With anxiety over the nation's direction approaching hurricane force, the 2008 election could blow away many of America's familiar political landmarks.
The collapse of public faith in Washington and the meltdown on Wall Street are generating gales of discontent that could reconfigure each party's electoral coalition and reorder long-standing patterns of support. Democratic presidential nominee Barack Obama is seriously competing for at least 10 states that President Bush carried last time -- including two, Indiana and Virginia, that haven't voted Democratic since 1964 -- and seeking to ignite a historic surge in turnout, particularly among young people and African-Americans. Republican presidential nominee John McCain, meanwhile, is battling these headwinds while trying to mend the GOP's frayed appeal among independent voters and restore the party's reach into states (such as Minnesota, New Hampshire, and Pennsylvania) where suburban swing voters have moved toward the Democrats under Bush.
With Obama now consistently leading in national surveys, McCain's best hope in this difficult climate may be the durability of the underlying partisan allegiances that have divided the country almost exactly in half between the parties for much of the past 15 years. Thirty-four states voted the same way in all four of the presidential elections from 1992 to 2004. And many of the demographic trends in this fall's polls follow familiar grooves. Most surveys, for instance, show McCain continuing to lead Obama among culturally conservative groups, such as working-class whites and regular churchgoers, that have resisted Democrats for three decades. Yet deepening economic anxiety may produce a demand for change powerful enough to tear even those patterns from their roots. After two presidential elections in which each party measured its gains in inches, polls now show Obama and his Democratic Party with the opportunity for a major breakthrough.
To understand the possibilities of sweeping change, and the barriers to it, National Journal has taken an unusually detailed look at the American electorate over the past 20 years. The project examines the results since 1988 from the general election presidential exit polls conducted outside polling places across the country by a consortium of newspapers and television networks. (The organization performing the exit polls has changed over the years. Edison Media Research/Mitofsky International conducted the most recent one, called the National Election Pool survey.)
Because these exit polls are huge (about 13,700 voters were interviewed for the 2004 national sample), they allow for analysis of much more narrowly defined groups than traditional polls do. Reliable conclusions can be drawn about the voting habits of young white men without college degrees, for example, or married Hispanics. In this report, National Journal is providing results for far more subgroups than have typically appeared in media accounts, thus permitting more finely grained assessments of how voters have sorted between the parties. Most political analysts know that white women, say, or seniors have oscillated in their preferences in recent elections; this analysis shows, with new precision, which members of those groups have moved the most. Results for more voting blocs will be released on NationalJournal.com through Election Day.
Ruy Teixeira, a leading public opinion analyst and author at the Brookings Institution and the liberal Center for American Progress, performed the calculations on the exit-poll data files to produce the results for each subgroup, more than 200 in all. National Journal, however, analyzed the results. Those findings display, in unprecedented depth, the pressure points in the electorate: the groups that lean solidly toward one party or the other, and those that swing between them; the groups that are trending toward one side, and those that are trending away. Taken together, the results provide a hidden history of the American electorate over the past two decades.
They also represent a road map to the challenges facing Obama and McCain in the next two and a half weeks. Each nominee is trying to accelerate some of these recent trends -- and reverse others. But as they navigate through the most turbulent political climate since at least 1994 and perhaps since 1968, both McCain and Obama are operating on the terrain that these complex dynamics have shaped.
What follows is an analysis of the electoral trends since 1988 among the key groups likely to decide the 2008 presidential election. These trends provide a yardstick that will measure either the durability of the electorate's divides, or the extent to which this extraordinary campaign reshapes them.
White Men and White Women
For all of the focus on whether Obama, as the first African-American nominee, can attract sufficient numbers of white voters, his struggle in that arena is hardly unique for his party's presidential standard-bearers. Over the past five elections, no Democratic nominee has carried even a plurality of white voters, although Bill Clinton came close in 1992 and 1996 when Ross Perot siphoned a substantial number of them away to his independent candidacy.
White men have been particularly cool to Democrats. Only once since 1988 has the Republican nominee amassed less than a double-digit lead among white men. (That was in 1992, when Perot reached his high point.) And in both 2000 and 2004, white men provided George W. Bush with crushing margins of about 25 percentage points over his Democratic opponents.
Looking at the results from another angle captures the Democratic difficulty among white men even more starkly: The past five presidential elections have involved very different Democratic nominees (from Michael Dukakis and John Kerry to Clinton and Al Gore) running in very different circumstances. Yet over that entire period, the Democratic share of the vote among white men has varied little: ranging between lows of 36 percent (in 1988 and 2000) and a high of 38 percent (for Clinton's 1996 re-election). That remarkable stability suggests a structural resistance to Democrats among these men that will be difficult for any single candidate to overcome.
