Battleground States and Outcome Scenarios

The 2004 normal vote predictions reported in Table 1 underscore the challenges facing the Republicans in the November election. The Democrats begin with normal vote advantages in states with 282 Electoral College votes. This is 100 more than the Republicans begin with and enough to secure the presidency. Overcoming deficits this large is not unprecedented in U.S. electoral history; Woodrow Wilson, John Kennedy, Lyndon Johnson, and Jimmy Carter all won the presidency by overcoming even greater normal vote deficits. For George W. Bush to join this select group, while generating the smallest set of departures from normal voting patterns, he will have to: 1) “hold his own” in states where the Republicans have normal vote advantages; 2) win all of the states in the competitive category (Colorado, Florida, New Hampshire, Nevada, Ohio, and Louisiana); and 3) score electoral upsets in two states in which the Democrats have the smallest normal vote advantage (Missouri and New Mexico). Thus, I consider these eight states to be the true battleground states in the 2004 election. If the Republicans win all of them they will capture the presidency; if the Democrats win any one of them they will win.

Given the post-1976 electoral patterns noted earlier, the Republican win scenario outlined above would not be extraordinary. The Democrats enjoy safe or secure electoral settings in 24 states; which suggests that two electoral upsets are in line with the 9% defection rate noted earlier. The defection of Missouri and New Mexico (both of which were categorized as competitive states in 1980), in conjunction with a uniform Republican shift among the six competitive states, would produce a Bush victory in the Electoral College similar to what he achieved in 2000. To shed light on the feasibility of this outcome scenario, the normal vote data must be joined with information on factors that are likely to drive electoral perturbations in the 2004 battleground states. There are two possibilities here. One is a uniform, national set of perturbations across battleground states (and, indeed, perhaps all states). The other possibility is that state-specific issues, problems, and electorates could generate crosscutting perturbations.

The issues that seem most capable of generating a uniform, national electoral shift in 2004 are the war in Iraq and/or concern with terrorism. A review of wars and the American electoral experience can shed some light on the prospects for such a uniform national shift; these prospects can be checked by contrasting President Bush's standing in the early polls with those of other recent incumbents seeking re-election. The most discussed domestic issues in the campaign thus far are unemployment and healthcare. The battleground states are differently situated on these issues: unemployment rates and the percent of the population lacking health insurance differ across them. These differences suggest that the receptivity of voters to issue-based campaign appeals may vary across states. Examining these different contexts may produce state-specific insights into departures from normal voting patterns. Examining state-specific surveys on presidential preferences will also help bring the 2004 election into clearer focus.

The Prospects for Uniform Electoral Shifts: Terrorism and the War in Iraq

An extensive analysis of the effect of war on deviations from normal voting patterns shows that wars have seldom benefited incumbents (Nardulli 2005, chap. 8). Indeed, wars are more likely to produce electoral catastrophes for incumbent parties than electoral dividends, particularly unpopular wars such as Korea and Vietnam. The results with respect to war hold true whether the nation was at war on Election Day or whether a war had been initiated and concluded before the election.

Consider first those situations in which the U.S. was at war on Election Day. There have been only seven elections held in the midst of a serious war: 1864 (the Civil War), 1944 (WW II), 1952 (Korea), and 1964, 1968, 1972 (Vietnam).

• In the elections of 1864 and 1944 the incumbent parties achieved little more than the normal party vote.
• In 1952 the Korean War contributed to the worst defeat for the Democrats at the presidential level since the 1920s.
• The Vietnam War was not a dominant issue in 1964 and, due to the abolishment of the draft and “Vietnamization,” it was receding as an issue in 1972. In 1968 the war was a dominant theme and its effect was sufficient to overcome a strong economic performance and seal the Democrats' defeat.

Examining the electoral aftermath of wars begun during a term of office, but concluded before Election Day yields similar conclusions.

• The Mexican American War ended just months before the 1848 election which the incumbent Democrats lost.
• In the election of 1900, which followed the conclusion of the Spanish American War, the incumbent Republicans again received only their normal vote.
• Despite a resounding military victory in World War I that made the world “safe for democracy,” the Democrats were routed in the 1920 election. No election since 1828 surpassed the Republicans' margin of victory in 1920 and no other inter-election swing in presidential voting compares to the 30-point swing experienced by the Democrats between 1916 and 1920.
• The Persian Gulf War generated negative electoral effects for the Republicans in 1992 that had a decisive impact on the election's outcome. The “rally” effects of military victory, so evident in public opinion polls, had little staying power. The election of 1992 was only the second time the Republicans lost a presidential election since the Vietnam War; the other defeat came in the wake of Watergate.

