The reluctant preference: communities of enthusiasts and the diffusion of atypical innovation

Abstract

Exploring the initial diffusion of innovation, this article investigates how early adopters reach a tipping point with a shared and emphatic preference for atypical products. In a community of enthusiasts where members mutually observes each other, we show that potential buyers are reluctant to express a preference for highly atypical products despite their vanguard positive attitude toward atypicality. We argue that such reluctance is driven by a membership-validating concern: while favoring highly atypical products, potential buyers still need to avoid atypical but low-quality alternatives that would undermine their sense of membership to the vanguard group. Consistently, we hypothesize that the endorsement granted by other community members to a basket of atypical products alleviates—and eventually removes—potential buyers’ reluctance. Between equally endorsed alternatives, potential buyers will then include the most atypical one in their displayed preference, thereby revealing their vanguard attitude and validating their membership to the in-group. We situate our analysis in a community of enthusiasts for electronic music recordings and find robust support to our hypotheses. By raising the bar of how a product must be to be distinctive, membership-seeking enthusiasts raise the threshold of atypicality. In so doing, they also expand the range of products deemed typical by the general consumer. These twined processes fuel the ongoing diffusion of innovation.

1. Introduction

In 1967, The Velvet Underground released one of the albums that revolutionized rock and alternative music. With the iconic banana cover signed by Andy Warhol, The Velvet Underground & Nico synthesized influences from experimental works, introducing tailor-made instruments, new sound-processing techniques, and dark and provocative lyrics (Bockris, 1995; Heylin, 2009). In the context of 1967s Summer of Love, however, the album’s maverick sound was completely ignored by the public, resulting in a terrible commercial flop. It took the album 10 years to receive critical and commercial appraisal. Ranking 13th on The 500 Greatest Albums of All Times by Rolling Stone Magazine, The Velvet Underground & Nico is now considered “the most prophetic rock album ever made” (Levy and Van Zandt, 2006).

The Velvet Underground’s story is illustrative of a recurring phenomenon in the history of innovation: groundbreaking ideas, often initially rejected by the audience “because they are perceived as weird, inappropriate, unworkable, or too risky” (Mainemelis, 2010), may eventually become not only acclaimed but unanimously consecrated as enlightened and revolutionary (Staw, 1995; Allen and Lincoln, 2004; Allen and Parsons, 2006).

Offering theoretical groundwork for this phenomenon, organizational scholars have focused on the role of vanguard consumers—in particular early adopters (Rogers, 1962; Katz et al., 1963)—as the driving force behind the initial diffusion of innovation, eventually underpinning subsequent general acclaim. Vanguard consumers (Koçak et al., 2014) are individuals with a leading interest in products that and producers who introduce innovative features, often combined in atypical ways (Rosenkopf and Nerkar, 2001; Katila and Ahuja, 2002; Goldberg et al., 2016; Lopez-Vega et al., 2016). Through their vanguard consumption, they fuel the formation of a critical mass whose presence is vital for innovative products to spread in the market (Rogers, 1962; Eger, 2002). However, the mechanism by which vanguard early adopters themselves converge to a “tipping point” (Gladwell, 2006) with a cohesive preference for innovative products remains largely unexplored.

To shed some preliminary light on this issue, we focus on the reception of innovation within a community of enthusiasts (Fiol and Romanelli, 2012; Koçak et al., 2014). Communities of enthusiasts bring together passionate individuals with a common desire to advance their topic of interest and who invest their identity “in the collective identity of a market” (Koçak et al., 2014: 777). We situate our analysis in a community of enthusiasts for electronic music, whose musical innovation proceeds through the recombination of style categories (Lena and Peterson, 2008; Askin and Mauskapf, 2017) and aims at producing atypical sounds—blends of styles that are not only diverse but also distant from one another in the sociocognitive maps used by participants to classify the field (Goldberg et al., 2016).

The choice of a creative setting for empirical investigation is not fortuitous. In most creative and cultural contexts, consumers look for surprising and novel sensations (Lampel et al., 2000; Zuckerman, 2007) but nonetheless devalue highly atypical items despite their greater creative potential (Hsu, 2006; Negro et al., 2010a). Although certain consumers’ traits might influence their receptivity to the atypical (e.g. in the case of omnivorous tastes, Bourdieu, 1984; Peterson and Kern, 1996), a generally ambivalent reaction to atypicality has been empirically documented in markets for theatrical productions (Uzzi and Spiro, 2005), industrial design goods (Hekkert et al., 2003), and popular music (Askin and Mauskapf, 2017). Across creative and cultural fields, audience members tend to reward moderately atypical products but assign penalties to those items in which the atypicality is particularly marked (see also, Uzzi et al., 2013; Leung, 2014).

In addressing this apparent contradiction, we depart from extant research in two ways. First, whereas a producer-side perspective pays attention to the qualities of the products to explain the ambivalent reception to atypicality (Hekkert et al., 2003; Uzzi et al., 2013; Kovács and Johnson, 2014; Askin and Mauskapf, 2017), we embrace an audience-side perspective and focus on the qualities of individual membership in a community of vanguard enthusiasts. Second, instead of focusing on actual consumption choices, we look at the preferences that potential buyers of innovative products make visible to other enthusiasts within the community. In this way, we propose a framework that relates the development of preferences to the context where the public display of such preferences operates as a mechanism of validation of in-group membership (Tajfel, 1974; Turner et al., 1987).

