Partisanship in a Social Setting
Department of Political Science
THIS IS A DRAFT.
PLEASE CONTACT THE AUTHOR TO CITE OR FOR FURTHER DETAILS.
No factor has proven more power in explaining vote choice, issue positions, and candidate evaluations than one’s partisanship. Yet, virtually all work on the origins and effects of party identification ignore how social context might condition its application – that is, not how partisanship affects the choice of discussion partners, but rather how social context can fundamentally change the influence of partisanship on preferences. In this paper, I fill that gap with a novel experiment in which I demonstrate that social settings can overwhelm the impact of partisanship, even among individuals who are primed to be strong partisans. The central implication is that, going forward, work that explores how citizens apply their party identification in making decisions must account for their social context.
Over the past half century, the study of partisanship has provided critical insights into how individuals’ identifications with a particular political party can color their perceptions of the world and influence their issue evaluations and policy preferences. Recent research on partisanship has re-visited the well-known idea put forth by Campbell et al. (1960: 133) that views partisan identification as a perceptual screen. This more recent work is done under the guise of work on partisan motivated reasoning (Taber and Lodge 2006) and it shows that partisanship can dramatically shape how individuals evaluate information. For example, a Democrat (Republican) may reject a policy sponsored by a Republican (Democrat) when they would support the exact same policy if endorsed by Democrats (Republicans) (e.g., Bartels 2002; Redlawsk 2002; Taber and Lodge 2006; Druckman et al. n.d; Lavine et al. 2012). Lavine et al. (2012) demonstrate that this partisan bias is particularly powerful when one feels strongly about their partisan identification.
Yet, over the past sixty years, virtually all research on how partisanship shapes opinions ignores what has proven to be a critical factor in determining opinions more generally: social networks (e.g., Berelson et al. 1954; Druckman and Nelson 2003; Huckfeldt et al. 2004; Mutz 2006; Scheufele et al. 2006; Sinclair 2012). Instead, partisanship is treated as though it operates in complete social isolation. In this paper, I correct this omission by incorporating social context into the process by which partisan identification shapes opinion formation. I show that one’s social setting – whether composed of ideologically like-minded partisans or a mixed heterogeneous group of partisans – fundamentally affects how one uses and applies their partisan identification in evaluating political information and arriving at an opinion. In sum, I find that studying how partisan identification influences information evaluation and opinion formation without accounting for social setting may result in a vast mischaracterization of the influence of partisanship.
I begin in the next section by discussing the known attitudinal consequences of partisanship, before turning to my hypotheses for how these consequences may differ in various social settings. I then describe an experimental design by which I varied partisan strength (a la Lavine et al. 2012) and the composition of social settings (i.e., no social interactions, homogenous partisan interactions, or heterogeneous partisan interactions). Perhaps the most overwhelming result I present in this study is that if one compares prior strength of partisanship against the impact of social settings, it is the latter that dominates preference formation. I conclude with a discussion of how my results have implications for the study of partisan identification and public opinion formation in general.
Attitudinal Consequences of Partisanship
Decades of research suggest that partisanship, an individual’s adherence to one particular political party or platform over another, is highly stable over one’s lifetime (Green and Palmquist 1994). It continues to be “a powerful and pervasive influence on perceptions of political events” (Bartels 2002, 120) and subsequent political preferences. Indeed, an extensive amount of work demonstrates that partisanship influences – among other things – issue preferences (Jacoby 1988), vote choices (Bartels 2002), evaluations of the economy (Lewis-Beck et al. 2008), competence of political parties (Gerber and Huber 2010) and the blame attributed to them (Tilley and Hobolt 2011), and even assessments as seemingly objective as the color of a candidate’s skin (Caruso et al. 2009).
