Daniel's been thinking about something that sounds like a contradiction. He lives in a country where the news is full of hostage rescues, terror alerts, military operations — and yet his day-to-day experience is that he feels pretty safe walking around. So he's poking at that gap between the headlines and the sidewalk. His first question: is his intuition correct that hostage situations are dramatically rare events, just cinematically loud ones? Second: does the incidence of crime actually track with how intensively a society is policed? And then he gets specific — he wants Israel versus the United States, not just on homicides but on the petty stuff too, pickpocketing, burglary. Is there any single international metric that lets you compare safety across societies? And underneath all of it: what actually makes a place safer — policing, or something else entirely?
That's five questions stacked inside each other. The hostage one is the hook, but it's also the easiest to dispatch, so let's do that first and then get to the meat.
Agreed. The hostage question is almost a category error — it feels like it should tell you something about a society, but it doesn't.
Right. The FBI runs something called the Hostage Barricade Database System — HOBAS — which tracked roughly two hundred to three hundred incidents per year in the United States. That works out to about zero point zero zero zero zero eight percent of the population involved annually. It's a rounding error. Globally, the UN Office on Drugs and Crime doesn't even break out hostage-taking as its own category — it gets folded into kidnapping statistics in most jurisdictions. So Daniel's intuition is dead on. These are statistically negligible events.
They dominate coverage because they have a structure that's built for television. A location, a perpetrator, victims, a ticking clock, a resolution. It's narrative crack.
And they tell you almost nothing about whether you're safe walking to the grocery store. A country could have zero hostage incidents and a sky-high burglary rate, or vice versa. So the real work starts when we move from the cinematic to the mundane — starting with the one crime statistic that's actually reliable across borders.
Homicide. A body is a body, and most countries count them.
Homicide data is the gold standard for cross-national comparison because the dark figure — the gap between reported and actual — is tiny compared to, say, sexual assault or theft. So let's put the numbers on the table. Israel's homicide rate, 2023 data from the Israel Police and UNODC, is roughly one point five per hundred thousand. The United States is about six point four per hundred thousand. That's more than four times higher.
It's hiding almost everything interesting. The US rate is driven by specific cities. St. Louis, Baltimore, New Orleans — these places have rates above fifty per hundred thousand. That's not a national crime problem, that's a hyper-localized one. Israel's rate is more uniform across its territory, though it does spike in Arab communities and during conflict periods.
So if you compare like with like — Tel Aviv versus New York City — what do you get?
Tel Aviv's homicide rate is about zero point seven per hundred thousand. New York City is around three point nine. Similar population density, both coastal cities, both have community policing and predictive analytics. And yet New York's rate is more than five times higher.
That's the number that makes you stop and ask what's actually going on. Because it's not like Tel Aviv has five times the police.
No, and that's where the policing question gets interesting. Israel has roughly three hundred and forty officers per hundred thousand people, including the Border Police. The US is about two hundred and forty per hundred thousand. So Israel does have more police per capita — but the relationship between police density and crime is not what most people assume.
More cops, less crime. It sounds like it should be linear.
It's not. Washington DC has over six hundred officers per hundred thousand — nearly double Israel's ratio — and it has one of the highest crime rates in the country. The mechanism isn't headcount. It's deployment strategy, community trust, and what those officers are actually empowered to do.
So Daniel's question — does crime vary with how well-policed a society is — the answer is... not in any straightforward way.
The word "well-policed" is doing too much work. It smuggles in assumptions about effectiveness. A better framing is: what kind of policing, deployed how, in what social context?
And that's before we even get to the petty crime question, which is where the data gets messy.
Homicide is clean because a dead body is hard to hide. A stolen wallet? Most people don't report it. Or they report it in some countries and not others. Or the police record it differently. So any cross-national comparison of pickpocketing or burglary is already swimming in noise.
Daniel specifically asked about the lesser stuff. What do we actually have?
The best apples-to-apples source is the International Crime Victims Survey, run by UNODC. For burglary, Israel is at roughly two point one percent of households per year. The US is about two point five percent. Those numbers are surprisingly close — much closer than the homicide gap would lead you to expect.
