by BARRY C. BURDEN
The Forum, November 15, 2004

In the immediate aftermath of the 2004 presidential election, a journalistic consensus emerged to explain George W. Bush's victory. Despite the sluggish economy and deteriorating situation in Iraq, voters supported Bush primarily because of his values. One prominently featured exit poll question showed "moral values" to be the most important issue for voters, ahead of terrorism, Iraq, and the economy. Backlash against the Massachusetts court ruling allowing gay marriage and attraction of Bush's appeals to Christian faith helped bring out socially conservative voters and cement Bush's second term. This explains why Bush won Ohio, for example, where an anti-gay marriage proposal was on the ballot. However compelling this story might be, it is wrong.

Instead, Bush won because married and white women increased their support for the Republican ticket. Though definitive evidence from the National Election Study will not arrive until spring, I hypothesize that women responded especially well to Bush's handling of national security in the post-September 11 environment. The increase in voter turnout in 2004 also helped Bush to a degree, but the Democrats' mobilization of young people kept Bush's margins down in battleground states. Yet it was women who helped Bush improve his vote share in most states. The shrinking of the gender gap and expansion of the marriage gap made the difference.

In this article I briefly account for the factors behind Bush's rise in the state-by-state popular vote between 2000 and 2004. This is not the same as identifying who elected Bush. That sort of analysis would put responsibility on white men since they voted 61-38 for Bush and comprise almost half of the active electorate. Instead, I focus on what changed between 2000 and 2004. In this view, it is white women who are responsible because they showed more aggregate change.

Identifying a cause for this shift looks for an explanation also in things that changed in the past four years. For example, John Kerry was not exactly Al Gore, so differences between Bush's two opponents could be a factor. But I suggest that such differences are dwarfed by a much larger intervention: the attacks of September 11. Turnout was up in 2004 because the perceived heightening of the stakes after 9-11 and because of intense competition between the candidates in a small number of battleground states. Higher turnout also appears to have helped Bush slightly. But it was the shift of married white women from the Democratic camp to the Republican camp that gave him the edge in 2004.

Little New in 2004

Simply put, Bush won the 2004 presidential election because he improved his share of the vote in a wide range of states. It was not a distinct surge in the Bible Belt or Midwest or any other socially conservative region that put the Republican ticket over the top. The overwhelming pattern is one of stasis. The geographic basis of the vote in 2004 closely replicated patterns exhibited in the 2000 election. As Table 1 shows, only three states changed partisan hands. The other 47 (and Washington, DC) went to the same party in both elections.

In the midst of great stability, even small changes can have important consequences. For example, reapportionment after the 2000 election helped Bush in an important way. The newly distributed Electoral College vote gave more weight to the Republican South and West, less to the Democratic Northeast and Midwest. As a result, many of the states Bush won in 2000 became worth more electoral votes in 2004. The 271 electoral votes Bush collected in 2000 were worth 278 electoral votes in 2004, a gain of seven before the 2004 campaign even began.

In addition to this initial advantage, Bush managed to win two new states in 2004 while only losing one that he had captured in 2000. The one state he lost--New Hampshire--was worth just four electoral votes. The two new states he won--Iowa and New Mexico--combined for 15 electoral votes.

Bush gained 2.9 points in the nationwide popular vote between 2000 and 2004. In the two-party vote, his gain was a smaller 1.6 points. As Figure 1 shows, that gain was spread across many states. The graph displays Bush's share of the two-party vote in both years. Most states are above the 45-degree line, demonstrating that Bush gained in more states than where he lost. Bush improved in 32 states, by an average of 2.5 points.

[FIGURE 1 OMITTED]

The few states where his support fell appreciably are those where Nader ran particularly well in 2000. The four circles farthest below the line are Vermont (where Nader got 7% in 2000), Alaska (10%), Montana (6%), and Oregon (5%). Most voters who chose Nader in 2000 either dropped out or decided to support Kerry in 2004. All told, Bush's share of the two-party vote fell in only 19 states, by an average of 1.3 points. This is roughly half the size of the average gain in the other 31 states.

Surprising is that his support dropped off most in traditional Republican strongholds in the Mountain West (Alaska, Montana, and Idaho) and upper New England (Vermont, New Hampshire, and Maine). His biggest gains came in some historically Democratic states such as Hawaii, New Jersey, Rhode Island, and New York. Bush even ran better in Kerry's home state of Massachusetts than he had against Gore in 2000.

The nearly uniform increase across the county challenges the notion that evangelicals or Latinos or any other concentrated group is responsible for his victory. Rather, a widely-distributed increase must be due to getting more votes from a widely-distributed demographic group.

I posit that women were the key swing constituency. Exit polls show the gender gap at just three points in 2004, less than half of what it was in 2000. Men actually voted for Kerry as much or more than they had for Gore, but women swung decidedly to Bush compared to 2000.

