Of the several socio-economic and demographic characteristics that potentially affect the level of support for the use of force, gender, age and education have received by far more systematic attention, followed by race and socio-economic status. The first, and most lasting, pattern to emerge is about gender. Since the first psychometric studies back in the 1930s, women have been systematically found less supportive of war and the use of force than men. The gender gap, as it was later dubbed, was first reported by Cantril (1940), in studies before and during World War II, and confirmed in the few available studies of the 1950s and 1960s, both among special groups, such as students (Garrison, 1951: 53; Farber, 1951: 220; Putney and Middleton, 1962: 658-665), and in the general population (Verba and Brody, 1970: 329; Modigliani, 1972; Suchman, 1972). The first systematic scientific research on support for the use of force, the Stanford study, found that, in 1966, men were systematically less pacifist, more in favor of escalation and less favorable to de-escalation than women (and this relationships held constant, also controlling for level of education) (Verba et al., 1967: 323).8
Studies that looked at the gender gap from a longer time perspective,9 all concur that on issues related to the use of force a systematic (although not dramatic) gap hovering between 6 and 10 percentage points exists among males and females; and that this gap remains stable over time, even in the presence of a steady decline in the proportion of 'don't know' responses among women (Shapiro and Mahajan, 1986: 56ff.). The size of this gap, however, varies also depending on the issue and content of the question. It is the largest on questions about support for the use of force in specific, real, circumstances, such as the Vietnam or Iraq war. It steadily decreases moving from questions about the use of force in hypothetical situations to questions about defense spending, to almost disappear for questions measuring general attitudes toward international relations (Benson, 1982). A few studies have explored the variation in the size of the gap as the situation in which force is used varies. In particular, Eichenberg (2003), in what is the most recent, systematic and detailed analysis on the topic, concludes that 'the type of military action, the principal policy objective, and the mention of casualties in the survey question each influences the gender difference when taken individually' (Eichenberg, 2003: 134-135). And, looking at these factors together, it emerges that women 'are relatively less likely to endorse violent (or escalatory) actions; they are relatively more sensitive to the loss of human life; but they are relatively more sensitive to humanitarian objectives' (Eichenberg, 2003: 136-137). This conclusion fits well with the explanation of the sources of the gender gap offered by Conover and Sapiro (1993), according to which women learn, in the early stage of socialization, 'to put off the use of violence until later in the course of a conflict than do men, to escalate its use more slowly, and to be more emotionally upset by it. ... Given such patterns of early learning, it would be surprising if women did not react somewhat differently to war than do men, particularly when it is "real" rather than merely hypothetical' (Conover and Sapiro, 1993: 1096).
Age is another variable found since the 1930s to be relevant by the literature in explaining levels of support for the use of force. However, both the meaning and the direction of the relationships between age and support for the use of force are quite controversial. At least three different factors have been suggested to explain differences between younger and older generations: life-cycle, generational and inter-generational effects.10 The first refers to the specific stage in life in which the respondent finds herself, with people changing their attitudes as they get older. Generational effects refer to the varying experiences different age groups live through and that have a lasting impact on the way they see issues for their whole life. Last, an inter-generation gap can be explained by the existence of issues on which generations can have conflicting stances and interests. As an example, issues like social benefits to the elderly and the draft can raise intergenerational conflicts.
The literature, although sparse and not systematic, has been oscillating both in assessing the existence of such an age-gap and in explaining what are the sources of it. The few studies before World War II were based on too narrow a sample to be of any generalizable use. In general, 'college students tested before the entrance of the United States into the war were mildly pacifistic or strongly pacifistic' (Carter, 1945: 344). Similarly, Cantril (1940) reported that, before World War II, young people were more reluctant to support a war against Germany, probably, he suggested, out of self-interest, since they were more likely to be doing the fighting in the case of war. However, attitudes toward war with Germany turned more belligerent once the war broke out (Stagner and Ross, 1944b; Cantril, 1942) for all age groups, defying any potential inter-generation gap due to the differential impact of the human costs of war.
Studies after World War II showed conflicting results. Some studies report young people more supportive of war (Farris, 1960: 61; Cutler, 1970; Rosenberg et al., 1970; Mueller, 1973: 136-137), also controlling for education. Gergen and Back (1965), on the contrary, found in a set of surveys carried out over the 1950s, that older people tend to prefer more violent and short-range solutions to more moderate and longterm ones. In a systematic review of all poll questions asked between 1936 and 1972 on 'warlike sentiments' that allowed for an age breakdown, Erskine (1972-1973) reported that before Vietnam, younger cohorts were systematically more supportive of 'aggressive policies' than older cohorts; after 1968, younger people (those under 30) turned more pacifist than older ones (over 50). However, Erskine (1972-1973: 617) also showed that the generational gap varied widely depending on the content of the question. Youth was less willing to use nuclear weapons, more likely to oppose air bombing and less likely to consider specific wars as mistakes (see also Erskine, 1970) than older generations, they were also less disposed to disapprove of the use of American troops abroad than older people.