The exit polls show few real openings for Democrats among white men. The past five Democratic nominees have averaged just 36.6 percent of the vote among white men with less than a college education (who tend to be blue-collar workers) and 36 percent among white men with a college education. Over that period, no Democrat won as much as 40 percent of either of those groups.
To assess the current inclinations of subgroups like these, National Journal compared the historic exit poll results with the cumulative findings for the daily Diageo/Hotline tracking poll for the entire month of September. Those cumulative results are large enough to be worth analyzing even after being sliced and diced into subgroups. We also compared that average with more recent results from the Diageo/Hotline survey.
In the case of white men, the September average showed Obama attracting levels of support among those with and without college degrees comparable to what earlier Democratic nominees received. Nearly 10 percent of each group, though, remained undecided -- offering Obama the possibility of exceeding those benchmarks.
White married men have also been extremely tough for Democrats: The party's recent high point among them was Clinton's meager 36 percent in 1996. Democrats have run somewhat better among single white men. Clinton, with an assist from Perot, carried them both times. Kerry's 46 percent among that group in 2004 was the highest share for Democrats over the past 20 years, and Kerry actually ran even among white, single college-educated men. Those voters are probably a fruitful target for Obama.
Over the past five elections, no Democratic presidential nominee has carried even a plurality of white voters, although Bill Clinton came close in 1992 and 1996.
Since 1988, white women have been both more open to Democrats and more volatile in their preferences. Only Clinton in 1996 won a plurality of white women, but Clinton in 1992 and Gore in 2000 ran about even with them. Over the past 20 years, the Democratic share of the white female vote has varied between 40 percent and 48 percent, and the Republican share, whipsawed by the Perot influence, has swung widely, between 41 percent and 56 percent. White women are voters who can be moved, which often makes them critical to the election's outcome. Although the September Diageo/Hotline average showed McCain leading among white women, for example, the survey's more recent results put Obama ahead.
The past 20 years' results show consistent lines of division among white women. In all five elections, Democrats have won a larger share of the votes of white women with a college education than of those with less schooling. White college-educated women, often liberal on social and foreign-policy issues, preferred the Democrats in 1992, 1996, and 2000, and split about evenly between Kerry and Bush in 2004.
White women without a college education, who tend to be more culturally conservative, have been tougher for Democrats. Clinton ran about even among them in 1992 and carried them in 1996, but they tilted toward Bush in 2000 and stampeded toward him in 2004: Bush beat Kerry among these working-class white women by a 3-2 ratio. Through September, the Diageo/Hotline tracking poll found Obama leading McCain narrowly among college-educated white women but trailing by double digits among noncollege white women; by mid-October, Obama had expanded his lead with the former and significantly narrowed the distance with the latter.
Marriage is another political dividing line among white women. Democrats carried white single women by double-digit margins in each of the past four elections; Republicans, meanwhile, carried white married women every time. Although Clinton stayed close among white married women both times, Bush re-established a resounding 3-2 lead among them in 2004. The most recent Diageo/Hotline results show both single and married white women moving toward Obama since September, but McCain still leading narrowly among the married women.
The interaction of educational and marital status compounds the effects and sharpens the picture. Single women are more Democratic than married women; college-educated women are more Democratic than noncollege women. Not surprisingly, then, Democrats have carried single, college-educated white women in each of the past five elections -- exceeding 60 percent in the last two. At the other end of the spectrum, married, non-college white women (the so-called waitress moms) have voted Republican five straight times; Bush carried almost two-thirds of them in 2004.
White women cross-pressured by their marital and educational status tend to be swing voters. Married white women with a college education split almost evenly between the parties in 1992, 1996, and 2000 but broke sharply for Bush in 2004. Single white women without a college education lean Democratic, but George H.W. Bush carried them in 1988 and his son ran almost even with them last time.
Other trends among whites will be discussed in subsequent sections. But overall, these results point toward a powerful class inversion among white voters. In the middle decades of the 20th century, when economic class served as the principal glue for the two parties' coalitions, Democratic presidential nominees Adlai Stevenson, John F. Kennedy, Lyndon Johnson, and Jimmy Carter all ran at least 13 percentage points better among white voters without a college education than among whites with college or post-graduate degrees, according to the University of Michigan's National Election Studies, an exhaustive post-election poll.