We advance that, in communities of vanguard enthusiasts, participants would be generally more likely to display a clear preference for atypical products that would validate their membership in the vanguard group (Koçak et al., 2014). However, confronted with highly atypical products imbued with great uncertainty (Caves, 2000), community participants need to avoid those highly atypical products that, instead of being ingeniously innovative, might merely reflect the dilettantism of their producers. Displaying a preference for such low-quality products, despite their attractive atypicality, would threaten the validation of individuals’ membership in the vanguard community. This trade-off determines a reluctant preference: a displayed preference that incorporates the reluctance of validation-seeking members to fully express their ideal attitude. In the context of a community of enthusiasts, such reluctance leads highly atypical products to be excluded from the displayed preferences of potential buyers.

To reinforce our argument, we also propose that community members’ reluctance is reduced–and eventually removed–when the perceived uncertainty about the quality of highly atypical products undergoes a change. Drawing on the notion of social influence (Burnkrant and Cousineau, 1975; Burt, 1987; Van Herpen et al., 2009), we argue that the observation of other community members’ previous choices (what we call community endorsement) relaxes validation-threatening uncertainty and makes potential buyers formalize their preferences according to their ideal attitude. Among equally endorsed alternatives, potential buyers will then include in their displayed preferences those products that have the highest level of atypicality—thereby validating their membership in the vanguard group. In this sense, community endorsement does not influence others’ choices directly (Burnkrant and Cousineau, 1975; Salganik et al., 2006; Salganik and Watts, 2008) but instead alters the conditions under which their attitude can be safely expressed through visible preferences.

We give preliminary groundwork for our hypotheses by drawing insights from a qualitative exploration that involves interviews, a questionnaire, and the analysis of conversations on the online forum of Discogs (discogs.com), a community of enthusiasts and a marketplace for electronic music (Montauti and Wezel, 2016). We then proceed to test our hypotheses on a sample of 25,518 electronic music recordings and find robust support.

Our results have major implications for understanding the mechanism behind the initial part of the S-shaped curve of the diffusion of innovation (Rogers, 1962) and the formation of tastes (Hennion, 2001). First, our study introduces an audience-side explicatory mechanism behind the formation of a tipping point of cohesive preferences within communities of experts. This mechanism, operating at the level of group preferences between subgroups of the same community, deepens our understanding of the curvilinear relationship of atypicality-appeal found in previous research on cultural consumption (Hekkert et al., 2003; Uzzi and Spiro, 2005; Askin and Mauskapf, 2017). Second, our study comments on the centrality of communities of enthusiasts in sustaining the diffusion of innovation in markets, also beyond the context of creative and cultural industries (Dahlander and Frederiksen, 2012). The adoption of increasingly atypical innovation within the boundaries of a community of enthusiasts can indeed reflect on the broader society, as community members often act as tastemakers for general consumers and sustain the formation of a consensus in the marketplace (Koçak et al., 2014). Finally, our results contribute to understanding how first-mover pioneers, sometimes initially opposed even by vanguard audiences, ultimately contribute to the advancement of their field by increasing the threshold of acceptability of innovation within communities of experts. In this process, first movers may eventually see their initial lack of recognition turn into unanimous consecration (Allen and Lincoln, 2004; Allen and Parsons, 2006), as The Velvet Underground experienced.

2. Theory and hypotheses

Since the acknowledgment that markets are imbued with uncertainty (Akerlof, 1970; Tversky and Kahneman, 1974), scholarly research has been engaged in unveiling how people make sense of and react to items that present an unusual set of characteristics—quintessentially, innovation. A prominent perspective on this issue is categorization theory (Hannan et al., 2007, 2019; Hannan, 2010). It argues that, especially under uncertainty, people evaluate and choose among alternatives by leveraging the “cultural infrastructure” (Vergne and Wry, 2014: 59) offered by social categories (e.g. Smith and Medin, 1981).

One of the most robust findings of categorization theory is that products and organizations that combine multiple and distant categories—and are therefore atypical to each of them (Kovács and Johnson, 2014; Kovács and Hannan, 2015; Goldberg et al., 2016)—receive lower attention or are poorly evaluated. Atypicality poses a challenge to individuals’ interpretive schemata (Giddens, 1979; Ranson et al., 1980; Uzzi and Spiro, 2005; Uzzi et al., 2013) and dilutes legitimacy by diverting the attention of specialized audiences (Zuckerman, 1999), communicating poor quality (Zuckerman et al., 2003; Phillips et al., 2013; Kovács and Johnson, 2014), and threatening the audience’s comprehension (Zuckerman, 2004; Kovács and Hannan, 2010; Negro et al., 2010b).

In creative production, however, atypicality is often a rule rather than an exception. Drawing from existing codified material and recombining it in unusual ways, creative producers pry and extend the boundaries of innovation (Fleming, 2001; Rosenkopf and Nerkar, 2001; Katila and Ahuja, 2002; Lopez-Vega et al., 2016). Consumers of creative material are therefore exposed to widespread category spanning and less likely to assign penalties to atypical products (Kovács and Hannan, 2010; Negro et al., 2010a). Depending also on their heterogeneity (Cattani et al., 2014; Goldberg et al., 2016) or interest in the artistic content (van Venrooij and Schmutz, 2018), some consumers might indeed hold a positive attitude toward atypical items that promise higher satisfaction of their curiosity (Lampel et al., 2000; Zuckerman, 2007).

