# A second look: situational and predispositions compared

We complete this chapter by exploring in greater detail all the determinants of support for the use of force using the Iraq case and their interaction as a test case for a combinatorial model. To do so, we explore three different sets of problems and we test our general theoretical models using block (or nested) regression analysis. The data are drawn from a set of experiments run by the Transatlantic Trends Survey between 2002 and 2004. More specifically, Table 6.12 reports the combination of the factors involved in the different scenarios we will be testing with the use of our theoretical model.

A few words first about the historical context in which these scenarios were explored. They have all been submitted in a period of tense

Table 6.12 Comparison of blocks of determinants of support (nested comparison - likelihood ratio test)

 EU US 2002 2003 2004 2003 2004 LR AIC Diff LR AIC Diff LR AIC Diff LR AIC Diff LR AIC Diff Situation 210,9 5443 - 62,5 5352,4 - 83,3 4514,6 - 20,06 759 48,71 690,8 - Predisposition 246,8 5210 233 579,2 4791,2 561 255,9 4276,7 238 128,5 711 110 17,66 691,2 0 Background 309,5 5087 282 301,0 4518,2 273 240,6 4064,1 213 104,5 564 85 43,8 667,4 24

LR = Likelihood Ratio; AIC and difference among situational, predispositional and background models. All tests are significant at level 0.001, except situation and predisposition in US in 2004. Model terms as in Table 6.11.

transatlantic relations. Therefore, we can safely (and credibly) assume that the Iraq experience guided the answers of most of the people to the experiments themselves. Moreover, the evolution over time between 2002 and 2004 allows for contextualizing different stages of the Iraq crisis: the pre-war stage, in which a battle was fought about the legitimacy of the issue, as we have seen before; the initial war stage, in which the sense of 'mission accomplished' might have influenced the response; and, finally, the post-war disillusionment from 2004 onwards, when it was clear that the fighting in Iraq was far from over. To what extent this situation affected our response can be inferred by a control variable that was included and which measures attitudes toward the use of force in Iraq.

Table 6.12 reveals that the different situational factors explored in the three experiments revolve around four different kinds of factors. A first factor, an identity substitution, is about the country to which the scenario refers. In two of the three experiments reference was made to real countries (Iraq, North Korea and Iran) and in one experiment to a hypothetical 'foreign' country.

The second factor is the nature of the situation, and, again, three kinds of situations were considered: one referred to terrorism, either in general (as in experiment 2004) or with specific reference to Osama bin Laden (experiment 2002). Another referred to weapons of mass destruction (experiments of 2002 and 2003). The last referred to a humanitarian crisis, in which the intervention would aim at establishing 'peace in a civil war in an African country' (experiment of 2004).

A third factor is related to the legitimacy issue, in which the approval of one or another international body is taken for granted. The 2002 experiment explicitly mentioned the lack or presence of UN approval, while the other two experiments offered a range of alternatives (legitimation by NATO, UN, US and allies, or the US acting alone).

A fourth and last factor mentioned was the likelihood of casualties. In 2002, the experiment made explicit reference to the fact that many or rather few casualties were expected.

In all experiments, support for the use of force constituted the dependent variable, but this option was defined differently in the different scenarios. The 2004 experiment asked simply about support for the respondent's country's use of armed forces. The 2002 experiment referred to the respondent's country taking part in the military action and the 2003 experiment explored whether the respondent would support the country's decision to take part in the military action.

Last but not least, we should mention that the 2002 experiment was submitted only to a set of six European countries and not to US respondents.

Each analysis is based on three different blocks of variables, following the multiple blocks model that was introduced in Chapter 2 and examined in the subsequent chapters. The first block is made of a set of three socio-demographics, such as education, gender and age (broken down in classes). The second block is made of predispositional variables and it includes both the four policy beliefs discussed in Chapters 3 and 4 (namely, feelings toward US and EU, Atlanticism, threat perception and attitudes toward the use of force) and the ideological preferences, measured by the Left-Right continuum. The last block of variables is represented by some of the situational factors discussed above.

In the next paragraphs, we first examine and discuss the nested1 6 model in its entirety and then we examine in detail the main results emerging from the different experimental applications. In our analysis, we first look at what each of the three blocks of variables contributes to the overall explanation of support for the use of force. Our full model includes all three blocks of variables - situational, predispositional and background.

