The Distribution and Determinants of Socially Supplied Political Expertise

Abstract

Recent work on social influence has highlighted the importance of socially supplied political expertise, crediting it with strengthening attitudes, resolving ambivalence, and encouraging political participation. However, in focusing on the consequences of socially supplied political expertise, scholars have made the implicit assumption that citizens have equal access to this resource and have largely ignored its distribution. Given that individuals are constrained by their social contexts, we are particularly troubled by this oversight, and thus use two nationally representative data sources to explore the distribution of expertise among and throughout the social networks of citizens. We find consistent evidence that existing resource inequalities reinforce the unequal distribution of expertise in social networks—a gender-moderated pattern that involvement in civil society may help remedy.

Keywords

social networks expertise gender civil society political knowledge

In a series of high-profile findings, Diana Mutz (2002a, 2002b, 2006) has posed a dilemma for democratic theorists: democracy can be either deliberative or participatory, but not both. There are many reasonable responses to this claim (e.g., Klofstad, Sokhey, & McClurg, 2013), though a particularly important one is that people may rely on political experts to provide them with information that is essential for political participation, regardless of the amount of disagreement that they encounter in their daily lives (e.g., McClurg, 2006). In this article we address this claim, describing the distribution of political expertise in the electorate, exploring the degree to which existing inequalities structure access to political experts, and considering the forces that may facilitate the acquisition of this important political resource.

Political expertise is important from a variety of well-established theoretical perspectives. Downs (1957) first argued that citizens’ social circles function as information shortcuts, and that people seek out socially supplied information from like-minded political experts. This assumption has been widely adopted (e.g., Huckfeldt & Sprague, 1995),¹ motivating the study of the effects of social access to political expertise, and enabling the argument that political experts undergird democracy by allowing an efficient division of labor in the broader electorate (Berelson, Lazarsfeld, & McPhee, 1954; Huckfeldt, 2001). Others have followed up on this point, presenting socially supplied expertise as a sort of democratic panacea that strengthens attitudes, resolves ambivalence, and encourages political participation (Huckfeldt, 2001; McClurg, 2006; Richey, 2008; Ryan, 2010; though see Sokhey & McClurg, 2012 for mixed evidence). As Huckfeldt succinctly describes it, “[S]ocial communication creates the potential for modest amounts of political expertise to go a long way in enhancing the performance of democratic politics” (2001, p. 425).

While socially supplied expertise appears to be an efficient method of information diffusion and a potential solution to Mutz’s (2006) democratic dilemma, we are troubled by the fact that the actual distribution of expertise in the mass electorate has received little scholarly attention. In reality, individuals face constraints in network selection (Huckfeldt & Sprague, 1995); networks are coercive and regulate individuals’ exposure to broader information environments (Granovetter, 1973; Huckfeldt, Beck, Dalton, & Levine, 1995); social contexts—like neighborhoods and churches (Djupe & Gilbert, 2009; Mutz & Mondak, 2006)—are broadly structured by socioeconomic status and race (Huckfeldt, 1986; Leighley & Matsubayashi, 2009). Thus, instead of an electorate that reaps the benefits of “trickle-down” knowledge, the reality may be unequal distribution. That is, political expertise may be concentrated in pockets, which may reinforce existing civic inequalities. What is perhaps more pernicious is that involvement in elements of civil society may be unlikely to unlock access to political experts in the same way as in the case of civic skill development (Burns, Schlozman, & Verba 2001; Djupe, Sokhey, & Gilbert 2007; Verba, Schlozman, & Brady, 1995). Therefore, we pursue two simple but important questions: Who has access to political experts, and what structures this access? Our answer, in brief, is that individuals with fewer politically relevant resources will have less access to political experts, and that the “gender system” and the contexts of civil society may reinforce or reduce that deficit depending on their structure. In the following sections, we first specify our theoretical motivations before introducing the data, methods, and results. Using two well-known studies—both of which are nationally representative and are separated by nearly a decade—we investigate access to expertise within the American public. We close with a discussion of the import of the findings as they pertain to both recent debates in the social influence literature and to citizens’ democratic potential more generally.

The Expertise Gap

Do all Americans have access to socially supplied political expertise? Extant work provides few answers, which is a peculiar oversight for a literature whose goal is to consider the social contingencies of citizenship (for brief exceptions, see Lake & Huckfeldt, 1998; McClurg, 2006; Richey, 2008, p. 540; for related work on the perceptions of expertise, see Mendez & Osborn, 2010).²Investigations into the distribution of other civic resources provide a starting place. Delli-Carpini and Keeter (1996; see also Luskin, 1987, 1990) detail how learning political knowledge depends on ability, motivation, and opportunity; in their civic voluntarism model, Verba et al. (1995) argue that political participation is furthered by the acquisition of civic resources, engagement with politics, and recruitment by others. While the particulars of the two approaches differ, the general frameworks are similar. We blend these into a similar three-part framework, which when applied to the questions at hand suggests that motivated individuals with existing resources will have greater access to political experts. Of course, our application of this framework is also informed by our question’s social nature.

We consider access to social expertise as an inherently relational problem and argue that much can be gained by asking a simple question: On what bases can a relationship with a political expert be sustained? Rather than thinking of a political expert as a heuristic that can be sought out at will, our approach to access to expertise suggests that discussant relationships are based on some mutually beneficial exchange (Homans, 1958). One way to understand this problem is via the adolescent retort, “It takes one to know one.” In employing this phrase, we do not mean that only experts can recognize other experts, but that experts know experts because they enjoy sharing their knowledge of a subject. Homophily based on interests and attributes is perhaps one of the oldest assumptions in the social sciences (see, e.g., Bentley, 1995 [1908]; Madison, 1787; Truman, 1951), and is fundamental to the study of social networks (e.g., McPherson, Smith-Lovin, & Cook, 2001). Under this assumption, the resource of expertise should remain locked up among the resourceful (experts).

