University of Amsterdam Faculty of Business and Economics
MSc Business Administration
Entrepreneurship and Management in the Creative Industries track Genre Spanning and Audience Appeal as Antecedents of Genre Consensus:
The Case of Electronic Dance Music DJs Master Thesis
June 29, 2015 Student: Valerie Bollen
10837949
1
Statement of Originality
This document is written by student Valerie Bollen who declares to take full responsibility for
the contents of this document.
I declare that the text and the work presented in this document is original and that no sources
other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of
2
Table of Contents
I. Abstract 4
II. Acknowledgements 5
III. List of Tables and Figures 6
1. Introduction 7, 8
2. Literature Review 9
2.1 An Introduction to Genre/Category Theory 9 – 11
2.2 Genre Consensus 11 – 13
2.3 Genre Spanning and Audience Appeal 14 – 16
2.4 Word-of-Mouth Theory: Creating the ‘Buzz’ 16 – 18
2.5 Hypotheses 18 – 20
3. Method 21
3.1 Sample 21
3.2 Data Collection 22
3.2.1 DJ Databases 22 – 24
3.2.2 Social Media & Ranking Lists 24 – 26
3.3 Genre Classification System Development 27 – 30
3.4 Variables and Measures 31 – 34
4. Results 35
4.1 Descriptive Statistics 35 – 42
4.2 Regression Analyses 43
4.2.1 Genre Spanning and Genre Consensus 43, 44
4.2.2 Audience Appeal and Genre Consensus 44
3
5. Discussion 46
5.1 Main Findings 46, 47
5.2 Implications 47, 48
5.3 Limitations 48, 49
5.4 Suggestions for Future Research 49, 50
6. Conclusion 51
References 52 – 58
Appendices
A. Sampling Frame DJ Mag Top 100 DJs 2010 – 2014 59 – 62
B. Sample DJ Mag Top 100 A-Z 63, 64
4
I.
Abstract
In the music industry, genre categorization systems play an important role in audiences’
evaluation of artists. If audiences are in agreement about an artist’s category-membership, this may positively affect the artist’s career. However, little research has addressed the predictors of genre consensus. Therefore, this study investigated the antecedents of genre consensus by
seeking to answer the question: “To what extent function genre spanning and audience appeal
as the antecedents of genre consensus?”
The Electronic Dance Music (EDM) industry was selected as empirical setting since it
is a largely neglected industry in scientific literature and because genres have a tremendous
influence on the EDM community. The study focused on the 187 most prominent EDM DJs
who had obtained a position in the DJ Mag Top 100 list in the period of 2010 – 2014.
Drawing on data from four major online DJ databases, several social media websites and
commercial music platforms, it was examined whether genre spanning and audience appeal
influenced genre consensus.
The results suggested that genre spanning negatively affected genre consensus, while
repeated appearances on the DJ Mag Top 100 list had a positive effect. Contrary to my
expectations, the artists´ years of experience in the music industry did not moderate the
negative effect between genre spanning and genre consensus.
This study contributes to theories of categorization and organizational ecology by
identifying the effects of genre consensus and audience appeal on genre consensus, and
5
II.
Acknowledgements
First and foremost I offer my sincerest gratitude to my supervisor Bram Kuijken MSc – thank
you very much for your guidance, encouragement and valuable feedback throughout this
research project.
Secondly, I would like to show my greatest appreciation to prof. dr. N. M. Wijnberg
for his illuminating insights that helped solving the puzzle of the rationale in this thesis.
Last but not least, I would like to express my appreciation towards my family and
friends who have supported me throughout the entire process.
6
III. List of Tables and Figures
Tables
Table 1 – Changes in Genre Classification Systems Based on Name Similarity 28
Table 2 – Genre Representation Across the Four Sources 29
Table 3 – Changes in Genre Classification Systems Based on Subgenre Identification 30
Table 4 – DJ Avicii Example Jaccard Similarity Coefficient Calculation 34
Table 5 – Genre Frequencies and Percentages per Source 36
Table 6 – Times Ranked in DJ Mag Top 100 List 2010 – 2014 39
Table 7 – Pairwise Consensus Comparisons Between Sources 40
Table 8 – Mean, Standard Deviation and Correlations of Study Variables 42
Table 9 – Linear Regression Analysis Genre Spanning and Genre Consensus 43
Table 10 – Moderation Model of Predictors of Genre Consensus 44
Table 11 – Linear Regression Analysis Times Ranked and Genre Consensus 44
Table 12 – Linear Model of the Predictors of Change in Genre Consensus Scores 45
Figures
Figure 1 – Histogram: number of DJs per country 35
Figure 2 – Histogram: genre spanning 37
Figure 3 – Box plot: years of experience 38
7
1.
Introduction
Nowadays, consumers make use of online platforms to discover, discuss and rate music, share
personal playlists and vote for their favourite artists. As a result, it has become more
important for artists to construct and manage an online identity (Koosel, 2013), and genres
help to facilitate this identity construction (Shocker, Bayus, & Kim, 2004). Consequently,
genre categorization systems are more important than ever because they capture the way
music is divided in the minds of consumers, and the way the production and distribution of
music is structured (DiMaggio, 1987). These categorization systems have an impact on how
individuals shape their music tastes and make sense of different artists and their identities
(Mattsson, Peltoniemi, & Parvinen, 2010), which influences their behaviour and consumption
patterns and thereby influences the chances of artists’ success (Zuckerman & Kim, 2003).
Genre categorization systems differ across societies and among their members
(DiMaggio, 1987). Even though audiences and producers collectively shape the structure of
these systems through interaction, they may still differ in how they apply category labels
(Rosa, Porac, Runser-Spanjol, & Saxon, 1999). In addition, an artist may even be associated
with different genres across audiences. However, if audiences are in agreement regarding the
genre profile of an artist, this is called genre consensus.
Some argue that genre consensus should be strived for because of the positive effect
on audience appeal (Hsu, 2006). That is, whether an offering is intrinsically appealing to the
members of the audience (Hannan, 2010). A lack of genre consensus may be a barrier to the
legitimation of an artist’s genre profile (Baumann, 2007), which may lead to devaluations
from the audience members (Zuckerman, 1999) and thereby negatively affect artist success.
Interestingly, little research has examined the determinants of genre consensus. Prior
research has found that consensus is influenced by the degree of connectivity between
8
turnover (Cattani, Ferriani, Negro, & Perretti, 2008). Also, it has been clarified how
consensus can be reached among the members of a cultural community through legitimation
and justification (Baumann, 2007), and how this leads to an increase in popularity and appeal
(Scott, 2012). Moreover, the effects of category spanning and category consensus on audience
appeal have been explored (Hsu, 2006).
However, to my best knowledge, genre spanning and audience appeal have not yet
been measured in the literature as antecedents of genre consensus, even though this may
provide some important insights for artists, and music marketing strategists. Therefore, this
study seeks to answer the following research question:
“To what extent function genre spanning and audience appeal as the antecedents of genre consensus?”
