Cultural Talk or Cultural Walk? Highbrow Tastes and Network Quality

.


Introduction
Network members, such as friends, acquaintances, or colleagues, often share tastes and interests, for example in music, food, and sports.Network members also offer help, for example when searching for a job, a home, or investors for a new business venture.Social science research has mapped the homophily of networks and the resources individuals can extract from them.This research has also addressed the processes through which individuals convert networks into material and symbolic resources, for example wealth, power, and reputation (e.g., Burt, 2004;Granovetter, 1973;Ibarra, 1993;Lin, 2000Lin, , 2001;;Putnam, 2000).The takeaway from this research is that networks provide tangible benefits.
A prominent strand of research within this larger literature addresses how networks shape, and are shaped by, cultural tastes (e.g., An and Western, 2019).This research shows that cultural tastes are positively associated with network quality (i.e., ties' socioeconomic status [SES] and resources; Erickson, 1996;Lizardo, 2013;Meuleman, 2021) and stability (Edelmann and Vaisey, 2014;Selfhout et al., 2009).In particular, the taste for highbrow culture, such as going to an opera, ballet, or art museum (e.g., De Graaf et al., 2000), has been argued to be a signal of high cultural and economic status that individuals can convert into high-quality networks (Bourdieu, 1984;Lamont and Lareau, 1988).If true, this implies that highbrow tastes provide access to network resources and assist in reproducing inequality.
While empirical research finds positive associations between cultural tastes and network quality and stability, it is largely uninformative about the origin of these associations (e.g., Erickson, 1996;Lizardo, 2006).In this paper, we propose that familiarity with highbrow culture is a necessary, but not a sufficient, condition for converting highbrow tastes into high-quality networks: individuals must also manifest these tastes socially.To address this hypothesis, we analyze if social manifestations of highbrow tastes, i.e., the ways in which individuals communicate their tastes in social interactions with network ties, mediate the positive association between highbrow tastes and network quality (Collins, 1988;DiMaggio, 1987DiMaggio, , 2011;;Edelmann and Vaisey, 2014;Lizardo, 2006Lizardo, , 2016)).We measure social manifestations empirically via panel data we collected in the Netherlands that include two manifestations of highbrow tastes: cultural talk (the frequency of conversations with network ties about highbrow culture) and cultural rendezvous (the frequency of shared visits to highbrow cultural performances and institutions).By measuring these manifestations empirically, we contribute by shedding new light on some of the potential mechanisms through which cultural tastes link to network quality and stability.
Our first contribution is that we address if cultural tastes, talk, and rendezvous yield different network returns.Having a personal taste for highbrow culture, for example by going alone to an opera or to an art museum, does not yield a network return because doing so does not involve a social manifestation of highbrow taste.On the other hand, talking about highbrow culture with network ties is more likely to yield a return because doing so communicates familiarity with high-status culture.Finally, participating in highbrow culture with network ties arguably yields the highest return because this activity not only communicates orally, but also demonstrates physically, a commitment to highbrow culture.In this way, individuals who "walk the walk" (cultural rendezvous) would be expected to have higher returns to highbrow tastes than individuals who only "talk the talk" (cultural talk).
Our second contribution is that we study the formation of new ties as well as the continuation of existing ties.We can do this because our data include two observations of the network (T1 = 2017 and T2 = 2019).We hypothesize respondents' highbrow tastes, and their social manifestations, to assist in creating a favorable first impression and thus in forming a new tie (i.e., a tie reported at T2 but not at T1).We propose two scenarios regarding continued ties.On the one hand, cultural tastes might not matter for maintaining an existing tie (i.e., a tie reported at T1 and T2) because other aspects than shared cultural tastes become more important over time (a scenario we label first impression).On the other hand, it might be that social manifestations of highbrow tastes continue to matter because shared cultural tastes help deepen the relationship (a scenario we label cultural bonding).We attempt to disentangle these scenarios in our empirical analysis.
Our third contribution is that we collect data on offline "core discussion networks" in a (largely) representative sample.We define the core discussion network as (non-kin) ties with whom individuals discussed important issues over the past six months (e.g., partners, friends, colleagues, group/club members, and/or neighbors; Marsden, 1987;McCallister and Fischer, 1978).Compared to existing research, most of which relies on student samples and/or online networks (e.g., Edelmann and Vaisey, 2014;Lewis et al., 2012;Selfhout et al., 2009), our data better capture the educational and occupational careers of individuals and the resources that their network ties possess.Moreover, core offline network relations are likely to be stronger and more easily mobilized than online network relations.
Our final contribution is that we address the different resources individuals may access through their network.Drawing on Bourdieu (1984Bourdieu ( , 1986)), we argue that social capital, i.e., "more or less institutionalized relationships of mutual acquaintance and recognition" (Bourdieu and Wacquant 1992, p.119), can be used to access other forms of capital.Individuals use (social manifestations of) highbrow tastes to attract and maintain high-quality networks that provide access to cultural and economic capital.We use ties' education as a proxy for cultural capital and propose that ties with high education, who tend to be very culturally active (Bennett et al., Fig. 1.Highbrow tastes and network ties. R. Meuleman and M.M. Jaeger 2009;Chan 2010), might be more likely to provide useful information, credentials, and legitimate culture and/or are more likely to know higher educated others who can.Similarly, we use ties' occupational status as a proxy for economic capital and propose that ties with high occupational status are more likely to (know others who) possess income and wealth and might be more likely to provide credit, property, and market access.In this way, individuals may convert highbrow tastes into social relations that might provide access to cultural and economic capital in different (but partly overlapping) subfields.
Our empirical analyses replicate the finding from existing research that cultural tastes (here measured by the frequency of participation in highbrow activities) are positively associated with network quality (ties' education and occupational status) (Erickson, 1996;Lizardo, 2013;Meuleman, 2021) and stability (Edelmann and Vaisey, 2014;Selfhout et al., 2009).Moreover, the frequency of highbrow talk, but not the frequency of highbrow rendezvous, is directly associated with network quality and mediates 20 to 50 percent of the positive association between highbrow tastes and network quality.We also find that highbrow tastes and talk are more strongly associated with new ties' education than with their occupational status; a finding suggesting that highbrow tastes signal (/provide access to) cultural rather than economic capital.Finally, in the analysis of continued ties we find that highbrow tastes and talk (but not rendezvous) are positively associated with continued ties' education, but largely unrelated with ties' occupational status.Moreover, the cultural bonding (rather than the first impression) scenario is more consistent with our results for ties' education.Overall, our results suggest that highbrow tastes operate in part via highbrow talk, link mostly to ties' education, and sustain relationships with highly educated ties.

