Last time we were talking about international law and crime, we brushed against a question that really deserves its own episode. What actually makes crime go up or down in a society? Daniel wrote in with a whole set of questions about this. He pointed out that Israel has relatively low domestic crime despite the geopolitical volatility, and he wanted to know which factors we assume matter — policing, education, economic development, drug use — versus which ones actually do when you look at the data. And then he asked whether the drivers differ for minor crime versus serious violent crime. So today we're going to take that question seriously — not with gut feelings, but with the best comparative criminology we have.
The variation across countries is enormous. The global homicide rate sits at about five point eight per hundred thousand, but that average hides a range from under one in Singapore and Japan to over sixty in parts of Central America and southern Africa. That spread — sixtyfold — is the puzzle. Why does one society produce sixty times more homicide than another?
And most people, if you stopped them on the street, would reach for poverty or policing or education. Those aren't wrong exactly. But they're incomplete in ways that matter, and some of what we reach for first turns out to be almost irrelevant once you control for other things.
Right. So the arc here — we'll look at what the data actually says about the major drivers, then tackle the misconceptions, then get to Daniel's stratification question. Because crime isn't one thing, and the factors that predict shoplifting are not the same ones that predict homicide.
Start with the biggest one. What's the single strongest demographic predictor of violent crime across societies?
The proportion of young men aged fifteen to twenty-nine in the population. The youth bulge hypothesis. It's remarkably consistent. Countries with large youth cohorts — and you see this across much of sub-Saharan Africa — have higher baseline violence, all else being equal. The mechanism is fairly straightforward. Young men are both more likely to commit violent acts and more likely to be victims of them. It's not cultural, it's not moral — it's a statistical regularity that holds across every society we have data for.
And the effect size?
Substantial. A one percentage point increase in the share of young men in the population is associated with a measurable increase in homicide rates. The exact magnitude varies by study, but the direction is so robust that it's essentially not debated in criminology. The debate is about what to do with that information, because you can't exactly legislate the age structure of your population.
No, but you can watch it coming. Countries with youth bulges now are going to have different crime profiles in twenty years as those cohorts age. That's a policy-relevant fact even if it's not a policy lever.
And that brings us to the second major factor, which is alcohol consumption per capita. The WHO and UNODC data show a robust correlation between per capita alcohol intake and both homicide and assault rates, and this holds independent of poverty levels. It's not just that poor countries drink more and have more crime — you see the relationship within income brackets.
What's the mechanism there? Is it pharmacological or social?
Both, but the pharmacological effect is primary. Alcohol impairs impulse control and escalates conflicts that would otherwise de-escalate. A shove becomes a punch, a punch becomes a weapon. The WHO estimates that alcohol is involved in roughly forty percent of homicides globally. And the per capita consumption numbers map onto that — Russia, with very high alcohol consumption, has homicide rates far above what you'd predict from its income level alone.
Compare that to Japan.
Japan is the inverse. Low alcohol consumption per capita, low homicide rate. Now, Japan also has low inequality, low youth unemployment, and a very different social structure — so you can't isolate alcohol as the sole variable. But when you run the regressions across countries, alcohol consumption keeps showing up as an independent predictor with a meaningful effect size.
So we've got demographics and alcohol. What's the third one — the one that sounds like a conspiracy theory until you read the literature?
Lead. The lead-crime hypothesis. Childhood lead exposure — from leaded gasoline, lead paint, industrial pollution — correlates strongly with later violent crime rates at the population level, with about a twenty-year lag. The phase-out of leaded gasoline in the nineteen eighties and nineties is one of the most plausible explanations for the dramatic crime decline in the United States and other developed countries starting in the nineteen nineties.
Walk me through the mechanism.
Lead is a neurotoxin. Childhood exposure impairs prefrontal cortex development — that's the part of the brain responsible for impulse control, judgment, and regulating aggressive behavior. The damage is permanent. So a child exposed to high environmental lead in, say, nineteen seventy is going to have impaired impulse control as an adult in nineteen ninety, when they're in the peak offending age range. Multiply that by millions of children, and you get a population-level effect.
