The last block of factors brings us more closely to the situation on the ground and how this may affect public opinion on the use of force. This implies not only a shift in emphasis from personal and ideological characteristics to contextual factors, but also a turn in theoretical perspective. When considerations related to events are used to explain the support for the use of force, and the list of factors being considered is quite diverse, with casualties usually in the first place, the underlying model is a cost-benefit calculation in which respondents engage when requested to express their opinion.
This model differs from the one suggested by others like Zaller (1992), Berinsky (2008) and Alvarez and Brehm (2001), according to which people form their opinions not on the basis of a rational calculation of the pros and cons of the intervention, but in reaction to the political discourse among the elites prevalent at the time. The two models might be closer than they appear, however, if we consider a few very recent studies showing that people on the one hand actually make calculations in determining their stance on the use of force, but that, on the other hand, these calculations are very much affected by their own ideological position and by the way the issue is framed by the media and the elites (Gaines et al., 2007).
We start by discussing the cost-benefit model and its main variants. Next, we explore a mediated cost-benefit model and we eventually conclude with the role of framing in explaining support for the use of force. The cost-benefit model, according to which 'decisions on the use of force are cost-benefit decisions' (Feaver and Gelpi, 2004: 98), is the underlying model of several scholars who, quite independently from one another, pointed to the concept of rational calculation as the way to better understand why and when people release or withdraw their support toward the use of force. The underlying assumption of all these models is that what counts for an explanation are the factors that affect costs and benefits respectively. Arraying them along two columns, the total sum produces either a positive or negative sign value that determines whether people are in favor or against the use of force in specific circumstances.
In listing what considerations people might have in mind in deciding whether to support the use of force, five factors are usually discussed in the literature (see Klarevas, 2002; Gelpi, Feaver and Reifler, 2009; Larson, 1993; Everts and Isernia, 2006): the goals of the mission, casualties, the multilateral nature of the action, the interests involved and the expectations of success. We shortly examine each of them. However, before doing that, it is important to reiterate that all factors discussed here assume an assessment of costs and benefits of the contextual situation, irrespective of the personal impact of the war on the respondent's personal interests. In other words, all cost-benefit models discussed in the literature, as far as support for the war is concerned, are definitely sociotropic in their perspective.
The reasons are related to the, perhaps, surprising result that the few available studies that have looked at the personal costs and self-interests, on balance, have been disappointing. It does not take a great stretch of imagination to hypothesize that perceived costs of a war, be they material or in terms of human costs (killed or wounded) are inversely related to the willingness to support it, particularly when the people involved perceive that they will also be the ones to bear the burden of the costs. Support would thus appear to be driven by self-interest. As mentioned before, it was Kant who claimed that support for the use of force was mostly shaped by individual self-interest. However, not only has this seemingly obvious relationship been quite rarely studied, it also does not appear to be borne out by the available evidence. The scattered set of empirical results available consistently shows no relationship between support for the war and the likelihood of bearing the personal and actual burden of the war. Cantril (1942) was the first to report no difference in attitudes towards war among persons with family members of military age and persons without such family members (Cantril, 1942: 148; Cantril, Rugg and Williams, 1940: note 1). Hahn (1970: 1194), during the Vietnam war, also found that family characteristics related to the proportion of draft-age people (social class) did not predict in a coherent way the position on Vietnam. In a more systematic comparison, Mueller (1973) found no relationship between support for the war and self-interest. Lau, Brown and Sears (1978) explicitly contrasted a self-interest explanation of support for the war in Vietnam with a symbolic politics approach (Sears et al, 1985). They found, if possible, exactly the opposite. Those who had a relative or a friend involved in the war, and were, therefore, more self-interested to the war, 'were somewhat more likely to feel the war was the "right thing" and to support a stronger stand in Vietnam.' A revised form of such self-interest is suggested by Gartner and Segura (1997), who argue that the public sociotropic judgement is formed locally. 'Respondents are likely to evaluate a war based on an assessment of societal costs, but this assessment is heavily influenced by the weight given to their proximate information' (Gartner and Segura, 1997: 674-675).
