Looking for Audience Costs in all the Wrong Places: Electoral Institutions, Media Access and Dispute Reciprocation

Transcription

1 Looking for Audience Costs in all the Wrong Places: Electoral Institutions, Media Access and Dispute Reciprocation Keywords: Audience Costs, Dispute Reciprocation, Parties, Media Access Matthew A. Baum Professor Harvard University Philip B. K. Potter Assistant Professor University of Michigan ABSTRACT For leaders to generate credibility through audience costs, there must be mechanisms in place that enable citizens to learn about threats and called bluffs. However, scholars have paid relatively little attention to variations among democracies in the extent to which the public is able to obtain this sort of information. We argue here that electoral institutions play this role by influencing the number of major political parties in a country, and, with it, the extent and depth of opposition to the executive. Multiple parties are associated with improved whistleblowing, which makes it more likely that that the public will actually hear about a leader s foreign policy blunders. The effectiveness of this whistleblowing, however, is conditional on the public s access to the primary conduit for communication between leaders and citizens: the mass media. We test these expectations statistically, and demonstrate that dispute initiators with higher numbers of political parties, conditional on public media access, face less reciprocation and have a greater likelihood of winning, than those with fewer parties. These findings suggest that democracies are not automatically able to generate credibility though audience costs, and that the domestic institutions and political processes that link the public and leaders must be taken seriously.

2 Are domestic audience costs central to establishing credibility in crisis bargaining? This question, introduced by Fearon (1994), has given rise to a virtual cottage industry of international relations scholarship, and the resulting discussion has fundamentally shaped the discipline s understanding of democratic conflict behavior. Yet, despite its volume, in many regards this literature still reflects the narrowness of the stylized theoretical puzzle that the argument was originally designed to solve. Audience cost theory holds that democratic leaders, by virtue of their electoral vulnerability, can enhance their bargaining positions in international negotiations or disputes by making public threats or promises (Fearon 1994; Smith 1998; Schultz 2001; Baum 2004b, 2004a; Schultz 1998). By risking political punishment at home for failing to follow through on a public commitment, a leader can effectively tie her hands, thereby enhancing her credibility to an adversary. Because backing down from a public threat is potentially costly for democratic leaders who are at the mercy of the ballot box, an adversary will recognize that these leaders are more likely than autocrats to follow through. Consequently, when a democratic leader makes a public threat, an adversary is more likely to view it as credible and therefore back down without a fight. But how might a mechanism like this actually function when transitioned from theoretical models to political reality? The concern in many circles (e.g. Gartzke and Lupu 2012) is that the answer likely falls somewhere between messily (at best) and not at all (at worst). In this article we attempt to unpack and resolve one particularly worrisome source of potential complication. Our specific concern is that the theoretical models that underpin the audience cost argument have relatively little to say about institutional heterogeneity within democracies. Just as Weeks (2008) differentiates among autocracies, finding a subset that are in fact able to 1

3 generate audience costs, we do much the same for democracies by establishing that not all are equally well-equipped to signal credibly. Specifically, drawing on the political knowledge and communication literatures, we identify two sources of variation in the capacity of democracies to generate credibility through audience costs: 1) the prevalence of whistleblowers positioned to ensure that information about a leader s foreign policy missteps will reliably become public, and 2) the extent of public access to such information. Leaders have clear incentives to hide their foreign policy blunders. To ensure that they cannot do so, there must be heterogeneous and autonomous political elites in positions of power that have both independent access to foreign policy information and the incentive to reliably blow the whistle, thereby transmitting that information to the mass public. We argue that opposition political parties are the most obvious candidates to fill this role in democracies and that the larger the number of opposition parties, the more efficient this mechanism becomes. Why is one opposition party sometimes insufficient to blow the whistle on the executive s foreign policy miscues? Fewer parties translate into diminished options for voters and a lesser likelihood that they will punish at the ballot box. Systems with larger numbers of parties are more likely to produce ideologically proximate alternatives for voters than are systems with fewer parties (Downs 1957). 1 In addition, electoral competition among a larger number of parties results in more competing policy frames (Milner 2002). Consequently, media in multi-party democracies are more likely than their counterparts with fewer parties to have 1 Two examples serve to illustrate this insight. Supposing that Obama were caught in a foreign policy bluff it is likely that relatively few of his median prior supporter would be willing to abandon him in favor of Romney because the two candidates differ so substantially on so may policy dimensions (though there might still be an impact on independents and swing voters that could be decisive in the election). In contrast, were Netanyahu caught in the same situation, he would presumably be more susceptible to defections to the far right religious parties and the centrist parties because these parties are more ideologically proximate to the median Likud voter. 2

4 access to, and hence to make available to citizens, competing frames including alternatives to the government s preferred frame (Sheafer and Wolfsfeld 2009). Heterogeneous political elites are necessary but insufficient to generate audience costs on their own. The public must also have access to the whistleblower s message, which is most likely with a free and accessible media. While several scholars have noted that a free media is an integral part of the audience cost process (Slantchev 2006; Potter and Baum 2010), it is also the case that democracy and the available measures of media freedom are nearly perfectly collinear, making any effects of media freedom difficult to differentiate from the more prevalent arguments about regime type. What has generally escaped notice, however, is the substantial variance within democracies in terms of actual access to these media that is, to the technologies of media reception (e.g., televisions, radios etc.). The implication of this variation is that that public scrutiny and the consequent potential political costs and benefits that go with it may be more likely in some states than in others for the simple reason that leaders (and opposition parties) have varying degrees of access to the public eye. Thus we argue that the ability of leaders to generate audience costs is contingent on the presence of independent political parties to serve as whistleblowers, conditioned on the extent of access to an independent media that can bridge the gap between political elites and the mass public. While audience cost arguments play an integral role in rationalist theories of war and peace, uncovering empirical support of the audience cost argument has proven difficult. The very nature of audience costs makes them difficult to detect. Leaders who successfully generate credibility do not incur costs (Schultz 2001), leading to potential bias stemming from partial observability. The solution that Schultz and others (e.g., Weeks 2008) employ is to explore the 3