And the evidence for this?
Jessica Reyes published a paper in two thousand seven that looked at the phase-out of leaded gasoline across US states. Different states phased it out at different times under the Clean Air Act, creating a natural experiment. She found that the reduction in childhood lead exposure explained more than half of the crime decline in the nineteen nineties. Other studies have replicated this across countries. The relationship shows up in the UK, in Canada, in Australia — basically everywhere that phased out leaded gasoline and tracked crime rates.
More than half. That's a staggering number if it holds.
It's contested, and I should say that. The exact magnitude is debated. Some researchers think Reyes overstates it. But the core finding — that lead exposure increases later criminality — is now supported by a substantial body of evidence, including neuroimaging studies that show the exact brain regions affected. This is not fringe criminology anymore. It's mainstream.
So let's put policing and incarceration on the table. Because that's where the policy debate usually lives.
More police and more prisons do correlate with lower crime. But the effect sizes are modest compared to the demographic and environmental factors we just discussed. The mass incarceration experiment in the United States — where the incarceration rate quadrupled between nineteen seventy and two thousand — produced diminishing returns. Each additional prisoner yields less crime reduction than the previous one, because you're incarcerating less serious offenders and reaching a point where the incapacitation effect plateaus.
There's also a criminogenic effect, right? Prison can make people more likely to offend.
Yes. Prison is a school for crime in many cases, and it disrupts employment, family ties, housing — all the things that reduce recidivism. So the net effect of incarceration is the gross incapacitation effect minus the criminogenic effect. At some point those lines cross, and you're making things worse.
And policing?
More police officers on the street does reduce crime, particularly property crime. But the effect is modest. A ten percent increase in police numbers reduces crime by maybe three to five percent, depending on the study. It's real, it's worth doing, but it's not the lever that explains why Japan has one homicide per hundred thousand and Honduras has sixty.
What about economic factors? Poverty, inequality — these are the ones people reach for first.
The relationship is nonlinear, and that's where most public discussion goes wrong. Very poor countries do have high violence — that's true. But among middle- and high-income countries, inequality — measured by the Gini coefficient — is a much better predictor than absolute poverty. Two countries with the same GDP per capita can have wildly different homicide rates depending on how that income is distributed.
Routine activity theory explains some of this.
Right. The framework is that crime requires three things — a motivated offender, a suitable target, and the absence of a capable guardian. Inequality creates more attractive targets — there's more to steal — and it reduces informal social control because communities are more fragmented. You get neighborhoods where nobody knows each other, nobody watches the street, and the opportunities for crime multiply.
So if you're a policymaker looking at crime, you should probably care more about the Gini coefficient than GDP growth. That's not the usual political instinct.
It's not, and it's politically awkward because reducing inequality is harder than announcing a policing initiative. But the data is what it is. Now, the US crime decline of the nineties is the perfect case study here because it produced a dozen competing explanations, and we can weigh them against each other.
Give me the list.
Lead phase-out, which we covered. The economic boom of the nineties. The expansion of policing — more officers, community policing strategies. The abortion legalization hypothesis — Donohue and Levitt's controversial two thousand one paper arguing that legalized abortion reduced the number of unwanted children born into disadvantaged circumstances, reducing the pool of potential offenders twenty years later. And then there's the incarceration surge, the crack epidemic winding down, changes in drug markets. It's a crowded field of explanations.
And which ones does the evidence actually favor?
The lead hypothesis has gained the most empirical support over time. The abortion hypothesis is still debated — the mechanism is plausible, but the effect size is hard to isolate from everything else happening simultaneously. The economic boom probably helped, but crime had already started declining before the economy really took off. Policing changes — CompStat, broken windows — probably contributed at the margins but don't explain the magnitude or the timing.