The idea that the Vietnam war represented a turning point among different age groups has been extensively discussed, although more in reference to elites than to the public (Roskin, 1974). Allison (1970-1971) for example, suggested that the younger generations were getting 'cooler' on foreign affairs since Vietnam 'magnified the conflict between costs of empire and needs at home' (Allison, 1970-1971: 158), with young Americans much more reluctant to bear the costs, both human and financial (Russett), of American hegemony. However, such a shift does not appear to be borne out by the available data on attitudes toward the use of force and on foreign policy in general.11
In what is probably the most exhaustive analysis of cohort effects, Mayer (1992: 151-166) concludes that 'cohort replacement has had very little effect ... on the state of public opinion toward foreign policy'. Similarly, Page and Shapiro, in their extensive review of aggregate trends in American public opinion, stressed that, of the three possible age effects (namely, life-cycle, generation and period), the most important in foreign policy is the period effect, as shown by the parallel trend in the ebb and flow of the percentages of those who responded that the Korean war was a 'mistake' (Page and Shapiro, 1992: 304-305; see also Holsti, 2004: 200-204 for an updated review of these data).
One systematic attempt to study the role of generational effects in foreign policy is by Schuman and Rieger (1992), who explored what was the most appropriate analogy, that of either Hitler or Vietnam, in describing attitudes toward the Gulf War and whether this is related to age. Predictably, they found that older people were more likely to choose the analogy with Hitler and the younger cohorts, to pick up Vietnam.12 More interestingly, they discovered that what analogy is chosen makes for a powerful predictor of support for the war. Ninety percent of those choosing Hitler supported the Gulf War, while only 55 percent of those who chose the Vietnam analogy supported that war. However, when analogy and age are considered together, age becomes weakly and negatively related to support (Schuman and Rieger, 1992: 323). They explain it with the cross-cutting impact of biological and social reasons. On the one hand, age has 'a negative direct effect on support for the war (older people are less supportive)' (Schuman and Rieger, 1992: 325). On the other hand, age has an indirect effect via the chosen analogy, that tempers the initial negative bivariate relationship between age and support: 'the appeal of the World War II analogy to older cohorts reduced somewhat their opposition to war in the Persian Gulf, while the appeal of the Vietnam analogy to younger cohorts tempered their tendency to give full support to military action in the Gulf' (Schuman and Rieger, 1992: 324).13
Since the beginning of empirical survey research on foreign policy, knowledge and education (or, more appropriately, the lack thereof) were singled out as key variables in explaining international attitudes. What Almond (1950: 127) called the 'attentive' public and Kriesberg (1949: 51) the 'informed' people, on the basis of their greater awareness, alertness and interest in politics, often have different foreign policy attitudes than those less attentive and uninformed. However, even if the average level of schooling in American population has dramatically increased,14 no aggregate change can be discerned over time in the level of knowledge of foreign policy issues among the population. In other words, no composition effect due to different levels of education can be discerned in attitudes toward the use of force (Mayer, 1992: 215-222).
In general, the better educated have been usually found to be more supportive of the use of force. This was as true in the 1950s (Suchman, Goldsen and Williams, 1953: 178-180; Putney and Middleton, 1962: 665-666; Back and Gergen, 1963a: 82-84; Rogers, Stuhler and Koenig,
1967: 246)15 as it is now (Mueller, 1994). However, studies have also found that the better educated are more likely to change their mind. Mueller (1973: 122-136), in an extensive analysis of support for the Korean and Vietnam wars among different educational groups records a greater sensitivity to war's developments among the better educated. This greater variability of the better educated contributed to narrowing down the educational gap during the Vietnam war. In Vietnam, before 1968, the less educated were more likely to consider the war a mistake. 'During 1967, however, the gap between the college and high-school educated narrowed, and in 1968, disappeared altogether. These groups then supported the war at roughly the same level until Gallup stopped asking the question in 1971' (Lunch and Sperlich, 1979: 37-38). Page and Shapiro (1992: 316) also report 'a slightly greater propensity to change opinions' among the better educated, but 'this tendency is small indeed', in part because it is offset by the lower propensity to change attitudes among those endowed with the greater stock of knowledge.