If we consider instead a wider conception of interests, the relationship with support for the use of force appears stronger and more robust. Russett and Nincic (1976) were probably the first to point to the importance of interests for the United States of the country attacked. They found that Americans were more likely to support the employment of American military force for countries geographically, economically and politically closer to the United States and that were externally attacked rather than internally subverted. This finding anticipated one of the most systematic analyses of the role of interests in support for the use of force, the Principal Policy Objective (PPO) hypothesis put forward and tested by Jentleson (1992; see also Jentleson and Britton, 1998; Larson, 1996, 1996a). Jentleson argues that the 'principal policy objective (PPO) for which military force is being used' is 'the most powerful and parsimonious explanation for the variation in support' (Jentleson and Britton, 1998: 396), for the use of force between the 1980s and the 1990s and, possibly, ever since the end of World War II (O'Neal et al., 1996).Jentleson argues that, in the 1980s and 1990s, support for the use of force varied, depending on whether it was for restraining an opponent 'engaged in aggressive actions against the United States, its citizens, or its interests' (Jentleson and Britton 1998: 397), or to bring about an internal policy change in another country 'whether in support of an existing government considered an ally or seeking to overthrow a government considered an adversary' (Jentleson, 1992: 49-50). Jentleson lists four reasons why the American public has 'a stronger disposition ... to support the use of military force to restrain rather than remake governments'
(Jentleson, 1992: 53). The more traditional nature of both the conflict ('interstate' versus 'intrastate') and the strategies to be used (military force versus political strategies, such as 'hearts and minds') in the foreign policy restraint operations, as well as their greater international and domestic legitimacy, make this type of operation easier to justify and to defend to the public than the Internal Political Change operations. In a later article, he added a third policy objective, humanitarian intervention, intended as 'the provision of emergency relief through military and other means to people suffering from famine and other gross and widespread humanitarian disasters' (Jentleson and Britton, 1998: 400). Analyzing nine cases in the period 1980-1991 and six cases in the post-Cold War 1992-1996 period, using cross-cases and intra-case variation in policy objectives and (with Britton, 1998) a multivariate analysis based on content analysis of question wording, Jentleson found that 'when the principal objective was to restrain an adversary, when that adversary had gone beyond simply posing a standing threat and initiated aggressive actions against American interests or citizens, the public was prepared to support military action. But when the principal objective was to remake the government of another country, the American public was disinclined to support the use of limited military force, either directly or indirectly' (Jentleson, 1992: 64). Jentleson is ready to admit that there is 'nonautomaticity' in the relationship between public support and policy objectives. 'Public support will not necessarily just be there; it must be cultivated and evoked through effective presidential leadership. But this evocation is far more likely to succeed when the principal objective is foreign policy restraint, even in the face of significant risks' (Jentleson, 1992: 71). In a subsequent analysis (with Britton, 1998), Jentleson made this qualification more precise, looking at the relative weight that other variables, such as interests, presidential cues, multilateralism and risk aversion, exert in shaping public support.
Another important consideration in peoples' minds is related to the costs of the military operation, usually measured in terms of human casualties. The obvious relevance of this element, for both the war itself and as part of the calculation people might make about the costs and the benefits of the war, however, hides the intrinsic difficulties people have in thinking of the merit of a military operation in terms of casualties. Asking people to consider a war or any other sort of military operation from the viewpoint of the human costs of it is, quite understandably, putting them in an odd situation. Not surprisingly, Rosen (1971) in studying the cost-tolerance for different foreign policy goals among students,25 concluded that
‘Lay actors seem to have fairly clear foreign policy priorities but not very clear cost-tolerance limits for these goals. They resist thinking in terms of human life costs for situations where such costs are inevitable. Cost- tolerance limits that are expressed are implausibly low: they do not seem to reflect the actual willingness to suffer. The limits expressed are also poorly related to the relative importance of the goals, as contrasted with the rigid logical prescriptions of utility theory.' (Rosen, 1971: 67)
These findings suggested to Rosen that most of the people set goals prescriptions unrelated to the calculation of costs and this then backfires when the costs become apparent.
