The Gender Cocktail. Part I: Learning to Sample

Sex. Gender. Gaming.
 

Historically, this hasn’t been the industry’s best cocktail. Even if we put aside for the moment that the US has some serious issues with female sexuality--we’re far more OK with eviscerated organs than breasts--gaming hasn’t exactly been at the forefront of the progressive movement.

As a kid playing video games I didn’t give this any thought. There was a princess, she was hot, and she needed saving. Game on. As a parent (with a daughter), it’s hard to not grow up a little. And like the violence debate, the great gender and gaming debate rears its head at pretty regular intervals--usually when a game does something truly egregious or on the border between cute and creepy.

Cards on the table, I’m moderate to slightly liberal, and I’m ready to be offended on behalf of the women in my life. Yet I’m also a researcher who wants systematic answers and hard data to base conclusions on. And on that score, there is evidence that hypersexualized portrayals of women leads to greater tolerance of sexual violence and even rape. But linking representations to attitudes (note that you can’t do a study to link to behaviors due to ethics) presupposes that we actually know how people are represented in the first place. So:

What is the state of representations in gaming?

Is it getting better or worse?

What fundamentally drives it?

It’s easy, and very tempting, to fire off answers to these questions based on the games you play and the people you know. After all, what could be less sexy, yet more important in all of this than sampling? I mean sampling, who even does that? Let’s start with what it is.

Sampling is the process of taking a few selected cases and then extrapolating up to a larger population. It’s how political pollsters can take 1,000 people and say they know what “America” thinks about something, within some margin of error. How can they do that? They know a lot about America thanks to the US Census. So, they know they need to get X amount of women, Y amount of Latinos, Z amount of people aged 30-35, etc. in order to get a representative sample.

Tackling issues of representation in video games causes a similar kind of exercise, only here it’s much, much harder. I’m going to cover the very simplest question first: What are the demographics of game characters?

Consider the following basic elements of that smallest question:

Do we care about playable characters, background characters, or both?

Do we count all games equally?

There are zillions of games. Do we pick them all?

How do we do this in an unbiased way so the results are beyond reproach?

The good news is that we actually have tackled this. I’m going to spell out the results, but you can find the full research paper here. The bad news is that we did it nearly nine years ago and games have changed a lot since then. Players, platforms, mechanics and business models especially have been all over the place. So what follows is essentially the “before” and there is no “after” yet. Aside: I run a game analytics company that’s starting up with its first customers right now. We measure characters, so at this time next year we can start providing more regular benchmarking.
 

Step 1: Find the list of what games were sold in a given time period. We used NPD data to track a single year of games sold across nine different platforms. We picked the top 150 games sold by retail, which covered slightly more than half of known dollars spent. Remember, back then this really was mostly a retail industry, so you could indeed track these things before the explosion of the App Store and microtransaction games.

Step 2: Don’t treat all games equally. Clearly Madden should carry more weight than Beyblade if for no other reason than that it sold a zillion times more copies. As a proxy for how much exposure content has, we weighted the results by sales. So, a game selling 7 million copies counts 7 times as much as one selling 1 million copies. Not perfect, but not terrible.

Step 3: Pick the characters. Hire a bunch of coders to view the games and record every single character in them. We looked at the first 30 minutes of every title and had at least two coders take independent records. We made a note of playable vs. non-playable. Our two coders agreed on the categories around 93% of the time. We covered over 5,000 characters.

Step 4: Throw it all into a database and look at the graphs.

So, what did we find? First, we decided to pick a baseline for comparison. You can argue what the game world should or should not be compared to and there is no perfect answer. We picked the US Census because we were looking at US sales only and because it stands to reason that the “real” world is a decent baseline. This lets us state that if a group shows up more or less often than the Census baseline, that group can be said to be over- or under-represented.

OK, results.

It may not be terribly surprising to see that there are more male characters than female, but the results were more skewed than we expected. First of all, the “real” world of the United States is 50.9% female, 49.1% male. If we were to see the “real” gender world show up in games, that’s what we would see in them. Overall, characters in games are heavily skewed male: 85/15.

But what about playable characters vs. the eye candy of the scenery? Here the skew is actually stronger, at almost 90/10.

You can make an argument that in recent years games have allowed for more character control. It’s often easier to customize your character now, and that can sometimes mean the gender. Unfortunately, in the wilds of the current marketplace it’s very difficult to give a systematic answer to that.

Does any of this matter? Here we have to be clear that we don’t have data on the effects of these patterns, just theories and research from other media. It’s pretty clear that the simple absence of a group is generally speaking not good for that group. Think about Blacks, Asians, Hispanics or Gays on TV. Before they were on, it was like they didn’t exist (the fancy academic term is symbolic annihilation), and then they had to go through a phase as villains or sidekicks before being able to be primary characters.

So no, I’m not claiming that the lack of female characters proves anything, but I suspect it probably doesn’t help much. Consider the young female player playing game X and thinking “where the heck are we?” At the time of the study, females were 38% of the playing audience, yet only 10% of playable characters. Is that related at all to the giant shortage of women entering computer science? Again, there’s no causal link in this study, but it probably doesn’t help much.

In Part II, I’ll go from simple numbers to actual body shapes and add some stats from the industry.