The powerful influence of partisanship can be attributed to directional motivated reasoning: a desire to reaffirm preexisting beliefs when processing new information (Kunda 1990). Psychological theories of cognitive consistency (Festinger 1957) explain this behavior by arguing that individuals are, above all, interested in reconciling new information with their existing beliefs. The consequence is that individuals rely on partisan cues to influence – and even determine – their political preferences (Rahn 1993). For example, a Democrat who supports President Obama and his Affordable Health Care Act will not only dismiss criticisms against it but will also be unlikely to tune into any radio or television programs featuring opponents of the Act, thus shielding himself from learning any new information that might have the potential to increase his understanding of the legislation or its alternatives. Democrats and Republicans are similarly likely to view the economy as doing well under their preferred political administration, even if they would be unhappy under the same conditions were the opposing party in power (e.g., Bartels 2002; Druckman et al. n.d.; Lavine et al. 2012).1
Thanks to partisan motivated reasoning, there are at least three distinct attitudinal responses we might see among individuals who hold a preexisting partisanship. First, they tend to support the option endorsed by the preferred party, regardless of how ideologically compatible that option may be (Lavine et al. 2012). Second, partisans may perceive their own party’s solutions to be highly effective, and to dismiss the opponent’s ideas as ineffective (Redlawsk 2002, 2004; Gerber and Huber 2010). And, finally, they are less likely to see out incongruent information that contradicts their priors (Redlawsk 2002). On one hand, partisan motivated reasoning provides helpful cues to guide decision-making (Druckman 2001), but some see these processes as normatively problematic. As Lavine and his co-authors explain (2012), this process of partisan motivated reasoning “raises deeply troubling questions about political representation… how can an electorate possibly reward or punish an incumbent party if it holds grossly distorted views of political conditions?” (chapter 5, p. 6).
Scholars have identified several factors that influence the degree to which individuals engage in partisan motivated reasoning, with a strict focus on individual-level attributes such as sophistication and prior opinion strength (e.g., Taber and Lodge 2006). Lavine et al. (2012) show that, among a host of these individual-level factors, the strongest moderator of partisan motivated reasoning is partisan ambivalence. Partisan attachments, like attitudes in general, are not strictly unidimensional but contain a mixture of both positive and negative evaluations (Martinez et al. 2005; Visser et al. 2006). Ambivalence, a distinct component of attitude strength, is defined as the degree to which an individual holds these conflicting evaluations. As Martinez and his coauthors explain (2005), an individual who is ambivalent may still “have strong opinions about a subject and perhaps even know more about it… but nevertheless feel conflicted” (2). Ambivalence plays a uniquely important role in the study of how partisanship influences in preferences. Partisans who are less ambivalent hold more consistently positive evaluations about their preferred party (Lavine et al. 2012). Their heightened sense of ingroup membership motivates them to defend their party in the face of counter-information (Nicholson 2012) and to engage in more partisan motivated reasoning (Lodge and Taber 2000; Taber and Lodge 2006; Lavine et al. 2012). As a result, partisans with more consistently positive evaluations of their party “show remarkably little learning or attitudinal updating during political campaigns or salient policy debates” (Lavine et al. 2012, Chapter 1, p. 22).
Preference Formation in Non-Social and Social Settings
The Effect of Partisan Strength on Motivated Reasoning
Partisan goals motivate individuals to “apply their reasoning powers in defense of a prior, specific conclusion (Kruglanski and Webster 1996)” (Taber and Lodge 2006, 756), which drives subsequent preferences towards preestablished party allegiances. This is all the more evident among partisans with strong party attachments, whose consistently positive evaluations of their party stimulate (a) a powerful preference for party-sponsored policies and (b) partisan motivated evaluations of political information.
Hypothesis 1a. Overall, strong (i.e., less ambivalent) partisans express a stronger preference for attitudinally congruent information – i.e. their own party’s policy – than do weak partisans.
Hypothesis 1b. Overall, strong (i.e., less ambivalent) partisans express perceive their own party’s policy to be more effective, and the opposing party’s policy to be less effective, than do weak partisans.
