Monday, July 24, 2017

"Roughing Up" and Violence

Remember the Donald Trump campaign and protesters getting pushed around? There's still a lawsuit pending on that incident, so it got me to wondering what people think about such actions.

Welcome to another round of Hollander Plays With Data.

The 2016 ANES has a couple of questions that kinda get to this issue. They ask:
When protestors [sic] get "roughed up" for disrupting political events, how much do they generally deserve what happens to them? Responses ranged from "not at all" to "a great deal" on a 5-point scale. 
How much do you feel it is justified for people to use violence to pursue their political goals in this country? Responses ranged from "not at all" to "a great deal" on a 5-point scale.
The first question clearly is aimed at Trump rallies and protesters, the second more a case of protesters using violence and seems, to me, more aimed at Democrats.

First off, there's a modest but statistically significant correlation between the two (r = .16, p<.001). That's not particularly strong. You see much stronger correlations with party identification and ideology. The more you leaned Republican, the more you liked roughing up protesters (r = .30, p<.001) but didn't like violence for political goals (r = -.06, p<.01). Or we can flip this and argue that the more you leaned Democrat, the less you liked roughing up folks but the more you agreed violence was sometimes necessary for political goals. that last one is pretty small, a correlation coefficient of .06. If not for the large N (3,572 respondents) it probably wouldn't even be statistically significant, so let's stick to the "roughing up" issue.

There's a lot to untangle here. It's no surprise that survey respondents who preferred Trump also tended to agree that sometimes you needed tough love with protesters (X2 = 420.0, df = 4, p<.001). To sum it up, for example, among those who strongly agreed that protesters got what they deserved, 75.5 percent supported Trump, 24.5 percent supported Clinton. The raw number of responses are seen below by Clinton/Trump support pre-election. It's hardly surprising that the red bars are longer at the agreement level, the blue bars are longer at the "not at all" end of the spectrum. In other words, it's all partisanship.

For fun I ran a quick-and-dirty multiple regression on the "roughed up" variable to see what factors are really driving this. Simply put, a regression puts all the factors into a model to compete with one another, with those that best explain the concept winning out and rising to statistical significance.

Some stuff does not play a role, when controlling for other factors. Age, sex, race, reading newspapers, watching television, news, or listening to radio for news are not factors. What is a factor? Less education and lower income are related to believing in roughing 'em up, which of course is a function of Trump's political support. Party ID and ideology are both factors in the expected direction. Using the internet for news is a negative predictor, so the more you use the net for news, the less you believe in roughing 'em up. Finally, for fun, watching Bill O'Reilly, even after controlling for all this other stuff, still means you think the protesters get what they deserve (beta = .06, p<.01 for you stats nerds out there). Fox News is always a special case.

Honestly, this may be worth a paper, if I can find time to really get into it.

No comments: