When People Use The Wrong Risk Tool - Tornadoes
66 tornado reports were filed across multiple states on Friday. Insurance companies will use one set of tools to think about this. Homeowners in tornado zones need a completely different set.
The Insurance Company's Game
An insurance company writing policies across Kansas, Oklahoma, and Nebraska is playing with large numbers. They have thousands of policies. Any individual tornado might destroy 50 homes, but the company's portfolio is spread across hundreds of towns. They can model frequency distributions, set premiums based on historical averages, and build reserves that absorb variation over time. The math works because there are many potential events. A bad year gets averaged out over a decade.
The toolkit: diversification, statistical modeling, reinsurance, reserves. All of it depends on having enough policies that the law of large numbers kicks in.
The Homeowner's Game
A homeowner in Moore, Oklahoma faces small-number risk. One house. One shot. Either the tornado hits or it doesn't. There's no portfolio to spread the loss across. There's no "average outcome" across ten years — if the house is gone this year, the game is over.
The homeowner can't use the insurance company's toolkit. Diversification doesn't exist when you own one house. Statistical averages don't help when the outcome is binary.
The toolkit that does work comes from A Spy's Guide to Taking Risks: identify necessary conditions for the bad outcome, then remove or layer them.
What the Small-Number Toolkit Looks Like Here
For the tornado to kill you, several necessary conditions must align:
A tornado forms near you AND
It tracks toward your location AND
It doesn't dissipate before reaching you AND
You're in its direct path AND
Your structure can't withstand the wind speed AND
You're in the house when it gets hit
You can't control the first three. But you can attack the last three.
Remove a necessary condition: Move. If you're not in tornado zones, the risk goes to zero. Extreme, but it's the cleanest option for taking a necessary condition off the board.
Or build a safe room or reinformed shelter. Then layer low-probability conditions on top of bad outcomes: Now for the tornado to kill you, it must hit your house and breach the shelter and catch you outside the shelter when it hits. Each added condition multiplies the probability down.
Create substitutes for good outcomes: Have multiple evacuation routes, multiple warning systems (weather radio, phone alerts, neighbors), multiple places to shelter. Elevate OR-operators — if Route A is blocked, Route B still gets you out.
Build tripwires: Install a weather alert system that wakes you at 3 a.m. if a tornado warning is issued. The tripwire buys time to act before the risk becomes irreversible.
Why People Use the Wrong Toolkit
Most discussion in the media about tornado risk uses the insurance company's language. "The odds of a tornado hitting any given home are low." "Statistically, you're more likely to die in a car accident." All true — and all irrelevant to the homeowner making a one-time decision about whether to build a shelter.
The homeowner isn't managing a portfolio. They're managing a single shot event with binary outcomes. The large-number toolkit doesn't help them. Worse, it can mislead them into thinking low probability means they don't need to act.
The Practical Move
If you're in a tornado zone, you're working with small-number risk. Build a simple model of the necessary conditions that would have to align for you or your family to be killed. Then attack one of those conditions — eliminate it, or layer additional low-probability steps on top of it.
Don't rely on averages.
Averages are for people with many events.
To survive a tornado, you only have one.