The empirical focus is on Electronic Dance Music (EDM) industry in which disk-jockeys
(DJs) represent the most prominent artist type. This industry is not only largely neglected in
scientific literature, but also one in which genres have a tremendous influence (McLeod,
2001). The sample consists of 187 EDM DJs who obtained a position in the DJ Mag Top 100
list in the period of 2010 – 2014. Secondary data from numerous online databases and
platforms are assessed, among which are DJ Mag, DJ Rankings, Partyflock, The DJ List, and
Top Deejays. These databases are used as they attract millions of visitors and make use of
genre categorization systems to classify artists.
The objective of this study is two-fold. First, it strives to contribute to categorization
theory and organizational ecology theory by identifying the effects of genre spanning and
audience appeal on genre consensus. Second, it seeks to offer insights for artists, labels, and
9
2.
Literature Review
“Art worlds typically devote considerable attention to trying to decide what is and isn’t art, what is and isn’t their kind of art, and who is and isn’t an artist” (Becker, 1982, p. 36).
2.1 An Introduction to Genre/Category Theory
Following DiMaggio (1987), genres are defined as “sets of artworks classified together on the basis of perceived similarities” (p. 441) and represent socially constructed categories. They are used to classify varieties of cultural products, particularly in the fields of visual art,
popular culture, film, literature, and music (Lena & Peterson, 2008). Within these fields,
categorization systems shape organizational dynamics and success (Hirsch, 1972). DiMaggio
(1987) explained that these so-called social artistic classification systems (ACSs) capture two
sets of processes: the way art is divided in the minds of consumers and the way institutions
structure the production and distribution of art. ACSs vary along four dimensions; (1)
differentiation, (2) hierarchy, (3) universality and (4) boundary strength. More specifically,
they differ in (1) the number of institutionally bounded genres, (2) the degree to which genres
are hierarchically ranked by prestige, (3) the degree to which classifications are similar among
subgroups of members, and (4) the degree to which tastes are clustered within ritual
boundaries.
If boundary strength is low, a category’s boundaries are called ‘fuzzy’ (Hannan,
2010). Fuzziness arises (1) when there is disagreement among producers and audiences about
which attributes and behaviour is typical for the category, and (2) when category members
have memberships in multiple categories (Vergne & Wry, 2014). The offerings of an artist or
10
boundaries, which may be challenging because boundaries change over time and differ across
audiences (Vergne & Wry, 2014). In addition, Lena and Peterson (2008) mentioned that
genres are constantly debated in dialogues among fans, artists, critics and marketing
strategists, which provides the opportunity for contesting musical quality and social prestige.
Due to these interactions, genres emerge, evolve and disappear over time (Lamont & Molnár,
2002). It also leads to the emergence of subgenres; subordinate categories within a particular
genre (McLeod, 2001). This boundary work is crucial because genres often compete for the
same resources such as fans, capital, media attention, and legitimacy (Lena & Peterson,
2008).
In their review of categorization literature, Vergne and Wry (2014) distinguished
between two types of organizational categorization theory: self-categorization and categorical
imperative. The self-categorization perspective, or cognitive psychological approach,
emphasizes how strategic managers perceive the external environment and their firm’s
position within that environment, and focuses on aspects such as power, politics, interest
seeking, and strategic co-optation (Porac, Wade, & Pollock, 1999). According to this
producer point-of-view, strategists construct mental models of the competitive environment,
and in turn these managerial perceptions determine the structure of the industry (Porac, Thomas, & Baden‐Fuller, 1989). In this case, producers pursue self-selection into a category through imitative behaviour and strategic use of linguistic tools. Therefore, category labels are
crucial as they help producers seek membership in an existing category (Vergne & Wry,
2014).
In contrast, when adopting the categorical imperative perspective, categories are
described from the audience-side. Audiences are homogeneous set of agents who have an
interest in a specific field and control over material and symbolic resources which affect the
11
categorical imperative is that audience members (e.g. critics, media, and consumers) attach
labels to categories, with which they associate a set of properties and rules that category
members (e.g. organizations or artists) are expected to follow. As a result, the labels facilitate
an evaluation process in which audience members determine which category an organization
or artist fits into and whether this matches their expectations (Vergne & Wry, 2014).
Therefore, categorization simplifies information processing and decision making for potential
consumers as it provides a context in which similarities and differences among cultural
products and producers can be highlighted (Shocker et al., 2004).
Hsu (2006) emphasized that an organization should ensure an intrinsic fit between its offerings and the audience’s taste preferences, and devote some level of engagement to make its offerings available to potential audience members. Without this, organizations are not able
to garner resources from them. This can be illustrated by an example from the music industry.
Audience members compare the attributes of artists that enter the music industry to a
collective system of categories and social codes (Mattsson et al., 2010). For instance, if an
artist´s music is profiled as a house music, and its attributes match with the expected category
characteristics and social codes, the artist will be perceived as a legitimate member of the
house genre. However, if audience members perceive a deviation in attributes, this may lead to devaluations (Zuckerman, 1999). In that case, people may reject artists, which is likely to
negatively influence their artistic careers.
2.2 Genre Consensus
In cultural industries, an agreement among audiences regarding the genre(s) of an artist is
referred to as genre consensus. Put differently, there is genre consensus when there is a
12
In general, consensus is issue-specific, which implies that any specific norm, belief or
practice is more or less consensual in a situation (Zelditch, 2001). Adopting the
self-categorization perspective, there is category consensus if organizations with common
attributes perceive themselves as members of a particular category (Vergne & Wry, 2014).
However, this study will assess consensus following other researchers, who have drawn upon
the categorical imperative, or ecological perspective.
From an ecological point-of-view, consensus among audiences is reached once the
features and activities of producers have become taken-for-granted elements (Cattani et al.,
2008). According to Zelditch (2001), this is achieved through justification, which is “an
argument made to explain how the unaccepted is in fact acceptable because it conforms to
existing, valid norms, values, or rules” (Baumann, 2007, p. 49). From the field of sociology it
can be learned that, once the elements of a social order are justified and seen as in agreement
with norms, values, and beliefs that individuals presume are widely shared, legitimation
occurs (Weber, 1978). Put simply, legitimation refers to the process of how categories or
category members gain and maintain acceptance (Vergne & Wry, 2014). Delegitimation
therefore refers to the process of losing this acceptance (Berger, Ridgeway, Fisek, & Norman,
1998). According to Hannan, Pólos, and Carroll (2007), legitimation from the ecological perspective is viewed as “conformity of feature values to schemata,” in the sense that it “grows with the level of consensus within the audience about the meaning of a label’ (p. 98). A genre only exists if it is recognized as a salient unit of analysis by a sufficient
number of members (e.g. artists) and audiences (e.g., critics, media, and consumers) (Vergne
& Wry, 2014). Ridgeway and Correll (2006) drew upon status construction theory in order to
explain how encounters between people within a social community spread status beliefs as
one teaches a previously acquired belief to another. Beliefs and knowledge of genres is
13
1995). Therefore it is reasoned that dialogues within such communities facilitate justification
for status beliefs regarding the characteristics and identities of producers and artists.
