Content uploaded by Cameron DeHart
Author content
All content in this area was uploaded by Cameron DeHart on Sep 24, 2020
Content may be subject to copyright.
MMS 4 (3) pp. 559–571 Intellect Limited 2018
Metal Music Studies
Volume 4 Number 3
© 2018 Intellect Ltd Article. English language. doi: 10.1386/mms.4.3.559_1
www.intellectbooks.com 559
ABSTRACT
Conventional wisdom suggests, and scholarship confirms, that the distribution of
heavy metal music across the world is uneven. Previous studies show there are
more metal bands per capita in Europe and North America than in other regions,
but it is not clear what country-level factors explain that distribution. Drawing
on data from the Encyclopaedia Metallum, I replicate a 2014 study and find weak
support to connect heavy metal and religion, legal history and other social factors.
In this article, I present an alternative model to explain the distribution of metal
bands and show that wealth and political freedom are highly predictive of metal
music, not only across the world, but also within regions.
Introduction
This article revisits the analysis of heavy metal production in volume 1, number
1 of this journal. Conventional wisdom suggests, and research confirms, that
the distribution of metal bands across countries is not equal: the genre was
introduced in Europe and North America before gaining popularity in Latin
America, Asia and, to a lesser extent, the Middle East and Africa. Previous
researchers have grappled with this distribution and the differing experiences
KEYWORDS
statistics
metal bands
replication
economics
metal politics
quantitative
CAMERON DEHART
Stanford University
Metal by numbers: Revisiting
the uneven distribution of
heavy metal music
mms
Metal Music Studies
Intellect
10.1386/mms.4.3.559_1
4
3
559
571
© 2018 Intellect Ltd
2018
ARTICLES
Cameron DeHart
560 Metal Music Studies
of metal fans and artists in non-Western settings (Harris 2000; Hsiang and de
Seta 2017; Wallach et al. 2011; Weinstein 2011a). Maguire (2014) contributes to
this growing literature with the field’s first regression analysis of metal around
the world.
Drawing on data from Encyclopaedia Metallum, known as the Metal-
Archives, Maguire explores the distribution of metal bands per capita across
countries with a series of models based on 55 variables. After testing several
permutations of the data with a technique known as stepwise regression,
Maguire finds that seven variables are correlated with metal:‘Scandinavian
Legal History’, latitude, the number of years under Marxist rule, the size of
the youth male population, concert halls per capita and the per cent of people
who are Catholic or non-religious.
In this article, I replicate Maguire’s (2014) analysis and show the results are
not robust to out-of-sample testing with conventional methods. I develop an
alternative model and show that nearly 80 per cent of the variation in metal
across the world can be explained by just four variables: income per capita,
level of democracy, region of the world and a binary indicator for Nordic
countries. The results are more consistent with the conventional wisdom:
metal bands are concentrated in wealthy and democratic countries in the West
where the genre originated. The advantage of this model is its simplicity. In
contrast to Maguire’s focus on the social and cultural correlates of metal, my
model performs well because it focuses on the macro-level factors that are
associated with a country’s capacity to record and consume music of any kind.
This article proceeds as follows: I discuss the research on the distribu-
tion of metal and summarize the original article. Second, I present the data,
describe the dependent variable and replicate the findings. After a discussion
of the alternative model and results, the article concludes with a comment on
the replication process in Metal Music Studies.
Background
Previous researchers have explored the global distribution of metal, including
data journalists in the United States and United Kingdom. Florida and Mellander
(2014) show a country’s metal bands per capita is correlated with economic
output, creativity and entrepreneurship (cf. Florida 2012).1 They conclude,
Though metal may be the music of choice for some alienated working-
class males, it enjoys its greatest popularity in the most advanced, most
tolerant, and knowledge-based places in the world. Strange as it may
seem, heavy metal springs not from the poisoned slag of alienation and
despair but the loamy soil of post-industrial prosperity.
(Florida and Mellander 2014)
Maguire’s analysis improves on these studies by employing multivariate
regression to look for relationships between several factors at once.
