Back to Basics
What unites explanations for the crime decline may be their unintended positive consequences
Let me spill some ink on one of my favorite papers, one that seems to me to explain much of the US experience with crime in the last 50 years and is also enormously useful in molding our future. The paper is about the original LoJack system, a technology now so obsolete it is more archeology than social science. Still, archeology always struck me as one of the most objective of the historical disciplines, since it gives equal voice to be the winners and the losers of history.
The goal of this essay is to apply some of the logic of Lojack to understand how crime declined and how that can improve crime prevention now that violence is increasing. But to do that, there is a little stage-setting that must come first.
For more than a generation, beginning roughly in 1991 and continuing up until the beginning of the pandemic years, crime and violence in the US were generally declining rapidly, declining slowly, or staying relatively stable. This was a very good thing and like all very good things in America, it was deserving of a tremendously evocative brand and was thus labeled the Great American Crime Decline. Search parties were sent—and continue to be sent—to every corner of the known data universe looking for clues as to its causes. The idea, of course, is that if we have a better understanding of the mechanisms that explain our history, we will have better tools to bend the arc of our future. This assumes that our histories and our futures are symmetrical, and while this is certainly up for debate, the journey seems to have a reasonable chance of at least some success. And so it goes.
Now, the blunt metaphor I am about to apply turns out to be really important, and I will return to it again. Most criminals are opportunists, not masterminds—which means they are sensitive to the risks and barriers they confront. The links between them and their victims tend to be a straight line. I suppose this is why true crime stories are so popular because most crime is blunt, not oblique and a good story needs some distance from mundane reality.
And so it is that the search for explanations for the crime decline has also tended to be linear. This is reasonable and fits Occam’s Razor, which beseeches society to look for simple explanations for complex problems. But it turns out there are lots of Razors and many of them tend to undermine Occam.
Another philosophical Razor is my Dad’s Razor, named after my father, who likes to say that “for every intractable problem, there is a solution that is simple, intuitive, and wrong.” This is a folksy take on Hume’s Razor, which when simplified, says that a big explanation is needed to explain big effects. Thus, if you have a big problem and a small answer for it, you probably have no more than a partial answer.
The problem with most explanations for the crime decline is that they are too small, by themselves, to explain this enormous phenomenon. And that is because our search for the crime decline has focused almost exclusively on direct links. We like Occam’s Razor which pats us on the back and says our intuition—what we think before we think about something—is really valuable. We like the story in Malcolm Gladwell’s Blink where art historians can spot a forgery in a, well, blink. We skip the later parts of the book that talk about the failures of intuition, like New Coke.
Many of the explanations for the crime decline seem to fail the Hume Razor test, at least by themselves. The abortion hypothesis—that crime fell in the 1990s because abortion became widely available in the 1970s post-Roe—seems to fail because abortion is a normal good with respect to income and thus people with more money will demand more of it, and thus the numbers of people born into high crime environments would be more limited than on first glance.
The lead hypothesis is limited in the same way. This theory hypothesizes that the removal of lead from consumer products—particularly gasoline—reduced lead in the bloodstream of people exposed to it. High blood levels can lead to brain injury, particularly in the parts of the brain that exert self-control over social behaviors. Thus, less lead exposure equals less crime. The theory states that those in particularly poor dense places are uniquely susceptible to lead poisoning and removal of lead thus leads to big reductions in criminal offending. Only, the original lead paper found no effects in the densest city in the US—New York City.
These explanations likely have some merit, but they seem to fall short of the Hume test. Each of the two dozen explanations for the crime decline has merit—their merit puts them on the list—but the idea alone is insufficient to explain the crime decline. The key is to figure out what attributes these ideas share.
But just this insight alone shows the folly of our default approach, which is to assume that only direct links matter. That only those things specifically designed to reduce crime can reduce crime. And each of these approaches makes an implicit assumption. In order to believe any single explanation for the crime decline, you have to believe that criminals tend to be masterminds, not opportunists. If criminals are masterminds, then you need only gather up the right people or, (hideously) abort them, in order for crime to decline for everyone.
If, on the other hand, you believe that most crimes are crimes of opportunity and that everyone has some propensity to offend given the right circumstances, then any single explanation for the crime decline is unlikely to satisfy Hume’s razor. No single explanation will be big enough.
If you are still not convinced, let’s return to the more than two dozen credible explanations for the crime decline. By itself, each appears to satisfy Alden’s Razor, which is also known as ‘Newton’s Flaming Laser Sword’ a label which is so unbelievably fantastic I can’t believe it hasn’t motivated an entire generation to study moral philosophy. The name derives from this razor being substantially sharper than Occam’s.
Anyway, Alden’s razor—the flaming laser sword—simply states that in order for an explanation to have credibility, it must be testable by experiment. This complements Popper’s falsifiability Razor which says that you should fuhgeddaboud any hypothesis that cannot be refuted. I should note that all of the two dozen-plus explanations for the crime decline appear to meet this standard or at least are testable using modern statistical methods.