What is more difficult to understand is why we might see unbalanced relationships, in which expertise is uneven in a discussion dyad. This is also the more important thing to focus on for democratic functioning. Such imbalances allow for expertise to diffuse, which can enable more effective decision making by voters. Given that both discussants are not experts, the key is to consider the alternate bases on which a relationship might be sustained. Of course, there are many answers to such a question, but we focus on two broad factors that may have macro consequences for the diffusion of social expertise—civil society and the gender system (Ridgeway & Smith-Lovin, 1999).

Though we develop these points further in the subsequent sections, the core of the argument is as follows: Some elements of civil society—especially organizations—have the potential to bring together a diverse set of citizens, force interaction among them, and thereby provide an alternate, shared interest that could sustain relationships across levels of political expertise. The gender system works in the same way. If “most interactions between men and women occur in the structural context of roles or status relationships that are unequal” (Ridgeway & Smith-Lovin, 1999, p. 191), then persistent interaction across lines of inequality provides a basis for the diffusion of political expertise. From here, we further develop the three-part framework (ability, motivation, and opportunity), discuss its overlap with the gender system, and together consider the social implications for access to political expertise.

Ability

Ability is approximated by traditional measures of education and income, though of course individuals may be naturally smart and able to learn effectively without traditional schooling (Delli-Carpini & Keeter, 1996). Still, one of the persistent findings over time is that education is related to political engagement, and a wide range of reasons are offered why. The value of political engagement may be learned in higher education as an essential component of adult life; in addition, the highly educated may have a firmer grasp of how the political process affects their interests. Moreover, education is not simply an individual measure, but is also a proxy for a place in the social structure (Huckfeldt & Sprague, 1995; Nie, Junn, & Stehlik-Barry, 1996). With greater formal education comes a host of social opportunities in terms of information and mobilization that would not otherwise be available. We should expect these measures of personal resources to be positively related to having access to a political expert.

Motivation

Engagement and motivation can mean different things—attitudinal orientations to politics, such as political interest, or information acquisition behaviors, such as media use. Because these factors are so deeply intertwined, Zaller (1992) argues that they should be considered in tandem, resulting in his political awareness measure grounded in factual information. Following this well-worn approach, we use political knowledge as a basis for a measuring the omnibus notion of political awareness. If it takes political information to carry on a conversation with someone who is already informed, a respondent’s level of political knowledge should predict having an expert in the discussion network.

One particularly important argument to confront is the claim that political knowledge is independent of access to expertise in the network (McClurg, 2006). Using Huckfeldt and Sprague’s (2007) Indianapolis-St. Louis (ISL) study (1996), McClurg finds no significant relationship between respondent expertise and network political expertise (see also Mendez & Osborn, 2010). This finding is important, and it warrants reexamination using nationally representative data. If the result holds for the broader electorate, then citizens can approach potential discussants independent of their own resource levels. In essence, a null relationship supports the “Downsian” perspective on the social information search (1957; see also Huckfeldt & Sprague, 1995). On the other hand, a positive relationship between respondent and network political expertise provides an estimate of the potency of resource differentials in structuring access to expertise, and points toward the model of exchange theory. In short, whether there is a relationship is quite important for understanding the role that network expertise may play in overcoming access barriers and rescuing citizens from political ignorance.

Motivation to seek out information should also apply to finding an expert. Accordingly, we expect that a higher level of political interest should lead to a higher level of political expertise within one’s own discussion network. But again, motivation is both an individual attribute and contextual indicator; it is contingent on having a reason to care about politics, which could be driven by the electoral environment (Lau & Redlawsk, 2006) or the social context (Djupe & Gilbert, 2009). The conditions that inhibit interest may also work to inhibit the supply of experts from which to choose.

Opportunity—Civil Society and the Contexts of Adult Life

“Opportunity” is a fairly broad and diverse category, and several individual attributes can be thought of as intrinsically related to opportunity. Of course, it is also common to think about opportunity as environmental contingency—the supply of information in an environment will affect the individual’s ability to acquire it. An environment can be as large as a campaign environment or media market, or something as intimate as a church or softball team. The essential point is that regardless of motivation, individuals are constrained by their placement in these environments (e.g., Huckfeldt & Sprague, 1995).

While we would ideally have a wide variety of measures of the actual supply of expertise in tightly drawn and diverse social contexts, these data do not exist.³ In the analyses that follow, we use the proportion of network discussants that an individual has in the church, workplace, and the neighborhood to approximate the “boundedness” of an individual’s choices (though these are clearly not the only possible sources of discussants). Voluntary organizations, like churches, are often quite diverse and typically gather individuals together with some purpose other than politics in mind (Djupe & Gilbert, 2009). When individuals participate in voluntary organizations, they bolster their chances of meeting a diverse group of individuals, though voluntary association membership does not level all differences as membership is also affected by resource differentials (Musick, Wilson, & Bynum, 2000; though see Verba et al. [1995, p. 320], who note that houses of worship are the most egalitarian voluntary associational type in the United States). One particularly important feature of associational life that makes access to political expertise more egalitarian is its formal organization. The bureaucratic structure of social affairs and governance in organizations serves to force social interaction among people who might not otherwise choose to interact (see Djupe & Gilbert, 2009; Neiheisel, Djupe, & Sokhey, 2009).

Thus, on balance, we suspect that involvement in civil society institutions, such as churches, will combat the unequal distribution of expert discussants driven by personal resource inequalities. This is essentially the story that Mutz and Mondak (2006) present regarding the workplace. Coworkers are unavoidable (Neuberger, 1996) and may be more diverse than potential discussion partners from other contexts (though see Sokhey & Mockabee, 2012). Moreover, because the workplace is well suited to discussion of public issues and corporations have adopted flatter organizational structures, Mutz and Mondak argue that “[t]here are beneficial societal consequences that flow from the cross-cutting conversation engendered by participation in the workforce” (2006, p.140; see also Conover, Searing, & Crewe, 2002). If these previous findings hold, then the workplace should also help to distribute political expertise throughout the population.

Contrary to this, neighborhoods most often do not have an explicit organizational structure that channels social interaction. The choice of discussants in a neighborhood is much more free form than in an association or workplace. Moreover, like the workplace, the supply of potential discussion partners in a neighborhood is much more constrained by social status than in associations, and hence should display more homogeneity in expertise. Together, these attributes of neighborhoods should inhibit the acquisition of experts for the resource starved, and bolster it for the more resourceful.

