Larry J. Sabato's Crystal Ball
Export date: Sun Oct 22 4:55:49 2017 / +0000 GMT


In less than three months millions of Americans will go to the polls to choose the next president of the United States. For the first time since 1952, neither the incumbent president nor the incumbent vice-president will be on the ballot. Instead, the Republican Party, which has seen its popularity and electoral fortunes plummet since 2004, will pin its hopes of keeping control of the White House on John McCain---an individual who has frequently clashed with his own party's leadership. And McCain's Democratic opponent will be Barack Obama, the first African-American ever to receive a major party presidential nomination.

The fact that neither George Bush nor Dick Cheney will be on the ballot along with the unusual characteristics of the Republican and Democratic candidates have led to considerable uncertainty among political observers about what to expect in November. There is uncertainty not only about the eventual outcome but about whether the electoral map will undergo a drastic change from the familiar blue and red hues of the past two contests. Both Obama and McCain have been courting independent voters and campaigning in states that have not supported their party's presidential candidates in many years. This has led some pundits to suggest that the 2008 electoral map could look very different from the 2000 and 2004 maps. According to these commentators, this year's election is likely to be fought on an expanded playing field with fewer red and blue states and more purple states.

Despite claims by both campaigns that they plan to compete throughout the country, however, there are good reasons to be skeptical about the possibility of a drastic change in the electoral map. That is because the colors of that map reflect the relative size of the Democratic and Republican electoral coalitions in each state and those coalitions do not change drastically from one election to the next. Thus, recent polls show that John McCain is doing very well among the same types of voters who strongly supported George Bush in 2004 while Barack Obama is doing very well among the same types of voters who strongly supported John Kerry in 2004. This can be seen in Table 1 which compares support for the Democratic and Republican presidential candidates among various voter groups in 2004 and 2008. The 2004 data come from the national exit poll while the 2008 data come from the Gallup tracking poll between June 9 and July 27. Because more than 40,000 registered voters were interviewed in the tracking poll over this seven week period, the results for even fairly small subgroups can be considered highly reliable. Moreover, the overall results of the tracking poll have been fairly stable over this period---Obama has led McCain during every week of polling by margins ranging from 2 points to 6 points.

Table 1. Democratic Margin among Voter Groups in 2004 and 2008

2004 2008 Change
Men - 11 - 5 + 6
Women + 4 + 11 + 7
18 - 29 + 9 + 27 + 18
30 - 49 - 7 + 1 + 8
50 - 64 - 5 + 3 + 8
65 + - 5 - 7 - 2
Whites - 17 - 11 + 6
Blacks + 78 + 87 + 9
Hispanics + 10 + 34 + 24
Northeast + 13 + 15 + 2
Midwest - 3 + 7 + 10
South - 16 - 9 + 7
West + 1 + 7 + 6
High School - 4 + 2 + 6
Some College - 8 + 2 + 10
College Grad - 7 0 + 7
Post College + 11 + 16 + 5
Whites Attending Church
Every Week - 43 - 37 + 6
Most Weeks, Monthly - 24 - 16 + 8
Seldom or Never + 8 + 10 + 2

Note: The last name of the winning candidate in each election is capitalized. An asterisk (*) indicates an incumbent president.

Sources: 2004 National Exit Poll and Gallup Tracking Poll, June 9 - July 27, 2008 .

The results in Table 1 show that while Barack Obama is doing better than John Kerry in almost every group, the pattern of support for Obama in 2008 is generally very similar to the pattern of support for Kerry in 2004. Like Kerry, Obama does best among younger voters, women, African-Americans, those with a postgraduate education, and less religious whites. And like Kerry, Obama's does best in the northeastern states and worst in the southern states. There are two important differences between Kerry's and Obama's electoral coalitions, however. First, there is a much larger generation gap in 2008 than in 2004: Barack Obama is doing far better than John Kerry among voters under the age of 30 but slightly worse than Kerry among those 65 and older. To an unusual extent, Obama's overall lead in national polls is based on his huge margin among voters under the age of 30. Second, Obama is also doing much better than Kerry among Hispanic voters, a key voting bloc in a number of swing states including Florida, Colorado, New Mexico, and Nevada. This may reflect a return to a more typical voting pattern for this group---George Bush's performance among Hispanic voters in 2004 was unusually strong for a Republican.

