1
Definitieve versie Bachelor eindproject
Door: Bart van de Giessen; s1864033
Docent: F.E. Bakker
Tweede lezer: Michael Meffert
Vak: Bachelor afsluitend project: Politieke psychologie binnen de Internationale Betrekkingen
Werkgroep: BAP werkgroep 2
Woorden: 7502
2
Testing the micro-foundations of the audience cost theory
Introduction
The study field of international relations studies how states interact with other states and why they interact in the way they do (Walt, 1998). Domestic politics influences the way states behave in the international politics; e.g. the internal structures of a state, what ideology the leadership of the state has, if the leadership is politically secure, all can influence the way a state interacts with other states (Lumsdaine, 1998; Kaarbo, 2015). The way the domestic politics influence states in their international relations has been studied in several middle-range theories. The middle range theory this paper uses to analyse the domestic-international interaction is the audience cost theory (ACT), as theorized by Fearon (1994, 1997). This theory analyses the interaction between the leadership and his audience in international crises. This thesis assesses one of the micro-foundations of the ACT. Micro-foundations are the assumptions a theory makes on the micro-level; about the psychologies and behaviours of individuals (Kertzer, 2017; Stein, 2017; Hoffman, Agnew, Lehmiller & Duncan, 2009).
The ACT has been coined by Fearon (1994, 1997). He theorized that states’ leaders and their audiences are in a principal-agent relation and that the principals are concerned whether “the leadership is successful or unsuccessful at foreign policy’’ (Fearon, 1994, p. 577). If the leadership is not successful he would be punished by his audience with the audience costs. In democratic states the audience would control the leadership on being consistent, therefore democracies should be more credible negotiating partners, than non-democratic states. This could explain phenomena like the democratic peace (Schultz, 2001a; Potter & Baum, 2010). This theory, as will be explained below more elaborately, is based on several micro-foundations about the leadership, their domestic audience and the leadership-audience relationship. The theory assumes this is most important for the audience to perceive the leadership as successful at foreign policy. This thesis analyses the ways a leader can achieve to be perceived as successful and the factors that are of importance for this end. The ACT assumes that the audience places the utmost importance in consistency of leadership behaviour, but this primacy has not been empirically tested. The research question: Is consistency in behaviour of the leader an important factor to be perceived as successful by the domestic audience?
First, a theoretical framework will be laid down, explaining the ACT and focussing on the influence of the domestic politics on the international politics according to this theory.
3
Secondly, the micro-assumptions this theory makes about the behaviour and psychology about the leadership, the audience and the audience-leadership relation will be discussed. Herein focussing mostly on the micro-assumptions about the audience as how it perceives the leadership is the question of this research. Third, the research method will be discussed. Fourthly, the found results will be discussed and their practical and theoretical implications. Lastly concluding that the evidence this research brings against the ACT is an interesting cause for sequel research.
4
Theoretical framework
The audience cost theory
Fearon (1997) theorized that leaders in international crises to make ‘costly signals’ about their preferences to their negotiating partner in order to be a credible negotiating partner. Non-costly signals are not taken seriously because the sender of the signal does not pay any form of costs for their signals and can therefore send as many signals as possible without consequences. These non-costly signals will be disregarded as ‘cheap talk’ by the negotiating partner (Fearon, 1997). Fearon (1997) identified two forms of costly signals leaders can make; tying their hands with audience costs or making sunk costs like mobilizing the military or stationing troops far abroad (p. 82). Both types of actions have sunk costs to them; they have costs that cannot be recovered. Public statements cannot be taken back without seeming inconsistent, mobilizing the military involves paying the salaries and transport costs. This thesis focusses on the audience costs. The audience costs are a form of ex ante sunk costs, meaning that the costs are felt after the crisis is resolved instead of during the crisis. The audience costs are the costs a leader makes when he speaks publically; his audience is listening to what he says and judges on whether they approve of what he said. This structure is used to explain why “crises are public events carried out in front of domestic political audiences’’ (Fearon, 1994, p. 577). If the leadership does not follow up with actions on what he has said, he will seem inconsistent to his audience, and the audience will disapprove of this. That is why the ACT theorizes that speaking publically creates a mechanism of audience control over the leadership.
The audience cares for this consistency because audiences disapprove inconsistent leaders and hold them in lower regard because “public demands or threats engage the national honor – thus creating audience costs that leaders would pay if they backed down’’ (Fearon, 1994, p. 581). This relationship is what makes speaking publically such a ‘costly signal’; the leadership makes audience costs. The speech ties the agent his hands to his words because his audience will judge him on his consistency. Inconsistent leaders are perceived to be incompetent because they handled the national honour carelessly; needless to say that principals do not want incompetent agents (Clare, 2007, p. 735; Fearon, 1994, p. 581; Schultz, 2001b: 34; Smith, 1998; Tomz, 2007). Creating a theoretical situation wherein states, that can generate more audience costs in a shorter time period, are more credible negotiating partners in crises. They have less possibilities to renege their commitments as they will be punished. Audience costs are not only threats to the other negotiating state, as Fearon (1994) theorized, they can take every possible
5
form of public statements of the leadership. Public promises about the outcome of a crisis or public commitments to international organisations can be regarded as public speeches tying a leader to his own words, and his audience controlling him on his consistency (Chaudoin, 2014; Janusch, 2016, p. 508, 2018, p. 222).
Democratic states have more possibilities for the audience to control whether the leader is consistent or inconsistent in his behaviour. Therefore Fearon (1994) theorized that leaders of democratic states make more audience costs when acting publically in international crises or negotiations. In democracies the domestic audience will have an incentive to remove an inconsistent leader who jeopardized the national honour to restore the national credibility (Guisinger & Smith, 2002). Hence, Fearon (1994) hypothesized that democracies would make more credible negotiating partners since their audience costs would be larger. Hereby Fearon (1994) suggested a solution to the democratic peace which had a lot of traction and was mistakenly taken as the main point of the ACT (Schultz, 2001a, 2012; Potter & Baum, 2010).
Micro-foundations of the audience cost theory Leaders
The ACT assumes that leaders in international crises or negotiations want to send signals to the other leaders or negotiators about their resolution to solve a crisis in their preferred way. Fearon (1997) theorized that only the ‘costly signals’ will be perceived as credible; not-costly signals are cheap talk. The incentive to bluff and thereby to over-signal resolve about the preferred outcome can scare off the less resolved negotiators and bet a better deal for the leaderships state. When the other state does not back down, these bluffs have tied the leaderships hands because he cannot back down himself, due to the audience costs. This over-signalling of resolve can trap a leader in a situation wherein it is preferable to go to war instead of backing out of the crisis due to the sunk costs in the previously made costly signals. Therefore, Snyder and Borghard (2011) doubt the rationale of the leadership that bluffs, since the leader ties his own hands and thereby decreases his own flexibility of possible actions. Sending a threat commits the sender to the execution of the threat by the audience costs mechanism. This is represented in the historical record: pure bluffs are “extreme rare and couched in ambiguity’’ (Snyder & Borghard, 2011, p. 439). Leaders usually keep their threats vague to avoid being tied to their threat by reputational costs they would suffer when they would back down. Fearon (1997, p. 83) acknowledged the usual vagueness of threats but argued that these vague threats would not count as form of costly signalling, hereby excluding the vague threats from being a part of the
6
negotiations. Nevertheless these are the only form of threats made throughout history (Snyder & Borghard, 2011; Trachtenberg, 2012).