But here’s the catch. Each of these hypotheses tests itself against the same null hypothesis. That is, each hypothesis tests against the same fact, that crime has declined. A baseball example helps to clarify (knowing anything about baseball or even caring for baseball is not required for this example to have meaning). Suppose that a team wins 20 more games in 2022 than it did in 2021. The team sets about understanding why this happened so it can win even more in 2023. One theory is that the defense improved, so it tests and finds that the team played better defense in 2022 and had more wins than in 2021, all else being equal. Another group posits that home runs led to more wins, and finds the team had both more home runs and more wins. Another group finds that pitchers threw more strikes in 2022 and again, associates this with the winning season. And more esoteric theories emerge, all plausible. Ticket prices went down and more fans were in the stands. It got hot temperature-wise in May rather than June and stayed hot. All of these theories and more create credible tests of the effects of causal mechanisms on wins.
Only, the 20 wins are doing a lot of work in this example. That’s a lot of theories claiming credit for the same thing in the same way that multiple theories claim credit for the crime decline. A reasonable reaction is to ask: should those theories share credit, instead of claiming it all?
The reasonable conclusion from all of this is that there is no single explanation for the crime decline, but rather that there were some combination of factors that jointly contributed. Now, the key question is, are there shared features of these explanations, such that they combine to describe a latent attribute? That is, are all of these crime decline explanations actually just parts of a larger whole?
Latent constructs are a key feature of our lives. If you talk about anything from cultural competency to alcoholism to motivation you are describing a latent construct. Latent constructs are more of a concept than a noun, where the whole is something more than the sum of the parts. Alcoholism is describable, for instance, by the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition, Text Revision (DSM-5-TR™) but it is a mismash of behavioral problems (does drinking lead to trouble at work or with the family?) and cognitive problems (do you feel guilty about your drinking?). Latent constructs are like pornography, something where you know it when you see it.
I have written that the latent construct that joins these two dozen explanations for the crime decline is prevention. Prevention takes many forms, but what I mean in this case is universal prevention, which is the removal of risk conditions and the strengthening of assets that affect everyone. To some extent, each of the crime decline explanations—from removal of lead from paint and gasoline to a more stable macro economy—either removes a large risk for everyone or contributes positive assets to everyone’s development.
Still, that begs the question, what is the mechanism that led to large-scale prevention in the 1990s that continued until relatively recently? And, what kind of mechanism could cause both increases and decreases in the amount of prevention in society?
And that leads me to what I really want to talk about.
***
And what I really want to talk about are externalities, and specifically, positive externalities. Twenty-five years ago Ian Ayres and Steve Levitt wrote an economics of crime paper that argued that “positive externality-generating unobservable self-protection”… “clearly...affects criminal behavior.” What that means is that if you invent a technology that makes it more likely a stolen vehicle will be recovered and the thieves will be arrested—and if that technology is invisible to the thieves—it reduces everyone’s risks.
This paper did not receive the attention it deserved, although it suffered from what I would call the Impenetrability Razor, which is that ideas have to be understood to be believed. And this article was pretty tough for the lay audience.
But still, it is a remarkable finding and critical to understanding what went right during the 30-year crime decline from 1990 to 2020 and what is going wrong today.
The underlying concept of externalities is straightforward. When two people engage in a transaction, it is sometimes that case that others are affected by the transaction. In the classic example, someone makes a product and I buy it, but the production of the product creates pollutants that affect people near the factory. That pollution imposes a cost on people who had nothing to do with the transaction—they are external to it. To be fair to everyone, the costs to the people affected by the pollutants should be included in the cost of the transaction. It should be internalized.
Public policy takes this idea one step further. It says that not internalizing the pollution costs—the externalities—warps the market in an important way. Because the costs do not reflect harm to people who were not part of a transaction, those costs are artificially lower than they should be. And because the costs are low, too much of the product will be produced. Society is probably willing to accept some level of pollution (almost all of us drive air-polluting motor vehicles after all) but the price should include the costs of fixing the problem. When the price does not, there will eventually be much more of the problem than intended.
Most people read a definition of externalities and immediately think about issues they are concerned about (bullets from guns hurting bystanders, high-cost real estate crowding out affordable housing, etc.). While we argue endlessly about the particulars, few would argue against the idea that a proper role of government includes policing negative externalities in the marketplace.
What the Ayres and Levitt paper does is to flip the externality problem on its head and ask—are there products that benefit people who are not part of the transaction? They find that indeed there are, at least with respect to crime. But they have a particular attribute that is unexpected—the products have to be invisible to potential criminals.
The product Ayres and Levitt focus on is LoJack. In the nineties, before cars and trucks became connected to the Internet of Things (IoT), devices to prevent car theft were mainly limited to locks. The Club, in particular, was popular. But unlike LoJack, it was easy for a would-be car thief to peak into your car and see if it had a Club. If it did, the thief would just move to the next car. That actually creates a negative externality for the rest of us—your purchase of a Club reduces your risk that your car is stolen but increases my risk that my car is stolen.