5 reciprocation of militarized interstate disputes as a second order implication of the audience cost argument. The logic is that the decision to reciprocate a dispute indicates the extent to which the threatened state finds the initiator credible. 2 All else equal, states that are able to generate audience costs should be more credible in their threats against adversaries and therefor face less reciprocation of those threats (Smith 1998; Schultz 1998). They should also be more successful in militarized disputes (Fearon 2004). We exploit this logic in our empirical tests, as well as assessing two alternative formulations not previously explored in the literature: violent reciprocation and victory. The remainder of this article proceeds as follows. We begin by discussing the logic of audience costs in more detail, with particular attention to the issue of mass political knowledge and engagement. We then clarify the role that electoral institutions play in linking the public to leadership in ways that make sanctions for bluffing more or less likely. Next we establish the importance of media access as a key requirement that allows audiences to monitor the foreign policy performance of their leaders. Finally, we test the resulting hypotheses in statistical models of dispute reciprocation and outcome, finding support for the proposed conditional relationship between whistleblowers, media access, and credibility. We conclude with a discussion of the implications of our findings for future work on audience costs. Audience Attention is not Automatic Most scholarship on audience costs shares an implicit assumption about the nature of democracies. This is simply that the actions and statements of democratically elected leaders are 2 Data come from the Correlates of War Project Militarized Interstate Dispute dataset. MID Reciprocation equals one when the hostility index score for the target (cwhost2) is greater than one. 4

6 immediately transparent to the voting public. This expectation, however, is at odds with a virtual consensus in the political communication literature, which holds that elites filter and distort the information they attempt to transmit to the public (Bennett 1990; Livingston and Bennett 2003; Bennett et al. 2006). Compounding the problem, the public is not equally attentive to all types of messages (Zaller 1992; Lupia and McCubbins 1998; Popkin 1993). While nothing in the audience cost argument requires the public to be perfectly informed, the theory does rest on the assumption that the public engages with foreign policy enough to be both aware of threats and consistent enough in its judgment regarding called bluffs to punish the behavior at the ballot box. Obviously, such a process is contingent on the public s capacity to gather and retain information, and to then use that information to formulate coherent opinions about the performance of their leaders. Existing research, however, suggests that the public is not particularly well-equipped to accomplish these tasks (Berinsky 2007; Baum and Groeling 2010). 3 Moreover, the prevailing evidence suggests that the public s attention to matters of foreign policy is generally quite low (Holsti 2004). This raises troubling questions about the extent to which nuanced information about foreign policy including information about threats and bluffs reaches individual voters. It is therefore quite plausible that, at least under some circumstances, democratically elected leaders can conduct foreign policy unencumbered by public scrutiny and any accompanying democratic constraints (Baum 2004b). Presumably, when they do so their signals are no more credible than those of autocracies. If this is the case, then audience costs are unlikely to arise mechanically and universally in democracies, but rather will be varied and context 3 But see Tomz s (2007) experimental research indicating that audience costs build as a crisis escalates, suggesting that citizens (especially politically attentive ones) respond rationally to events, and that citizens do care about their nation s international reputation and are hence motivated to punish failure in foreign policy. 5

7 dependent. Thus, our contention is not that the audience cost argument is incorrect, but rather that mapping the model to actual politics requires that we carefully consider which institutions are required for the mechanism to function well. Elite Dissention, Parties and Audience Costs Given the established low baseline of public attention, what prompts the mass public to sometimes pay enough attention to foreign policy matters to impose political costs? Drawing on the prevailing view in political behavior research, we argue that voters use heuristics to help them determine both when to engage with foreign policy issues and what to think about them when they do. Chief among these is the role played by trusted political elites in the formation of mass opinion. Research has shown that in many situations citizens are able to make rational decisions with relatively little information by employing informational shortcuts (Sniderman et al. 1991; Popkin 1993), most notably by relying on the opinions of trusted political elites (Larson 2000; Krosnick and Kinder 1990; Iyengar and Kinder 1987). 4 This insight is particularly salient when considered in the light of the extensive literature on public opinion and foreign policy that finds that elite discord plays a central role in drawing a typically disengaged public into the foreign policy fray. Schultz (1998) notes the key role for opposition parties, but conceives of this role in binary terms: the presence of a single opposition party represents a shorthand for democracy, while its absence represents a shorthand for autocracy. However, the expansive comparative literature on democratic governance makes it clear that the number of opposition parties matters a great deal. Downs (1957) links the number of political parties to the nature of political debate, arguing that fewer parties introduce 4 Popkin (1993) terms this process low information rationality. 6

8 pathologies including ineffective government in the absence of political consensus among voters and ambiguous platforms and positions. In contrast, he argues that multiparty democracies tend toward sharply defined ideological distinctions between the parties. Similarly, Lijphart (1999) demonstrates a relationship between the effective number of parties in a democracy and the number of issue dimensions (see also Cox 1990; Dow 2001). Recent research into the relationship between institutional forms of democracy, media diversity, and citizens political knowledge (e.g. Schmitt-Beck 2003; Baum and Zhukov 2012) suggests that not all democracies are alike in the quality of information they provide to their citizens. In particular, media outlets in multi-party electoral systems tend to report on a wider range of policy issues, and to do so in greater depth, than media in two-party systems. In other words, multiparty systems are associated with higher quality political information where quality is defined as information voters can use to inform party choices across contests (local, state, and federal) and across time (Moosbrugger n.d. p. 13) and, hence, more politically sophisticated electorates, than two-party systems (Kumlin 2001; Schmitt-Beck 2003). Hence, all else equal, citizens in multiparty systems ought to be better equipped to hold their leaders accountable than their counterparts in two-party systems. In general, elite discord serves as a heuristic which signals to voters that they should engage with the policy process. Zaller (1994) notes that elite discord can produce a polarization effect by sparking more critical media coverage, resulting in heightened attention from the public. Through this mechanism, opposition parties can act as whistleblowers who inform the public when leaders are not performing, including when their foreign policy bluffs are called. This possibility meshes closely with the received wisdom about the role of elite discord in the formation of public opinion about foreign policy. For instance, in studies of the U.S. media, 7

