Daniel sent us this one — he's asking about the riskiest jobs in the world, mentions Alaskan crab fishing as the obvious contender, and wants to know what else is out there that's unusually perilous. And look, I get why crab fishing is the poster child. Deadliest Catch, freezing water, giant waves, guys yelling on deck. It makes great television. It's also, statistically, not even in the top three by fatality rate per hundred thousand workers.
It's not. And this is one of those cases where the thing we all think we know — the thing that feels intuitively right — is just wrong. Not completely wrong, fishing is absolutely dangerous, but wrong in the ranking. The BLS just released the twenty twenty-five summary of the Census of Fatal Occupational Injuries, and logging comes in at eighty-two point one deaths per hundred thousand workers. Commercial fishing overall is at seventy-five point two. Aircraft pilots and flight engineers are at forty-eight point six. Roofers at forty-seven. The national average across all occupations is three point five.
A logger is roughly twenty-three times more likely to die on the job than the average worker, and nobody's making a reality show about that.
Nobody wants to watch a guy with a chainsaw in the Oregon rain for twelve episodes.
Which is exactly the problem. But before we get into the specific jobs, we need to talk about how we measure risk — because the numbers tell a very different story than the headlines.
And this is where most coverage falls apart. There are three ways to measure occupational danger. Fatality rate per hundred thousand workers — that's your annual risk if you're in that profession. Total number of deaths — that's the body count, the raw toll. And injury severity — which captures the non-fatal but life-altering stuff, lost limbs, chronic pain, the things that don't make the BLS fatality tables at all.
Depending on which metric you pick, a completely different job looks like the deadliest. Can you walk us through an example of how that plays out in practice?
Take commercial fishing versus trucking. If you sort by fatality rate, fishing looks terrifying — seventy-five deaths per hundred thousand workers. Trucking is around twenty-five per hundred thousand. So fishing seems three times more dangerous. But if you sort by total deaths, trucking killed over a thousand people last year and fishing killed maybe forty. So which one is actually more dangerous? It depends entirely on whether you're asking "if I take this job, what are my personal odds of dying this year?" or "which occupation produces the most corpses annually?
The injury severity metric flips it again.
If you sort by long-term health destruction, you're looking at firefighters with cancer rates fourteen percent above the general population, or first responders with PTSD and suicide rates two to three times the national average. A firefighter might survive thirty years on the job and then die of a cancer that never appears in any workplace fatality statistic. So the question isn't really "what's the riskiest job." The question is "which risk are we measuring, and why do we ignore most of them.
The media coverage skews toward dramatic, episodic risks. A crab boat sinks, six guys die in sixty seconds, that's a story. A truck driver falls asleep on I-Eighty and crosses the median — that's a Tuesday. Happens so often it's not news anymore.
There's a term for this in risk perception research — it's called the availability heuristic. We judge how likely something is by how easily we can recall examples of it. Plane crashes get wall-to-wall coverage, so people think flying is dangerous. Car crashes are a paragraph on page six of the local paper, so people think driving is safe. The same thing happens with occupational fatalities. A crab boat sinking is a dramatic, narratively satisfying tragedy. A roofer falling off a two-story house in suburban Phoenix is just... sad and routine. No helicopters, no Coast Guard rescue, no dramatic soundtrack.
The invisible carnage. So let's start with the job that tops the fatality rate charts every single year, and the one that nobody talks about at dinner parties.
Eighty-two point one deaths per hundred thousand workers in twenty twenty-four. That number has barely budged in twenty years. And the mechanism is almost always the same — something falls on you. A tree, a limb, a dead branch that snaps off a canopy hours after the initial cut.
That's the actual term. And it's not folklore — it's a recognized phenomenon in forestry safety. A branch gets hung up during felling, it looks stable, the logger moves into the drop zone to work on the next tree, and then a gust of wind or just gravity does its thing. Forty percent of logging fatalities in the Pacific Northwest in twenty twenty-three involved a struck-by incident from a falling tree or limb. OSHA published that report. It's not a mystery why this keeps happening.
