Results tagged “polls”

Election Debriefing

 Posted by  Larry Bartels

I’ve already seen lots of excellent political science post-mortems on the election. John Sides has had a particularly good series of posts at Monkey Cage attempting to debunk various myths about what happened and why. Here are a few points along the same lines, from today’s CSDP roundtable on the election.

First, as always, much of the focus of commentators in the immediate aftermath of the election has been on the quality of the two candidates’ campaigns. The New York Times made Obama’s campaign the lead, headlining their main morning-after analysis (by Adam Nagourney, Jim Rutenberg and Jeff Zeleny) “Near-Flawless Run Is Credited in Victory”. On the other hand, my Princeton colleague Julian Zelizer put the focus on McCain in a Newsweek piece headlined “Worst Campaign Ever?”. (Julian may be this year’s winner of my quadrennial award for first use of the phrase “worst campaign ever”; I believe the last time there was no award was in 1976.)
Of course, it is very hard to know—or even how to know how we would know—which candidate ran a better campaign. Many decisions on both sides were significantly shaped by the fundamental contours of the election year. It seems well worth noting, however, that the result turned out very much as one would have expected based on political scientists’ early forecasts—and this despite Obama’s big fundraising advantage and the “fortuitous” timing of the financial meltdown, which focused voters’ attention even more single-mindedly on the economy than would otherwise have been the case.
There is certainly plenty of evidence in the exit polls suggesting that the election was primarily a referendum on the state of the country under President Bush. In the country as a whole and in several key swing states, half the voters thought national economic conditions were “poor,” and they went for Obama by 2-to-1 margins. A similar number strongly disapproved of Bush’s performance, and they went for Obama by even wider margins. With 90 to 95% of the votes he needed to be elected coming from people who strongly disapproved of the incumbent government, Obama didn’t have to win much support from the other half of the electorate (many of whom disapproved of Bush’s performance, but less strongly) in order to win.
Retrospective Voting in 2008 Exit Polls

National Economic Conditions Poor
(Obama Margin)
Strongly Disapprove of Bush’s Performance
(Obama Margin)
49% (66-31)
51% (82-16)
53% (66-33)
52% (82-16)
43% (66-32)
43% (86-13)
North Carolina
49% (64-35)
47% (84-15)
53% (68-30)
52% (83-16)
44% (72-27)
53% (86-13)
43% (67-33)
50% (84-15)

There has also been lots of talk about whether the election outcome marks a major shift in the standing of the parties and the nature of their supporting coalitions. The detailed presentation of election results in today’s New York Times was headlined “In a Decisive Victory, Obama Reshapes the Electoral Map.” A piece by John Judis in The New Republic is even bolder, announcing that “Obama’s Victory Marks a Radical Realignment in American Politics”. This, too, is a common theme in commentary whenever either party wins an election by more than a few percentage points—and it, too, is often much-overblown.
In a brief pre-election post I suggested looking at the continuity of state-level election returns from 2004 to 2008 as a way to gauge whether this election produced any major political shifts. The simple answer turns out to be that, by this measure, nothing very unusual happened. The following table reports the slope, intercept, and standard error of the bivariate regression of 2008 popular vote margins on 2004 vote margins in the 50 states and DC. (Omitting DC, or Hawaii, which is a notable outlier, makes no real difference to the statistical results.)
Shifts in State-Level Voting Patterns

1980-2004 (Average)
Continuity of Previous Partisan Pattern
(regression slope)
National Swing
(regression intercept)
Local Forces
(standard deviation)

In every respect, the results from 2008 look much like those from other recent presidential elections. The slope of the regression line is 1.00, which suggests that 100% of the existing partisan pattern of support from 2004 persisted in 2008. The intercept reflects a national swing of 9 points (remember, these are vote margins rather than vote shares)—only slightly larger than the average for elections since 1980. And the standard deviation of the regression (a measure of how much specific states departed in either direction from the national vote swing) was less than 6 points, again well within the normal range for recent elections.
For purposes of comparison, the table also shows the corresponding regression results from 1932, a genuine realigning election. In that year, one-third of the previous partisan pattern disappeared, the national swing in the popular vote margin was 29 points, and the standard deviation was 16 points—all vastly larger numbers than in 2008 or any other recent election. If 1932 was a realigning election, 2008 was one-third of a realignment—but then, most elections are.