Cattani et al. (2008) contributed to this rationale as they discovered that consensus is
affected by the network structure because this facilitates interaction between audiences and
organizations. They specified that audience members have an influence on the organizations’
survival as they reach, reinforce, and preserve consensus about the organizations’ features and
behaviour. Furthermore, they suggested two conditions which account for the creation of
consensus among audiences about which characteristics and attributes firms must show in
order to be accepted or excluded from a particular category. They identified (1) the degree of
connectivity of the network between audiences and organizations, and (2) the degree of
repeated interactions, as antecedents of consensus. In addition, they found that audience
turnover destabilizes consensus (Cattani et al., 2008).
Other research in the field of organizational ecology has focused on genre consensus
as a predictor rather than an outcome. For instance, Hsu (2006) highlighted that when
offerings clearly fall within a certain category and establish a clear fit with the targeted taste
position, this allows the audience to understand the characteristics that these offerings have in
common, which has a positive effect on audience appeal.
In this study, genre consensus is considered to be a desired outcome as it is reasoned
to legitimate art forms, and contribute to appeal and an artist’s survival within the field.
However, it must be mentioned that consensus will never be absolute, as there is never
complete consensus within a society about anything (Baumann, 2007). As a result, consensus
at the collective level – and not necessarily at the individual level – is considered to be
sufficient to reach legitimation (Baumann, 2007; Zelditch, 2001).
14 2.3 Genre Spanning and Audience Appeal
An organization engages in category spanning when it has simultaneous membership in two
or more categories (Vergne & Wry, 2014). Correspondingly, an artist engages in genre
spanning through membership in two or more genres at the same time. The concept of niche is relevant here, which refers to a small group of customers with similar needs and
preferences (Dalgic & Leeuw, 1994). In niche marketing, a company focuses on a certain
category to fulfil those customers’ needs. Spanning multiple categories indicates a wider niche width, which is expected to attract a broader audience (Hsu, 2006). But whether an
offering is intrinsically appealing to the members of the audience (that is, whether audience
appeal is high) depends on how well the offering matches with the taste of that audience (Hannan, 2010).
Niche width measures the range of environmental dimensions across which an
organization exists (Carroll, 1985). Taking the newspaper publishing industry as illustration,
Carroll (1985) explained the divide between generalist and specialist firms. While the former
are organizations that seek to exploit a wide array of sources, the latter focus on only one or a
limited few domains. When addressing the topic of category spanning, researchers often make
this division between generalists and specialists (Hsu, 2006; Zuckerman, Kim, Ukanwa, &
von Rittmann, 2003), in the sense that generalists engage in category spanning, while
specialists do not.
From the self-categorization perspective, it is argued that spanning categories may
foster organizational success through competitive differentiation (Porac et al., 1989). In some
cases, strategic manipulation of multiple identities can be a source of power (Padgett &
Ansell, 1993). Moreover, categories with many subcategories may allow their members more
leeway to innovate than categories with few sub-categories (Brewer, 1993). However,
15
several roles with respect to the same audience, and this often outweighs its advantages.
Prior research which adopted the categorical imperative point-of-view has addressed
that organizational identities are built around codes and rules which audiences regard as
standards for a producer or firm (Hsu & Hannan, 2005). So, membership in categories
constitutes part of the identities of producers and organizations, and these identities may or
may not appeal to the members of the audience. McKendrick, Jaffee, Carroll, and Khessina
(2003) discovered that producers who seek membership in multiple market categories are less
likely to construct clear and appealing identities to relevant audiences. Besides lower appeal,
category spanning was also found to result in less legitimation (Zuckerman, 1999). On the
other hand, focused or specialized identities have an advantage because they facilitate
valuation and legitimation (Zuckerman et al., 2003).
According to Negro, Hannan, and Rao (2010), category spanning may lower audience
appeal because of: (1) partiality of category memberships (atypicality), (2) categorical
contrast, and (3) expertise or capability. Regarding atypicality, it is argued that category
spanning may lower the appeal of offerings in categories because it confuses audiences (Hsu,
2006). As mentioned before, when offerings clearly fall within a certain category, this allows
the audience to understand the characteristics which these offerings have in common.
Consequently, if the offerings are too complex to match with audience members’ expectations
and perceptions of a particular category, they will have more difficulty interpreting the
identity of the producer or organization, and such blurred identities lead to confusion and
lower appeal (Hsu, 2006).
Second, spanning categories is also found to negatively influence appeal through
lowered categorical contrast – that is, a decrease in the sharpness of a category’s boundaries.
This leads to a growing disagreement about the category and thereby reduces appeal (Negro et
16
Third, category spanning has been theorized as a potential influencing factor on the
perception of expertise or capability, in the sense that producers or organizations who span
multiple categories are expected to develop less expertise in each category in comparison to
category specialists, which would result into a lower fit with the category schemas and
subsequently lower audience appeal (Hsu, Hannan, & Koçak, 2009). Even actors who manage
to develop a high level of expertise in multiple categories face the difficulty of convincing
audiences of this (Kovács, Hannan, & Sorenson, 2015). However, Hsu et al. (2009) found
empirical evidence that generalists were not devalued because spanning categories indicated
poor skill, but because they were not perceived as genuine full members of a particular
category.
On the contrary, an advantage of generalists is that spanning facilitates flexibility
(Zuckerman et al., 2003). They attract larger audiences, and spreading risk across multiple
regions of the environment may help them to outlast specialists (Hsu, 2006). On top of that,
Pinheiro and Dowd (2009) found that aesthetic generalism (that is, being conversant in
multiple genres), had a positive effect on the earnings and national recognition of jazz
musicians. Finally, Hsu (2006) hypothesized and found that genre consensus mediates the
negative effects of genre spanning on audience appeal. But, in order to reach genre consensus,
interaction (Cattani et al., 2008), justification, and legitimation is needed (Baumann, 2007).
This is where word-of-mouth communication comes into play.
2.4 Word-of-Mouth Theory: Creating the ‘Buzz’
Word-of-mouth (WOM) is the process of conveying information from person to person and
plays a major role in customer buying decisions (Richins & Root-Shaffer, 1988). It functions
through social networking and trust, because people tend to share opinions, reactions and
17
perceive as trustworthy (Chevalier & Mayzlin, 2006). Concerning the quality of experience
goods, information from others is valued because the product quality is unknown prior to
consumption (Carare, 2012; Peltoniemi, 2014).
During the past decade, electronic word-of-mouth has substituted and complemented
other forms of business-to-consumer and offline word-of- mouth communication about the
quality of products (Chevalier & Mayzlin, 2006). It has been shown that people are influenced
by online reviews of other customers, even when they are not acquainted. For instance, in
their study on the effects of online user reviews, Duan, Gu, and Whinston (2008) found
evidence that movies’ box office sales were significantly influenced by the volume of online reviews. They considered this awareness effect to online user reviews as an indicator of
underlying word-of-mouth.