No matter how difficult and uneasy the situation the respondent is put in, survey research has nevertheless not been shy to ask, in quite different ways, whether people are prepared to accept casualties, and how many they are ready to take before giving up on an operation, and a pattern emerges from these data. Mueller (1971, 1973), in a landmark study, looked at support for the Korean and Vietnam wars and found that, in the aggregate, irrespective of the different patterns and pace of casualties, the trends in support were very similar in both (Mueller, 1971: 366-367): taking the logarithm of the total number of American casualties suffered at the time of the poll (1971: 366) in both wars, support dropped approximately 15 percentage points ‘every time American casualties increased by a factor of 10'. The major implication of such a pattern is that the public becomes increasingly hardened to war's costs: ‘They [i.e., the people] are sensitive to relatively small losses in the early stages, but only to large losses in later stages' (Mueller, 1971: 367). Interestingly enough, this study is classified among those that claim an unmediated effect of situational factors on support, while one of its major conclusions is that ‘events do not seem to have set up major perturbations in these trends' (Mueller, 1971: 365). Mueller stresses that in both the Korean and the Vietnam war events on the ground - for example, a military victory or defeat - did not move up or down the same way casualties did.26 In a further study of the Gulf War of 1991, Mueller (1994: esp. 76-78) found the same logarithmic pattern in questions about hypothetical US casualty levels in the Gulf War. Larson has also confirmed the logarithmic nature of the downward trend in support using only battle deaths, instead of overall casualties, for several wars, including Korea and Vietnam (Larson, 1996b and especially Figure 2.2, p. 9).
Mueller's study in 1971 (further enlarged in 1973) has been credited for its seminal and consequential effects well beyond the academic community. The study 'fixed in the public mind the idea that support for Vietnam buckled as the body-bag toll mounted, and this gradually hardened into the conventional wisdom that the public is reflexively casualty phobic' (Gelpi, Feaver and Reifler, 2009: 9). Following this study, the discussion in the literature has proceeded along two different paths. A first group of studies has discussed, criticized and qualified the functional nature of the relationship between casualties and support, as initially suggested by Mueller. A second group has included these cost considerations in a more general model of cost-benefit calculations, criticizing the inexorability implicit in Mueller's argument about the downward trend in support. While the first group of studies does not deny the inexorability logic, but only contests its pace, the second line of research argues that, under certain conditions, support does not decrease or the decline is contrasted by other considerations.
Among the first kind of studies, we select three for discussion. One is a study by Gartner and Segura (1997, 1998) that qualified Mueller's conclusion about Vietnam. Gartner and Segura criticize Mueller's conclusions based on the log of cumulative casualties because methodologically flawed. The log of casualties ' homogenizes conflicts with very different patterns of casualty accumulation, and it underestimates the importance of turning points, decisive events, and exogenous shocks to opinion' (Gartner and Segura, 1998: 280). According to them, casualties affect support differently, depending on how the war is fought; 'marginal casualties are a better predictor of opposition during periods of escalation or continuous fighting, but cumulative casualties are more likely to serve that purpose during periods of de-escalation' (Gartner and Segura, 1998: 286). Given the different pattern of marginal accumulation of casualties in Korea and Vietnam, much faster at the beginning in Korea and at the end in Vietnam, they find that marginal casualties were more important in Vietnam and cumulative casualties were more important in Korea. More recently, Gartner and Segura (2005) have framed this hypothesis in a wider theory of public opinion deliberation about war, arguing that the evaluation of costs is constantly updated based on three different measures of casualties: recent casualties, the trend in casualty cumulation and the casualty context.