The timing point is important. If crime started dropping in the early nineties, and leaded gasoline was phased out starting in the mid-seventies, that twenty-year lag lines up almost perfectly.
It does. And the crime decline wasn't just a US phenomenon. Canada, the UK, Australia, much of Western Europe — all saw similar declines on roughly similar timelines, and all phased out leaded gasoline around the same period. That cross-national pattern is hard to explain with US-specific factors like abortion law or policing strategies.
Let's shift to the misconceptions. What do people think matters that the data says doesn't?
Immigration is the classic. Across dozens of studies, immigrants commit crime at rates equal to or lower than native-born populations. This is one of the most robust findings in criminology — it's called the immigrant paradox. And it holds across countries, across time periods, across different immigrant groups. First-generation immigrants are particularly law-abiding.
Yet public perception consistently overestimates immigrant crime.
Dramatically. Surveys in the US and Europe consistently show that people believe immigrants increase crime, and the gap between perception and reality is enormous. The mechanism for the paradox is probably selection effects — people who uproot their lives to move to a new country tend to be ambitious, risk-averse in certain ways, and invested in not getting deported. The US-born children of immigrants have slightly higher crime rates than their parents, but still lower than third-generation Americans.
So the data says immigration reduces crime, if anything, and the public believes the opposite. That's a pretty spectacular failure of information transmission.
It's a failure of political incentives. Scapegoating immigrants for crime is electorally useful, and the media amplifies individual immigrant crimes in ways that distort risk perception. A native-born offender doesn't make headlines the same way.
What about drug use? That's another one people assume is a direct driver of violent crime.
The relationship is real but indirect, and it's weaker than most people assume once you control for poverty and policing practices. Drug addiction correlates with property crime — people steal to fund habits. That link is fairly well established. But the connection to violent crime is mostly through drug markets, not drug use itself. Illegal markets generate violence because participants can't use the legal system to resolve disputes. It's the prohibition, not the pharmacology.
So if you legalized and regulated, you'd expect property crime from addiction to remain, but the market violence to drop.
That's what the evidence from places that have decriminalized or legalized suggests. Portugal decriminalized all drugs in two thousand one and saw a drop in drug-related violence and HIV transmission. The property crime question is more complicated — addiction still drives acquisitive crime regardless of legal status.
And education? That's the other one people reach for.
More education correlates with lower crime. But the effect is mostly through delaying entry into the labor market and increasing the opportunity cost of crime — if you've invested years in education, you've got more to lose from a conviction. It's not that schools teach morality and that reduces crime. The causal effect is real but smaller than the raw correlation suggests, because the same factors that lead to more education — stable families, higher income — also reduce crime directly.
So keeping kids in school reduces crime while they're in school, and gives them more to lose afterward. It's a temporal displacement plus an incentive effect.
That's the mechanism. And it means that the quality of education matters less than the duration and the labor market it connects to. A vocational program that leads to employment reduces crime more than a liberal arts degree that doesn't, all else equal.
Now let's get to Daniel's stratification question. Do the same factors drive minor and serious crime? Because that's where a lot of policy goes wrong — people assume crime is crime.
The data shows a split. Property crime — theft, burglary, shoplifting — is more responsive to economic opportunity and policing. It's rational and opportunistic. When unemployment spikes, property crime tends to rise because the opportunity cost of stealing drops. When police presence increases, property crime drops because the risk of getting caught goes up. It responds to incentives in fairly predictable ways.
And violent crime?
Violent crime — homicide, assault — is more driven by demographics, alcohol, and environmental toxins. The youth bulge, lead exposure, alcohol consumption. These are deeper structural factors that don't respond quickly to changes in policing or economic conditions. In fact, economic downturns sometimes decrease homicide because people go out less, drink less, and have fewer of the risky social interactions where violence occurs.
Wait, say that again. Recessions reduce homicide?