Theoretical framework
This section presents our theoretical framework.We begin by conceptualizing highbrow tastes as signals of cultural and economic status.We then link highbrow tastes and their social manifestations to network quality and stability, arguing that they assist individuals in creating and maintaining high-quality networks.We illustrate our theoretical framework in Fig. 1.

Cultural tastes as status signals
We begin from the notion that high-SES parents transmit cultural tastes and behaviors to their children (Bourdieu, 1984).Since highbrow cultural tastes and behaviors are more common in high-SES families than in low-SES families, they act as signals of cultural and economic status (e.g., DiMaggio, 1987;Nagel et al., 2011).In addition to signaling status, highbrow tastes are a scarce or specialized resource from an ecological perspective (DiMaggio, 1987;Lewis and Kaufman, 2018;Lizardo, 2006).Therefore, in social interactions highbrow tastes are particularly suited to distinguish (high-status) individuals familiar with legitimate cultural codes, modes of conduct, and use of language from (low-status) individuals that lack knowledge and skills to reciprocate (DiMaggio, 1987).

Cultural tastes and network ties
We build on cultural reproduction theory (Bourdieu, 1984) to argue that individuals use cultural tastes and behaviors to infer about others' status, family background, and desirability as a network tie.This means that highbrow tastes act as signals of both cultural and economic status.Indeed, theory and empirical research has put forward that, in the absence of tangible information, people use diffuse signals about status, for example signals implied by cultural tastes, to infer about economic traits (Veblen, [1899] 1994; Weiss and Fershtman, 1998).Moreover, high-status individuals, for example those with high education or high occupational status, are argued to be more likely to form a tie with others if they perceive them to be of similar (or higher) status than if they perceive them to be of lower status (Bourdieu, 1984;Lamont and Lareau, 1988).These processes are self-reinforcing because highbrow tastes create group boundaries by self-selecting people with similar tastes into socially homophilous groups (Bourdieu, 1984;McPherson et al., 2001).In support of this idea, empirical studies show that individuals with similar cultural tastes are more likely to establish and maintain network ties (e.g., Edelmann and Vaisey, 2014;Lewis and Kaufman, 2018;Nagel et al., 2011;Selfhout et al., 2009).

Cultural tastes and network returns
Individuals extract resources from their network (Bourdieu, 1984;DiMaggio, 1987;Lamont and Lareau, 1988;Lewis and Kaufman, 2018;Lizardo, 2006).We conceptualize these resources via Bourdieu's (1984Bourdieu's ( , 1986) forms of capital.Network ties, which themselves comprise social capital, provide access to cultural (e.g., legitimate taste, information, and credentials) and economic (e.g., wealth, credit, and property) capital.We illustrate these two forms of capital in Fig. 1 in the box labeled New Tie (we address the box labeled Continued Tie below).In the figure, ties' education and occupational status indicates which type of capital they are most likely to possess (/provide access to).We argue that ties with high education are more likely to possess (/provide access to) to cultural capital, while ties with high occupational status are more likely to possess (/provide access to) economic capital.The reasoning behind this argument is that highly educated ties tend to possess more institutionalized, but also more embodied and objectified, cultural capital and tend to affiliate with others with high education and who are embedded in subfields (e.g., education, art, and media) in which institutionalized and objectified cultural capital is the main currency.Conversely, ties with high occupational status are more likely to possess wealth, credit, and property and tend to affiliate with others with similar status and who are embedded in subfields (e.g., finance, philanthropy, and industry) in which economic capital is the main currency.Thus, individuals may convert their highbrow tastes into social relations that might provide cultural and economic capital.