We start with a reduced model with situational variables only and we then add the predispositional predictors. In 2002, the reduced model includes three situational variables, to which we add seven predispositional variables and the last 14 background variables (excluding the intercept). In 2003 and 2004 only nine predispositional variables could be included since we do not have the typology on the use of force measurements for 2002. Moreover, in 2002 we have data only for the EU countries.

As Table 6.12 shows, in all cases we can reject the null hypothesis that the additional variables do not contribute to the support for the use of force.17 All three blocks, in all scenarios and for the EU countries as well as the United States, add significantly to the explanatory power of our models. This is particularly true for the EU model, although a comparison across models over time is somewhat debatable. By comparing the log likelihood of the three different models we can test whether the differences among them are statistically significant. If this is the case, then our more general model would fit the data significantly better than the more restrictive ones.

It is quite clear from the results that background conditions, situational and predispositional variables all together significantly improve our ability to predict support for the use of force, under different conditions.

Moving to the substantive results emerging from this set of regressions, Table 6.13 reports the coefficients for the different models, comparing the US and the EU countries. We report here the results with robust standard errors and clustered by country. We also stress that in these results the possibly biasing effect of the Iraq war is controlled for. The control variable measuring support for the Iraq experience - always significant, both in Europe and the United States - takes this biasing effect out. Therefore, the different terms should be interpreted as 'net' of this effect.

Starting with the situational variables, legitimacy is always an important factor in improving support for the use of force. In 2002 (asked only in Europe), not surprisingly, the UN approval is a key factor in boosting support. In 2003 again, the impact of UN Security Council and NATO support is also significant, while the 'coalition of the willing' is not. This is a clear consequence of the Iraq experience at that time. Similarly, the nature of the threat also makes a difference. Mentioning Iran (2003) and Terrorism (2002 and 2004) - as compared to North Korea and Weapons of Mass Destruction - has a positive effect on support, while humanitarian intervention (2004) has a depressive effect. Quite interestingly, the role of casualties was, at least in 2002, quite negligible in shaping European attitudes toward the use of force in the two scenarios presented to respondents.

However, limiting ourselves to the situational variables is not enough. Predispositions add important information about whether people will or will not support the use of force, as shown below.

All policy beliefs and the general ideological orientation are important shapers of attitudes toward the use of force, with some interesting differences between Europe and the US, depending on the scenario involved. Realist threat perceptions are important, and positively so, when the issues are related to terrorism or war against a rogue state. In 2004, when the issue refers to terrorism or to civil war, perceptions of threat arising from a global issue rather than from realist threats play a significant role, and only so in Europe. Feelings toward US (in Europe) and about the EU in the United States are also important, especially when the scenarios are about military or terrorist threats. The index of Atlanticism, moreover, is always important in Europe. In the US this is only the case in 2004.

Last, the typology of attitudes on the use of force plays an important role, especially in Europe, and it works in the expected direction.

 Variables Europe 6 United States 2002 2003 2004 2003 2004 Situational variables Without UN = 0; With UN = 1 2.742“ (0.193) UNSC = 1 2.037“ (0.212) 2.433* (0.844) NATO = 1 1.514“ (0.157) 0.744** (0.0714) 2.768** (0.916) 0.797 (0.198) US+Allies = 1 1.189 (0.125) 1.004 (0.276) Allies = 1 0.717“ (0.0683) 0.790 (0.200) Casualties Tew = 0 - Many = 1 1.043 (0.0722) WMD = 0; Terrorism = 1 1.291“ (0.0899) North Korea = 0; Iran = 1 1.353“ (0.0991) 2.176“ (0.497) Terrorism = 0; Civil war Africa = 1 0.512“ (0.0397) 0.223“ (0.0475) Predispositional variables Threat realism (0-3) 1.445“ (0.133) 1.632“ (0.148) 0.976 (0.0905) 2.026** (0.510) 1.041 (0.285) Threat global (0-3) 0.957 (0.0874) 0.920 (0.0644) 1.416“ (0.128) 0.787 (0.160) 1.173 (0.272)

(Continued)

