Elites, Events and British Support for the War in Afghanistan

Transcription

1 Elites, Events and British Support for the War in Afghanistan Douglas Kriner Assistant Professor of Political Science Boston University Graham Wilson Professor of Political Science Boston University Paper prepared for presentation at the annual meeting of the American Political Science Association, Washington, D.C., September 2-5, 2010.

2 Abstract Studies of the dynamics driving public support for war have largely split into two camps, one emphasizing the importance of unmediated conflict events and the other of elite discourse. To test the explanatory power of each perspective, we marshal a wide range of aggregate, individual-level and experimental survey data concerning British support for the war in Afghanistan. We find little support for event-response theories. With respect to elite theories, the data is more mixed. At the individual level, we find modest evidence consistent with elite opinion leadership; politically aware partisans are more supportive of the war in Afghanistan than are their peers who are less attuned to politics and therefore less exposed to elite cues. Similarly, our experiments show that even modest elite cues continue to influence support for war among some segments of the public. However, the aggregate level of war support is strikingly low throughout most of the period, despite the strong elite consensus in support of the war effort. Thus, elite discourse does appear to influence opinion, but only at the margins around a surprisingly low mean level of public support. 1

3 Few topics in political science have received as intensive and sustained attention over the past thirty years-five years as the dynamics governing popular support for war. At least since Kant (1795), theorists and political analysts alike have posited that public opinion in a democratic republic provides one of the strongest checks on politicians impulse to adopt aggressive and costly military policies. Many contemporary theories of the influence of domestic politics on military policymaking whether they emphasize domestic institutional constraints on foreign policy executives (Morgan and Campbell 1991; Maoz and Russett 1993; Siverson 1995; Clark 2000; Howell and Pevehouse 2007; Kriner 2010), the domestic audience costs that leaders face when they fail to follow through on a threat to use force (Fearon 1994; Smith 1998), or the ultimate threat of removal from office at the ballot box (Bueno de Mesquita and Lalman 1992; Bueno de Mesquita and Siverson 1995; Bueno de Mesquita, Morrow et al. 1999; Bueno de Mesquita, Morrow et al. 2003) rely on public opinion, either explicitly or implicitly, as one of the most important mechanisms through which political pressure is brought to bear on decision-makers to moderate their conduct of military affairs. Two main schools of thought dominate the wartime opinion literature. The first emphasizes the role of unmediated conflict events in driving fluctuations in popular support for war. Citizens are held to update their cost-benefit calculations in response to casualties and other battlefield events when deciding whether to back an ongoing war. 1 The second strand stresses the critical importance of the cues sent by political elites in shaping popular attitudes toward military conflict. While a growing number of studies has begun to probe the conditional nature of elite influence on war support (Berinsky 2007, 2009; Groeling and Baum 2008; Baum and Groeling 2009; Kriner and Howell 2010), one of the most important and consistent empirical 1 Although the military commonly uses the word casualty to refer to both fatal and non-fatal casualties, following the standard usage in political science (e.g. Gartner and Segura 1998, Gelpi, Feaver and Reifler 2009), this paper uses the term to refer only to military fatalities. 2

4 findings of this literature is that elite consensus bolsters public support for war, while elite dissension causes the erosion of public support for a military conflict. The proposed causal processes of both of these theories of wartime opinion formation and change should hold generally in any democratic nation with a free press and an institutionalized opposition party. However, most empirical assessments of these theoretical perspectives have focused on the American case. There are reasons for this, not the least of which being the paucity of comprehensive opinion data in many other contexts. However, because of this lacuna, scholars have lost an important opportunity to assess theories of opinion formation in different political and social contexts. This paper marshals a wide range of public opinion data on British support for the war in Afghanistan. At both the aggregate and individual levels, we look for evidence consistent with theories of elite opinion leadership. We also assess whether patterns of British war support in the aggregate match the expectations generated by an event-response perspective emphasizing the importance of casualties. Finally, we complement this analysis of observational data with an original survey experiment to assess the influence of different elite cues on different partisan publics support for continuing the British military commitment in Afghanistan. The paper proceeds in five parts. Part 1 reviews the relevant literatures emphasizing conflict events and elite discourse as the principal drivers of war support among the general public. Part 2 applies these perspectives to the British case in Afghanistan and develops precise theoretical expectations for each theory. Part 3 assesses these competing expectations against all available aggregate-level opinion data on British support for the war. Part 4 analyzes individuallevel survey data from the British Election Study to test further empirical expectations derived from the elite opinion leadership hypothesis. Specifically, it examines whether respondents more 3