Two other features of networks beyond the source of discussion partners are important to consider. Both their size and the degree to which they host political agreement should boost the proportion of experts in the network. More expansive networks are probabilistically more likely to host an expert, and expanded networks of political discussants may also send a strong signal of political engagement (McClurg, 2006). Political agreement has been associated with more political discussion and political participation, which would seem related to expertise in the network (e.g., McClurg, 2006). Huckfeldt (2001) has also identified agreement as informing a bias in the perception of an expert. Likewise, it is possible to see agreement as a motivation to learn from a source (e.g., Chen, Duckworth, & Chaiken, 1999; Downs, 1957), allowing an individual to acquire information about a discussant’s level of expertise.

The Gender System

It is important to acknowledge that gender overlays our entire theoretical scheme. Despite considerable progress, women still lag behind men in educational attainment, income, and other politically relevant resources such as political interest, knowledge, and efficacy (Burns et al., 2001; Delli-Carpini & Keeter, 1996; Schlozman, Burns, & Verba, 1994; Verba, Burns, & Schlozman, 1997). If “it takes one to know one” from an exchange perspective, then women are at a disadvantage in finding and maintaining relationships with political experts. As Huckfeldt and Sprague describe it, “[M]en will not turn to women for information and ideas about politics if women’s perspectives are different and constantly devalued” (Huckfeldt & Sprague, 1995, p. 194; see also Sapiro, 1983). In the often-cited 1984 South Bend Study, men choose women as political discussion partners at lower rates, even among spouses (Huckfeldt & Sprague, 1995); in the 1996 Indianapolis-St. Louis Study, Mendez and Osborn (2010) find that both men and woman downplay women’s political expertise, and that men choose female discussion partners at lower rates.

However, while these findings further our understanding of gendered perceptions of relationships, they do not do as much to shed light on how men and women choose their discussants. In fact, we will see that a slim majority of women’s discussion partners are men (though this rate is lower than the rate among men), and that the overall rates of expertise in women’s networks are higher than in men’s networks. Even though men tend to devalue women’s views and knowledge bases, and may not perceive them to be discussion partners (dynamics that are not salutary), men still communicate about political matters with women, and women can benefit from access to political experts of whatever gender.

Ridgeway and Smith-Lovin (1999) refer to these dynamics as the “gender system.” Even if the behaviors of men and women of equal status tend not to differ greatly, men and women still interact across lines of status, and they interact with the underlying belief that women have less political competence. While persistent interaction governed by these beliefs serves to maintain the gender system, it also has the paradoxical effect of exposing a broader selection of society to what is perceived to be political expertise.

Though gender dynamics cut across all categories of explanation (ability, motivation, and opportunity), many previous explorations of access to expertise do not include a control for gender, let alone pursue a more systematic analysis (Huckfeldt, 2001; Lake & Huckfeldt, 1998; McClurg, 2006; though see Richey, 2008, p. 540, where a gender variable is included but not discussed). Of course, Burns et al. (2001) carefully examine the relationship between gender and participation through the lens of the Civic Voluntarism Model (Verba et al., 1995), but do little more than mention the idea of network effects (appropriate question batteries were not included in the 1990 Citizen Participation Study). To a certain degree, this part of our theory builds on Mendez and Osborn’s (2010) contributions using the 1996 ISL data, though we focus our attention on the broader American electorate, and consider the distribution and determinants of access to expertise, rather than its perceptions and the consequences of those perceptions for discussion rates.

Hypotheses

Here, we summarize our major hypotheses:

Hypothesis 1: Those with greater resources stocks (e.g., education) should have more expertise in their networks.

Hypothesis 2: Those with greater political knowledge and political interest should have more expertise in their networks.

Hypothesis 3: Larger networks should host more political expertise.

Hypothesis 4: More agreeable networks should host more political expertise.

Hypothesis 5: The more that discussion partners are sourced from churches, the more expertise the less resourceful should have, and the less expertise the more resourceful should have. This should work in the opposite way for neighborhood discussants, augmenting network expertise for the resourceful and weakening it for the less resourceful. (Civil Society Hypothesis)

Hypothesis 6:To the extent that men and women have male discussion partners, their networks should contain more expertise. (Gender System Hypothesis)

Data

Ideally, we would test our theory using contextual estimates of the supplies of political expertise to which people have access. However, aside from the 1984 South Bend Study, these data do not exist. Instead, we turn to two well-known, nationally representative sample surveys that contain measures of the presence of political expertise in social networks. In 1992, a consortium of scholars assembled the Cross-National Election Studies (CNES; Beck, Dalton, & Huckfeldt, 2000), from which we use the American component. Interviews were conducted with 1,318 main respondents generated from a cluster sample of 39 counties. Essential for our purposes, the CNES asked a battery of questions—including perceived expertise—for up to five discussants. The procedure used is referred to as a multiplex name generator; first, up to four discussants were obtained using an “important matters” generator (Marsden, 1987),⁴ and then up to one additional discussant was acquired via a secondary, “political” generator. While Klofstad, McClurg, and Rolfe’s (2009) analysis of the CNES parses out the 5th, “political” discussant, we include that person—if named—in our analyses that follow. As Figure 1 will show, this has important consequences for estimates of expertise in the electorate. The second nationally representative sample we employ is the 2000 ANES (American National Election Study; Burns, Kinder, Rosenstone, Sapiro, & The National Election Studies, 2001),which included interviews with 1,807 respondents. The ANES also hosted a network battery using a “political” name generator; this gathered data on up to four political discussants.

Figure 1.

The distribution of expertise in political discussion networks, 1992 CNES & 2000 ANES.

To be clear, by “access to political expertise,” we mean the political expertise ratings of listed political discussion partners. Both data sets measure discussant expertise in the same manner, maintaining continuity from the early 1984 South Bend study (Huckfeldt & Sprague, 1995): “Generally speaking, how much do you think < discussant name > knows about politics? Would you say: a great deal, an average amount, or not much at all.” Huckfeldt (2001) has provided solid evidence that the objective measure is valid and predictable; this was done in the 1996 ISL data by matching political knowledge data from main respondents with independent, follow-up interviews conducted with named discussants. This is not to say that there are not effects due to misperception, or that the perception of expertise is not relative.⁵ Instead, existing work suggests that actual knowledge levels are tightly connected to perceived knowledge levels.