Secondly, the ACT assumes the signal-receiving leaders understand the costliness of the audience costs signals the signal-sending leader has made. The receiving leadership must understand that the sending leadership has made costs and cannot back down, and must asses the audience costs as a credible signal. Yarhi-Milo, Kertzer and Renshon (2018) found that the signal-receiving leaders understand that costliness of the signal. It does not matter whether the signal-sending state is an authoritarian state or an democracy, both are understood as equally costly to the receiving state. Even though the receivers personal psychology is relevant; more experienced leaders can better asses the signals; audience costs signals are less effectively against more hawkish inclined leaders (Yarhi-Milo, Kertzer & Renshon, 2018, p.2171).
Audiences
According to the ACT is the domestic audience is judging the leadership on whether he is successful in his actions. Tomz (2007, pp. 835-836) showed that foreign policy is commonly ranked as one of the most important criteria for being perceived as successful as a leader. Fearon (1994) suggested that the most important way to be perceived as successful by the audience is consistency (p.581). If the leadership is caught being inconsistent to other states, the other states would take the state of the leadership as less credible. Therefore, is consistency the engagement of the national honour. The domestic audience acknowledges this engagement and disapproves when a leader handles the national honour carelessly. Several experiments have validated the phenomena of increasing disapproval when a leader is unsuccessful (Tomz, 2007; Chaudoin, 2014; Levendusky & Horowitz, 2012). These experiments have not proven however, that the consistency of leadership behaviour is the most important factor for explaining the levels of approval of the audience. Fearon’s (1994, 1997) focus on consistency as the primary source of being perceived as successful is problematic. Clare (2007) argued that consistency is important to the audience, but that its importance can be outweighed by other more salient aspects of the international crisis. He argues that being perceived as successful is different from being perceived as consistent. The ‘being perceived as successful’ depends on whether the audience gives more salience to the policy outcome or to the consistency; when the audience gives much salience to the policy outcome of the crisis, the consistency is less important. The audience judges the leader’s competence on the basis of whether he defended their interests successfully. If they give no salience to the topic of the crisis, the audience judges him on his consistency (Clare, 2007, p. 735). Hereby Clare (2007) created a second basis from what the approval could
7
originate. Brutger and Kertzer (2018) add to this that the individuals’ perception of the leader’s actions and the leader’s competence depends on ideological predispositions they already have, independent of the information or the situation. They argue that the reputational logic behind arguments as ‘consistency of leadership behaviour is the engagement of the national honour’ is a unidimensional approach to the influence the reputation has (Brutger & Kertzer, 2018, p.694; Kertzer and Zeitzoff, (2017, p. 554). Including the salience the individual gives to the topic of the crisis and the individual’s previous held predispositions into the analysis the reputational logic, give the ACT a multidimensional approach to the reputational logic whereon the ACT is based. Encompassing these psychological “cogs and wheels’’ of the of the individuals broadens the theory (Kertzer, 2017, p.81; Clare, 2007; Brutger & Kertzer, 2018; Kertzer & Zeitzoff, 2017).
A second assumption the ACT makes about the domestic audience is that the logic behind the audience costs is a homogenous logic. Every individual has different psychological predispositions, as above discussed, and these predispositions influence the way the individual processes and interprets the situation (Mercer, 2010; Jervis, 2006). Brutger and Kertzer (2018), Kertzer and Brutger (2016) and Bakker (2017; 2018) have shown that there is a relevant difference in the interpretation of international politics by hawkish and dovish respondents. These two ideal types are different in several ways; hawks tend to be less trusting of other international states and more inclined to embrace deterrence mind-sets. Therefore, it is important for hawks to have a reputation of following up on threats, so they will be taken seriously when making threats. In a perfect typology, doves are less militant, more trusting, more internationally trusting an cooperating (Brutger & Kertzer, 2018, p. 701). These characteristics combined make doves more prone to disapprove for reasons like belligerence than consistency. This divides the audience in the more hawkish inclined audience and the more dovish inclined audiences. Kertzer and Brutger (2016) showed that the level of hawkishness influences the way the respondents react to different leadership behaviours in an international crisis as theorized by the ACT. This divide influences the way the audience costs can be studied by experiments. With this divide the reasons for approval differ between groups in the audience. In a situation wherein leader gets involved in an international crisis, makes threats and backs down later, the doves would disapprove ef the involvement the international crisis at all, because they dislike belligerence; the hawks would disapprove of the reneging on the made threats, for they fear that in the future their state would be taken less seriously in their threats (Kertzer & Brutger, 2016; Brutger & Kertzer, 2018, p. 701). These differences in reasons for
8
disapproval of leadership behaviour make the audience costs mechanism tested by Tomz (2007) and Chaudoin (2014) a double barrelled one. To understand that the audience disapproves of the leadership behaviour it is important to study why the audience disapproves of the leadership behaviour. The reason why the audience disapproves can be related to the salience a respondent gives to the topic of the crisis or be related to the preferences he has for a certain set of behaviour of the leadership (Clare, 2007; Brutger & Kertzer, 2018; Kertzer and Brutger, 2016). Note that ‘perceiving the leadership as successful in international politics’ can originate from every facet of international politics the respondent finds salient, not excluding to different reasons than the reputational logics or policy outcomes of the international politics.
The audience-leader relationship
The audience costs theory assumes that audiences punish leaders who renege their threats, but the empirical record does not support that assumption (Hoffman et al., 2009; Trachtenberg, 2012). This could be for the several reasons. As mentioned above, leaders make vague threats instead of committing ones (Fearon, 1997; Snyder & Borghard, 2011). Levendusky and Horowitz (2012) showed that the domestic political complexities influence whether the audience penalizes the leadership for inconsistencies. Justifying leadership actions by pointing out the national interests or point to new information decrease the influence of the audience costs mechanism that penalizes inconsistencies. Thereby Levendusky and Horowitz (2012_ showed that the reasons for absence of a historical record of the audience costs mechanism can be accounted by the possibilities the leadership has to play down the audience costs or even sidestep them (Snyder & Borghard, 2011; Levy, 2012; Schultz, 2001b, 2012). As an auxiliary hypothesis Fearon (1994) stated that democracies would be more credible negotiating partners because their audience would easier punish their leaders. Hereby he suggested that the ACT mechanism could be an explanation for the democratic peace. Weeks (2008) proved him wrong; leaders in non-democracies do have audience costs but they have a different relevant audience; different principals that punish them. This still does not reject the argument that audience costs in democracies and autocracies work differently. She argued that studying the audience costs on the basis of ‘non re-election costs’ is a too simplistic approach, also because that assumes that foreign policy is the only basis of re-election in democracies, that is an unjustified assumption.