Scholars have attempted to identify the social and cultural conditions that
foster the genre. Although the original study refers to common assumptions
about the subject, the analysis does not rest on firm theoretical ground. The
process for selecting variables in the first study relied too heavily on outlier
cases, especially Scandinavia, that possess far more metal bands per capita
than their peers elsewhere. Metal music scholars have historically focused on
the conditions that gave rise to the genre initially, but the project at hand is
1. Media reports on the
distribution of metal
bands include The
Atlantic (Grandoni
2012), Foreign Policy
(Keating 2012), The
Guardian (30 May
2014), Invisible Oranges
(8 October 2015)
and The Daily Mail
(Zolfagharifard 2014).
Metal by numbers
www.intellectbooks.com 561
slightly different. We want to shed light on the contours of metal’s distribution
across the world, and that requires broadening our scope beyond the outliers.
The Metal-Archives data tell a familiar story: metal is concentrated in
the West, with some variation across other regions of the world. Tables 1 and
2 show the top and bottom 25 countries according to the number of metal
bands per capita. Only two of the top 25 countries are non-European, Chile
and Canada, and all the Nordic countries are in the top fifteen. Just five of the
top 25 metal-per-capita producers are in the top ten countries as ranked by
total metal bands: Germany, Italy, Sweden, Finland and Canada. Among the
Country or region
Bands per
100,000 people
Rank by number
of metal bands
Svalbard† 152 117th
Finland† 62 8th
Sweden† 43 7th
Iceland† 31 63rd
Norway† 30 20th
Liechtenstein 30 96th
Faroe Islands† 24 95th
Monaco 20 104th
Greece 16 16th
Luxembourg 15 66th
Denmark† 15 29th
Malta 14 70th
Estonia 14 51st
Gibraltar 14 114th
The Netherlands 12 14th
Austria 12 27th
Czech Republic 12 21st
Slovenia 12 45th
Germany 12 2nd
Switzerland 11 28th
Portugal 10 23rd
Belgium 10 25th
Hungary 10 26th
Chile 9 19th
Canada 9 9th
Italy 9 3rd
Note: The symbol † denotes Nordic countries and regions. Data on the number of
metal bands were provided by Metal-Archives in May 2015. Population data from
the World Bank. The values for bands per capita were rounded to the closest whole
number for clarity.
Table 1: Top 25 countries by metal bands per capita.
Cameron DeHart
562 Metal Music Studies
top 25 overall metal-producing countries, just nine are located outside Europe
(see Table 6 in the Appendix). In contrast, there are no European countries
among the bottom 25 metal producers per capita, which are concentrated in
Africa, Asia, the Caribbean, the Middle East and North Africa. These patterns
suggest we should be cautious about a research strategy that over-emphasizes
Western factors that are commonly associated with metal.
The proliferation of metal in the Nordic countries is an interesting pattern
that scholars should continue to interrogate, but there is a limit to what we
can learn about the distribution of metal in the rest of the world if our models
are too narrowly focused on those exceptional cases. The first study found that
the number of bands per capita is linked to‘Scandinavian Legal History’, but
Country or region
Number of
metal bands
Bands per
100,000 people
Ethiopia 1 0.00
Afghanistan 1 0.00
Myanmar 2 0.00
Mozambique 1 0.00
Uganda 2 0.01
Zambia 1 0.01
Kenya 3 0.01
Angola 2 0.01
India 153 0.01
Cambodia 2 0.01
China 234 0.02
Uzbekistan 6 0.02
Pakistan 49 0.03
Bangladesh 46 0.03
Libya 2 0.03
Egypt 26 0.03
Jamaica 1 0.03
Iraq 11 0.04
Madagascar 9 0.04
Vietnam 40 0.04
Laos 3 0.05
Saudi Arabia 13 0.05
Nepal 17 0.06
Turkmenistan 3 0.06
Azerbaijan 6 0.06
Note: Countries with zero metal bands are excluded. Data on the number of metal
bands were provided by Metal-Archives in May 2015. Population data from the World
Bank. The values for bands per capita were rounded to the nearest hundredth.
Table 2: Bottom 25 countries by metal bands per capita.
Metal by numbers
www.intellectbooks.com 563
2. The author offered to
clarify the instructions
for compiling the data,
but declined to share
the original dataset.