Same Old, Same Old?

 Posted by  Larry Bartels

One of the things to watch when the election returns (finally!) come in tomorrow night is how the voting behavior of specific states and demographic groups differs from what it has been in recent elections.  In a 1992 book chapter on "The Impact of Electioneering in the United States," I used a scatterplot of state presidential votes from 1984 and 1988 to underline the importance of enduring partisan loyalties in voting behavior.  In a 1998 piece in Electoral Studies  I showed that the stability of voting patterns from one election to the next has been greater since the 1980s than in the previous several decades.

 A recent post by Andrew Gelman shows the corresponding scatterplot for 2000-2004, which shows very modest deviations despite the intervening 2000 election controvery, massive redistributive tax cuts, 9/11 terrorist attacks, and war in Iraq.  Gelman also compares Obama's standing in recent state polls with Kerry's performance in 2004.  Not surprisingly, there is a considerable upward shift, but also a good deal more scatter.  How much of that is due to noise in the polls, and how much to shifts in the underlying partisan landscape?  The answer to that question may provide an early clue to whether 2008 will mark a significant departure from the evenly balanced, highly polarized electoral status quo of the past 20 years.

 Posted by  Lynn Vavreck

It’s the question on everyone’s minds in the last week before the election. Are these poll numbers real? Another way to ask this is to ask: what would have to be true for us to believe that Obama is enjoying this level of support from the American electorate? And, what does all of this have to do with race in America?