So on the serious stuff, Israel looks significantly safer. On property crime, it's basically a wash.
And then there's the perception layer, which is a whole different thing. Numbeo runs a crowd-sourced Crime Index — it's not victimization data, it's how safe people say they feel. Israel scores about thirty-one point five, which they classify as low crime. The US scores forty-eight point seven, moderate crime.
Perception data is slippery. It measures fear as much as incidence.
It also gets distorted by what's salient. If you live in a country with bus bombings, you might report feeling unsafe even if your actual risk of being mugged is near zero. Meanwhile, someone in a quiet American suburb with no terrorism but a high property crime rate might report feeling perfectly safe.
The security theater problem.
We'll get to that. But first, let me land the policing point with some counterexamples, because this is where the data really undermines the simple story.
Go ahead.
Japan. Zero point two homicides per hundred thousand. About two hundred police officers per hundred thousand — fewer than the US. Extremely low crime, extremely low police density. What Japan has is off-the-charts social trust and economic equality.
And the flip side?
South Africa. Forty-two homicides per hundred thousand. Three hundred and thirty police per hundred thousand — more than Israel, way more than Japan. And it's one of the most dangerous countries on earth by violent crime metrics. Extreme inequality, low institutional trust.
So policing intensity is almost a decoy variable. It correlates with crime, but it doesn't seem to cause low crime.
The causality probably runs the other direction. High-crime societies hire more police. The police don't necessarily make them low-crime.
Which brings us to Daniel's fourth question. Is there a single metric? One number you can look up that tells you whether a country is safe?
The honest answer is no. But there are attempts, and they're instructive in their failures. The most ambitious is the Global Peace Index, from the Institute for Economics and Peace. It aggregates twenty-three indicators — homicide, violent crime, political instability, militarization, incarceration rates, arms imports, the works.
And where does Israel land on that?
One hundred and thirty-fourth out of one hundred and sixty-three countries. That's the 2024 ranking. Very low.
Wait. Israel has one point five homicides per hundred thousand and it ranks a hundred and thirty-fourth?
Because the GPI heavily weights conflict and militarization. Israel scores terribly on those dimensions — for obvious reasons — and those drag the composite down. So you end up with a ranking that says Israel is less peaceful than, say, Nicaragua or Bangladesh.
Which is absurd if what you mean by peaceful is "can I walk home at night without getting stabbed."
The GPI is measuring something real — the presence of organized violence and military capacity — but it's not measuring what Daniel means by safety. It's measuring something closer to "how likely is this country to be involved in armed conflict."
So it's a peace index, not a safety index.
And those are not the same thing. The most commonly used single metric is the UNODC's intentional homicide rate. It's clean, it's comparable, and it's the least bad option. But it misses property crime, sexual assault, white-collar crime, domestic violence. A country can have a low homicide rate and be terrifying for women. Or have low violence and rampant corruption.
Numbeo's Crime Index is the crowd-sourced alternative.
Which captures the stuff homicide misses — people rate their fear of mugging, car theft, burglary, drug use, being attacked. But it's perception, not incidence. And perception is shaped by media, by recent events, by what your neighbors tell you. A bombing that killed three people will spike fear scores more than a burglary wave that affected three thousand.
Every index embeds a value judgment about what counts as safety.
That's the core insight. There is no neutral metric. Choosing to weight homicide over property crime, or violence over corruption, or perception over incidence — those are political choices dressed up as methodology.
So what actually makes a society safer? If it's not police headcount, and it's not captured cleanly by any single number, what are we left with?
The research converges on three factors that consistently outperform policing intensity. First: economic equality. The Gini coefficient — the standard measure of income inequality — correlates with crime rates more strongly than police presence across OEC countries. More unequal societies have more crime, full stop.
Second?
Social trust. The kind measured by surveys that ask "do you think most people can be trusted?" Countries with high generalized trust — the Nordics, Japan — have low crime regardless of policing levels. And this isn't just correlation. High-trust societies have more informal social control. Neighbors watch each other's houses. People intervene when they see something wrong.