9 scholars (e.g. Bennett 1990; Zaller and Chiu 2000; Entman 2003) have found substantial evidence that news content, especially in times of war, tends to be indexed to the tenor of elite debate. This means that such news coverage reflects the tone of policy debate in Washington. A hypothesis follows concerning how domestic institutional structures might influence the ability of leaders to generate audience costs: H1: Ceteris paribus, as the number of parties increases in the initiator state the probability of reciprocation by the target state will decrease. Media Access Of course, the presence or absence of whistleblowers means relatively little unless the public hears them. Most work in this area assumes that the statements and actions of leaders and the opposition are immediately transparent to the public. This may have been a defensible assumption prior to the 20 th Century, when participation in politics and diplomacy was typically limited to a handful of elites within a state. But in an era dominated by mass democracies, in which many millions of individuals hold the franchise, this assumption is problematic. Instead, theories that rest implicitly on political communication must grapple with the role of media as an institution that links leaders, opposition, and the public. A free press is a defining characteristic of liberal democracies. Many autocracies have elections, legislatures, and the outward trappings of representation, but few tolerate open dissention from the press corps. 5 This nearly perfect coincidence of democracy and press freedom has obscured the systematic variation in the transmission of information to the public 5 The correlation between the widely used Polity IV measure of relative democracy and the Reporters without Borders Press Freedom Index is No country with a Polity score less than 5 in 2003 (on the -10 to +10 DEMOC-AUTOC scale, where +10 is most democratic) scored higher than 81 st in the world in press freedom. That country was Cambodia. 8

10 that might alter the audience cost mechanism. 6 Yet while press freedom is relatively consistent across democracies, access to that press is not. For instance, among democracies in our data set, the mean number of televisions per 1,000 population across all years is 354, or about one television for every three inhabitants. Norway in 1980 fell very close to this average, at 350 TV s per 1,000 people. However, the standard deviation around this mean is 245. This means that in roughly 1/3 of the democracies in our data set, there is at most one television for about every 10 inhabitants. The Philippines in 1998 fell near this level, with, on average, about 110 TV s per 1,000 residents. In 1996, TV access was just over half that level in India, at 63 TV s per 1,000 residents. This represented a sharp rise from, say, 1985, when there existed only about one television for every 250 residents. The U.S. and U.K. have the highest levels of TV access among democracies in our dataset, at 831 (1997) and 850 (1999) TV s per 1,000 inhabitants, respectively. Access to mass media is integral to the process of generating audience costs because, when access is limited the public has a diminished capacity to reliably determine whether its leaders have bluffed or failed. In such a setting, the leaders in power, regardless of how they gained office, can control the flow of information in a way that renders hollow the threat of electoral sanction. Even if there are opposition politicians with an interest in exposing the failure of an incumbent leader, absent broad media access their ability to communicate with the public is severely constrained. An additional hypothesis follows: H2: Ceteris paribus, as media access increases in the initiator state the probability of reciprocation by the target state will decrease. 6 A few scholars have investigated the role of the free press in the audience cost argument. See, (Choi and James 2006; Slantchev 2006; Potter and Baum 2010) 9

11 In an absolute sense, whistleblowers without access will never reach the public. Similarly a public with universal media access, but without a whistleblowing elite, will never hear about bluffing. The implication is that the risk of whistleblowing will only inhibit leaders from bluffing in the face of a credible threat that the public will hear the whistle being blown. This requires that a substantial portion of the public have access to media reporting about government policy. Conversely, public access to the media is only likely to concern leaders considering bluffs given a credible threat that some elite will blow the whistle. Consequently, we anticipate an interaction between party systems and media access, with the strongest credibility enhancing effects emerging in the presence of a whistleblowing elite, a press corps inclined to cover the blowing of whistles, and a public likely to receive such messages. A third set of hypotheses follows: H3a: Ceteris paribus, at low levels of public media access in the initiator state, variations in the number of parties will not significantly affect the likelihood of reciprocation by target states. H3b: Ceteris paribus, beyond some threshold of media access in the initiator state, as the number of parties increases, greater public access to the media will be more strongly associated with decreased reciprocation by target states. Even in democracies with fewer parties there remains an opposition and thus some audible (to the public) baseline for elite discord. Similarly, all democracies have media access the question is how much. Level of Reciprocation Of course all reciprocation is not equal. In particular, low-level, nonviolent reciprocation, such issuing threats, may be less influenced by variations in initiator credibility in part because it has a lesser chance of leading to uncontrolled escalation. Indeed, such tepid reciprocation may, in fact, represent a hedging strategy from a target state that believes that the initiator is indeed 10