5 attuned to politics, and therefore more receptive to elite cues, are more supportive of the war than their peers with otherwise identical characteristics who are less attentive to political happenings. Finally, part 5 presents the results of an original survey experiment conducted with a representative sample of the British public in August 2010 to assess the influence of elite cues about the proper course in Afghanistan on support for war in the new era of the Coalition government. Events vs. Elites as Drivers of War Support Many scholars conceptualize the wartime opinion formation process as the result of a cost-benefit calculation akin to those made by policymakers. Battlefield events inform both citizens estimate of the current costs of a military action and of those that likely must be paid to achieve the mission s objectives (Gelpi, Feaver and Reifler 2009, Eichenberg 2005). Since Mueller s seminal work (1973), much, though not all, of this literature has focused on the importance of combat casualties in driving opinion change. According to Scott Gartner (2008, 99), casualties represent the primary information individuals use to evaluate war, assess past costs, estimate future costs, and formulate their positions. In a host of major American military actions from the Korean, Vietnam and Iraq wars to smaller engagements including the peacekeeping mission in Somalia and the 1991 Persian Gulf War, studies have shown very strong inverse correlations between the log of cumulative American casualties and public support for the military action (Mueller 1973, Larson 1996, Mueller 1994, Voeten and Brewer 2006, Eichenberg, Stoll and Lebo 2006). Other studies have emphasized the importance of short-term shocks in casualties to understanding fluctuations in war support (Gartner and Segura 1998), while still others have shown that differences in local 4

6 casualty rates significantly affects spatial variance in war support across the country (Gartner, Segura and Wilkening 1997, Kriner and Shen 2010). According to this event-response literature, battlefield events and, perhaps most importantly because of their prominence, combat casualties drive war support by shaping citizens cost-benefit calculations. Small casualty totals and other signs of tangible military success on the ground bolster public support for war; by contrast, mounting casualties and military setbacks in theater tilt the scale in the opposite direction and cause war support to wane. Other scholars emphasize instead the importance of elite cues to driving patterns in popular support for war. Most such theories begin with the venerable Almond-Lippmann consensus (Almond 1950, Lippmann 1992): the public is largely a tabula rasa on foreign affairs. Whereas in domestic policy matters most citizens have some background information or can at least rely on pocketbook and other heuristics to guide their opinions, public knowledge of and attention to foreign affairs is extraordinarily limited. As a result, citizens look to political elites both for information on a war s progress and for an interpretive framework to guide their processing of such events. One of the most important theoretical and consistent empirical results to emerge from this elite-centered literature is that elite consensus bolsters support for the use of force while elite dissension undermines it. For example, in an analysis of dozens of rally around the flag events from the 1940s through 1980s, Brody (1991) finds that when Congress supported the president s actions, the public rallied behind the Commander in Chief. However, when opposition leaders in the legislature did not back the president s handling of a crisis, no such rally materialized (see also Lian and Oneal 1991). Zaller s RAS model of information processing leads to a similar prediction for war support more generally. Elite consensus produces a one-sided information 5

7 flow in which identifiers of both parties receive cues from trusted elites supporting the war effort. This, in turn, produces a mainstream pattern in which support increases among members of both the president s party and the opposition party as respondents political awareness increases. By contrast, elite dissension leads to a two-sided information flow, as the media reports conflicting cues from party elites concerning war conduct. This, in turn, leads to a polarization pattern in which war opinion diverges between identifiers of the major parties as political awareness increases. Citizens whose co-partisan elites vocally support the war become more supportive as political awareness increases, while those whose co-partisan elites oppose it become less so. Zaller finds evidence of both patterns within the Vietnam War. During the period of bipartisan consensus before the Tet offensive, war support remained high and little gap emerged between party identifiers. By contrast, after Tet when elites began to diverge in their assessments of the war, a polarization pattern began to emerge. Recently, Berinsky (2007, 2009) has adapted this perspective and applied it to the very different contexts of World War II and the war in Iraq. Exploiting variance in party positions in the lead up to World War II, Berinsky shows that a significant partisan gap emerged in support for American involvement in World War II when elites were divided, but that this gap disappeared when the Republicans nominated the interventionist candidate Wendell Wilkie in Berinsky also argues that this elite consensus explains the high levels of public support throughout the war, despite the massive number of American casualties sustained. In Iraq, Berinsky finds strong evidence of partisan polarization in war support even before Democrats took a strong stand against the war. The lack of consensus and vocal Democratic elite support for the war, Berinsky shows, allowed Democratic identifiers to define their opinions in 6