Why hasn't it been solved? We've had decades of safety improvements. Hard hats, training protocols, mechanized harvesting.
Mechanized harvesting is actually the answer, and it's also the problem. In places where the terrain is flat and the trees are uniform — think pine plantations in the Southeast — you can use feller-bunchers, these enormous machines that grab a tree, cut it, and stack it, all with the operator inside a reinforced cab. Fatality rates are much lower there. But in the Pacific Northwest and the Mountain West, you're dealing with steep slopes, irregular old-growth timber, and conditions where machines just can't operate. So you still need a guy on the ground with a chainsaw.
That guy is often working alone, or with one other person who might be out of sight.
Out of communication range. And possibly hours from the nearest trauma center. The other big killer in logging is equipment rollovers. Skidders and crawlers on uneven terrain, no seatbelts in older machines, and if the machine goes over, the operator gets crushed. The fatality rate is so stubbornly high because the fundamental working conditions haven't changed — you're putting a human body next to enormous falling objects in remote locations. There's only so much a hard hat can do.
It's the Swiss cheese model but all the holes are lined up by geography.
That's actually a good way to put it. And here's a concrete case that illustrates the problem perfectly. In twenty twenty-two, a logger in southern Oregon — forty-seven years old, twenty years of experience — was felling a Douglas fir on a forty-degree slope. He made his notch cut, his back cut, the tree started to go, and then it hung up on a neighboring tree about thirty feet up. This happens all the time. The protocol is to use a winch or a wedge to bring it down from a safe distance. But the winch cable was back in the truck, a ten-minute hike each way. He made the calculation that thousands of loggers have made before him — he'd just give it a few more cuts at the base, get it moving. A limb from the neighboring tree, dislodged by the vibration of his saw, came down and killed him instantly. His spotter was two hundred yards away and didn't see it happen. It took forty minutes to get a medevac helicopter to the site.
That's not a guy who didn't know the rules. That's a guy who knew the rules and made a cost-benefit calculation in the moment.
Which is exactly how most workplace fatalities happen. It's rarely ignorance. It's usually a series of small, rational-seeming shortcuts that, in isolation, almost always work out fine. Until they don't. Now, fishing — this is where the public imagination lives, and it's not undeserved. Seventy-five point two deaths per hundred thousand workers. The Bering Sea crab fishery specifically is a subset of this, and it's worse than the average. But the thing that makes fishing uniquely terrifying isn't the rate, it's the speed.
The F/V Destination.
Twenty twenty-four, Bering Sea. Six crew members. The vessel was a hundred and ten feet long, which sounds big until you realize the waves out there can reach forty feet. The Destination was hit by a single wave that either broke over the stern or caught the vessel broadside — the investigation isn't fully settled — but what we know is that the boat capsized and sank in under sixty seconds. No EPIRB signal that lasted long enough to get a fix. The Coast Guard found debris and an empty life raft.
You're asleep in your bunk, or you're in the galley having coffee, and then you're in thirty-four degree water.
The survival time in water that cold is maybe ten to fifteen minutes before you lose consciousness. Your body goes into cold shock — you gasp involuntarily, and if your head is underwater when that happens, you drown immediately. If you make it past that, your extremities stop working within minutes. You can't climb into a life raft. You can't even hold onto debris.
Even if you get out of the boat, you're probably dead.
Getting out of the boat is the hard part. Crab boats carry these enormous steel pots — seven hundred pounds each, stacked on deck. When a boat starts to capsize, those pots shift. Hatches get blocked. The deck is a maze of cables and hydraulics that become death traps the moment the boat isn't upright. There's a reason the Coast Guard calls it the thirty-second window. If you're not on deck and clear of the vessel in half a minute, you're not getting out.
What does the crew actually do during those thirty seconds? Is there a protocol?