3. Encyclopaedia
Metallum is
maintained by a team
of moderators with
discretion over the
admission process, and
the true count of metal
bands per country is
no doubt larger than
reported. Maguire
(2014) notes that bands
in the Japanese genre
Visual Kei are not listed
in the archives. Other
noteworthy genres
that are absent include
nu-metal (Slipknot,
Linkin Park), glam rock
and hair metal, core
bands, and metal-
adjacent genres like
gothic rock (HIM, The
69 Eyes) and industrial
rock (Rammstein).
that result is an artefact of these cases skewing the analysis. Similarly, it is
neither surprising nor helpful to learn that latitude is correlated with metal. By
selecting variables based on the top 1 per cent of cases, the first study revealed
little about the distribution of metal bands within Africa, the Caribbean or the
Middle East, where countries have lower latitudes and no Scandinavian legal
history, but also differ from Europe and North America in many other ways.
Without a stronger theoretical foundation for the link between religion
and metal, it was inappropriate to include these variables in the original study.
In particular, the cross-sectional nature of the data does not allow us to specu-
late about how people behave over time. To be sure, the first study could not
test the claim that Catholic metalheads leave their religion at lower rates than
Protestant metalheads (Maguire 2014: 163). The per cent of people who have
no religion is a plausible correlate of metal if secular societies tolerate more
extreme forms of music, but that claim was not tested before the variable was
included in the model. Other factors, such as liberal democracy, are also asso-
ciated with tolerance for extreme views even in countries where rates of secu-
larism are low. Both happen to be concentrated in the West.
Of the seven variables, the number of concert halls per capita is the most
plausible explanatory variable, but it is unclear to what extent the venues
counted in the dataset, such as Carnegie Hall and the Royal Academy of
Music, are ever used by metal bands. Given that metal music is often labelled
extreme and underground (Kahn-Harris 2006), we might even expect the
number of concerts halls to be unrelated to the production of metal music if
artists are eschewing mainstream venues for unconventional ones. Either way,
the capacity to build infrastructure for musical performances is correlated with
other factors, like wealth, that are more plausible.
Data
I followed the instructions in the original article and compiled data from over
50 sources. It was difficult to know if the new dataset matched the original
without access to the author’s data.2 As will be discussed later, I was able
to approximately replicate the results from the specifications favoured by
Maguire (2014), models 36 and 44. After replicating the results, I added the
new data for metal bands per country (Table 7 in the Appendix shows the
summary statistics). As Maguire points out, we should consider the limitations
in the Metal-Archives data.3 In this section, I focus on three limitations of the
dataset: the cumulative nature of the dependent variable, the unit of analysis
and the absence of countries with zero bands.
The dataset counts all bands a country has produced without regard for
how many of those bands are currently active. This is a problem because not
all countries produced bands during the full range covered in the data set,
1964–2015, and a country’s measures of metal bands per capita may differ
from year to year. To illustrate this limitation, consider two hypothetical cases:
Country A has 100 bands that have existed for 30 years and have each released
one album a year, totalling 3000 albums. Country B also has 100 bands but
each band existed for one year and produced one album each, totaling 100
albums. The dataset would measure A and B the same, although we might
think their metal scenes are qualitatively different.
It is also important to note that the unit of analysis, country, is not clearly
defined in the original data. The dataset includes polities that could be coded
as part of another country including Guam, Puerto Rico, Svalbard, Gibraltar,
Cameron DeHart
564 Metal Music Studies
the Faroe islands and Greenland. The Metal-Archives dataset also includes
semi-autonomous polities like the Isle of Man and Åland Islands, which are
dependent upon Great Britain and Finland, respectively, but are not consid-
ered‘countries’. This article presents the dependent variable as-is to avoid the
debate for now, but it is worth noting that coding the data differently could
change the results: three semi-autonomous regions, three European microstates
and Luxembourg are among the top 25 metal-per-capita producing countries.