9 Studying the micro-foundation of public preference
Levendusky and Horowitz (2012) showed that the disapproval originating from the inconsistency of leadership behaviour can be mitigated by other factors. Snyder and Borghard (2011) and Levy (2012) argued that the predominance of consistency of leadership behaviour for explaining audience, as suggested by the ACT, is debatable. Other factors like policy outcome, the justification for the leader’s inconsistency, the audience’s predispositions or salience the audience gives to the topic, play a role in whether the audience approves or disapproves of the leadership’s behaviour. My research analyses what these factors are and how influential they are. Therein expecting to find at least some disapproval of the audience when they are confronted with inconsistency, in line with the results of Tomz (2007), Chaudoin (2014) and Levendusky and Horowitz (2012).
H1: The audience disapproves of inconsistent leadership behaviour
This research also expects to find that the hawkish inclined respondents are more likely to disapprove inconsistent leadership behaviour. As Brutger and Kertzer (2018) argued, hawks are expected to be more concerned with the reputation of being consistent with military threats. Therefore, I expect to find an interaction between the hawkishness of a respondent when he is exposed to some form of inconsistency by the leadership. More hawkishness respondents should, according to Brutger and Kertzer (2018), be more disapproving of inconsistency than doves.
H2: Hawkish audiences disapprove inconsistent leadership behaviour more than dovish audiences
10
Method
This research studies the effects of inconsistent leadership behaviour on the approval levels of respondents. As an auxiliary it analyses if the more hawkish inclined respondents disapprove more of inconsistent leadership. Therefore, it is measuring both the approval levels and the hawkishness of respondents in two different situations. In one situation the leadership has made public promises about the outcome of an international crisis; the other situation where the leadership has not made these promises. In both situations the crisis ends with the same compromise. For studying the effects of the inconsistency on the approval levels of the audience, an experimental approach is used. This approach enables studying the causal mechanism of one singled out variable without the interference of variables out of the researchers control (Hyde, 2015; McDermott, 2002; Mintz, Yang & McDermott, 2011). That is the reason why this method is useful for this experiment. The causal relation between the dependent variable: ‘approval of the audience’ and the independent variable: ‘presence or absence of audience costs’ can be singled out and examined. The research instrument that was created for this purpose and the relevant variables will be discussed below.
Research instrument
This inquiry of the effects of audience costs on a respondents approval of leadership behaviour was designed and carried out as an experimental survey. The experiment was distributed to 64 respondents. The respondents got a story describing a hypothetical international border dispute. The border dispute is between two very similar countries existing on an island somewhere in the ocean, one of these countries is ‘Mycountry’ the other is ‘Othercountry’. In the disputed area are oil fields, the oil could stimulate the economies of both countries. The border dispute between the two countries is longer standing, but since oil has been discovered in the disputed area the dispute has become more salient and the countries want to define the borders now so they can begin to exploit the oil fields. In the hypothetical scenario a new prime minister was recently elected in ‘Mycountry’. This new prime minster had made all kind of promises about how he would manage the situation in ‘Mycountry’ during his election campaign.
Treatments
The first treatment, and the ‘baseline’ of the research, is describing the election of a new prime minister who has run his successful election campaign on promises of economic reforms. This leader wants to make the economy futureproof and independent from foreign investors by developing the high-tech and tourist industries. He does not only want to stimulate the economy,
11
but he also wants to decrease the citizens’ economic burdens. This is added to not alienate and find appreciation for his plans from all sides of the political spectrum. This treatment is used as the baseline because the leadership has not made any form of audience costs about the border dispute.
The second treatment, the ‘audience costs treatment’, describes the election of a new prime minister who has run his successful election campaign on the promise of a hard-line approach to the border dispute. He argues that the attainment of the entire area is of vital importance to ‘Mycountry’. He makes several speeches in which he makes promises about the outcome of the crisis, he promises that he will attain the whole disputed area for ‘Mycountry’. The prime minister even announces, as ‘pure bluff’, that he is willing to send the military in if he is not content with the outcome of the negotiations. Both the speeches and the threat to send in the military create a situation in which there would be audience costs if he does not delivers on his promises to attain the whole disputed area.
Compromise
The border dispute ends with a compromise between the two nations. ‘Mycountry’ attained one third of the disputed territory, ‘Othercountry’ two-thirds. The prime minster of ‘Mycountry’ states that he is content with the outcome of the negotiations. Due to this compromise the respondents from the audience costs treatment will face a situation in which ‘Mycountry’ has not attained the whole disputed area as the prime minister promised. Since, they are faced with an inconsistency between the promises the leadership made about the outcome of the crisis and the actual outcome they should disapprove of the outcome and the leadership behaviour.
In theory the more hawkish inclined respondents should approve less of this compromise because they are less trusting of other states and more militant assertive. Especially when ‘Mycountry’ got less than half of the disputed area, the hawkish respondents should think of him as not assertive enough. In the case of the audience costs treatment, more was promised, creating an interaction between the audience costs and the hawkishness. Doves should theoretically approve more of the prime minister’s behaviour because he handled the crisis diplomatically due to their cooperative inclinations and trusting predispositions (Brutger & Kerzter, 2018, p. 701)
12 Variables
Approval of the audience
The approval of the audience and the reason for the approval of the audience are the study subjects of this experiment. Previous research asked only generally “Do you approve, disapprove, or neither approve nor disapprove of the way the U.S. prime minister handled the situation?’’ (Tomz, 2007, p. 825). Levendusky and Horowitz (2012) measured ‘approval’ also as “approval of the president’s decision’’ (p. 328). Hereby they ignored Snyder and Borghard (2011) and Levy (2012) their criticism: the approval of the audience can be determined by the outcome of the crisis. Therefore, the approval of the respondents was measured twofold in this experiment. Firstly, by asking the respondents whether they approve of the behaviour of the leadership. Secondly, by asking whether the respondents approve of the outcome of the international crisis. The dependent variable questions were: “Do you approve, disapprove or neither approve or disapprove of the outcome of the crisis as in the story above described?’’ and “Do you approve, disapprove or neither approve or disapprove of the behaviour of the leadership as in the story above described?’’. Both questions are asked on a one to seven scale, for the extent to which they agree or disagree with the question. One was described as “totally disapproving’’, seven was described as “totally approving’’, and four was described as “neither approving nor disapproving’’. Measuring approval twofold excludes the possibility that the approval is only caused by the outcome of the border dispute or that the audience cares more for the way the leadership behaved. Both factors are measured to control on different dimensions for the origins of the approval.
