Elsevier, publishers of Journal of Adolesence, sent a newsflash today announcing a journal edition devoted to the effects of videogame violence on children. The newsflash claims that a consensus has emerged among the scientists that violence in media, especially videogames, causes violence in users. Their words follow. The journal is available here.
In my opinion, researchers in this field establish correlations, not causality. Still, peer-reviewed research is hard to ignore.
From the Elsevier Press Office:
DO VIOLENT VIDEO-GAMES AFFECT ADOLESCENTS' ATTITUDE AND BEHAVIOUR?
Yes, is the conclusion of the most recent research.
A special issue of Elsevier's Journal of Adolescence, just published, addresses this familiar question and moves the debate forward, through important new research conducted in several countries.
The conclusion, say the editors, is that "the case for the negative impact of violent video-gaming may have been under rather than over-stated".
The summaries of published papers in this email are intended to be read as accessible signposts towards the original abstracts and articles, which of course have been peer-reviewed. However, the summaries in this email have not been peer reviewed, nor have they been approved by the authors of the articles or their editors.
+44 (0)1865 843277
Castronova: In reviewing existing work on this topic, my conclusion has been that most research in this area is simply ham-handed, paying too little attention to issues of causality and substantive significance. These are issues that econometricians have worked on for decades, and my feeling is that some of the researchers in the videogame violence field are just not up to speed. It seems that most conclusions in this literature are based on experiments where you put a bunch of teens in a room, let them play Quake, then answer a questionnaire asking whether they hate their mom. And the ones who just played Quake say "yeah, I hate her guts!" more often. But there are serious problems going from this kind of data to any real world implications.
First, surveys about opinion are notoriously fishy. So is in-lab behavior. Has anyone tied videogame violence to a real consequence, i.e., actually hitting anyone in anger?
More important, though, the reverse causal explanation for these outcomes is equally (if not more) plausible. Let's say I'm a teen, and my world is violent, brutal, neglectful; I love videogames because of that; if you put me in a room with kids from loving homes, I'll not only be more inclined to get into Quake, but I will immerse myself in its thought-world just to get some relief from the rage I feel, rage that the other kids just don't have to deal with. The causality doesn't run from videogame to violent teen to parent-as-victim. It runs from parent-as-criminal (yes: criminal, in greater or lesser doses - neglective, abusive, demanding, tyrannical, invalidating, narcissistic, and generally unwilling to work on themselves long enough to become capable of meaningful love relationships with spouses or children) to violent teen to videogame.
For example, from the summary of one of the articles:
> In the same study, they also found that past history of exposure to violent video games is positively associated with aggressive self-views. One implication of their work is that the effects of violent media exposure on self-image can occur entirely without awareness of change, a finding that helps explain why violent video gamers so stoutly deny that their hobby has any impact on them. They simply cannot notice these subtle changes in self.
Note the words "positively associated." Now consider the fact that the height of men is positively associated with the height of their sons. Does that mean that father height causes son height? Nope. Cut all of the dads' heads off, and the sons will be no shorter, will they?
Isn't there a clear and obvious reverse explanation in the above case? Isn't it equally plausible that people with aggressive self-views tend to seek out violence in videogames? I haven't read the article yet, so perhaps there's been a careful effort to establish causation. But I doubt it. Vidoegame violence researchers just don't get very far into the tricky issues of connecting theories of causality to data outcomes, at least, not as much as econometricians do. And if you don't go into these issues, you lose the inferential purchase of the data. Completely.
Now, why would everyone be so accepting of one causal interpretation of these data points and not the other? Fact is, contemporary society is extraordinarily unwilling to consider the secondary causal line. No, no, no, it can't be parents. No. Parents are fine. It's the kids who are broken, not the parents, not the marriages. It's the kids and their darned videogames. Well, the literature, as currently focused, will never produce anything to challenge this view. It will just keep finding a videogame/violence correlation. Over and over. And parents and governments will continue to avoid looking in the mirror, where the real problem lies.
Third, why don't we see studies comparing the effect of videogames on violence to the effect of other things, such as bad parenting? If we are going to become concerned about something, let's be concerned about the things that have the greatest effect on childhood violence. I could do a study where I gave kids ice cream and then noted the level of aggressiveness. I imagine that ice cream would have a sedating effect. And if that's the only factor I am looking for, I will dutifully report "Study shows ice cream reduces violence in children." But how MUCH? Is it enough to be concerned about? The answer to this, in most of this research, is to report that some effect is 'statistically significant.' They then conclude that the effect is also substantively significant, i.e., it matters. Anyone who does that, however, simply doesn't understand the concept of statistical significance. It has nothing to do with substance. Yet an extraordinarily large number of researchers in the social sciences, including economists, just don't understand the meaning of the statistical tools they deploy. If you're interested in this, check out Dierdre McCloskey's writings on how scholars improperly make inferences of substantive importance from statistical tests (I had to do it myself on many occasions to satisfy editors, because I don't have tenure.... but stay tuned). The basic message, if you're clued into this jargon, is that when someone says "the difference was statistically significant at the 5 percent level" and then moves on without comparing the actual size of the effects, they don't know what they are doing (and neither did the journal editor). For example: suppose in a sample of 1 million basketball shots for each player, I found that Allen Iverson's shooting percentage was 48.9 while Gary Payton's was 48.7. With a million data points, even the tiniest difference will show up as statistically significant. But substantively, they are equal. People in the science lit would announce: "Iverson shown to be better shooter than Payton." People in the NBA would laugh in their face.
Facit: are videogames an important cause of actual violence? The existing literature here is just a premeditated balm, designed, conducted, and construed to support a theory that minimizes blame on the real culprits in this story. This literature doesn't tell us anything about real violence, or about causality, or whether any of the effects are important. What it does do, is provide cover for a culture whose relationships skills, especially involving parents and children, have withered almost to nothing.