8 apposition to those of President Bush, whose support among Democrats was already strikingly low (see also Kriner and Howell 2010). Across multiple contexts, then, this literature argues that the cues transmitted by political elites guide the opinions of a public ill-equipped to judge the merits of foreign policy events independently. Theoretical Expectations Events According to event-response theories, public support for war should largely follow the development of events on the ground. When military progress is being made and the costs of the mission are low, public support should remain high and stable. However, as the military meets setbacks and the costs of a conflict increase, support should begin to wane. In this paper, we focus exclusively on combat casualties, which provide the most-often used metric of conditions on the ground that might affect popular support for war. If casualties drive public opinion, then support for the UK mission in Afghanistan should remain high until spikes in the casualty rate or the gradual accumulation of battle deaths cause it to erode. Figure 1 presents both the number of British casualties sustained in Afghanistan in each month as well as the cumulative death total for British forces throughout the nine-year course of the war. Perhaps the most striking aspect of the figure is the very low number of British casualties in the war prior to the fall of The first real spikes in British casualties occurred in August and September of that year. This corresponds with the beginning of Operation Mountain Fury, which was designed to clear Taliban insurgents out of their strongholds in the eastern provinces of the country. After this turning point, monthly casualty counts are 7

9 considerably larger, though there is significant monthly variance; by extension, the cumulative death toll begins to rise during this period, and it accelerates with the heavy fighting of 2008 and The public repatriation ceremonies in Wootton Bassett, which received extensive media coverage, also began in 2007; these ceremonies heightened media interest in and attention to the war and transmitted to the public highly visible images of flag-draped caskets in hearses that served as a stark reminder of the war s human toll. 2 << Figure 1 About Here >> If similar relationships hold between casualties and British support for the war in Afghanistan as have been observed in multiple American wars, then at the aggregate level support for the war should remain high through This support, however, should begin to erode in the war s later stages as both cumulative and marginal British casualties increase. Elites By contrast, if elite opinion leadership is the driving force behind patterns of war support, then the presence or absence of elite consensus behind the war effort should govern the expected level of public approval for the Afghan war at any moment in time. This leads to a very different expected temporal pattern of British war support than that posited by event-based theories emphasizing the importance of combat casualties. From its start in October, 2001 onwards, the British political elite was united in support of the Afghan War. Leaders of all the major political parties were clear in their support for military action against Al Qaida and the Taliban regime in Afghanistan that was sheltering it. It is perhaps worth quoting from Hansard for the 7 th of October when the Prime Minister 2 See, for example, Steven Morris, Bodies of British Soldiers Killed by Afghan Policemen Brought Home. The Guardian, November 10,

10 announced that military action had begun. Mr. Blair, the Prime Minister and Labour Party leader, argued that British and American armed forces with the support of other allies had started their attacks and argued that the 9-11 attacks on the United States was not just an attack on American but on civilized values everywhere. Britain would be resolute until the victory was secured that would mark not the victory of revenge but the victory of justice over the evil of terrorism. (Column 811) Mr. Ian Duncan Smith, Leader of the Conservative Party, immediately offered his support. No one should doubt the determination of the British people to see this through to a successful conclusion. Our future security and well being require no less. (Column 815) The Leader of the Liberal Democrats, Mr. Charles Kennedy, was also clear that this was a necessary and just war..may I fully associate Liberal Democratic Members with the support and concern for our armed forces that have been so properly expressed at this most difficult and dangerous of times?...the military strikes of the last 24 hours are sad. Indeed, we all consider them tragic but they are none the less inevitable.we are correct to pursue military action but unlike the terrorists, we will continue to display mercy. (Column 817.) This united front was unsurprising. Britons had been genuinely shocked and horrified by the 9-11 attacks. Indeed, sixty seven of the victims were British, making 9-11 one of the major losses of British lives to terrorism. Important traditions of British foreign policy also supported active British military support for the United States. First, the American alliance had nearly always been the primary foreign policy priority of the UK. Second, the UK had long been an active proponent of collective security and NATO in particular; in responding to 9-11, for the first time in its history, NATO invoked the an attack on one of us is an attack on all of us clause in its founding treaty. The Afghan War therefore began with the political leaders of the nation united, even though they also recognized that the war could be lengthy. 9

11 Equally importantly, this unanimity was not limited to the immediate aftermath of the mission s commencement, but has continued through multiple election campaigns to the present day. In 2010, the Labour Manifesto argued that if Afghanistan fell to the Taliban, Al Qaeda would regroup and Pakistan s stability and our national security would be threatened. It also emphasized that Labour had met every request for extra equipment for Afghanistan. Similarly, the Conservative manifesto emphasized the Party s commitment to the transatlantic alliances which is as necessary in 2010 as ever in order to make progress in Afghanistan. In 2009, Liam Fox, the Conservative Party spokesman on foreign policy issues, argued that We are in Afghanistan out of necessity, not out of choice. (28 September 2009) The Liberal Democratic manifesto also committed the party to be critical supporters of the Afghanistan mission. The military surge must be accompanied by a strategy to secure a more legitimate government tackle corruption and win over moderate elements The party leaders expressed similar sentiments during the 2010 campaign debates. Thus the British political elite united behind the Afghan War early, and this elite consensus has remained unshaken over the intervening years. In stark contrast to the demonstrations and backbench revolts that helped shape British support for the Iraq War (Clarke, Sanders, Stewart, and Whiteley 2009), significant elite opposition never materialized against the war in Afghanistan. At the aggregate level, theories emphasizing the importance of elite consensus versus elite dissension would therefore lead us to expect strong, stable levels of British public support for the Afghan war throughout its course. At the individual level, we should also expect to find evidence consistent with Zaller s mainstream pattern. That is, as political awareness increases, respondents of all three major parties should become more likely to support the war because they 10