Besides online customer reviews, also ranking lists and sales charts are known to
affect customer behaviour and buying decisions because they provide visibility (Yoo & Kim,
2012). It is widely accepted that charts attribute value to entities that would otherwise have
remained unrevealed (Attali, 1985). For example, Sorensen (2007) found that book sales
increase followed the rankings in the New York Times bestseller list. It was reasoned that
appearing in a chart such as a bestseller list serves as a signal of quality. As the quality of
experience and cultural goods are unknown to the consumer prior to consumption, ignorant
potential consumers tend to believe that the rankings reflect other buyers’ perspective on the
quality of the cultural good (that is, audience appeal). This phenomenon of people using
information of popularity of products as a signal of quality is referred to as observational
learning (Hedström & Swedberg, 1998).
Also Carare (2012) found that the public information about the past popularity of
products in the form of bestseller lists significantly affects customers’ purchase decisions. His
18
to pay of consumers was approximately $4.50 greater for a top ranked app than for the same
unranked app. In addition, Salganik and Watts (2008) discovered that participants’
preferences were shaped by the music download choices of other participants in their
experiment with an artificial online music market.
According to Scott (2012), online ranking charts are central to building buzz because
they function as a template for comparing, valuing and ordering music producers, and thus
they act as a proxy for popularity and market potential. He referred to buzz as “the infectious
power of rumours and recommendations circulating through dense cultural intermediary
networks” (p. 244), which implies that people are influenced in their tastes and purchasing decisions by their social environment. A positive buzz generates excitement and enthusiasm
and can be used to form an audience, stimulate consumption, and generate marketable values
(Scott, 2012), and this is what word-of-mouth marketing seeks to achieve.
Becoming aware of and consuming music can be seen as a social process when people
listen to music together and form opinions about music based on others’ assessments. This
intangible social value formed from and within specific contexts helps to create a buzz around
that particular art/culture (Caves, 2000; Currid, 2007), which influences the economic value
of a cultural good such as music. Therefore, artists, labels, booking agencies and other
cultural agents construct music communities online in order to provide a platform for buzz,
since this is expected to lead to higher audience awareness and appeal.
2.5 Hypotheses
All in all, several determinants of consensus have been addressed in scientific literature. From
a sociological point-of-view, it was indicated that justification and legitimation precede
consensus (Baumann, 2007). From an ecological perspective, it was found that spanning
19
targeted positions (Hsu, 2006), which is expected to result in less consensus about the
organization’s position in the market. Moreover, the lack of focus that comes with a wider
niche width might prevent a clear identity from forming in the audience (McKendrick et al.,
2003; Negro et al., 2010). All things considered, it is assumed that successful justification and
legitimation becomes more difficult to achieve as spanning increases. So, in context of the
music industry, it is expected that the more genres are spanned, the more difficult it becomes
to justify and legitimize the genre classification of artists, which should result in a lower
consensus across audiences.
In addition, Phillips and Zuckerman (2001) argued that high-status actors have less of
a need to conform to broad cultural codes in order to construct an identity that appeals to the
audience, because their status affords allow them to deviate to some degree. Also Mattsson et
al. (2010) argued that artists making their first entry are likely to face higher penalties by
audiences if they deviate from existing genres. This could imply that high-status artists who
are experienced in the music industry are less vulnerable to the negative effects of spanning
on the genre consensus among audiences than newcomers. In the same line, Zuckerman et al.
(2003) found that the trade-off between a generalist or specialist identity is greater among
novice actors, because novices have yet to go through the audiences’ selection process that
differentiates the skilled from the unskilled. Altogether, it is hypothesized that:
H1a: Genre spanning has a negative effect on genre consensus. H1b: This effect is moderated by the artist’s years of experience.
Nowadays, social media may significantly impact a firm’s reputation, sales, and even survival
(Kietzmann, Hermkens, McCarthy, & Silvestre, 2011). In the context of artists, social media
20
relevant indicators of reputation and popularity. Social media facilitate knowledge sharing
(Yates & Paquette, 2011) and ranking lists provide visibility (Carare, 2012; Yoo & Kim,
2012). Thus, it is assumed that if artists appear on ranking lists and their social media profiles
display high audience appeal, this would imply a considerable amount of word-of-mouth
communication and buzz surrounding these artists and their genre profiles (Scott, 2012). An
increase in dialogue and interaction (Cattani et al., 2008) by the right consumers, marketing
strategists, critics and media could facilitate legitimation and justification of the artists’ genre
profiles (Baumann, 2007) and spread status beliefs (Ridgeway & Correll, 2006), which should
lead to a higher genre consensus across audiences. Thus, it is expected that;
21
3.
Method
3.1 Sample
The Electronic Dance Music (EDM) industry is chosen as empirical setting because it
represents a largely neglected creative industry in scientific literature, despite its impressive
growth during the past decade (EVAR, 2012). Disc jockeys (DJs) represent the most
prominent actors in this industry, and the genre labels that are adopted to classify them have a
tremendous influence on the EDM community (McLeod, 2001). It is even claimed that “the
continuous and rapid introduction of new subgenre names into electronic dance music
communities is equalled by no other type of music” (McLeod, 2001, p. 60). All this indicates
the relevance of genre labelling for DJs in the EDM industry.
The sampling frame is derived from the DJ Mag, which is a monthly magazine
dedicated to EDM that has been published since 1991. Its DJ Top 100 list represents the
outcome of the world’s leading DJ poll which attracts over 350,000 votes a year. The poll
asks the audience members to list their five favourite DJs. Thus, since the list is widely
recognized because of its popularity and determined by public vote (EVAR, 2012), it can be
assumed that it accurately displays which DJs are appealing in the eyes of the audience.
This study focuses on the DJs who have been listed in the DJ Mag Top 100 between
2010 and 2014 (appendix A on p. 59-62). Each DJ who appeared on the list at least once was
included, which led to a sample of 187 DJs (appendix B on p. 63-64). The large sample size is
considered to have a positive effect on the generalizability of the study, whereas the clearly
22 3.2 Data Collection
Secondary data were obtained from online DJ community databases, several social media
websites, and commercial music platforms which display the popularity of DJs and genres.
This form of data collection was convenient because the raw data were easily accessible
online and usually complete. Personal accounts were created if these were necessary to get
access. Next, a dataset was built by integrating information taken from the artist pages of all
DJs (N = 187) from 11 online sources which are described in detail below.
3.2.1 DJ Databases
Genre spanning and genre consensus data were derived from four online DJ databases: DJ
Rankings, Partyflock, The DJ List, and Top Deejays. In order to determine the DJs’ years of
experience, data were used from Discogs.
DJ Rankings
DJ Rankings is an online DJ community, established in Japan in 2012. The website provides a
top 10,000 DJ list, which is based on an advanced algorithm that considers DJ fees and
salaries, media presence, chart data from music releases, airplay from radio stations and
followers on large social networks. Their expert jury team makes the final adjustments based
on their perspective on the DJs’ technical skills and craftsmanship (DJ-Rankings, 2015). The list displays the artists’ name, nationality, and associated genres. The DJ rankings can also be selected per country or genre. In addition, the website hosts remix competitions and provides
a DJ swap feature which connects resident DJs around the world.