Audience costs
In the research instrument above described, the leadership of the first situation has made no promises or threats concerning the crisis. Therefore this situation will be described as the baseline of the experiment. In the second situation the leadership has made audience costs by acting inconsistent with the promises he made. Therefore this situation will be discussed as the audience costs treatment. The respondents of the audience costs treatment should punish the leadership with the audience costs for this inconsistent behaviour, as he made promises and threats concerning the crisis but did not keep his promises. The audience costs in the treatment consist of the costs in the inconsistency between the public promises of the leadership to attain the whole disputed territories and the threat to invade the other country if he is not content with the outcome of the international crisis. Accepting the compromise is inconsistent with the
13
promised outcome. If H1 is correct, the audience in the audience costs treatment should approve less of this leadership behaviour. Furthermore, the more hawkish inclined respondents should approve less of the inconsistency of the behaviour of the leadership in the second treatment.
Hawkishness
This experiment wants to measure how the hawkishness of the respondents interacts with the audience costs treatment. The set of measurements of ‘hawkishness’, made by Kertzer and Brutger (2016), is used. Hawkishness herein is a combination of the respondents predispositions about militant assertiveness, international trust, national chauvinism and political orientation. All questions measuring hawkishness were asked on a one to seven scale whether they agree or disagree with a sentence. This is the same one to seven scale as was used to measure the approval of the respondents. The question about political orientation is measured on a socialist-liberal-conservative scale, since it impractical to measure that on an approval scale. If H2 is correct there should be an interaction between hawkishness and the audience costs treatment. When confronted with the audience costs treatment, the more hawkish respondents will disapprove more because they fear that future threats will not be taken serious.
The questions from Kertzer and Brutger (2016) have been adjusted to a non-United States context in national chauvinism and political orientation. Note that the questions are translated to Dutch and are not a perfect copy, this was done by myself because there were no official translations in Dutch.
Procedure
The experiment was conducted in November 2019 and consisted of a convenience sample of personal acquaintances. These respondents were evenly and randomly assigned one of the treatments discussed above. Half of the respondents have been assigned the audience costs treatment and the other half the baseline treatment. The sample of respondents was a convenience sample. I distributed the experimental survey over WhatsApp to mostly group chats. I also asked some friends and my parents to help me distribute the link, herein asking politely to carefully read the story and answer the questions. The experiment was prepared and distributed in Qualtrics, with the treatments randomised, evenly presenting the elements and randomly presenting one of the treatments.
14
Results
Descriptives of the respondents
There were 64 valid respondents for this experiment. From these 64 respondents, 34 identified as ‘male’, 28 as ‘female’ and 2 respondents identified themselves as ‘other’. The educational levels were: 45 respondents have had some form of university education, 7 are in VWO, 9 are doing HBO, 1 in MBO and 1 in VMBO. The average respondent was 28.5 years old (M= 28.5, SD=14.77).
The analytic model and the preparation of the data
To analyse the influence of the independent variables of the audience costs treatment and the hawkishness of the respondents on the dependent variables of approval for the leaderships behaviour or crisis outcome, the data need to be prepared for a linear regression analysis. The linear regression model is used to find what influences a change in the independent variable has on average on the dependent variables In this research the predictors are the audience costs treatment and the hawkishness of the respondent (Field, 2013, pp. 293-329). The audience costs treatment and the hawkishness variables were recoded for this model. What treatment the respondents got was recoded as ‘0’ for the baseline treatment and for the audience costs treatment ‘1’. The hawkishness of the respondents was measured with the questions from the research of Kertzer and Brutger (2016). The scales of the second question on military assertiveness and the second variable on national chauvinism are reversed (appendix 1). The hawkishness variables were tested on internal consistency; whether they measure the same psychological inclinations of the respondents. The variables had a high Cronbach’s alpha of 0.826 and measure the same psychological inclinations in the respondents, therefore they are merged into the variable ‘hawkishness’. Thus the level of hawkishness a respondent has is the average of what he responded on the six questions that measure hawkishness (table 1). The level of ‘hawkishness’ was on average 3.15 (M= 3.15, SD= 1.18).The approval of the respondents for the behaviour of the leadership and for the outcome of the crisis correlates strongly, Pearsons r = 0.675, p<0.01). The difference was found for the approval of the behaviour of the leadership behaviour (M = 3.5; SE = 1.497) and the approval of the outcome of the crisis (M = 3.5; SE = 1.522). This finding was not statistically significant, t(63) > 0.001, p = 1.000.
15 Table 1. Corhnbahs alpha of the items used for the Hawkishness scale
Linear regression results
In the first model only the predictor of the audience costs treatment is incorporated in the analysis, in the second model the audience costs treatment and the hawkishness are used as predictors. Only the hawkishness has an effect on the approval of the respondents. The predicting value from interaction between the hawkishness and the treatment variable was neglectably small (b = 0.086) and not significant (p = 0.788). Due to its collinearity with the audience costs variable it was taken out of the linear regression model (Table 2.1). When analysing the analysis of variance of the of both models there is no significant predictive value of predictors on ‘the approval of the behaviour of the leadership’; model 1 (F = 0.472(4,57) p = 0.756) and for model 2 (F = 1.907(5,56) p = 0.108). These non-significant F-values and the low R2 indicating that the predictors in model 1 only explain 3.2% of the variation in the
Items a if item deleted
De beste manier om vrede te behouden is door sterke militaire capaciteiten, hierdoor kunnen andere landen je niet bedreigen of manipuleren.
,815
Het gebruik van militairen maakt problemen vaak erger en lost internationale conflicten vaak niet op. (scale reversed)
,809
Oorlog is een vervelende situatie, maar soms is het de enige oplossing voor internationale problemen.
,807
Andere landen zijn over het algemeen niet te vertrouwen, ze zullen als ze kunnen proberen hun territorium en macht proberen uit te breiden.
,785
Hoe veel beter vindt je jouw eigen land (in de echte wereld) ten opzichte van andere landen)
,799
Er zijn dingen van mijn eigen land (in de echte wereld) waar ik me voor schaam. (scale reversed)
,788
Als je aan politieke stromingen denkt, hoe zou je je eigen standpunt omschrijven?
,815
16
approval of the leadership (R2 = 0.032). For model 2 this is neither a significant R2 (R2 = 0.146). Therefore, this model does not have value for explaining the variance in the dependent variable.