This is what happens in communities of enthusiasts (Kovács and Hannan, 2010; Fiol and Romanelli, 2012; Koçak et al., 2014), where vanguard participants—differently from specialized experts (Zuckerman, 1999) or peer practitioners (Cattani et al., 2014)—hold a common interest in the innovative practices of their field (Freiberger and Swaine, 1999; Fiol and Romanelli, 2012). In such communities, members openly share not only their consumption choices but also their preferences, in the form of wish lists, rankings, and/or active participation in topic-specific conversations. The alignment between individual preferences and the standards of evaluation (Stone and Cooper, 2001) operating within the community ultimately constitutes the basis for the validation of individual membership. In fact, central to communities of enthusiasts is the link between participation and individuals’ social identity (Tajfel, 1974; Turner et al., 1987). As happening in most social groups, “[e]nthusiasts have their personal identities invested in the collective identity of a market and regard market engagement as not only an economic activity but also a social one” (Koçak et al., 2014, p. 777). Gathering in communities, enthusiasts share their values and interests and adopt community-shared codes to enact their membership (Cattani et al., 2014). In a community of enthusiasts, where innovation is central to the reasons underpinning participation (West and Lakhani, 2008), members should therefore be more likely to display a preference for atypical products that publicly reflect their vanguard attitude toward innovation.

However, the social desirability of displaying a preference for atypicality can collide with the uncertainty associated with the consumption of previously unknown products (Ram and Sheth, 1989). Especially under the “nobody knows” principle that characterizes markets for creative products (Hirsch, 1972; Caves, 2000), community members might be reluctant to acclaim products whose features display a particularly high level of categorical atypicality (Hekkert et al., 2003; Uzzi and Spiro, 2005; Uzzi et al., 2013). Extant research has diffusely shown that highly atypical products are generally perceived as the result of inexperienced or low-quality producers (Zuckerman et al., 2003; Phillips et al., 2013; Kovács and Johnson, 2014) and are therefore less attractive to consumers. Within the boundaries of a community of enthusiasts, however, the uncertainty surrounding highly atypical products not only increases members’ risk of choosing a low-quality item, more importantly, it also threatens the perception of alignment between them and the shared vanguard attitude of the community, thereby invalidating their membership in the group.

In regard to publicly displaying their preferences, therefore, members of a community of enthusiasts face a trade-off. On the one hand, to express their vanguard attitude, in general, they would publicly display a preference for more atypical products. On the other hand, however, exposure to uncertainty makes them refrain from publicly acclaiming highly atypical products that could disconfirm their membership in the vanguard community.

The combination of positive attitudes toward atypicality and threats to membership validation determines a reluctant preference: a displayed preference that incorporates the reluctance of validation-seeking members to fully express their ideal attitude. In practical terms, community members will include in their displayed preference products with a moderate level of atypicality over poorly atypical alternatives but exclude those products with highly atypical (and therefore more uncertain) traits. This leads to our base hypothesis.

H1: In a community of enthusiasts, there is an inverted U-shaped relationship between the level of atypicality of a product and its likelihood of being included in the displayed preferences of community members.

Embedded in a network of mutual observation, participants in a community of enthusiasts constantly observe the behaviors of other participants and react to such observations accordingly (Kaplan, 2011; Hutter and Stark, 2015). Mutual observation activates a mechanism of social influence that has diffusely been shown by scholarly research and affects evaluation and choice. In general, individual evaluation is prone to informational social influence, defined as the “influence to accept information obtained from another as evidence about reality” (Deutsch and Gerard 1955, in Burnkrant and Cousineau, 1975: 206–207; emphasis in the original). The effect of social influence on evaluation is particularly acute when the quality of a product is uncertain (Dean, 1999; Salganik et al., 2006), when individuals have a high perception of risk (Tan, 1999; Kunze and Mai, 2007), or when the value of a good can be assessed only through personal experience (Nelson, 1970). In the field of music, for instance, Salganik et al. (2006) and Salganik and Watts (2008) have demonstrated that being exposed to other consumers’ choices makes evaluators more likely to choose cultural products that have been chosen by the majority (see also Kretschmer et al., 1999). In this framework, social influence fuels a bandwagon effect (Abrahamson, 1996) that makes individuals align their consumption choices to the observed trends, eventually enhanced by endorsing sponsors (Ross et al., 1984).

Social influence, however, can also have a subtler effect on evaluation. In addition to directly influencing the choices of consumers, the observation of others’ past choices also alters the condition under which individuals make their preferences visible to their relevant group (Ram and Sheth, 1989; Mitchell and McGoldrick, 1996). Within a community of enthusiasts, we refer to others’ past choice regarding a product (e.g. previous purchase) as a community endorsement granted to that focal product. We propose that such an endorsement indirectly sustains the alignment between the attitude and the displayed preferences of community participants by reducing the uncertainty surrounding alternatives and thereby lowering the perception of threats to the validation of in-group membership.

Under perceived uncertainty, the displayed preference of community members will exclude those highly atypical products whose complexity might merely reflect poor quality and undermine the validation of their membership in the vanguard community. However, the community endorsement granted to a focal product can serve as marker of the product’s quality or community-level desirability and can thereby relax uncertainty-driven reluctance. Taking into account the information embedded in others’ endorsement, participants in a community of enthusiasts will then include in their displayed preference those products, among other equally endorsed ones, with the highest level of atypicality—which will be, in fact, those that better fit with their underlying attitude toward innovation.

