Interpersonal Discussions and Attitude Formation on Foreign Policy: the Case of Polish Involvement in the Iraq War

Abstract

This paper examines Polish support for the country's participation in the Iraq war in 2004. I argue that interpersonal discussions are a driving force behind emergent attitudes on foreign policy, such as support or opposition to war. I identify three mechanisms through which political discussions can influence individual's views on the war and develop hypotheses about the impact of kinship ties and frequency of discussions on strengthening the influence. I test my argument using the first large-N data on interpersonal discussions and foreign policy outside of the US context. Findings demonstrate that having a pro-war conversation partner greatly increases the probability that one will adopt similar views. They also show that when one's social environment is taken into account as the source of information about the policy, the impact of mass media diminishes.

In 2003, the Polish government became one of the most vocal supporters of the US-led coalition in Iraq. Polish troops were stationed in the region until 2008. Since 2004, the country has assisted with NATO's mission in Afghanistan, and in 2011, it took over the rotating presidency of the European Union. Despite the country's growing international presence and its strategic ties to the United States, we know little about ways in which the Polish public makes sense of these efforts and, in turn, decides to support or oppose specific foreign policy. This paper investigates the formation of Polish views on one such policy, the country's participation in the military intervention in Iraq, by relying on a socially based approach to explore the role of interpersonal discussions in shaping attitudes on the war.

Previous research on Polish public opinion is limited as is empirical research on public opinion and foreign policy in Eastern Europe in general. Most of what we know about attitude formation on foreign policy in this region is based on the studies of public support for European integration. Scholars have usually imported variables deemed significant from extensive research on US public opinion on foreign policy. I argue, however, that citizens of East European countries differ somewhat from those living in the West in how they make judgments about foreign policy in general because of their post-Communist legacy. In the United States, individuals have had an opportunity to openly engage in various types of public and private political activities for years. This has allowed them to receive political cues through diversified sources, including informal discussions with friends but also from attending meetings with local leaders or volunteering at political rallies to name a few. By contrast, the authoritarian regimes in Eastern Europe forced many people to disengage from public political debate, in effect pushing them to rely almost exclusively on informal social networks to obtain information about the government's abuses (Gibson 2001). The idea of ‘kitchen circles’ or informal discussions among friends about art and politics carried in the kitchens of one's apartment are considered by many as unique features of Communist societies (Wedel 1986; Greenfeld 1992:23). This tradition of engaging in interpersonal political discussions in a private, informal manner remains alive today and is considerably more common than in the United States. Indeed, Gibson (2001) shows that the percentage of individuals engaging in informal political discussions daily is much greater in Bulgaria, Poland, and Russia than in the United States. The same is true when comparing discussions with friends in Eastern Europe and Western democracies in the 1990s. Janmaat (2006) shows that when it comes to various types of civic engagements, political discussion with friends is the only indicator on which Eastern Europe scores slightly higher than Western democracies. As such, I posit that existing approaches of attitude formation on foreign policy based mostly on research in the US context are likely to offer less explanatory power for understanding public attitudes in Eastern Europe.

The study's theoretical contribution is to offer a new lens through which we study attitude formation on war policy in Eastern Europe. I explore ways in which one's immediate social environments can influence attitude formation. I argue that while some rely on elite and media cues, many prefer to turn to friends and family for information about a policy or to discuss an issue because people are more likely to identify with their immediate social groups than they are with distant elites. These conversations, however, are more than just idle talk. They have been and continue to be an integral part of daily life, helping people make sense of complex policies, such as the war in Iraq, and offering ‘translations’ of elite and media messages.

In his research on public opinion and war policy, Gartner (2008) shows that context is vital in giving meaning to numbers such as war casualties. Without context, information has little meaning to average citizens. Like Gartner, I argue that context matters. Yet rather than emphasizing the importance of news stories in revealing casualty trends, I show that interpersonal discussions serve as a lens through which individuals form their views on policies.

Empirically, the study contributes to our knowledge of public opinion and foreign policy in the East European context. It presents novel data on interpersonal discussions and public support for Polish involvement in the Iraq war in 2004 and sheds light on how individuals' conversation partners shape their perceptions of costs and benefits associated with ongoing war involvement. Yet the utility of social approach is not limited only to Poland. It can be generalized to areas beyond those of Eastern Europe to regions where discussions are a norm among the population. Findings in this paper suggest that future research on public opinion and war policy needs to account for individuals' embeddedness in their social networks to better understand the process of attitude formation.

Next, I present a brief overview of the research on public support for military intervention, highlighting this study's contribution. I then develop a theoretical story that links social interactions—interpersonal discussions—to the process of opinion formation by emphasizing three mechanisms through which such interactions shape people's views on foreign policy. In section 4, I describe my research design. I present results from empirical analysis in section 5 and compare the socially based approach to existing models of opinion formation. I conclude with suggestions for future research.

Public Support for Military Intervention

Existing research on individuals' inclination to become supporters or opponents of military interventions explores either the impact of individual characteristics on shaping attitudes or the ways in which external environment delivers information about foreign policy, thereby allowing people to engage in a cost-benefit analysis. Works emphasizing the importance of individual characteristics examine factors such as gender, age, income, education, occupation, and political knowledge as predictors of internationalist or isolationist views among the public (Hughes 1978; Wittkopf 1990; Nincic 1997; Sinnott 2000; Isernia 2001; Colton 2002). Professionals with high incomes, findings demonstrate, favor cooperative internationalism but reluctantly support military interventions. On the contrary, those with lower educational levels show greater propensity toward militant and isolationist policies in the United States (Wittkopf 1990; Nincic 1997). In the context of gender, studies indicate that women favor peace-oriented policies (Goldstein 2001; Sahliyeh and Deng 2003), though findings are still inconclusive, showing that gender gaps diminish through time.

Another set of individual-based explanations for attitude formation incorporates beliefs and perceptions. Core beliefs serve as heuristics through which citizens form ideas about world affairs. Early on, Almond (1950) has argued that ‘psycho-cultural’ elements define national character, and in countries such as the United States, where arbitrary aggression is negatively evaluated, the use of force abroad might not be easily accepted. Outside of the United States, beliefs about threats associated with Western institutions such as NATO have been cited as the main cause of anti-American views in Russia (Shiraev and Zubok 2000). There has also been some focus on human psychology and personality to determine the level of support for military policies. In his study of German public opinion on issues ranging from support of the US-led invasion of Iraq to support for the Euro, Schoen (2007), for example, shows that personality traits shape goals and motivations—factors that determine people's responsiveness to external stimuli such as war-related news.