The American experience with war suggests that being a “war president” is unlikely to generate electoral dividends; it has not in nearly two centuries. If saving the union, making the world safe for democracy, defeating fascism, and containing communism were not enough to provide the victorious party with electoral dividends, then it is unlikely that President Bush will be rewarded for having ousted Saddam Hussein. Watching the drama of the Iraq War unfold (the Abu Ghraib prison scandal, the hangings, mutilations, and decapitations of captured Americans, daily armed attacks) provides keen insights into why war has had such consistently negative electoral effects.

Two factors suggest that these historically based findings on war's electoral effects will be highly relevant for the 2004 election. First, most of the historical assessments recounted here were generated before women emerged as a powerful and independent electoral force; their emergence is likely to enhance the historically negative electoral impact of war. Second, recent technological advances in media coverage vividly bring the costs of the Iraq War into every American home as never before. As television exacerbated the electoral impact of the Vietnam War, so the Internet will exacerbate the salience of the Iraq War as an electoral issue.

The Republicans can possibly negate, and perhaps reverse, the routinely negative reactions of the American electorate toward war by underscoring the ties between Saddam Hussein and domestic terrorism. But information emanating from the presidential commission investigating the September 11 attacks, as well as congressional investigations into pre-war intelligence, has begun to erode what was once widely considered a key administration strength: its handling of terrorism. These investigations are making it more difficult to tie the Iraq invasion to the September 11 terrorist attacks.

It is difficult to tell, at this point, whether the Iraq War and/or the concern with terrorism issue will generate a discernible national shift within the electorate. Some insights into this question can be gained by examining trends in national preference polls conducted during the 2004 campaign. To do this I examine a variable termed incumbent preference margin, which is calculated by simply subtracting the proportion of respondents in a survey who prefer the challenger from the proportion that prefers the incumbent. Thus, positive numbers indicate a preference for the incumbent; negative numbers indicate a preference for the challenger. To put the 2004 trendline in perspective, I compare it with preference trends in the last six incumbent bids for re-election. Three of these bids were successful (Nixon, 1972; Reagan, 1984; Clinton, 1996); three were unsuccessful (Ford, 1976; Carter, 1980; George H. W. Bush, 1992).

For each of these election campaigns I used survey results for a 270-day period before the election, beginning roughly in the first week of February.⁴

Figure 2(a) Trends in Incumbent Preference Margin, Successful Re-election Bids

Figure 2(b) Trends in Incumbent Preference Margin, Unsuccessful Re-election Bids

The data presented in Figure 2 indicate that there is no sign of an electoral catastrophe in 2004. Indeed, the available 2004 data indicate no discernible departure from national normal voting patterns. The average of the 77 preference surveys that form the basis for the 2004 entries in Figure 2 average to “.5,” which is very close to the projected 2004 national value of the normal vote. Thus, as it has in several past war-related elections (1864, 1900, 1944), it appears from the available data that the electoral impact of the Iraq war has been to solidify partisan attachments. But extrapolations from the available 2004 survey data must be made cautiously. As Figure 2 demonstrates, most of the movement in incumbent preference margins occurs after the nominating conventions.

While the 2004 data for incumbent preference margin are limited to the March through August period, Figure 2(a) makes it clear that President Bush is heading into the fall in a far weaker position than any of the last three presidents who won re-election. For the comparable time frame in 1972 Nixon had a 15 point average preference margin and a 7 point comfort margin. Thus, he was doing about 8 points better than expected given the relative size of the Republican's electoral base. In 1984, Reagan had a 12 point average preference margin and a 6 point comfort margin, suggesting he was doing 6 points better than expected. Both Nixon and Reagan experienced strong fall rallies and generated impressive departures from national normal voting patterns (16 and 12 points, respectively). During the comparable period in 1996, Clinton's average preference margin was 14 points. While the two parties had comparably sized electoral bases in 1996, Clinton translated this edge into a 10 point electoral perturbation.

Comparing President Bush's current standing with the last three incumbents who failed in their re-election bids produces mixed results. He is well ahead of where Ford was at this juncture in 1976, but Ford began to rally in late summer. Bush in 2004 is behind where Carter was for most of the comparable period in 1980. Carter began to fade in late May, only to rally again in the fall— even though his actual votes failed to match his late surge in most polls; he ultimately lost by a 10 point margin. Bush is, on average, quite a bit behind where his father was at this point in 1992 (8 points). But George H. W. Bush faltered in August and was never able to recover.

As useful as these candidate comparisons are, perhaps the most important insight produced in Figure 2 concerns the power of incumbents to generate a fall surge in the polls. Nixon, Ford, Carter, and Reagan all mounted impressive fall rallies in the polls, even though Carter's rally did not extend into the voting booth. On the other hand, Clinton does not manifest much of a fall surge in 1996 and George H. W. Bush's surge was too little, too late. If the current president can avoid the type of free fall experienced by the three unsuccessful incumbents even a relatively small fall surge (5 points) could be decisive.