Table 2.1. Model of the linear regression of the amount of approval for the leadership behaviour 1
Model 1 Model 2
(Constant) 2.443 4.637**
(1.316) (1.485)
Audience costs treatment -0.147 -0.373
(0.384) (0.373) Hawkishness -0.497** (0.182) Level of education 0.231 0.164 (0.212) (0.203) Age -0.009 -0.024 (0.014) (0.015) Gender 0.077 0.534 (0.382) (0.399) R2 0.032 0.146 R2 adj. -0.036 0.053 N 64 64
Note: MLS-regression coefficients with standard errors in parentheses. ***p <0.001, **p <0.01, *p <0.05.
The two models are constructed the in the same way as the 2.1 table and the interaction of hawkishness and the audience costs treatment are excluded for the same reason, with an influence of (b = -0.058) and a significance of (p = 0.859). What does stand out in this table with respect to the 2.1 table is that the level of education influences the approval of the audience in the first model (p <0.1), and the influence of age in the second model on the approval levels (p <0.1). When analysing the analysis of variance of the of both models there is no significant
1 The model of linear regression of the amount of approval for the leadership behaviour with the interaction
17
predictive value of predictors on ‘the approval of the outcome of the crisis ’ in the first model; model 1 (F = 1.401(4,57) p = 0.245). For the second model the predictors can explain a significant portion of the variance in model 2 (F = 2.894(5,56) p = 0.022). This second model of the influence of the predictors on the approval of leadership behaviour is the only model wherein the predictors can actually explain the amount of approval of the audience.
Table 2.2. Model of the linear regression of the amount of approval for the outcome of the crisis 2
Model 1 Model 2
(Constant) 2.008 4.261**
(1.298) (1.456)
Audience costs treatment -0.278 -0.509
(0.378) (0.366) Hawkishness -0.510** (0.179) Level of education 0.387* 0.319 (0.209) (0.199) Age -0.014 -0.028* (0.014) (0.014) Gender 0.210 0.259 (0.377) (0.391) R2 0.090 0.205 R2 adj. 0.026 0.134 N 64 64
Note: MLS-regression coefficients with standard errors in parentheses. ***p <0.001, **p <0.01, p <0.05.
In both the 2.1 table and the 2.2 table the intercept is significant in the second model. The other variable that stands out is the variable for hawkishness, as a change of one in the hawkishness levels for a respondent predicts 0.474 less approval for the leadership behaviour and 0.543 less approval for the outcome of the crisis. The analysis of variance of the models is notable only in
2 The model of linear regression of the amount of approval for the outcome of the crisis with the interaction
18
the second model of the dependent variable of approval for the outcome of the crisis predicts the variance in the model correctly.
19
Discussion
This experiment set out to examine the influence of the audience costs on the approval of the audience. And as an auxiliary, it examined whether there is an interaction of the audience costs treatment and the hawkishness of the respondents on the approval of the audience. Herein the audience was presented with a fictitious international crisis and outcome of this crisis and they were randomly presented with either a prime minister that made audience costs or a prime minster that had not made audience costs. Herein, based on previous research by Tomz (2007), Chaudoin (2014) and Levendusky and Horowitz (2012), an interaction was expected between the approval of the audience and the audience costs. Taking the criticism of Levy (2012) and Snyder and Borghard (2011) on the ACT was used to split the dependent variable of approval in two. One dependent variable measuring the approval for the
leaderships behaviour, the other dependent variable measuring the approval for the outcome of the crisis. With both dependent variables of the experiment there were no effects of the audience costs treatment measured. Based on the research of Kertzer and Brutger (2016) and Brutger and Kertzer (2018) an interaction was expected between the hawkishness of the respondents and the audience costs treatment. No such interaction was measured with both dependent variables. The possible reasons and meaning of the absences of the expected results will be discussed here.
This research separated the dependent variable of ‘the approval of the audience’ into the ‘approval of the audience of the leadership behaviour’ and into the ‘approval of the audience of the crisis outcome’. This separation created a possibility to study what the reason for the overall approval is. The difference between both dependent variables was not significant. The regression coefficients were similar for the lack of influence of the audience costs treatment and the similar predictive value of the change of one on the hawkishness scale of the
respondents from -0.510 for the approval of the outcome of the crisis to respectively -0.497 for the approval of the leadership behaviour. Nonetheless, in the models for the approval of the crisis outcome the control variables of age and education levels seemed to play a role as predictors. This was not the case in the models for the approval of the leadership behaviour (p<0.1). The separation of the two concepts of approval created a possibility to examine whether the approval of the audience stems from the outcome of the crisis or from the leadership behaviour, in accordance with the criticisms of Levy (2012) and Snyder and Borghard (2011). As the both dependent variables correlated strongly, had no differences in a paired t-test and both variables had predictors of similar strength and significance in a
20
multivariate regression analysis, it is reasonable to expect an overlap in the cause for these approval variables. This experiment did not study the cause of the overlap these variables had. It is possible that the approval for the leadership behaviour is based on the crisis outcome, it is also possible that the approval for the crisis outcome is based on how the leadership attained this outcome. However, the both dependent variables did not differ from each other
In theory, the approval of the respondents for the leadership behaviour or crisis outcome is expected to be related to the salience the respondent gives to the crisis. If the salience the respondent gives to the crisis outcome is higher than the salience the respondent gives to consistent leadership behaviour, in such a case the effects of the audience costs should be less (Clare, 2007). This experiment failed to include a control for how salient the respondents perceived the hypothetical international crisis to be, nor encompassed whether the audience disapproved of the audience costs treatment for the reason of the inconsistent leadership behaviour. It is therefore possible that the reason for the absence of the expected audience costs mechanism is caused by the higher perceived salience the respondents gave to the topic of the international crisis than they gave to whether the leadership was consistent or not. Due to the multidimensionality of the cause of the approval of the respondents it is possible that it stems from another factor in the experiment not controlled for.
The hawkishness of the respondents was measured with the same questions used by Kertzer and Brutger (2016). These measured a cohesive set of psychological inclinations, as the high Crohnbach’s alpha indicates. The more hawkish inclined respondents approve less of the outcome of the crisis and of the behaviour of the leadership. This is due to their higher militant assertiveness, their lower trust in other states and international cooperation and their higher national chauvinism. This is not, as Brutger and Kertzer (2018) expected due to their reputational logic. They theorized that hawks would be more disapproving of inconsistency due to their concern with the effectiveness of future threats (Brutger & Kertzer, 2018, p. 701). Therefore, an interaction was expected between the effects of audience costs treatment and the hawkishness of the respondents. This interaction was not found in the results. This could be caused by the absence of the audience costs mechanism, which can be absent for above discussed reasons. This interaction could also be absent due to faulty expectations of the focus on consistency of hawkish respondents; the measurements of hawkishness focus on militant assertiveness, international trust, national chauvinism and political ideology do not directly measure a focus on consistency with previously made threats. This focus was expected to be there due to the hawks militant assertiveness; if a state reneges on its threats, the threats the
21
state makes in the future will be taken less seriously (Brutger & Kertzer, 2018, p. 701). The militant assertiveness of the hawkish respondents could be more instrumental and result in an understanding of the possibility to use a threat as an assertive negotiating instrument to get a better result. Therefore, the hawkish respondents could even approve of this pragmatic militant-assertive strategy, instead of disapproving of the inconsistency. Coming back to the omission to incorporate questions to control the supposed audience costs mechanism; in this case whether the respondents disapprove of the inconsistency in the audience costs treatment.