Partyflock
23
create a personal profile, add dance events to their agenda, create a list of friends, add pictures
and indicate their favourite genres. The website allows to search for a list of upcoming and
past events per artist, venue, organization, or city, and contains numerous forums that
function as a platform for its users to engage in dialogue with each other about music and
events. It also offers interviews with artists, reviews about new music and the possibility to
watch photos and videos taken at events. Partyflock has approximately 200,000 active
members and around 750,000 unique visitors every month (Partyflock, 2014).
The DJ List
The DJ list is an online DJ database and platform based in the United States, dedicated to
promotion and awareness of Electronic Dance Music. It contains over 510,000 DJ profiles
and almost 900,000 registered users. Users can create a personal profile, follow their favourite
DJs, and are informed of EDM news, reviews, interviews, events and contests. The website
also offers the possibility to filter DJs by country or genre, and displays each DJ’s global ranking and global ranking per genre. Furthermore, it recommends 10 related artists per DJ
(TheDJList, 2015).
Top Deejays
Top Deejays is an online DJ database, founded in Slovenia. The website uses an algorithm to
calculate DJs’ social media influence by combining Facebook, Twitter, SoundCloud, MySpace, Last.fm and YouTube fans, subscribers and followers, in order to construct a DJ
ranking list. DJs can be filtered by country, genre or social network. The website also
provides information about new artists and lists of the seven most popular genres and
countries. Visitors can access and create DJ profiles which contain social media statistics,
24
Discogs
Discogs is the largest online user-built database and marketplace for mainly electronic music.
It contains almost 6 million releases, of almost 3.9 million artists, from almost 750,000 music
labels. Discographies can be filtered per genre, style, format (e.g. vinyl, album, or CD), country, or decade. Artists’ profiles contain a biography, an overview of their discography (e.g. albums, singles, compilations and DJ mixes), and links to the marketplace where related
CDs and vinyl can be purchased and sold. The website also offers user-created and managed
groups which discuss music and other related subjects (Discogs, 2015).
3.2.2 Social Media & Ranking Lists
Statistics from social media sources such as Facebook, SoundCloud, Spotify, Partyflock,
Twitter and YouTube, as well as rank data from the DJ Mag were obtained in order to
measure audience appeal.
Facebook is the world’s most popular online social networking service with over a billion of
registered users. The site allows its users to create a personal profile, post information about
themselves, and leave messages on their friends’ profiles (Raacke & Bonds-Raacke, 2008).
Moreover, it is possible to upload pictures and videos, exchange private messages and keep in
touch with friends, family and colleagues. Users can also create groups and events, or build a
public page around a topic of interest.
SoundCloud
SoundCloud is a social music platform which allows users to create, share and discover
25
upload, listen, like and repost songs, follow their favourite artists, create a personal profile
and access their listener statistics. The uploaded audio can easily be shared privately with
friends or publicly with the entire SoundCloud community. To a certain extent, the service
outwits music piracy as the songs can only be downloaded if permission is given by the artist
(SoundCloud, 2014).
Spotify
Spotify is a commercial music streaming service which provides the possibility to listen
millions of tracks; legally and unlimited. Users can browse music by artist, album, genre and
record label. It also includes other features such as creating and sharing playlists, or choosing
a playlist which matches the user’s mood. Consumers can choose to either use the service for
free, or pay a monthly “Premium” subscription which removes all advertisements, and
provides the options to download music and listen offline. Spotify has over 15 million paying
subscribers, over 60 million active users and offers more than 30 million songs (Spotify,
2015).
Twitter is an online social network on which users can register and upload text-based posts of
up to 140 characters (the so-called “tweets”). Each personal profile shows an overview of the
person’s tweets, with the most recent one on top. Personal accounts are characterized by the “@” in front of the name. Tweets can include words preceded by the symbol “#”, this hashtag is used to mark keywords. Clicking on a keyword directs the user to a complete list of all
tweets containing that topic. Users can follow others or be followed themselves, and choose to
either keep their tweets private or make them public. Moreover, Tweets can be retweeted,
meaning that the message is copied onto the user’s personal profile and made available for the user’s followers, which is a very effective resource for electronic word-of-mouth (Jansen,
26
Zhang, Sobel, & Chowdury, 2009). Twitter connects 302 million active users every month
and 500 million Tweets are sent on a daily basis (Twitter, 2015).
YouTube
YouTube is the world’s largest video-sharing website with billions of users. It offers visitors
without an account the chance to watch videos and access video channels. A registered
account allows to create a personal channel, subscribe to favourite channels, and upload, rate
and discuss videos. Videos can be filtered based on categories such as popular, music, sport,
games, films and news. The website also offers personalized recommendations based on the user’s viewing history and is actively used as a distribution platform of original content creators and advertisers (YouTube, 2015).
DJ Mag
DJ Magazine, or DJ Mag in short, is a monthly dance music magazine from the United
Kingdom. The related website offers Electronic Dance Music news, interviews and reviews,
and publishes two Top 100 lists each year; one for DJs and one for clubs, both based on
public votes (DJMag, 2010). Despite the criticism (after all, the list measures to what extent
the DJs appeal to the audience, but not necessarily who the ‘best’ DJ is), the DJ Top 100 list
27 3.3 Genre Classification System Development
The specificity of genre classifications varied across the four sources: (1) DJ Rankings, (2)
Partyflock, (3) The DJ List, and (4) Top Deejays (appendix C on p. 65). As each of the online
databases and platforms was founded in a different country, the variation of genre
classification systems was to be expected because artistic classification systems tend to differ
across societies (DiMaggio, 1987). In order to accurately measure genre spanning and genre
consensus, a genre classification system was developed.
First, it had to be determined how and by whom the classification systems on each
source was constructed, whether the data were up-to-date, and who decided on the DJs’ genre
profiles. Therefore, all platforms were contacted via e-mail or via a question form on the
website itself. A representative of Top Deejays replied and indicated that the genres on their
website are updated over time, and if applicable set according to the DJ’s personal request. In
addition, the DJ profiles are created by audience members but compared to profiles on other
social media websites by members of the Top Deejays team in order to confirm the profiles’ validity.
Despite repeated requests, no responses were received from the other platforms. Thus,
in order to warrant the validity of the final genre classification system, two EDM experts were
consulted concerning its development via telephone. Both experts had more than 10 years of
working experience in the EDM scene, with organizing and promoting events, and also
composing line-ups.
The process of data-cleaning occurred in several steps. First, the genres that clearly
indicated the same but were spelled differently (for instance, ‘psy-trance’ and ‘psychedelic trance’) were aligned. This narrowed down the number of distinct genres to 46. An overview of the names that were changed is shown in table 1 (p. 28).