But as the disputes between the parties have increasingly revolved around cultural and foreign-policy issues since the 1960s, Republicans have gained among blue-collar whites and Democrats have improved among white-collar professional whites. The result is that upscale and downscale whites have reversed their political tendencies. In 1988, 1992, and 1996, the exit polls show, the Democrats' nominee ran almost exactly as well with college-educated white voters as he did among those without a college degree. In 2000, Gore ran 4 points better among college-educated whites than among noncollege whites; in 2004, Kerry ran 6 percentage points better among college-educated whites than those without college. And in 2004, that class inversion was evident not only among college-educated white women but also their white male counterparts; in 2000, Gore ran about the same among white men in both categories.
Today's economic maelstrom is providing Democrats with their best opening among working-class white voters since 1992; but given Obama's strength among the upscale, and the cultural resistance he has faced from blue-collar whites, the class inversion could persist or even expand on November 4.
Race, class, gender, and marital status all shape the political choices of independents, critical swing voters prized by both parties. Independents gave George H.W. Bush a resounding advantage in 1988, preferred Clinton in 1992 and 1996, tilted (by just 2 percentage points) to George W. Bush in 2000, and then edged narrowly (by a single point) back to Kerry in 2004, the exit polls show.
Distinct patterns swirl beneath those overall results. The gender gap is important: Democrats have carried independent women in four of the past five elections, while Republicans have carried independent men three times during that period. The marriage gap is also key: On average, Republicans have run 13 percentage points better among married independents than among single ones since 1988. The class inversion is evident among independents, too: Democrats have done better among college-educated independents than among their noncollege counterparts in all five races, by an average of nearly 5 percentage points overall.
The most-powerful factors among independents, though, are ideology and race. Liberal independents have preferred the Democratic nominee to the Republican one by more than 3-to-1 in each of the past four elections. Conservative independents have backed the GOP nominee by comparable margins. Independents who consider themselves moderates -- perhaps the very center of the electorate -- voted for George H.W. Bush (narrowly), Clinton twice (substantially), Gore (narrowly), and Kerry (narrowly).
Looming just as large among independents is race. Nonwhite independents have preferred Democrats overwhelmingly, with three-fifths (or more) voting Democratic in each of the past five elections except 1992, when many opted for Perot. White independents have been much more unsettled. George H.W. Bush romped among them in 1988. Then Clinton narrowly carried them in both of his contests. In 2000 and 2004, they broke toward George W. Bush by narrow 6- and 5-point margins, respectively.
The class inversion has been especially pronounced among white independents. Democrats have consistently run more strongly among upscale "Bill Bradley" white independents than among their downscale "Lou Dobbs" counterparts. In all five elections, Democrats performed better among college-educated white independents than among their noncollege equivalents. That pattern has held among both men and women. It's at Smith & Hawken, not Wal-Mart, that Democrats now find their most tempting targets among independents.
Pervasive dissatisfaction with Bush creates an opening for Obama among white independents, but McCain is a formidable competitor for their votes. The Diageo/Hotline average for September showed McCain narrowly leading among white independents, although more-recent results gave Obama the edge with these critical voters.
Less significant than independents, but also worth watching, are the potential defectors in each party. Over the past three elections, Republicans have won an average of 13 percent of white Democrats; that number rises to 16 percent among non-college white Democratic men, a group that could cause Obama headaches, too. Republican defections have been slightly smaller in recent years, but the exit polls show some inroads for Democrats among GOP moderates and the small cadre of Republican liberals. Both groups might present some openings for Obama.
Democrats have carried Hispanics in each of the past five elections. Only in 2004, when Bush made a major play for their support, did the Republican nominee attract as much as 40 percent of their votes. But since the Republican Party hit its low point with Latinos in 1996 (when Bob Dole embraced an array of conservative anti-immigration proposals), the GOP has regained substantial ground.
In each of the past two elections, Bush improved the GOP vote among Hispanic women and men. In 2004, he reached 43 percent among the men. In 2000 and 2004, Bush similarly improved the GOP vote among married and single Latinos. And in 2004, he achieved 43 percent among the former. (A change in the method through which the exit poll identified Hispanics between 1996 and 2000 contributed to Bush's gains, analysts believe, but do not fully explain them.)
Religion is another significant dividing line among Hispanics. The Democratic nominee has run at least 10 points better among Hispanic Catholics than among Hispanic Protestants (many of whom are evangelicals) in each of the past five elections. Compared with 1996, Bush gained ground with both groups, but Gore still won nearly two-thirds of Hispanic Catholics and Kerry carried more than three-fifths of them. By contrast, Bush ran about even among Hispanic Protestants in 2000 and then won 54 percent of them in 2004. The pattern appears to be persisting this year, though with an overall tilt in Obama's direction: The September average of the Diageo/Hotline poll shows Obama winning nearly two-thirds of Hispanic Catholics but trailing slightly among Hispanic Protestants.