Overall this research expanded the notion of the ‘approval of the audience’ into the two segments of ‘approval of leadership behaviour’ and ‘approval of outcome of the crisis’, finding no difference between both variables. Both dependent variables found that the
hawkishness of the respondent is the only variable having an influence on the approval of the respondents. The audience costs treatment and the interaction between the hawkishness of the respondent and the audience costs treatment both had no influence. The exact reason for this absence of influence is unclear due to the omission of control variables controlling for whether the audience costs mechanism was perceived by the respondents as was intended.
22
Conclusion
This thesis analysed the micro-foundations of the ACT. The leadership acts as the agent of the audience and needs to be perceived as successful by the audience to be kept in power (Fearon, 1994). How the leadership handles international crises and negotiations is one of the main subjects whereon the audience judges the leadership on his competence (Tomz, 2007). The audience keeps an eye on the behaviour of leadership to control him on whether they approve of his behaviour. Fearon (1994, 1997) theorized that speaking publically is therefore a
credible signal to the negotiating partner; allowing you to be controlled by your audience. This audience costs mechanism, that the audience disapproves of inconsistent leadership behaviour, has only been proven in experimental research designs (Tomz, 2007; Levendusky & Horowitz, 2012; Chaudoin 2014). Even though the audience disapproves of inconsistency, this mechanism has no historical record (Trachtenberg, 2012).
This absence of empirical record can be understood when studying the micro-foundations of the theory. The theory has assumptions about behaviour of the leadership, the audience and the audience-leadership relations that do not match with reality. The leader will behave in such ways that he will try to bypass the audience costs (Snyder & Borghard, 2011). The predominance of consistency as the ground for assessment of the leadership is unwarranted, audiences find other aspects of leadership also salient, e.g. the outcome of the international crisis (Clare, 2007). The audience is not a homogenous audience and there exist subgroups in the audience that find other aspects of leadership more salient and therefore, more audience logics exist (Kertzer & Burger, 2016; Brutger & Kertzer, 2018). Levendusky and Horowitz (2012) showed how domestic political complexities influence the audience costs mechanism, Kertzer and Zeitzoff, (2017) even argue that although audiences are influenced by political elites, they fall back on their own predispositions when in doubt. Weeks (2008) showed that the ACT does not only work in democratic states as Fearon (1994) and Schultz (2001a) theorized.
Taking the criticisms of Levy (2012) and Snyder and Borghard (2012) into consideration the ‘approval of the audience’ was separated into the two dependent variables of ‘approval of leadership behaviour’ and ‘approval of the outcome of the crisis’. Both variables did not differ, and the regression models were similar their predicting variables. This leads to the conclusion that although this is a more multidimensional tool to study the effects of the ACT mechanism, it did not found anything different between both variables. This extra tool could
23
show results in a sequel experiment testing the effects of different crisis outcomes combined with an audience costs mechanism on both the measurements of the approval of the audience and examine the difference.
This thesis showed, in contrast with the previous experimental studies, no effect of the
audience costs mechanism on the approval of the audience; in contrary to the hypothesis. The only effect the experiment found was the effect of the hawkishness of the respondents on the approval for either the leadership behaviour or the crisis outcome. The hypothesis, based on the Kertzer and Brutger (2016) Brutger and Kertzer (2018) descriptions of hawkishness, that there would be an interaction between the audience costs treatment wherein the leadership acts inconsistent and the hawkishness of the respondents. The hawkish respondents would be concerned with the reputation of the own state as consistent with its threats, and therefore disapprove extra when confronted with inconsistent leadership behaviour. The expected interaction effects were not found. The reason for these absences is ambiguous for the omission of control variables testing for whether the international crisis was perceived as salient nor for if the reneging of the leadership on his promises and threat is perceived as inconsistency of leadership behaviour. It would be prudent to replicate this experiment with control variables to appreciate the findings correctly.
24
Bibliography
Bakker, F. E. (2017). Do liberal norms matter? A cross-regime experimental investigation of the normative explanation of the democratic peace thesis in China and The Netherlands. Acta Politica 52(4), pp. 521-543.
Bakker, F.E. (2018). Hawks and doves. Democratic peace theory revisited. Leiden. (dissertation).
Brutger, R. & Kertzer, J. D. (2018). A dispositional theory of reputation costs. International Organization, 72(3), pp. 693-724.
Chaudoin, S. (2014). Promises or policies? An experimental analysis of international agreements and audience reactions. International Organization, 68(1), pp. 235-256.
Clare, J. (2007). Domestic audiences and strategic interests. The Journal of Politics, 69(3), pp. 732-745.
Fearon, J. D. (1994). Domestic political audiences and the escalation of international disputes. American Political Science Review, 88(3), pp. 577-590.
Fearon, J. D. (1997). Signaling foreign policy interests: Tying hands vs sinking costs. Journal of Conflict Resolution, 41(1), pp. 68-90.
Field, A. P. (2013). Discovering statistics using IBM SPSS statistics. London; Thousand Oaks, California: Sage publications.
Guisinger, A. & Smith, A. (2002). Honest threats: The interaction of reputation and political institutions in international crises. Journal of Conflict Resolution, 46(2), pp. 175-200.
Hoffman, A. M., Agnew, C.R., Lehmiller, J. J., & Duncan, N.T. (2009). Satisfaction, alternatives, investments, and the microfoundations of the audience cost models. International interactions, 35, pp. 365-389.
Janusch, H. (2016). The Breakdown of International Negotiations: Social Conflicts, Audience Costs, and Reputation in Two-Level Games. International Negotiation, 21(3), pp. 495-520.
Janusch, H. (2018). The Interaction Effects of Bargaining Power: The Interplay Between Veto Power, Asymmetric Interdependence, Reputation, and Audience Costs. Negotiation Journal, 34(3), pp. 219-241.
25
Jervis, R. (2006). Understanding beliefs. Political Psychology, 27(5), pp. 641-663.
Kaarbo, J. (2015). A foreign policy analysis perspective on the domestic politics turn in IR theory. International Studies Review, 17(2), pp.189-216.
Kertzer, J. D. & Brutger, R. (2016). Decomposing audience costs: bringing the audience back into audience cost theory. American Journal of Political Science, 60(1), pp. 234-249.
Kertzer, J. D. and Zeitzoff, T. (2017). A bottom-up theory of public opinion about foreign policy. American Journal of Political Science, 61(3), pp.543-558.