Increasingly, scholars have begun to emphasize the role of information in shaping individual views on military interventions. The number of casualties is perhaps the most widely recognized piece of information that people rely on to develop preferences on military policy. Empirical research supports this argument (Gartner, Segura and Wilkening 1997; Kriner and Shen 2010), although Gartner (2008) has shown that numbers are less meaningful in predicting public support for a policy. Instead, understanding the context matters. Gartner (2008) shows that trends in casualties shape people's perceptions about the expectations of future wartime losses and thus are a more accurate predictor of public opinion than information about casualties in the absence of context.

Because the media often play an instrumental role in delivering information about battle deaths and war-related issues, it is not surprising that scholars have investigated the media–public opinion linkage. Most often the media are seen as outlets for political elites to present commentaries and information and set the agenda for news coverage (e.g., Zaller 1992; Soroka 2003). Increasingly, however, there is more emphasis on the media's independent effect on public opinion. Gelpi (2010) argues that people respond to surprising news stories and these, more than elite cues, influence how they update information about wars. This effect is more pronounced in the post-9/11 environment—citizens concerned about terrorism are more likely to support hawkish policies when exposed to threatening news stories (Gadarian 2010).

Studies of public opinion formation and war in the Polish context are limited. Researchers analyze the media's coverage of the war (Meyer and Zdrada 2006) or explain the government's reluctance to respond to public opposition to the country's involvement in Iraq (Radziszewski and Wolfe 2012). Empirical research on Polish public opinion focuses predominantly on attitudes toward the EU. Most of the works emphasize variation in individual characteristics as predictors of pro- or anti-EU views, exploring the role of material calculations, party preferences, and cultural values in a way that has been done in the literature on public opinion in the United States (Cichowski 2000; Slomczynski and Shabad 2003).

Because systematic empirical research on public support for military interventions in the East European context is missing, this project attempts to fill this gap in the literature. Instead of relying exclusively on individual-based explanations or focusing on the role of the media, I examine attitude formation through the lens of one's social environment. Just like Gartner (2008), I argue that context in which information is delivered matters. Yet unlike Gartner, I argue that one's social networks play an important role in giving meaning to information particularly in Poland where engagement in political discussions is common. One's immediate social circles diffuse information about policy, encourage learning about policy, and often define what an ‘acceptable’ opinion should be. Through these mechanisms, they can influence the malleability of attitudes.

Existing research on the social aspect of opinion formation finds that one's political discussants influence voting behavior and support of candidates during campaigns (Huckfeldt and Sprague 1995; Druckman and Nelson 2003). This research focuses exclusively on the United States. Here, I demonstrate the benefit of incorporating such an approach to the study of individual views on foreign policies in a region where interpersonal conversations about various issues, including international politics, are common as is the case in Poland. To explain why informal conversations could impact individual views, I expand on the theoretical story of influence presented in the research on American voting behavior by identifying three distinct mechanisms through which influence could occur.

Social Circles, Influence, and the Shaping of Individual Views on Polish Involvement in the Iraq War

To examine ways in which social interactions affect individual views on foreign policy, I begin with two assumptions. First, I assume that individuals employ a rational calculation of the costs and benefits associated with a particular policy. If the benefits of supporting the policy outweigh the costs, individuals are more likely to support that policy (Page and Shapiro 1992). Second, I assume that individuals are embedded in some form of social communities. Whether it involves attending work, socializing with friends, or at the very least living within the boundaries of neighborhoods, individuals are members of social communities (Huckfeldt and Sprague 1995; Watkins and Danzi 1995). Although the extent to which individuals are actively embedded in social communities varies, most individuals engage in some basic social activities. Integrating the two assumptions, I argue that when deciding whether to support or oppose the war, individuals will start by evaluating the costs and benefits of adopting a specific view in a consistent manner—a logic that resonates with the rational actor model—yet social environments in which individuals are embedded will identify what these costs and benefits are, at times, defining the appropriate way of thinking or even transforming one's identity in the process—a logic consistent with constructivism.1 While individuals make rational decisions, the discussions with friends—interactions with one's social environment—define the costs and benefits not just in terms of material considerations but also by invoking the logic of norms and appropriateness that will underpin the decision about supporting or opposing the war. Depending on the social environment's role in one's life, its impact in shaping attitudes could be limited simply to providing information about the policy or be as significant as molding one's identity and transforming attitudes in a more substantial way. While individuals may be exposed to messages from the media and national elites, discussions with one's social circle create an opportunity to dissect the policy's potential impact on one's life in a way that exposure to the elites and media rarely permit. Consequently, it is likely that throughout their lives, individuals will frequently turn to their social circles before deciding how to think about foreign policy.

I now explore different mechanisms through which social interactions, specifically interpersonal interactions, can mold individual views on foreign policy. My focus is on the mode through which political conversations emerge, which, I argue, constitutes a critical element in the story of influence. In a basic way, social circles can influence people to support a policy when discussions provide information about an issue. In a more engaging manner, social circles can encourage individuals to pursue a deeper exploration of the topic during conversations, or put pressure on an individual to think about the policy in a specifically defined manner. Delineating these different paths of influence helps to uncover how ordinary people make sense of foreign policy and how one's social circles affect this process.

Transmission of Influence: Information Provision

The most basic way in which interpersonal interactions facilitate opinion formation is by creating opportunities for people to share information about a policy, a step that helps individuals make sense of the costs and benefits associated with supporting the issue. Individuals are generally less informed about foreign policy than they are about domestic policy (Iyengar and Morin 2006). As such, uninformed individuals may turn to their social circles for information because, unlike the media, such networks could provide information about foreign policy that is most relevant to the person who seeks it. Furthermore, sometimes people could be exposed to information even if they do not explicitly look for it, but happen to join a conversation during which political issues are discussed. In such a way, information randomly registers with an individual (Huckfeldt and Sprague 1995). Thus, whether or not one directly seeks information about a policy, political discussions deliver cues with the potential to influence one's opinions. This is because uninformed individuals are not only exposed to information during conversations but also more likely to accept the way people with whom they discuss politics, and who they believe are more informed than they are, interpret information.