12 are increasingly exposed to elite discourse and cues championing the need to stay the course in Afghanistan. Aggregate-level opinion data Because no single polling outlet has asked the same question or questions of the British public regularly throughout the course of the war in Afghanistan, it is necessary to use data from multiple polls to gain the most accurate picture possible of patterns in British war support over time. Figure 2 pools all relevant polls querying British support for the Afghan war from October 2001 to June This and all subsequent figures also present a LOWESS-smoothed trend to illustrate the general pattern in changing public support over time. << Figure 2 About Here >> In the immediate aftermath of the September 11 th terrorist attacks and the initial bombardment of Taliban positions in Afghanistan, British support for the mission hovered in the mid-60s in most polls. This relatively high level of initial support is likely a function both of public horror at the recent attacks against the World Trade Center and Pentagon and the almost unanimous support of political leaders for British participation in the war effort. For example, in no poll did a majority of Britons ever support for the use of force in the Suez crisis, which was sharply criticized by Labour party leaders (Epstein 1960). Unfortunately, the paucity of polling data between the flurry of polls in the immediate wake of the war s initiation and the fall of 2006 makes it difficult to say precisely when and how quickly support for the war in Britain fell. However, an August 2002 poll suggests that, less than a year into the war, support for the 11

13 conflict had already dropped significantly. 3 And by mid-2006 British support for the war had fallen precipitously to well below 40% in most polls. 4 This pattern stands in stark contrast to trends in war support in the United States. As shown in Figure 3, the war in Afghanistan enjoyed almost universal support among Americans in its earliest stages. Levels of support approaching 90% are unsustainable for any conflict, and the percentage of Americans supporting the war did decrease throughout the presidency of George W. Bush. However, even as Bush prepared to vacate the Oval Office, the war in Afghanistan continued to enjoy the support of a clear majority of the American public. Only in 2009 and 2010 did a significant number of polls begin showing a majority of Americans opposing the war effort. << Figure 3 About Here >> These early trends in aggregate British war support do not fit neatly with expectations generated by either event- or elite-driven theories of wartime opinion formation. Casualties cannot explain the downward arc of British war support in these early years, because support had eroded to very low levels before British casualties began to mount. Looking beyond casualties, it is also hard to argue that events on the ground caused the strikingly low levels of popular support for the war. While some objectives remained unmet, coalition forces enjoyed 3 Unfortunately, the August 2002 ICM question wording is very different from others and the specific invocation of civilian casualties in Afghanistan may be responsible for the low level of support. The question asked: The bombing in Afghanistan had the effect of replacing the Taleban Government and weakened the Al Qaeda terrorist organisation somewhat but also resulted in a number of civilian deaths. Some people say that these civilian deaths, although regretted, were a price worth paying to defeat the Taleban and weaken Al Qaeda. Do you agree or disagree? Only 43% agreed that the civilian casualties incurred were worth it. While this wording certainly complicates the issue, it is worth noting that the question does imply that the coalition had largely succeeded in defeating the Taliban and weakening Al Qaeda as a result of the military action in Afghanistan. Given the literature emphasizing the importance of assessments of success in driving support for war (e.g. Gelpi, Feaver and Reifler 2009, Eichenberg 2005, Kull and Destler 1999), we might have expected this wording to have yielded a large percentage of respondents approving of the war. 4 For example, in a July 2007 ICM poll a poll taken when the UK had suffered fewer than 15 total casualties in Afghanistan only 23% of Britons responded that the UK military presence in Afghanistan was helping to improve the situation there. 12

14 considerable success in the war s early stage, and proponents of the conflict could point to a number of achievements that illustrated the allies progress in the war on terror. In short, it is difficult to argue that raw assessments of conditions on the ground alone could produce such low levels of support; indeed, a strong majority of Americans continued to support the war during this period. Theories of elite opinion leadership fare no better in explaining early patterns in British public support for the war in Afghanistan. As mentioned previously, leaders of all three major parties forcefully backed the war at its outset and continued to stand by it throughout the campaign. This, past theory suggests, should lead to strong and stable public support for a military conflict. The disjunction between this expectation and the very low levels to which British war support sank in the mid 2000s could not be more striking. The beginning of Operation Mountain Fury and the resulting spike in coalition casualties twice as many British soldiers died in Afghanistan in August and September of 2006 as had died in the mission to that point prompted renewed interest on the part of pollsters in the public s support for war. Figure 4 presents the aggregate support data from all polls taken from 2006 to 2010, and it also breaks it down into specific subsets of questions asking respondents generally whether they supported or opposed the war in Afghanistan, whether they believed the war was succeeding/would succeed, and whether they supported an expeditious withdrawal of British forces. While there is considerable variance across polls, which undoubtedly reflects differences in question wordings and sometimes long temporal gaps between polls, along most of the dimensions the data is at least suggestive of a slight increase in war support over time. Beginning in the upper left-hand corner of Figure 4, the LOWESS smoothed trend on the scatter plot including all relevant polls suggests a gradual increase over time (albeit with high variance 13