Based on the evidence presented in Table 1, we would expect the 2008 electoral map to differ somewhat from the 2004 map. Opinions of President Bush and the Republican Party are much less favorable now than they were four years ago. As a result, the entire map should show a red-to-blue shift. However, since the GOP has lost ground among almost every voting group in the electorate, the relative level of support for the Democratic and Republican candidates in the states should be fairly consistent with the relative level of support for the Democratic and Republican candidates in 2004: the strongest Bush states in 2004 should be the strongest McCain states in 2008 while the strongest Kerry states in 2004 should be the strongest Obama states in 2008. In fact, this is exactly what the Gallup tracking poll data show. Over seven weeks of polling between June 9 and July 20, Barack Obama had an average lead of 17 points in the "blue states" that John Kerry carried by more than 5 points while John McCain had an average lead of 9 points in the "red states" that George Bush carried by more than 5 points. Obama held an average lead of 7 points in the "purple states" that were decided by a margin of 5 points or less in 2004.

In order to estimate the potential for change in the electoral map, I used the results of recent state polls to compare the level of support for McCain and Obama this year with the level of support for Bush and Kerry in the 2004 presidential election. For 2008, I used estimates of the current level of support for the presidential candidates in all fifty states from the website. These estimates are based on the results of public opinion polls in the states over the past several months.

As expected, the results of this comparison show a very high degree of continuity between the 2004 and 2008 electoral maps. Barack Obama is leading in all 19 states that were carried by John Kerry while John McCain is leading in 26 of the 31 states that were carried by George Bush. Moreover, the five states carried by Bush where Obama is currently ahead include two, Iowa and New Mexico, that were carried by Al Gore in 2000. Obama is leading in only three of the 29 states that George Bush carried in both 2000 and 2004: Ohio, Virginia, and Montana. In all three states, Obama's lead over McCain is less than four points.

Beyond simply looking at which candidate is leading in the polls, we can compare the level of support for McCain and Obama this year with the level of support for Bush and Kerry in 2004. Figure 1 displays a scatterplot of this relationship. The x-axis on this graph represents George Bush's margin vs. John Kerry in 2004 while the y-axis represents John McCain's margin vs. Barack Obama in 2008. Each dot in represents a state and the diagonal line is the regression line that describes the relationship between the 2004 and 2008 results across all fifty states.

Figure 1. Current McCain Margin by 2004 Bush Margin

The data displayed in Figure 1 reveal a very close relationship between Bush's margin in 2004 and McCain's margin in 2008. McCain is generally doing well in the same states where Bush did well four years ago and he is generally doing poorly in the same states where Bush did poorly four years ago. The correlation between Bush margin in 2004 and McCain margin in 2008 is .92 which means that Bush margin in 2004 explains 84 percent of the state-to-state variation in McCain margin in 2008. By comparison, the correlation between Bush margin in 2000 and Bush margin in 2004 was .96. So the relationship between McCain margin in 2008 and Bush margin in 2004 is almost as strong as the relationship between Bush margin in 2004 and Bush margin in 2000. This is especially impressive when one considers that the correlation between the 2000 and 2004 results was the strongest between any pair of elections in the past sixty years and that our measure of McCain support is subject to some measurement error since it is based on the results of public opinion polls rather than actual votes. In some states, the 2008 estimates were based on only one or two polls that were taken several months ago. It is therefore very likely that the final correlation between Bush's vote in 2004 and McCain's vote in 2008 will be even stronger.

The data in Figure 1 suggest that in addition to the 2004 Bush vote, two other factors may be influencing support for the presidential candidates in 2008. One is the residual impact of the divisive Democratic nomination campaign. John McCain appears to be doing better than expected in many of the states that Hillary Clinton won during the Democratic primary race. In addition, McCain appears to be doing better than expected in several southern and border states regardless of who won the Democratic primary. This could indicate that resistance to an African-American candidate among white voters is greater in these southern and border states than in the rest of the country.

In order to provide a stronger test of the divisive primary and racial resistance hypotheses, I conducted a multiple regression analysis of support for McCain in 2008. The dependent variable is McCain's estimated margin against Obama in each state. The independent variables are support for Bush in 2004, a dummy variable (Hillary) coded as 1 for states won by Hillary Clinton during the Democratic nomination campaign and 0 for states won by Barack Obama, and a regional dummy variable (South) coded as 1 for 15 southern and border states (the 11 states of the old Confederacy plus Kentucky, West Virginia, Missouri, and Oklahoma) and 0 for all other states. The results are displayed in Table 3.