Those who seek to deepen their knowledge about a policy's potential impact may also receive sophisticated information about the policy during informal discussions. Seeking new knowledge can be costly particularly if it involves gathering specialized information. Unlike uninformed citizens, whose first priority might include acquiring basic information, or political experts, who are highly knowledgeable, some may be interested in expanding their rudimentary knowledge to assess how a given policy might affect their lives. Yet in many instances, specialized information is difficult to obtain. Take the case of Polish involvement in the Iraq war. The most controversial foreign policy decision in 15 years, Polish intervention in Iraq, was weakly justified, and the analysis of the policy's costs and benefits was limited (Wągrowska 2004). It was not clear, for example, to what extent Polish commitment as an ally to the United States would strengthen the country's position internationally or weaken it by alienating the country's European allies. Other issues associated with the benefit of undertaking the policy, such as the removal of visa requirements for Poles traveling to the United States, remained ambiguous. Therefore, it is not surprising that relying on one's social circles could present an attractive option for those in need of information, especially when such circles consist of credible experts, people who are interested in foreign policy and have the information to share (Bikhchandani, Hirschleifer and Welch 1992).2

Interpersonal discussions provide a quick and cost-efficient way through which knowledge can be acquired. More importantly, such discussions could be helpful in translating difficult concepts and bridging the link between foreign policy and one's daily life. Individuals with weakly formed preferences on the policy who are exposed to such messages may either become stronger supporters of the policy if the messages are consistent with prior views or change their views when faced with more sophisticated and convincing information.

Transmission of Influence: Social Pressure

The second mechanism through which social interactions can shape individual opinions and alter individual calculations of policy preferences is rooted in the role of social circles as providers of material and social goods. Whether they help with finding employment or serve as an anchor in times of tragedy, friends, family, and coworkers provide support in times of material or social needs (Wellman and Wortley 1990; Uehara 1994). Individuals might embrace a particular stand on foreign policy, such as opposing a controversial war, not necessarily because the facts suggest the war would be costly, but rather because they have an interest in preserving the benefits they receive from their social circles and fear opposing the popular views. This mechanism is similar to the logic behind groupthink, a phenomenon sometimes observed in decision making in small groups where individuals feel pressure to reach a consensus in a group (Janis 1982). Like in groupthink, individuals in this case are aware of the emerging norms in their environment.3 The desire to maintain the relationship with the group then drives people to adopt the norm, even if they do not necessarily agree with it.

The extent to which one is willing to conform to the beliefs of one's social circle, or at the very least to the beliefs of key member(s), will depend on the benefits one derives from socializing with friends, family, and so forth (e.g., how significant is popularity to one's esteem (Bernheim 1994). In such instances, informal discussions are not merely about delivering information. Instead, during such interactions people could communicate the ‘acceptable’ way of thinking about a policy and put pressure on the individual to be pro- or anti-war. In other words, a strong norm of appropriateness is communicated, and the individual deciding whether or not to support the war must consider the cost of violating the norm. Yet the pressure of adopting specific views could be reduced in cases when individuals have multiple and unrelated family and friendship ties and thus feel less pressure to conform to particular views. Even when some relations dissipate due to opinion differences, having alternative social circles can mitigate the loss.

Transmission of Information: Social Learning

One's social circles can also shape opinions by encouraging deeper learning about policies. Social learning, as defined by Checkel (2001), refers to a social process of persuasion aimed at utilizing argumentation and discussion to convince one to form a specific view on an issue. In this sense, learning involves a great deal of ‘argumentative persuasion’ and debating in the absence of coercion (ibid 562). Social learning, however, is more complex than just obtaining information. It should be manifested by a genuine and deep belief in the cause to a point where individuals who engage in social learning internalize their preferences and are less likely to react to new information, for example, by changing their opinions (Risse, Ropp and Sikkink 1999). During initial stages of socialization, individuals may evaluate the cost and benefit of supporting/opposing a policy based on discussions with one's social circle, but as the process of persuasion continues, individuals begin to internalize the norm and no longer have the need to engage in the cost/benefit analysis. In the final stages of socialization, supporting or opposing the policy becomes automatic. Unlike the information-provision mechanism I described earlier, social learning is not just about receiving information, but about forming a sense of attachment to a specific world view that emerges through in-depth discussions.

Unlike the mechanism of social pressure described earlier, social learning occurs in the absence of pressure to conform to social standards. Social learning, as conceptualized by Checkel (2001), is about convincing people to think about a policy in a specific way by creating an opening for critical analysis. Ultimately, this form of learning allows individuals to gain new understanding of interests in a way that may change one's identity, for example, by making one a supporter not only of Polish involvement in Iraq, but a genuine believer in a policy of foreign intervention in military conflicts. Such a process can occur through meaningful interactions during which one's social environment encourages discussions about foreign policies that involve analysis and debating, leaving one with a good sense of why he/she wants to support/oppose a policy. If the interactions take place within a larger social circle, some friends or family members may emerge as more vocal proponents/opponents of a policy. Such individuals, which Hudson (2005) calls entrepreneurs, may frame the initial tone of the debate, yet such agents will still rely on persuasion rather than coercion to convince individuals to take a specific stand on a policy and will invite others to participate freely in the discussions.

Of the three mechanisms I outlined, the social-learning mechanism illustrates the most pronounced way in which social environments can shape individual views. Not only is information exchanged during social learning but the process of critical analysis that characterizes such learning is likely to help individuals internalize their preferences on a policy so that individuals become strong supporters/opponents of a policy. While the three mechanisms demonstrate different ways in which social environments can shape views on foreign policy, they all suggest the following:

Hypothesis 1: Discussing Polish involvement in the Iraq war with a pro-war individual increases the probability that one will also exhibit pro-war views.

