« Duffy: How Not to Experiment in Virtual Worlds | Main | You've Been Blerped! Now What Do You Do? »

Jul 23, 2009



>Latinos, who play more intensely than whites, are 12% of players, but only 2% of characters.

We get pretty well the same games in the UK that you get in the USA, and this figure of 2% Latino characters corresponds with our roughly 0% Latino population.

One country's under-representation can be another country's under-representation.

Do WoW dwarf characters count as Scots? They have Scottish accents...



First off, on the whole this is awesome work. One point that you mention at the end (around WoW) involves player created avatars. I'd be curious to know how playership of games in which avatars are customizable stacks up against playership of games where you're handed a player of a specific gender/age/ethnicity. Clearly there's more work to be done here, but in the mean time, did I mention how awesome this is?


err character


I would like to follow up on Richard's comment:

In Europe, the underrepresented minority is unambiguously those of middle Eastern descent, but this doesn't exist in your survey.

Many of the games you discuss are developed elsewhere than the US (say, Japan).

Well, that's the question: How would you situate the study in a global context?


Another thing that might be worth looking at is the breakdown of the characters used for advertising purposes and comparing that with the breakdown for the game itself. This could be interesting because it impinges on the conclusions that people might draw.

For example, MMOs are regularly advertised using female characters, but inside the game female NPCs are almost invariably outnumbered by male NPCs. Why would that be? If female avatars are used to draw people into the MMO, why would you not also populate the MMO with female avatars? Why does are bait and the reality at odds?



@Jesper. We were looking only at US distribution and the US population, and so we used the US Census data as our comparative baseline. To do this globally, you'd need the equivalents: Global sales data and global population data. Given that markets, populations and norms vary so widely, I'm not sure that a clear picture would come out of it. It might be better in the global case to limit the baseline group to the playing population rather than the general one. In the US, the penetration rates are so high that the players and the general population are now similar, whereas that might not be the case in other nations. So, perhaps localized studies make more sense. As always, the challenge is getting both sales and population data.

You raise an interesting point about where the games are made and by whom. If you go to nearly the end of the paper, I have profile data on the developers. That population is much closer to the character population than the players or the general public. For example, females are 38% of players and 51% of the general population, yet are only 10% of playable characters. What causes that large discrepancy? Our study doesn't prove anything, but I think it's quite telling that females are 11% of the game industry workforce.

@Richard. I think you're talking about marketing here, and maybe it's just that females sell better than males--at least in the US and Britain. Or maybe that's just an assumption of marketers. I wonder if the 38% female player rate would go up or down if the female advertising character rate were to drop.


Dmitri>I think you're talking about marketing here

Marketing is the excuse for using pictures of (usually hypersexualised) female characters to draw people into MMOs, yes. Why is that not carried through into the game world, though? Within the game world, the argument seems to be that the larger proportion of male characters reflects society's attitudes towards women ... yet the marketing reflects young straight men's attitudes towards women (or at least panders to what they might wish them to be). Why is there a dissonance there?

>I wonder if the 38% female player rate would go up or down if the female advertising character rate were to drop.

You may be able to make a prediction based on your stats, which is why I mentioned this. If you can find a correlation between (in this example) female images in advertisements or on official web pages and female character representation within the virtual world, that could point towards a trend.



مرحبا الأعزاء ،
وقد يجلس رائعة ، لأنه بفضل ،
لقد زرت موقعك ، وهذا أمر جيد ولكن يكفي أن يتمتع أكثر من أي واحد فقط من زيارة هذا الموقع والتمتع الألعاب الكثير من هذه الألعاب.


I think this is very interesting work, but Richard has a good point: the market for computer games is international.

Games produced by US developers are usually sold overseas as well, so (hypothetically) if the game-characters represented the demographics of all the customers for the game, they likely wouldn't match US demographics.

Were all the games you looked at from US companies? If the game was originally produced for the Japanese market and then translated into English, there's even less reason for the characters to match US demographics.


This is a US-centric study. The weighting data reflect what was bought by US audiences only. So, for example, while soccer games are big sellers internationally, they are dwarfed in the US by Madden Football. Our data lend much more weight to Madden Football purely by virtue of US sales. So, this is just a scope issue. I wouldn't make claims about the potential impacts in other countries where the sales figures and the demographics would differ.

The producers of the games were not only US companies. We didn't research each company, and the only data on the developers we touch on is the demographic profile the IGDA runs. Befitting the "I," it's not exclusive to the US, and probably is representative of the firms making the 150 titles we looked at.