All the while, there is a “social flow” to political information (Huckfeldt et al. 1987) that is highly consequential for preference formation, yet completely neglected in the study of motivated reasoning. Throughout the course of a day, individuals are surrounded by others; we are, as Sinclair argues (2012), “social citizens”. Some of our networks, for example voluntary associations, tend to be homogeneous (Popierlarz and McPherson 1995; Mutz 2006), while other networks are more commonly quite diverse (Sinclair 2012). Huckfeldt et al. (2004) argue that “disagreement and heterogeneous preferences are the rule rather than the exception within the micro -environments surrounding individual citizens” (p. 17) and further research demonstrates that “the likelihood of encountering disagreement rises considerably as conversational networks become larger in size” (Price, Capella, and Nir 2002, 98; also see Granovetter 1973; Weimann 1982, Bramoulle and Rogers 2010). The degree to which individuals’ networks are homogeneous or heterogeneous is therefore a subject of some debate, but we can agree that our environments are highly social. Our varying social groups are often not formed on the basis of political preference or partisan compatibility (Sinclair 2012), yet these interactions frequently give way to political conversation nevertheless (Walsh 2004).
Scholarly attention has given considerable attention to the existence of social settings in so far as their partisan and issue preference composition (for e.g. Mutz 2004; Huckfeldt 2004) and their effects on participation (Mutz 2002) and persuasion (Sinclair 2012). Yet, beyond these studies, virtually no empirical work has examined how social setting interferes with an individual’s motivation to reconcile new information with his partisan allegiances. I will next turn to my hypotheses for how social settings – both homogeneous and heterogeneous in partisan composition – may do just that.
The Effect of Homogenous Groups on Motivated Reasoning
Evidence on the influence of attitudinal homogeneity on preference formation generally indicates that the like-mindedness of a social setting has a reaffirming effect. Druckman and Nelson (2003) invite participants to interact in a social setting wherein all group members are provided with the same issue frame. The authors find that certainty in this frame is affirmed and crystallized as a result of the homogeneous setting (also see Druckman 2004). This finding supports observational work on social context, which similarly demonstrates that discussions without dissent affirm common arguments reiterated within the group, as homogeneous discussants encourage one another’s mutual viewpoints (Mutz 2002). Participants in these settings become even more convinced by their agreed upon preference (see Isenberg 1986: 1141, Mendelberg 2002: 159). This can be explained by the reinforcement of a common group identity within the group. As the identity strengthens, it becomes more salient to the group members who then reconcile their identity with incoming information via directional motivated reasoning (Nicholson 2012; Druckman et al. n.d.).
I thus expect that respondents overall should engage in more partisan motivated reasoning when discussing political issues in a group composed of partisans who share a common partisan identity. The identity will become reinforced in this ideologically like-minded context and partisans will become even more likely to prefer their own party’s policy and will rate the opponent’s policy as more ineffective, as a result of the like-minded social setting.
Hypothesis 2a: Respondents in homogeneous groups (i.e., groups that include only fellow partisans) engage in more partisan motivated reasoning and prefer policy solutions that more closely represents their own party’s policy, as compared with those in non-social settings.
Hypothesis 2b: Respondents in homogeneous groups (i.e., groups that include only fellow partisans) perceive their own party’s policy to be more effective, and the opposing party’s policy to be less effective, as compared with those in non-social settings.
The Effect of Heterogeneous Groups on Motivated Reasoning
In diverse social settings, on the other hand, I expect individuals to respond quite differently. Observational data suggest that diverse discussion networks provoke more even-handed and considered electoral choices (Nir 2005). Cross-pressures from within one’s network causes attitudes to become more ambivalence (Mutz 2002), which in turn weakens the group cues that guide decision-making (Druckman and Nir 2008). Druckman (2004) finds that heterogeneous groups effectively introduce new perspectives and, with Nelson (2003), finds that these groups minimize the influence of preexisting issue frames over respondents’ evaluations of issues.
I thus expect that heterogeneity of discussion groups have a powerful influence over prior beliefs. This suggests that respondents who discuss political issues in heterogeneous networks should engage in less partisan motivated reasoning than do those in non-social settings. I expect that respondents in heterogeneous networks will weaken their partisan identity and become less defensive of their own party’s policy and less critical of the opponent.
Hypothesis 3a: Respondents in heterogeneous groups (i.e., groups that include other partisans) engage in less partisan motivated reasoning and prefer policy solutions that more closely represents the opposing party’s policy, as compared with those in non-social settings.
Hypothesis 3b: Respondents in heterogeneous groups (i.e., groups that include other partisans) perceive their own party’s policy to be less effective and the opposing party’s to be more effective, as compared with those in non-social settings.
