The “Who” Matters: Types of Interpersonal Relationships and Avoidance of Political Disagreement

Abstract

Research consistently finds that we discuss politics most often with our strong ties (i.e., our close, intimate others). As our strong ties tend to be more politically similar to us than not, the conclusion is that everyday political discussions are overwhelmingly characterized by real or perceived political agreement. However, this scenario may paint only a partial portrait of everyday political discussion. Neglected is the distinction between politically similar discussion partners, on the one hand, and similarity of views expressed during conversation, on the other. Although our strong ties may be more politically similar to us than not, they may, paradoxically, be just the people with whom we are likely to express disagreement. Indeed, this study illustrates that although discussion with strong ties increases the probability of agreement, it simultaneously increases the likelihood of discussing disagreement.

Keywords:

• interpersonal relationships
• political disagreement
• political discussion

Notes

1. Although some study designs (e.g., Huckfeldt & Sprague, 1995) capture actual rather than perceived agreement (i.e., measures from respondents as well as their discussants), financial and other logistical reasons often restrict researchers to perceptions alone (i.e., measures from only the respondent). Most research devotes little attention to the distinction between actual political agreement and political congruency. Indeed, political communication research comparing perceptions with actual agreement suggests that accuracy may be sufficient to employ measures of perceived agreement as actual agreement (Huckfeldt & Sprague, 1995). However, decades of research in social psychology have demonstrated errors in perception of others—both individuals and collectives (see Eveland & Glynn, 2008). Furthermore, congruency is conceptually and operationally distinct from actual agreement. Perceptions may be more relevant for some research questions (Mutz, 2006), reality may be the focal point of other inquiries, and comparison of perceptions to reality may be important to still other research endeavors. For instance, research examining the influence of disagreement on political or social tolerance may find it most profitable to focus on perceptions. On the other hand, research examining how individual opinions contribute to the outcome of a group decision-making process may be more interested in the actual composition of individual opinions prior to discussion. Thus, we feel it is important that researchers begin to make distinctions between political congruency and actual agreement, including use of appropriate terminology.

2. Some might argue that prior disagreement is necessary for avoidance of disagreement to be relevant. That is, how do we know we disagree with someone if we haven't discussed the disagreement before? However, expression of disagreement is not a necessary condition for the suspicion that disagreement exists. That is, we often make inferences about others based on observation or indirect indicators. For instance, we often assume professors are liberal and business owners conservative; we assume African Americans are Democrats and southern Caucasians are Republicans. These presumptions are not necessarily accurate as applied to any specific instance. Nevertheless, we often use these sorts of heuristics, based on profession, geography, religiosity, or physical appearance, to name a few.

3. Our gender distribution, based on a quota, is nearly perfectly split. The mean income in our study is closest to the category “$50,000–$75,000,” and the American Community Survey 2006–2008 reports a mean of $63,000. Our sample tends to underrepresent those in the lowest education categories—less than a high school diploma and high school diploma—and overrepresent those with some college or an associate's degree, but our study matches the American Community Survey very closely in terms of percentages in the highest education categories of bachelor's degree and graduate degree. The mean age in our study, at 47.8, is nearly identical to the 2008 American National Election Study (ANES). And our percentage of self-reported Democrats also matches nearly perfectly the results of the ANES, although our study slightly underestimates Republicans (approximately 5%) and overestimates Independents (approximately 8%) compared to the ANES.

