Correlation is not Causation, clearly

Repeat after me:

Internet Explorer vs Murder Rate Will Be Your Favorite Chart Today

We discussed causation and correlation in my Methods class this morning. I generally use the ice cream sales and murder rates example, but since this has been floating around the internet lately, I figured I would throw it in. It got a few chuckles out my class, from those who also wanted to insist that ice cream made people deranged and thus more likely to murder someone, but a good reminder nonetheless. A regression of murder rates on ice cream sales or internet explorer market share will have a positive and statistically significant coefficient estimate, but it doesn’t mean that either is causing more murders to occur.

Source

Cartoon Violence

There are all sorts of studies that claim that cartoon violence, particularly of the video-game variety, encourages children to be violent, but this morning I was treated to a cute little analysis of violence that most likely isn’t actually engendering any violence. It does bring up questions of exactly what is being represented, though. Does violence in New Yorker cartoon articles correlate with rates of violence in the real world?

There are myriad questions, of course, that one could ask that might strengthen or weaken the relationship. Whether the total number of cartoons (violent or not) is constant over decades, turnover in the cartoonists, who the cartoonists are, etc.

I’ve never really known any cartoonists, so I can’t say much about their average temperaments, or sources of inspiration, or how much they read the news, or how much their cartoons reflect other trends in society. It’s reasonable to assume that, however inaccurately, they have some idea of what’s going on, even if they don’t metaphorically have their finger on the pulse of the nation. But more than rates of actual violence or murder in real life, I wonder if the increases can be associated with differing levels of depression or other mood disorders. A quick google search did not reveal an easy way to get depression statistics in the same format as the New Yorker cartoon violence data (and even if it did, it’s unlikely that there are enough data points to get a statistically significant answer). I’m sure a health economist friend (or even my mom) could help me out with this, were I to pursue it.

Then, I thought, maybe confidence in government? This WSJ graph has historical presidential approval ratings. The discontinuities would make it difficult to analyze, but  could be aggregated over each decade. There might be a story, or even better, might be a story about stability. Perhaps more volatility in presidential approval ratings over time means more violence. Or some combination of level and volatility?

Regardless of the outcome of such a search, this is clearly not a case of causation. Even asking whether cartoon violence is predictive or reflective of actual trends in depression or presidential approval ratings would take much finer data and many more assumptions. But I think it’s fun to try to link trends in media to trends in other facets of life. I used to tell a friend, who complained often about seemingly obvious journalism–“People use technology to do stuff”-type articles, for instance–that they’re necessary for the historical record. They’re also for nerdy economists to read and try to find patterns over time.