Besides focusing on the mode through which political discussions emerge and influence attitudes, the characteristics of the people with whom individuals discuss politics could also matter in opinion formation. Some scholars (Katz and Lazarsfeld 1955; Putnam 1966; Kenny 1998) demonstrate that close relationships between friends and family are most conducive for influencing individuals. People may adopt the views of their family because of greater intimacy and trust that often characterize such relationships. It is also close friends and family who often constitute the main source of information on issues (Straits 1991). Granovetter (1973), however, argues that greater information flow increasingly taking place at work facilitates the importance of ‘weak’ ties, individuals who are not a close kin but who may be highly connected because they are linked to other groups. These individuals may be useful in delivering new information especially if one seeks more sophisticated knowledge because such people may be connected to different groups with a variety of specialized information on the topic. Thus, discussions among people with weak ties could play an increasingly important role in shaping opinions because of such individuals' greater ability to rely on new and relevant information than might be seen in more intimate yet closed social circles. This leads to two different predictions:

Hypothesis 2a: Discussing the war in Iraq with pro-war family/friends makes one more likely to adopt pro-war views than discussing the war with weak ties.

Hypothesis 2b: Discussing the war in Iraq with pro-war coworkers/neighbors makes one more likely to adopt pro-war views than discussing the war with strong ties.

We should also expect that as the frequency of conversations between political confidants increases, so do the odds that opinion conformity on foreign issues will occur. Political discussions reinforce ideas and clarify uncertainties. The more exposed an individual is to certain ideas, the more likely he/she is to adopt them. Hence,

Hypothesis 3: As the frequency of discussions between an individual and a pro-war conversation partner increases, so does the likelihood that one will exhibit pro-war views.

Research Design

Data and Case

I examine interpersonal discussions of Polish population. Among East European countries, Poland has one of the strongest traditions of informal political discussions dating to the emergence of underground opposition in the 1980s (Osa 2003). The government's control of the media forced individuals to rely on the word of mouth to disseminate information and discuss the regime's abuses. Because of the strong legacy of underground communication and the importance of social networks in mobilizing people to support Solidarity's cause (ibid 2003), the case allows me to examine the importance of political discussion on Polish attitudes toward the country's involvement in the Iraq war. Focusing on Poland also matters because as the third largest country from the former Soviet bloc and the sixth most populous country in Europe, its foreign policy has the potential to affect EU politics. Its close ties to the United States also make Poland a strategic trans-Atlantic ally. Thus, understanding opinion formation on foreign policy in Poland merits attention. To test my hypotheses, I designed a survey of public opinion on the war in Iraq, arguably the second most active and critical foreign policy of post-Communist Poland after EU negotiations. The survey data in this study are the first large-N data exploring interpersonal discussions and attitudes on foreign policy.

Sampling

The target population in the telephone survey consists of Polish citizens aged 18 and above. The random probability sample4 includes 1000 respondents. I chose telephone surveys rather than face-to-face or written interviews because the former offers some benefits. Although the disadvantage of phone surveys involves the risk of omitting those respondents who do not own phones, this format usually ensures greatest response rate over a relatively short period of time (Czaja and Blair 1996). Telephone surveys also represent a cost-efficient way of conducting interviews (ibid 1996).

First, respondents were selected from 64 stratified units given that Poland has 16 provinces and four types of cities/villages (cities with: 200,000 inhabitants or greater, 50,000–200,000 inhabitants, 50,000 inhabitants or less, and villages). Second, to determine the number of interviews for each province, I took the percentage of population for each province relative to total population in the country and multiplied it by 1000—the desired sample size. Third, I determined the number of interviews required in each village/city type within each province. This was done by multiplying the percentage of specific city type population relative to total province population for all city/village sizes by the number of interviews needed in the particular province for all city/village types. Fourth, I randomly selected villages and cities for interviews within each province and city/village type. Fifth, telephone books served as a sampling frame to select respondents.

To ensure variation in individuals' education, gender, and age, a quota was established to choose the number of respondents with different socioeconomic backgrounds reflecting national distribution. Thus, the number of interviews in each age category was chosen by establishing the percentage that the population in a specific age group constituted relative to total population for all age groups and then by multiplying the percentage by the number of required interviews. Within each age category, there was a quota for males and females established based on the same formula. The quota for level of education was conducted in the same way as that for age, independently of age and gender. Quota for place of inhabitance was established independently of age and gender and was discussed earlier.

Computer-assisted telephone interviews (CATI) were used to conduct interviews with the assistance of IMAS International, a research company in Wrocław, Poland. Interviews began on August 20, 2004, and ended on September 8, 2004. Each interview lasted approximately 16 minutes. The response rate was 23 percent.

Dependent Variable

Support or opposition for Polish participation in the Iraq war is the dependent variable as captured by the following survey question: Which of the following best describes how you feel about Poland's participation in the war in Iraq? I coded the responses as dummy variables, with 1 assigned to supportive views for the country's involvement and 0 to neutral and nonsupportive ones. Given that very few respondents expressed neutral views, I merged the neutral and nonsupportive categories.

Independent Variables

Below, I introduce the main explanatory variables included in the model of attitude formation on Polish participation in the Iraq war. These variables explore the views of one's conversation partners (discussants), the relationships between the discussants, and the frequency of conversations.

Social Interactions Variables

• 1.
  Perceived Discussant Views: The variable measures the viewpoints on the Iraq war of the people with whom survey respondents discuss the war. The respondents are asked the extent to which they think their conversation partners support Polish involvement in the war. The survey asked: Think about the person with whom you most often discuss politics. It could be someone from your family, a neighbor, a friend, or someone you know from school, work, or meet in church/other organization. Which of the following best describes this person's attitude toward Poland's participation in the war in Iraq? I coded the responses as ordinal variables ranging from −2 to 2 (definitely unsupportive, rather unsupportive, neutral, rather supportive, definitely supportive) for the first discussant in Model 1 and from −1 to 1 in Models 2 and 3. The second discussant's views are coded from −1 to 1. I merged the measure for the second discussant because the full measurement did not yield enough observations in each category. I did so to avoid potential bias in estimation. I did the same for the first discussant in Models 2 and 3 where I use the variable as part of an interactive term. Individuals without a discussant were coded as zero, the same value assigned to individuals who named a discussant with a neutral view. Given that I seek to explain attitude formation of the general population rather than just studying the views of individuals who have a discussant, such coding avoided the problem of excluding those individuals who did not name a discussant yet should be included in the initial model. I used the same coding in cases where the respondent named a second conversation partner (variable Discussant 2).