Social manifestations of cultural tastes
Some aspects of individuals' cultural tastes are visible without social interaction, for example via appearance, dress, and etiquette.However, for the most part highbrow tastes are embodied and need to be socially visible to be effective in enhancing network quality.Fig. 1 illustrates two social manifestations of highbrow tastes: cultural talk (discussing highbrow culture with network ties; Lizardo, 2006) and cultural rendezvous (attending highbrow performances and institutions with network ties; Daenekindt and Roose, 2013).In the following, we describe these social manifestations, including how they differ in terms of signaling value.
Cultural talk signals familiarity with highbrow culture by communicating knowledge and cultural capital to others through conversation (e.g., Collins, 1988;DiMaggio, 1987DiMaggio, , 2011;;Edelmann and Vaisey, 2014;Lizardo, 2006Lizardo, , 2016;;Friedman and Laurison, 2020).Consequently, cultural talk operates via shared interest in leisure activities and specialized styles of communication, accent, or speech.Thus, cultural talk "… gives strangers something to talk about and facilitates the sociable intercourse necessary for acquaintanceships to ripen into friendships" (DiMaggio 1987, p.443).In Fig. 1, we use the arrow B to capture the expected positive association between cultural talk and the quality of a new tie.
Cultural rendezvous implies a physical commitment to highbrow culture via shared attendance at highbrow performances and institutions.We argue that ties interpret cultural rendezvous ("walk the walk") as a stronger social manifestation of highbrow tastes than cultural talk ("talk the talk") because it is costlier (in terms of time and money), more difficult to engage in superficially, and more socially visible.In addition, while cultural talk involves social interaction between two individuals, cultural rendezvous provides a fellow audience with which to interact and a physical space dedicated to highbrow culture.Thus, one has to choose an appropriate cultural venue/performance, adhere to appropriate norms regarding dress and etiquette, and engage with the art and one's network tie (and potentially other visitors).This means that cultural rendezvous, being challenging and difficult to fake, provides a better opportunity than cultural talk for "showing off" familiarity with high-status culture, thus "… gaining prestige, having the opportunity for networking, being in the company of someone" (Daenekindt and Roose, 2013, p.313).In Fig. 1, we use the arrow C to capture the expected positive association between cultural rendezvous and the quality of a new tie.It follows from our theoretical framework that, in relative terms, we expect C to be higher than B.

Network stability
In this last part of our theoretical framework, we link highbrow tastes and their social manifestation to network stability.In Fig. 1, we include boxes to capture the network comprised of new ties (upper part of the figure) and the network comprised of continued ties (lower part of the figure).Research shows that similarity in cultural tastes is positively associated with the formation of new ties and with the continuation of existing ties (e.g., Edelmann and Vaisey, 2014;Nagel et al., 2011).For example, Lewis et al. (2012) find that students with similar tastes in music and movies are more likely to form and maintain friendships.Yet, it might be that highbrow tastes are only effective in making a good first impression and, over time, characteristics of the relationship itself (e.g., quality or conflict) become more important in maintaining the relationship (Lewis and Kaufman 2018).For example, Selfhout et al. (2009) find that similarity in music preferences links to friendship formation but not to friendship (dis-)continuation.
Building on these arguments, we propose two scenarios describing how highbrow tastes and their social manifestation might affect networks over time (we outline the empirical implications of each scenario below).We label the first scenario first impression.In this scenario, highbrow tastesand their social manifestation via cultural talk and rendezvousare more important in establishing a new tie than in maintaining an existing tie.Here, highbrow tastes are useful as signals when an individual's cultural and economic status are unknown at the start of a relationship; however, once a network tie gets to know the individual, other qualities of the relationship become more important.
We label the second scenario cultural bonding.In this scenario, highbrow tastes per se do not assist in maintaining a network tie, but their social manifestations do.That is, cultural talk and cultural rendezvous are as relevant in maintaining a network tie over time as they are in forming a new tie.In this scenario, social manifestations deepen the social and emotional rewards (e.g., affection, friendship, and shared cultural interests) associated with the tie.Consequently, the returns to maintaining a tie are expressive (i.e., intrinsic pleasure and social support) as much as instrumental (i.e., extrinsic and using social capital to access resources).

Empirical hypotheses
Based on our theoretical framework, we now present four empirical hypotheses pertaining to expected associations between, on the one hand, cultural tastes, talk, and rendezvous and, on the other hand, the quality of new and continued ties.As we use a descriptive research design, our hypotheses pertain to associations rather than to causal effects.We propose two hypotheses regarding new ties.
• H1: Highbrow tastes are positively associated with new ties' education and occupational status; i.e., we expect a positive association for arrow A in Fig. 1.We also expect the positive association to be stronger for ties' education than for ties' occupational status, capturing that highbrow tastes are stronger signals of cultural status than of economic status.• H2: Highbrow talk and rendezvous mediate some of the positive association between highbrow tastes and new ties' education and occupational status; i.e., we expect associations B and C to account for some of the association A.Moreover, we expect C to be higher than B in relative terms since highbrow rendezvous is a stronger social manifestation of highbrow tastes than highbrow talk.
We propose two hypotheses regarding continued ties.
• H3: First impression.Highbrow tastes, talk, and rendezvous are more strongly associated with new ties' education and occupational status than with continued ties' education and occupational status.In this scenario, we expect associations B and C (capturing the first impression) to be stronger than associations D and E (capturing continued cultural interactions).• H4: Cultural bonding.The positive associations for highbrow tastes, talk, and rendezvous are the same for new and continued ties' education and occupational status.In this scenario, we expect associations D and E (capturing continued cultural interactions) to be the same as associations B and C (capturing the first impression).