15 about the trend) from an average level of support in the low 30s in to just over 40% by June The pattern in the general approval questions in the upper right hand corner of Figure 4 is perhaps more ambiguous, given the number of polls in 2009 and 2010 showing very low levels of public support. However, the questions measuring public assessments of the mission s prospects for success shows an uneven upward trend over time. Finally, the percentage of Britons opposing an expeditious withdrawal from Afghanistan also appears to have increased, on average, since 2007, although there is still considerable variance across polls. << Figure 4 About Here >> Thus, at the aggregate level British war support from also fails to accord with theoretical expectations derived from event-response theories emphasizing the importance of combat casualties to citizens cost-benefit calculations. Recall from Figure 1 that cumulative casualties increased sharply during this period from less than 20 in June 2006 to more than 300 in July Marginal casualties also increased during this period, as Taliban resistance intensified and coalition strategy shifted to engage insurgents more aggressively in the hopes of building greater stability in troubled parts of the country. The surprisingly low level of British war support before the fighting escalated and casualties increased may blunt the expected downward effect of rapidly increasing casualties on war support, simply because support for the war did not have too much room to fall further. However, in sharp contrast with theoretical expectations from event-response theories, the aggregate data suggests that British support for the Afghan actually increased, albeit gradually and modestly, even as the war s human and financial costs mounted. Elite opinion leadership may be able to explain, at least in part, this somewhat surprising trend. First uncertainty over whether Prime Minister Gordon Brown would call an early election 14

16 and then the lead-up to the 2010 general election campaign itself significantly increased the number of elite cues supporting the war transmitted to the public. Leaders of all three parties reaffirmed their commitment to the British military commitment in Afghanistan throughout the campaign and pre-campaign period. This, in turn, provided their partisan identifiers with important cues on the war, and it may also have helped distinguish the war in Afghanistan, which still enjoyed strong support from political leaders of all stripes in London, from the highly unpopular Iraq War, which had long cast a pallor on popular evaluations of the Afghan war. While the suggestive, slight upward trend in British war support since 2006 is consistent with expectations derived from the elite opinion leadership perspective, it is important to remember that a clear majority of Britons continue to oppose the war in Afghanistan and to support the expeditious withdrawal of British troops from the region. Thus, while the aggregate data may be partially consistent with elite-centered theories of opinion leadership, even a strong cross-partisan elite consensus failed to produce majority support for British involvement in the war. The data suggests that elite cues may have influenced popular war support, but only at the margins. Individual-level opinion data Most event-response theories do not lead to clear predictions about variance in support for war among individuals. 5 For example, the number of casualties in a military venture is posited to provide information about a war s costs that all citizens can incorporate into their costbenefit calculations. For some, this information may be enough to tip the balance and lead them 5 An important exception is the growing literature examining the influence of local casualties on support for war and patterns of political behavior (Gartner, Segura and Wilkening 1997, Karol and Miguel 2007, Grose and Oppenheimer 2007, Kriner and Shen 2010). 15

17 to oppose the war, while for others the benefits will continue to outweigh the costs. However, all citizens are held to process the new information in essentially the same way. Elite theories, by contrast, do offer precise expectations about differences in opinions across citizens that can be tested empirically with individual-level survey data. A lengthy literature in social psychology demonstrates that individuals are most responsive to cues sent by trusted political elites with whom they share basic political predispositions (e.g. Lupia 1994, Sniderman, Brody and Tetlock 1991). Thus, elite-theories predict that the opinions of different segments of the mass public may diverge substantially if their co-partisan elites take different positions on a war, even though all citizens receive the same information about casualties and major conflict events in theater. For example, Labour party leaders sharply criticized the Eden government s use of force in the Suez crisis in 1956, and public opinion largely split along partisan fault lines. Whereas 68% of Conservative identifiers supported the use of force in a November 1956 poll, only 16% of Labour party identifiers did so (Epstein 1960, 221). However, if major party elites share the same basic position on an issue, we would not expect to see major opinion gaps between partisan mass publics. Second, individuals who are the most attentive to politics should be the most likely to converge to the policy preferences articulated by their party s leaders. For example, the first tenet of Zaller s RAS model (1992) is that for an elite cue to influence opinion the hearer must receive the message. The likelihood of receiving policy cues transmitted by political elites is a function of the degree of attention a respondent pays to politics. Those who are more attentive to politics, thus, should mirror the patterns of elite discourse more than those who are inattentive. In the current context, elites from all three major parties, Labour, Conservative and Liberal Democrat, throughout the war have transmitted cues to their co-partisans encouraging 16