The Influence of Partisan Strength Versus The Influence of Social Settings
Hypothesis 1 states that strong partisans engage in more partisan motivated reasoning but does not take social setting into account. Hypothesis 2 and 3 introduce the element of social setting to the study of motivated reasoning but do so without distinguishing between strong and weak party identifiers. An important question that has yet to be answered (and has yet to be even asked) in existing work is whether the influence of partisan strength is resistant to the influence of social setting. How does strength of partisanship drive preference formation in the face of differing social contexts? No existing theory provides for why one of these elements might overshadow the other, but there is empirical work that suggests the power of the social setting dominates the influence of partisan strength.
Sinclair (2012) demonstrates that the power of social context is considerable. She controls for factors that influence an individuals’ choice of social network and, in so doing, demonstrates that the network itself has a powerful influence on an individual’s vote choice. Sinclair makes the case that political preferences among individuals in a common social network become more compatible as a result of the social interaction, but common preferences do not form the basis of the network a priori – but her work does not take partisan strength into consideration. Huckfeldt et al. (2004) help in this regard, arguing that even “strong partisans are not immune to the political messages that are filtered through networks of political communication” (63). I thus expect that when both strong and weak partisans find themselves in social settings, the power of the social setting will overcome the influence of partisanship and will determine the degree to which motivated reasoning influences preferences. Whereas strong partisans tend to engage in more partisan motivated reasoning than do weak partisans, I expect that weak partisans in homogeneous settings will engage in more partisan motivated reasoning than will strong partisans in heterogeneous settings.
Hypothesis 4a: Weak partisans in homogeneous social settings engage in more partisan motivated reasoning and prefer policy solutions that more closely represents their own party’s policy, as compared with strong partisans in heterogeneous networks
Hypothesis 4b: Weak partisans in homogeneous social settings perceive their own party’s policy to be more effective, and the opposing party’s policy to be less effective, as compared with strong partisans in heterogeneous networks
In sum, I expect that partisan strength will determine the degree to which motivated reasoning biases preferences, when social context is not taken into account: strong partisans will be more biased than will weak partisans, just as previous scholars find (e.g. Taber and Lodge; Redlawsk; Lavine et al.) (H1). When we look only social context, homogeneous network discussants will engage in more ideologically driven motivated reasoning than will those in without groups (H2) and those in heterogeneous networks will engage in less motivated reasoning (H3). When partisan strength and social context face each other, however, the influence of the social network will prove dominant. Weak partisans in homogeneous networks will be more biased that will strong partisans in heterogeneous networks (H4). I thus expect, overall, that partisanship without taking social context into account does not provide sufficient information to predict the degree to which motivated reasoning influences preferences.
A final measured outcome in this study has to do with the “downstream effects” of the various treatment groups. Downstream effects are those residual effects of experimental treatments that may subsequently influence respondents’ behavior of attitude (Green and Gerber 2002). One effect that non-social, homogeneous or heterogeneous social settings may have on discussants is their subsequent interest in participating in homogeneous or diverse social settings in the future. A known consequence of strong partisanship as opposed to weak partisanship is a relative disinterest in seeking out opposing information (Redlawsk 2002). In Redlawsk’s study, in which respondents are presented with the option to seek out information in isolation, strong partisans demonstrate a preference for learning information about candidates they already support. I thus expect to find that my hypotheses hold with respect to downstream effects. Strong partisans, as opposed to weak partisans, will prefer a more homogeneous discussion group for future discussions (H1). Partisans in like-minded social settings will have a greater preference for like-minded discussion groups, as compared to those in non-social settings (H2), and those in heterogeneous settings will prefer a more bipartisan group (H3). Finally, weak partisans in homogeneous settings will have a greater preference for homogeneous discussion groups, as compared to strong partisans in heterogeneous settings who will express a greater preference for bipartisan social settings (H4).