Previous research has shown that high-quality products may benefit from being atypical (Phillips and Zuckerman, 2001; Kovács and Johnson, 2014) since they attract omnivorous individuals (Peterson and Kern, 1996) looking for exclusive consumption patterns (Bourdieu, 1984). However, in a community of enthusiasts, we advance that the desirability of elevate atypicality is contingent on the behaviors of other community members. As Simmel (1957, 1991) would pose it, by selecting from the endorsed alternatives, “the individual is freed from the worry of choosing and appears simply as a creature of the group, as a vessel of the social context” (Simmel, 1957). However, “[w]hatever is exceptional, bizarre, or conspicuous … exercises a peculiar charm upon the man of culture” (Simmel, 1957). By choosing the most atypical product among equally endorsed alternatives, community members can therefore validate the vanguard attitude that made them participate in the community. This leads to a positive moderating effect of community endorsement on members’ displayed preference for atypicality. At low levels of endorsement, displayed preferences will still exclude highly atypical products. However, as the endorsement granted to a product increases, the displayed preferences will be profoundly altered by including products with consistently increasing levels of atypicality. In a community of enthusiasts, community endorsement therefore helps members reveal their vanguard attitude through displayed preferences.

H2: In a community of enthusiasts, the level of endorsement granted to a product positively moderates the curvilinear relationship between its level of atypicality and its likelihood of being included in the displayed preferences of community members.

3. Electronic music: empirical setting, data, and research design

For a long time, music has been a privileged setting to gather insights into many socioeconomic phenomena. In particular, electronic music has attracted scholarly attention as a fruitful field for exploring a wide variety of organizational, geographical, and socioeconomic issues (Hesmondhalgh, 1998; Gilbert and Pearson, 1999; Lange and Bürkner, 2013; Formilan and Stark, 2020). As we shall see, electronic music is representative of contemporary contexts where innovation proceeds through the recombination of sharp subgenre categories (Lena and Peterson, 2008).

In its most popular form, electronic music developed in some industrialized areas of Europe and the USA, where it musically synthesized the noise of assembly lines and factory plants (Reynolds, 1998). Quickly elected as the sound of a worldwide subcultural movement (Hebdige, 1979), electronic music fueled the Second Summer of Love in 1988–two decades after The Velvet Underground’s opening story (Thornton, 1996; Gilbert and Pearson, 1999). The cultural roots of electronic music have not been perverted by its recent mass popularization (O’Malley Greenburg, 2013), and the genre remains a breeding ground of political, social, and musical experimentations (Hofer, 2006; St John, 2006). As stated by the American producer Porter Robinson, “Electronic music is at its best and its healthiest when new, exciting, unexpected things are happening. This is a genre that thrives on novelty” (Rafter, 2018).

This latter dimension is central to our study. In fact, while electronic music’s novelty resided primarily in the chip-generated sound of the machines themselves (Nelson, 2015), the musical elements reproduced through the machines were adopted from other music genres—especially from rhythm and blues, funk, gospel, and 1970s disco music (Reynolds, 1998). As a consequence of its genre-recombinant origin, the classificatory schema of electronic music is populated by a large amount of highly distinct styles, and the audience is largely accustomed to products that are atypical of any single style. Although representing subgenre categories, electronic music styles work as sharp classificatory tags (Montauti and Wezel, 2016) that organize the macrogenre into nonoverlapping music scenes (McLeod, 2001; Formilan and Stark, 2020). As reported in Discogs’ submission rules, “Style is only required when using the Electronic genre”: without style specification, electronic music remains an excessively general denomination that would offer little guidance to consumers to navigate the genre.

As the background for this study, we conducted a qualitative exploration and triangulated information from different sources (Jick, 1979) to outline the general attitude of field participants toward categorical atypicality (Figure 1). As part of a broader project, we conducted more than 30 open-ended interviews with electronic music artists in Berlin and New York and distributed a questionnaire to electronic music enthusiasts through the online forum of Discogs (discogs.com), “the most reputed source for electronic music” (Montauti and Wezel, 2016: 958). Discogs is a platform that collects detailed information on audio recordings contributed by enthusiastic users, most of them experts in the field (fine collectors, artists, record labels’ founders, owners of records stores, and music students).

Discogs is composed of three primary areas: Explore, Marketplace and Community. The first area is the catalog, where users can scan the database and search for recording details. The second area is the marketplace, where registered users can buy, sell, and rate recordings. In this section, each recording profile page reports a statistics table with the number of users who claimed to own the recording, the average rating assigned to the recording, and the number of users who assigned a rating (Figure 2). The organization of recording profile pages is relevant to our research design in three ways. First, the statistics for each recording summarize the number of community participants that included the record on their wish list. The public display of preferences comes in many forms, such as ratings, rankings, and lists (Stark, 2011). Discogs users’ wish lists, in this sense, represent a visible marker of participants’ preferences. Second, the market statistics are immediately visible to any community member consulting the page of a recording. They represent a quickly accessible estimate of a community’s endorsement of a recording, which community participants can leverage to gain indirect information about the quality or desirability of the recording. Third, the figures summarizing how many Discogs users own or want a focal recording reflect mutually exclusive groups of community members, allowing us to disentangle others’ past behaviors (current possession of a recording) from the display of preferences (inclusion of a recording on a wish list). These features make the use of Discogs’ data particularly grounded for our analytical purposes. From the Explore and Marketplace sections of Discogs, we collected a large dataset of music recordings to test our hypotheses.