A potential problem exists with employing an indirect measure of the discussants' opinions as respondents may sometimes have inaccurate perceptions of others' views. Despite the potential shortcoming, I rely on the respondents' perceptions for several reasons. First, given that I am interested in how interpersonal discussion affects opinion formation, what might matter more is the discussant's perceived message rather than the actual one (Mutz and Martin 2001). Second, even if we accept the potential bias in the respondent's perceptions, the inaccuracies are likely to be small as found by Beck, Dalton, Greene and Huckfeldt (2002) in their study of voting behavior in presidential elections. This suggests that reliance on perceptions is not as problematic as sometimes assumed.

• 2.
  Dummy 1 (No Discussant): This variable distinguishes between those individuals who named a discussant and those who did not and allows me to observe whether those who named a discussant are different in their attitude formation from those who do not converse with anyone. I coded the variable 1 for no discussant and 0 for the presence of a discussant.
• 3.
  Dummy 2 (No Discussant): Same as above, except it looks at people who named a second discussant.
• 4.
  Relationship Ties * Perceived Discussant Views: This interaction variable measures if one is more likely to support or oppose Poland's participation in the Iraq war when the discussant supports or opposes the policy and happens to be a strong or weak tie. Relationship ties are coded from 3 to 1, with 3 denoting interactions with family and 1 denoting interactions with neighbors and coworkers. Discussant views were coded from −1 indicating opposing views to 1 indicating supportive views with 0 as neutral. The interactive term ranges from −3 to 3 and focuses on the first discussant.
• 5.
  Frequency of Political Discussions * Perceived Discussant Views: The frequency of political discussions, an ordinal variable ranging from daily to once a month and less than that and coded from 5 to 1, is multiplied by discussant views. Discussant views were coded from −1 indicating opposing views on the policy to 1 indicating supportive views with 0 as neutral. The interactive term ranges from −5 to 5.

Control Variables

Socioeconomic Factors

1. Age: Research has shown that older generations would favor Polish involvement in the war because of their traditional sentiment toward the United States (Applebaum 2005). I coded the variable from 3 to 1, with 1 denoting ages 18–39 or younger generation, 2 denoting ages 41–59 or middle-aged individuals, and 3 denoting ages 60 and greater or the elderly. I used the simplified coding to increase the number of observations in categories.
2. Education: Studies on US foreign policy have demonstrated that support for militant policies is stronger among those with limited educational background (Wittkopf 1990; Nincic 1997), possibly suggesting that less-educated individuals would approve of Polish involvement in the war. Level of education is an ordinal variable coded from 3 to 1, where 3 = higher than high school, 2 = high school or equivalent, 1 = elementary.
3. Gender: Work on gender and foreign policy has shown that women favor peace-oriented initiatives (Goldstein 2001; Sahliyeh and Deng 2003). This might imply that Polish women should oppose their country's presence in Iraq. Gender is a dummy variable coded 1 for female and 0 for male.
4. Occupation: When considering the impact of occupation, Wittkopf (1990) has demonstrated that those in professional and managerial occupations have been supporters of ‘cooperative internationalism,’ while blue-collar workers and farmers have been considered hardliners and supporters of ‘militant internationalism.’ Given the findings, I created several dummy variables: 1 for professionals (doctors/professors/etc.) and 0 others, 1 for those who work in trade/service and 0 for others, 1 for qualified manual labor and 0 for others, 1 for farmers and 0 for others. Base category is others, which includes unemployed, housewives, students, nurses, small business owners, unqualified manual labor, and the retired.
5. Income Change: Individuals with declining income could attribute their situation to the country's costly foreign policy. Income change is an ordinal variable ranging from 2 to −2, (2 = financial situation has improved greatly, 1 = has improved somewhat, −1 = has deteriorated somewhat, −2 = has deteriorated greatly). A value of 0 indicates no income change.
6. Place of Inhabitance: A dummy variable distinguishes between rural and urban areas, with 1 indicating villages and 0 indicating towns/cities.

Partisanship

1. Level of Support for the Government: Anderson and Kaltenthaler (1996)have shown that individuals evaluate foreign policies through the lens of domestic developments. A favorable view of the government should generate support for Polish involvement in the Iraq war. Support for the government's handling of the economy is an ordinal variable and coded as 1 = very satisfied about the government's handling of the economy or somewhat satisfied, −1 = somewhat unsatisfied or very unsatisfied, and zero indicating ambivalence.
2. Party Support: Sahliyeh and Deng (2003) have shown that those who support pro-war parties will also exhibit pro-war attitudes. Respondents were asked to indicate the party they would vote for if parliamentary elections were held today. I classified each party according to its level of support for the war in Iraq. I then coded the ordinal variable as, 1 = strongly pro-war or somewhat pro-war, −1 = somewhat anti-war or strongly anti-war, and zero indicating neutrality. I merged the supportive measurements and the opposition measurements to have enough observations in each group. I relied on newspaper articles and official party Web sites to obtain information about each party's stand on the issue. The table illustrates party classification according to levels of support for the policy at the time of the interviews.

Party Support for the War

Party	Level of Support
Prawo & Sprawiedliwość	Strongly supportive
Platforma Obywatelska	Strongly supportive
Sojusz Lewicy Demokratycznej	Strongly supportive
Unia Pracy	Somewhat supportive
Samoobrona	Strongly unsupportive
Polskie Stronnictwo Ludowe	Somewhat supportive
Liga Polskich Rodzin	Strongly unsupportive
Unia Wolności	Strongly supportive
Socjaldemokracja Polska	Strongly unsupportive
Polska Partia Narodowa	Strongly unsupportive

Level of International Trust

1. The extent to which individuals exhibit positive perceptions of other nations can translate to specific preferences for policies (Brewer, Gross, Aday and Willnat 2004). Those who trust the United States and UK should display supportive attitudes toward Poland's participation in the Iraq war. Trusting France, Germany, and Russia—states that opposed the war—should be associated with suspicion of US actions in Iraq and thus limited support for Polish participation in the mission. Respondents were asked whether Poland should trust selected nations or whether the country should be cautious when dealing with those nations. I examine five variables: level of trust for (i) Russia, (ii) France, (iii) Germany, (iv) UK, and (v) United States. In the survey, I coded each variable as 2 = definitely trustful of the country, 1 = somewhat trustful, −1 = somewhat cautious, −2 = definitely cautious. I merged the categories in the analysis and recoded the variable, with −1 denoting distrustful opinions, 1 denoting trustful ones, and 0 assigned to neutral responses. I merged the groups to have enough observations in each category.