For reference, here are the titles (duplicates on platforms reduce the effective number to 134):
1. 50 Cent Bulletproof
2. Advance Wars Dual Strike
3. Age of Empires
4. Animal Crossing
5. Animal Crossing (DS)
6. 50 Cent Bulletproof
7. Advance Wars Dual Strike
8. Age of Empires
9. Bey Blade V Force
10. Blitz: The League
11. Bratz: Rock Angels
12. Burn Out 3
13. Burnout Legends
14. Call of Duty 2
15. Call of Duty: Big Red One
16. Castlevania Dawn of Sorrow
17. Chronicles of Narnia
18. Civilization 4
19. Coded Arms
20. Condemned:Criminal Origins
21. Crash Bandicoot 2
22. Crash Bash
23. Crash team racing
24. Dead or Alive 4
25. Sims Deluxe
26. Disney's Chicken Little
27. Disney Princess
28. Donkey Kong Country 3
29. Doom 3
30. Dragon Ball GT Final
31. Dragon Ball Z: Budakai
32. Dragon Ball Z: Ultimate Battle
33. Fable
34. Fight Night Round 2
35. Fight Night Round 3
36. Final Fantasy IX
37. Final Fantasy Tactics
38. Final Fantasy VII
39. Final Fantasy VIII
40. Fire Emblem: Sacred Stone
41. Forza Motorsport
42. God of War
43. Gran Turismo
44. Grand Theft Auto: San Andreas
45. GTA: Liberty Stories
46. Guild Wars
47. Gun
48. Halo 2
49. Halo: Combat Evolved
50. Harry Potter Goblet of Fire
51. Kameo: Elements of Power
52. King Kong
53. Kirby Airride
54. Kirby Canvas Curse
55. Lego Star Wars
56. Luigi's Mansion
57. Lumines
58. Madagascar
59. Madden '06
60. Major League Baseball 2K5
61. Mario and Luigi Partners in Time
62. Mario Golf
63. Mario Kart
64. Mario Kart Double Dash
65. Mario Party 7
66. Mario Party Advance
67. Mario Superstar Baseball
68. Medal of Honor: European Assualt
69. Metal Gear Acid
70. Midnight Club Dub
71. MVP Baseball 2005
72. Namco Museum
73. NBA 2k6
74. NBA Live '06
75. NCAA Football 2006
76. Need for Speed: Most Wanted
77. Need for Speed: Underground
78. NFL Gameday 2005
79. NFL Street Unleashed
80. Nintendogs: Chihuaha
81. Nintendogs: Dachschund
82. Nintendogs: Labrador
83. Perfect Dark Zero
84. Pokemon Coliseum
85. Pokemon Dash
86. Pokemon Emerald
87. Pokemon Firered
88. Pokemon Leafgreen
89. Pokemon XD: Gale of Darkness
90. Project Gotham Racing
91. Quake 4
92. Ratchet Deadlocked
93. Resident Evil 4
94. Ridge Racer
95. Rollercoaster Tycoon 3
96. Shadow of the Hedgehog
97. Simpsons Road Rage
98. Sims 2
99. Sims Deluxe
100. Sims University
101. Sims: Nightlife
102. Socom 3 US Navy Seals
103. SOCOM US Navy Seals: Fireteam Bravo
104. Sonic Heroes
105. Sonic Mega Collection
106. Sonic Rush
107. Soul Calibur 3
108. Splinter Cell Chaos Theory
109. Spongebob Supersponge
110. Spyro: Year of the Dragon
111. Star Wars Battlefront
112. Star Wars Battlefront 2
113. Star Wars III: Sith
114. Star Wars: Republic Commando
115. Super Mario 3
116. Super Mario 64
117. Super Mario Strikers
118. Super Mario Sunshine
119. Supersmash
120. Tiger Woods
121. Tony Hawk's American Wasteland
122. Tony Hawk: Underground 2 Remix
123. Twisted Metal: Head On
124. Untold Legends
125. Wariowear Touched
126. Wipe Out Pure
127. World of Warcraft
128. WWE Smackdown
129. Yoshi Touch and Go
130. Zelda
131. Zelda Windwalker
132. Zoo Tycoon
133. Zoo Tycoon DS
134. Zoo Tycoon: The Complete Collection


In your other paper (on body shape) you ask:

Why do female characters at low levels of photorealism have such large heads?

This is a well-known characteristic of the manga/anime style, and is frequently copied in games. Among other reasons, making the head unrealistically large lets you put more detail on the face, which is one of the most information-carrying parts of the body. (But also: children have a larger head to body ratio than adults, so there may also be an element of taking a child's body as the ideal of beauty.)


Some of those games have their NPC demographics set by the subject matter. In Civilisation 4, for example, they're hamstrung by their requirement to use actual historical leaders and their geographic location. Ancient Rome was ruled by men, and you have to go to Septimus Severus to find an emperor who had any African ethnicity.

Likewise, all those sports games you cite are going to have NPCs that reflect the RL make-up of teams, coaches and what have you. If "whites, males and adults" are over-represented in the NFL relative to the general population, then having equivalent numbers in Madden NFL '06 will mean that the game, too, is over-represented relative to the general population; however, it is not over-represented with respect to the NFL - it's a reasonably true reflection of the reality (I assume - I haven't ever played it).