Data
We analyze data from the Family Survey Dutch Population (FSDP) 2017 and 2019 (see dataarchive.lissdata.nl).This survey, which addresses the life course and living situation of Dutch speaking citizens in the Netherlands between the ages of 18 and 70, has been ongoing since 1992.In 2017, we drew a representative sample from the Longitudinal Internet Studies for the Social Sciences panel (LISS), CentERdata.This panel is representative of adult Dutch individuals who participate in monthly web surveys.It is a true probability sample of households drawn from the population register and consists of approximately 4500 households and 7000 individuals.Evidence on panel attrition within the LISS shows that attrition is higher amongst older (rather than younger) respondents, higher among those whom LISS did not provide internet or a PC, and higher among households without disabled persons (see lissdata.nl).Panel attrition does not vary by other socio-demographic characteristics, for example sex, size of household, education, employment status, immigrant status, and urban vs. rural residence.Moreover, refreshment samples ensure that the LISS panel remains nationally representative over time.In our dataset, 70% (N = 2128) of respondents sampled from the first wave (N = 3051) were successfully re-interviewed in the second wave.Additional analyses using the original sample (i.e., respondents who participated at T1) yield the same substantive results as the results we present below (which uses the sample with respondents who participated in both T1 and T2; results available upon request).
For the FSDP 2017, we randomly selected one respondent per household (net sample size: 4447).In total, 3099 respondents participated in the fall of 2017 (net response rate: 70%) and were asked about their cultural tastes and social network. 1 In the fall of 2019, we reinterviewed respondents in the FSDP 2017 specifically for this study and repeated the questions on cultural tastes and social networks (net sample size: 3051).In total, 2128 respondents participated in the FSDP 2017 (net response rate: 70%).Consequently, the FSDP data consist of two waves: T1 (2017) and T2 (2019).

Variables
Existing research shows that the influence from cultural tastes to network characteristics ("culture conversion model") is stronger than the reverse ("traditional network model"; Lizardo, 2006;Meuleman, 2021).Moreover, cultural similarity between individuals affects the likelihood of forming friendships later on (e.g., Edelmann and Vaisey, 2014;Selfhout et al., 2009).Building on these studies, and in line with the causal pathways we outline in the theoretical framework, we measure the dependent variables (i.e., ties' education and occupational status) at T2 and all individual and tie characteristics at T1.We measure characteristics of new ties at T2, because, logically, we have no information on these ties at T1. Finally, we include variables pertaining to ties (level 1) and respondents (level 2).Table 1 shows descriptive statistics for all variables in the analysis.

New and continued ties
The FSDP (2017 and 2019) uses a name generator (McCallister and Fischer, 1978) to identify respondents' network ties.This measurement instrument, standard in research on social networks (e.g., Campbell and Lee, 1991; Lin and Erickson, 2008; Marsden,   1 To reduce total (web) questionnaire length, we asked respondents to report on their networks in a separate questionnaire (also in the fall of 2017) (N = 3,038, response rate: 68%).R. Meuleman and M.M. Jaeger 1987;McPherson et al., 2006), uses self-reports to identify the core discussion network.In the name generator, respondents list the first names of the (maximum five) people with whom they discussed important issues over the past six months.In our sample, we excluded 180 respondents (8.5%) who did not name any (codable) ties in 2017 and 2019.The remaining respondents mentioned on average about three ties.We assigned a unique identification number to each tie.By comparing ties' identification numbers in 2017 and 2019, we distinguish new ties (only mentioned in 2019) and continued ties (mentioned both in 2017 and 2019).
Respondents in the FSDP provided information on several characteristics of their ties, such as sex, age, migration background, and their relationship to the respondent.Based on this information, we exclude kinship ties (i.e., parents, children, and other family members) to avoid, to the extent possible, that our results pick up intergenerational transmission within families.Moreover, we exclude continued ties whose reported characteristics in 2017 (in terms of kin relationship, sex, and immigration background) were inconsistent with the characteristics reported in 2019.

Ties' education and occupational status
The FSDP includes information on new and continued ties' education and occupational status.We measure ties' education via years of completed schooling.In the survey, respondents provided information on each tie's highest educational attainment, ranging from 1 "not (yet) completed any education" to 7 "university."We recoded these educational levels into the minimum years of schooling required to graduate in each educational track, ranging from six to 18 years. 2 We measure ties' occupational status via the International Socio-Economic Index 2008 (ISEI).Respondents provided information on each tie's occupation using a list of categories from 1 "higher intellectual or independent profession" to 9 "agrarian profession."We recode occupational categories into ISEI scores, ranging from 16 to 90 (Ganzeboom, 2010), and average ISEI scores for self-employed and salaried professions within occupational categories.
We exclude respondents (and ties) with missing values on education or occupational status.The final sample on new ties' education includes 2005 ties reported by 1135 respondents, while the sample for new ties' occupational status includes 1363 ties reported by 850 respondents.The final sample on continued ties' education includes 1307 ties reported by 838 ties, while the sample for continued ties' occupational status includes 876 ties reported by 603 respondents.