A second study, by Gelpi, Feaver and Reifler (2009) criticizes Muller's functional relationship as misspecified. Analyzing Mueller's data again, they found that casualties had a differential impact on support depending on whether the war was successful or not. Although the paucity of data available does not allow for firm and robust conclusions, both multivariate analysis and cross-tabulations point to the fact that casualties have a greater depressing impact on support when the war goes badly than when it is successful, in both Korea and Vietnam.
A third, less-known and never published study by Garnham (1994), is the only comparative one in replicating Mueller's approach in France during the Indochina (1946-1954) and Algerian (1954-1962) wars and in Israel during the Lebanon war (1982-1985). Although complicated by lack of comparable questions, by imperfect measurement of the main independent variable, by a limited number of cases, as well as by the different nature of the military force, namely an all-volunteer military force employed in Indochina and Algeria, Garnham found a strong correlation with casualties in both Indochina and Algeria, also controlling for other factors, such as duration and financial costs. However, Garnham also found that 'political cultures seem to make a difference' (Garnham, 1994: 21). Israelis had a higher cost tolerance than did French public opinion, and Garnham also stresses that events and stakes matter. 'If a war's stakes are small (Indochina) or unachievable (Lebanon), then the public is prone to become frustrated and impatient. Under these conditions, any level of continued casualties is likely to undermine public support' (Garnham, 1994: 21-22).
While the studies discussed so far focus exclusively on the bivariate relationship between casualties and support, another line of research points to casualties as the cost element in a more general cost-benefit model. In this vein, Larson was the first to offer a fully specified cost- benefit model. According to Larson, a person will be supportive until his utility for the military operation is above a minimal threshold (in formal ways, Ui >Umini) and this utility, Ui, depends on the ratio between the value of the operation (weighted by its probability of success) and its anticipated or actual costs. In Larson's model, each of the three parameters of the model depends on the messages and cues coming from political leaders and the media. For this reason, Larson combines this micro-level model with a 'social process model', heavily influenced by Zaller's RAS model (Larson, 2005: 25-27) and he concludes (Larson, 1996, 1996a, 2000; Larson and Savych, 2005) that the cost-benefit model seems well-corroborated by the empirical facts. Gartner and Segura (2005) have suggested a comparable model based on two parameters, the value of the war aims and the costs expected to be incurred to achieve them. The debate here is about the relative role of these variables in the model, with authors taking a different position on the issue. On the one hand, Larson suggests that the value of the operation is the crucial factor in explaining support, while Feaver and Gelpi stress the prospect of success. Both positions, however, agree that the relative weight of these three factors - probability of success, values of the operation and costs - might change over time, within each operation, as well as across different operational situations. And their diverging conclusions are based on slightly different evidential bases. Larson (Larson and Savych, 2005) grounds his conclusion on a systematic comparison of different cases between the 1950s and the early 2000s, while Feaver and Gelpi (2009) rest on evidence marshaled using experimental data on hypothetical situations and from the Iraq war.
Larson, in a set of studies spanning the last decade, has studied eleven major cases of military operations27 and does not find a radical change in the overall level of tolerance for casualties over time - a conclusion already reached by O'Neal et al. (1996), comparing a larger set of cases from the early 1950s and 1960s up to the late 1980s - while variations in the interests at stake and the values pursued in the military operations seem to make a lot of difference in explaining support. Comparing the operations of the 1990s with the most recent Afghanistan, Iraq and Terrorism operations (that he labels GWOT = Global War on Terrorism), Larson and Savych (2005: 215-216) conclude that the crucial difference is in the perceived stakes, esteemed to be much higher in the most recent GWOT operations compared to the peace-keeping operations of the 1990s. And this difference explains why both the prospects of success and the costs of the operation are relatively less important. Larson attributes a much greater role to stakes in upholding public support and in contrasting the negative impact of human costs on support than Feaver and Gelpi do. Larson and Savych claim that 'in virtually every case, our respondent-level modeling suggested that the perceived stakes were a more important predictor of support than beliefs about the prospects for success' (Larson and Savych, 2005: 225-226).