12 credible, but is perhaps playing for time and information or seeking to minimize the costs from its own domestic audience that might follow immediate acquiescence. In contrast, higher-level reciprocation, including the actual use of violence, is a clearer indication that the target is challenging the initiator s credibility. It may therefore be the case that violent reciprocation might drive any credibility-based effects for overall reciprocation (combining both low- and high-level types). This leads us to a fourth set of hypotheses. H4a: Ceteris paribus, as the number of parties increases in the threat initiating state the probability of violent reciprocation by the target state will decrease. H4b: Ceteris paribus, as media access increases in the threat initiating state the probability of violent reciprocation by the target state will decrease. H4c: Ceteris paribus, at low levels of public media access in the initiating state, variations in the number of parties will not significantly affect the likelihood of violent reciprocation by target states. H4d: Ceteris paribus, beyond some threshold of media access in the initiating state, as the number of parties increases, greater public access to the media will be more strongly associated with decreased violent reciprocation by target states. Dispute Outcome Our first four hypotheses all address the relationship between democratic intuitions and dispute reciprocation as an indirect measure of the ability to generate audience costs. There are, however, limitations to the utility of reciprocation as an indicator of the presence or absence of a functional audience cost mechanism. Foremost among these is the reality that reciprocation is a second order implication of the audience cost argument. That is, the domestic calculations that fuel the audience cost mechanism are assumed rather than tested because leaders who face audience costs have obvious incentives not to incur them. Second, reciprocation is difficult to measure; even if it is a conceptually sound indicator of variations in audience costs, the coding 11

13 may be substantially flawed. These concerns aside, models of reciprocation have become standard in this literature in large part because no equally plausible alternatives have since emerged. Hence, since we are making several novel adaptations to audience cost theory we begin by applying them to the standard measure in order to allow comparability. In order to further raise confidence that our theoretical mechanism is indeed driving our findings we explore a separate theoretical implication outside the standard tests of reciprocation. Specifically, it follows directly from Fearon (1994) that states with greater credibility generated through audience costs should win their disputes more often then their counterparts that are unable to generate credible commitments. This leads us to three additional hypotheses, which are direct parallels of those already introduced. H5a: Ceteris paribus, as the number of parties increases, initiators become more likely to win disputes. H5b: Ceteris paribus, as media access increases, initiators become more likely to win disputes. H5c: Ceteris paribus, beyond some threshold of media access, as the number of parties increases, greater public access to the media will be more strongly associated with increases in the likelihood that initiators will win disputes. RESEARCH DESIGN We initially test our hypotheses with adaptations of Schultz s (2001) models of militarized interstate dispute reciprocation. Our measure of dispute reciprocation is equal to one when the target state responds to a threat militarily (and zero otherwise). We assess the robustness of this formulation of the dependent variable by splitting reciprocation into two varieties and analyzing each separately and in conjunction: (a) any militarized reciprocation and (b) violent reciprocation entailing escalation to the actual deployment or employment of military force. In a worst-case scenario this means that we are uncovering important variation in dispute 12

14 reciprocation (driven by a mechanism other than audience costs), which we argue is interesting and important in itself. However, this begs the question of what that alternative data generating mechanism might be. This study adds to a growing literature tying reciprocation to the domestic politics of the initiating state, which, when taken together, increasingly crowd out mechanisms aside from the audience cost argument. The logic of audience cost theory ought also to offer leverage into predicting outcomes. Hence, to further increase confidence in our argument, we expand the analysis to investigate the outcomes of the disputes. We thus turn to a different facet of the correlates of war militarized interstate dispute dataset. Specifically, we code an event as a win for the initiating state if Side A obtains its preferred outcome. Our period of analysis spans the years We chose this period because it corresponds with the television age, and access to television serves as one of our primary measures of media access. In addition to this change in the period of analysis (Schultz s models span from 1945 to 1999), we make a several additional departures. First, since we are concerned with distinctions within democracies (rather than between democracies and autocracies), we restrict the models to democracies and exclude Schultz s key independent variable (a dichotomous measure of democracy/autocracy). In its place, we consider variables designed to capture the nature of the electoral system specifically the number of politically meaningful parties within each state. We initially operationalize this variable using Golder s (2005) measure of the Effective Number of Parliamentary Parties (ENPP). Golder (2005) defines ENPP based on the following formula, from Laakso and Taagepera (1979): 1 Σs!! 13

15 where s i is the percentage of the vote won by the ith party, with Independents or others again coded as a single party. For our primary analyses, we account for media access with counts of televisions per 1000 people. We focus on television for two reasons, one conceptual and one practical. Beginning with the latter, far more data are available on TV access than for other indicators of mass media. More significantly, television remains by far the most important form of media worldwide for presenting political information to mass audiences. According to a 2009 Pew Center survey of 25 countries 7 spanning nearly every region of the globe and level of economic development an average of 72% of respondents named television as their primary source of news about national and international affairs. Newspapers came in second at less than 10%, while the Internet took fourth place on the list, at 7.9%. It is reasonable to anticipate that the dominance of television during much of our period of analysis was at least as high, if not higher. Nonetheless, in a robustness test reported below we replicate our results while also taking radio access into account. Following Schultz (2001), we control for the power dynamic within the initiator/potential reciprocator dyad, as well as measure of whether those states are contiguous, whether they are allies, and the degree of similarity in their alliance portfolios. We also include several measures of the nature of the initiator s challenge. Finally, we include one additional control variable that is not typical in models of reciprocation: child mortality. We use this widely employed measure to account for each country s relative level of development (Lake and Baum 2001). This variable reduces the likelihood that our media access variables will pick up any development effect that 7 The countries included in the survey were: Canada, United States, Argentina, Brazil, Mexico, Britain, France, Germany, Spain, Poland, Russia, Egypt, Israel, Jordan, Lebanon, Palestinian territories, Turkey, China, India, Indonesia, Japan, Pakistan, South Korea, Kenya, and Nigeria. 14