In addition to updating the count of metal bands, I have expanded the
dataset to account for polities that do not have any bands. At least 67 of 203
countries/polities were not included in the original dataset because they did
not have any bands. Sixty-one of those countries had zero bands as of 2017,
and six were found to have at least one: Mauritius, Zambia, Afghanistan,
Ethiopia, Cambodia and Trinidad and Tobago. These cases take on the mini-
mum value of our variable of interest, and it is important to include them in
the statistical analysis in order to properly account for variation at the low end
of the distribution.
Analysis
I employ the same cross-sectional linear regression model in the original arti-
cle, where Y is the number of metal bands per capita, β0 is the intercept term,
βi is the coefficient on independent variable Xi for i number of variables, and
εi is the error term.
Yi=β0+β1∗X1+···+βi∗Xi+∈i
The technique employed in the first study is known as stepwise regression
and involves testing several model specifications by systematically adding
and removing independent variables and comparing the goodness of fit. We
should interpret these results carefully, and this strategy is sometimes derided
as data mining or star seeking. Stepwise regression at this stage should be
considered exploratory, and we should avoid making claims about what the
data can tell us about causality (see Harrell 2015).
One reason we should be cautious about stepwise regression is that
models tend to perform poorly out of sample. If the data for any of the vari-
ables were to change, the results may not be robust. This is true of any data
analysis, but we should be especially concerned here about overfitting: if we
pick the best fitting model based on one set of data, we risk losing the abil-
ity to describe data outside of the set. The stepwise method does not perform
well when the model is applied to data outside the sample. That is to say, the
six variables that appeared significant in regression #36 with the 163 observa-
tions may be null when we include the other thirteen countries in the original
dataset and the 27 additional countries I identified as missing.
Setting aside our concerns about the validity of stepwise regression, the
analysis as described was not carried out systematically enough to say that
all‘significant’ variables were found. The protocol calls for adding and removing
variables that return a p value of 0.05 or less in an iterative series of models that
can include up to fifteen variables. There are nearly 12 trillion possible combina-
tions of fifteen variables from a pool of 55 variables, but fewer than 100 models
were estimated in the original study. Perhaps one of the variables that was
dropped early in the analysis because it did not achieve a p value of 0.05 in the
first set of specifications could have achieved significance in one of the several
trillion specifications that were not run. The second round of eliminations based
on a p value of 0.03 was also unconventional and arbitrarily applied.
Metal by numbers
www.intellectbooks.com 565
Another way to compare models is the R2 statistic, a measure of how much
of the total variance in the dependent variable is explained by the model. The
stepwise regression technique has a major pitfall, especially with model speci-
fications with up to fifteen variables: the inclusion of more variables tends to
bias the R2 statistic upward. Models that are not parsimonious can achieve
high R2 values by chance, in which case the adjusted R2 metric may be more
appropriate.
Tables 3 and 4 show the replicated models 36 and 44, respectively. The
results are approximately the same given slight differences in how the data
Dependent variable
Bands per
100,000 people
Bands per
100,000 people
2013 2015
(1) (2) (3)
Scandinavian legal history 24.64*** 28.49*** 27.44***
(2.25) (2.01) (2.10)
Catholic, per cent of pop. 0.04*** 0.04*** 0.03**
(0.01) (0.01) (0.01)
Concert halls per 100,000 people 0.01 0.03 0.03
(0.02) (0.02) (0.02)
Latitude (absolute value) 0.06*0.08*** 0.01
(0.03) (0.02) (0.03)
No religion, per cent of pop. 0.05 0.04 0.07**
(0.04) (0.03) (0.03)
Years of Marxist rule −0.03 −0.03 −0.01
(0.02) (0.02) (0.02)
Male youths, per cent of pop. −0.86*** −0.86*** −0.28
(0.27) (0.23) (0.29)
GDP per capita (avg. 2008–13) 0.00
(0.00)
Democracy 0.22
(0.21)
Constant 6.73** 6.42** 0.97
(3.21) (2.51) (3.34)
Region fixed effects ✓
Observations 126 191 159
R20.71 0.75 0.81
Adjusted R20.70 0.74 0.79
Note: Coefficients in bold. Standard errors in parenthesis.
*p<0.1; **p<0.05; ***p<0.01.
Table 3: Replication of model 36.