Finally, the Community area is devoted to discussion groups, the Discogs forum, and a blog. On the Discogs forum, users debate music and discography-related topics and contribute to the maintenance and development of the community itself. We leveraged the Discogs forum to distribute a questionnaire to electronic music enthusiasts, aiming at soliciting direct and explained insights into the attitude and preference of community participants.

Sixty-three electronic music enthusiasts voluntarily participated in our exploratory questionnaire. The sample, composed of self-selected participants, included 48 males and 13 females (2 not declared), aged between 20 and 60 (20–30: 45; 31–40: 12; 41–60: 6), from 9 different countries. Of them, 46 were electronic music lovers, and 17 were practitioners (11 producers, 4 DJs, 1 event manager, and 1 journalist). We structured the questionnaire into three sections, with both fixed and open-ended questions (outline reported in  Appendix A). We initially asked the respondents to choose the most appealing recording out of a set of four alternatives purposefully designed to display different levels of atypicality. We then asked them to explicate their choice in an open-ended question format. Finally, we posed some general questions on their attitude toward musical experimentation and style recombination.

As reported in Figure 3, most of the respondents (87%) chose one of the three tracks that spanned multiple styles. When asked to briefly explain their choice, their answers revealed a marked desire to consume music products that are able to raise curiosity and surprise.

Respondent. More is always better. Less is boring.

Respondent. It’s harder to understand how it's going to be. So, it's more interesting. And it gives me the idea that we're going beyond something.

Respondent. I would be interested in what they could have come up with combining all those styles.

The explanations provided by the questionnaire respondents were consistent with the answers they gave to other atypicality-focused questions (Figure 4; see also  Appendix A), confirming a general positive attitude toward style recombination.

Despite their general attitude and stated preference for atypicality, most of our respondents (52% of the overall respondents; 60% of those who did not choose the single-category track) still chose a music track that presented an intermediate level of atypicality. Figure 5 shows the variation between declared attitude and actual choice. Although purely illustrative, the largest discrepancy regards those respondents who declared themselves to be experimental listeners, coded on the basis of their statements’ intensity as reported in Figure 4.

Intriguingly, when attitude and choice are divided by level of expertise of the respondent (DJs, promoters and producers are the ones with the highest level of expertise in the field), the results preliminarily suggest that nonexpert consumers hold the highest level of reluctance—that is, the largest discrepancy between attitude and choice (Figure 6). In fact, more expert consumers have a more profound interest in the field and are therefore more likely to resolutely prefer highly atypical records (van Venrooij and Schmutz, 2018). Although not statistically significant, this preliminary evidence is intriguing. We will return to it in the concluding discussion.

In addition to directly distributing our questionnaire through the Discogs forum, we also explored the forum using a set of keywords to identify patterns of mutual observation within the community (Figure 1). In the Discogs forum, the presence and relevance of mutual observation is evident. Several discussion threads are aimed at solving problems of selective observation. In particular, the importance of mutual observation in our setting vividly emerged ∼6 years ago, when the Discogs administrators removed a feature that allowed users to see the complete list of community members who owned, wanted, or rated a focal record. This change in the structure of the records’ profile page raised a notable discussion in the forum.

Senior User. This is a very big problem that has been causing me a headache for the past few months. I know several dozen people on Discogs and I like to ask about artwork, label stamps, and other variations of records before I buy from a seller who I might not trust… Not being able to see the full list causes a headache, because I have to look at people's collections manually, and look for the record among sometimes several thousand items!!! Is there any way you guys can fix this?

User. Agreed. There are users whose taste I trust when deciding to buy a release and if they own it/rate it that will affect my decision to purchase a record.

Discogs’ members observe each other to reduce the uncertainty associated with purchase decisions. Observation involves not only person-to-person relationships but also wider attention to how much others have endorsed a particular music product.

User. I’d like to know if there is a way to find out what are the releases most owned at Discogs. Is it possible to find out? Is there still a stats page?

User. Is it possible somehow to browse records on Discogs in a view that sorts them according to number of owners? I.e., rank records in the order of how many people have the records in their collection?

The acknowledgment that consumers’ intentions and actions may not be congruent is not surprising (Bagozzi and Dholakia, 1999; Gollwitzer, 1999). However, within our community of enthusiasts, our explorative data suggest that the incongruence between attitude and displayed preference is tightly linked to the lack of information about the behavior of other members.

4. Sample and statistical model

To construct our sample, we identified all the artists that published at least one recording with an electronic music label based in Berlin, acknowledged worldwide as an important place for electronic music (Reynolds, 1998; Lange and Bürkner, 2013). To compile the initial list, we relied on Resident Advisor (residentadvisor.net), one of the leading magazine and community platforms dedicated to electronic music. We identified all the labels based in Berlin, created the lists of artists in their respective rosters, and then considered the whole discographies of these artists as included in Discogs. We set 1974 as the lower temporal boundary since it is considered the birthday of popular electronic music when Düsseldorf-based Kraftwerk released their LP Autobahn (Reynolds, 1998). The final sample included 25,518 recordings published by 3053 artists between 1974 and 2007 (the last year we could access data) and categorized through 151 different style categories.

4.1. Dependent variable

As measure of the displayed preference for a recording, we used the logged number of Discogs’ users that formally included a focal recording in their wish list, summarized in the “Items I Want” section of a focal recording profile page.