18 them to support the war and staying the course in Afghanistan. Thus, we should not expect to see large partisan divergences in war support as, for example, have been documented in the American context with respect to the war in Iraq (Jacobson 2006). Moreover, a respondent s probability of supporting the war should increase with his or her level of attention to politics. The end result should be what Zaller termed a mainstream pattern in which politically attentive members of all three major parties are more supportive of the war in Afghanistan than their less politically attentive peers. To test the predictions offered by elite opinion leadership theories, the analysis turns to individual-level survey data drawn from the Continuous Monitoring Survey (CMS) of the British Election Study (BES). The BES surveys comprise a nationally representative sample of Britons. In the cumulative file, 24% of respondents identified with the Conservatives, 35% with Labour, 9% with the Liberal Democrats, 8% with other parties, and 23% affiliated with no party. Two questions afford insight into patterns of support for the war in Afghanistan at the individual level. The first question, which was included on CMS surveys from June to September of 2008, asked respondents for their level of agreement or disagreement with the following statement: All British troops in Afghanistan should be brought home immediately. The second question, included on the January-April 2010 CMS surveys, asked respondents whether they approved or disapproved of Britain s involvement in the war against the Taliban in Afghanistan. The ordered logit regressions that follow model war support on both dimensions as a function of respondents partisanship, attention to politics, the interaction of the two, and several demographic control variables. 6 The key expectation drawn from elite theories of opinion 6 Specific factual knowledge questions, such as those used in many other studies to create scales of respondents level of political information (e.g. Delli Carpini and Keeter 1997, Zaller 1992), were not included in the BES CMS surveys. As a result, the analysis relies instead on a self-reported political attention measure. This item has a mean of 6.2 and a standard deviation of

19 leadership is that the coefficients for all three party-attention interactions should be positive and statistically significant in the both models. As a respondent s attention to politics increases, so, too, should his or her probability of approving of the war and disapproving of an immediate withdrawal. << Table 1 About Here >> Full results for both ordered logit models are presented in Table 1. The results are mostly consistent with theoretical expectations. In the model of Afghan war approval, the coefficients for all three partisan attentiveness interactions are in the expected direction, and two of them (for Conservatives and Labour) are statistically significant. In the withdrawal model, all three coefficients are again in the expected direction, and two (for Labour and Liberal Democrats) are statistically significant. To ease the interpretation of the interaction effects, Figures 5 and 6 present predicted probabilities, obtained from simulations, of war support for each partisan group as political attention increases. The results in Figure 5 are generally consistent with the mainstream pattern posited to hold in periods of elite consensus. There is no great divide between identifiers of the major parties on the Afghan War, which is consistent with the underlying cross-partisan elite consensus that Britain must remain committed to the central front of the war on terror. Moreover, the probability of approving of the war does increase slightly for Labour and Conservative party identifiers as attention to politics increases, though the increase generated by even a shift in political attentiveness from its lowest to highest level is exceedingly modest (8% for Conservatives and only 4% for Labour identifiers). Among Liberal Democrats, however, war support did not vary with political attentiveness at all. Thus, the data does suggest that Labour and Conservative party identifiers who were most attentive to politics were more likely to 18

20 support the war than their peers who were less attentive to politics presumably because these respondents received more cues from their party s leaders supporting the war. However, even among the most attentive respondents, support for the war hovers below 50%, a surprisingly low figure given the level of elite consensus behind the conflict. << Figure 5 About Here >> The effect of the interactions between partisanship and attention to politics is substantively stronger in the support for withdrawal models. As shown in Figure 6, consistent with the posited mainstream pattern, for all three partisan groups the probability of an individual respondent supporting an immediate withdrawal of British troops from Afghanistan decreases substantially as his or her attention to politics increases. The effect is most pronounced for Liberal Democrats and Labour party identifiers. For example, increasing a Liberal Democrat respondent s position on the political attention scale from one standard deviation below its mean to one standard deviation above it decreases the predicted probability of supporting withdrawal from over 50% to approximately 35%. The effect for Labour party identifiers is roughly equivalent in magnitude, whereas the estimated effect of increasing attention to politics for Conservatives is smaller, and the relevant coefficient in the regression failed to reach conventional levels of statistical significance. 7 << Figure 6 About Here >> In both models, we see evidence consistent with the mainstream pattern predicted by elite theories of opinion leadership to emerge in periods of partisan elite consensus. However, as with the aggregate data, the individual-level analysis again reminds us of limitations on the 7 The un-interacted coefficient for attention to politics is statistically significant in the withdrawal model. Thus, there is a statistically significant decrease in support for withdrawal for Conservative identifiers as attention to politics increases; however, the magnitude of this effect is no greater or less than that observed for respondents who did not identify with any party. 19