The protocol is to get to the life raft, which is supposed to deploy automatically when the vessel sinks. The problem is that if the boat capsizes fast, the raft can get tangled in the rigging or pinned under the boat. And even if it deploys correctly, you have to get into it. In thirty-four-degree water, with waves breaking over you, wearing boots and heavy rain gear that immediately fill with water and pull you down. The Coast Guard did a study where they put experienced fishermen in cold-water survival suits and asked them to climb into a life raft in calm conditions. It took most of them over two minutes. In a storm, it's functionally impossible.
Yet people sign up for this.
Because the money is extraordinary. A deckhand on a Bering Sea crabber can make fifty thousand dollars in a few months. For someone without a college degree in a remote coastal community, those are life-changing numbers. It's a calculated gamble — high risk, high reward, compressed into a short season.
Which brings us to the counterintuitive one.
Forty-eight point six deaths per hundred thousand. And this is where the aggregate statistic is deeply misleading, because "aircraft pilots and flight engineers" lumps together airline captains flying seven-thirty-sevens with a guy in a fifty-year-old Piper Cub dusting crops at a hundred and fifty feet.
Right — the crop duster is doing aerobatics below power line height, and the airline pilot is sitting in a chair that's statistically safer than your living room couch.
Commercial aviation in the United States hasn't had a fatal crash involving a major carrier in over fifteen years. The fatality rate for airline pilots specifically rounds to zero. But crop dusters, bush pilots in Alaska, helicopter medevac crews, firefighting pilots — these are the ones dying. The Swiss cheese model applies differently in general aviation because there are fewer layers. No co-pilot, no dispatcher, no airline safety department, often no air traffic control in remote areas. One mechanical failure, one weather misjudgment, one moment of spatial disorientation, and there's nobody to catch it.
Medevac helicopters have their own special category of danger.
Night flights into unprepared landing zones, pressure to accept missions in marginal weather because someone's life is on the line, and a safety culture that has historically lagged behind the airlines by decades. The NTSB has been pushing for terrain awareness warning systems and night vision goggles on medevac helicopters for years, and compliance is still spotty.
The pressure to fly because someone's dying is the exact thing that gets the crew killed.
It's the same psychological trap that plays out in fishing. The financial pressure to haul one more pot, the medical pressure to take one more flight — the risk is rationalized in the moment because the alternative feels like failure. There's a well-documented case from twenty twenty-three — a medevac helicopter in Oklahoma was dispatched to a car accident victim in a rural area. Weather at the destination was below minimums. The pilot had already flown two missions that shift and was approaching his duty limit. The dispatcher knew the weather was bad but the patient was a child. The pilot accepted the mission. The helicopter flew into a fog bank, the pilot lost visual reference, and the aircraft hit a hillside three miles from the landing zone. Three crew members dead. The NTSB report cited pressure to accept the mission as a contributing factor. And the thing is, you read that report and you can't even be angry at anyone in particular. Everyone was trying to do the right thing.
The system set them up to fail.
And then we get to roofing and construction.
Forty-seven per hundred thousand for roofers.
The mechanism is almost boring in its predictability. Roofers fall off roofs. It sounds absurdly simple, but the BLS data shows that falls account for the overwhelming majority of roofer deaths, and the "fatal four" in construction — falls, struck-by-object, electrocution, caught-in-between — account for sixty percent of all construction deaths annually.
We know exactly what kills construction workers. We've known for decades. And we still can't stop it.
The technology exists. Personal fall arrest systems — harnesses, lanyards, anchor points — are cheap and effective. The problem is compliance. Small residential roofing crews, especially in the Sun Belt where a lot of the work is done by immigrant labor, often skip fall protection entirely. It slows them down. The foreman doesn't enforce it. The general contractor looks the other way. And a guy goes off a two-story roof onto a concrete driveway.
How many of those deaths are undocumented workers?