16 might then be collinear with media exposure, the object of our theoretical interest. At the same time, development may be associated with increased demands for political responsiveness and, through this mechanism, have a relationship with our dependent variable. Results Table 1 presents the results of our primary regressions. Model 1 explores the effect on all dispute reciprocation. Model 2 limits the analysis to violent reciprocation (that is, the actual deployment or employment of military force). Model 3 enables us to assess the hypotheses regarding the effects on victory. [Table 1 here] Because it is difficult to substantively interpret log likelihoods and conditional coefficients, we transform the coefficients on the key causal variables in Table 1 into probabilities of any reciprocation, violent reciprocation, and victory, as media access and the number of parties varies, with all other controls held constant at their mean values. Figure 1 plots the results. In each graphic, we separately plot the predicted probability of reciprocation, violent reciprocation, or victory as TV access increases from its lowest to highest values among Democracies (0 to 1000), with the expected number of parliamentary parties (ENPP) set one standard deviation below and above its mean value, also among Democracies (1.7 to 4.34). [Figure 1 here] Our first two hypotheses predict that as the number of parties (H1) or media access (H2) increases, reciprocation and violent reciprocation (per H4a and H4b, respectively) in MIDs by target states will become less frequent. The results in Figure 1 support both predictions. Beginning with H1 and H4a, across all three graphics, and at nearly all values of TV access the 15

17 exception being at the low-end of TV access the green curves (representing one standard deviation below the mean number of parties) are 21 or 20 percentage points on average, for reciprocation (H1) and violent reciprocation (H4a), respectively, higher than the red curve (representing one standard deviation above the mean number of parties). These average differences are highly significant. Turning to H2 and H4b, here the appropriate test is to observe the slope of the curves as media access increases. In fact, both the green and red curves are downwardly sloping for both the all reciprocation (consistent with H2) and violent reciprocation (consistent with H4b) graphs. Both red curves are statistically significant at p<.01, while, as expected, the green curves are considerably shallower and statistically insignificant. In short, increased media access is indeed associated with reduced reciprocation and violent reciprocation, albeit far more strongly and significantly so (as expected) given a relatively high number of parties. This brings us to H3a and H3b, as well as H4c and H4d, which predict that the overall reciprocation-repressing effects (H3a and H3b) and specifically violent reciprocation-repressing effects (H4c and H4d) of increased media access ought to heighten as the number of parties rises (H3b and H4d), but only beyond some minimum threshold of media access (H3a and H4c). As suggested above, this is clearly the case in both graphics. In the low-enpp cases (the green curves), the downward slopes associated with increased TV access are all relatively shallow, and, as noted, insignificant. In sharp contrast, all three high-enpp cases (the red curves) are sharply downward sloping and significant at p<.01 for reciprocation and violent reciprocation. 8 8 Significance levels are based on the full range of TV access. The significance levels remain p<.01 when TV access varies from one standard deviation below to one standard deviation above the mean. 16

18 Moreover, consistent with H3a (all reciprocation) and H4d (violent reciprocation), in all cases the effects of variations in the number of parties are statistically indistinguishable at the low end of TV access. This supports our theory concerning the necessity of access; increased whistleblowing has no statistically identifiable effect when the mass public is relatively unlikely to get the message. The effects become statistically distinguishable at a little below one television per three residents, which is quite close to the overall mean level of TV access among democracies in our data of about one TV per 2.8 residents. H5a-H5c predict that as the number of parties (H5a) or media access (H5b) increases, initiator states will be more likely to win the MIDs they initiate. Beginning with H5a, the bottom-left-hand graphic in Figure 1 indicates that at nearly all values of TV access the exception again being at the low end the red curve (representing one standard deviation above the mean number of parties) is higher by an overall average of 27 percentage points than the green curve (representing one standard deviation below the mean number of parties). This difference is again highly significant. Turning to H5b, here the appropriate test is to observe the slope of the curves as media access increases. Consistent with H5b (as well as implicitly with H3a), the red curve is upwardly sloping for the victory graph (p<.01). 9 As expected, the green curves are considerably shallower and statistically insignificant. In short, per H5b, increased media access is indeed associated with an increased probability of victory, albeit again, as H5c predicts, only for high party states. Moreover, also consistent with H5c, these effects become statistically distinguishable only when 9 Significance levels are based on the full range of TV access. The significance level becomes p<.05 when TV access varies from one standard deviation below to one standard deviation above the mean. 17

19 media access exceeds about one television per 2.5 residents, which is just above the overall mean among democracies in our data (which is, as noted, one TV per about 2.8 residents). Table 2 summarizes the probabilities at one standard deviation above and below the mean values of parties operationalized through both ENPP (discussed here) and ENEP (discussed in robustness test section, below) and media access. [Table 2 here] Ordered and Multinomial Models To determine whether or to what extent high-level (violent) reciprocation is driving the observed relationship with combined low- and high-level reciprocation indicator, we convert our dependent variables into a single scaled variable where 0=not reciprocated, 1=reciprocated with maximum of threat, 2=reciprocated with maximum of force deployment or employment. This change in the dependent variable requires a shift to an estimator that allows for more than two values on the dependent variable. At first glance the reformulation of the dependent variable appears best suited for an ordered logit analysis. However, reviewing the Tau Beta coefficients and standard errors from this model suggests potential overlap in the first two categories. The implication is that the categories may not be entirely distinct and evenly distributed, which violates an assumption of ordered logit. Consequently, we also present a second analysis employing unordered multinomial logit. The latter estimator has the benefit of relaxing the assumption of an ordered, evenly distributed scale. Table 3 presents the results from both models. [Table 3 here] 18