The Distribution of Socially Supplied Expertise

We begin with a description of the distribution of exposure to socially supplied political expertise. Figure 1 presents several views of social networks from both the 2000 ANES and the 1992 CNES. Panels A and B show the proportion of respondents who named 0 to 4/5 discussion partners (dark bars; panel A from the ANES, panel B from the CNES), as well as the percent of each named discussant that could be labeled a political expert (light bars; these were, for purposes of this presentation and further analyses, discussants who were perceived to know a “great deal” about politics).

Though the proportion of respondents able to name a discussant declines as more discussants are prompted, the proportion of discussants who are identified as being “expert” remains stable in both data sets across the range of the first four discussants. For the ANES (panel A), over a third of each named discussant is perceived to be an expert about politics, and over 40% of first-named discussants are experts; for the CNES (panel B), just under 30% of each named discussant is an expert. The exception is the 5th named discussant in the CNES—about 60% were able to name this discussant, and over one third of these were named experts.

Panels C and D present a slightly different story; they show the proportion of respondent networks (y-axis) with varying proportions of expertise (x-axis). About 55% of ANES networks and roughly 50% of CNES networks contain no experts (i.e., no discussants that know a great deal about politics). In the ANES, a sizable minority does have access to at least one political expert. Overall, we see considerable overlap between the networks in the two data sets—a finding that jells with investigations into variation produced by these name generators (Huckfeldt & Mendez, 2008; Klofstad, McClurg, & Rolfe, 2009; Mendez & Osborn, 2010; for a comparison among political name generators, see Sokhey & Djupe, in press).⁶ People may be prompted to choose political specialists for their networks (the ANES political generator), but their close confidants with whom they discuss important matters (the CNES important matters generator) appear to be generalists that are more than capable of engaging political topics. In other words, core networks look fairly similar in terms of political expertise, regardless of which of these prompts is used.⁷

Based on the results in Figure 1, it seems hard to claim that access to socially supplied expertise is widespread. Small majorities in both data sets do not have access to a political expert, thus making it harder to claim that expertise can solve the problems of low political participation and ignorance in America. However, it remains unclear from these figures whether this reflects the unequal distribution of politically relevant resources, differences in motivation to acquire political information, or social-structural barriers.

Thus, in Table 1, we use the 2000 ANES and the 1992 CNES to examine how the portions of networks composed of at least one political expert are distributed, and whether they are statistically distinguishable amongst subgroups. Though we must exercise a bit of caution as resources tend to cluster, we focus on individual covariates that traditionally divide the population by their resource bases; we also focus on salient political resources. In-line with work on political participation (e.g., Verba et al., 1995), expected differences emerge. The differences are statistically significant for most variables, though the nature of these differences is quite distinct for the 1992 CNES. For every variable examined in 2000 (save education), a majority of the networks in both groups contain at least one political expert—that is, the significant differences are between majorities of networks. In every split examined for 1992 (save race), the significant differences are between a majority of networks in one group, and a minority of networks in the other.

Table 1.

The Proportion of Networks With an Expert, by Subgroups.

Variable (ANES N/ CNES N)	2000 ANES	1992 CNES
Whites (946/1139)	.60	.50
Non-Whites (206/174)	.52	.45
p =	.03	.24
Some college or more (803/803)	.63	.57
High school graduate or less (347/513)	.49	.36
p =	.00	.00
US$65,000/US$75,000 and above a year (153/150)	.71	.65
Below US$65,000/US$75,000 a year (881/1078)	.56	.48
p =	.00	.00
Women (628/736)	.61	.51
Men (519/582)	.55	.46
p =	.07	.07
High political knowledge (565)	.64	—
Low political knowledge (586)	.53	—
p =	.00
Very interested in politics (406/744)	.69	.55
Somewhat, not very interested (746/571)	.52	.41
p =	.00	.00
Independents, incl. leaners (426/492)	.62	.44
Partisans (711/797)	.57	.52
p =	.10	.00

Source: 2000 ANES and 1992 CNES.

Note: CNES = Cross-National Election Studies. ANES = American National Election Study.

In the ANES, about 8% more of White respondents have access to at least one expert compared to non-White respondents—a difference that is not statistically significant in the 1992 data, and that (as we will see) fades once controls for added for education and income in the multivariate results. The difference in access between those with only a high school education versus those with more advanced schooling is about 15 points in the 2000 data and over 20 points in the 1992 CNES. In both samples, a similarly large gap in access to at least one political expert—roughly 15 points—appears between the top 10% of earners (making over US$65,000) and the bottom 90%. Whether it is because of differences in the social opportunity structure of living and working around people with political knowledge, or because of an expanded interest in seeking out experts, these fundamental resources allow for expanded access to network sources of expertise.

Surprisingly, given the differences in socioeconomic status between men and women, in the 2000 ANES we find that women have more political experts in their networks than do men; this is also the case in the 1992 data. Since women have lower levels of political knowledge than men (a difference of 1 point out of 6 using the ANES, p = .000; results not shown), it is possible that women merely perceive greater levels of expertise. To examine this possibility, we used the 1996 ISL, as it included snowball interviews of discussants, and directly assessed discussants’ political knowledge (results not shown).⁸ In that study, there is no aggregate difference between male and female respondents in the perceived political expertise of their networks (as there is in the 2000 ANES), but there is a significant difference in the objectively measured political knowledge of discussants. In the ISL, men’s discussants have higher levels of objective knowledge (a difference of .12 on a 3-point scale, p = .016, n = 1,470), which tells us that women may over report or men may under report the political expertise of their discussants to at least a small extent (see also Mendez & Osborn, 2010).