The Media

1. Perceived Media Bias: Whether they set the agenda, disseminate facts, or grant air time to the elites, the mass media have been studied extensively as the tool for shaping opinions (Zaller 1992; Iyengar and Simon 1994). I examine whether exposure to media outlets is associated with individual support or opposition to Poland's participation in the Iraq war. Respondents were given a set of TV programs and newspapers and asked to select those they read/view regularly. They indicated the extent to which they thought each program/newspaper displayed pro- or anti-war leanings in their coverage. Variable for perceptions of each medium bias is ordinal and coded in the survey as, 2 = media definitely supportive of the policy, 1 = somewhat supportive, −1 = somewhat unsupportive, −2 = definitely unsupportive, 0 = neutral.6

To capture total exposure to TV programs and newspapers, I added the values assigned to the opinion of each newspaper/TV program for the total opinion score for print and TV and then recoded all positive values (supportive views) as 1, all negative (opposing views) as −1 and assigned a value of 0 to neutral views. The outcome is an ordinal variable ranging from −1 to 1. In the analysis, I only include the variable for print as TV and print are highly correlated and inclusion of both could lead to biased estimators.

Analysis

Over 78% of Poles in my study opposed their country's participation in the Iraq war in 2004, while 17% supported it.7 Half of the respondents reported discussing the war with one person while 11% conversed with two people. Half of the respondents did not discuss the war with anyone. The majority of conversations took place a few times during the week (49.5%), followed by once every 2 weeks (18.7%) and once a month (12.9%). Six percent discussed politics daily. The majority of respondents also reported that their conversation partners lived in the same village or city as they (44.2%), while 31.8% lived in the same neighborhood. Overall, the numbers indicate that many engage in informal discussions about the war and do so frequently.

Turning to the analysis, I use logistical regression to test the hypotheses as my dependent variable is dichotomous. My analysis focuses on three models. Model 1 examines hypothesis 1—it tests the extent to which having a pro-war discussant increases the odds that one will exhibit pro-war views. This model includes the full sample of surveyed individuals (N = 1000) that is those who discuss the war with another person and those who do not engage in such discussions. Models 2 and 3 delve deeper into political discussions and thus include only the sample of individuals who reported discussing the war with someone (N = 503). Model 2 focuses on the role of relationship ties between the discussant and the respondent and how these ties affect opinions about the war (hypothesis 2a&b). Model 3 examines how the frequency of conversations impacts war attitudes (hypothesis 3).

Results from Model 1 (Table 1) show that having a pro-war discussant has a statistically significant impact on the respondent's pro-war attitudes. This supports hypothesis 1. The impact only holds for the first discussant, the first person the respondent named when asked about the individual with whom she/he most often discusses politics. It may be that the first discussant is considered the most credible person with whom one chooses to converse on foreign policy, and when influence occurs, it is because the first discussant, more so than the second one, is considered an expert on politics or at least a credible source of information. The model also shows that those who do not discuss the war with anyone are less likely to be pro-war than those who have discussants (variables Dummy 1 & 2). This is because individuals who have pro-war discussants are more likely to be exposed to pro-war messages than those without any discussants. The latter are likely to be more influenced by the national mood and media coverage, which in this case were opposed to the country's participation in the war.

Table 1. Logistical Analysis of Polish Support for the Country's Participation in the Iraq War

	Model 1	Model 2	Model 3
^a (Notes: *p < .1,** p < .05,***p < .01,N = 1000 (Model 1), N = 503 (Models 2& 3); two-tailed test).
Social Interactions
Discussant 1	1.083 (0.118)***	2.035 (0.496)***	2.121 (0.507)***
Discussant 2	−0.119 (0.348)	−0.050 (0.393)	−0.079 (0.393)
Dummy 1 (no 1st discussant)	−0.838 (0.217)***
Dummy 2 (no 2nd discussant)	0.276 (0.352)	0.491 (0.403)	0.443 (0.402)
Disc 1 × Frequency			−0.122 (0.146)
Frequency			−0.083 (0.139)
Disc 1 × Relationship		−0.124 (0.023)
Relationship		−0.316 (0.199)
Socioeconomic Factors
Education	−0.198 (0.201)	−0.312 (0.332)	−0.224 (0.341)
Gender	−1.028 (0.218)***	−1.106 (0.361)***	−1.263 (0.354)***
Professionals	0.751 (0.515)	0.593 (0.716)	0.535 (0.716)
Trade/Service	0.661 (0.527)	0.653 (0.919)	0.765 (0.858)
Age	0.039 (0.151)	0.099 (0.243)	0.057 (0.242)
Skilled Labor	0.171 (0.262)	0.618 (0.402)	0.705 (0.390)*
Farmer	−0.050 (0.666)	0.157 (0.981)	0.434 (0.928)
Income change	0.001 (0.110)	−0.136 (0.175)	−0.092 (0.173)
Place of Inhabitance	0.644 (0.232)***	−1.090 (0.363)***	−1.104 (0.365)***
Partisanship
Government Support	0.360 (0.187)*	0.759 (0.295)**	0.718 (0.298)**
Party View on Iraq	0.477 (0.172)***	0.459 (0.254)*	0.453 (0.254)*
International Trust
UK	0.064 (0.146)	−0.170 (0.218)	−0.160 (0.218)
France	−0.247 (0.139)*	−0.467 (0.216)**	−0.469 (0.214)**
USA	0.489 (0.137)***	0.830 (0.207)***	0.812 (0.210)***
Russia	0.034 (0.160)	−0.148 (0.256)	−0.112 (0.251)
Germany	0.020 (0.143)	0.182 (0.218)	0.134 (0.214)
Media
Print	0.268 (0.178)	0.237 (0.238)	0.221 (0.236)
Constant	−0.153 (0.664)	0.559 (1.037)	0.086 (1.044)

Predicted probabilities provide additional information about the variables' impact on attitude formation. Having a strongly pro-war discussant means there is a 63% probability that the respondent will harbor similar views (Table 2). When the discussant strongly opposes Polish involvement in the war, the probability that the respondent will be anti-war reaches 98%. This suggests a strong impact of the first discussant particularly in the direction of anti-war attitudes and is expected because local conversations take place in the midst of national debates that often set the tone for local discussions. It is, thus, easier to adopt an anti-war stand if your discussant is also anti-war because this reinforces the emergent national mood on the policy. When the discussant's view clashes with those dominating in the country, individuals are still more to likely follow the discussant, but some will also choose not to do so.