Respondents' highbrow tastes
The FSDP asked respondents how often they visit classical concerts, opera, and ballet performances; serious theatre performances (drama, dance); and art museums (visual arts, exhibitions).Response categories range from 0 "never" to 3 "more than 6 times a year."We calculated the mean of these items to measure respondents' highbrow tastes (Cronbach's alpha is 0.71).

Social manifestations of highbrow tastes
We added questions to the name generator in the FSDP to measure social interactions with network ties.First, we asked respondents how often they talked with each tie about classical concerts, opera, and ballet performances; serious theatre performances (drama, dance), and art museums (visual arts, exhibitions).The response categories range from 1 "never" to 7 "almost every day."We calculate a mean score for these items for each tie to capture cultural talk (Cronbach's alpha ranging from 0.80 to 0.84 per tie).
Second, we asked respondents how often, together with each named tie, they visited classical concerts, opera, and ballet performances; serious theatre performances (drama, dance); and art museums (visual arts, exhibitions).The response categories range from 0 "never" to 3 "more than 6 times a year."We calculate a mean score for these items for each tie to capture cultural rendezvous (Cronbach's alpha ranging from 0.59 to 0.67 per tie).We measure cultural talk and cultural rendezvous at T1 for continued ties and T2 for new ties.

Control variables for ties
We control for ties' sex, age, migrant background, and the frequency with which respondents talk with each tie.We include indicators of the tie's sex (0 "male" and 1 "female"), age (categorical variable with 13 categories recoded into midpoints), and migrant background (0 "Dutch" and 1 "migrant background").Moreover, we include an indicator of how often the respondent talks with each tie in general (labeled general talk), with response categories 1 "less than once per month," 2 "at least once a month," 3 "at least once a week" and 4 "almost every day."We include this variable to control for the overall frequency with which respondents communicate with ties.We measure each control variable at T1 for continued ties and T2 for new ties.Finally, we impute missing values for the control variables pertaining to network ties using all relevant information in the dataset except the dependent variables (maximum 3% per variable).

Research design
Our objectives with regard to new ties are to estimate the empirical association between highbrow tastes and the quality of new ties (H1) and the extent to which highbrow talk and highbrow rendezvous mediate this association (H2).Our objectives with regard to continued ties are to compare the first impression (H3) and the cultural bonding (H4) scenarios, each claiming different associations between highbrow talk/rendezvous and the quality of new and continued ties.
We use multilevel linear regression models because our dependent variables (ties' education and occupational status) are metric and our data include observations of ties nested within individuals.In our baseline model specification (labeled M1 in Table 2 below), we model new/continued ties' education/occupational status as a function of respondents' highbrow tastes and the socio-demographic controls.This model addresses arrow A in Fig. 1 (and H1).We then add highbrow talk and rendezvous as explanatory variables, both individually (M2, M3) and jointly (M4), and analyze the extent which they mediate the association between highbrow tastes and new/ continued ties' quality (i.e., the change in A attributable to B and C, cf.H2).
In the analysis of continued ties, we compare regression coefficients across models for new and continued ties to address the first impression (H3) and the cultural bonding (H4) scenario.To address the first impression scenario, we test if the regression coefficients on the associations between highbrow tastes/talk/rendezvous and new ties' quality (i.e., B and C in Fig. 1) are stronger than the same associations for continued ties' quality (i.e., D and E).In the cultural bonding scenario, we expect these associations to be the same for new and continued ties' quality (i.e., B = D and C = E). 3 As our data include clustering (network ties nested within respondents), we use multilevel models with random effects to account for clustering.We standardize all metric variables to have a mean of zero and a standard deviation of one to facilitate interpretation.Finally, as our analysis is descriptive rather than causal, we refer to associations rather than to causal effects.

Results
We now present the main empirical results.We begin by analyzing the baseline association between highbrow tastes and new ties' education and occupational status.These analyses enable us to address H1 and H2.We then analyze associations with existing ties' education and occupational status, thus addressing H3 and H4.Table 2 summarizes main results.(Please see Appendix A.1 and A.2 for additional information).