 Variables Europe 6 United States 2002 2003 2004 2003 2004 Feelings US (0-100) 1.013“ (0.00175) 1.007“ (0.00169) 1.002 (0.00176) 1.018“ (0.00518) 0.994 (0.00615) Feelings EU (0-100) 0.999 (0.00180) 1.004* (0.00171) 1.007“ (0.00187) 0.984** (0.00525) 0.999 (0.00554) Index of Atlanticism (2003: US/EU leadership) (0 = Low - 3 = High) 1.287“ 1.352“ 1.255“ 0.830 1.452** (0.0600) (0.111) (0.0577) (0.256) (0.179) More important economy = 0; military = 1 1.250* (0.142) Hawks = 1 2.610“ (0.489) 1.836* (0.447) 1.546 (0.793) 1.133 (0.611) Pragmatists = 1 2.737“ (0.412) 1.683** (0.277) 2.617* (1.273) 1.114 (0.575) Doves = 1 1.143 (0.172) 0.839 (0.132) 0.873 (0.483) 0.843 (0.481) Left and Right (1 = extreme Left -7 = extreme Right) 1.114“ (0.0309) 1.091** (0.0340) 0.987 (0.0341) 0.961 (0.0853) 0.813* (0.0693) Socio-demographics and background variables Gender 1.442“ (0.101) 1.570“ (0.117) 1.205* (0.0951) 1.431* (0.202) 0.885 (0.183) Age 25-34 1.909“ (0.203) 1.134 (0.127) 1.513“ (0.190) 0.870 (0.196) 1.474 (0.435) Age 35-44 1.593“ (0.164) 1.201 (0.129) 1.208 (0.141) 0.762 (0.150) 2.334** (0.717)
 Age 45-54 1.359** (0.146) 1.013 (0.115) 0.881 (0.102) 0.962 (0.201) 1.757 (0.520) Age 55-64 1.125 (0.130) 1.000 (0.122) 0.862 (0.105) 0.811 (0.182) 1.263 (0.401) Some high school 0.985 (0.0980) 1.097 (0.134) 0.961 (0.121) 0.711 (0.162) 8.13e-09" (l.lle-08) Graduation high school 0.977 (0.101) 1.222 (0.164) 1.071 (0.128) 0.865 (0.164) 1.13e-08" (1.50e-08) Graduation college 0.867 (0.0999) 1.136 (0.154) 1.086 (0.149) 0.780 (0.177) 1.15e-08" (1.54e-08) Post-graduate degree 0.691 (0.205) 0.837 (0.190) 0.861 (0.146) 9.91e-09" (1.34e-08) Germany 0.368*** (0.0431) 0.588*** (0.0756) 0.406*** (0.0570) - France 1.032 (0.126) 1.574*** (0.213) 1.395* (0.198) - Italy 0.771* (0.0951) 0.450*** (0.0587) 1.062 (0.166) " ? Netherlands 0.869 (0.103) 0.726** (0.0873) 0.996 (0.141) - Control Iraq (1 = not invade, 3 = invade US alone) 1.973*** (0.135) 0.377*** (0.0342) 0.518*** (0.0575) 0.0977*** (0.0251) 0.272*** (0.0691) Pseudo R2 0.1359 0.1744 0.1263 0.3292 0.1506 Observations 4076 3925 3660 665 665

Exponentiated coefficients; Standard errors in parentheses

* p < 0.05, ** p < 0.01, *** p < 0.001

Source: Transatlantic Trend Survey, 2002, 2003, 2004.

In the United States, on the contrary, it plays no role in 2003. Last, a difference between Left and Right emerges in Europe as well as the US. In Europe, those on the Right are more likely to support the use of force (but in 2004 the coefficient is not significant), while in the US the opposite is the case. This suggests again - as we discussed already in Chapter 4 - that the Left-Right ideological cleavage is relevant but plays out differently in Europe and in the United States. In Europe, besides policy beliefs about the use of force, Atlanticism, threat perceptions and feelings toward the US, ideological preferences continue to have a role in shaping attitudes toward the use of force, while in the US this is not the case, and when the coefficient is significant, it has a depressing role. This is a sign that once controlled for ideological factors, what is left of the ideological orientation in the US are the isolationists, mostly on the Right, who are shy of any intervention (see the significant and less than 1 coefficient in 2004).

Among the last block of variables, those related to the socio-demographic background, gender and age still play a role - with women less supportive of the use of force both in Europe and the US and younger generations more favorable to the use of force than older generations - while education has no role, once controlled for policy beliefs.

What the findings reported in this chapter tell us about the nature of the Atlantic divide at the level of public opinion and the structural reasons why politicians in the US and EU often find it difficult to agree on the use of military force are discussed in the next and final chapter.