The Political Fortunes of War: Iraq and the Domestic Standing of President George W. Bush

Richard C. Eichenberg, Tufts University and Richard J. Stoll, Rice University

This website contains supplementary materials for our monograph of the same name that is published by The Foreign Policy Centre in London, England.

Updated Analysis Through Oct 11, 2004. Posted October 29, 2004

Our analysis of the impact of casualties on the political fortunes of George W. Bush was published on July 13, 2004.  This brief note updates the analysis through October 11, 2004. For complete details on our procedures and results, please consult the original full report.  

The analysis of approval of the President's handling of Iraq has been updated through the middle of October, 2004. The analysis of the President's job approval has been updated through the end of August for the full model including real disposable income per capita (August is the last month for which these data are available). However, because disposable income is statistically insignificant in our model, below we also report results for job approval through October 11, 2004 with disposable income excluded from the statistical model.

The following are updated analysis, date, and results files.

Bush Job Approval Compared to Previous Incumbents Seeking Re-Election

 

An updated version of Figure 2, which compares the job approval rating of past incumbents seeking reelection, is available here (note: to view the figure, you must have Adobe Acrobat).

By our calculations, Bush's approval rating remained essentially static in June, July, and August at 47-48 percent (average of CBS/NYT, Gallup/CNN/USA Today, ABC/Washington Post, and Pew Center for the People and the Press).  As a result, his approval rating was lower than the approval rating of incumbents who successfully sought re-election in the past, but it was higher than the approval of past incumbents who lost. 

However, in September, Bush's approval rating improved  to an average of 51 percent --a  change that contradicts what we would expect from our earlier analysis. Eighty American soldiers died in Iraq during September, a number above the average for the previous three months. Yet the President's approval rating increased.

We discuss the reasons for this unexpected change below. Here we would note that, through the middle of October, Bush's approval rating had once again declined to 47 percent --a level that puts his re-election at risk given the historical patterns revealed in Figure 2 above.

 

Table 2: Impact of Casualties on Approval of President's Handling of Iraq

 

In the original report, we estimated that each 100 deaths of American service personnel lowered approval of the President's handling of Iraq by 2.9 percentage points.  The revised estimate is a drop of 2.0 percentage points for each 100 deaths.

 

Table 4: Impact of Casualties, Disposable Income, and Major Events on Presidential Job Approval Ratings.

 

The original estimate (on data through April 30, 2004), was that each 100 deaths of American personnel lowered the President's job approval rating by 1.2 percentage points. Equally important, we found no significant impact of per capita disposable income on the approval rating.

 

The revised estimates (through August 31, 2004) show that each 100 deaths of American personnel lowered Bush's  job approval by 1.5 points between March 19, 2003 and August 31, 2004.   Since disposable income has no discernible statistical impact on job approval, we also replicated the analysis for the period ending October 11, 2004 (now ignoring disposal income, but including a dummy for the impact of the Republican Convention).  The results are almost exactly the same: for each 100 battle deaths in Iraq, the President's job approval rating has declined by 1.48 percentage points.

 

Discussion

 

Our conclusion in the original paper was that the Bush presidency since the beginning of the Iraq war has been essentially a war presidency. Bush's job approval is affected (negatively) by casualties in Iraq; economic conditions have had no statistically significant effect. Only the most dramatic international events (such as the rally at the start of the war) have any impact on his approval ratings.

 

The revised estimates through October, 2004 do not change these conclusions.  The economy has had no appreciable influence on the President's approval rating, and the handover of sovereignty to the Iraqi government in July had no upward impact on Bush's ratings (it was not, for the public, a truly dramatic event).

 

Thus, our original conclusion continues to hold. The Iraq war has rendered this election more competitive than it would otherwise be. Based on our estimates, Bush's approval rating would be on the order of 58 percent were it not for the casualties suffered in Iraq.  This would place Bush easily within the range of comfortable re-election, based on past history.

 

The anomaly for our analysis occurs in September, when the President's general approval rating (and indeed approval of his handling of Iraq) actually increased, despite the fact that the casualty rate actually increased over the level of June and July.