This is a useful set of data you have, but it's going to be very easy for people to misinterpret should they choose to do so.

Demographically speaking, by the way, a large proportion of the population of the USA is made up of dead people. Games seem to feature mainly live ones, though. Is that deadist or what?


"Ethnically, [game developers] are 83.3 percent white, 7.5 percent Asian, 2.5 percent Hispanic and 2.0 percent black."

But yur reference for this, (Gourdin, 2005), says:

"Further, due to very low response rates from most non-English speaking countries, only responses from those living in the USA, Canada, the UK and Australia were used"

So Gourdin's figures are only applicable to game developers living in those (Western) countries. But many of the actual games you looked at were not developed in the West. According to Wikipedia, Final Fantasy IX (for example) was designed by Hironobu Sakaguchi and Hiroyuki Itō at Square.

According to your figure 2, "Asian/Pacific Islander" is the only group apart from "White" that is overrepresented in games relative to the US population. (Though I'm not sure if the difference is statistically significant). This makes a lot more sense once you realise that (a) many of games are developed by Japanese companies, who likley have considerably more Japanese employees than the respondents to Gourdin's survey and (b) many of the games are aimed at the Japanese as well as US market


@Richard. The games in the sample frame are what they are. We can focus on why the characters in a sports game or Civ or any one title may be high or low on some dimension, but the takeaway point of the paper is about the universe of characters across games in the broadest sense.

Note that comparing the universe of game characters to that of televised sport doesn't change the implications about identity and representations. The TV versions can also be skewed or proportional, and noting how games diverge (or don't) is to note how games are also skewed or proportional. Taking the US football example you point out, I know that many sociologists have lamented that the dominant image of African Americans available through media is in sport. Many have thought that this may ultimately be a negative since young African Americans should aspire to do more than only grow up to play pro ball. It's one reason why our current president is such an important figure for the racial and identity politics of the US--he's not an athlete. You could make the case that if there were more African American presidents in media, Obama's election wouldn't have been quite so stereotype-jarring. As silly as it may sound, it was probably an important step for movie and TV presidents to start being black over the past 10 years, e.g. Morgan Freeman in Deep Impact, Dennis Haysbert on 24. Are there any black presidents in games?

@Susan. Yes, some of the games in the sample were made outside of the US. I honestly don't know what proportion, or what sub-proportion were made in Japan, or by primarily Japanese developers (as apposed to 2nd-party titles commissioned by, say Nintendo or Sony). There are also developers in Europe and elsewhere. Glancing down the list, I suspect that the large majority are North American-made, but that's just a guess. It wasn't the focus of our study, but is an interesting point. If you dig up something more systematic about the list here, let me know.


I take issue with the conclusions being drawn from this study. Video games are not the same as TV or movies, players do not look to the characters to represent themselves or the real world. I am female, but that does not mean I want to choose a character that is me, I want to be a robot or superhuman force that can battle zombies. Video games allow us to transport ourselves to other worlds and be someone else, the opposite of who we are. In addition, there are a huge number of games in which I can choose to be a female character, such as racing games and fighting games. One small sample of the thousands of games out there is a shallow measure, and does not give a true picture of the depth of video game characters.


Carla, you raise some key points, and it's important to note that anything I say about the implications of the results is purely speculative. This is not a study of causes, or of effects. We also note that more and more games allow players to customize their character. Some of those choices are broad, some are limited, and some do not exist at all.

As for the sample, I invite you to read the paper. It's not small (8,500 characters), and is not of a handful of games. Indeed, it's the majority of all games bought during a year, and features all of the most popular games. Any game in the long tail of titles that didn't make the sample sold fewer than 160,000 copies. Compare that to the top seller, Madden Football, which sold over 6 million copies in the same time. The average number of titles sold in the sample was about 900,000 copies sold. So I believe the sample very much does give a true picture of the depth of game characters. That's why and how we picked it.

Lastly, note that we excluded all non-human characters from this analysis. I am sure that how and why people select non-human characters and the range of choices they have is probably indicative of something, but it wasn't what we studied here.


Heh, this could get very interesting if there's a content analysis study on video game tropes. Some of the tropes might explain why certain character types (e.g. females) are fewer than others.


Dmitri>the takeaway point of the paper is about the universe of characters across games in the broadest sense.

No, not the broadest sense. You have focused on computer games. You haven't looked at sports (which are games, and which for physiological reasons are almost universally split by gender) and you haven't looked at board games (eg. the demographics of chess pieces). You haven't looked at playground games, or wargames or mind games or any other non-computer games.

However, when there are computerised versions of those games available, you look at these and then present them in such a way as to encourage readers to draw conclusions about computer games - as if the source material were irrelevant.