Table 2. Selected Predicted Probabilities for Supporting Polish Participation in the Iraq War

	Model 1
(Notes: Predicted probabilities show how a change in the variable of interest from minimum to maximum value affects the dependent variable while holding all other variables at their mean. Predicted probabilities are for those variables that are statistically significant in the first, main model).
Social Interactions
Discussant 1	(0.02) (0.63)
Dummy 1	(0.15) (0.07)
Socioeconomic Factors
Gender	(0.17) (0.07)
Place of Inhabitance	(0.13) (0.07)
Partisanship
Government Support	(0.10) (0.19)
Party View on Iraq	(0.06) (0.15)
International Trust
France	(0.13) (0.08)
USA	(0.07) (0.17)

Results from Model 2 (Table 1) demonstrate that relationship ties between the respondent and his/her discussant do not play an important role in the process of shaping attitudes during conversations about the war. Discussing the war with pro-war family is not more likely to increase the odds of adopting pro-war attitudes than discussing the subject with pro-war coworkers and neighbors. Neither is this true for weak ties. Thus I do not find support for hypothesis 2.

Findings from Model 3 (Table 1) show that increase in the frequency of discussions with pro- or anti-war people has little impact on shaping war attitudes. The interaction variable is not statistically significant, and I do not find support for hypothesis 3. It could be that the conversations' content and the credibility of the source are more important in shaping one's attitudes than the frequency of conversations about the war. My findings do not mean that the respondent and his/her discussant do not interact frequently, only that frequency of discussions about the war is not important in strengthening the influence. It is entirely possible that during political conversations about issues other than the war—and these types of conversations are common according to my survey data—one forms a perception about the discussant's credibility so that this credibility matters during subsequent discussions about specific topics regardless of how frequent these discussions are. Insights from the US context seem to support this assertion. For example, Kenny (1994) argues that perceptions of discussant's knowledge could affect the dynamics of discussions. Mendez and Osborn (2009) suggest that even if men and women frequently discuss politics with each other, perceptions that women are less credible as discussion partners could have an impact on opinion formation in discussion networks, thereby lending support to the idea that perceptions of discussants' credibility could affect the strength of influence and thus explain why the frequency of discussions is not as important in this study as initially assumed. This could also explain why I did not find statistical significance in Model 2 for the idea that relationship ties affect the strength of influence. If perceptions of credibility may be more important, then whether one talks to a spouse or a coworker should then matter less in the process of influence.

Turning to control the variables, I find that several variables are statistically significant in the expected direction across all models, including gender, place of inhabitance, and partisanship. Females are less likely than males to exhibit pro-war views as are the inhabitants of villages. Those who support the government's handling of the economy are more likely to support the war as are those who favor a political party with a pro-war platform. In terms of international trust, trusting the United States increases the probability that one will support the war, while trusting France is related to negative views of the war. Perhaps the most surprising finding is the lack of statistical significance for the media's effect. It seems that local discussions with one's social circles are more important in shaping attitudes on the war than exposure to the media's coverage of the policy, a finding that runs contrary to previous research on public opinion and foreign policy. This is because scholars focus predominantly on the US context where the extent of informal political discussions is not nearly as high as in Eastern Europe (Gibson 2001). Consequently, this would explain why the media's role in shaping attitudes has been seen as significant in previous research, but not so here. While majority of Poles in this study admit to regular media exposure, it is possible that information from the media is either ignored by some people in favor of information coming from interpersonal discussions or used by others as a starting point for these discussions. What follows is that during such discussions, individuals dissect information from the media, build upon it, or dismiss it, ultimately making the interactions more critical in molding individuals' views than the media exposure.

I now discuss the robustness of my findings. There is a potential for a selection bias. It may be the case that respondents choose to discuss the war with individuals who share their views in the first place. If that is the case, then the causal mechanism of influence is absent. Such possibility, however, is small for several reasons. First, findings on political disagreement among discussants show that level of disagreement is not lower among people who discuss politics with each other than among people who discuss nonpolitical issues with their discussants. Thus, while people may interact more frequently with those who share their general political preferences, they disagree on specific policies and do not necessarily avoid discussions with such people because of political disagreement (Huckfeldt and Mendez 2008). Mollenhors, Volker and Flap (2008) also argue that while people may seek to interact with people who are similar to them on key social dimensions when considering marriage partners, they care less about similarity when interacting with friends, coworkers, and neighbors. While the clustering of similar individuals does occur, especially with respect to race/ethnicity, gender, and education (McPherson, Smith-Lovin and Cook 2001), disagreement among those who discuss politics is not routinely shunned as often assumed. If that is the case, then it is unlikely that most individuals in my study would be concerned about systematically picking discussion partners whose views on the war resemble their own. It also implies that observed agreement among discussants is likely the result of influence that occurs during conversations.

Second, even if we were to assume that individuals select discussants who reflect their own views on the war, this would imply that individuals have an infinite pool of people to choose from to discuss a specific policy with a specific person who holds the same view on that issue. The demand for like-minded conversation partners on every political dimension, such as abortion, the EU, and so forth, is constrained by the opportunities people have to socialize with those who see eye to eye with them on different issues. The composition of our conversation partners is shaped by the social context in which we meet them. As such, Blau (1977) argues that demand for similarity in discussion partners is by itself not sufficient to explain who it is that we end up interacting with. Thus, even if people sought to discuss the war with pro-war friends, it is possible they may not be surrounded by friends who are pro-war that is the opportunity may be limited to consistently select among people those individuals who will agree with one on various political dimensions.

The data in my study support this claim. The respondents were asked whether they discuss politics with anyone, and if they said yes, they were asked whether they discuss the war with the person whom they identified as the political discussant. Over 80% of the people who named a discussant said they only discuss politics with one person and also discuss the war with that particular person, while 12% mentioned having two discussants. This shows that majority of respondents in my study do not have multiple political discussion partners that they could screen for specific political views to discuss the war with only those who share their own views. Therefore, I conclude that while a possibility exists that some individuals choose to discuss the war with those who already share their views on the war this is unlikely to be the case for majority of the respondents in this study. Instead, it is more plausible that individuals have been influenced to adopt specific views on the war.

