% after weighting). The remaining 77 (5% after weighting) had no opinion or were supporting other candidates.

The original sample of 1,567 consisted of 585 who identified as Republicans (35% after weighting), 554 who identified as Democrats (37% after weighting), and 428 who identified as independents (27% after weighting).

First,

1. Which of the following best describes your reaction to the outcome of the presidential election - [ROTATED: very pleased, somewhat pleased, neither pleased nor upset, somewhat upset, (or) very upset]?

2. Next, as you may know, many people in the United States are unable to vote in elections because they have moved their place of residence, or they are physically unable to go to a polling place, or they are away on business on Election Day. Were you, personally, able to vote in this year's election, or did something come up that prevented you from voting?

3. Which of the following applies to you -- you voted in-person on Election Day itself, you voted in-person before Election Day, taking advantage of early voting opportunities in your state, or you voted by absentee ballot or mail?

COMBINED RESPONSES (Q.2-3)

4. Were you aware of any problems related to people's votes in the location where you voted, or not?

BASED ON 905 RESPONDENTS WHO VOTED IN-PERSON; ±4 PCT. PTS.

5. How about you? Did you encounter any of the following issues when you voted, or not? How about -- [RANDOM ORDER]?

BASED ON 905 RESPONDENTS WHO VOTED IN-PERSON; ±4 PCT. PTS.

6. NOT ASKED

7. NOT ASKED

8. Next, can you please tell me which candidate for president you voted for? [OPEN-ENDED]

[Did you vote for John Kerry, the Democrat, George W. Bush, the Republican, or someone else?]

[Your response to this question is extremely important to the accuracy of the survey. We want to assure you that your response will only be used to estimate the candidates' share of the vote. Your responses will be kept completely confidential.]

BASED ON 1,054 RESPONDENTS WHO VOTED; ±3 PCT. PTS.

9. What would you say were the one or two most important reasons why you voted for [INSERT RESPONSE FROM Q.8]? [OPEN-ENDED]

BASED ON 1,054 RESPONDENTS WHO VOTED; ±3 PCT. PTS.

10. How important were each of the following as reasons why you voted for [INSERT RESPONSE FROM Q. 8] -- extremely important, very important, somewhat important or not important? How about [ITEMS A-M ROTATED, AS APPROPRIATE]?

BASED ON 1,054 RESPONDENTS WHO VOTED; ±3 PCT. PTS.

A. His leadership characteristics

B. His intelligence

C. His policies for dealing with terrorism

D. His policies for dealing with the situation in Iraq

E. His policies for the economy and taxes

F. His policies on moral values issues

G. His policies on domestic issues such as healthcare and education

H. His choice of a vice presidential running mate

I. His convention speech

J. His performance in the debates

K. Because he is a Democrat

L. Because he is a Republican

M. Because he was not [Bush/Kerry]

Q.10 CONTINUED

SUMMARY TABLE: IMPORTANT REASONS TO VOTE (TOTAL VOTERS)

SUMMARY TABLE: IMPORTANT REASONS TO VOTE

BUSH VOTERS

SUMMARY TABLE: IMPORTANT REASONS TO VOTE

KERRY VOTERS

11. When did you make up your mind to vote for [INSERT RESPONSE FROM Q.8] -- in the voting booth, on Election Day, in the last few days before the election, in the last week, or earlier than that?

12. Thinking about this a little further, did you decide that you were going to vote for [INSERT RESPONSE FROM Q.8] -- after the three presidential debates were held, during the time the three presidential debates were held, after the two parties' conventions, or before the two parties' conventions?

BASED ON 1,054 RESPONDENTS WHO VOTED; ±3 PCT. PTS.

COMBINED RESPONSES (Q.11-12)

Turning to a different topic,

13. Do you approve or disapprove of the way George W. Bush is handling his job as president?

14. Next, we'd like to get your overall opinion of some people in the news. As I read each name, please say if you have a favorable or unfavorable opinion of these people -- or if you have never heard of them. How about -- [ITEMS A-B READ IN ORDER]?

A. George W. Bush

B. John Kerry

15. Have you attended church, synagogue or other religious services within the past six months, or not?

16. At any of the services you attended, did the priest, minister or rabbi talk favorably or unfavorably about a specific presidential candidate, or not?

17. Did the priest, minister or rabbi urge the members of the congregation to vote for a specific presidential candidate, or not? [Which one?]

BASED ON 776 RESPONDENTS WHO ATTENDED CHURCH IN THE PAST SIX MONTHS; ±4 PCT. PTS.

COMBINED RESPONSES (Q.16-17)

Finally,

D9. In politics, as of today, do you consider yourself a Republican, a Democrat, or an independent?

*Â鶹´«Ã½AV weights down the sample of likely voters to match the projected turnout, so the 78% were weighted down to 60%, with those scoring highest on the likely voter scale receiving a full weight, and those scoring next highest getting a fraction of their weight.

**Asking people who already voted how they voted seemed to increase the refusal rate on the vote preference question slightly compared to what it is on the standard "if the election were held today" question. Again, this is a trade-off that needs to be considered versus asking respondents appropriate questions to their situation.

***While inconsistency in responses is referred to as "change" in preference, the authors acknowledge that other explanations (coding error, measurement error, misremembering by respondents) could produce inconsistency. So the estimates here are likely upper limits of actual change in voter preferences.

****It should be noted that the 1996 election panel was conducted over a shorter timespan, just 11 days (Nov. 7-17, 1996).