Download data and study materials from OSF
Principal investigator:
Adam Berinsky
Massachusetts Institute of Technology
Email: berinsky@mit.edu
Homepage: http://web.mit.edu/berinsky/www/
Sample size: 1231
Field period: 07/23/04-08/2/04
The most dominant area of research on public opinion and war concerns the role of casualties in determining support for military action. This work has been carried out almost exclusively at the aggregate level. As a result, we know little about the role that casualties play in the microfoundations of support for opinion. In this study, I measure public knowledge of war deaths in Iraq and investigate the effect of informing subjects of the correct casualty figures on support for the current intervention in Iraq. I find that while many individuals knew the correct number of casualties, more individuals did not know the correct level of war dead. I find that misperceptions of the extent of war deaths are influenced by respondents' engagement with the political world. I also find that estimates of war deaths were also influenced by partisan attachments; Republicans were more likely than Democrats to underestimate the number of soldiers killed. Finally, I find that these perceptions of war deaths did not influence attitudes toward war, and correcting respondents' misperceptions of casualty levels had little effect on support for war. Thus, contrary to conventional wisdom, there is little evidence to suggest that perceptions of war deaths drive war support.
- Great variation will exist in knowledge of the number of war dead. Politically engaged individuals will give more accurate estimates of the number of casualties. Partisanship will influence the casualty estimates as well.
- Introducing correct information about the casualty rates will have a smaller impact on support for war than differences in the partisanship of the respondents (e.g. the gap between Republicans and Democrats on support for the war will be larger than between individuals who are only asked to guess the number of casualties and those who guess the number of casualties and are then told the correct casualty rates).
I use a fully crossed 2X2 design. The first treatment is a measure of information concerning the number of American casualties in Iraq and the second treatment concerns the provision of the correct causality information. The experimental design is detailed in Figure 1:
Information Provision
Casualty Guess
No Information Provided
Correct Casualty Count Provided
Respondents do not estimate number of casualties
Condition 1:
Control
Condition 2:
Informed
Respondents estimate number of casualties
Condition 3:
Casualty Estimate
Condition 4
Estimate and Correction
I used two measures of war support
1. Do you think the U.S. made the right decision or the wrong decision in using military force against Iraq?
a. Right
b. Wrong
1a. Do you feel strongly or not strongly that the U.S. made the [right/wrong] decision?
a. Strongly
b. Not Strongly
2. All in all, considering the costs to the United States versus the benefits to the United States, do you think the current war with Iraq has been worth fighting, or not?
a. worth fighting
b. not worth fighting
2a. Do you feel strongly or not strongly that the war in Iraq [has/has not] been worth fighting?
a. Strongly
b. Not Strongly
The mean estimate of deaths in the sample was 952 deaths. However, the accuracy of the median respondent disregards the large variation in the casualty estimates. 42% percent of respondents underestimated the number of war deaths, while 11% overestimated the number of war dead.
I find that higher levels of general political information and greater attention to news about Iraq both increase the probability that a respondent will give the correct answer. However, estimates of war deaths are also influenced by partisanship. As compared to strong Republicans, strong Democrats are more likely to overestimate the casualty levels and are less likely to underestimate those levels.
I find that introducing the correct casualty information does not affect support for war. My experimental design allows me to compare levels of support for the war between two comparable groups: (1) the respondents in the "casualty estimate" condition who underestimate casualties and (2): the respondents in the "Estimate and correction" condition who underestimate war deaths. I can make a similar comparison for respondents who overestimate casualties. This is a powerful comparison, because the "correct information" treatment was randomly assigned. The only difference between the "estimate" group and the "corrected" group is that respondents in the "corrected" condition were subsequently told the true casualty rates. I find no difference in levels of support for the war between these two groups, either for under-estimators or over-estimators.
These results undermine the central tenets of the casualty hypothesis. Perceptions of war deaths are influenced by the respondent's engagement with the political world, but they are also determined by an individual's partisan attachments. Furthermore, the perceptions of war deaths do not influence attitudes toward war, and correcting respondents' misperceptions have little effect on support for war. Whatever inconsistent effects that do exist pail in comparison to the effects of partisanship. In the control condition, where there was no mention of casualties -- respondents were neither asked nor told about the number of war deaths -- the gap in support for the war between Democrats and Republicans was 46 percentage points on the "wrong decision" question and 38 points on the "worth fighting" question.
For a full write-up of the findings, see:
http://web.mit.edu/berinsky/www/war.pdf
Berinsky, Adam. 2007. "Assuming the Costs of War: Events, Elites, and American Public Support for Military Conflict." Journal
of Politics. 69: 975-997.