Cameron DeHart
566 Metal Music Studies
were compiled: the original model 36 explained approximately 80% of the
variation in the dependent variable, and the replication model explains about
75% (according to the R2 statistic). Model 44 explained approximately 72% of
variation and the replication model explains between 42 and 47%. Most of the
models in the first study explain between 70 and 80% of the variance, but it
is not clear that any particular model performs better than any other without
theory to guide us.
Alternative model
Next I test an alternative model of the global distribution of metal. For variable
selection, I rely on Florida (2012) and Florida and Mellander (2014) who find
Dependent variable
Bands per
100,000 people
Bands per
100,000
2013 2015
(1) (2) (3)
Catholic, per cent of pop. 0.02 0.02 −0.03
(0.02) (0.01) (0.02)
Concert halls per 100,000 people 0.11*** 0.13*** 0.10***
(0.03) (0.03) (0.04)
Latitude 0.16*** 0.16*** 0.12**
(0.04) (0.03) (0.05)
No religion, per cent of pop. −0.03 −0.04 −0.05
(0.05) (0.04) (0.05)
Years of Marxist rule −0.06*−0.06** −0.02
(0.03) (0.03) (0.03)
Male youths, per cent of pop. −0.58 −0.62*0.44
(0.39) (0.32) (0.42)
GDP per capita (avg. 2008–13) 0.00***
(0.00)
Democracy 0.59*
(0.30)
Constant 2.11 3.12 −8.06*
(4.50) (3.61) (4.84)
Region fixed effects ✓
Observations 126 191 159
R20.42 0.47 0.58
Adjusted R20.39 0.45 0.53
Note: Coefficients in bold. Standard errors in parenthesis.
*p<0.1; **p<0.05; ***p<0.01.
Table 4: Replication of model 44, excluding the Nordic cases.
Metal by numbers
www.intellectbooks.com 567
that metal bands per capita are positively correlated with economic condi-
tions. Table 5 shows the results of the model with regional fixed effects and
just three variables: income per capita, democracy and an indicator variable
for Nordic countries. Income is measured by the average annual GDP per
capita averaged over 2008–13, and democracy is measured from 1 to 10 (The
Economist 2013). Model (1) uses the 2013 measure of metal bands per capita,
model (2) uses the 2015 measure, model (3) excludes the countries with zero
bands and model (4) excludes the Nordic countries.
The alternative model performs as well as the published result accord-
ing to R2, explaining 73 to 79 per cent of variation, but it has several advan-
tages. The model is more parsimonious, my dataset includes more countries
including the zero band cases and I account for the Nordic outliers without
dropping them from the analysis altogether. Regional fixed effects also allow
us to account for the difference in metal bands per capita due to differences
between regions of the world. When we include regional fixed effects, rather
than latitude, the coefficients for other variables can be interpreted as the vari-
ation that occurs within regions.4
Economic and political conditions that relate to a country’s capacity to
record and consume music appear to be better predictors of metal bands per
capita than social factors like religion, young males and legal history. These
results are more consistent with Florida and Mellander (2014) than Maguire
(2014).
[While] new musical forms may spring from disadvantaged, disgruntled,
or marginalized groups, it is the most advanced and wealthy societies
4. The regions are
Africa, Asia, Europe,
the Caribbean,
Central America,
South America, North
America, Middle East
and North Africa, and
Oceania.
Dependent variable
Bands
per 100K
Updated bands
per 100K
(1) (2) (3) (4)
GDP per capita 0.00** 0.00*** 0.00** 0.00***
(0.00) (0.00) (0.00) (0.00)
Democracy 0.56** 0.46** 0.66** 0.44***
(0.24) (0.19) (0.27) (0.11)
Nordic country 21.35*** 25.73*** 25.58***
(1.92) (1.85) (2.15)
Constant −3.04*−2.09** −3.94*−2.02***
(1.82) (0.96) (2.04) (0.55)
Region fixed effects ✓ ✓ ✓ ✓
Observations 118 163 122 158
R20.77 0.79 0.78 0.73
Adjusted R20.74 0.77 0.76 0.71
Note: Model (3) excludes countries with zero bands. Model (4) excludes the Nordic
countries.