Sometimes, yes. The mechanism is that bars close, people stay home, social interactions decrease. Homicide often happens in the context of social gatherings, nightlife, alcohol-fueled confrontations. If people are staying home because they can't afford to go out, those interactions don't happen. Property crime goes up because people are desperate, but homicide can go down.
That's counterintuitive enough that it's worth underlining. The same economic shock pushes property crime and violent crime in opposite directions.
Which means that crime policy that works for one category may fail for another. Broken windows policing — targeting minor disorder and petty crime — may reduce visible disorder and improve quality of life. But the evidence that it reduces homicide is weak to nonexistent. Conversely, reducing lead exposure or raising alcohol taxes may lower violence but won't do much about shoplifting.
So a mayor who runs on reducing "crime" without specifying which crime is essentially saying nothing. The policy toolkit for reducing burglary is different from the toolkit for reducing murder.
And some of the most effective tools for reducing murder — lead abatement, alcohol regulation — aren't even in the criminal justice budget. They're public health measures. That's the implication that makes a lot of people in law enforcement uncomfortable. The biggest levers for reducing violent crime may be at the EPA and the liquor control board, not the police department.
What about the middle category — robbery, aggravated assault? Where do those fall?
Robbery is interesting because it straddles both. It's property-motivated but involves violence or the threat of violence. The drivers are a mix — economic factors matter, but so do demographics and alcohol. Aggravated assault looks more like homicide in its drivers — alcohol is huge, youth demographics are huge, inequality matters. The more violence is involved, the more the structural and environmental factors dominate.
So if you plotted crimes on a spectrum from purely instrumental to purely expressive, the economic factors would dominate the instrumental end and the demographic-environmental factors would dominate the expressive end.
That's a good way to think about it. Shoplifting is mostly instrumental — it's about acquiring goods. Homicide is often expressive — it's about anger, status, impulse. And the policy implications are different at each end of that spectrum.
One more thing before we bring Hilbert in. We've been talking about developed countries mostly. Does the same framework hold in developing countries?
The framework holds, but the weights shift. In very poor countries, absolute poverty becomes a stronger predictor because the baseline of deprivation is so low that survival crime becomes rational for a larger share of the population. State capacity also matters more — if the police are corrupt or absent, the deterrent effect of law enforcement collapses. And in countries with very young populations — median age under twenty — the demographic driver swamps everything else.
So the policy prescription for reducing crime in Norway versus Nigeria are going to look very different, even if the underlying framework is the same.
Right. Norway should focus on alcohol regulation and inequality. Nigeria should focus on economic development, state capacity, and waiting for the demographic transition to age the population. Same analytic lens, different policy conclusions.
Hilbert.
Hilbert: You've got the mechanism right and the scale wrong.
Which part?
Hilbert: Lead and crime. It's more tightly linked than you described. I spent six months in two thousand three as a data entry clerk for the UK Home Office's Offenders Index. The old paper-based system, before they digitized. My job was to transcribe conviction records from the nineteen seventies and eighties into a database.
So you were literally reading the raw data.
Hilbert: Day after day. Birth year, offense type, conviction date. After about three weeks you start noticing patterns. The cohorts born in the mid-sixties — the peak lead exposure years in the UK — had conviction rates that jumped off the page. You could look at a birth year, estimate the lead exposure from the gasoline lead content that year, and predict the crime rate twenty years later. It was eerie.
And the Home Office knew this?
Hilbert: The analysts knew. This was two thousand three — the academic literature on lead and crime was just emerging. Reyes hadn't published yet. But the analysts could see it in their own data. They just didn't publish it.
Why not?
Hilbert: Politically awkward. If the crime drop is mostly an environmental accident — we stopped putting lead in petrol and twenty years later crime fell — then what do you tell the public? That all the tough-on-crime policies were theater? Not a great election platform. So the analysis stayed in internal memos.
They sat on it.
Hilbert: They didn't sit on it exactly. They just didn't shout about it. The Home Office published crime statistics every quarter. They never published the lead-crime correlation. It wasn't suppressed in the sense of someone ordering it buried. It was just... not pursued. The implications were inconvenient.