Simon Jackman and I are the Principal Investigators of The Cooperative Campaign Analysis Project (CCAP) and we are in a unique position of having data that helps to answer these questions. CCAP is a six-wave panel study of 20,000 registered voters. The survey started in December of 2007 and will end after the election in November.   It is a cooperative project of 30 universities and nearly 75 scholars. 
During the last weeks of March and September we asked people to answer a battery of questions to measure racial antipathy toward African Americans. We also asked people whom they thought they would vote for in the election. 
To measure racial antipathy, we use the traditional racial resentment battery (Kinder and Sanders) that asks people questions about affirmative action, work ethic, and government assistance for African Americans relative to other racial groups. The four racial resentment items are scaled using IDEAL. In the primaries, the distribution of the racial antipathy scale for voters casing ballots in the Democratic primary for Clinton or Obama looked like this:
Distribution of Racial Resentment Scale, Clinton and Obama Primary Voters
(Measured in March 2008)
Racial Resentment.jpg 
How does this scale relate to Democratic primary voters’ choices between Clinton and Obama in the primaries? 
The results are striking. Increasing levels of antipathy toward blacks are strongly related to vote choice even within the Democratic primary. This result is robust to demographic controls such as gender, age, income, race, ideology, and party ID. 
In the primary election, this means that people with higher levels of antipathy toward blacks were more likely to vote for Hillary Clinton. These are the now infamous group of white, lower income voters living in the Mid-western regions of the country that we heard so much about during the primaries. Another interesting finding to emerge from this relationship is that white voters with very low levels of racial antipathy are voting for Obama in very strong numbers. This relationship among racially tolerant people is much stronger for Obama than it has been in the past for other African American candidates such as Jesse Jackson (Tesler 2008).
 Obama over Clinton, March.jpg
                                                                     Level of Racial Antipathy
When Obama clenched the Democratic party nomination, many (31%) of these Clinton supporters said they would not vote for Obama and would instead vote for John McCain. 
In September, over 2/3rds of those Clinton supporters who said they would defect have come home to the Democratic party. Only 19% (of the 31%) of Clinton supporters who said they would vote for McCain continue to hold this position. Another 20% of the 31% are still undecided.
Racial antipathy toward blacks remains a strong predictor of vote choice in the general election match-up between McCain and Obama. Using all white voters this time, here is the distribution of racial resentment among white, general election registered voters by their vote choice. These data come from our September wave of the survey. 
Distribution of Racial Resentment among White Voters by Vote Choice
  Obama McCain Racial Resentment.jpg
Once again, as antipathy increases the probability of voting for Obama decreases. Among people who voted in the Democratic primary, however, the relationship is mitigated by the power of party identification. Nearly all Democratic-primary voters are predicted to vote for Obama based on their levels of antipathy toward blacks as a group.   This explains the difference between the two lines below.
 Obama Over McCain.jpg
                                                          Levels of Racial Antipathy
To underscore these findings, we asked people whether race was a factor in their vote decision. Roughly 7% of registered voters voting against Obama reported that Obama’s race was a reason they would vote against him. We estimate that this number is understated by a factor of 1.6. Through a series of experiments described below, we conclude that nearly 11 percent of registered voters who are not supporting Obama are voting against him because of his race.
Conversely, 12% of registered voters supporting Obama said they were voting for Obama because of his race. Using list experiments we estimate that this number underestimates the true value by a factor of 3. We believe more than a third of registered voters supporting Obama are attracted to him because of his race.
Basic List Experiment with 3 Conditions
We randomly divide the sample into groups that receive different survey questions. The first group (OPEN LIST) is asked to tell us WHICH conditions are important as reasons to vote for (or against) Barack Obama. We provide a list of items. 
In the second group (CONTROL), we simply ask people to tell us HOW MANY of the things are important as they make up their minds. People respond with a number from zero to five. In this group, we leave one item OFF THE LIST – the item mentioning Obama’s race. Note that people do not have to tell us WHICH THINGS, only how many. 
In the third group (TREATMENT), we do the same thing as in group two, but we include Obama’s race as an item in the list. People in this group respond with a number between zero and six. 
The difference between the average value in Group Two and the average value in Group Three is the percentage of people who added Obama’s race to the list of reasons to vote for or against him. For example, if the average value in Group Two is 3.1 items and the average in Group Three is 4.1 items, we know that everyone (100 percent) included the race item in Group Three (because all the other items are the same and the groups are randomly assigned.) 
This methodology allows people to anonymously reveal when socially undesirable considerations (things they don’t want to admit to) are important to them. Results are below.

Which things matter?

Economic Plan      38
Party                      33
Iraq Policy              57
Health Care Plan 52
Speaking Ability    16
He’s black              7
How many things matter?
Economic Plan     
Iraq Policy
Health Care Plan
Speaking Ability
How many things matter? (TREATMENT)
Economic Plan     
Iraq Policy
Health Care Plan
Speaking Ability
He’s black
 Difference between CONTROL and TREATMENT = 11%

Which things matter? 

Economic Plan      74
Party                      61
Iraq Policy              76
Health Care Plan 79
Speaking Ability    64
He’s black              12
How many things matter?
Economic Plan     
Iraq Policy
Health Care Plan
Speaking Ability
How many things matter? (TREATMENT)
Economic Plan     
Iraq Policy
Health Care Plan
Speaking Ability
He’s black
 Difference between CONTROL and TREATMENT = 30%
We conclude from these data that many people may not be comfortable expressing that race is a factor in their decision to vote against Obama, but many more voters are uncomfortable saying that race is a factor in their decision to vote for Obama. This relationship works in both directions. To underscore this, among black respondents (almost all voting for Obama) 25% tell us in the open list (Group One) that race is a factor in their decision to vote for Obama. But, the list experiment shows that 75% of African Americans reveal that race is a factor in their decision to vote for Obama. The open list results underestimate this number by a factor of 3. In other words, even black respondents are reluctant to admit openly that race is driving their vote. 
We ran the same type of experiment for McCain’s age and found dramatic results. Many more people list McCain’s age as a reason to vote against him when they can do so anonymously than list Obama’s race as a reason to vote against him. 
For more information on these data and analyses contact:
(650- 305- 9424 LVAVRECK@UCLA. EDU)
(650- 387- 3019 JACKMAN@STANFORD. EDU)