To test these hypotheses, I required (i) a sample population of partisans and (ii) policy issues for the partisans to discuss in social settings. For participants, I turned to the undergraduate population at a large Midwestern university between October 2011 and February 2012. Three hundred and seventy-nine undergraduate students participated in the study in return for academic course credit. In line with previous experimental work on partisan behavior (e.g. Levendusky 2010; Druckman et al. n.d.), I excluded the 35 respondents who identified as pure independents (also see Bullock 2011) and I grouped leaning independents with the party they profess to prefer. Leaners behave like partisans when it comes to opinions and vote choice (Lascher and Korey 2011; Magleby et al. 2011), justifying this typical approach (which is the norm in studies of partisan elite influence). This left me with three hundred and forty-four research subjects who identified as a Democrat, a Republican, or as a leaning independent with a preference for one of the two parties. Among them, an overwhelming proportion identified as Democrats (79%) as opposed to Republicans (21%). Homogeneous and non-social settings were thus comprised of all Democrats, while heterogeneous groups were comprised of both Democrats and Republicans. The analyses in this study therefore focus exclusively on responses among Democratic respondents. This limits the generalizability of the findings to some extent, but there exists neither theoretical nor empirical evidence to suggest that Republicans would respond differently. The student population has variance on the characteristics of study and thus provides a suitable sample for this study (see Druckman and Kam 2011). Students’ political identities are perhaps not as crystallized as an adult population, and we may therefore see more movement in preferences among students than we would among an adult sample but, if this is the case, then these results indicate an upper-bound of the effects one would find in a non-student sample.
I will now turn to a detailed description of the study, which involves two policy issues that respondents were asked to evaluate. I will begin by describing these policy issues, followed by the experimental manipulations of partisan strength and group composition. I will next outline the precise experimental conditions, and, finally, the dependent variables that were used to operationalize the results of this study.
I selected two prominent public policies for participants in this study to evaluate and consider: energy policy and health care policy. By using two issues, instead of just one, I was able to test my hypotheses in two different settings, which provides robustness. I selected energy policy and health care policy as discussion topics because they are high-profile issues with which most Americans are familiar, including undergraduate political science students. A brief content analysis of the top five print newspaper publications in the United States during the one year leading up to this experiment revealed 1,707 articles that include both the terms “energy policy” and “gas prices,” and over 3,000 articles that include both the terms “health care policy” and “health insurance.” This is significantly more coverage than for other policy areas; articles including the term “immigration policy" numbered at only 401, and articles including the term “education policy” numbered only 298. By incorporating salient issues with relevance to young adults, sample bias based on the all-student population is tempered, since salient issues are more difficult to manipulate with treatments.
Both issues also provide conservative tests of my hypotheses due to the fact that there are topics about which the population is highly polarized. Whereas less heated debates (for example regarding less well-known local issues, or even fictitious issues generated solely for the purpose of the experiment) might allow for more malleable shifts in opinion, these issues are on which individuals may already have strong opinions; indeed, Democrats and Republicans are somewhat polarized on these issues, consistently reporting opposing viewpoints. For example, the Pew Research Center found in 2011 that 83% of Democrats and Independents leaning toward the Democratic party supported alternative fuels as an energy policy solution. Conversely, favorability among Republicans and independents leaning toward the Republican party was only 53% during this same time period.2
This study incorporates two unique experimental treatments. First, I employ actual social interaction into the experimental design. Whereas scholars often rely on self-reports of social network composition (for e.g. Mutz 2002; Huckfelt and Sprague 2002) or experimentally manipulate the degree to which respondents anticipate social interaction but never actually engage in such interaction (Tetlock and Kam 1987; Tetlock 1983; Groenendyk 2012), this experimental design involves randomly assigned interaction in small groups. Second, I experimentally manipulate party strength by inducing either ambivalent or univalent evaluations of the respondent’s preferred party. I am thus able in this experiment to identify the causal relationship between social context and partisan motivated reasoning. I now turn to describing each experimental treatment in detail.
Social Setting Treatment
One month before attending the experimental study, respondents began their involvement in this study by first completing a brief anonymous survey, in which they provided their demographic traits, including their political party identification. Based on their self-reported partisanship, I randomly assigned respondents to one of three social setting conditions: a homogeneous group of 8 Democrats; a heterogeneous group of 4 Democrats and 4 Republicans; or a non-social setting, in which group members do not interact with one another. When the respondents were called to the study center, they arrived one group at a time, without knowing the ideological composition of their group or even that they would interact with the group at all.