4.2. Independent variables

In line with previous research (Goldberg et al., 2016), we operationalize atypicality using a formula that includes the number of categories spanned by a focal recording alongside the cognitive distance among these categories. Mathematically,

Atypicality=1-11+log⁡1+D(x)lx-1

where D(x) reflects the cognitive distance among the x categories combined by the focal product, and l_x measures the number of x categories combined by the focal product. This measure has a value of 0 when focal product i’s categorization comprises only one style and grows to 1 as the combination of style categories becomes increasingly unusual.

We calculated the cognitive distance among the combined styles following the co-occurrence procedure largely used in categorization research (Kovács and Hannan, 2015). We constructed a measure of stylistic similarity as the cross-product of the relational matrix between recordings and styles. For each observation, the sum of paired-style similarity reflects the within-recording similarity (Rao et al., 2005). The measure of cognitive distance is then the reciprocal of similarity.

4.3. Moderating variables

To test Hypothesis 2, we constructed recording i’s measure of community endorsement as the logged number of Discogs’ users who declared to own the recording, as reported in the recording profile page. For interpretive purposes, we mean-centered the independent and moderating variables.

4.4. Control variables

We included a number of control variables in our regression models to control for confounding effects at the level of the artist, the recording, and the releasing label. At the recording level, we controlled for the “single,” “album,” or “other” formats of the recording (Recording Format) since different formats may be differently appealing to community members. We considered the geographical location where the recording was first published (Recording Location) to rule out the effect related to different markets’ relevance and a 5-year temporal window of recording publication (Recording Period) to exclude temporal effects. The decision to use a 5-year temporal window was informed by the temporal pattern of atypicality in our sample (Figure 7). However, the inclusion of the individual-year control variable did not alter the regression results.

We also controlled for the first style assigned to the recording (Recording Main Style), which could play a major role in orienting people’s understanding (Gregan-Paxton et al., 2005). Since Discogs is a community largely populated by collectors, we also constructed a variable that accounts for the type of diffusion experienced by a focal recording. In fact, some recordings might be poorly diffused on the market either because they are considered to be low-quality products or because they are rare. Conversely, community members might declare an interest in a particular recording because it is a must-have (highly diffused) or because it is a high-quality but rare product. This variable, labeled Diffusion Type, is composed of four levels: popular recordings that experienced high diffusion (above the mean) and received high rating (above the mean); unpopular recordings, poorly diffused and poorly rated; rare recordings, poorly diffused but highly rated; and despised recordings, highly diffused but poorly rated. In addition to accounting for the type of diffusion experienced by the recording, this variable also controls for the influence of the bandwagon effect of social influence (Abrahamson, 1996). Finally, in the statistical models, we also accounted for the raw number of spanned styles (Number of Styles).

At the label level, we included the betweenness centrality degree of the label (Label Centrality) to control for the influence it might exert from central or peripheral positions (Burt, 2004; Cattani et al., 2014), as well as its size in terms of volume of published recordings (Label Size). We also coded each label as “major” or “independent” (Major Label), following the classification of the 2011 Billboard International Buyers’ Guide that has been used in previous research on music (Peterson and Berger, 1971; Lopes, 1992; van Venrooij and Schmutz, 2018). Major labels usually have higher budgets for promotion and advertising and tend to experience higher visibility on the shelves.

Finally, at the artist level, we controlled for the number of years each artist has been active in the field (Artist Career Length) and the number of recordings he or she has published (Artist Release Volume), both of which we logged.

Table 1 shows the descriptive statistics and Pearson correlation of all variables. We controlled for multicollinearity using two different diagnostics—Variance Inflation Factor and the more sensitive Condition Number (not reported for brevity). Both diagnostics confirmed that multicollinearity was not a concern in our primary model specification.

4.5. Estimation procedure

We used fixed-effect (FE) ordinary least squares regression models to account for unobservable features and unit heterogeneity at the artist level that could influence the displayed preferences of community members. We compared the Akaike Information Criterion (Akaike, 1974) of ordinary least square (57,705), maximum likelihood (49,350), and artist-level FE (40,439) estimators and assessed the consistency of FE estimates over random-effect estimates using the Hausman specification test (Hausman, 1978). The result of the test led us to reject the null hypothesis of consistency of the random-effect estimates (χ² = 4,105; P = 0.000).

5. Results

Table 2 reports the results of the four models that progressively include the relevant regressors.

5.1. The reluctant preference: reward atypicality, refrain from the extremes

Models 2 and 3 show that despite a general tendency of atypical recordings to be included in the displayed preferences of community members (model 2), the preference for atypicality is essentially nonmonotonic (model 3). This confirms Hypothesis 1. As visually depicted in Figure 8, community participants have a positive attitude toward products that combine style categories in an atypical way but are reluctant to include in their displayed preferences those products whose level of atypicality is particularly high—the turning point is at ∼0.197 within the interval (−0.392, 0.516).

5.2. Community endorsement effect: abandon the reluctance

Model 4 shows the moderating effect of community endorsement. At low levels of endorsement, the relationship between product atypicality and likelihood of being included in displayed preferences remains curvilinear. However, as the community endorsement granted to a focal recording increases, the curvilinear relationship becomes strictly positive. This confirms Hypothesis 2. Figure 9 shows the relationship between atypicality and likelihood of being included in preferences at low and high levels of endorsement (below and above the average endorsement, respectively). The table Critical Points quantifies the practical significance of the moderating effect.