*p<0.1; **p<0.05; ***p<0.01.
Table 5: Alternative models.
Cameron DeHart
568 Metal Music Studies
that have the media and entertainment companies that can propagate
new sounds and genres, as well as the affluent young consumers with
plenty of leisure time who can buy it.
(Florida and Mellander 2014)
Conclusion
A number of scholars have expressed doubts about whether the field of metal
studies can build a space where researchers revisit and expand upon each
other’s work (Kahn-Harris 2011). The trajectory of metal studies, thus far,‘has
not usually been the result of a concern with building upon previous academic
work […] the literature that metal studies proposes to gather up and extend
does not display a line of development towards a theoretical synthesis or the
systematic compounding of mutually criticized and systematically interrelated
research studies’ (Weinstein 2011b: 244).
This article looks back with a critical eye. Conventional methods with
updated data give us results that are consistent with what most of us
know intuitively: metal flourishes in wealthy and politically open countries.
Additional research is needed to confirm and expand these results, and both
quantitative and qualitative researchers should continue exploring the nuances
of the global distribution of metal. To that end, I will provide the data and
replication file for this article (available on IngentaConnect) so that readers
may corroborate and extend this analysis. I also encourage the field of metal
studies to adopt this practice for quantitative studies in the future.
Appendix
Country or region
Bands per
100,000 people
Rank by bands
per capita
United States 7 33rd
Germany 12 19th
Italy 9 26th
Brazil 2 57th
France 7 4th
United Kingdom 6 35th
Sweden† 43 3rd
Finland† 62 2nd
Canada 9 25th
Poland 8 30th
Russia 2 58th
Spain 6 37th
Mexico 2 65th
The Netherlands 12 15th
Australia 9 27th
Table 6 (Continued)
Metal by numbers
www.intellectbooks.com 569
Country or region
Bands per
100,000 people
Rank by bands
per capita
Greece 16 9th
Argentina 4 48th
Japan 1 73rd
Chile 9 24td
Norway† 31 5th
Czech Republic 12 17th
Colombia 3 55th
Portugal 10 21st
Indonesia 0.5 82nd
Belgium 10 22nd
Note: The symbol † denotes Nordic countries and regions. Data on the number
of metal bands were provided by Metal-Archives in May 2015. Population data
from the World Bank. The values for bands per capita were rounded to the
closest whole number for clarity.
Table 6: Top 25 countries by total metal bands.
Statistic NMean St. Dev. Min Max
Metal bands (2013) 135 616.61 1811.73 1 17,25
Metal bands (2013) per
100,000 people
135 4.41 8.37 0.00 53.20
Metal bands (2015) 203 504.05 1842.29 0 21,43
Metal bands (2015) per
100,000 people
203 4.10 12.88 0.00 152.28
Young men as per cent
of total population
201 8.81 1.765 4.70 14.71
Years of socialism 202 8.19 17.29 0 69
Latitude 201 26.02 17.69 0 78
Democracy (1–10) 167 5.52 2.19 1.08 9.93
Corruption (2012) 177 42.67 20.10 0.00 90.00
Protestant (per cent) 192 14.70 20.24 0.00 96.00
Catholic (per cent) 192 27.92 30.67 0.00 96.00
Orthodox (per cent) 192 6.87 20.06 0.00 94.70
Judaism (per cent) 192 0.44 5.28 0.00 73.10
Islam (per cent) 192 25.00 36.15 0.00 99.56
No religion (per cent) 192 7.23 11.63 0.00 75.75
Concert halls per capita 203 10.931 17.98 1 64
Annual GDP per capita
(avg 2008–13)
192 13,784.56 21,595.56 202.21 165,164.70
Table 7: Summary statistics for key variables.
Cameron DeHart
570 Metal Music Studies
REFERENCES
Anon. (2013), ‘Democracy index 2012: Democracy at a standstill’, The
Economist: Intelligence Unit, www.eiu.com/democracy2012. Accessed 30
June 2016.
——— (2014),‘Metal doesn’t soundtrack alienation – it’s the music of pros-
perity’, The Guardian, 30 May, https://www.theguardian.com/music/2014/
may/30/heavy-metal-music-of-prosperity/. Accessed 29 June 2016.