Hard to get precise numbers because the BLS doesn't track immigration status, but multiple studies suggest that Latino workers, particularly foreign-born, have a fatality rate in construction that's significantly higher than the industry average. Language barriers, fear of reporting unsafe conditions, and the fact that they're disproportionately assigned the most hazardous tasks.
The fatality rate isn't just about the physical danger. It's about who has the power to say no to unsafe work.
That's the through-line in almost every dangerous occupation. The people dying are the ones who can't afford to walk away. There was a case in Texas a few years ago — a roofer, undocumented, twenty-three years old, working on a three-story apartment building. The crew had harnesses in the truck but the foreman said they were behind schedule and didn't want to waste time setting up anchor points. The kid fell. He left a wife and a two-year-old daughter. OSHA investigated, fined the contractor twelve thousand dollars, and the contractor declared bankruptcy and reopened under a different name six months later. That's the system working as designed.
Now, those are the jobs with the highest rates. But if we look at total deaths, a very different picture emerges — one that's happening on highways near you right now.
One thousand and five deaths in twenty twenty-four. That's the single highest total of any occupation by a wide margin. For comparison, commercial fishing kills about thirty to forty people a year total. Crab fishing specifically, maybe eight to ten. But we made a TV show about the crab fishermen and not the truckers.
Because a truck crash on I-Forty in Oklahoma is a local news story that lasts thirty seconds.
It's almost always framed as a traffic accident, not a workplace fatality. A truck driver crushed between a trailer and a loading dock — that's a workplace death. A truck driver who falls asleep and drifts into oncoming traffic — that's somehow just a car crash in the public mind.
The fatigue problem.
The FMCSA report from twenty twenty-four found that sixty-five percent of fatal truck crashes involved driver fatigue. And the electronic logging device mandate — which was supposed to fix this by enforcing hours-of-service limits — has not reduced the rate. The ELD tells you when you've hit your eleven-hour driving limit, but it doesn't stop a dispatcher from pressuring you to "make it work," and it doesn't stop you from driving tired within your legal window.
The eleven-hour limit loophole.
You can drive eleven hours in a fourteen-hour on-duty window. That sounds reasonable until you realize that those eleven hours might start at three AM after a lousy night's sleep in a truck stop parking lot, and hour ten is in dense urban traffic with the sun in your eyes. The regulation assumes all driving hours are equal. They're not.
How bad is the sleep situation for long-haul drivers, practically speaking?
There was a study that put sleep monitors on long-haul truckers for two weeks. The average driver got about five hours of fragmented sleep per night. Not five hours of good sleep — five hours total, interrupted by reefer unit noise, idling engines, parking lot lights, and the constant low-grade anxiety of cargo security. And then they're expected to pilot an eighty-thousand-pound vehicle through mountain passes and rush-hour traffic. The cognitive impairment from chronic sleep deprivation is comparable to being legally drunk. We've criminalized drunk driving but we've normalized exhausted driving, even though the impairment is similar.
The gig economy has created a whole new layer of this.
Amazon vans, UPS trucks, and especially the app-based food delivery people on scooters and bikes. A twenty twenty-five study from the University of Washington found that gig delivery drivers have a forty percent higher injury rate than traditional couriers. But they're not classified as employees, so their injuries don't show up in BLS data the same way, and they don't have workers' comp in most states.
The data we have is probably undercounting the real toll.
And the gig platforms have every incentive to keep it that way. If a DoorDash driver gets hit by a car, DoorDash's position is that he's an independent contractor and it's not a workplace injury. The hospital codes it as a traffic accident. It vanishes from the occupational safety statistics.
The invisible worker with the invisible death.
That brings us to the other invisible category — the psychological toll. Firefighters and police officers have lower fatality rates than roofers or loggers, but the long-term health outcomes are devastating. The twenty twenty-five Lancet study on firefighter cancer rates was a landmark. Fourteen percent higher risk of dying from cancer than the general population.
That's not from running into burning buildings once. That's cumulative.