Most important, perhaps, are the class divisions among Latinos. Viewed through that lens, Hispanics vote today the way non-Hispanic whites did two generations ago. In each of the past five elections except 1992, Hispanics without a college degree have voted more Democratic than Hispanics with a college degree, the reverse of the new pattern among whites.
Since the Republican-controlled Congress rejected comprehensive immigration reform in 2006, the overall trend among Hispanics has been Democratic. McCain may have compounded his Latino problem during the GOP primary season when he renounced his support for his own comprehensive immigration proposal. In the Diageo/Hotline September average, McCain was attracting less than one-third of Hispanics, although many remained undecided. If McCain is to improve on that figure, his best chance is probably among well-educated Hispanics. But through September, he trailed Obama among those voters by 2-to-1.
Among African-Americans, the divisions evident elsewhere in the electorate are largely eliminated by an overwhelming preference for Democrats. Over the five elections, Republicans have performed slightly better among African-American men than among black women (by about 5 points on average), but that's little comfort: No GOP nominee over the period has carried more than 15 percent of black men. Likewise, Republicans have run slightly more strongly among married African-Americans than among single blacks (also by about 5 points), but no GOP nominee has carried more than 16 percent of married African-Americans. As with Hispanics, blacks follow the traditional class alignment: Republicans do their best among the well-educated. But the effect is small: Republican nominees have run only about 3 points better among college-educated African-Americans than among their non-college counterparts, and exceeded 11 percent of college-educated blacks only in 1996 and 1988. In any case, Obama's electric appeal to black voters appears likely to sweep away all of these considerations: The September Diageo/Hotline average showed McCain winning just 4 percent of African-Americans.
Voters under 30 moved sharply toward the Democrats in 2004, giving Kerry a 9-point margin over Bush. And Obama appears poised to do even better: The September Diageo/Hotline average gave him a 14-point lead among young voters. But the exit polls show that the story is much more complex than those numbers suggest.
Among white voters under 30, the Democratic position has been much more tenuous. Clinton ran even among young whites in 1992 and narrowly carried them in 1996. But George H.W. Bush won 61 percent of them in 1988, and his son carried 55 percent of whites under 30 in each of his two elections. The September Diageo/Hotline average showed McCain leading among whites under 30 although other national surveys have Obama flipping this group sharply toward the Democratic column.
Young white men without a college education have been especially difficult for Democrats. Kerry's modest 40 percent showing among them was the highest total for any Democrat over the five elections; George W. Bush twice won about three-fifths of them. Democrats haven't done much better among young, white college-educated men, although Kerry spiked to 46 percent. Young women have been more of a swing group: Noncollege white women under 30 voted for Clinton twice, but then voted twice for Bush twice. College-educated young white women voted for Clinton both times and also for Gore before switching to Bush in 2004.
Nonwhite voters have been the foundation of the Democrats' strength among young people. Democrats have won between 72 percent and 81 percent of their votes over the five elections, and they account for Obama's lead among the young this year in the Diageo/Hotline poll as well.
A similar pattern is evident among seniors. Democrats have won at least two-thirds of nonwhite seniors in all five elections. But Republicans carried white seniors in three of them; only Clinton broke the trend in 1992 and 1996. Kerry's 44 percent among white seniors was the worst showing for any Democrat in the period.
The trend toward the GOP has been sharpest among blue-collar white seniors. The Republican vote among white seniors without a college education soared from 42 percent for Dole in 1996 to 56 percent for Bush in 2004; among college-educated seniors, the GOP vote improved from 50 percent to 56 percent. The Republican gain was especially large among senior white men without a college education -- fully 64 percent of them backed Bush in 2004. (By comparison, Kerry narrowly carried senior white women without a college education.) Those older downscale men look like a tough group for Obama, too. They gave McCain a substantial lead in the September Diageo/Hotline average.
Among whites, Democrats have often faced the greatest resistance from voters in their prime years of work and family formation. Even with the deterioration among seniors, the 2004 election produced a kind of bell curve result for Democrats: Kerry performed best among whites under 30 or over 65. In three of the past five elections, the Democrats' weakest performance has come with whites ages 30 to 44.