21 explanatory power of elite opinion leadership theories. Particularly in the basic war approval model, the modest increase in war support that results from increasing attentiveness to politics is modest and comes against a backdrop of surprisingly low levels of public support for the war. The effects of the partisan-attention interactions are stronger in the withdrawal models. However, even at the very highest levels of attention to politics, roughly a third of respondents are predicted to support immediate withdrawal, despite the forceful opposition of elites of all three major parties to a premature exit from Afghanistan. Experimental data It is exceedingly difficult to establish a causal relationship between elite discourse and change in public support for war from observational data alone. For example, the slight increase in the British public s support for war from , even in the face of mounting casualties and setbacks on the ground, is consistent with an elite-centered story emphasizing the continued cross-partisan elite consensus behind the war and the increasing prominence accorded it by political leaders in the long lead up to the 2010 general election. However, other factors, not elite discourse (after all, elites were solidly behind the war earlier in its course, as well, when support for the mission plummeted), could be driving the observed pattern. In a similar vein, the individual-level evidence of mainstream patterns in war support, in which support increases with political attentiveness, is also consistent with elite opinion leadership. However, the correlations could also be spurious, as any other factor that is correlated with both war support and a respondent s level of attention to politics could be producing the observed pattern. We cannot be sure that it is greater receptivity of and responsiveness to elite cues sent by partisan elites that is driving the observed increase in support among these respondents. 20

22 To overcome this difficulty and to assess the potential of elite cues to shape popular support for the war in Afghanistan almost a full decade after its inception, we embedded an original survey experiment within a YouGov survey with a nationally representative sample of 2,000 British adults. Respondents were randomly assigned to one of three groups. All respondents received the following prompt: The Conservative Prime Minister David Cameron has argued that British troops should stay in Afghanistan until Afghan troops are ready to assume security responsibilities. He hopes that this will happen by Respondents randomly assigned to the first group, the control group, were told no additional information. Respondents in the second group received the same initial prompt as those in the control, but were also told about the presence of a wider partisan elite consensus behind the need to keep British troops in Afghanistan for the foreseeable future. These respondents in our first treatment group were told: Labour and Liberal Democrat MPs support the Conservatives plan to keep British troops in Afghanistan. These MPs agree that withdrawing British troops prematurely would undermine the campaign against global terror. Finally, respondents in the second treatment group were also given the baseline prompt, but were then told that Labour and Liberal Democrat MPs opposed the Prime Minister s plan: Labour and Liberal Democrat MPs oppose the Conservatives plan to keep British troops in Afghanistan. These MPs have raised concerns about rising casualties and costs. They support an earlier withdrawal of British troops. All respondents were then asked the same question: Thinking about the British troops now in Afghanistan, do you think the Government should? followed by a number of options ranging from immediate withdrawal, regardless of conditions on the ground in Afghanistan to withdrawing British forces only after Afghan troops are ready to assume security responsibilities, even if this does not occur until after To ease interpretation of the results, respondents 21

23 were collapsed into two categories, those who supported withdrawal prior to 2015, regardless of conditions on the ground, and those who did not. 8 Because respondents were randomly assigned to the control or either of the two treatment groups, the differences in means across groups are unbiased. Results are presented in Table 2 below. << Table 2 About Here >> The first row presents the percentages favoring withdrawal before the PM s 2015 target in each of the three groups for all survey respondents. The elite consensus treatment, which told respondents that Labour and Liberal Democrat MPs supported the Conservatives plan, decreased support for an earlier withdrawal by 5%. From one perspective, the magnitude of this decrease is modest, even if statistically significant. However, the size of the effect is somewhat impressive given that, even after nine years in which war opinions logically should have hardened, a modest elite cue supporting the British commitment can still sway the opinions of a segment of the public. By contrast, the elite dissension treatment, which told respondents that Labour and Liberal Democrat MPs opposed the Conservatives plan to stay until 2015, had no statistically significant effect on war support. By 2010, an overwhelming majority of Britons wanted the troops to return home. In such an environment, the addition of a single, simple elite cue favoring a more expeditious withdrawal appears to have had little influence on respondents as a whole, most of whom had almost certainly long since made up their minds on the war. The next three rows disaggregate the results for Labour, Conservative and Liberal Democrat party identifiers. Consistent with other recent research emphasizing the differential responses of various partisan groups to the same elite cues (Groeling and Baum 2008, Berinsky 8 Alternatively, Wilcoxon-Mann-Whitney tests can be used to examine whether the distributions of responses on the full five point ordinal scale are significantly different across pairs of conditions. These nonparametric tests yield very similar results to those presented here. 22