But, LoJack is invisible. A thief peeking into a car has no idea if it has LoJack. It is invisible. But effective. The car thief peeks into the car, sees no likely problems, steals the car and drives it to the chop shop. You realize your car has been stolen, call the cops, who flip on the LoJack locator, find the car and in the process of recovering your stolen car, they break up the chop shop. This, as you might imagine, is really bad for business. In 1990, there were over 146,000 cars stolen in New York City. In 2021, it was about 10,000…
Graham Farrell and I wrote an article in 2002 arguing that the government should invest in crime prevention tools with positive externalities. I’d like to take that idea much further today.
Today, it appears that many of the explanations of the crime decline reduced crime not through a direct effect, but rather through a positive externality. Let me give you just a couple of examples.
I argued earlier that the abortion hypothesis was limited because it did not consider that wealthier people were more likely to demand an abortion. Here, I want to argue it is also incomplete because it considers abortion in isolation. Roe v Wade also brought a new focus on women’s reproductive health in general. The post-Roe era saw massive growth in birth control beyond the pill and including all types of contraception. This process of de-stigmatizing contraception lead to huge reductions in teen pregnancy. Thus, there may have been fewer children being born into desperate poverty in economically isolated places, but these are mainly externalities from a broader focus on maternal health.
The lead exposure story is similar. Proponents of lead employ Occum’s Razor thinking and focus on the removal of lead from gasoline. But during the same period, all sorts of environmental protections were undertaken to remove all kinds of harmful chemicals from consumer products and greatly reduce all kinds of harmful pollution. Many of these have the positive externality of reducing crime.
Here again, is the list of credible explanations for the crime decline. Explanations where positive externalities played an outside role are in bold.
· Mass incarceration (Truth In Sentencing/Three strikes/Mandatory minimums)
· Problem-Oriented Policing/HotSpot Policing
· Economic Stability/Consumer Sentiment
· More community-based organization serving at-risk neighborhoods
· Increase in alternatives to incarceration/diversion
· Therapeutics for mental health
· Decreased drug use/alcohol consumption for 16 and 17 year-olds.
· Better policing investigational technology (CODIS/NIBIN)
· Better policing operational technology (COMPSTAT)
· Crime Prevention Through Environmental Design
· Increased average educational attainment
· Dramatic declines in teen births
· Widespread use of medication (Ritalin, anti-anxiety, anti-psychotic, anti-depressant)
· In-home entertainment (internet, video games, pornography, cable)
· Under-reporting as crime moves online
Conclusion
The point of this essay was simply to make the claim that positive externalities played an outsized role in the Great American Crime Decline. Down the road, there is much more to say about the role of prevention in creating positive externalities directly and by incentivizing private uptake of products with positive externalities.
To conclude then, one other idea seems relatively obvious. Although little has been written about the 2010s, this was a decade where crime and violence seem to have steadied such that additional crime declines were difficult. This suggests two things.
First, that there is a limit to how much crime and violence can be prevented with our usual array of crime and justice tools that take on crime directly—2014 seems to be the year where these gains were optimized. And immediately thereafter, the negative externalities from those tools, particularly from wanton over-policing, became obvious.
And second, that most of the gains from the positive externalities from changes (both public and private) in the 1970s, 80s and 90s, have begun to wane.
This suggests that it is time to actively search for new positive externalities. The difference between the beginning of the Great Crime Decline in the early 1990s and today is that we now know to look for them. And they are there to be found.
The summer jobs literature provides a great example. Summer jobs were thought to produce a positive externality for underemployed and high-risk teens. Teens would commit fewer crimes through a positive externality—more money in their pocket reduced their need for larceny, spending time outside their neighborhood sequestered them from trouble, etc. But mismanagement and a lack of focus led to widespread findings of no effect. But a simple tweak—combining summer jobs with cognitive-behavioral training developed from the prevention literature—changed the mechanism, and new studies consistently find important effects.
There are more externalities to be found.
Great read. I was fortunate to leave a 10 year Chief tenure on a violent crime decline in an era of increases across the country. But, was it my strategies? Was it luck? Probably a little of both, or maybe neither? A broken clock is always accurate twice per day so maybe my timing was just good! Of all of the plans and programs we attempt in policing I still think our impact on crime overall is minimal, no matter how successful they may seem. However, raise a child in a stable environment from birth with solid nutrition and access to education and healthcare and most of your crime problems will dissipate. But, that is the long game and society is built around the short game!
Great analysis, John. You have a way of stimulating our thinking. So I started to think about poverty rates among young adults (which increased recently) and how the Covid19 relief funds may have produced more awareness of being poor. Then I thought, do externalities interact in ways that enhance some externalities and not others, and if one externality is internalized would it change the effects of others? I love the way you continue to tackle this complex set of questions.