Levendusky, M. S. & Horowitz, M. C. (2012). When backing down is the right decision: partisanship, new information and audience costs. The Journal of Politics, 74(2), pp. 323-338.
Levy, J. S. (2012). Coercive threats, audience costs, and case studies. Security Studies, 21(3), pp. 383-390.
Lumsdain, D. (1998).The intertwining of international and domestic politics. Polity, 29(2), pp. 299-306.
McDermott, R. (2002). Experimental methods in political science. Annual Review of Political Science, 5(1), pp. 31-61.
McDermott, R. (2011). New directions for experimental work in international relations. International Studies Quarterly, 55(2), pp. 503-520.
Mintz, A., Yang, Y., & McDermott, R. (2011). Experimental approaches to international relations. International Studies Quarterly, 55(2), pp. 493-501.
Hyde, S. D. (2015). Experiments in international relations: Lab, survey, and field. Annual Review of Political Science, 18, pp. 403-424.
Mercer, J. (2010). Emotional beliefs. International Organisation 64(1), pp. 1-31.
Schultz, K. A. (2001a). Democracy and Coercive Diplomacy. Cambridge: Cambridge university press.
Schultz, K. A. (2001b). Looking for audience costs. Journal of Conflict Resolution 45(1), pp. 23-60.
26
Schultz, K. A. (2012). Why we needed audience costs and what we need now. Security Studies 21(3), pp. 369-375.
Smith, A. (1998). International crises and domestic politics. American Political Science Review, 92(3), pp. 623-638.
Snyder, J. & Borghard, E. D. (2011). The cost of empty threats: A penny, not a pound. American Political Science Review, 105(3), pp. 437-455.
Tomz, M. (2007). Domestic audience costs in international relations: An experimental approach. International Organization, 61(4), pp. 821-840.
Trachtenberg, M. (2012). Audience costs: An historical analysis. Security Studies, 21(1), pp. 3-42.
Walt, S. M. (1998). International relations: one world, many theories. Foreign Policy, (110), pp. 29-46.
Weeks, J. L. (2008). Autocratic audience costs: Regime type and signalling resolve. International Organisation, 62(1), pp. 35-64.
Yarhi-Milo, K., Kertzer, J. D., Renshon, J. & Horowitz, M. (2018). Tying hands, sinking costs, and leader attributes. Journal of Conflict Resolution, 62(10), pp. 2150-2179.
27 Appendix 1
Experiment opzet:
Begintekst
In de volgende enquête lees je een scenario over een fictief internationaal conflict. Lees deze tekst a.u.b. goed door, want daarna wordt er naar jouw mening gevraagd over hoe jij vindt dat leiders zich moeten gedragen in een conflict. Je antwoorden zijn anoniem en zullen enkel voor academische doeleinden gebruikt worden.
Algemene tekst
Een nieuw conflict is opgelaaid tussen twee landen op een eiland midden in de oceaan. Stel je voor dat je je hele leven in Mijnland woont, het andere land op het eiland is Anderland. Mijnland en Anderland zijn de enige twee landen op dit eiland en ze zijn gescheiden door een bergrug. Mijnland is door zijn gunstige ligging in de oceaan een rijk, goed ontwikkeld met grote havens. Mijnland heeft door zijn natuurlijke bronnen een overvloed aan mineralen waardoor de high-industrie zich goed heeft kunnen ontwikkelen. Ook zijn er grote gasvelden in de oceaan die met boorplatforms geëxploiteerd worden. Door de vruchtbare grond is Mijnland een dichtbevolkt land met ongeveer 9 miljoen inwoners.
Anderland is, door geografische redenen nooit veel in contact geweest met Mijnland, maar heeft een heel vergelijkbaar ontwikkelingsproces doorgemaakt. De havens zijn op exportgericht en het land heeft een grote toeristensector. Voor de rest is het land, net als Mijnland, een dichtbevolkt welvarend land met zeer veel natuurlijke bronnen en een aantal grote gasvelden in de oceaan. Er is tussen beide landen altijd een redelijk neutrale relatie geweest, voornamelijk gericht op goede handelsrelaties.
Stel je nu voor dat er in deze situatie in een bergregio een grensconflict oplaait. Dit conflict gaat over een redelijk klein gebied, waar maar weinig mensen wonen. Voor de inwoners van dit gebied, die er wel zijn, verschilt het per vallei bij welk land ze willen horen en er is geen consensus over bij welke natie ze horen. Nu er recent olievelden in dit gebied zijn gevonden, staat dit gebied plotseling in de schijnwerpers. Zowel Mijnland als Anderland vinden dat ze aanspraak maken op dit gebied. Binnenkort zijn er onderhandelingen over de verdeling van dit gebied. De nieuwe premier van Mijnland heeft een ruim mandaat in het parlement om zijn plannen uit te voeren op het moment dat hij de onderhandelingen aangaat.
28
(treatment 1 of 2 willekeurig en gelijkmatig ingedeeld) Treatment groep 1
Stel je je voor dat er in de recente verkiezingen in Mijnland een nieuwe premier is gekozen. Een charismatische leider, die Gerard heet, heeft tijdens een zeer succesvolle verkiezingscampagne beloofd het land te hervormen om het nog meer internationaal op de kaart te zetten. Door de dreigende internationale recessie is het van het grootste belang om de economie bestendiger te maken, de huidige situatie is te veel erop gericht op de bodemschatten exporteren en nog niet genoeg om deze te ontwikkelen met grote internationale merken die met bedrijven als Apple, Samsung en Shell kunnen concurreren. Door investeringen in de economie wil hij bedrijven stimuleren die zich met deze internationale reuzen kunnen meten. Ook moeten binnenlands de lasten worden verlicht voor de gewone hardwerkende man. Een nieuw sociaal programma moet de binnenlandse economie eerlijker maken om zo de binnenlandse bevolking rijk te houden en de economie duurzaam te hervormen zodat hij toekomstbestendig is. Jij gelooft in Gerards plannen en hebt ook op hem gestemd.
Treatment groep 2
Stel je je voor dat er in de recente verkiezingen in Mijnland een nieuwe premier is gekozen. Een charismatische leider, die Gerard heet, heeft tijdens een zeer succesvolle verkiezingscampagne beloofd ervoor te zorgen dat het gehele betwiste land bij Mijnland zal komen. Dit gebied en vooral de olievelden zijn volgens hem van groot belang voor Mijnland aangezien dit de economie van Mijnland zou stimuleren en onafhankelijk zou maken van olie-importen die zo belangrijk zijn voor de industrie. Ook is het belangrijk om de olievelden voor Mijnland te winnen omdat Anderland misschien de militaire overhand zou krijgen op het eiland. Jij gelooft in Gerards plannen en hebt ook op hem gestemd. Tijdens zijn tijd als president heeft hij meerdere speeches gegeven waarin hij duidelijk maakt dat hoe vastbesloten hij is om hier succesvol uit de onderhandelingen te komen. Gerard geeft zelfs aan dat als de onderhandelingen niet volgens plan verlopen, of als Anderland niet genoeg toegeeft, hij zelfs bereid is militairen in te zetten.