In the remainder of Table 1, we concern ourselves with differences in network expertise levels for those with high and low levels of significant political resources. Even allowing for potential over-reporting bias from those with lower political knowledge, in the 2000 data there is still a significant, 11-point network expertise gap between those with high- and low-levels of political knowledge—a gap that grows to around 15 points in both data sets when examining those very interested in politics versus those who report being less interested. Interestingly, in the ANES a few percentage points more independents (including “leaners”) report having an expert discussant when compared to partisans; this pattern is reversed in 1992, which may speak to the dynamics of that year, and its legitimate third-party candidate.

The Determinants of Socially Supplied Network Expertise

As a next step, in Table 2 we present the estimates from logistic regressions predicting whether respondents in the ANES and CNES have access to at least one political expert. In Appendix B, we report OLS models predicting average levels of network expertise.⁹ In the discussion of the results that follows, we focus our attention on this first, dichotomous measure of access (Table 2), as we see the crucial distinction being whether people have access to any expert. Put differently, we argue that levels of expertise relative to some accessible amount are less critical than access to any expertise relative to none—this is particularly true if the influence of expertise in a network declines when more experts are present (Richey, 2008). That said, the results are quite similar (and consistent across data sets), whether we model access to at least one expert or average network expertise.

Table 2.

The Determinants of Having At Least One Political Expert in a Network (Logistic Regression Estimates).

	2000 ANES		1992 CNES
Network indicators	Coeff	(S.E.)	Coeff	(S.E.)
Network size	.50	(.06)***	.52	(.05)***
Network agreement	.44	(.18)**	.03	(.18)
Network prop same sex as respondent	−.55	(.26)**	−.52	(.30)*
Male × net prop same sex	1.55	(.42)***	.88	(.44)**
Network proportion of neighbors	−2.44	(1.66)	1.14	(1.59)
Education × net prop of neighbors	.22	(.12)*	−.09	(.12)
Network prop of church members	4.75	(1.91)**	4.05	(1.83)**
Education × net prop of church members	−.35	(.14)***	−.28	(.13)**
Network prop of coworkers	−.21	(1.26)	−1.54	(1.67)
Education × net prop coworkers	.01	(.09)	.08	(.12)
Personal resources
Partisan strength	−.16	(.08)**	.09	(.07)
Political interest	.16	(.06)***	.33	(.11)***
Political knowledge	.15	(.05)***	—	—
Attend church	−.04	(.16)	−.39	(.19)**
Organization member	.21	(.14)	.18	(.21)
Education	.04	(.05)	.12	(.04)***
Income	.04	(.03)*	.00	(.05)
Age	−.01	(.00)**	.01	(.00)*
Male	−1.61	(.31)***	−.73	(.29)***
White	.20	(.19)	−.04	(.21)
Constant	−1.44	(.74)*	−17.14	(8.42)**

Sources: 2000 ANES, 1992 CNES. *p < .10 . **p < .05. ***p < .01(two-tailed tests)

Note: CNES = Cross-National Election Studies. ANES = American National Election Study.

0 = no experts in network. 1 = 1 or more experts in network. All estimates are rounded to two decimal places.

ANES results: Pseudo R² = .13, LL = −602.51, n = 1,018

CNES results: Pseudo R² = .14, LL = −655.55, n = 1,102

In the model specifications (identical for Table 2 and the online Appendix tables), we include independent variables that fall into two camps: network indicators and personal resources. We also include three interactions between resources and network measures: (a) education and the proportion of discussants from the church, (b) education and the proportion of discussants from the neighborhood, and (c) respondent gender and proportion of the network that is the same gender as the main respondent. Following our theoretical framework and stated hypotheses, these interactions are designed to test the conditional effects of particular contexts (“opportunity”) on access to expertise, given resource levels (abilities) of respondents. To restate from the previous section, we expect greater concentrations of church-member discussants to reduce unequal access by lowering the likelihood of having an expert among the highly educated and by increasing that likelihood among those with lower than average levels of education. We expect the opposite pattern with respect to the neighborhood (Civil Society Hypothesis). Finally, with respect to gender, we expect the likelihood of having an expert discussant to increase for both men and women as the portion of male discussants in the network increases (Gender System Hypothesis).¹⁰ The same logic holds when modeling the amount of expertise in networks (online Appendix Tables A.1 and A.2).

Looking first at network properties, network size is a consistent predictor of having an expert discussant across both data sets. Whether the name generator asks for “important matters” (CNES) or explicitly political discussants (ANES), larger core discussion networks predict the presence of at least one expert.

Though ANES respondents in more agreeable networks are more likely to have at least one expert discussant (the coefficient is similarly signed, though insignificant, in the CNES), we remain speculative about why this is the case. Mutz, (2006) reports that those with more education have more agreeable networks (an indication that bias shades the coloration of expertise); it is also possible that expertise only emerges after prolonged conversation, which is something that disagreement potentially inhibits. Unfortunately, better—and ideally, longitudinal—data are needed to fully address this point.

Civic Society Results

We also examine how discussant context affects the likelihood of finding an expert, conditional on an individual’s personal resources (as captured by formal education). Does the context exacerbate resource differentials, or does it level the playing field? We see extremely consistent results across the two data sets. The interaction term between education and the proportion of the network from the church is negative and significant in both cases; the interaction term between respondent gender and the proportion of the network sharing the gender of the respondent is significant and positive in both cases. Of course, interaction terms themselves are less than revealing (and require proper hypothesis testing—Brambor, Clark, & Golder, 2006). Thus, in Figure 2 we graph the marginal effects for both the ANES and CNES estimates to aid in interpretation and to demonstrate the ranges of statistical significance.¹¹

Figure 2.

Estimated interaction effects predicting the presence of an expert in the network.

The graphs serve to confirm much of the “Civil Society Hypothesis” advanced previously. Panels A and B display the effect of an increase in the proportion of network discussants that are church and neighborhood discussants, respectively, on the probability of having access to at least one political expert.¹² Values are calculated across the full range of formal education (with all other variables in the model held to their mean values), and 90% confidence intervals appear as gray lines.