 Posted by  Robert S. Erikson

With about three weeks to go before the election, Obama leads McCain by about eight points. While an upset remains possible, Obama is clearly poised as the likely presidential winner. What can we expect from the polls over the next three weeks, and how well will they predict the actual outcome? 

For poll-watchers, the temptation is to treat every new poll as a decisive piece of new evidence, as if any departure from the current trend might indicate a change that will carry forward to Election Day.    But the next outlier we see will probably be an artifact of routine sampling error rather than a harbinger of true change. True change in preferences occurs slowly, especially during the final weeks of a campaign. Observe the following graph of the Bush-Kerry vote in the polls during the final 28 days leading up to the 2004 campaign  
2004 polls in last 30 days.jpg

The Polls During the Final 30 Days of the 2004 Presidential ElectionEach dot represents the poll-of-polls or average of all polls whose coverage centered on that date.  The curved blue line represents a lowess trend line. The dashed orange bar represents the Election Day outcome. 

This graph shows the poll-of- polls for the final 28 days in 2004. (Polls are assigned to the date that is the midpoint of their coverage.)   Although a trend line can be forced through the data as the wobbly curved line, the distribution of the observations shows no real pattern and is within the range that, according to sampling theory, would occur by chance if there was no actual change during the final 38 days. In other words, technically, we cannot reject the null hypothesis that all observed variation in the final month of the 2004 campaign was sampling error, with voter preferences constant throughout. (The standard deviation of the observations is a mere 1.01.)  Probably Bush was slightly ahead through this period, and the occasional poll that showed Kerry in the lead was a statistical illusion.
We can extract information about the final few weeks in the polls from elections going far back—as far back as 1944. The next graph shows the week-to-week movement in the polls leading up to Election Day and then the shift from the final week’s polls to the Election Day verdict.    Poll verdicts represent the poll-of-polls for the week, with polls assigned to a particular week according to the middle date of their polling period.    Observations are based on polls from 1944 through 2004, although not all election years are represented by polls for the given week. 
Weekly Poll Margins, 1944-2004.jpg
Weekly poll margins by lagged weekly poll margins in the latter weeks of the campaign, 1944-2004. Observations are based on weekly polls-of-polls.. For some weeks of some election years, there were no pollsThe diagonal lines represent lines of equality between the poll margin and the lagged poll margin, not regression lines.
This graph’s obvious feature is the incredible stability of the polls from one week to the next. And then the final polls predict the vote quite well, although the size of the lead in the final polls typically shrinks by about 30 percent on Election Day.    If we regress the poll verdict on the lagged verdict for weeks T-1, T-2, or T-3, the adjusted R squared is .96 or higher. If we regress the Election day verdict on the poll verdict for the final week (ignoring pre-1952 quota-sample polls), the adjusted R squared is .95. Thus, we see that in past campaigns, in the runup to the election, weekly change has come in small increments. 
Still, the slim movements from one week to another carried some meaning. As the next figure shows, from week to week the polls became increasingly accurate as a predictor of the final outcome. We also see that the poll margins exaggerate the size of the vote margins. (Cases cluster in the “West by Southwest” and the “East by Northeast” octants.)   The most remarkable fact is the near absence of cases in the off-diagonal—where the leader in the poll-of-polls ends up losing. The two exceptions of late-campaign comebacks are Truman’s famous surge in 1948 (partially an artifact of  bad polling) and 2000 (Gore’s futile popular vote comeback). 
Vote by Poll Margins, 1944-2002.jpg
 Election Day vote by poll margins in the four weeks leading up to the election, 1944-2004.  Observations are based on weekly polls-of-polls. For some weeks of some election years, there were no pollsThe diagonal lines represent lines of equality between the poll margin and the vote, not regression lines.
So where does this leave us regarding 2008? As of this writing, about three weeks before the election, Obama leads McCain by about 8 points. My back-of-the-envelope calculations based on the regression of the vote on polls on this date in earlier years suggests about an 86 percent chance of an Obama victory, giving McCain one chance in seven of pulling it out (slightly more optimistic for Obama than the current betting markets have it). . This forecast is based on polls alone, without considering how the economic crisis aids Obama.   Then there may be other unknowns this year that bear on the final outcome. So let's continue to keep an eye on the polls.  
The author is a professor of Political Science at Columbia University. The data and analysis presented here is from joint work with Christopher Wlezien of Temple University.