And third?
Legitimacy of institutions. When people believe the police and courts are fair, they cooperate with them. They report crimes. They testify. They obey laws voluntarily rather than under threat. That's a completely different mechanism than deterrence through patrol density.
Israel is an interesting test case for all three of those.
It's fascinating. High social cohesion within communities — especially Jewish communities — but low trust between Arab and Jewish citizens. So you get a bifurcated safety experience. If you're Jewish and living in a Jewish neighborhood, you probably feel quite safe. If you're Arab and living in an Arab town with under-resourced policing, your experience is different. The homicide data bears this out — the spike in Arab communities is real.
And then there's the paradox you mentioned earlier. Israel is exposed to conflict, terrorism, military operations — and yet its interpersonal violence rate is low.
The counterintuitive finding. Security threats — terrorism, war — can actually suppress certain types of crime. They increase surveillance, community vigilance, state capacity. There's more police on the streets, more cameras, more bag checks. That visible security apparatus deters some kinds of opportunistic crime.
But that comes at a cost the Global Peace Index captures and the homicide rate doesn't.
Right. You're trading one kind of safety for another. Israel's low homicide rate is real. But the militarization, the surveillance, the erosion of civil liberties — those are also real. And they show up in the GPI ranking. The US has the opposite profile. Low conflict exposure, high interpersonal violence. Driven by inequality, gun availability, fragmented policing jurisdictions.
So Daniel's question — what makes a society safer — the answer seems to be: it depends on what kind of unsafe you're trying to prevent.
And for whom. Safety isn't one thing. It's a bundle of different risks, distributed unevenly across a population. A single number can't capture that.
The burglary clearance rate in Israel is under ten percent.
That's a real statistic, and it complicates the "Israel is safe" narrative. The police are heavily focused on national security and counter-terrorism. Property crime is not the priority. So you have this strange situation where the state is extremely capable at preventing bombings and extremely incapable at solving break-ins.
Which means the safety people feel from terrorism might not translate to the safety they feel about their apartment.
And that connects to something I want to bring in — the distinction between security theater and actual safety.
Security theater.
Visible, deterrent-focused measures that make people feel safer. Armed guards at mall entrances. Bag checks. Metal detectors. These work for terrorism because they raise the cost of an attack. But they do almost nothing for the kind of crime that actually affects daily life — burglary, car theft, domestic violence.
And in some cases, they might even create opportunities for petty crime.
That's... Actually a very specific insight. I want to come back to that. But first let me land the structural argument. The countries that are safe — by any measure — tend to have three things in common. Low inequality, high social trust, and legitimate institutions. Policing matters, but it's downstream of those factors. You can't police your way out of a trust deficit.
Japan has fewer police per capita than the US and almost no crime. South Africa has more police than Israel and is drowning in violence. The variable that explains the difference isn't the number of uniforms on the street.
It's whether people believe the system is fair. Whether they have a stake in social order. Whether they trust their neighbors. Those things are slow to build and fast to destroy.
And they're almost invisible to the kind of metrics we've been discussing. The GPI can't measure whether your neighbor will watch your house while you're away.
Which brings me back to the security theater point. In Israel, the security apparatus is everywhere. Soldiers with rifles on buses. Guards at every mall entrance. Bag checks at the cinema. That's not policing — it's deterrence theater. And it works for its intended purpose. It makes the cost of a terrorist attack very high.
But it doesn't stop someone from breaking into your car.
Or picking your pocket while you're distracted by the bag check.
That's... wait. Say more about that.
Think about the queue at a mall entrance. Everyone's emptying their bags, showing their ID, going through the metal detector. Their attention is on the security process. Their wallet is in their hand or their open bag. It's the most profitable thirty seconds in the building for a pickpocket.
So the security measure designed to keep people safe is creating the vulnerability it's supposed to prevent.
Not the same vulnerability. It's preventing a bombing and enabling a theft. Different threat models. But from the perspective of the person whose wallet just got lifted, the distinction doesn't feel very meaningful.
That's a perfect illustration of the gap between perceived safety and actual safety.
Hilbert: I timed it.
