Why did crime decline in the 1990s? I think you all know my answer. But yet another theory was offered recently by Lena Edlund and Cecilia Machado in an NBER working paper: cell phones. Their paper is complicated and I’ve only skimmed it, but their basic theory is simple: cell phones changed the nature of drug dealing, making it less turf-based and therefore less violent. This reduced the murder rate, mostly in urban areas.
I don’t have a strong opinion about this, other than the observation that the crack epidemic of the late 80s has long been considered responsible for an increase in urban crime, which then declined as crack itself declined. In any case, I don’t think the cell-phone theory is incompatible with lead as a driver of crime. Take a look at the chart below of the US murder rate:
Please note that I’m not offering this as an explanation, just as a possibility. Lead might well have been the primary driver of the murder increase, doubling it between the mid-60s and mid-70s. But other things affect crime too, and you can see that there are several humps in the murder rate. The late-80s hump might very well be caused by the crack epidemic, and it’s possible that cell phones reduced it. Lead then did the rest.
It’s also possible that the crack wars simply burned themselves out, and it’s only a coincidence that this happened at the same time that cell phones were starting to enter the market. This has always been the most popular theory, primarily because drugs are faddish and we know for a fact that crack distribution declined in the 90s as it was replaced by marijuana as the illicit drug of choice. It’s never been clear to me why crack distribution is inherently more violent than other drugs, and naturally I suspect that crack just happened to hit the market at the same time that lead poisoning peaked among 18-year-olds. But that’s just a hunch on my part, not something backed up by evidence.
Aside from this, I’d warn generally that there aren’t all that many murders in the US: only about 20,000 per year even at its peak. This means that the change in the murder rate over a short period is always based on a fairly small sample size and can be fairly noisy, especially when you rely on state-by-state or city-by-city comparisons. Caveat emptor.