We don't know what causes wide spread crime trends

Note: This is one of a series of posts I created for the American Society of Evidence Based Policing, called The Criminal Justician. With the migration to the new site, these posts were not migrated, so putting them back here on the Crime De-Coder site for exposure. Encourage everyone to join ASEBP! Originally posted in early 2023, but the lesson is still the same, it is difficult if not impossible to disentangle the reasons we observe general crime trends.

Timing a post about general crime trends is not per se best done at the new year – historically the FBI’s Uniform Crime Report for the prior year came out in the fall, in which case many news articles would run about (not so recent) crime trends. I don’t want to do a recap of crime trends though in the past year, you can find reasonable overviews from others, such as Jeff Asher (1, 2) or German Lopez. In short, the recent gun violence up-ticks that happened around the start of the pandemic have started to wane in many jurisdictions, but are still high relative to before the pandemic.

Discussion of the numbers, whether crime is going up or down (or flat), is not entirely trivial. I have written about in the past how people can be fooled by short term volatility, especially in city level homicide numbers (Wheeler & Kovandzic, 2018). But the recent nationwide increases are unlikely just typical ups-and-downs, I characterized the 2016 homicide increase as akin to a “1 in 100 year” event (Yim et al., 2020), and the 2020 increase was even more abnormal.

There is a bigger issue though in even just discussing the numbers, ultimately people do not want to just know whether crime is going up or down, but to also attribute those changes to specific causes. This is pretty much impossible to do in any rigorous way for short term (or often even long term trends), although many criminologists try (Berk, 2005; Rosenfeld, 2018)!

The typical standard we hold of say a new type of policing innovation to promote it to be “evidence based” is it should be evaluated using preferably experimental designs and replicated by several teams/jurisdictions. Neither of these standards can be held in cases of widespread changes in crime trends. The problems tend to by myriad with such macro-criminological speculation.

First, there is a difference in a pre-defined experiment, in which you look at an intervention and then see its effects, vs seeing a change in crime trends and then looking for its root cause in the past. There will always be something that happened in the past that the timing corresponds to the changes in the trends.

You may be thinking at this point – Andy, you are just being a pedantic academic, duh the Covid pandemic clearly “caused” many of these changes in recent violence trends. My response would be what exact part of the pandemic caused the increase? German Lopez’s most recent Times article lists reductions in social services, police pull back, and social unrest (both due to George Floyd and January 6th). When looking to the past to find causes, you will ultimately find many plausible causes that cannot be untangled.

I don’t want to be entirely pessimistic – there is some well done macro criminological research, but it relies on quasi-experimental research design that has a control group. So for one example, Bartos & Kubrin (2018) look at California’s Proposition 47 and subsequent changes in California’s crime trends relative to trends generated from other states. The difference between this study with most current Covid (and social unrest) explanations on crime trends is that they don’t have a clearly defined group that the intervention does and does not apply to – Covid happened everywhere.

It is possible that future researchers will identify more specific changes related to the pandemic and formulate stronger quasi-experimental designs. For example, it may be possible to identify differences in firearm access trends in different areas and relate those to differences in crime trends. But we have one similar example in the past on widespread macro level crime changes – the general crime increase in the 70/80’s and the crime decline in the 90’s – that still 30 years later has not produced consensus among criminologists on what caused the increase or the decrease (Blumstein & Wallman, 2000).

Look, I want to know what causes widespread crime trends just as much as anyone. It is important to understand the past to be able to shape the future. But I suspect 10 years from now our understanding of what caused the general uptick in shooting violence starting in 2020 will not be any clearer than it is now. It is really the reason we advocate for experimental designs, in those circumstances we can be relatively sure that some specific intervention we did caused a subsequent outcome. We will just never have that level of certitude with what causes more general crime trends. It doesn’t make sense to demand a particular level of rigor for policing interventions to count as evidence based and then accept more superficial analysis for macro crime patterns.

References