20 In Figure 2 we separately plot the probabilities of non- reciprocation, non- violent reciprocation, and violent reciprocation for the ordered (top 3 charts) and unordered multinomial (bottom 3 charts) logit models, respectively. [Figure 2] The results in Figure 2 essentially replicate those from our prior analyses, in that variations in TV access have a large and statistically significant effect, in the predicted directions, in high- party states but not in low- party states. Interestingly, however, both models represented in the Figure suggest that, as anticipated, the effects of variations in media access are greater for the movement from non- reciprocation to violent reciprocation than for movement from non- reciprocation to non- violent reciprocation. Though this is evident in both models, it is particularly so in the unordered multinomial logit, where the probability of low- level reciprocation varies hardly at all with TV access. The effect is somewhat larger and statistically significant in the ordered logit model, though still substantially more modest than with respect to violent reciprocation. This distinction is not evident in the dichotomous specifications, presumably because the all reciprocation dependent variable treats both violent and non- violent reciprocation as a positive occurrence. Hence, as expected, these alternative models suggest that the interaction of party systems and media access matters more for violent than for non- violent reciprocation, with the strongest constraining effect apparent with respect to the latter, more serious types of disputes initiated by high- party systems with high media access. Replications and Robustness Tests 19

21 Taken together, the preceding results indicate strong support for each of our hypotheses across multiple dependent variables. However, to further establish the robustness of our findings we introduce alternative operationalizations of our key independent variables: the number of parties and the extent of public media access. For an alternate formulation of the number of politically autonomous parties we turn to Golder s measure of ENEP, representing the expected number of electoral parties. Golder (2005) defines ENEP, in turn, with the following formula, which like the measure of ENPP is taken from Laakso and Taagepera (1979): 1 Σv!! where vi is the percentage of seats won by the ith party, with Independents or others coded as a single party. The difference between ENPP and ENEP stems from the fact that not every party that competes in a given election wins a sufficient vote share to gain seats in the parliament. ENPP only takes into account those parties that win seats, which itself depends on electoral rules translating votes into seats. These, in turn, vary widely across countries. We believe ENPP, because it discounts small, outlier parties with no meaningful chance of winning seats or a place in a coalition government, is the better indicator. Nonetheless, because this is an uncertain judgment, we replicate our findings using ENEP. We present the results of the logit equations in Table 3. To conserve space and because logit coefficients are difficult to interpret we report robustness findings only as graphed probabilities (see top half of Figure 3) rather than coefficient tables. (We invite interested readers to consult the online appendix for the full sets of regression results.) [Figure 3 here] 20

22 Comparing Figures 1 and the top half of Figure 3, it immediately becomes apparent that the results are virtually indistinguishable. The effects of increased TV access in high party states are slightly stronger in the ENEP reciprocation and violent reciprocation models, while the low- ENEP curves are somewhat flatter, as TV access varies. As indicated in Table 2, both curves representing all reciprocation or violent reciprocation, as TV access varies are highly significant (p<.01) given a high number of parties and insignificant given a low number of parties. The probabilities of reciprocation and violent reciprocation vary slightly more strongly, in the predicted direction. Finally, the effects of variations in TV access on the probability of victory, though a bit less pronounced than in the ENPP model (and with correspondingly wider confidence intervals), are also substantial in magnitude and statistically significant at p<.01 given a high number of parties. Conversely, the corresponding effects are tiny and insignificant given a low number of parties (see Table 2). Moreover, as in the ENPP models, the distinct effects of media access for low- vs. high-party states only materialize beyond a threshold level of TV access, with the threshold again lying near the mean level of TV access among democracies in our data set. In short, each of our hypotheses is supported when we employ ENEP in place of ENPP as our indicator of the number of political parties. To further assess the role of media access we investigate a second operationalization, combining television and radio access. Because in some developing democracies access to television was, and continues to be, limited, we combine our TV access variable with an indicator of the number of radios per 1,000 population. Simply put, households in less developed nations are relatively more likely to own a radio than a television. Including this indicator is particularly helpful in the early years of our data set, when relatively few citizens in developing nations had access to television, even as radios were far more prevalent. To illustrate this point, 21

23 among democracies, the correlation between time and our TV access indicator is.35, indicating a strong increase in TV access over time. The corresponding correlation for radio access is -.06, indicating at most a slight, and in this case negative, over-time trend. The point is that radio penetration had largely peaked by the late 1960s, while TV ownership increased over the course of the ensuing decades covered in our data. The bottom half of Figure 3 presents the final set of probabilities and again indicates that that our findings hold for this alternative formulation of media access. The results shown in the bottom half of Figure 3 replicate those from Figures 1 and the top half of Figure 3 in every aspect. In fact, any differences indicate even stronger magnitudes of effects, consistent with each of our hypotheses, for the TV + radio ownership operationalization of media access. Table 4 presents the corresponding probabilities of reciprocation, violent reciprocation, and victory, as TV+Radio access increases from one standard deviation below to one standard deviation above the mean values among Democracies in our data set. The results are presented separately for states a standard deviation above and below the mean number of parties (based on the ENPP indicator). [Table 4 here] The results indicate that among high- ENPP states, increased TV+Radio access is associated with large and statistically significant declines in non- violent and violent reciprocation, as well as increases in the probability of victory for the initiator. As before, the corresponding effects for low- party states are small and insignificant. As in our prior investigations, the low- and high- party curves become significantly distinct beyond some 22

24 threshold of media access located near the overall mean level of media access. These results once again support each of our hypotheses. Finally, we replicate ordered and multivariate logit modes for the same variations in the measures of parties and media access. We plot the results in Figures 4-6. [Figures 4-6 here] The findings are generally quire consistent with those uncovered in Table 1 and graphed in Figures 1-3. As with Figure 2, the one instance where the models employing dichotomous dependent variables differ from the 3- category models concerns the probability of non- violent reciprocation, which varies far less in all of the 3- category models than does the probability of violent reciprocation. These results once again imply that the effects of variations in media access primarily concern movement from non- reciprocation to violent reciprocation. CONCLUSION The ability of leaders to use their citizenry to generate credibility in their international interactions is variable and context dependent. Specifically, electoral institutions that give rise to a whistleblowing opposition, combined with sufficient media access to make it likely that citizens will be aware of any such whistleblowing, facilitates credibility. Put differently, if the domestic institutions in a democracy are such that the public is unlikely to hear about backing down should it occur, or that what they do hear does not challenge the leader s policies, then public threats do not carry with them much potential political cost to leaders. Without costs, there is no increased credibility. 23