Feaver and Gelpi reach a different conclusion using a different database to Larson. Their analysis of the historical cases (and they examine six of them, including the most recent Iraq war) leads them to conclude that they 'cannot answer this question [i.e., "what aspects of the context are most important?"] definitely with the aggregate survey data most of the literature relies upon' (Feaver and Gelpi, 2009: 65). They instead ground their conclusion on a set of experiments run in the fall of 2004 and on a trend analysis of several surveys conducted between October 2003 and October 2004. Based on their data, Feaver and Gelpi conclude that expectation of success is as strong and often even stronger than costs in determining support for the war, both in hypothetical situations and in the very concrete Iraqi context.28 Both Larson, and Feaver and Gelpi, are careful, however, to stress that their model should be embedded in a wider social context, in which most of the information and cues come from the elite and the public political discourse, but still they both stress that, given this external context, most of the public29 makes independent rational calculations in deciding its positions and that these calculations also sound rational to an external observer.
Another stream of research, however, challenges this very assumption, arguing that most of what people base their assessment on is socially 'mediated' by factors such as ideology, the media and the political elites. This 'mediated' model of public support for the use of force can be traced back to the landmark study of Zaller (1991, 1992) on the nature of public opinion. Using a simplified version of the McGuire Model of attitude change, Zaller suggested that citizens' opinions are crucially dependent on the nature of the information provided by the elite and the media (whether balanced or unbalanced, conflictual or consensual), and the ways these elite-cues shape opinions depends on individual level variables such as attention to political discourse (in turn, a function of political awareness) and individual political predispositions. Based on the interaction between these three elements, Zaller was able to show what explains attitude formation and attitude change on political issues. In particular, Zaller focused on two possible situations generated by the interaction of different media and discourse environments with political predispositions and differential levels of political awareness in the general population: the mainstream effect and the polarization effect. They vary in the nature of the elite message context. In the mainstream situation, elites have a consensual policy position, and as a consequence, communications are all in the same direction. In such a context, we would expect an increasing level of support among the public as political awareness increases, since those more aware are also more receptive to elite cues and are then swayed by the direction the messages overwhelmingly have. On the contrary, in the polarization situation, elites are divided along partisan lines, and this is reflected in the messages they issue. 'Thus, in the case of an evenly divided partisan elite and a balanced flow of partisan communications, the effect of political awareness is to promote the polarization of attitude reports as more aware liberals gravitate more reliably to the liberal position and more aware conservatives gravitate more reliably to the conservative position' (Zaller, 1992: 101-102). For our purpose, it is interesting that Zaller applied his model of attitude change to the case of the Vietnam war. Using the data from the SRC between 1964 and 1966, and examining the patter of media messages between 1964 and 1970, Zaller concluded that 'public attitudes on major issues [such as Vietnam] change in response to changes in the relative intensities of competing streams of political communication, as filtered through the reception-acceptance model' (Zaller, 1992: 190). Zaller's conclusion about Vietnam departs in two ways from the cost-benefit models and their variation examined so far. First, 'public attitudes ... are a response to the relative intensity of competing political communications' (Zaller, 1992: 210) and not to actual events on the ground. Second, contrary to the assumption of a rational calculator, explicitly supported by most of the studies examined in this section, Zaller suggests that 'survey responses are a function of immediately accessible' considerations' where the flow of information in elite discourse determines which considerations are salient (Zaller, 1992: 36), implying that the crucial determinant of public attitudes is the nature and balance flow of messages coming from the political discussion, as mediated by levels of political awareness and political ideology.