As hypothesized, the displayed preference for atypical products undergoes a profound change when the level of community endorsement is considered. At intermediate levels of atypicality, the likelihood of inclusion is the same for both highly and poorly endorsed recordings (+7.5%). However, poorly endorsed recordings remain excluded from potential buyers’ wish lists as atypicality exceeds such an intermediate level, while the trend is the opposite for highly endorsed recordings. Marked community endorsement, in other words, deeply impacts the reluctant preference for intermediate atypicality, unveiling potential buyers’ preference for those recordings whose level of innovativeness is higher.

5.3. Robustness and sensitivity checks

We ran different specifications of the base model to check the robustness of our results (not reported for brevity). First, we clustered standard errors at the artist level to control for serial correlation within each artist. Second, we considered a subsample composed only of recordings that combined less than six different styles (N = 24,592), since the average number of combined styles in our sample is 2.18, and 95% of all recordings span less than eight styles. We also ran the same model on a subsample that excluded the recordings published before 1990 that might have biased our results (N = 24,629). Finally, we excluded all nonatypical recordings to test whether the community endorsement mechanism remained effective when considering only atypical recordings (N = 16,058). In all these robustness check models, coefficients underwent no significant change, and the results were still consistent with the main model.

To further test the sensitivity of our findings under different assumptions, we developed two additional models that account for alternative explanations and possible confounding effects (Table 3).

First, we acknowledge that the preference for a recording might largely depend on the reputation an artist has developed over his or her career (Sorenson, 2014). To control for the effect of past recordings on the displayed preferences for the focal recording, we ran an additional model that explicitly included the community endorsement granted to the previous three recordings published by the same artist (model 5). Although the three lags of endorsement have all positive and significant coefficients, they did not alter the effects of our main regressors.

Second, concerns might arise with respect to the influence exerted by the releasing label of a recording on its likelihood of being included in the display preferences of community members. Labels can affect the preferences for their new recordings by investing in a high level of marketing resources (Jones, 2006). Moreover, as emerged during our interviews, electronic music labels act as style-specialized communities on the scene, facilitating the recognition and legitimation of artists and recordings. To control for the influence of the label, we ran label-level FE regression, including Artist Release Volume and Artist Career Length, to maintain the artist’s invariant features (model 6). In spite of lower statistical significance, the results remained consistent with the main model.

6. Discussion

Innovative products that combine categorical material in unexpected and atypical ways are often initially opposed by consumers (Mainemelis, 2010) but might be consecrated later as prophetic pioneers (Staw, 1995; Allen and Lincoln, 2004; Allen and Parsons, 2006). In this article, we drew on the notions of atypicality and social influence to shed preliminary light on the mechanism behind this phenomenon.

Focusing on a community of enthusiasts and the shared attitude of its vanguard members (Fiol and Romanelli, 2012; Koçak et al., 2014), our study introduced the notion of reluctant preference as an audience-side mechanism that informs research on the reception of atypicality. Departing from a producer-side argument, the notion of reluctant preference helps to explain why novelty-seeking audiences looking for surprising and unusual sensations (Zuckerman, 2007) still shy away from marked atypicality (Hekkert et al., 2003; Uzzi and Spiro, 2005; Askin and Mauskapf, 2017). Participating in communities, consumers look for products that promise to satisfy of their curiosity. Importantly, they also seek validation of their membership in the community itself (Turner et al., 1987). Our results showed that this is also the case among expert consumers of innovative products because the “nobody knows” uncertainty that permeates the field constantly threatens the vanguard traits individuals seek to validate (Caves, 2000). In this sense, not only casual consumers but also vanguard-oriented experts constantly observe the behavior of their peers and, when reassured by others’ endorsement, become more likely to publicly display their preferences to validate to their vanguard membership.

Operating as “cultural infrastructure” (Vergne and Wry, 2014: 59), categories intervene in the interstice between the individual and the collective. They organize the sociocognitive space (Bandura, 1989, 2001; Goldberg et al., 2016) and standards of evaluation (Stone and Cooper, 2001) that help people make sense of their environment. However, such standards are localized. In markets for securities, for example, the attention of evaluators is driven by sharp categorical fitness (Zuckerman, 1999), whereas hiring committees evaluating the job market sometimes hold opposite criteria of evaluation (Merluzzi and Phillips, 2016). Whereas market-makers actively look for atypical ideas (Pontikes, 2012), cinema and restaurant goers do not always appreciate atypicality (Hsu, 2006; Kovács and Johnson, 2014). The presence of diverse standards of evaluation partially explains why social objects as diverse as securities, candidate profiles, movies, and restaurants keep spanning multiple categories. As Wry et al. (2014) showed, it is not the fact of spanning categories per se that might matter to audiences. In this article, we argued instead that evaluators express preferences for that which can contribute to validating their membership in social groups. The same categories used to classify other market participants are leveraged by audience members to self-classify their own participation (Turner et al., 1987). Vanguard audiences attentive to vanguard producers might hold standards of evaluation that vary dramatically from those held by casual or specialized consumers. In fact, despite different levels of expertise, both casual (e.g. restaurant goers; Kovács and Johnson, 2014) and specialized evaluators (e.g., security analysts; Zuckerman, 1999) leverage evaluative standards that favor categorically focused candidates. In this sense, although hybrid producers might be exposed to threats of illegitimacy in some contexts, the coexistence of multiple standards of evaluation ensures spaces for innovation (e.g., Wry et al., 2014).