4. Each respondent was asked to report his or her gender (50% men, 50% women, as an equal distribution was secured by quotas), age in years (M = 48.8, SD = 14.8), highest level of education completed (M = 3.9, between “some college” and an “associate or technical degree”; median = 3, “some college”), and family income (M = 3.6, between “$25,000 to $49,999” and “$50,000 to $74,999”; Mdn = 3, “$25,000 to $49,999”). Political interest was captured by a single question asking respondents to agree or disagree on a 5-point Likert scale to the statement, “I'm very interested in politics and public affairs” (M = 3.8, SD = 1.1; strong agreement = 5). Respondents were also asked how they identified themselves politically (36.8% Democrats, 31.8% Republicans, 24.7% Independents, 6.8% other). In addition to political party affiliation, respondents were asked two questions concerning their political ideology, one regarding social issues and the other regarding economic issues. The answers to these two questions were averaged for each individual to derive a single measure of political ideology (7-point scale, 7 = very conservative; M = 4.1, SD = 1.7, r = .80). Ideological strength was measured by folding the scales for social and economic political ideology at the midpoint. The scores for social and economic political ideological strength were averaged for each individual to form a single measure of ideological strength (M = 1.4, SD = 1.0). Two different measures were considered to serve as variables for political discussion network size. An estimate of size could be derived from the number of individuals listed by the respondents in the discussant generator, or, alternatively, a separate question asked participants about how many people they discuss politics with during a typical week. As the discussant generator was capped at three discussants, and strong ties tend to be listed first in name generators, the overall network size question is more likely to capture the periphery of an individual's network, or his or her weak ties. Thus, political discussion network size was measured by asking respondents, “In a typical week, about how many people do you talk about politics and public affairs with?” (Min. = 0, Max. = 100; M = 4.2, SD = 5.6). Respondents were also asked to report their overall political discussion frequency (independent of any single discussion partner), captured using a single item that asked respondents, “How often do you talk about politics and public affairs?” (0 = not at all, 7 = very often; M = 3.1, SD = 2.1).

5. Although this 10.4% figure departs considerably from the results of the name generator in the 2000 ANES—in which considerably more report no discussion partners—it matches almost perfectly with the findings of Huckfeldt and Sprague's (1995) South Bend study, which found 11% unable to report a single discussion partner. Moreover, our mean network size (derived from this name generator) of approximately 2.45 corresponds closely with the means of 2 to 2.5 reported across several other studies (see Klofstad, McClurg, & Rolfe, 2009). In any case, the number of alters named via name generator questions is heavily influenced by both interviewer characteristics (Marsden, 2003) and respondent level of cooperation (McPherson, Smith-Lovin, & Brashears, 2009). Unlike the ANES and the South Bend studies, our study is not susceptible to interviewer effects because it was self-administered.

6. Granovetter (1973)argued that strong ties were particularly likely to have highly overlapping networks. That is, the stronger the tie between two individuals, the more ties these two individuals will have in common. Thus, strong ties comprise a relatively dense component of an individual's interpersonal network. This observation is critical to Granovetter's argument that strong ties are likely to provide redundant information, whereas weak ties are more likely to convey new or novel information. In this study, we are not particularly concerned with the extent of overlap in an individual's political discussion network. That is, our hypotheses are not based on the assumption that strong ties are more likely to know and interact with many of the same people. Rather, we speculate that the intimacy component of strong ties makes them particularly likely to exhibit political congruency as well as elicit discussion of disagreement. We operationalize strong and weak ties accordingly.

7. Differences in significant predictors between the two avoidance measures were only apparent for the influence of perceived knowledge of the discussion partner. Perceived knowledge was a significant predictor of avoidance of issues, but it was not a significant predictor of avoidance of specific issue aspects. Age and political ideology were marginally significant predictors of avoidance of issues; however, they did not approach significance for avoidance of specific issue aspects.

8. The following formulas may be used to change log odds into predicted odds and predicted probabilities. To calculate predicted odds: exp(η). To calculate predicted probability: .

9. Dyadic-level controls include perceived level of discussant political knowledge and frequency of political discussion with the discussant. Perceived knowledge may influence political congruency, as we may be more inclined to agree with those we perceive as sophisticated, intelligent, or otherwise politically informed than with those we perceive as politically ignorant or incompetent. Frequency of discussion with a particular discussion partner could lead to either real or perceived congruency over time. Furthermore, frequency of discussion itself is also typically considered to be an indicator of the strength of the relationship.

10. Dyadic-level controls include perceived level of discussant political knowledge, frequency of political discussion with the discussant, and congruent party affiliation. Hesitancy to express disagreement may be stronger when a discussant is perceived as politically knowledgeable. Frequency of discussion could function similarly to relationship strength, such that the more we discuss with a particular discussion partner, the more comfortable we are with this person and the more willing we are to discuss matters of disagreement. Political congruency should influence conflict avoidance, as the more perceived agreement, the less need or opportunity to avoid disagreement.