Is It Over?

 Posted by  Larry Bartels

Historically (since 1948), about 75-80% of the margin in a typical October poll has lasted until Election Day. If that holds true this year, the current best forecast of the popular vote based on the polls is that Sen. Obama will win by about 6 points.

Historically, best guesses of this sort have been off by an average of about 5 points. Part of that, about 1 point, is due to pure sampling error in the polls. The rest, about 4 points, is due to other sources of error -- most importantly, changes in voters' preferences between the poll and Election Day. Compilations of many polls (such as average out most of the pure sampling error in individual polls, so the relevant average error is probably about 4 points.

Statistical theory suggests that if Sen. Obama is ahead by 6 plus or minus 4, his current chance of winning (the popular vote) is a little over 90%. This calculation ignores (at least) three important factors. First, there is a fairly strong tendency for late shifts in preferences to favor the incumbent party when the economy is strong and the out-party when the economy is weak; that makes Sen. Obama's position stronger than it looks in current polls. Second, there is more uncertainty than usual this year about who will actually turn out to vote, suggesting that the polls may be more wrong than usual; however, given the pattern of new registrations and the apparent strength of the two campaigns' voter mobilization efforts, Sen. Obama seems more likely to benefit than to lose support from unexpected turnout. Third, the possibility that undecided white voters may, in the end, be unable to bring themselves to vote for Sen. Obama on racial grounds makes his position somewhat more precarious than it would otherwise be. It is very difficult to estimate the importance of the racial factor (that is, the extent to which racial antipathy is not already reflected in current polls), but my guess is that this is less important than the other two factors, both of which seem likely to work in Sen. Obama's favor.

These comments were also posted today in Ezra Klein's blog on The American Prospect.

Pollsters routinely point to their industry’s long record of success in predicting the outcome of the popular vote for the U.S. president. However, it is the statewide elections that matter most, since the president is chosen through the Electoral College and not through a national popular vote. In close national elections, reliable state polls and insightful state-level voting analyses are essential parts of predicting electoral votes and the result of the Presidential election, as well as control of the Congress. Given the likelihood of yet another close election in 2008, our ability to predict the ultimate outcome depends greatly on polls taken at the state level. But who is conducting these polls and how much does their reliability vary from state to state? What are the key demographic and social trends that really drive voting trends in the battleground states? Moreover, how will the statewide polls be affected by having, for the first time in history, a major party candidate who is African American?

On October 7, CSDP co-sponsored a panel discussion on the reliability of state polls in this election.
Christopher Achen, Department of Politics, Princeton University
Larry Hugick, Chairman, Princeton Survey Research Associates International
Andrew Gelman, Departments of Statistics and Political Science, Columbia University
Joe Lenski, Executive Vice President and Co-Founder, Edison Media Research
Co-sponsors: Woodrow Wilson School of Public and International Affairs, the Center for the Study of Democratic Politics, The New York and New Jersey chapters of the American Association for Public Opinion Research

Race in the Race

 Posted by  Larry Bartels

You may have noticed that one of this year’s presidential candidates is African-American. How will that affect the election outcome?