New ties
First, we analyze the association between respondents' highbrow tastes and new ties' education.Net of socio-demographic controls, M1a in Table 2 shows that highbrow tastes at T1 are significantly and positively associated with new ties' education at T2 (B = 0.19, p < 0.001).This result replicates results from existing research (e.g., Meuleman, 2021).M2a shows that the frequency of highbrow talk with new ties is also significantly and positively associated with ties' education (B = 0.10, p < 0.001).That is, in addition to having highbrow tastes, the extent to which individuals manifest these tastes via conversations with ties also correlates with ties' education.Moreover, highbrow talk mediates around one-fifth ((0.19-0.15)/0.19*100= 21.1%) of the association between respondents' highbrow tastes and new ties' education, as evaluated by the reduction in the regression coefficient on highbrow tastes when comparing M1a and M2a. 4 Similar to highbrow talk, the frequency of highbrow rendezvous (i.e., "walk the walk") is significantly and 3 We use the proposed research design because it is straightforward and results are easy to interpret.However, the analytical sample we rely on for continued ties is a selective one because each continued tie began as a new tie.If selection into a continued tie depends on unobserved (including cultural) characteristics that also affect tie quality, our results will be biased.In addition to the results presented above, we have estimated Heckman selection models to assess the impact of sample selection on our results.In the first step, we estimated the extent to which individuals with stronger highbrow tastes (T1) were more likely to form a new high-quality tie (T2).In the second step, we corrected for sample selection (i.e., individuals with highbrow tastes self-selecting into the sample of continued ties) by including the inverse Mills ratio as an explanatory variable when estimating regression models for continued ties' quality.Substantive results from these selection models were very similar to the ones we present in the main analyses (Table 2).We also analyzed if including highbrow tastes at T2 (rather than at T1) as an explanatory variable, thus adjusting for change over time in respondents' highbrow tastes, affected our results.It did not. 4We present results from a formal mediation analysis in Appendix A.3.Here, we show that highbrow talk is positively associated with highbrow tastes (and with highbrow rendezvous), which is a necessary condition for it to be a mediator of highbrow tastes.To save space, we present the main results from the mediation analysis in the text and relegate details to Appendix A.3.We use the same mediation approach in the analyses of highbrow rendezvous, ties' occupational status, and continued ties presented below.We note that as our models also include the indicator of the frequency with which respondents talk with the tie in general, we control directly for loquacity.
positively associated with new ties' education (M3a) and mediates around ten percent (10.5%) of the association between highbrow tastes and new ties' education.Finally, when we include highbrow tastes, talk, and rendezvous in the same model (M4a), highbrow tastes and highbrow talk remain significant while highbrow rendezvous does not.Together highbrow talk and rendezvous mediate around one-fifth (21.1%) of the baseline association between highbrow tastes and new ties' education.Second, we analyze the association between highbrow tastes and new ties' occupational status (Table 2, M1-4b).Net of sociodemographic controls, highbrow tastes at T1 are significantly and positively associated with new ties' occupational status at T2 (B = 0.09, p < 0.01, M1b).Moreover, highbrow talk is significantly and positively associated with new ties' occupational status (B = 0.11, p < 0.001, M2b) and mediates 44.4 percent of the association between highbrow tastes and ties' occupational status (and renders highbrow tastes insignificant).Although the coefficient is in the expected direction, highbrow rendezvous is not statistically significantly associated with new ties' occupational status (B = 0.05, p > 0.05; M3b).Finally, when we include highbrow tastes, talk, and rendezvous in the same model (M4b), highbrow talk is the only indicator that is statistically significant.These results suggest that the social manifestation of highbrow tastes, in the form of highbrow talk, but not highbrow tastes themselves, is associated with new ties' occupational status.
Overall, our results support H1 stating that highbrow tastes are positively associated with new ties' quality.Yet, our results differ for new ties' education and occupational status.The regression coefficient on highbrow tastes is higher in the models for new ties' education than in the models for occupational status, both in the baseline (M1a/M1b; 0.19 vs. 0.09, p < 0.001) and final (M4a/M4b; 0.15 vs. 0.05, p < 0.001) model specification.Moreover, we find that the association between highbrow tastes and new ties' education (but not new ties' occupational status) remains statistically significant after we control for social manifestations of highbrow tastes.Together, these results suggest that highbrow tastes are not equally important across different dimensions of ties' SES and, consistent with H1, they appear to be stronger signals of (/access to) cultural than of economic status (/resources).One interpretation of the positive association net of social manifestations (M4a) is that highbrow tastes capture signals not communicated via social interactions (e.g., appearance, dress, and etiquette) that ties with high education (but not those with high occupational status) understand and appreciate.
Our results also support H2 stating that highbrow talk and rendezvous mediate some of the association between highbrow tastes and new ties' quality.Again, results differ for ties' education and occupational status.While highbrow talk mediates some of the association for new ties' education (while highbrow rendezvous does not), most of the direct association persists and remains statistically significant.By contrast, highbrow talk mediates a much larger share (44.4%) of the baseline association in the model for new ties' occupational status (M4b) and renders highbrow tastes insignificant.These findings suggest that highbrow tastes per se are not enough when attracting new ties with high occupational status: individuals must actively signal these tastes via highbrow talk.