28
Table 1. Changes in Genre Classification Systems Based on Name Similarity
Source Original classification New classification DJ Rankings Hardcore techno
Indie / Underground
Traditional house
Hardcore / Hard techno
Indie dance House Partyflock Deephouse Electro Hardcore Progressive Techhouse Deep house Electro house
Hardcore / Hard techno
Progressive house
Tech house
The DJ List Indie dance / Nu disco Psy-trance
Indie dance
Psychedelic trance
Top Deejays Breaks Hard techno
Psy-trance
Breakbeat
Hardcore / Hard techno
Psychedelic trance
Next, it was assessed which genres were represented on two or more online sources. The four
sources are all founded in different countries (Japan, the Netherlands, United States and
Slovenia). Thus it was reasoned that if a genre appears on at least two sources, this would
indicate an international consensus on the classification to some extent. As displayed in table
2 (p. 29), only 16 genres appeared on multiple platforms and databases: (1) breakbeat, (2)
deep house, (3) drum & bass, (4) dubstep, (5) electro house, (6) electronica, (7) hard dance,
(8) hardcore / hard techno, (9) house, (10) indie dance, (11 ) minimal, (12) progressive house,
29
Table 2. Genre Representation Across the Four Sources
Genres Number of sources Genres Number of sources
2-step 1 Hardcore / Hard techno 4
Acid 1 Hardhouse 1
Ambient 1 Hardstyle 1
Breakbeat 2 Hardtrance 1
Chill out 1 Hiphop 1
Classics 1 House 4
Club 1 Indie dance 3
Commercial dance 1 Jungle 1
Darkcore 1 Latin 1
Deep house 2 Minimal 4
Dirty house 1 Moombahton 1
Disco 1 Pop 1
Drum & bass 4 Progressive house 4
Dubstep 4 Psychedelic trance 3
Eclectic 1 R&B 1
EDM 1 Raw hardstyle 1
Electro house 4 Soul 1
Electronica 3 Tech house 4
Funk 1 Techno 4
Garage 1 Trance 4
Goa 1 Trap 1
Groove 1 Tribal house 1
30
In collaboration with the two EDM experts, it was assessed whether the remaining 30 genres
could be classified as one of the 16 main genres. A list was made of the subgenres that could
be labelled according to their main genre, which was approved of by the EDM experts. The
eight changes that were made are displayed in table 3 below.
Finally, the genres that were left consisted of: Acid, Ambient, Chill out, Classics,
Club, Commercial dance, Disco, EDM, Eclectic, Funk, Garage, Groove, Hardhouse,
Hardtrance, Hiphop, Jungle, Latin, Pop, R&B, Soul, Trap, and Urban. These 22 genres could
not be classified as one of the 16 main genres with certainty because of three reasons. First, a
few represent an entirely different genre (e.g. Acid and Jungle). Second, a few genres had a
dyadic character. For instance ‘Hardhouse’ shares characteristics with House but also with
Hard dance. Third, a group of genres could not be classified as a genre of Electronic Dance
Music in the first place (e.g. Funk, Hiphop, Latin and Urban). For these 22 genres, dummy
variables were created (see p. 33).
Table 3. Changes in Genre Classification Systems Based on Subgenre Identification
Source Original Classification New Classification
Partyflock 2-step Breakbeat
Darkcore Hardcore / Hard techno
Dirty house House
Goa Psychedelic trance
Hardstyle Hard dance
Moombahton Electro house
Raw hardstyle Hard dance
31 3.4 Variables and Measures
Independent variables
Genre spanning. Genre spanning is calculated by adding up the total number of distinct
genres associated with every DJ across the four online databases; (1) DJ Rankings, (2)
Partyflock, (3) The DJ List, and (4) Top Deejays. For instance, if a DJ is classified as house,
electro house, house, and progressive house, the total genre spanning is assigned a value of 3.
In addition, a total of eight variables were adopted to measure audience appeal.
Average ranking (reversed). This variable is calculated by taking the sum of the DJ Mag
positions between 2010 and 2014, divided by the number of years that the DJ had a position
on the list. Next, based on ordinal ranks Rt = [1, 100], a DJ's inverse listing (101 - Rt)
approximates his yearly popularity (Keuschnigg, 2015). Using this approach, a higher score
indicates higher popularity, which makes it easier to interpret the results. As an example,
Knife Party appeared on the list in 2012, 2013 and 2014. Across these three years, they
obtained position 33, 25, and 53. The sum of these values, divided by three gives an average
position of 37. Lastly, 101 – 37 = 64, which gives the final average ranking value.
Times ranked. This variable measures the number of times a DJ obtained a position on the DJ
Mag Top 100 list across five years (2010 – 2014). It is considered to be an indicator of
audience appeal over time as the list is based on public vote.
Facebook likes. The number of likes on the DJ’s Facebook page.
Partyflock fans. The number of fans on the DJ’s Partyflock artist profile.
SoundCloud followers. The number of followers of the DJ’s SoundCloud profile. Spotify followers. The number of followers of the DJ’s Spotify page.
32 Twitter followers. The number of followers of the DJ’s Twitter account.
YouTube subscribers. The number of subscribers on the DJ’s YouTube channel.
Dependent variable
Genre consensus. Genre consensus measures audiences’ consensus on a DJ’s fit with targeted
genres across the four online databases; (1) DJ Rankings, (2) Partyflock, (3) The DJ List, and
(4) Top Deejays. Following Hsu (2006), the average pairwise similarity was calculated
between each source using Jaccard’s similarity coefficient. This formula has the following
form;
JS = the Jaccard similarity coefficient that lies between JS = 0 (dissensus) and JS= 1
(consensus);
a = the sum of positive genre matches between the two pairs;
b = the sum of genres which was mentioned by the first source, but not by the second; c = the sum of genres which was mentioned by the second source, but not by the first; For instance, if a DJ is classified as trance and hard dance on website 1, but on website 2
categorized as trance, house and electro house, then JS = 0.25.
As the data were obtained from four different sources, the Jaccard coefficients for each of the
six pairwise comparisons were calculated. Next, these six values were added up and then
divided by six in order to find the genre consensus value for that DJ. For example, DJ Avicii JS = a a + b + c JS = 1 1 + 1 + 2
33
is classified as progressive house by DJ Rankings, as house by Partyflock, as house and
progressive house by The DJ List, and finally as progressive house and electro house by Top Deejays. Six pairwise comparisons are made as displayed in table 4 (p. 34). In the case of
missing data, for instance if a DJ was classified on three of the four sources, only three
pairwise comparisons were calculated and the sum was divided by three. Thus, listwise
deletion was not used on purpose, in order to make utmost use of the data.
Moderator
Years of experience. This measure is based on data from the DJ’s profiles on the online music
database Discogs. It is calculated by taking the year of artist entry, meaning the year of the
first publication of recorded music by an artist (Mattsson et al., 2010) and subtracting this
from the current year, 2015. Pinheiro and Dowd (2009) used a similar measure for human
capital; their measure of experience involved the number of years elapsed since each
respondent first began playing musical instruments. In this study, artist entry for each DJ is
based on the year of the first registered release, which can be either a single or an album.
Control Variables
Number of archival sources. Following Hsu (2006), I controlled for the number of archival
sources in which the DJs were classified, because the number of genres under which a DJ is
categorized is likely to increase with the number of different sources in which the DJ is listed.