Our findings also contribute to current research on social influence and the diffusion of innovation. First, departing from a multiple-audience framework where different audiences influence each other’s evaluation (Fini et al., 2018), our article shows how mutual observation also exerts its effects within the same audience, contributing to the formation of consensus about how to evaluate innovative practices. Second, drawing on a peer-influence framework, but departing from choice as the primary marker of social influence (Salganik et al., 2006; Salganik and Watts, 2008), our study highlights how actual choices—the ones constructing a critical mass of early adopters–are not the only outcome of mutual influence. More subtly, the observation of others’ past choices deeply alters the public display of the preferences of evaluators. The mechanisms we outlined do not need to correspond to actual consumption patterns—and, in fact, attitudes and behaviors are not always consistent with one another (Bagozzi and Dholakia, 1999; Gollwitzer, 1999). Nonetheless, by altering the display of preferences within a community of enthusiasts, mutual observation contributes to the systemic diffusion of innovation. In fact, the diffusion of atypical products within the community is transferred to a broader audience, since members of communities of enthusiasts are often active professionals of the field. In our case, for instance, DJs can push their renewed taste for atypicality during club events, or label owners might start looking for artists who can fit with their increased preference for an atypical recombination. Since—even analytically—the perception of atypicality is contingent on the frequency in which diverse categories are combined (Kovács and Hannan, 2015; Goldberg et al., 2016), the diffusion of atypical recombinations in society increases the co-occurrences of previously disconnected categories and thereby raises the threshold at which atypical innovation will be opposed.

As Fiol and Romanelli (2012) noticed, “Some ideas attract proponents or enthusiasts, which begin the formation of a collective that may sometimes gain recognition and legitimacy among a larger audience.” Departing from proponents as the primary agents of change, our findings note the centrality of enthusiasts in initiating the well-known S-shaped curve by which innovation spreads throughout society. Most industries comprise vanguard communities similar to the one we studied. Beta testers of videogames, magazines focused on technological innovation and high-end fashion, blogs spreading new lifestyle recommendations and contemporary societal issues and other such actors are all central nodes in their respective communities of enthusiasts. Examples of community-based alterations of enthusiasts’ preferences are also widespread, with tradeshows across different industries being a notable instance. From technology fairs to fashion weeks, tradeshows bring together pioneers and enthusiasts and collectively expose them to innovative products, processes, and practices. Leaving the show, the fashion enthusiast will spread his new taste for previously unseen outfits, and the technology enthusiast will display her updated preferences for cutting-edge technologies. In extending the boundaries of what is typical, communities of enthusiasts are then a central engine of innovation in all those settings where group practices represent a valuable resource for extending the boundaries of the known (Dahlander and Frederiksen, 2012).

The same process can be read from the perspective of producers, shedding some light on The Velvet Underground’s opening story. Central to industrial economics are questions regarding the relative advantage of first versus second movers in acquiring market share and achieving superior performance (Rasmusen and Yoon, 2012). Especially under uncertainty and lack of market knowledge, producers might benefit from delaying the release of a new product and leave a first-mover role to their competitors. Our findings suggest that such a second-mover advantage might also derive from a social mechanism of the alteration of preferences following the action of pioneering producers: their innovation might not gain immediate market attention but certainly contributes to raising the threshold of acceptability of atypical features. Market entrants can then benefit from consumers’ lower resistance to change (Ram and Sheth, 1989; Laukkanen et al., 2007), capturing a second-mover competitive advantage. At the same time, however, first-moving pioneers might be later acclaimed as groundbreaking innovators. If this happens, they would experience not simply critical or commercial acclaim but an unanimous consecration (Allen and Lincoln, 2004; Allen and Parsons, 2006) that secures them a market share well above the one gained by newcomers.

Although enlarging the current understanding of the reception and diffusion of innovation, our study also raises a number of issues for future research. We provisionally showed that less expert enthusiasts (general electronic music listeners) are the ones that mostly present a discrepancy between their positive attitude toward atypicality and their preferences (Figure 6). We suggested that these listeners, despite their enthusiasm for atypicality, might have a higher perception of uncertainty compared with more experienced individuals. Future research might explore reluctance and the influence of community endorsement on and across different groups of community participants and between diverse communities. Moreover, while we were able to distinguish potential buyers from other community members, we could not access information on the temporal functioning of the endorsement mechanism. This, combined with the structure of our data where products’ features are treated as a basket of similar alternatives, opens up room for further enquiry. Future research could develop suitable research designs to investigate how the community endorsement mechanism operates dynamically and over time. We believe experimental protocols could serve the purpose while also making it possible to situate the reluctant preference mechanism on a product-evaluator match instead of on an aggregated basket of similar alternatives.

Longing to innovations that stand out and are deemed outstanding (De Vaan et al., 2014), early adopters looking for membership validation in vanguard communities raise the threshold of atypicality. In so doing, they also expand the range of products deemed typical—and therefore accessible—by the general consumer. Proceeding through the careful socialization of preferences, these twined processes fuel the diffusion of innovation.

Acknowledgments

We are thankful to Simone Ferriani, David Stark, Casey Jun, Fredrik Tell, Michael G. Jacobides, and two anonymous reviewers for their valuable comments and support. We are grateful to Simone Santoni who shared a large part of the raw data analyzed in the study. Any inaccuracy in the use of data is solely ours.

Funding

This work has been supported by the European Research Council (“Blindspot” [Grant: 695256]).

References

Appendix A. Questionnaire outline