>Note that comparing the universe of game characters to that of televised sport doesn't change the implications about identity and representations.

If that were truly the case, you would be wasting your time looking at computer games; after all, you can get all the information you need from televised sport, for which a vast mass of statistical data is already compiled.

What I'm complaining about here is the ease with which the motives of game developers as a group can be called into question based on an agglomeration of figures that fail to take into account source material. I'm also not all that happy that it doesn't (yet - you may be publishing more on this) account for how these NPCs are presented. For example, a warband of amazon warriors would increase the representation of females of African descent, but whether it's in a positive or negative light you'd never know from the data.

Also, you count only humans. Now whereas this may make sense in terms of identifying race (you don't see many Jewish elves), gender is almost always apparent. If a game has humans as mainly male and elves as mainly female, whether for aesthetic reasons or political reasons (drow are matriarchal), well you're not going to pick that up. You'll just see a game that has a ton of male characters and no female characters.

>Are there any black presidents in games?

No idea. Are there any presidents in games? Do you have to be specific about presidents, or can any leader of a country work?



I feel I'm being a bit too harsh on this paper, because although I have some doubts about the methodology, I think the research questions are interesting.

But anyway, to carry on nitpicking the methodology...

So I believe the sample very much does give a true picture of the depth of game characters.

But the paper says that:

In other words, gamemakers created games that heavily featured male characters, but the games that were actually purchased were even more heavily male.

From your data, we know that the highest-selling titles are more skewed towards male characters than the slightly less popular ones. This would be consistent with the hypothesis that the "long tail" of the distribution - games that individually didn't get many sales, but taken all together form a respectable portion of the market - is slightly less skewed towards male characters than the games in your sample. Or - more generally - that lower budget, lower sales indie games can take more risks and include a more diverse range of themes and character types. The posssible counterargument is that the "long tail" isn't that long - that all the low sales volume games, even when all added together, don't amount to enough sakes to much affect what gamers are, on average, seeing.


Your long tail suggestion is a good one. I just don't know what's in it because we only had the resources for the top 150. There were hundreds and hundreds more in that long tail, down to games selling only 1,500 copies. In the aggregate they still don't add up to a majority, but of course they matter. Are the titles in that tail more or less skewed than the ones in our sample? I don't know. Some may skew more, some less.

Keep in mind, though, that even though there is a gap between the skew in games-as-produced vs. games-as-purchased, it isn't a substantively large gap. My speculation is that it's male gamers picking the titles off the shelf vs. game producers hoping to add some more females to the buying ranks. Not a virtuous feedback loop, I think, and perhaps even a sign that the developers are trying to broaden their appeal despite the market forces. I'm sure a lot of female buyers are less likely to look at the games shelf, assuming that it's all men blowing things up (which for the record, I quite enjoy).

Richard, you raise the point that many games draw from existing, offline cultural objects that have existing baggage and attitudes attached to them. I agree with you. In some cases, the video game version is challenging the offline version, but in most it is likely reinforcing it. In that sense, whether you find it normatively "good" or "bad" depends on how you feel about the original offline thing. Some, for example, might feel that the over-representation of minorities in real-world sports is a good thing, while others may lament it as a poor goal for aspiration among youths.

I think that TV and movies have made a lot of progress at catching up to our multi-racial country. Grey's Anatomy is a hit show where the doctors in charge are black. When that exists, then black athletes become part of the mix, rather than all of the mix. I think what the study results suggest is that there isn't a corresponding mix in video games just yet.

I've been reading the comment threads on news posts like Kotaku's coverage, and it's overwhelmingly angry young men indignant that some liberal professor is trying to make their games all PC. I got some hate mail this morning saying that if Hispanics wanted a Hispanic game, game makers would make a game for driving across the country to work on a farm. The writer genuinely had no clue that this was stereotypical, and that Hispanics might simply want to play a Hispanic superhero/adventurer/doctor/whatever like anyone else. The absence of the other racial group had never occurred to him because, I think, it's hard to notice what (or who) isn't there.

What I think I've found is that video games aren't as good a mirror of offline society as TV and movies are, and I suggest that it's not a terrible goal for them to try. It's in their self-interest to expand their market.


Dmitri>What I think I've found is that video games aren't as good a mirror of offline society as TV and movies are, and I suggest that it's not a terrible goal for them to try. It's in their self-interest to expand their market.

So the designers of Madden NFL should attempt to expand their market in what way?



I recently came across your blog and have been reading along. I thought I would leave my first comment. I don't know what to say except that I have enjoyed reading. Nice blog. I will keep visiting this blog very often.




There have been a couple of recent blog posts on avatar customization, that seem related to this (Martyn Zachary cites Dmitri's survey):


Body Armour and the Problem of “Avataritis”

The comments to this entry are closed.