The Democrats' problem has been especially acute with blue-collar, middle-aged white voters. Those working-class families, many of them with children, are arguably the prime target for the economic populist message that Democrats have stressed since Clinton left office. But in each of the past five elections, Republicans have amassed a bigger advantage among noncollege whites ages 30 to 44 than among noncollege whites in any other age group. Bush carried those voters by fully 2-to-1 in 2004. Democrats haven't done much better among noncollege whites moving toward the empty-nest years (those ages 45 to 64): They also gave Bush almost two-thirds of their votes last time.
On the other hand, in 2000 and 2004, Democrats ran better among college-educated whites in each age category through the working years -- 18 to 29, 30 to 44, and 45 to 64 -- than among their non-college counterparts. The September poll average suggests that Obama may be more likely to extend those gains than to dramatically reverse the Democratic decline among working-class whites in the central years of family life.
Historically, the critical link between religion and politics was religious affiliation. White Catholics traditionally preferred Democrats; the Republican Party was founded as a party of Northern Protestants. Since the 1960s, Southern evangelical Protestants have substantially switched their loyalty from the Democrats to the GOP.
Traces of that old alignment are still evident in the exit polls since 1988. In each of the past five elections, the Democratic vote among white Catholics has been at least 9 points greater than the party's vote among white Protestants, a group now split between overwhelmingly Republican Southern evangelicals and closely divided Northern mainliners.
But increasingly, religious practice is the dividing line between the parties. The University of Michigan's National Election Studies show little difference between the voting preferences of Americans who attended religious services regularly and those who did not during the 1950s and 1960s. But a religious attendance gap appeared in the 1970s and has remained constant since: The more often voters attend religious services, the more likely they are to vote Republican.
That effect is only slightly apparent among African-Americans. But in the exit polls, which asked about religious attendance in 2000 and 2004, Bush ran better each time among Hispanics who attended church regularly than among those who did not. Among whites, the attendance gap holds for both Catholics and Protestants. In 2000 and 2004, Bush performed about 20 points better among white Protestants who attended church at least once a week than among those who went monthly or less. Among Catholics, the attendance gap was smaller but still substantial: 11 points in 2000 and 8 in 2004. In 2004, even white Protestants who rarely attended church strongly preferred the GOP. But white Catholics who go to church infrequently are a key swing group: They backed Gore narrowly in 2000, then tilted slightly to Bush in 2004.
Another measure of the same phenomenon is the growing number of Americans -- 10 percent of voters -- who claim no religious affiliation. Democrats have won substantial majorities even among white voters in that category in all five elections.
As on many other fronts, polls this year are showing similar patterns. In the Diageo/Hotline average for September, McCain had a commanding lead among white Protestants and white Catholics who attend church at least weekly. Obama was winning narrowly among white Catholics who attend church irregularly; trailing among white Protestants who do so; and capturing a solid majority of white voters without any religious affiliation.
Those results, like many others this fall, testify to the consistency of the parties' coalitions for much of the past two decades. But the stability of those familiar electoral alignments will be severely tested by the wave of discontent and unease now cresting as Election Day approaches.
Surveys used in this analysis
In addition to the Diageo/Hotline 2008 poll, surveys used in this analysis include CBS News/New York Times: 1988 National Election Day Exit Poll, 11,645 respondents; Voter Research and Surveys: 1992 National Election Day Exit Poll, 15,490 respondents; Voter News Service: 1996 National Exit Poll, 16,637 respondents; Voter News Service: 2000 National Election Day Exit Poll, 13,225 respondents; and National Election Pool: 2004 National Election Day Exit Poll, 13,719 respondents.
Margin of error
The huge size of the exit polls means that subgroups are much more likely to be representative samples than in a typical poll. However, there is still a margin of error associated with exit-poll estimates. The smaller the number of respondents in a given subgroup, the greater the margin of error. The margin of error, as typically defined, is the number that can be added or subtracted from a poll finding to get the 95 percent confidence interval for the finding. For example, if 30 percent of a group are found to have voted Democratic and the margin of error is 3, you can have 95 percent confidence that the true figure for the entire voting population in that group was between 27 percent and 33 percent.
In the exit polls, the margin of error ranges from 1, when the sample size for a subgroup is over 8,000, to a minimum of 6, when the sample size is 100 or less. We flag subgroups with sample sizes under 100 with an asterisk to indicate that their margins of error are relatively large and that those results should be viewed with caution.
Go to NationalJournal.com for dozens more charts that plot the political history of a wide number of voting blocs -- ranging from Hispanics over age 45 to independents who regularly attend religious services.
This article appears in the October 18, 2008, edition of National Journal Magazine.