Following this line of argument and exploiting the recent developments of research on framing (see Druckman), a stream of studies in the 2000s has emphasized the importance of the way issues are framed in the public domain, especially through the media (Entman, 2004; Berinsky and Kinder, 2006; Gaines et al., 2007), as well as the role of more traditional political variables, such as partisan cues, in explaining support and opposition to military operations (Berinsky, 2009). This literature has made several important points that qualify, and sometimes openly revert, the conclusions reached by the cost-benefit model of support. A first line of criticism takes a stand against the view that the public is able to make appropriate judgements on military events as they unfold on the ground. Criticisms are here of two, often interrelated, kinds: people lack the appropriate or correct factual information on which their judgements could be based (Berinsky, 2007); and even if they are provided with such information, their posterior update of their prior beliefs is far from perfect.
Several studies (e.g., Berinsky, 2007; Gaines et al., 2007; Kull, Ramsay and Lewis, 2003-2004) have shown that people have only a scant amount of policy-specific information on which to rely in making their own judgements. More specific studies have shown that, for example, people vary widely in their estimation of the actual number of casualties in Iraq (Berinsky, 2007: 979-981). Moreover, they also clearly misperceived several facts about the Iraq war, such as the alleged links between Al Qaeda and Iraq, the discovery of weapons of mass destruction (WMDs) and general world public opinion orientation toward the US intervention in Iraq (Kull, Ramsay and Lewis, 2003-2004). Berinsky (2007) found that in World War II as well as the Iraq war most of the people had only a vague idea of what was happening on the ground in the different battlegrounds in Europe and the Pacific during the 1941-1945 period or in Iraq in 2003-2004. Similarly, Gaines et al. (2007) found that factual beliefs count less than the way they are interpreted in explaining policy preferences. And, going round in the opposite direction of the cost-benefit model, they also conclude that 'those who acquire the most information about a policy and its consequence are also the ones most likely to rationalize their existing opinions' (Gaines et al., 2007: 972). Other studies (e.g., Berinsky and Kinder, 2006) have shown that 'ordinary citizens' understanding of politics depends in systematic and intelligible ways on how information is presented to them' (Berinsky and Kinder, 2006: 654). Berinsky and Kinder (2006) report, for example, an experiment in which the way the Kosovo crisis was framed affected the opinion about the intervention. When the Kosovo 1999 intervention was framed in a more positive way (the humanitarian story), people were more ready to support the operation than when it was painted in a gloomier fashion (the 'risk to America' story). Several studies (Boettcher and Cobb, 2006; Gartner, 2008; Myers and Hayes, 2010) have also pointed to the impact of different framings of the casualties on support, emphasizing the point that casualties can affect support in differential ways, depending on how they are framed.
A second element emerging from these studies is that partisan cues are far more important than events in explaining variation in support. Following Zaller, several of these studies suggest that 'elite cue theory' is a better predictor of the pattern of opinion in wars as diverse as World War II and the Iraq war. Support and opposition in such a polarized political environment as the one in which the Iraq war was carried out, produced a typical pattern of polarization among Democrats and Republicans, irrespective of what was happening on the ground.
There are, however, elements that point to a possible reconciliation among the two perspectives. In particular, some recent studies (see Perla, 2005, 2005a, 2011 in press) show that framing effects related to the narrative of the conflict's objectives affect public support for the operation. When the narrative of the conflict stress a Foreign Policy Restraint (FPR) objective, support for the operation is higher than when the narrative is stressing an Internal Political Change (IPC) objective. These studies not only indirectly confirm the importance of the mission's objective on support, but also they suggest a differential way in which people calculate costs and benefits. Different students (e.g., Nincic, 1997; Perla, 2010) have stressed the underlying link between the Principal Policy
Objective theory and the Prospect theory of decision-making. Nincic was the first to suggest that the nature of the objectives could be related to the differential evaluation of gains and losses, as predicted by the Prospect theory. Nincic found that in 18 military operations between 1950 and 1994, the public was 'more willing to reward the president for foreign policy actions intended to preserve or restore a situation that had already been attained than for those meant to pursue a new gain or to create a new outcome' (Nincic, 1997: 115). More recently, Perla (2005, 2005a, 2010) has articulated the Framing Theory of Policy Objectives' that explains the differential level of support for FPR and ICP missions on the basis of prospect theory.