25 This study contributes to an emerging literature explicating the domestic nuances of the audience cost argument (e.g. Horowitz and Levendusky forthcoming; Tomz 2007). We believe that scholars of international conflict need to take seriously domestic processes that underpin the audience costs argument. Since the link between leader s actions and the public s response is far from automatic, it behooves us to understand the institutions that shape this relationship. This includes the critical intervening role of the mass media as the primary vehicle for transmitting information between leaders and citizens. Failing to account for the implications of differences in media and electoral institutions for information transmission seems likely to perpetuate the seeming disconnect between theoretical propositions and empirical evidence that has bedeviled researchers interested in employing audience cost theory. It also limits scholars capacities to fully explicate the implications of domestic audience costs for when disputes are, or are not, likely to escalate to war, as well as the role of democratic institutions in mediating this likelihood. There are several extensions of this project that are ripe for future work. First and foremost, the way people access media is changing rapidly. It would be interesting to replicate the findings uncovered for more recent years while including a measure of internet access, which would serve to assess the extent to which new media and online communication alter the relationship between individuals and the government. Unfortunately, the Internet as a means of political discourse was in its infancy in the years covered by our data, and so was highly unlikely to influence the generation of audience costs in any meaningful way. 10 Similarly, were one to investigate more recent periods it would be worth exploring the extent to which access to 10 We nonetheless re-ran our models, including controls for Internet access. This indicator predictably proved insignificant and failed to mediate the key relationships in all model specifications. 24

26 satellite television has changed the equation. The data needed to accomplish these extensions are not readily available, but as they become so it may be possible to uncover important insights about shifts in the nature of media consumption and changes in the extent to which different nations and leaders are able to signal the credibility of their international actions and commitments. 25

27 References Baum, M. A. 2004a. "Going Private: Presidential Rhetoric and the Domestic Politics of Audience Costs in U.S. Foreign Policy Crises." Journal of Politics 48 (October): b. "How public opinion constrains the use of force: The case of Operation Restore Hope." Presidential Studies Quarterly 34 (2): Baum, M., and T.J. Groeling War stories: The causes and consequences of public views of war. Princeton, NJ: Princeton University Press. Baum, M., and Y. M. Zhukov "What Determines the News About Foreign Policy? Newspaper Ownership, Crisis Dynamics and the 2011 Libyan Uprising." In International Studies Association Annual Meeting. San Diego, CA. Bennett, W. L "Toward a theory of press- state relations in the United States." Journal of Communication 40 (2): Bennett, W. L., R. G. Lawrence, and S. Livingston "None Dare Call It Torture: Indexing and the Limits of Press Independence in the Abu Ghraib Scandal." Journal of Communication 56 (3): Berinsky, A. J "Assuming the costs of war: Events, elites, and American public support for military conflict." Journal of Politics 69 (4): Choi, Seung- Whan, and Patrick James "Media Openness, Democracy and Militarized Interstate Disputes." British Journal of Political Science 37: Cox, G. W "Centripetal and Centrifugal Incentives in Electoral Systems." American Journal of Political Science 34 (4): Dow, J. K "A Comparative Spatial Analysis of Majoritarian and Proportional Elections." Electoral Studies 20:

28 Downs, A An Economic Theory of Democracy. New York: Harper. Entman, R. M Projections of Power: Framing News, Public Opinion, and U.S. Foreign Policy. Chicago: University of Chicago Press. Fearon, J. D "Domestic Political Audiences and the Escalation of International Disputes." American Political Science Review 88 (3): Gartzke, E., and Y. Lupu "Still Looking for Audience Costs." Security Studies 21 (3). Golder, M "Democratic electoral systems around the world, " Electoral Studies 24 (1): Holsti, O. R Public Opinion and American Foreign Policy. Ann Arbor MI: University of Michigan Press. Horowitz, M. C., and M. Levendusky. forthcoming. "When Backing Down Is the Right Decision: Partisanship, New Information, and Audience Costs." Journal of Politics. Iyengar, S., and D. R. Kinder News that Matters: Television and American Opinion. Chicago: University of Chicago Press. Krosnick, J. A., and D. R. Kinder "Altering the Foundations of Support for the President through Priming." American Political Science Review 84 (2): Kumlin, S "Ideology- Driven opinion formation in Europe: The case of attitudes towards the third sector in Sweden." European Journal of Political Research 39 (4): Laakso, M., and R. Taagepera "The Effective Number of Parties: A Measure with Application to West Europe." Comparative Political Studies 12 (1):3. Lake, D. A., and M. A. Baum "The Invisible Hand of Democracy: Political Control and the Provision of Public Services." Comparative Political Studies 34 (August):

29 Larson, E. V "Putting Theory to Work: Diagnosing Public Opinion on the US Intervention in Bosnia." In Being Useful: Policy Relevance and International Relations Theory, ed. M. Nincic and J. Lepgold. Ann Arbor: University of Michigan Press. Lijphart, A Patterns of democracy. New Haven: Yale University Press. Livingston, S., and W. L. Bennett "Gatekeeping, Indexing, and Live- Event News: Is Technology Altering the Construction of News?" Political Communication 20: Lupia, A., and M. D. McCubbins The Democratic Dilemma: Can Citizens Learn what They Need to Know?: Cambridge University Press. Milner, H Civic Literacy. Hanover: Tufts University Press. Moosbrugger, L.. n.d. "Institutions, Information, and Political Sophistication: A Causal Model." University of California, Santa Barbara. Popkin, S. L "Information Shortcuts and the Reasoning Voter." In Information, Participation, and Choice: An Economic Theory of Democracy in Perspective, ed. B. Grofman. Ann Arbor: University of Michigan Press. Potter, P. B., and M. Baum "Democratic Peace, Domestic Audience Costs, and Political Communication." Political Communication 27 (4). Schmitt- Beck, R "Mass communication, personal communication and vote choice: The filter hypothesis of media influence in comparative perspective." British Journal of Political Science 33 (02): Schultz, K. A "Domestic opposition and signaling in international crises." American Political Science Review 92 (4): "Looking for Audience Costs." Journal of Conflict Resolution 45 (1):