It's the turnout gear. The protective equipment that saves them from burns is treated with PFAS — forever chemicals — that leach into their skin. It's the smoke they inhale at every call, even the minor ones, because modern houses are full of synthetic materials that produce carcinogenic compounds when they burn. It's the diesel exhaust from the truck bay. And then there's the mental health side.
PTSD and suicide.
Firefighters and police have suicide rates two to three times the national average. The number of firefighters who die by suicide each year now exceeds the number who die in the line of duty. And this doesn't show up in the BLS fatality tables at all. It's not counted as an occupational death, even though the causal link to the job is about as clear as it gets.
The BLS number for firefighter fatalities — whatever it is, maybe five or six per hundred thousand — is telling maybe a third of the story.
And this is the fundamental limitation of the fatality rate metric. It captures acute, traumatic deaths that happen on the clock. It misses the slow deaths — the cancers that show up twenty years after retirement, the suicides, the lungs destroyed by silica dust in construction, the backs ruined by years of loading trucks.
There's an analogy here that I think helps. It's like measuring the danger of smoking by only counting people who die in house fires caused by cigarettes. You'd get a number, and it would be real, but it would completely miss the lung cancer that kills orders of magnitude more people.
The BLS fatality tables are counting the house fires, not the lung cancer. And if we zoom out globally, the picture gets even starker.
Oh, it's not even the same conversation. In the United States, the average worker has about a one in twenty-eight thousand chance of dying on the job in a given year. In Qatar, during the World Cup construction boom, construction workers — mostly migrants from Nepal, Bangladesh, and India — had an estimated one in one thousand chance. That's twenty-eight times worse.
Twenty-eight times. And that's not because Qatari construction is inherently twenty-eight times more dangerous. It's because the regulatory framework is different.
It's what some analysts call risk arbitrage. Multinational corporations operate in countries with weak labor protections, minimal OSHA-equivalent enforcement, and a workforce that can't complain because their visa is tied to their employer. The same construction firm that follows safety protocols in Houston doesn't bother in Doha, because nobody's going to fine them and nobody's going to stop them.
The kafala system in the Gulf states makes this uniquely exploitative. Your employer sponsors your visa, which means your legal right to be in the country is contingent on your continued employment. If you complain about unsafe conditions, you can be fired. If you're fired, your visa is canceled. If your visa is canceled, you're deported. And many of these workers have taken on significant debt to pay recruitment fees to get the job in the first place. So the choice is: accept the unsafe conditions, or lose everything and go home in debt.
The workers know the risks.
They know some of them. But when the alternative is subsistence farming in rural Nepal, a construction job in the Gulf — even a dangerous one — is still the best option on the table. It's the same calculus as the crab fisherman in Dutch Harbor. The risk looks irrational from a comfortable distance. Up close, it's just the cost of feeding your family.
What do we do with all this? The data is clear, but the policy and personal implications are where it gets interesting.
The biggest takeaway for me is the risk perception gap. We massively overestimate dramatic, visible dangers — crab boats in storms, sharks, plane crashes — and we massively underestimate chronic, invisible ones. Trucking kills thirty times more people than crab fishing every year, but nobody's afraid of truckers. Construction falls kill hundreds of workers annually, and it barely registers in the public consciousness.
This has real resource-allocation consequences. OSHA's budget, its inspection priorities, its enforcement focus — all of this is shaped by political pressure, which is shaped by public perception, which is shaped by what makes good television.
Logging has the highest fatality rate in the country, and when was the last time you heard a politician talk about logger safety? It's not a voting issue. It's not a donor issue. It's a few thousand guys in rural counties that nobody campaigns in.
The other thing that jumps out from the data is that most of these deaths are preventable with existing technology. We know how to stop roofers from falling — harnesses work. We know how to reduce truck driver fatigue — stricter enforcement and better scheduling. We know how to protect firefighters from PFAS — alternative gear materials exist. The barrier isn't technical. It's cost and culture.