Were there other patterns you noticed? Things that didn't make it into the public analysis?
Hilbert: The alcohol link showed up in the assault data very clearly. Convictions for assault spiked on weekends, and the spike correlated with pub closing times. Again, not a surprise to anyone who's ever been near a pub at eleven p.m., but the magnitude was striking. Something like forty percent of weekend assault convictions had alcohol mentioned in the police report.
And the youth bulge?
Hilbert: That one was so obvious it wasn't even interesting. The conviction rate for males aged eighteen to twenty-four was something like five times the rate for males over thirty. The analysts treated that as background noise — it was so consistent it didn't feel like a finding.
I'm struck by the institutional reluctance to embrace the lead hypothesis. This is now fairly mainstream criminology. But you're saying the Home Office had the data to confirm it before the academics did.
Hilbert: Had the data, had the analysts who spotted it, had no incentive to publish. Because if crime is caused by lead and alcohol and demographics, then the Home Secretary doesn't get to take credit for the crime drop. And the opposition doesn't get to blame the government for the crime rise. The whole political theater of crime policy collapses.
And the flip side — if crime goes up, and it's demographics and environmental factors driving it, then the Home Secretary also can't fix it with a policing initiative.
Hilbert: Right. The levers that actually work are slow and unglamorous. Lead abatement takes decades. Changing alcohol culture takes generations. Demographics you just have to wait out. That's not what voters want to hear.
So the political incentive is to pretend the levers are policing and sentencing, even when the data says otherwise.
Hilbert: That's been the incentive structure for fifty years. It's not a conspiracy. It's just... people respond to incentives. The analyst who writes the memo saying "crime will drop regardless of what we do because we banned leaded petrol twenty years ago" is not the analyst who gets promoted.
Did you keep any of the data?
Hilbert: No. It was all on paper, and it stayed at the Home Office. But I remember the numbers. The nineteen sixty-four birth cohort in London had a conviction rate for violent offenses that was roughly double the nineteen seventy-four cohort. Same city, same policing, same laws. The only difference was the amount of lead in their blood as children.
Double. That's...
Hilbert: It's a lot. And that's just one city. The pattern repeated across the country. The analysts I worked with — there was a woman named Margaret who'd been there since the seventies — she said she'd watched the conviction rates rise and fall like a wave moving through the birth years. She knew it was environmental before anyone had a name for it.
She never published either.
Hilbert: She was a civil servant. Civil servants don't publish.
If you take one thing from this episode, it's that the factors that actually drive crime are not the ones we build policy around. Demographics, lead, alcohol — these are slow, structural, and mostly outside the criminal justice system. The things we argue about — policing levels, sentencing guidelines, immigration — have effects that are real but modest compared to the big structural drivers.
The stratification point. Property crime and violent crime respond to different levers. A policy that reduces burglary might do nothing for homicide, and vice versa. Anyone who talks about "crime" as a single thing is already missing the picture.
The open question that leaves me with — if the biggest drivers are demographic and environmental, things societies can't easily change in the short term, what does that mean for criminal justice policy? Should we be shifting resources from policing to public health — lead abatement, alcohol regulation, prenatal care? The evidence points that way, but the institutional incentives point the other direction.
As the global youth bulge shifts — it's moving toward Africa and South Asia now, as other regions age — we're going to see crime patterns shift in ways current policies aren't designed for. Countries that are aging rapidly, like Japan and much of Europe, will probably see continued declines in violent crime almost regardless of what they do. Countries with young populations will face headwinds that policing alone can't overcome.
The next time someone tells you crime is caused by one thing — immigration, poverty, video games — ask them what the cross-national data says. The answer is almost always more complicated than the slogan.
Thanks to our producer Hilbert Flumingtop. This has been My Weird Prompts. Find us at my weird prompts dot com, or email the show at show at my weird prompts dot com. We'll be back soon.