Continued ties
Having established that social manifestations of highbrow tastes, in the form of highbrow talk, are associated with new ties' education (and, to some extent, occupational status), we now analyze their association with continued ties' quality.We first present baseline associations and then assess the two competing (first impression/cultural bonding) scenarios.The lower part of Table 2 summarizes results from multilevel linear regressions of continued ties' education and occupational status (M1-4c and M1-4d).
In M1c, we estimate the association between highbrow tastes and continued ties' education.Here, we find a significant, positive association between highbrow tastes at T1 and continued ties' education at T2 (B = 0.22, p < 0.001).Moreover, based on estimates from M2c we calculate that highbrow talk at T1 (B = 0.12, p < 0.001) mediates around a quarter (27.3%) of this baseline association.Similarly, highbrow rendezvous at T1 is significantly and positively associated with ties' education at T2 (B = 0.10, p < 0.01, M3c) and mediates around one-fifth (22.7) of the association between highbrow tastes and continued ties' education.When we include highbrow tastes, talk, and rendezvous in the same model (M4c), highbrow tastes and highbrow talk remain positively associated with continued ties' education, while highbrow rendezvous is no longer significant.These results, which are similar to our results for new ties, indicate that highbrow talk is the only social manifestation of highbrow tastes that mediates part of the association between highbrow tastes and continued ties' education.
Next, we analyze the association between highbrow tastes and continued ties' occupational status.M1d shows that highbrow tastes at T1 are significantly and positively associated with continued ties' occupational status at T2 (B = 0.11, p < 0.01).However, and unlike our results for continued ties' education, neither highbrow talk nor highbrow rendezvous at T1 are associated with continued ties' occupational status (cf.M2b and M3b).Similar to our results for new ties' quality, these results suggest that highbrow tastes link more strongly to continued ties' education than to their occupational status.We now proceed to addressing the first impression and cultural bonding scenarios, as outlined in H3 and H4.

First impression or cultural bonding?
Our first impression hypothesis (H3) states that highbrow talk and highbrow rendezvous are more strongly associated with new ties' quality than with continued ties' quality.The motivation behind this hypothesis is that a good first impression might wane over time as other aspects of a relationship become more important.Alternatively, the cultural bonding hypothesis (H4) states that highbrow talk and highbrow rendezvous are equally important for new and continued ties' quality.The motivation behind this hypothesis is that social manifestations of highbrow tastes continue to matter as the relationship deepens.To distinguish each scenario empirically, we test if the regression coefficients on highbrow talk and highbrow rendezvous are the same in the models for new (Table 2, M1-4a/b) and continued (Table 2, M1-4c/d) ties' education and occupational status.
We begin with highbrow talk and ties' education.The regression coefficient on highbrow talk in the full model is 0.09 for new ties (p < 0.001; cf.M4a and arrow B in Figs. 1) and 0.10 for continued ties (p < 0.01; cf.M4c and arrow D), i.e., practically identical and not statistically significantly different (z-stat = − 0.073, p > 0.05).Similarly, the regression coefficients on highbrow rendezvous (B = 0.01 and B = 0.05, both p > 0.05), cf.Arrows C and E, are not different from each other (and from zero; z-stat = − 1.61, p > 0.05).These results support the cultural bonding hypothesis (H4) in that the positive association between cultural talk and ties' education is the same for new and continued ties; i.e., cultural talk, but not cultural rendezvous, continues to matter over time.
The regression coefficients on highbrow talk in the regressions of ties' occupational status are 0.11 (p < 0.01; M4b) for new ties and 0.04 (p > 0.05; M4d) for continued ties.These coefficients are not significantly different from each other (z-stat = 1.86, p = 0.06).The regression coefficients on highbrow rendezvous (B = − 0.01 and B = 0.02, both p > 0.05) are not statistically different from zero, nor from each other (z-stat = − 0.76, p > 0.05).To some extent, these results resemble our results for ties' education in that associations for social manifestations of highbrow tastes are similar for new and continued ties.However, far fewer coefficients are statistically significant, suggesting that highbrow tastes, and their social manifestations, are not particularly relevant for attracting and maintaining ties with high occupational status.From a theoretical perspective, these results indicate that, as a source of social capital, highbrow tastes mainly signal (/provide access to) cultural (rather than economic) capital.We return to these points in the final discussion.

