The Monkey Cage

In comments on my earlier post, Andy and then Michael suggest that the proper comparison is between the Gallup item (“Is there more crime in the U.S. than than there was one year ago, or less?”) and the change in the crime rate. So here are three more scatterplots. They plot the percent saying “more” against the change in the crime rate vs. the previous year, 2 years ago, and 3 years ago. With each additional “lag,” particular years are lost, of course — and there is missing survey data in some years as well — but this comparison is still instructive. I put the correlation and associated p-value under each plot.

The challenge here is to figure out what people’s frame of reference is when they answer this question. My original post showed that, from 1989-2001 there was an extremely strong correlation (r=.97) between the violent crime rate and the percent saying “more.” However, after 2001, this relationship disappeared and a larger percent of people said “more” even as the crime rate remained flat.

If people are answering this question as it’s worded, then the topmost figure above — which depicts the change in the crime rate vs. the previous year — should show the strongest correlation. It does not. The strongest correlation is actually with the crime rate three years ago, although this correlation shrinks dramatically (to r=.43; p=.17) when 1992 and 1993 are excluded. So I’m not all that confident in this relationship. Such are the limitations of small datasets.

But let’s assume that this correlation vs. 3 years ago is real. It is puzzling, to say the least. In other areas — for example, the economy — voters are much more myopic, willing even to discard years of bad economic performance and reward incumbents for the most recent 6 months of economic gains (see this paper by Larry Bartels and Chris Achen). So why would perceptions of crime change so slowly?

All in all, these three plots illustrate my original point: there is a striking gap between perceptions and reality. You have periods with movement in opinion, such as in 2002-2008, but no real change in the crime rate. Moreover, consider the absolute level of opinion. There are many years in which the majority of the public says “more,” but crime has decreased relative to the previous year, 2 years ago, and 3 years ago.

Nothing in either of these posts is conclusive. But I’m still struck by the disjuncture I noted in the first post, and by how much perceptions simply don’t correspond that closely to crime rates.