Unclassified genres. Genre dummy variables were included for the 22 genres that could not
be listed in the genre classification system (see p. 30) in order to control for category effects
(Hsu, Negro, & Perretti, 2012). For each genre, the DJs (N = 187) were assigned a 1 if the
34
Table 4. DJ Avicii Example Jaccard Similarity Coefficient Calculation
Genre Sources Jaccard Similarity Calculation 1. DJ Rankings – Partyflock
= 0 JS =
0 0 + 1 + 1
2. DJ Rankings – The DJ List
= 0.5 JS =
1 1 + 0 + 1
3. DJ Rankings – Top Deejays
= 0.5 JS =
1 1 + 0 + 1
4. Partyflock – The DJ List
= 0.5 JS =
1 1 + 0 + 1
5. Partyflock – Top Deejays
= 0 JS =
0 0 + 1 + 2
6. The DJ List – Top Deejays
= 0.33 JS = 1 1 + 1 + 1 Genre consensus = 0.31 JS = 0 + 0.5 + 0.5 + 0.5 + 0 + 0.33 6
35
4.
Results
4.1 Descriptive Statistics
The sample included 187 DJs from 30 different countries. Only 4 DJ acts consisted of
females, which is not surprising because the EDM scene has always been dominated by men
(McLeod, 2001). Most DJs are from the Netherlands (n = 45), followed by the United
Kingdom (n = 27), the Unites States (n = 23), Germany (n = 13), and Sweden (n = 12), from
which can be derived that 67.2% of the DJs represent one of the top 5 countries. A complete
overview is displayed below in figure 1.
36
Genre Popularity
Table 5 below displays the genre frequencies and percentages per source. It shows that
progressive house is the most popular genre with 263 classifications, followed by trance, electro house, house, and techno. Together, these five most popular genres represent 75.51% of all genre classifications. In addition, breakbeat is clearly the least represented genre in the
sample.
Table 5. Genre Frequencies and Percentages per Source
DJ Rankings Partyflock The DJ List Top Deejays Freq. % Freq. % Freq. % Freq. %
1. Breakbeat 0 0 3 .08 0 0 1 .03
2. Deep house 0 0 3 .08 0 0 2 .06
3. Drum & bass 4 1.7 6 1.6 4 1.6 4 1.1 4. Dubstep 5 2.1 8 2.1 6 2.4 6 1.7 5. Electro house 40 16.8 60 15.6 43 16.9 67 19 6. Electronica 4 1.7 0 0 3 1.2 11 3.1 7. Hard dance 8 3.4 17 4.4 8 3.1 17 4.8 8. Hardcore / Hard techno 8 3.4 2 .05 8 3.1 16 4.5 9. House 17 7.1 82 21.3 40 15.7 39 11.1 10. Indie dance 6 2.5 0 0 1 .04 7 2 11. Minimal 1 .04 6 1.6 1 .04 1 .03 12. Progressive house 52 21.8 51 13.2 60 23.6 100 28.4 13. Psychedelic trance 5 2.1 2 .05 5 2 5 1.4 14. Tech house 9 3.8 7 1.8 9 3.5 8 2.3 15. Techno 11 4.6 28 7.3 11 4.3 9 2.6 16. Trance 57 23.9 54 14 55 21.7 54 15.3 Total 227 95.4 329 83.16 254 100 347 98.6 Others 11 4.6 56 16.84 0 0 5 1.4 Total 238 100 385 100 254 100 352 100
37
Genre Spanning
The genre spanning varied across the four databases. On DJ Rankings (n = 181), the majority
of the DJs (n = 124) were associated with only one genre, followed by a spanning of two
genres for the remaining 57 DJs (M = 1.31, SD = 0.47). The genre spanning on Partyflock (n
= 167) ranged between values 1 to 5 which occurred most in descending order, and displayed
one outlier of 8 (M = 2.3, SD = 1.2). Similar to DJ Rankings, the spanning on The DJ List (n
= 176) and Top Deejays (n = 184) differed merely between one or two genres. On the DJ List,
the divide between a spanning of one (n = 98) or two genres (n = 78) was almost equal (M =
1.44, SD = 0.5). In contrast to DJ Rankings and The DJ List, DJs were related more often to
two genres (n = 168) instead of only one (n = 16) on Top Deejays (M = 1.91, SD = 0.28). In
addition, most DJs were related to a total of two, three or four different genres across the four
sources (Mdn = 3, range = 8), as displayed in figure 2 below.
38
Years of Experience
The DJs (N = 187) differed in their number of years of experience in the music industry from
1 to 30 years (M = 11.06, SD = 6.2), as is shown in figure 3 below. For illustration, years of
experience appeared to be non-normally distributed with skewness of 0.56 (SE = 0.18) and
kurtosis of -0.44 (SE = 0.35). However, tests of normality have little relevance in this study
because “in large samples, they can be significant even for small and unimportant effects”
(Field, 2013, p. 184).
Figure 3. Boxplot: years of experience
Audience Appeal
The DJ’s reversed ranking positions differed between 2 and 99 (Mdn = 38, range = 97). The
number of DJs who shared the same final average reversed ranking value did not exceed the
value of 4. In addition, the DJs varied in their presence on the DJ Mag lists between 2010 and
2014 (M = 2.66, SE = 1.49). 31% had appeared only once, while 22.5% had been listed two
39
Table 6. Times Ranked in DJ Mag Top 100 List 2010 – 2014
Times Ranked Frequency Percent Cumulative Percent
1 58 31 31 2 42 22,5 53,5 3 27 14,4 67,9 4 25 13,4 81,3 5 35 18,7 100 187 100
In general, the DJs had a tremendous amount of likes on their Facebook profiles (Mdn =
680186, range = 55197304). A more detailed look at the data showed that 36.9% of the DJs
scored between 100000 – 500000 likes, and seven DJs even had over ten million likes on their
artist profile.
The number of Partyflock fans was considerably less (Mdn = 472, range = 17755),
which is probably due to the smaller amount of active users. In addition, 30.5% of DJs had
between 101 and 500 fans, and only 3 DJs scored above 10000.
Third, 19 DJs had more than a million followers on SoundCloud (Mdn = 97602.5,
range = 5625012), and 32.6% scored between 100000 and 500000 followers.
Fourth, eight DJs were followed by more than one million individuals on Spotify, and
33.7% had between 10000 – 50000 followers (Mdn = 29916, range = 6615991).
With regard to Twitter, the most popular category of 100000 – 500000 followers was
represented by 36.9% of the DJs. Moreover, there were three DJs with over five million
followers (Mdn = 146121, range = 17150364).
Sixth, the number of YouTube subscribers ranged from 144 to almost ten million
(Mdn = 37489, range = 9858756). In total, 14 DJs had more than one million subscribers on
40
Genre Consensus
Average consensus scores ranged between 0 and 1 (M = .52, SD = .21). Hereby, a value of 0
would imply there is no consensus at all, while a value of 1 would indicate pure consensus.
There was partial consensus on the genre categorization for the majority of DJs. For some,
there was no consensus at all, see figure 4 below.
Figure 4. Boxplot: genre consensus
Table 7. Pairwise Consensus Comparisons Between Sources
Source Comparisons N M SD
1. DJ Rankings – Partyflock 161 .36 .32
2. DJ Rankings – The DJ List 174 .87 .26
3. DJ Rankings – Top Deejays 179 .51 .31
4. Partyflock – The DJ List 159 .42 .31
5. Partyflock – Top Deejays 165 .42 .28
41
Table 7 (p. 40) provides an overview of the six pairwise comparisons that were calculated in
order to measure the average genre consensus between each pair of sources. It appeared that DJ Rankings and The DJ List agreed more on the DJ’s genre profiles than any other
combination of sources (M = .87, SD = .26).
Correlations
To study the antecedents of genre consensus, the correlation matrix is presented (table 8, p.
42). From the correlations it appeared that in general, genre consensus is only significantly
related with the number of times a DJ is ranked and genre spanning. This is in line with
previous research on the positive effects of visibility on popularity (Duan et al., 2008; Scott,
2012) and the negative effects of genre spanning (Hsu, 2006; Negro et al., 2010). However,
contrary to my expectations, genre consensus did not correlate with audience appeal on social
media or years of experience in the music industry. The matrix also shows that this latter
variable correlates with times ranked, from which could be derived that DJs who have more
industry experience appear more often on the DJ Mag Top 100 list.
With regard to the social media platforms, the table displays that the Facebook likes,
SoundCloud followers, Spotify followers, Twitter followers and YouTube subscribers are all
significantly related, and also with average reversed ranking and times ranked. This signals
that the higher and the more often a DJ is ranked, the more popular the DJ is on social media
(and vice versa). However, the number of Partyflock fans did not correlate with audience
appeal measures of other social media platforms. This is presumably due to country bias,
because Partyflock is mostly used by Dutch people. Another interesting fact is the significant
negative correlation between years of experience in the music industry and SoundCloud
followers. Perhaps SoundCloud is more actively used by well-established artists than
42 Table 8. Mean, Standard Deviation and Correlations of Study Variables
M SD 1 2 3 4 5 6 7 8 9 10 1. Average ranking (reversed) 41.65 24.7 2. Times ranked 2.66 1.5 .627** 3. Facebook likes 1929341.3 4942898.8 .457** .312** 4. Partyflock fans 1495.5 2541.5 .179** .294** .057 5. SoundCloud followers 483307.6 1094067 .373** .212** .317** -.077 6. Spotify followers 199048.5 681265.1 .399** .295** .927** .034 .380** 7. Twitter followers 545681.1 1657972.1 .396** .302** .802** .024 .416** .738** 8. YouTube subscribers 306994 1108186.6 .390* .254** .821** .039 .448** .9** .616** 9. Years of experience 11.06 6.2 .033 .222** .104 .082 -.149* .082 .115 -.021 10. Genre spanning 3.2 1.2 .043 -.028 .089 -.105 -.009 .067 .040 .052 .120 11. Genre consensus .52 .21 .061 .213** -.101 .049 .039 -.079 -.079 -.075 .115 -.527**
**. Correlation is significant at the 0.01 level (2-tailed). *. Correlation is significant at the 0.05 level (2-tailed).
43 4.2 Regression Analyses
4.2.1 Genre Spanning and Genre Consensus
First, it was hypothesized that genre spanning had a negative effect on genre consensus. As
expected, the linear regression analysis (see table 9 below) showed a significant negative
effect between genre spanning and genre consensus, β = -0.53, t(183) = -8.42, p < 0.01. In
addition, genre spanning explained 28% of the variance in genre consensus scores R2 = 0.28,
F(1, 184) = 70.87, p < 0.01. Therefore, I found support for H1a.
Table 9. Linear Regression Analysis Genre Spanning and Genre Consensus
Second, it was predicted that the effect between genre spanning and genre consensus would
be moderated by the number of years of experience in the music industry. In order to
calculate moderation in SPSS, the process tool was installed (Hayes, 2013). A moderation
analysis was done which included genre spanning as predictor, genre consensus at outcome
variable and years of experience in the music industry as moderator, F(3, 182) = 19.57, p <
0.001. As displayed in table 10 (p. 44), the interaction between genre spanning and years of
experience was not significant, which suggests that H1b is not supported.
Variable N β R2 F
44 Table 10. Moderation Model of Predictors of Genre Consensus
(95% bias corrected and accelerated confidence intervals reported in parentheses)
Note. R2 = .31
4.2.2 Audience Appeal and Genre Consensus
The correlation matrix (table 8, p. 42) showed that of all measures of audience appeal, only
times ranked was significantly correlated with genre consensus (r = .213, p < 0.01).
A linear regression was done to calculate whether times ranked had a positive effect on genre
consensus (see table 11 below). A significant effect was found between the number of times a
DJ appeared on the DJ Mag Top 100 list and the consensus of the DJ’s genre profile, β = 0.21,
t(183) = 2.96 , p < 0.01. A small portion of variance in genre consensus scores was explained, R2 = 0.05, F(1, 184) = 8.74, p < 0.01. Thus, H2 is partially supported.
Table 11. Linear Regression Analysis Times Ranked and Genre Consensus
Variables b SE B t p
Constant .59 (.49, .54) 0.01 41.56 < 0.01
Genre spanning -.09 (-.12, -.07) 0.14 -6.78 < 0.01
Years of experience .01 (.001, .01) 0.002 2.64 < 0.01
Genre spanning x years of
experience -.002 (-.01, .002) 0.002 -1.03 0.33
Variable N β R2 F
45 4.3 Robustness Checks
Several additional checks were performed in order to test the robustness of the results. First, I
controlled for the number of archival sources. The DJs who had profiles on all four sources (n
= 156) formed the baseline group. 23% of the DJs was mentioned by three sources, followed
by 7 % who were listed on two sources. Only one DJ appeared on merely one source, and was
therefore automatically excluded from the analysis by SPSS. Table 12 below displays that for
both genre spanning (β = -.2) and times ranked (β = -.03), the change in genre consensus
scores goes down as a DJ changes from being included on four sources to only two sources.
Table 12. Linear Model of the Predictors of Change in Genre Consensus Scores
Second, as described before (p. 30), dummy variables had been created for each of the 22
presumably irrelevant genres that appeared in only one of the four original genre
categorization systems (appendix C, p. 65) and could not be clearly classified. Still, regression
analyses were executed in order to confirm the absence of any categorical effect. The results
showed that the inclusion of the dummies did not drastically change the significant effects of
genre spanning (β = -.59, p < 0.01) and times ranked (β = 0.15, p < 0.01) on genre consensus,
which confirms the robustness of the previous results.
Variables Genre spanning Times ranked
b SE B β p b SE B β p Constant -.89 .04 <.01 .47 .03 <.01 4 archival sources vs. 3 archival sources -.03 .04 -.2 .33 -.02 .05 -.03 .33 4 archival sources vs. 2 archival sources -.27 .06 -.34 < .01 -.26 .08 -.23 <.01