The impact of question wording: turning a liability into an asset
Public opinion polls are often blamed or considered with profound skepticism because differences in question wording can have a distor- tive effect on the outcome of the poll, that is, in our case support for the use of military force. When we add to this the sheer lack of knowledge of and attention to political issues by many respondents and the oddity of the interview experience, the warning, issued by Mueller on several occasions,1 that a single response percentage has not much meaning in itself, is well justified. Other researchers too have pointed at the impact of question wording and have warned, for instance, against measuring support for war as a dependent variable by taking just one or two questions.2 Eichenberg (2005) concluded: ' ... a single question on any issue will be a misleading gauge of the public mood ... [and] ... a reliable analysis requires the study of many survey questions that employ a variety of wordings.'
These constraints should weigh particularly heavily when we are dealing with comparing differently worded questions asked in different countries at different points in time. To the extent that we compare and generalize across countries we shall have to bear in mind not only the possible disturbing effects of time or differences in question wording, but also of different national politics. Governments and political elites in the countries involved in military operations tend to react rather differently, as was clear also in the cases examined in more detail in Chapter 5 and this was reflected in the domestic political debates in these countries as well as in their public opinions.
In short, to get an idea of the forest, one has to walk deep inside it, looking at the different trees, rather than stopping at only the first tree one meets. To understand the public mood on any political issue and in order to deal with interpretive problems arising from what the public mood is on an issue, it is particularly useful to adopt a comparative approach3 and, more specifically, to compare identical questions over time and differently worded questions at the same time.
In fact, this analysis is built on the assumption that rather than just concluding that survey data taken in isolation may be misleading, we can turn the liability of variations in question wording into an asset, assuming that this very sensitivity to question wording might be helpful in clarifying the conditions under which a certain stimulus produces its effects. We do so also because we follow the tradition that sees the formation of public opinion on international affairs and foreign policy largely as a function of whether, and the ways in which, elites frame foreign policy issues. Differences in question wording by polling institutes generally reflect these frames and in turn help to shape the public debates. We treat these differences as stimuli that reflect the various considerations that people (including politicians) are exposed to or take into account when making up their minds about the appropriateness and acceptability of military force. While differences in question wording may reduce the possibility of making valid comparisons across cases, these liabilities can also be turned into assets, in the sense that they may allow us to recreate, as it were, the societal discourse.
Although attitudes are undoubtedly related too to individual personality and demographic characteristics, we do not look in this section at the impact of the variables/conditions specified in our model at the individual level. Lacking the possibility to undertake a poll that inquires into the effect of all the specified variables, let alone one that allows an analysis across countries, we are forced to accept a next-best solution, which is detailed below.
Recent wars, however tragic for those directly involved, have been a blessing in disguise for those interested in the study of the determinants of public support for the use of force. Compared to earlier international conflicts, they have given rise to a flood of polls, both national and international, offering various possibilities not only to assess the impact of differences in question wording but also to make comparisons across space (exploring the differences in national discourses) and time (exploring the dynamics of support as a factor of the progress of time (such as the effect of the 'rally around the flag' syndrome versus increasing fatigue when results are not forthcoming). Therefore, these cases offer an excellent opportunity to examine the thesis of the gap in public opinion across the Atlantic.