——— (2015),‘7 reasons it feels like every metal band is from Scandinavia’,
Invisible Oranges, 8 October, http://www.invisibleoranges.com/7-reasons-
scandinavia/. Accessed 29 June 2016.
Florida, Richard (2012),‘Where the heavy metal bands are’, CityLab, 9 April,
www.citylab.com/design/2012/04/where-heavy-metal-bands-are/1714/.
Accessed 30 May 2016.
Florida, Richard and Mellander, Charlotta (2014),‘How heavy metal tracks
the wealth of nations’, CityLab, 26 May, www.citylab.com/equity/2014/05/
how-heavy-metal-tracks-the-wealth-of-nations/371473/. Accessed 30
May 2016.
Grandoni, Dino (2012),‘A world map of metal bands per capita’, The Atlantic, 29
March, https://www.theatlantic.com/entertainment/archive/2012/03/world-
map-metal-band-population-density/329913/. Accessed 30 May 2016.
Harrell, Frank (2015), Regression Modeling Strategies: With Applications to
Linear Models, Logistic and Ordinal Regression, and Survival Analysis, Cham:
Springer.
Harris, Keith (2000),‘“Roots”?: The relationship between the global and the
local within the extreme metal scene’, Popular Music, 19:1, pp. 13–30.
Harrison, LeighMichael (2010),‘Factory music: How the industrial geogra-
phy and working-class environment of post-war Birmingham fostered the
birth of heavy metal’, Journal of Social History, 44:1, pp. 145–58.
Hsiang, Chiu and de Sata, Gabriele (2017),‘Headbanging in Taiwan’, Metal
Music Studies, 3:2, pp. 335–44.
Kahn-Harris, Keith (2006), Extreme Metal: Music and Culture on the Edge,
Oxford: Berg.
——— (2011),‘Metal studies: Intellectual fragmentation or organic intellec-
tualism?’, Journal for Cultural Research, 15:3, pp. 251–53.
Keating, Joshua (2012),‘Map: Finnish metal is the densest of all metal’, Foreign
Policy, 30 March, http://foreignpolicy.com/2012/03/30/map-finnish-metal-
is-the-densest-of-all-metal/. Accessed 30 May 2016.
Maguire, Donald (2014), ‘Determinants of the production of heavy metal
music’, Metal Music Studies, 1:1, pp. 155–69.
Wallach, Jeremy, Berger, Harris M. and Greene, Paul D. (eds) (2011), Metal
Rules the Globe: Heavy Metal Music around the World, Durham: Duke
University Press.
Weinstein, Deena (2011a),‘The globalization of metal’, in J. Wallach, H. M.
Berger and P.D. Greene (eds), Metal Rules the Globe: Heavy Metal Music
Around the World, Durham: Duke University Press, pp. 34–59.
——— (2011b),‘How is metal studies possible?’, Journal for Cultural Research,
15:3, pp. 243–45.
Zolfagharifard, Ellie (2014),‘The sign of a wealthy country? Heavy metal:
Music is a surprisingly accurate indicator of a nation’s prosperity’, The Daily
Mail, 27 May, http://www.dailymail.co.uk/sciencetech/article-2640661/
The-sign-wealthy-country-HEAVY-METAL-Music-surprisingly-accurate-
indicator-nations-prosperity.html/. Accessed 30 May 2016.
Metal by numbers
www.intellectbooks.com 571
SUGGESTED CITATION
DeHart, C. (2018), ‘Metal by numbers: Revisiting the uneven distribution of
heavy metal music’, Metal Music Studies, 4:3, pp. 559–571, doi: 10.1386/
mms.4.3.559_1
CONTRIBUTOR DETAILS
Cameron DeHart is a Ph.D. candidate in political science at Stanford
University. His research focuses on American political development, state and
local politics, public policy, and institutions.
Contact: Stanford University, 616 Serra Street, Stanford, CA 94305, USA.
E-mail: cdehart@stanford.edu
Cameron DeHart has asserted his right under the Copyright, Designs and
Patents Act, 1988, to be identified as the author of this work in the format that
was submitted to Intellect Ltd.