Table 2. Regression Analysis of 2008 McCain Support on 2004 Bush Vote, Democratic Primary Dummy Variable, and Southern/Border State Dummy Variable

Source: and data compiled by author

With the addition of the Democratic primary and regional dummy variables, the regression equation now explains 88 percent of the state-to-state variation in support for John McCain. A comparison of the standardized regression coefficients indicates that support for George Bush in 2004 is by far the strongest predictor of support for John McCain in 2008. There is much more evidence of continuity than change in these data. However, the results do show evidence of a small residual effect of the divisive Democratic nomination contest---McCain is doing about three points better than expected in states carried by Hillary Clinton. And despite the fact that George Bush carried all fifteen southern and border states by comfortable margins in 2004, McCain is doing about five points better than expected in those states. While no definitive explanation for this finding is possible, based on the history of race relations in the United States, a reasonable guess is that McCain is doing better than expected in the southern and border states because of greater resistance among white voters in those states to an African-American candidate.

Conclusion: The State of the Race with Less than 90 Days Left

Despite the evidence of a residual effect of the divisive Democratic nomination contest and resistance to Barack Obama by some white voters in the southern and border states, the overall impression that emerges from this analysis of the 2008 polling data is that Obama is entering the final three months of the campaign with a modest but significant lead over John McCain. He is losing the white vote by a smaller margin than any Democratic presidential candidate since Bill Clinton and he is doing better than John Kerry in almost every demographic group, and substantially better among younger voters and Hispanics.

Both the Gallup national tracking poll and recent state polls show Obama leading McCain by a wide margin in the blue states, maintaining a modest lead in the purple states, and narrowing the Republican advantage in the red states. With his own electoral base secure, Obama can afford to shift campaign resources into a number of once solidly Republican states that now appear to be in play such as Virginia, North Carolina, and Indiana. At a minimum this will force the McCain campaign to devote scarce resources to these states---resources that might otherwise be used in Florida, Ohio, and Pennsylvania. Less than three months from Election Day, the 2008 presidential race appears to be Barack Obama's to lose.

Alan I. Abramowitz is the Alben W. Barkley Professor of Political Science at Emory University, and the author of Voice of the People: Elections and Voting Behavior in the United States (2004, McGraw-Hill). He can be contacted via email at

The Crystal Ball's Larry Sabato adds:

Many of our readers may be understandably confused by the Gallup tracking poll that is released daily. Rasmussen Reports also does daily tracking, and we follow its numbers as well. The results are breathlessly analyzed on cable TV, and every jot and tittle is invested with great meaning in the tradition of daily journalism--"McCain scored on this issue, pushing his number up," followed the next day by, "Obama has a terrific new ad on the air, and so that must have added a couple points to his total," and so on. Yet the over-emphasis on each daily tree, separately, makes it difficult for the audience to see the broader forest.

Prof. Abramowitz has provided two graphs (below). The first shows the daily ups-and-downs in the Gallup tracking since the end of the Democratic primaries. What a dynamic campaign! There are great swings back and forth, as though millions of Americans can be pushed and pulled across the party column by every new development (and as though they were really paying such close attention). But in the second graph, Prof. Abramowitz plots out the Gallup trackings on a ten day average. Now the more stable trend becomes clear--Obama has a steady lead of about 3 points. The random statistical noise of daily polling has been eliminated, and a smoother reality becomes apparent.

One can criticize Gallup or Rasmussen or any other poll from many perspectives. Maybe the sampling technique or the turnout model is wrong, including too many of this kind of voter and not enough of some other. Maybe Obama is really ahead by more because young people are under-sampled, for example. Obama has a large lead among the young, and their enthusiasm for him could increase their turnout rate in November. Or perhaps McCain will do better than the polls suggest because of a hidden factor called "racial leakage", which might prod some whites who say they are voting Democratic to quietly abandon the African-American candidate in the voting booth. We don't know, and no one can. And truth be told, the polls aren't going to matter much until the VP candidates are selected, the tickets are complete, the conventions are held, and the bounce that each party gets from the convention subsides--sometime in mid-September. Patience, dear readers.

Trends in Gallup Tracking Poll since End of Democratic Primaries

A. Daily Tracking Poll Results

Analysis: Look--Obama is surging! Oh my gosh, Obama's support is collapsing. No wait, he's surging again! Oh no, he's collapsing again! What happened to that 9 point lead? McCain's strategy must be working. All is lost for Dems--the end is near! They should have nominated Hillary!

B. Ten-Day Average of Gallup Tracking Poll

Analysis: Never mind.

Post date: 2008-08-07 00:00:00
Post date GMT: 1970-01-01 04:59:59

Export date: Sun Oct 22 4:55:49 2017 / +0000 GMT
This page was exported from Larry J. Sabato's Crystal Ball [ ]
Export of Post and Page has been powered by [ Universal Post Manager ] plugin from