The Swiss cheese model makes this really clear. Every dangerous job has layers of protection — training, equipment, regulations, supervision. An accident happens when holes in all the layers line up. The logger's training told him to check for widowmakers, but he was tired. His hard hat was on, but a two-hundred-pound branch doesn't care about your hard hat. His supervisor was supposed to inspect the site, but the supervisor was covering three crews that day. The regulation said he should have a spotter, but the spotter was on the other side of the ridge.
Each layer has a hole, and the holes shift over time. Fatigue widens one hole. Budget cuts widen another. A dispatcher in a hurry widens a third.
Eventually the holes line up and someone dies. The insight from the Swiss cheese model is that you don't need to eliminate all the holes — you just need to make sure they don't all line up at the same time. Redundancy is the whole point.
For someone listening who works in one of these fields — or who employs people who do — what's the practical takeaway?
Know your fatal four. Every high-risk occupation has a small number of hazards that cause the vast majority of deaths. In construction, it's falls, struck-by, electrocution, and caught-in-between — those four account for sixty percent of fatalities. In logging, it's struck-by incidents from falling objects. In fishing, it's vessel disasters — capsizing, sinking, flooding. In trucking, it's fatigue and highway collisions.
If you're a roofer, the single most impactful thing you can do is wear your harness and make sure it's tied off to a proper anchor point. That one decision eliminates the biggest hole in the cheese.
If you're an employer, the highest-ROI interventions are not the expensive ones. Fall protection in construction is cheap. Fatigue management in trucking — actually respecting hours-of-service limits and not penalizing drivers who stop when they're tired — costs nothing in equipment and saves lives. Mental health support for first responders, which means reducing the stigma around asking for help and providing access to counseling, is dramatically cheaper than the cost of losing a twenty-year veteran to suicide.
The frustrating thing is that we've known all of this for decades. The BLS has been publishing these numbers every year. The OSHA reports are publicly available. The NTSB and Coast Guard investigations are meticulous. And yet the fatality rate in logging is basically flat for twenty years.
Because safety is not a technical problem. It's an economic and cultural problem. As long as the incentives push toward speed and production over safety — and as long as the workers bearing the risk don't have the power to push back — the numbers won't move.
One last thing that doesn't get talked about enough — as AI and automation start replacing the most dangerous jobs, the risk frontier shifts.
This is the open question that keeps me up. Autonomous trucks could eliminate driver fatigue deaths entirely. Robotic logging equipment could get human beings out of the drop zone. Drone delivery could take gig workers off the roads. But new risks will emerge. The people maintaining the autonomous systems. The cybersecurity vulnerabilities in a fleet of driverless trucks. The psychological toll of a workforce that's been displaced and has to retrain at fifty.
The physical risk becomes a systemic risk. Instead of one truck driver dying in a crash, you get a software failure that affects a thousand vehicles at once.
The data infrastructure to track those new risks doesn't exist yet. We're still using a fatality classification system designed for a world where workers died from falls and equipment accidents, not from algorithmic mismanagement and AI failures.
The riskiest job in the world might be the one you don't know is dangerous yet — because the data hasn't caught up.
And if there's one thing I want listeners to take from this, it's that the numbers are available, they're surprisingly clear, and they almost certainly contradict whatever job you think is the deadliest. Check the BLS CFOI tables. Look at your own industry. Know what's actually killing people, not what makes the best documentary.
Now: Hilbert's daily fun fact.
Hilbert: In the 1880s, researchers on the Yamal Peninsula observed that jellyfish polyps produce a faint acoustic signal during strobilation — the process by which they asexually clone themselves into stacks of baby jellyfish — and the frequency of this signal shifts depending on water temperature, effectively making the polyp a biological thermometer that sings at a pitch no human ear can detect.
A singing, self-cloning, biological thermometer.
I'll never look at a jellyfish the same way.
This has been My Weird Prompts. Thanks to our producer Hilbert Flumingtop. If you enjoyed this episode, leave us a review wherever you get your podcasts — it genuinely helps. We'll be back next week.