30 Sheafer, T., and G. Wolfsfeld "Party Systems and Oppositional Voices in the News Media A Study of the Contest over Political Waves in the United States and Israel." The International Journal of Press/Politics 14 (2): Slantchev, B. L "Politicians, the Media, and Domestic Audience Costs." International Studies Quarterly 50 (2): Smith, A "International crises and domestic politics." American Political Science Review 92 (3): Sniderman, P. M., R. A. Brody, and P. E. Tetlock Reasoning and choice. Cambridge: Cambridge University Press. Tomz, M "Domestic Audience Costs in International Relations: An Experimental Approach." International Organization 61 (Fall): Weeks, J.L "Autocratic audience costs: regime type and signaling resolve." International Organization: Zaller, J The Nature and Origins of Mass Opinion. New York: Cambridge University Press "Positive Constructs of Public- Opinion." Critical Studies in Mass Communication 11 (3): Zaller, J., and D. Chiu "Government's Little Helper: U.S. Press Coverage of Foreign Policy Crises, " In Decisionmaking in a Glass House, ed. B. L. Nacos, R. Y. Shapiro and P. Isernia. New York: Rowman and Littlefield. 29

31 TABLE 1. Logit Analysis of Likelihood of Dispute Reciprocation, Violent Reciprocation and Initiator Victory as ENPP and Television Access Vary Model 1 Model 2 Model 3 Reciprocation Violent Recip. Victory β/(se) β/(se) β/(se) ENPP * (0.255) (0.246) (1.105) TV Access (0.003) (0.003) (0.006) ENPPxTV ** ** 0.005* (0.001) (0.001) (0.002) Major - Major Dyad (0.973) (0.969) Minor - Major Dyad (0.647) (0.699) (1.207) Major - Minor Dyad * (0.655) (0.649) (1.400) Initiator Capability Share (0.725) (0.771) (2.118) Contiguity *** (0.419) (0.396) (0.479) Ally (0.604) (0.594) (1.052) Alliance Portfolio Similarity (0.875) (0.944) (1.268) Status Quo Initiator (1.409) (1.496) (1.290) Status Quo Target (1.466) (1.559) (2.151) Territory (0.499) (0.513) (1.404) Regime * (0.790) (1.244) (1.825) Policy (0.415) (0.438) (1.471) Other (1.210) (1.447) Child Mortality (-0.006) (-0.006) (-0.018) N <0.10, * p<0.05, ** p<0.01, ***p<0.001 Note: Estimates are maximum likelihood coefficients obtained from logit equations with the militarized intersate dispute as the uit of anlaysis. Standard errors (clustered by initiator) are parentheses. Constants are suppressed to conserve space. 30

32 TABLE 2. Probability of Reciprocation, Violent Reciprocation, and Victory, as Number of Parties and TV Access Vary from One Standard Deviation Below to One Standard Deviation Above the Mean Values Among Democracies All Reciprocation Low TV Access High TV Access Diff Low-ENPP High-ENPP ** Low-ENEP High-ENEP ** Violent Reciprocation Low-ENPP High-ENPP ** Low-ENEP High-ENEP ** Victory Low-ENPP High-ENPP * Low-ENEP High-ENEP ** *p<.05, **p<.01 31

33 TABLE 3. Ordered and Unordered Multinomial Logit Analyses of Effects of Variations in Parties and Media Access on Dispute Reciprocation Ordered Logit Multinomial Logit a Non-violent Reciprocation Violent Reciprocation ENPP (0.228) (0.405) (0.266) TV Access ( ) ( ) ( ) ENPP x TV Access ** ** ( ) ( ) ( ) Child Mortality ( ) ( ) ( ) Major-Major Dyad *** (0.991) (1.103) (0.998) Minor-Major Dyad (0.629) (0.941) (0.727) Major-Minor Dyad (0.617) (0.931) (0.696) Initiator Capability Share (0.629) (1.089) (0.807) Contiguity (0.399) (0.653) (0.428) Ally (0.569) (0.942) (0.637) Alliance Portfolio Similarity (0.859) (1.433) (0.978) Status Quo Initiator (1.293) (1.593) Status Quo Target (1.390) (1.914) (1.679) Territory (0.442) (0.675) (0.554) Regime (0.794) (1.031) (1.235) Policy (0.406) (0.678) (0.449) Other *** (1.388) (0.742) (1.367) Constant (1.164) (2.211) (1.367) Constant * (1.173) N Robust standard errors in parentheses; *** p<.001, ** p<.01, * p<.05, + p<.10 a omitted category is non-reciprocation 32

34 TABLE 4. Probability of Reciprocation, Violent Reciprocation, and Victory, as Number of Parties and TV+Radio Access Vary from One Standard Deviation Below to One Standard Deviation Above the Mean Values Among Democracies All Reciprocation Low TV +Radio Access High TV+Radio Access Diff Low-ENPP High-ENPP ** Violent Reciprocation Low-ENPP High-ENPP ** Victory Low-ENPP High-ENPP p<.10, **p<.01 33

35

36 1

37 2

38 3

39 4

40 5