24 2009, Kriner and Howell 2010), the results show that responses to the experimental treatments varied significantly across partisan groups. First, the disaggregated results suggest that the effect of the elite consensus cue was strongest on Labour party identifiers. This accords with past research in several respects. First, for these individuals the cue that Labour party MPs support the Conservatives plan is both trusted (Crawford and Sobel 1982, Druckman 2001) and costly (Calvert 1985, Groeling and Baum 2008, Kriner and Howell 2010) in that opposition party elites are backing the Government. 9 Moreover, Labour partisans may be the most conflicted on the war, with many having supported it for years while their party was in power despite other predispositions to oppose interventionist military policies. With the war having become the Conservatives, Labour identifiers may be particularly apt to look to their party s elites for cues about whether they should continue to support the war even now that the party is in opposition. Labour and Liberal Democrat opposition increased the percentage of respondents supporting an expeditious withdrawal, but the increase is not statistically significant. Conservative identifiers appear unresponsive to cues transmitted by Labour and Liberal Democrat MPs. There is no evidence of any statistically significant effect for either treatment on the opinions of Conservative respondents. This unresponsiveness to the positions taken by elites of other parties is perhaps unsurprising; however, one striking result was the very low level of support for the Prime Minister s timetable among his co-partisans. Conservatives were more supportive of withdrawing British troops before 2015, regardless of conditions on the ground than any other partisan group. Finally, among Liberal Democrats the elite consensus cue had no statistically significant effect on support for withdrawal. However, the Labour and Liberal Democrat MP opposition cue did increase the percentage of LibDem respondents supporting withdrawal from 54.9% in the 9 Of course, given Labour s long-standing support for the war, a reversal could also be politically dangerous. 23

25 control group to 63.5% in the opposition treatment group. This 8.5% increase is substantively large and only narrowly misses conventional levels of statistical significance. 10 This result is at least suggestive of an underlying tension within the Coalition; if Liberal Democrat MPs were to speak out against plans to stay in Afghanistan for the long term, it could lead to a surge in war opposition among Liberal Democratic ranks. Thus, the experiment shows that elite cues still hold the power to shape British support for the war in Afghanistan in modest, but significant ways, even though the opinions of many may be hardened after nine years of exposure to the war. Labour identifiers appear to be the most responsive to signals of elite consensus behind the Government s plan. Liberal Democrats, by contrast, appear to be more responsive to signals of opposition from their own and Labour party elites. Conclusion Two of the most frequently made claims in the wartime opinion literature are that mounting casualties erode public support for a military mission, while elite consensus bolster it. When applying these ideas to the case of British support for the war in Afghanistan, however, important divergences from theoretical expectations emerge. First, casualties offer little insight into patterns of support for the Afghan war in Britain over time. Support for the conflict plummeted to very low levels even before British casualties began to rise in the late summer and fall of And as casualties began to mount with the intensification of the fighting in , British support seems to have increased slightly. 10 Moreover, a Wilcoxon Mann-Whitney test suggests that the difference in the full distributions of responses in the opposition treatment and control groups is statistically significant (p <.10). 24

26 With respect to elite opinion leadership theories, the empirical data is decidedly more mixed. The most striking result is the low levels of British support for the war, despite strong, cross-partisan elite support for the conflict from its initiation to the present day. Contra expectations, British elite consensus behind the Afghan war has failed to produce majority support for the mission. It may be that many Britons simply viewed the war in Afghanistan through the lens of the much more unpopular and contentious conflict in Iraq. Alternatively, widespread distrust of President Bush and American foreign policy may have dampened support for British involvement in the American-led mission. 11 Future research should investigate these and alternative hypotheses to understand the puzzlingly low level of support for the Afghan war in Britain. However, both the individual-level survey and experimental data do show evidence of the British public s responsiveness to elite cues. Moreover, it appears that elite opinion leadership may be strongest on the public s prospective policy preferences regarding the war. While the relationship between attentiveness to politics and war support were weak in the general war approval model, the relationship was considerably stronger in the withdrawal models. Labour and Liberal Democrat identifiers with high levels of attention to politics were significantly more likely to oppose an expeditious withdrawal than were their peers who are less attentive to politics. At the aggregate level, we observed the greatest upward trend in support from 2007 to 2010 in polls measuring Britons beliefs in their opposition to immediate withdrawal and in their beliefs in the mission s capacity to succeed. Finally, our experiment, which showed strong 11 For example, as early as November 2003, in an ICM poll 48% of Britons judged George W. Bush a threat to world peace. Similarly, a 2007 BBC poll found that 57% of Britons believed that the United States primarily played a negative role in the world, versus only 33% who believed it played a positive one. Kevin Sullivan, View on U.S. Drop Sharply in World Poll. Washington Post, January 23,