In the ANES, for those possessing higher levels of formal education, having a network with more church-based discussants drops the probability of having an expert discussant (Panel A, graph B)—the result is statistically significant for individuals who graduated from college and beyond (the right-most part of the graph), which is just under a third of respondents. In the CNES, this effect is not statistically significant for those with higher education (the confidence intervals straddle zero), though substantively it is quite similar. Oppositely, across both data sets, having more church-based discussants boosts the probability of having an expert for those with low levels of formal education. In both the CNES and the ANES, the effect holds through high school graduation (or nearly in the case of the ANES), which corresponds to roughly 40% of each of these samples.

Turning to Panel B, we see partial support for our expectations regarding resources, neighborhood discussants, and access to expertise. The results are not significant—and are incorrectly signed—–for the 1992 sample. However, for those with higher levels of education in the ANES, having a higher proportion of neighbors in one’s core network increases the probability of having access to a political expert thus potentially exacerbating inequalities. The results are statistically distinguishable from zero beginning with those who report 15+ years of formal education (about 38% of respondents). We note too that the workplace exerts no statistically significant effects—main or interactive—in either data set.

Looking for a moment at the bottom portion of Table 2, we see that many of the personal resources included in the model structure access to network expertise. As the CNES did not include political knowledge items, we were only able to include such a measure in the model for the 2000 ANES. The significant, positive coefficient marks a relationship between personal expertise and socially supplied expertise. This has serious implications for the availability of this network resource in the American electorate (a point to which we will return shortly).¹³ Partisan strength is a negative predictor in the ANES and a positive one in the CNES, but only emerges as significant in the case of the former. Political interest boosts the likelihood of having at least one expert in both samples. Men (with no male discussants; the interpretation of the main effect given the interaction) are less likely to have an expert in their network. Race and organizational membership have null effects on the likelihood of having at least one political expert. Those who attend church are less likely to have an expert in their network (significant only in the CNES), while age has opposing—but substantively tiny—effects across the two data sets.

Results for Gender

Returning to the interactions, we consider our “Gender System Hypothesis.” Simply stated, we observe strong and consistent evidence of gender-based distinctions in the distribution of political expertise across these two studies. In the models, a sizable and statistically significant relationship emerges between the gender composition of networks and the reported presence of a political expert, conditional on respondent gender. Controlling for a host of other resource-based explanations, the results indicate that gender does indeed moderate the relationship between network gender homogeneity and network political expertise. Figure 2 (Panel C) graphs this interaction term, presenting the effect of being male on the probability of having at least one expert discussant, conditioned on the portion of the network is male (90% confidence intervals appear in grey). In the CNES, the effect is statistically significant until the network is 60% of the same gender as the respondent; in the ANES it holds up until the 80% mark.

The plots flesh out the story told by the Table 2 estimates, helping to convey important and expected findings: women are more likely to have access to at least one political expert in their network (given the coefficient on “male”); both men and women have a higher probability of having an expert discussant when their networks are composed of more men (given the coefficients on gender proportion of the network and the interaction term). Men’s probabilities of having an expert discussant are pushed up as the proportion of men in the network increases; for women, the probability declines as the proportion of women in the network increases.

That women have lower levels of political interest and political knowledge in these (and other contemporary) samples necessitates that we qualify the argument that it takes an expert to know an expert. However, it does not lay bare the reason why experts more densely populate women’s networks. In our estimation, three possibilities stand out: First, psychologists have found that—contrary to popular belief—“traditional” gender differences in personality traits have only emerged in Western, developed nations (see, e.g., Schmitt, Realo, Voracek, & Allik, 2008); thus, it is possible that political specialization may be a byproduct of this sexual dimorphism. Second, and related, because of resource gaps between men and women, women may tend to seek out political information subsidies which political experts can supply. From this perspective, the presence of experts in women’s networks is the result of the rational pursuit of information. However, there is also a third possibility—that women’s access to political expertise comes through relationships with men that are not characterized by patterns of mutual exchange (Djupe & Sokhey, 2009). If motivation is important to fully process information from a source, then one-sided discussion may be less than fruitful. In the next and final section, we reflect further on these and the other results, noting their import for work on social influence and for democratic functioning more generally.

Discussion and Conclusion

Existing work on socially supplied expertise has either focused on its perception (e.g., Huckfeldt, 2001) or behavioral consequences (e.g., McClurg, 2006), and has generally excluded gender from consideration (e.g., Lake & Huckfeldt, 1998; though see Mendez & Osborn, 2010). In this article we step back to offer a critical assessment: socially supplied expertise cannot benefit those who do not have it, and those who do not have it do not constitute a random sample of the population.

Looking at the American electorate in two distinct snapshots, we find that slim majorities of the citizenry have no access to this important shortcut—one that individuals can use to supplement meager stocks of political information. If all individuals without access to an expert freely choose to avoid this resource, we could write it off as apathy, mistrust, or political disinterest. Unfortunately, we find evidence that social ignorance is systemic and that some individuals who lack experts are locked into this scarcity by their own lack of knowledge and other social-structural factors. For many, political expertise cannot function as civic welfare because a system of delivery simply does not exist.

In our ANES model, we find that those with more political knowledge have a greater concentration of expertise in their social networks (see also the online Appendix). This is an important counter to previous findings of a null relationship between political knowledge and social expertise (e.g., McClurg’s [2006] and Mendez & Osborn’s [2010] examinations of the 1996 Indianapolis-St. Louis data). In short, our results indicate that access to expertise is structured by a variety of factors, and speak not only to social-structural position (e.g., living in the right neighborhood), but the needs of maintaining relationships. This finding also has normative importance: social expertise is not randomly distributed and tends to be concentrated in resource-rich portions of the electorate—a pattern that reinforces rather than levels existing maldistributions of political resources.

In this effort we have also inquired about the impact of gender on distributions of expertise. Women—who tend to be slightly less politically informed and interested than men—have access to expertise in higher concentrations than men. Women also have more male than female discussion partners, though men have more male discussion partners in absolute terms.¹⁴ The fact that women have been found to pursue homophilous individual and associational affiliations with other women (e.g., McPherson & Smith-Lovin, 1986; McPherson et al., 2001) suggests that the incidence of expertise in women’s networks may not always be entirely welcome (Djupe & Sokhey, 2009); whether or not it is always entirely politically useful is another question (e.g., Atkeson & Rapoport, 2003; Hansen, 1997).

