“The ‘Red’ vs. ‘Blue’ Crime Debate and the Limits of Empirical Social Science”

The title of this post is the title of this notable new Issue Brief from the Manhattan Institute authored by George Borjas and Robert VerBruggen. Here is how it gets started:

For the past two years, several think tanks on opposite sides of the political divide have waged war over whether “red” or “blue” America has a worse crime problem.  Commentators on the left have pointed out that red states have higher homicide rates than blue states, while those on the right have noted that the relationship is more nuanced and can easily flip at a more local level: red-state crime problems are often concentrated in blue cities, and red counties have lower murder rates than blue counties.

In this brief, we do two things.

First, we highlight this debate as an example of how seemingly minor research decisions — such as whether to analyze data at the state or at the local level — can drastically change results.  If we look at the county level, Democratic areas seem particularly murder-ridden; but when we look at the state level, Republican states are clearly more violent.  Casual consumers of empirical social science research often fail to appreciate all the ways in which researchers can manipulate the data to say whatever they want.

Second, we want to move the debate forward by showing how the correlation between crime and partisanship changes after adjusting for differences in social characteristics that could affect both crime rates and partisanship, such as the age, income, and racial composition of the region.  Previous analyses have sometimes noted the importance of these potential confounders, but few have addressed the problem clearly and compellingly.

The upshot is that models with control variables — in other words, models that compare states or counties with roughly similar demographic and economic characteristics — tell a much less spectacular story than those without. In fact, by adjusting for differences in basic demographic and economic characteristics, we can easily make the red–blue difference in homicide rates disappear.  Perhaps further research with more advanced and complex designs could make additional progress on the question.  However, given the sensitivity of the conclusions to how the researcher chooses to analyze the data, we suspect that such effort would be better spent studying and debating concrete policies, as opposed to figuring out which political party has the most violent constituents.

The authors of this issue brief also have this short City Journal piece headlined “More Crime Analysis, Less Crime Politics: Both sides of the debate manipulate data for their own purposes.”  It starts this way:

Since crime spiked in 2020, politicians and pundits have scrambled to figure out what they think is the root problem: whether Republicans or Democrats are more to blame.

Conservatives blame soft-on-crime policies in big cities, noting that many Democratic-run cities have long suffered from high crime rates and that many such places experienced particularly large spikes over the past four years. Liberals counter that, at the state level, it’s Trump-voting states that have higher murder rates, which they largely blame on irresponsible gun laws.

In a new Manhattan Institute brief, we try to calm things down a bit — and urge those worried about crime to think about which policies work, not about whether the politicians implementing them happen to be Democrats or Republicans.