More specifically, married white women are almost entirely responsible for the change in the Bush vote. White women split equally between the major party candidates in 2000, and had favored the Democrats by five points in 1996. But in 2004 exit polls show that they reversed course, voting for Bush 55-44. As 36% of the voters in 2004, white women's five point swing could theoretically be responsible for most of Bush's gain (5 points x 36 percent [approximately equal to] 1.8 points). Married women showed a similar shift to the Republicans. They voted 46-48 against Bush in 2000 but 55-44 for Bush in 2004.

In the larger context of so much geographic and demographic stability, these are seismic shifts. The shifts are found exclusively among white and married women. (1) Improvement among these demographic groups is consistent with the widespread bounce Bush got in 2004. It is also tentative support for the popular "security mom" thesis, though concerns about security also affected women without children. Though we cannot be certain from exit poll data, they suggest that concerns about national security helped improve Bush's standing among white women.

While it is impossible at this point to rule out any effect of "moral" issues in driving up Bush's support among women, their influence is suspect. Most importantly, surveys show that women are consistently more liberal than men when it comes to gay marriage and other cultural issues. A Pew study conducted a year before the election showed women supporting both gay marriage and civil unions more than men did. (2)

In addition, there is not much evidence that social conservatives turned out disproportionately more in 2004. Protestants made up 54% of the voters according to exit polls in both 2000 and 2004. Those who attend church weekly were 27% of voters in 2000 and 26% of voters in 2004, hardly a surge. In addition, these two groups gave Bush as much or more of their vote in 2000 than they did in 2004.

Voter Turnout

Voter turnout was up appreciably in 2004, but the increase was far from historic. As Figure 2 shows, by one measure turnout increased 4.3 percentage points over 2000 to 58.9%. (3) This comes in just below the 1992 level. It is an improvement in particular on 1996, when turnout hit its lowest point in decades.

[FIGURE 2 OMITTED]

But 2004 does not touch the higher turnout elections of the 1950s and 1960s. Though an improvement, it should not be regarded as record-setting in any sense.

Since 1952, the range between the highest turnout election (1960) and lowest (1996) is a mere 11.2 points. But the variation among states in any given year is far greater. State turnout in 2004 ran from a low of 47.9% in Hawaii to a high of 78.0% in Minnesota. This is a range of some 30 points, which far exceeds the small changes in overall turnout from one election year to the next.

Changes in voter turnout between 2000 and 2004 varied dramatically across the states depending on whether the presidential election was closely fought there or not. For example, though overall turnout rose, it actually fell in California and New York, where the outcome was not in doubt. It was driven up more than 10 points in South Dakota due to the high-powered U.S. Senate race there. More generally, the battleground states saw turnout up by 7.4 points on average while the remaining states saw turnout rise by only 4.6 points. (4) It was only the slide in turnout among some noncompetitive states that kept nationwide turnout from creeping higher.

A more rigorous analysis of changes in voter turnout is presented in Table 2. The dependent variable is the change in turnout in each state between 2000 and 2004. The explanatory variables account for the presence of minor party candidates, the two-party distribution of the vote, the intensity of the major-party race, and whether initiatives banning gay marriage were on the ballot. The first column includes all states and DC; the second column excludes the 16 states where Ralph Nader was not on the ballot. Both models are estimated using ordinary least squares regression to evaluate these influences simultaneously.

Least surprising in these results is that turnout was higher by two to three points in battleground states. The parties concentrated their advertising efforts in a small number of media markets in 2004. A rich literature in political science has already pointed to the relationship between elite mobilization and voter turnout (Rosenstone and Hansen 1993). The face-to-face canvassing that was so prominent in 2004 is an especially effective way to boost turnout (Green and Gerber 2004). These techniques were used extensively in swing states such as Ohio, Florida, and Wisconsin, but much less elsewhere.

Contrary to much media analysis, the results show that mobilization around gay marriage ballot measures had no aggregate effect on turnout. Any additional stimulus among the evangelical right was offset by those opposed to the gay marriage bans. This result contrasts quite sharply with claims such as the Atlanta Journal-Constitution's Jim Galloway's conclusion that "the presidential election just behind us was all about God, values, gay marriage and turnout of the white, conservative Christians in unprecedented numbers." Though gay marriage or President Bush's appeals to "faith" might have increased turnout among conservative religious adherents, it made no more (or less) of a contribution to turnout than did increases among the Democratic base.

Finally, there is scant evidence that the two most prominent minor party candidates increased turnout. Libertarian Michael Badnarik appeared on all but one state ballot and drew almost as many votes as Nader did. But in the end he was not responsible for any of the increase in voter turnout. Nader not only failed to bring out new voters but turnout actually increased less in states where he ran well, namely politically lopsided states in the Northeast and Mountain West. This was in stark contrast to 2000 where roughly half of Nader's support came from new voters. The reason for this negative effect is probably that the closeness of the vote in some states encouraged strategic voting by Nader supporters (Burden 2003).