The empirical results discussed so far do not support the 'Enlightenment' model (Gamson and Modigliani, 1966), according to which, in line with Kant, when the choice of fighting is left to the people, they will be reluctant to wage wars, because they know what costs they will have to put up with. Since education makes the prospective assessment of these costs easier, one should therefore expect a negative correlation between education and support for the use of force (Gamson and Modigliani, 1966: 187). The prevalent explanation of the education gap has to do with what has been called the 'mainstream' model, according to which education 'brings with it, not so much better understanding of the world as greater participation in it and attachment to the mainstream. The politically educated are not better analysts of complex situations but are simply more aware of what official US policy is' (Gamson and Modigliani, 1966: 188). According to this model, the better educated would be more likely to support government policies, irrespective of whether they are belligerent or pacifist. This tendency of the more educated strata of society to follow governmental policies has been confirmed also by other studies (Putney and Middleton, 1962; Mueller, 1973: 123-124). Zaller (1992) has framed the argument in the theoretically most articulated way, in explicit relation with the Vietnam war, connecting education - or what he calls political awareness - with political predispositions and political debate. We will discuss it in the next section, once we explore the role of predispositions.
Much less attention has been paid by the literature to socio-economic status.16 Social status as an independent determinant of attitudes toward war has produced mixed results. In part, this is a consequence of the high correlation between income, social class and level of education (Hughes, 1978: 44). Studies explicitly devoted to socio-economic status are few, and all concentrated during the Vietnam war as part of a wider debate about middle-class authoritarianism in explaining support for the use of force. The results are inconclusive, however. Some studies found no, or a weak, relationship between status (differently measured) and support for the use of force (Verba et al., 1967; Modigliani, 1972: 962-964). Other studies found instead a positive relationship (Robinson and Jacobson, 1969: 67; Hahn, 1970: 1190, 1192; Hamilton, 1968; Galtung, 1964).
A last background variable to consider here is race, and, in particular the black-white distinction.17 Several American studies conducted during the Vietnam war reported blacks to be systematically less supportive of the war than whites (Hamilton, 1968; Hahn, 1970; Verba et al., 1967; Mueller, 1973). Verba et al. (1967: 325) found that, among the sociodemographic variables, race was the strongest predictor, also controlling for level of education (see also Converse and Schuman, 1970: 22; and for a contrary view, Zaller, 1991: 1228-1229). Lunch and Sperlich (1979: 35) remarked that the opposition of blacks to the Vietnam war was 'the most widely noted demographic aspect of public opinion about the war in Vietnam'. The pattern is more systematic. It was present before that war too, and continued into most recent wars as well. Hamilton (1968) compared attitudes toward the Korean and Vietnam war and found that blacks in both conflicts were more supportive of 'pulling out' or negotiation and less supportive of 'bombing China' and of a 'strong stand' than whites. Similarly, Mueller reports (1973: 147-148) that blacks in both wars were both more likely to judge war a 'mistake' and less likely than whites to support escalatory moves to bring it to an end. This race gap appears to hold also in later wars. For the Persian Gulf War, Holsti (2004: 228-229) reports data showing that 'race may well be emerging as one of the more powerful sources of foreign policy cleavages'. The gap between blacks and whites is defined as 'enormous'; and such a gap holds also in the 2003 Iraq war (Holsti, 2004: 230). Nincic and Nincic (2002) suggest in their analysis of four cases (Korea, Vietnam, Desert Shield and Desert Storm) that the gap may have even widened over time. Quite interestingly, Holsti (2008: 228) remarks that support for intervention in South Africa, Haiti and Somalia among blacks was much higher than for Bosnia, suggesting that the race gap might also be affected by the specific situation in which force is used. Some other studies have shifted their attention from the direct impact of race on support of war to its mediated role in shaping the impact of events on opinion. Two studies in particular single out this mediated role of race: Zaller (1991) and Gartner and Segura (2000), with mixed results.
These four variables - age, gender, race and education - although widely studied in the past, have known a common evolution over time. Most of the earlier studies looked at these characteristics for their individual, bivariate contribution to explaining patterns and differences in support for the use of force. However, with time, attention has moved from assessing their independent contribution - either in isolation or partialling out their effect from other confounding variables - to their moderating effects, in combination with other variables. These analyses, of which distinct examples are Eichenberg (2003) for gender, Gartner and Segura (2001) for race and Zaller (1990, 1992) for education and race, have looked at socio-demographic factors in interaction with other variables closer to the situation, namely predispositions and cost-benefit calculations, to which we now turn.