Crosscutting Electoral Perturbations: Unemployment, Healthcare and State Preference Polls

If the highly polarized partisan setting in 2004 prevents any widespread fall surge, then the election's outcome will likely be decided by contests in the eight battleground states. Understanding relevant differences across these states can produce insights into the plausibility of the Republicans sweeping these states. A rough picture of the differences in issue contexts and poll results in these states is presented in Table 2.

The first set of entries address unemployment rates. Of the eight battleground states, two rank among the lowest third of all states in terms of unemployment rates: New Hampshire (11) and Nevada (13). Another three of the battleground states rank among the highest one-third of states with respect to unemployment: Louisiana (43), Ohio (37), and New Mexico (36). All of the battleground states except Nevada experienced an increase in unemployment rates during the Bush administration; all have also seen drops over the past year. Information on the proportion of the state's population lacking health insurance is available from the CDC; the latest data available are for 2003. With respect to this measure, two battleground states rank in the top third of all states: Missouri (16) and New Hampshire (3). In contrast, five battleground states were among the bottom third of all states with respect to health insurance coverage: New Mexico (50), Louisiana (49), Nevada (48), Florida (43), and Colorado (36).

The last four columns of Table 3 report state poll results; a column that reports a transformed normal margin of victory variable (from Table 1) precedes these entries. Normal margin of victory was transformed to be comparable to the incumbent preference margin (% preferring Bush –% preferring Kerry). The state incumbent preference margin variable was calculated using only survey results that included Ralph Nader that were conducted after May 1. Thus, no data are available for Nevada.

In Table 2 states were categorized as “Republican” or “Democrat” if the margin of preference favored one party and exceeded the survey's sampling error (usually 4 points); they were categorized as “Leaning” toward one of the parties if the margin of preference favored one party and was equivalent to the survey's sampling error. State preference margins that were within the latest survey's sampling error were categorized as “Competitive;” states with no recent polls were categorized as “Indeterminate.”

The polling data show two battleground states in which the Republicans lead (Colorado, Louisiana), although 12% of Louisiana voters polled were undecided. Four other states with available data are categorized as competitive. New Mexico and New Hampshire categorized as Democratic.

Conclusion

Because of the impact of persistent secular changes in presidential voting patterns over the course of the past three decades, the Democrats approach the 2004 campaign in a surprisingly strong position. In terms of the relative size of their electoral base and its distribution across states, the Democrats have not been so advantaged in nearly 60 years. However, the fact that the Democrats won three of nine presidential elections (1960, 1964, 1976) during this period facing considerable disadvantages in the size of their electoral base demonstrates the impact of election-specific events on electoral outcomes. While one of these electoral upsets was by the narrowest of margins, and two others were marked by unusual circumstances (the Kennedy assassination, Watergate), the normal vote barriers the Republicans face in retaining the White House are not insurmountable by historical standards. Moreover, they fall within the norms of post-1976 electoral patterns.

But joining the electoral base factors with policy relevant considerations makes the Republicans' task in the 2004 campaign more sobering. While no other war president has been able to muster much of an electoral boost in the aftermath of a war, President Bush is in dire need of such a boost. Moreover, six of the battleground states Bush must win in order to retain the presidency rank among the bottom third of all states on either unemployment or health insurance coverage—the two most discussed domestic issues. Two states (Louisiana and New Mexico) rank in the bottom third of all states for both unemployment and health insurance coverage; only New Hampshire ranks in the top third for both indicators. Missouri is middling on both indicators. In New Mexico the Republicans must overcome a normal vote deficit, a poll deficit, and policy-based difficulties. Also, the situation in Florida is compounded by the lingering hostilities and attention rooted in the 2000 election; in Nevada the Republicans must deal with the potential ramifications of the recent efforts aimed at stockpiling nuclear waste in the state.

Countering this rather dismal assessment of the Republicans' prospects are the data presented in Figure 2 demonstrating the capacity of incumbents to generate impressive fall surges in the polls. These data also show that President Bush is well positioned to benefit from such a surge—unlike Ford, Carter, and George H. W. Bush who found themselves in a deep trough by Labor Day. This notwithstanding, there are serious questions about the likelihood of such a surge in 2004. The two incumbents with the least impressive rallies were George H. W. Bush and Clinton. Clinton's inability to mount a fall surge could be a reflection of the highly polarized partisanship that characterizes the current era; if so this polarization may also hinder President Bush's 2004 efforts. Equally important, however, are the implications of the American electoral experience with war. George H. W. Bush's failure to generate a fall surge is consistent with the experience of other war presidents.

If George W. Bush can achieve a war surge, something other war presidents have been unable to do in nearly two centuries, it could have a decisive impact on the election's outcome. A 5 point national shift would capture the necessary Electoral College votes for re-election. If national polls yield no fall surge, then careful monitoring of polls in the eight battleground states will yield insights into the election's outcome.