An interesting side note is that support for Badnarik and Nader does not appear to run along ideological lines. Though one might have suspected that the two candidates would appeal to quite different constituencies given their contrasting platforms, in practice they drew support from similar constituencies. The correlation between their state vote shares is a robust .84. Even limiting the analysis to states where Nader was on the ballot, the correlation is .64. This suggests that support for the two candidates was driven at least as much by protest against the two major parties as it was for their specific policy agendas.

The Partisan Consequences of Turnout

The 2004 election challenged the conventional wisdom that Democrats always benefit from higher voter turnout. While most partisans believe this to be true, political science research suggests that most turnout effects are negligible, though high turnout tends to help the Democrats in more elections than not (Citrin, Schickler, and Sides 2003). Yet before the election nearly every observer expected that high turnout would help Kerry. Independent 527 groups such as America Coming Together hired canvassers in battleground states to register and turn out Democratic voters. Young people are notoriously difficult to turn out, but they were expected to favor Kerry. In short, higher turnout should have hurt Bush.

This reasoning appears dead wrong. Table 3 explains Bush's state share of the vote in 2004 as a function of turnout and other factors. Based on the stability in the vote demonstrated above, I assume that the Bush vote in 2004 is largely a replication of his vote in 2000. I also include the change in turnout from 2000, whether the state was a battleground, and whether anti-gay marriage initiatives and Ralph Nader appeared on the ballot.

The sharpest finding in Table 3 is that, if anyone, Bush not Kerry benefited from higher turnout. Though the efforts of the 527 groups and the Democrats themselves surely increased turnout, it was the Republicans who were advantaged in the end. For every four percentage points that turnout increased between 2000 and 2004, Bush gained a full point.

At the same time, Bush generally did worse in battleground states than elsewhere. (This result holds even if the model controls for the South, where Bush did best.) So while turnout per se helped Bush, mobilization by pro-Kerry groups in battleground states kept Bush's margins down. This might be because the Democrats so narrowly targeted the swing states while mostly ignoring the rest of the country. Again, there are no effects of anti-gay marriage initiatives. These ballot measures had no effect on Bush's vote share, nor did the presence of Nader on the ballot hurt Kerry this time.

Concluding Thoughts

Based on the rudimentary analysis reported here, several of the media's immediate conclusions about the 2004 presidential election appear to be in doubt. First, the election was not fundamentally about mass support of right-wing religious values. Though evangelical conservatives turned out at a higher rate than in 2000, so did just about every other demographic group. There is no evidence that the presence of gay marriage initiatives on 11 state ballots had any effect on Bush's vote share or turnout more generally.

It is ironic then that the increase in voter turnout (mostly in non-battleground states) helped Bush rather than Kerry. Perhaps the multifaceted mobilization effort waged by the Kerry campaign, the Democratic National Committee, 527 groups, and other anti-Bush forces was not coordinated enough (either by law or practice) to provide much of a net advantage. In this view, the unified field effort of the Bush campaign was simply more efficient at turning out its voters. Even so, the Republicans were not turning out Christian conservatives disproportionately. Rather, Bush's strength was distributed quite widely. His vote share actually increased more in blue states than red states, which makes the centrality of "moral values" in his victory difficult to accept.

A more plausible explanation is that white married women, who unlike other subpopulations are distributed rather evenly across the states, are responsible for almost all of Bush's increase between 2000 and 2004. Men voted almost identically to how they had in 2000, but women gave the Republicans four more percentage points in 2004. This was enough to provide the popular vote victory for Bush that he was denied four years earlier.

Precisely why white women moved is a matter to be resolved with further study. It is probable that at least some women decided to stick by the incumbent administration as a source of stability in a time of international turmoil. These so-called "security moms" (not all of whom were actually mothers) showed concern about future terrorist attacks and believed that the military action in Iraq was a reasonable part of the "war on terror." Bush's trespassing on traditional Democratic issues such as education did not hurt his cause among women either.

Kerry might have made a stronger mark on these issues had he not been forced to explain what seemed like inconsistent positions. This lack of clarity might be an inherent part of Kerry's personality or a strategic mistake made by the campaign. More likely, it highlights a general liability faced by senators who run for president (Burden 2002). Senator Kerry voted for the Iraq invasion and admitted that he would have done it again later, even after learning that weapons of mass destruction and links with Al Qaeda were unsubstantiated. He could not adequately explain his decision to vote against $87 billion in supplemental funding for the war or why he opposed No Child Left Behind despite voting for it in the Senate. Even though female voters disagreed with many of Bush's actions, enough preferred the certainty of a second term to the ambiguity of the Kerry message.