Besides addressing the issue of selection bias, it is necessary to consider the possibility that the respondent and the discussant shape each other's views on the war. In such a case, the respondent could be influencing the discussant's views, while also being influenced. To isolate the discussant's independent effect on the respondent, I used instrumental variables to estimate the discussant's pro-war views based on exogenous demographic characteristics (education and occupation). I then obtained predicted probabilities for the discussants' pro-war views and captured the probability of the discussants' pro-war views free from the respondents' influence and used them to analyze the respondents' war attitudes. The results remained the same, thereby suggesting that the discussants have an independent influence on the respondent's pro-war views.

Conclusion

Research on attitude formation on foreign policy focuses predominantly on individual's socioeconomic background or the media's effect on shaping opinions. Yet understanding the impact of one's social circles, those individuals with whom people discuss politics, remains elusive. In this study, I identified three mechanisms through which discussions about foreign policy could shape attitudes. Specifically, I argued that individuals often turn to their friends, family, coworkers, and so on, to receive basic information about foreign policy and to learn about the policy's potential impact on their lives. Occasionally, social circles may also exert pressure on individuals to adopt specific views on the policy.

I tested my argument by analyzing Polish attitudes on the country's participation in the Iraq war in 2004. Empirical research on public opinion and foreign policy in Eastern Europe is limited, and large-N data on interpersonal discussions about foreign issues is nonexistent. It is often assumed that models of attitude formation can be easily imported from the US context to the rest of the world. Yet in some parts of the world, reliance on one's social circles for information and support is much greater than in others. Gibson's (2001) study on the pivotal role of informal social networks in helping individuals engage in politics and learn about democracy in Russia shows that such networks are more extensive there than in the US and thus have vast implications for human behavior. This study builds on such research by showing that in Poland, political discussions matter a great deal in shaping individual views on foreign policy. As such, this study not only advances theoretical research on public opinion and foreign policy by focusing more on one's social environment but also provides substantive insights about the process of attitude formation in the understudied part of the world.

Findings show that when interpersonal discussions are taken into account to explain Polish attitudes, the media's impact becomes less significant, although the latter is considered an important predictor of attitudes in the United States (Soroka 2003). It is because in the East European context informal political discussions are more prevalent (Gibson 2001), and as this study has shown, this has implications for how people make sense of information. Previous research has rarely tapped into these issues because of lack of systematic data on interpersonal discussions on foreign policy in Eastern Europe that would allow scholars to incorporate social networks into the study of attitude formation on foreign policy. Furthermore, even though interpersonal discussions have been studied in the context of US voting behavior (Huckfeldt and Sprague 1995), their impact on shaping foreign policy views has not been analyzed systematically. It is likely that such interactions matter in the United States as well when it comes to foreign policy views, but there is a strong likelihood the effect would be smaller than in Eastern Europe where these discussions are a norm.

Results from this study indicate that individuals most likely form their views by discussing the war with someone from their social circle with whom they already discuss politics regularly. Interestingly, relationship ties do not play a role in shaping attitudes. This most likely suggests that the credibility of the discussant, already established from previous conversations, matters more than the person's relationship to that discussant. The study's larger implication is that social interactions have the potential for shaping mass opinions. After all, if individuals form attitudes on foreign policy when conversing with others, then they, in turn, will be able to influence someone else. In other words, such discussions give rise to a network effect with vast implications for shaping the country's opinions on a massive scale.

While this study established the impact of interpersonal discussions on attitude formation on foreign policy, future research could focus on several issues. First, a comparative study of several countries could examine variation in influence that social circles exert on opinion formation. Here, I argued that such influence should be grater in Eastern Europe than in the United States because of greater engagement in informal political discussions in post-Communist societies. Empirical evidence is necessary to assess the utility of the approach in explaining views on foreign policy across different contexts. Second, one could also examine how the different mechanisms of influence that I identified here could shape the longevity of specific attitudes. For example, are individuals more likely to oppose wars in the long term because their discussions with others about the merits and perils of such policies involved extensive debates that allowed them to form deeper beliefs on the issue? Future studies could examine how the dynamics unfolding during informal discussions affect the durability of views.

1. 1

   Although the two approaches are often considered contradictory, researchers are increasingly recognizing the value of relying on insights from both paradigms, arguing that in real life, actors may behave in a way that is consistent with both logics (Risse 2002; Jupille et al. 2003). Jupille et al. (2003), for example, argue that utility is an ‘open-ended concept,’ leaving it open for altruistic/normative arguments to enter decision making, even if rationalist scholars do not commonly use such arguments (p.13).

2. 2

   Political experts are likely to be resistant to social persuasion, and with strongly developed preferences, they are likely to remain loyal to their view regardless of what happens during interpersonal discussions (Huckfeldt and Johnson 2002). Often, they serve as opinion leaders.

3. 3

   Individuals need not discuss the policy with everyone in the group to understand the emerging consensus. During discussions with a dominant member of the group, for example, one may be told about the opinions of others.

4. 4

   The advantage of using a national probability sample in studying the impact of social contacts on attitude formation is that such samples mostly generate independent social contacts among the respondents, thus minimizing the odds of spillover from one social contact to another (Fowler, Heaney, Nickerson, Padgett and Sinclair 2011).

5. 5

   The discussants' views in the interactive terms are coded from −1 to 1 instead of from −2 to 2 as is the case with the main explanatory variable in Model 1 because the interactive term with the expanded coding produces too many categories for which I lack enough observations. Hence, in all of the interactive variables, the discussants' supportive views are merged together as are unsupportive ones.

6. 6

   I rely on perceived media bias as reported by the discussants. While it would be ideal for the researcher to measure the bias directly, logistical problems prevent me from conducting content analysis. Given that some respondents named local publications, locating such periodicals becomes problematic. Since the media's impact is not the core focus of this paper, content analysis of local press is beyond the scope of the project.

7. 7

   This is consistent with other opinion polls conducted in Poland. For example, the Center for Public Opinion Analysis (Centrum Badania Opinii Społecznej) found that in August of 2004, that is the month during which I conducted my survey, 73% of Poles opposed their country's involvement in Iraq, while 23 supported it (CBOS 2004).