Lastly, we find consistent evidence of a nuanced role for civil society when it comes to the distribution and determinants of expertise. Lake and Huckfeldt (1998) found that organizational memberships augment expertise in networks—a finding we partially replicate with regard to the church. But, the particular nature of the context supplying discussants is critically important. While Mutz and Mondak (2006) find the workplace important because of its potential to expose people to disagreement, we find no evidence that this context facilitates access to expertise. By contrast, we find that the neighborhood potentially exacerbates disparities between the resource rich and poor (as measured via formal education, and at least in the 2000 ANES).

It is important to remember that since organizational contexts—like churches—direct social interaction through small, purposive groups, that individual motivation to pursue expert discussants is constrained. Viewed from another angle, civil society serves to distribute expertise. If the highly educated with discussants sourced from churches have the same level of network expertise as those with low levels of education, then we may be witnessing how diverse relationships are created and expertise is distributed through civic organizations. Without an organization channeling social interaction (e.g., in a context such as a neighborhood), individual difference is allowed to dictate choice, and the expected disparities in network expertise in the population are present.

While the exploration of whether individuals can acquire crucial resources from social intimates is an important task with many remaining questions, this analysis points to the importance of remembering that networks are embedded in larger contexts (civil society) and systems (a gendered social order), and that those contexts both give shape to (Huckfeldt, 1986; Huckfeldt & Sprague, 1995) and affect the influence of networks (Djupe & Gilbert, 2009; Mondak, 1995). The integration of multiple levels of politics may complicate the analysis, but it brings essential empirical and normative democratic problems into view.

Footnotes

Acknowledgements

We wish to thank Editor Gaines and the reviewers for a thorough, thoughtful review process that improved the final product. We also thank Ted Jelen, Jeannette Mendez, and Tracy Osborn for comments on an earlier version of the article. All errors are our own.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Author’s Note

Final version submitted to American Politics Research, March 22, 2013.

Notes

Author Biographies

Paul A. Djupe is an associate professor of Political Science at Denison University, Granville, Ohio, specializing in religion and politics, social networks, and public opinion. He is the coauthor of The Political Influence of Churches (2009, Cambridge) and the forthcoming book God Talk: Experimenting with the Religious Causes of Public Opinion (Temple).

Anand Edward Sokhey is an assistant professor of Political Science at the University of Colorado, Boulder, specializing in the role that social influence plays in voting behavior, political participation, and opinion formation, with a special focus on how people employ their networks when acquiring and processing information.

References

Ahn

T. K.

Huckfeldt

Ryan

J. B.

(2010). Communication, influence, and informational asymmetries among voters. Political Psychology, 31, 763-787.

Atkeson

L. R.

Rapoport

R. B.

(2003). The more things change the more they remain the same: Examining gender differences in political attitude expression, 1952-2000. Public Opinion Quarterly, 67, 495-521.

Beck

Dalton

R. J.

Huckfeldt

(2000). Cross-National Election Studies: United States Study 1992 (2nd ICPSR version) [Computer file]. Columbus, OH: Paul A. Beck, Ohio State University/Irvine, CA: Russell J. Dalton, University of California/Bloomington, IN: Robert Huckfeldt, Indiana University [producers], 1993. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor].

Bentley

(1995). The process of government. New Brunswick, NJ: Transaction.

Berelson

Lazarsfeld

McPhee

(1954). Voting. Chicago, IL: University of Chicago Press.

Brambor

Clark

W. R.

Golder

(2006). Understanding interaction analyses: Improving empirical analyses. Political Analysis, 14, 63-82.

Burns

Kinder

D. R.

Rosenstone

S. J.

Sapiro

& The National Election Studies. (2001) National election studies, 2000: Pre-/post-election study [data set]. Ann Arbor, MI: University of Michigan, Center for Political Studies [producer and distributor].

Burns

Schlozman

K. L.

Verba

(2001). The private roots of public action. Cambridge, UK: Harvard University Press.

Chen

Duckworth

Chaiken

(1999). Motivated heuristic and systematic processing. Psychological Inquiry, 10(1), 44-49.

10.

Conover

P. J.

Searing

D. D.

Crewe

I. M.

(2002). The deliberative potential of political discussion. British Journal of Political Science, 32, 21-62.

11.

Delli-Carpini

M. X.

Keeter

(1993). Measuring political knowledge: Putting first things first. American Journal of Political Science, 37, 1179-1206.

12.

Delli-Carpini

M. X.

Keeter

(1996). What Americans know about politics and why it matters. New Haven, CT: Yale University Press.

13.

Djupe

P. A.

Gilbert

C. P.

(2009). The political influence of churches. New York, NY: Cambridge University Press.

14.

Djupe

P. A.

Sokhey

A. E.

(2009). Examining the gendered nature of political networks. Paper prepared for delivery at the annual meeting of the APSA, Toronto, Canada.

15.

Djupe

P. A.

Sokhey

A. E.

Gilbert

C. P.

(2007). Present but not accounted for? Gender differences in civic resource acquisition. American Journal of Political Science, 51, 906-920.

16.

Downs

(1957). An economic theory of democracy. New York, NY: Harper and Row.

17.

Erisen

(2012). The effect of social networks on the quality of political thinking. Political Psychology, 33, 839-865.

18.

Granovetter

(1973). The strength of weak ties. American Journal of Sociology, 78, 1360-1380.

19.

Hansen

S. B.

(1997). Talking about politics: Gender and contextual effects on political proselytizing. Journal of Politics, 59(1), 73-103.

20.

Homans

G. C.

(1958). Social behavior as exchange. American Journal of Sociology, 63, 597-606.

21.

Huckfeldt

(2001). The social communication of political expertise. American Journal of Political Science, 45, 425-438.

22.

Huckfeldt

Beck

P. A.

Dalton

R. J.

Levine

(1995). Political environments, cohesive social groups, and the communication of public opinion. American Journal of Political Science, 39, 1025-1054.

23.

Huckfeldt

Mendez

J. M.

(2008). Moths, flames, and political engagement: Managing disagreement within communication networks. Journal of Politics, 70(1), 83-96.