Temporary shifts like the ones that occurred among white and married women following 9-11 are not likely to be good predictors of future electoral coalitions. Whether Karl Rove's plan to transform the two Bush victories into lasting Republican dominance comes to fruition is an open question. But it is clear that 2000 and 2004 are something of a crossroads. These two elections are the closest consecutive presidential elections since the 1880s. And the Republican margins in the Congress are the slimmest in the history of the American two-party system. Time will tell us whether post-September 11 politics are fleeting or they usher in a new era of party alignments.

Table 1: Few States Change Hands
 
                  Winner in 2000
 
                  Bush   Gore
 
Winner    Bush      29      2
in 2004   Kerry      1     19
 
Table 2: Turnout Up in Battleground States,
Down Where Nader Does Well
 
Nader's 4-Way Share of the Vote      -3.24 **    -3.41 **
                                     (1.40)      (1.49)
 
Badnarik's 4-Way Share of the Vote    4.11        3.07
                                     (4.21)      (4.50)
 
Kerry's 4-Way Share of the Vote       -.06 *      -.06
                                      (.03)       (.04)
 
Battleground State                    2.53 **     2.35 **
                                      (.83)       (.98)
 
Gay Marriage Initiative on Ballot      .71         .004
                                      (.82)      (1.03)
 
Constant                              8.21 ***    8.70 ***
                                     (1.71)      (2.22)
 
Adjusted [R.sup.2]                     .32         .35
Number of Cases                         51          35
 
Ordinary least squares regression coefficients with standard errors
in parentheses. * p < .10, ** p < .05, *** p < .01, two tailed test
 
Table 3: Bush Gains Where Turnout is Up,
Loses Where the Race is Closely Fought
 
Bush Share of 2-Party Vote in 2000            .93 ***
                                             (.03)
 
Change in Voter Turnout from 2000 to 2004     .25 *
                                             (.14)
 
Battleground State                          -2.19 **
                                             (.91)
 
Gay Marriage Initiative on Ballot            -.14
                                             (.82)
 
Nader on Ballot                               .24
                                             (.72)
 
Constant                                     3.92**
                                            -1.73
 
Adjusted [R.sup.2]                            .95
Number of Cases                                51
 
Ordinary least squares regression coefficients with standard errors
in parentheses. * p < .10, ** p < .05, *** p < .01, two tailed test

(1) The presentation of exit poll data after the election do not allow for the simultaneous examination of marital status, race, and gender. The National Election Study will allow study of the interaction between sex and marital status in a more sophisticated manner.

(2) http://pewforum.org/publications/surveys/religion-homosexuality.pdf

(3) Throughout I use McDonald and Popkin's (2001) turnout calculations based on the "voting eligible population." How turnout is measured--particularly what denominator is used to calculate the percentage--is a source of some debate.

(4) I use the 11 battleground states identified by Nagourney and Seelye (2004).

References

Burden, Barry C. 2002. "United States Senators as Presidential Candidates." Political Science Quarterly 117:81-102.

Burden, Barry C. 2003. "Minor Parties in the 2000 Presidential Election" In Models of Voting in Presidential Elections: The 2000 U.S. Election, ed. Herbert F. Weisberg and Clyde Wilcox. Stanford, CA: Stanford University Press.

Citrin, Jack, Eric Schickler, and John Sides. 2003. "What if Everyone Voted? Simulating the Impact of Increased Turnout in Senate Elections." American Journal of Political Science 47:75-90.

Galloway, Jim. 2004. "Religious Vote Fuels Victory for GOP." Atlanta Journal- Constitution. November 4.

Green, Donald P., and Alan S. Gerber. 2004. Get Out The Vote! How to Increase Voter Turnout. Washington, DC: Brookings Institution Press.

Michael P. McDonald, and Samuel Popkin. 2001. "The Myth of the Vanishing Voter." American Political Science Review 95:963-74.

Nagourney, Adam, and Katharine Q. Seelye. 2004. "Bush and Kerry Focus Campaigns on 11 Key States." New York Times. October 24.

Rosenstone, Steven J., and John Mark Hansen. 1993. Mobilization, Participation, and Democracy in America. New York, NY: MacMillan.

Barry C. Burden *

* Barry C. Burden is Associate Professor of Government at Harvard University. He is co-author of Why Americans Split Their Tickets: Campaigns, Competition, and Divided Government (University of Michigan Press) and editor of Uncertainty in American Politics (Cambridge University Press). Burden has also published articles in the American Political Science Review, American Journal of Political Science, British Journal of Political Science, Legislative Studies Quarterly, and elsewhere on American electoral politics and representation.

* Harvard University, burden@fas.harvard.edu

Gale Document Number: A134619149