A major new survey conducted by Knowledge Networks for Stanford, the Associated Press, and Yahoo examined prospective voters’ racial prejudices from several different angles. AP’s Ron Fournier reports that “Statistical models derived from the poll suggest that Obama’s support would be as much as 6 percentage points higher if there were no white racial prejudice,” but goes on to add that “in an election without precedent, it’s hard to know if such models take into account all the possible factors at play.”
Back in February some of my starry-eyed liberal friends argued that race would not be an issue, since the people unwilling to vote for an African-American candidate would already be voting Republican anyway. That sounded optimistic to me, so I pulled out the data from the 2004 National Election Study survey and examined the relationship between racial attitudes and voting behavior.
The NES survey asked respondents to rate “blacks” and “whites” on a 100-point “feeling thermometer.” I somewhat arbitrarily decided that people who rated blacks more than 10 points lower than whites on the feeling thermometer—despite the presumed social undesirability of expressing “racist” feelings—would be pretty unlikely to vote for a black presidential candidate. The proportion of Kerry voters for whom that was true: 17%. Uh-oh.
On the other hand, 9% of Bush voters in 2004 rated blacks more than 10 points higher than whites on the NES feeling thermometer. By my logic, those people should be pretty likely to support a black candidate, given the chance. Still, adding together these off-setting effects implied a 3.7% drop in the Democratic presidential vote. That is a huge potential swing in the contemporary electoral context—probably insurmountable except in a year in which the “fundamentals” (the economy, incumbent tenure, presidential approval, organization, and campaign funds) seem to be heavily stacked in favor of the Democrats.
Incidentally, I hear with some regularity that racism will depress Obama’s performance among “downscale white voters.” But the Kerry voters I identified as likely defectors from Obama due to racial antipathy were not particularly “downscale.” They were only slightly poorer and slightly less educated than other Kerry voters. What they were, mostly, is older. One interpretation of that difference is that older people simply feel less constrained about expressing racial antipathy. Another, more hopeful interpretation is that racial antipathy really is less prevalent among younger people. In the latter case, the generation gap in vote intentions evident in recent polls should be attributed not only to the obvious generational contrast between a Reaganesque Republican and a Kennedyesque Democrat, but also to evolving attitudes toward blacks and whites in American society.

McCain for President?

 Posted by  Nolan McCarty

A conservative-liberal oscillation cements his 'maverick' reputation. August 31, 2008

This election is about Barack Obama not John McCain. This is true for a couple of reasons. The first is that in a year so tilted toward the Democrats, Obama wins easily so long as he doesn't run too far behind the rest of his party. But it is also true because after a long stretch in the public eye as a heroic prisoner of war, a senator and a presidential candidate, the voters feel like they know John McCain. Consequently, McCain's polling numbers are stuck around 43 to 44 percent while all the action is in Obama's numbers....

Read the rest of this opinion piece by Nolan McCarty, Keith Poole, and Howard Rosenthal at


About this site

The mission of Princeton’s Center for the Study of Democratic Politics at the Woodrow Wilson School is to promote empirical research on democratic processes and institutions.  That broad mandate has attracted a diverse collection of faculty, students, and visitors pursuing a wide variety of research topics. However, the American electoral process has been a recurrent focus of interest for many of the scholars associated with CSDP and a frequent topic of conferences, colloquia, and other events sponsored by the Center.  As the 2008 campaign unfolds, we thought it might be helpful and fun to collect the election-related research, analyses, and offbeat insights of our extended scholarly community, both for our own edification and as a resource for others interested in how political scientists are thinking about the election.  We welcome contributions, comments, and suggestions. For more about the people and activities of CSDP, please visit our website, To post a comment, click the "speech bubble."

  — Larry M. Bartels, Director