Compromis
Na lange en moeilijke onderhandelingen wordt er een afspraak gemaakt over de grenzen in het grensgebied. Mijnland krijgt 1/3de van het betwiste gebied. Gerard presenteert dit als een zeer succesvol verlopen onderhandelingen en geeft aan dat hij tevreden is met de uitkomst.
29 De vragen
Ik de vind manier waarop de premier van Mijnland de crisis heeft opgelost goed.
Ik vind deze uitkomst van de onderhandelingen goed.
[1: helemaal mee oneens. 7: helemaal mee eens. ]
(militaire assertiviteit)
De beste manier om vrede te behouden is door sterke militaire capaciteiten, hierdoor kunnen andere landen je niet bedreigen of manipuleren.
Het gebruik van militairen maakt problemen vaak erger en lost internationale conflicten vaak niet op.
Oorlog is een vervelende situatie, maar soms is het de enige oplossing voor internationale problemen.
[1: helemaal mee oneens. 7: helemaal mee eens. ]
(internationaal vertrouwen)
Landen zijn over het algemeen niet te vertrouwen, ze zullen als ze kunnen proberen hun territorium en macht proberen uit te breiden.
(nationaal chauvinisme)
Hoe veel beter vindt je jouw eigen land (in de echte wereld) ten opzichte van andere landen)
[1: helemaal niet beter. 7: helemaal wel veel beter. ]
Zijn er dingen over jouw eigen land (in de echte wereld) waar je je over schaamt
[1: helemaal geen dingen waar ik me over schaam. 7: heel erg veel dingen waar ik me over schaam. ]
(Om politieke ideologie mee te meten: )
Als je aan politieke stromingen denkt, hoe zou je je eigen standpunt omschrijven?
30
Appendix 2.
Descriptives
mijn gender is:
Frequency Percent Valid Percent
Cumulative Percent Valid Male 34 53,1 54,8 54,8 Female 28 43,8 45,2 100,0 Total 62 96,9 100,0 Missing anders 2 3,1 Total 64 100,0
het hoogste niveau van onderwijs dat ik heb afgerond / nu mee bezig ben:
Frequency Percent Valid Percent
Cumulative Percent Valid VMBO 1 1,6 1,6 1,6 MBO 1 1,6 1,6 3,1 HAVO 1 1,6 1,6 4,7 HBO 9 14,1 14,1 18,8 VWO 7 10,9 10,9 29,7
Een universitaire opleiding 45 70,3 70,3 100,0
31
Hoe oud ben je? - Click to write Choice 1
Frequency Percent Valid Percent
Cumulative Percent Valid 16 3 4,7 4,7 4,7 17 2 3,1 3,1 7,8 18 3 4,7 4,7 12,5 19 6 9,4 9,4 21,9 20 4 6,3 6,3 28,1 21 5 7,8 7,8 35,9 22 9 14,1 14,1 50,0 23 9 14,1 14,1 64,1 24 4 6,3 6,3 70,3 25 1 1,6 1,6 71,9 26 3 4,7 4,7 76,6 28 1 1,6 1,6 78,1 30 1 1,6 1,6 79,7 38 1 1,6 1,6 81,3 50 1 1,6 1,6 82,8 52 1 1,6 1,6 84,4 53 2 3,1 3,1 87,5 55 3 4,7 4,7 92,2 57 2 3,1 3,1 95,3 61 2 3,1 3,1 98,4 80 1 1,6 1,6 100,0 Total 64 100,0 100,0
No significant outliers or influential cases
Statistics Cook's Distance for Approval Uncentered Leverage Value for Approval N Valid 64 64 Missing 0 0 Mean ,0194 ,0469 Std. Deviation ,02857 ,02619
32 testing normality
34 The regression tables including the hawkishness*audience costs treatment for both variables: Coefficientsa Model Unstandardized Coefficients Standardize d Coefficients t Sig. Collinearity Statistics B Std. Error Beta Toleranc e VIF 1 (Constant) 2,443 1,316 1,856 ,069 wel treatment = 1 -,147 ,384 -,051 -,383 ,703 ,976 1,024
het hoogste niveau van onderwijs dat ik heb afgerond / nu mee bezig ben:
,231 ,212 ,145 1,086 ,282 ,958 1,044
Hoe oud ben je? - Click to write Choice 1
-,009 ,014 -,085 -,645 ,521 ,982 1,019
Gender_Dummy ,077 ,382 ,026 ,202 ,841 ,992 1,008
2 (Constant) 4,530 1,611 2,811 ,007
wel treatment = 1 -,193 1,065 -,066 -,181 ,857 ,116 8,630
het hoogste niveau van onderwijs dat ik heb afgerond / nu mee bezig ben:
,171 ,208 ,107 ,821 ,415 ,914 1,094
Hoe oud ben je? - Click to write Choice 1
-,024 ,015 -,220 -1,619 ,111 ,844 1,184
Gender_Dummy ,538 ,403 ,184 1,334 ,188 ,815 1,227
Hawkishness -,474 ,222 -,387 -2,133 ,037 ,472 2,120
Hawk_Treat -,058 ,325 -,065 -,179 ,859 ,117 8,573
35 Coefficientsa Model Unstandardized Coefficients Standardize d Coefficients t Sig. Collinearity Statistics B Std. Error Beta Toleranc e VIF 1 (Constant) 2,008 1,298 1,547 ,127 wel treatment = 1 -,278 ,378 -,094 -,735 ,465 ,976 1,024
het hoogste niveau van onderwijs dat ik heb afgerond / nu mee bezig ben:
,387 ,209 ,239 1,848 ,070 ,958 1,044
Hoe oud ben je? - Click to write Choice 1
-,014 ,014 -,122 -,957 ,342 ,982 1,019
Gender_Dummy -,210 ,377 -,071 -,557 ,580 ,992 1,008
2 (Constant) 4,419 1,579 2,798 ,007
wel treatment = 1 -,772 1,044 -,261 -,740 ,463 ,116 8,630
het hoogste niveau van onderwijs dat ik heb afgerond / nu mee bezig ben:
,309 ,204 ,190 1,515 ,135 ,914 1,094
Hoe oud ben je? - Click to write Choice 1
-,028 ,015 -,252 -1,930 ,059 ,844 1,184
Gender_Dummy ,253 ,395 ,085 ,639 ,525 ,815 1,227
Hawkishness -,543 ,218 -,436 -2,493 ,016 ,472 2,120
Hawk_Treat ,086 ,319 ,095 ,270 ,788 ,117 8,573
