Comment by doctorpangloss | Hacker News Reader

Comment by doctorpangloss | Hacker News Reader

Parent Full thread

You can figure this for anything, and prepare to be shocked. The variables are the number of incidents per period, the population during that period, and the expected duration an incident could occur. For example, in San Francisco, there are about 4700 car related injuries and deaths per year; about 800000 people; and the life expectancy is high at 83. This gives you 41 years for a cumulative 25% chance to have been involved in your first car injury.

It is crazy. Everyone should be pissed off. Do this for guns, diabetes, etc. Think about the people you know, and how the further you venture into acquaintances, you know someone who has been arrested, injured by a car, injured by a gun, has diabetes, etc.

Some will be looking for math flaws. It’s an exponential model for time until first event, “constant hazard.” It’s actuarial. Nothing unorthodox. If you couldn’t calculate the durations until events insurance couldn’t work. The biggest factors are numerator, denominator and eligible period (ie rate of incidence). The biggest factors are NOT other things…

The reason you doubt this stuff, that you think it’s crazy, isn’t because the math is wrong. You believe very strongly that people involved in crimes and car crashes and guns have personal agency. The discount for panopticon approved safe driving is a pathetic 5-10%! If agency is all it was, auto insurance would be way cheaper.

This is all about the myth about what kinds of agency matter. You cannot be much of a “better” driver. You can’t be a “safer” criminal. Even if you never drove, due to the rate of injuries, it takes 41 years. I didn’t give the figure for causing injuries, just being involved in one, on purpose: we’re primed to assume it’s all about agency. Consider if there were only ONE criminal who commits the same rate of crimes: they’re still going to arrest a ton of people! Or ONE driver who hits everybody: you’re still going to be involved.

The math is easy to understand but people absolutely refuse to question their beliefs about agency.

> For example, in San Francisco, there are about 4700 car related injuries and deaths per year; about 800000 people;

800k permanent residents.

> In 2019, prior to the COVID-19 pandemic more than 26 million visitors travelled to San Francisco.

That changes your statistics quite a bit.

If you include all 26m visitors, the duration that an event could occur correspondingly gets much smaller, from 83 years to whatever the average stay is, which might be days. So it turns out to be about the same.

Anyway, you can Google “how does insurance modeling work” and see for yourself. There are whole teams of people who do this modeling. I am just sharing a simple and representative model, better than to an order of magnitude.

I don't think it does unless you assume that San Francisco roadways are on aggregate dramatically safer or more dangerous than other places.

Indeed. If we assume two days per visitor, that would be 140k additional people.

> For example, in San Francisco, there are about 4700 car related injuries and deaths per year; about 800000 people; and the life expectancy is high at 83. This gives you 41 years for a cumulative 25% chance to have been involved in your first car injury.

It doesn't account for the fact that people can be injured more than once. Taking that into account, the number would be closer to 50 years. It also makes agency more relevant if most of the injuries are concentrated in a smaller group of people (ex: bikers).

> The discount for panopticon approved safe driving is a pathetic 5-10%!

Well, in France, the maximum discount for safe driving is officially -50%, and the maximum penalty for unsafe driving is +250%, calculated based on your history of at-fault accidents. That's a 7x difference! Not only that but if you get into too many accidents, the insurance company may cancel your contract, usually forcing you to go for insurance companies that specialize in high risk customers, which are even more expensive. I don't know about the insurance system where you live, but a 5-10% difference seems crazy.

> 5-10% difference seems crazy.

Well. That’s what it is. It’s crazy only because of how deeply you believe in agency.

If you instead think deeply about the numbers, and then the fact that you are comparing two completely different driving environments, one thing becomes clear: the way the driving environment is designed and driving culture is administered in France might be much better and safer! Then, agency plays a relatively larger role, by definition, even if it is absolutely small everywhere. Whereas in America, such as in San Francisco and many other communities, a lot of injuries are probably attributable to the environment and culture. This is at least the expert consensus is, not just the consensus of insurance companies.

Reply on news.ycombinator.com