24.

Huckfeldt

R. R.

(1986). Politics in context: Assimilation and conflict in urban neighborhoods. New York, NY: Agathon Press.

25.

Huckfeldt

Sprague

(1995). Citizens, politics, and social communication. New York, NY: Cambridge University Press.

26.

Huckfeldt

Sprague

(2007). Indianapolis-St. Louis election study, 1996-1997. ( ICPSR02962-v2). [Computer file]. Bibliographic Citation: Bloomington, IN: Indiana University Center for Survey Research [producer], 2000. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor]. doi:10.3886/ICPSR02962.

27.

Klofstad

C. A.

McClurg

S. D.

Rolfe

(2009). Measurement of political discussion networks: A comparison of two “name generator” procedures. Public Opinion Quarterly, 73, 462-483.

28.

Klofstad

C. A.

Sokhey

A. E.

McClurg

S. D.

(2013). Disagreeing about disagreement: How conflict in social networks affects political behavior. American Journal of Political Science, 57(1), 120-134.

29.

Lake

Huckfeldt

(1998). Social capital, social networks, and political participation. Political Psychology, 19, 567-584.

30.

Lau

R. R.

Redlawsk

D. P.

(1997). Voting correctly. American Political Science Review, 91, 585-599.

31.

Lau

R. R.

Redlawsk

D. P.

(2006). How voters decide: Information processing during election campaigns. New York, NY: Cambridge University Press.

32.

Leighley

Matsubayashi

(2009). The implications of class, race and ethnicity for political networks. American Politics Research, 37, 824-855.

33.

Luskin

(1987). Measuring political sophistication. American Journal of Political Science, 31, 856-899.

34.

Luskin

(1990). Explaining political sophistication. Political Behavior, 12, 331-361.

35.

Madison

(1787, Nov 22). No. 10: The Union A Check on Faction. Daily Advertiser.

36.

Marsden

P. V.

(1987). Core discussion networks of Americans. American Sociological Review, 52(1), 122-131.

37.

McClurg

S. D.

(2006). The electoral relevance of political talk: Examining disagreement and expertise effects in social networks on political participation. American Journal of Political Science, 50, 737-754.

38.

McPherson

J. M.

Smith-Lovin

(1986). Sex segregation in voluntary associations. American Sociological Review, 51(1), 61-79.

39.

McPherson

J. M.

Smith-Lovin

Cook

J. M.

(2001). Birds of a feather: Homophily in social networks. Annual Review of Sociology, 27, 415-444.

40.

Mendez

Osborn

(2010). Gender and the perception of knowledge in political discussion. Political Research Quarterly, 63, 269-279.

41.

Mondak

J. J.

(1995). Nothing to read: Newspapers and elections in a social experiment. Ann Arbor: University of Michigan Press.

42.

Musick

M. A.

Wilson

Bynum

W. B.

Jr. (2000). Race and formal volunteering: The differential effects of class and religion. Social Forces, 78, 1539-1570.

43.

Mutz

D. C.

(2002a). Cross-cutting social networks: Testing democratic theory in practice. American Political Science Review, 96, 111-126.

44.

Mutz

D. C.

(2002b). The consequences of cross-cutting networks for political participation. American Journal of Political Science, 46, 838-855.

45.

Mutz

D. C.

(2006). Hearing the other side: Deliberative versus participatory democracy. New York, NY: Cambridge University Press.

46.

Mutz

D. C.

Mondak

J. J.

(2006). The workplace as a context for cross-cutting discussion. Journal of Politics, 68(1), 140-155.

47.

Neiheisel

J. R.

Djupe

P. A.

Sokhey

A. E.

(2009). Veni, vidi, disseri: Churches and the promise of democratic deliberation. American Politics Research, 37, 614-643.

48.

Neuberger

(1996). Relationships between colleagues. In Auhagen

A. E.

von Salisch

(Eds.), The diversity of human relationships (pp. 269-288). New York, NY: Cambridge University Press.

49.

Nie

N. H.

Junn

Stehlik-Barry

(1996). Education and democratic citizenship in America. Chicago, IL: University of Chicago Press.

50.

Richey

(2008). The autoregressive influence of social network political knowledge on voting behavior. British Journal of Political Science, 38, 527-542.

51.

Ridgeway

C. L.

Smith-Lovin

(1999). The gender system and interaction. Annual Review of Sociology, 25, 191-216.

52.

Ryan

J. B.

(2010). The effects of network expertise and biases on vote choice. Political Communication, 27(1), 44-58.

53.

Sapiro

(1983). The political integration of women. Urbana: University of Illinois Press.

54.

Schlozman

K. L.

Burns

Verba

(1994). Gender and the pathways to participation: The role of resources. Journal of Politics, 56, 963-990.

55.

Schmitt

D. P.

Realo

Voracek

Allik

(2008). Why can’t a man be more like a woman? Sex differences in big five personality traits across 55 cultures. Journal of Personality and Social Psychology, 94(1), 168-182.

56.

Sokhey

A. E.

Djupe

P. A.

(in press). Name generation in interpersonal political network data: Results from a series of experiments. Social Networks.(Advance online publication). doi: https://dx-doi-org.web.bisu.edu.cn/10.1016/j.socnet.2013.02.002

57.

Sokhey

A. E.

McClurg

S. D.

(2012). Social networks and correct voting. Journal of Politics, 74, 751-764.

58.

Sokhey

A. E.

Mockabee

(2012). Reexamining political discussion and disagreement in church networks: An exit poll assessment. Politics and Religion, 5, 253-279.

59.

Truman

(1951). The governmental process. New York, NY: Alfred A. Knopf.

60.

Verba

Burns

Schlozman

K. L.

(1997). Knowing and caring about politics: Gender and political engagement. Journal of Politics, 59, 1051-1072.

61.

Verba

Schlozman

K. L.

Brady

H. E.

(1995). Voice and equality: Civic volunteerism in American politics. Cambridge, MA: Harvard University Press.

62.

Zaller

J. R.

(1992). The nature and origins of mass opinion. New York, NY: Cambridge University Press.