Conclusion and discussion
How do cultural tastes shape networks?In this paper, we collected panel data on offline networks in a representative sample to analyze the extent to which social manifestations of highbrow tastes, in this case cultural talk and cultural rendezvous, mediate the positive association between highbrow tastes and network quality.Our theoretical argument is straightforward: individuals must not only possess, but must also manifest, their highbrow tastes to attract and maintain high-quality ties.They can do this via conversations about highbrow culture with network ties ("talk the talk") or via shared visits to highbrow cultural performances and institutions ("walk the walk").
Our empirical analysis shows that highbrow tastes are positively associated with ties' quality (as captured by their education and occupational status) and with the stability of the network over a two-year period.We interpret this result as an indication that ties use highbrow tastes to infer about an individual's status and desirability as a network tie.Our results are particularly clear for ties' education, which suggests that highly educated ties are more responsive to individuals' highbrow tastes than are ties with high occupational status.We interpret this difference as evidence that highbrow tastes are better signals of cultural than of economic status and are more appreciated (and potentially rewarded) by ties with similar tastes.Moreover, social manifestations of highbrow tastes, in particular via highbrow talk, are directly associated with network quality and stability (especially ties' education) and mediate some of the positive association between highbrow tastes and network quality/stability.This result supports our theoretical argument that individuals must manifest their highbrow tastes to benefit from them, and adds to previous debates on the importance of communicating cultural tastes in conversation (Collins, 1988;DiMaggio, 1987DiMaggio, , 2011;;Edelmann and Vaisey, 2014;Lizardo, 2006Lizardo, , 2016;;Friedman and Laurison, 2020).
Contrary to theoretical expectation, we find that highbrow rendezvous does not appear to be a stronger social manifestation of highbrow taste than cultural talk.Although unexpected, this finding makes sense in light of evidence that communicating cultural tastes via conversation is a particularly strong signal of social position and smoothens social interactions (Collins, 1988;DiMaggio, 1987DiMaggio, , 2011;;Edelmann and Vaisey, 2014;Lizardo, 2006Lizardo, , 2016;;Friedman and Laurison, 2020).Moreover, cultural talk is more flexible and less costly than cultural rendezvous (both for individuals and ties) since it is easy to initiate (e.g., in-person or via phone or social media), is not restricted physically by travel time to and from cultural performances/institutions, and need not cost money.Consequently, it might be that cultural talk is the better strategy when seeking to convert highbrow tastes into high-status networks, while cultural rendezvous is better suited for pleasurable cultural interaction.We welcome future research that addresses this question.
Finally, we argue that highbrow tastes might either serve to make a good first impression (but do not matter in the long run) or, alternatively, they create cultural bonding that sustains relationships over time.Our empirical results support the cultural bonding scenario to some extent: the positive associations between highbrow tastes/cultural talk and ties' education is the same for new and continued ties.By contrast, we find no clear associations for new and continued ties' occupational status.
How do our results speak to broader debates on cultural tastes and social networks?Building on Bourdieu (1984Bourdieu ( , 1986)), we provide new evidence on the convertibility of different forms of capital into each other.Our results suggest that while individuals can convert highbrow tastes into social capital, the associated exchange rate of this social capital is higher for ties' education (and thus for access to cultural capital) than for ties' occupational status (and thus for access to economic capital).To us, these differences highlight that the exchange rate between different types of capital varies across subfields.Although this argument is not new (1984,1986), our study is one of the few to demonstrate empirically the value of cultural capital in generating social capital.Moreover, and unlike existing research (Edelmann and Vaisey, 2014;Lewis et al., 2012;Selfhout et al., 2009), we analyze (largely) representative data, offline networks, and address new and continued ties.Finally, our study is one of the first in this field to use a "name generator" to identify the core discussion network.Reassuringly, our findings using this approach are in line with existing studies that use (measurement instruments similar to) a "position generator" to capture social networks (e.g.Erickson, 1996;Lizardo, 2013;Meuleman, 2021).
We end the paper by highlighting several limitations in our data and analysis.First, although we have two-year panel data there is too little within-individual variation in cultural tastes, talk, and rendezvous to estimate fixed effects models.Nevertheless, we do demonstrate that individuals' earlier cultural tastes are positively associated with the education and occupational status of new ties formed two years later.These findings are in line with previous studies showing that individuals with similar cultural tastes are more likely to establish (and maintain) network ties (e.g., Edelmann and Vaisey, 2014;Lewis and Kaufman, 2018;Nagel et al., 2011;Selfhout et al., 2009).They are also in line with research showing that the relationship between cultural tastes and social networks is affected more by selection processes ("culture conversion model") rather than influence or diffusion processes ("traditional network model"; Lizardo, 2006;Meuleman, 2021;Lewis et al., 2012).
Second, while we have detailed information on cultural talk and cultural rendezvous for up to five network members, we do not measure what is actually communicated or demonstrated, nor how this is received and valued by network ties.We only examine what people consume, but not how they display or appropriate culture in different ways (Jarness, 2015).For instance, one might commit a faux pas if one is unaware of, or misinterprets, certain cultural codes when engaging in highbrow cultural talk.It would be interesting to delve deeper into how cultural talk and cultural rendezvous operates "in action."Moreover, it would be fascinating to analyze if individuals from privileged socioeconomic backgrounds, who arguably master sophisticated cultural distinctions, are better at converting social manifestations of cultural tastes into high-quality networks than individuals from less privileged backgrounds.
Third, our results suggest that highbrow tastes and their social manifestations promote access to network ties with higher status positions.Yet, we do not know which specific resources individuals actually access through their high-quality networks, nor if these resources are mobilized or converted into other forms of capital in specific subfields.For example, do highbrow tastes and their social manifestations facilitate ties that may help individuals find a new job, home, educational opportunities for their child, or venture capital?We encourage future research that addresses this important issue.
Finally, the way in which we measure cultural tastes is restrictive and does not take into account that non-highbrow tastes, for example omnivorous or voracious tastes (Erickson, 1996;Katz-Gerro and Jaeger, 2013), have higher status than highbrow tastes in some subfields.In addition, we do not measure the ways in which individuals present their cultural tastes (e.g., via appearance, dress, or etiquette) in social interactions.While beyond the scope of this study, we would welcome more refined measures of cultural taste, in particular by including alternative dimensions of cultural tastes.Despite these limitations, our study is one of the first to collect panel data on social manifestations of cultural tastes, including how they relate to network quality and stability.We welcome future studies that shed light on social expressions of cultural tastes and their consequences.Source: FSDP 2017/2019; ***p < 0.001, **p < 0.01, *p < 0.05 (two-tailed).Note: All independent variables measured at T1, only at T2 for new tie characteristics.All interval/ratio variables were standardized to a mean of zero and standard deviation of one.
All variables measured at T1 with the exception of the dependent variables and the new tie characteristics.Variables before standardization.

Table 2
Results from multilevel linear regressions of new and continued ties' education and occupational status (T2).