...Timed what?
Hilbert: The queue at the Dizengoff Center. Tel Aviv, two thousand three. I spent six months there as a security consultant. My job was to design the bag-check and metal-detector flow. Had a clipboard and a stopwatch. I timed how long it took people to get through security.
What were you optimizing for?
Hilbert: Throughput. The mall wanted people inside spending money, not standing in a queue getting annoyed. So I'm standing there with my stopwatch, and I notice something. The security guards were terrible at spotting actual threats. They were looking for weapons. But the real crime was happening right in front of them. Tourists getting their wallets lifted while they were distracted by the security theater.
So you're saying the queue at the bag check is the most profitable spot in the mall for a pickpocket.
Hilbert: I timed it. Thirty-seven seconds was the average dwell time with a wallet in hand. That's an eternity for someone who knows what they're doing.
That's a perfect real-world example of what I was getting at. The visible security apparatus deters one kind of threat and creates an opening for another.
Hilbert: Everyone talks about policing and security like they're the same thing. They're not. In Israel, the security apparatus is everywhere — soldiers on buses, guards at every entrance, bag checks at the cinema. That's not policing. That's deterrence theater. And it works for terrorism. It makes the cost of an attack very high. But it does almost nothing for the kind of crime that actually affects people's daily lives.
Burglary, car theft, domestic violence.
Hilbert: The police in Israel are actually pretty bad at solving property crimes. The clearance rate for burglary is under ten percent. So you have this weird situation where people feel very safe from terrorism but not particularly safe from having their apartment broken into.
That's not a contradiction — it's two different systems operating in parallel. The counter-terrorism apparatus is highly effective. The routine policing apparatus is under-resourced and deprioritized.
Hilbert: The queue I was timing had three armed guards and a metal detector. But nobody was watching the queue itself. I pointed this out to the mall manager and he said that's not what we're paying for.
Security theater is a product, and the customer is buying the feeling, not the outcome.
Hilbert: They were paying for the appearance of safety. The actual safety was somebody else's problem.
This connects directly to the structural factors we were discussing. If you invest in visible deterrence rather than in the slow work of building trust and reducing inequality, you get exactly this outcome. People feel safe in some dimensions and are vulnerable in others.
Hilbert: The pickpockets understood this better than the security consultants did. They didn't need a stopwatch. They just needed to stand in the queue.
Daniel's question — does more policing make a society safer — your answer from the Dizengoff Center is: it depends on what the policing is actually doing.
Hilbert: What it's not doing. Three guards at the door, zero guards in the parking lot. That's a choice.
It's a resource allocation choice that reflects priorities. And in Israel, the priority for decades has been counter-terrorism. That's not wrong — the threat is real. But it means other kinds of safety get shortchanged.
The metrics we use to compare countries don't capture any of this texture. The Global Peace Index punishes Israel for militarization without crediting the low homicide rate. The homicide rate doesn't capture the burglary clearance problem. Numbeo captures fear but not the specific texture of what people are afraid of.
Which leaves us with Daniel's deepest question — what actually makes a society safer — answered, but not resolved. We know policing intensity is a weak predictor. We know inequality, trust, and institutional legitimacy are strong predictors. We know security theater can create its own vulnerabilities. But how should we actually measure whether a place is safe?
For whom. That's the part that sticks with me. Safety isn't evenly distributed. Your experience of safety depends on who you are, where you live, what you look like, what language you speak. A single number can't capture that.
As AI-driven predictive policing and surveillance expand — and they're expanding fast in both Israel and the US — the gap between perceived safety and actual safety may widen. You can make people feel safer with cameras and algorithms while doing nothing about the structural conditions that produce crime.
Or you can make people less safe while convincing them the opposite.
Hilbert: The queue at the Dizengoff Center had more cameras by the time I left. Didn't change the pickpocket rate.
Of course it didn't.
This has been My Weird Prompts, with thanks to our producer Hilbert Flumingtop. If you have a weird prompt you want us to dig into — something about crime, safety, or the gap between perception and reality — email the show at show at my weird prompts dot com. We'll be back soon.