 

Our explanation for this anomaly hinges on the dynamics of the Presidential campaign and our speculation on the impact of the press coverage that follows. The Republican convention culminated on September 1, 2004 with a concentrated effort to focus the country's attention on the "war against terror" and to equate the war in Iraq with the war against terror. Public opinion surveys suggest that the President succeeded. During September, the "war against terror" increased slightly as the most important electoral issue (ABC/Washington Post), and the Iraq war declined slightly (these trends are especially prominent among women, as described here).  As the salience of the terrorism issue increased, so did the President's approval rating (51 percent in September), his Iraq approval rating,  and his standing in the horserace polls against Kerry.

 

The situation in October was the reverse.  Beginning on September 20, Senator Kerry launched a sustained criticism of the President's handling of the Iraq war, and the criticism was amplified from the platform of the first two presidential debates on September 30 and October 9.  During this period, two things happened: 1.) Iraq rose once again to exceed terrorism as the most important issue (again, especially among women); 2.) the President's approval rating dropped once again (to an average of 47 through October 11).

 

In short, the renewed focus on Iraq dropped the approval rating to a level closer to what our model would predict. Thus, we might argue that even the "anomalous" improvement in the President's September approval rating confirms the logic of our model: it happened only when attention was diverted from the Iraq issue, and it dropped once again when the focus switched back to the topic of Iraq and the problems there. As we argued in the original report, this Presidency --and this election-- are dominated by Iraq.  Were it not for the war and the casualties suffered there, the election would not be as close as it is.

A Methodological Note on Measuring Casualty Totals and Their Impact on Presidential Approval

Much of the work on the relationship between casualties and support for war has used the log (base 10) of casualties, rather than (as we do) the total of  battle deaths.  In John Mueller’s classic work he justifies using the logged total as follows:

"...one assumes that the public is sensitive to relatively small losses at the start of the war but only to rather large ones toward its end. Specifically, one does not expect casualties to affect attitudes in a linear manner with a rise from 100 to 1000 being the same as one from 10,000 to 10,900. Rather, a rise from 100 to 1000 is taken as the same as one from 10,000 to 100,000. Thus the distance between the numbers, 10, 100, 1000, 10,000, 100,000 and 1,000,000 is made equal." [John Mueller. 1973. War, Presidents, and Public Opinion. New York: John Wiley, p. 60.]

In essence, Mueller believes that as casualties increase, it takes greater and greater thresholds to have an impact on public opinion.

We believe that since – fortunately – the death toll in Iraq has been much lower than in Korea or Vietnam (the two wars Mueller studied), so the public reaction to casualties in the initial stages is based on the total of deaths. This point of view is supported by our initial analysis. But as time goes on and deaths increase, the relationship could change to the logged relationship. Below we compare the coefficients for total battle deaths and logged total battle deaths for the equation predicting Bush’s Iraq approval, and for the equation predicting Bush’s overall job approval after the start of the war. We show the comparisons for analysis ending July 31, and analysis ending September 29.

Bush's Approval on Iraq

                                                      Through July 31                                               Through September 29
Measure Coef. Std. Err. z P>|z| Measure Coef. Std. Err. z P>|z|
Total Battle Deaths -.030406 .0080214 -3.79 0.000 Total Battle Deaths -.0197 .0074 -2.66 0.008
Log10 Total Battle Deaths -14.32631 9.42274 -1.52 0.128 Log10 Total Battle Deaths -13.3085 9.0051 -1.48 0.139

Bush's Overall Job Approval

                                                      Through July 31                                               Through September 29
Measure Coef. Std. Err. z P>|z| Measure Coef. Std. Err. z P>|z|
Total Battle Deaths -.014113 .0043097 -3.27 0.001 Total Battle Deaths -.0160 .0022 -2.51 0.012
Log10 Total Battle Deaths -9.357726 3.892377 -2.40 0.001 Log10 Total Battle Deaths -15.5111 2.6827 -5.78 0.000

As you can see, in terms of Bush's approval on Irag the simple total of deaths outperforms the logged total. But for overall job approval, the logged total battle deaths now outperforms the total number of battle deaths. This suggests to us that for many people when they evaluate the overall presidency of George Bush, they have begun to distance themselves from the immediacy of individual deaths in Iraq. However, if the focus is on Iraq, the level of (unlogged) deaths is the predominant influence.

Analysis Files for the Original Paper

The following are files that contain the analysis discussed in the original paper or will allow you to replicate it.

Other Documents