Thursday, September 13, 2012

Standard unit of risk

The Massachusetts Department of Public Health has determined that Amherst is at "High" risk of mosquito-borne illnesses, and recommends that we all stay inside at dusk.

Yeah... but I like going outside at dusk.

The incentives for the Health Department are to exaggerate the risks, because when somebody gets Eastern Equine Encephalitis and dies they can (correctly) point to their High Risk Warning and say "not our fault, we told you so."

But maybe the risk really is high, and we ought to slather ourselves with DEET before walking downtown for dessert.

I wish there was a standard, easy-to-understand unit of risk, so the Heath Department could say something like "We estimate spending an evening in downtown Amherst with no mosquito repellant is 4-10 RUs (risk units)."

Where 1 risk unit is something we can all understand-- maybe "the risk associated with driving 100 miles on the highway."

Or "1 risk unit is the overall risk of dying in an accident in any given year."

I don't know what the right baseline would be; it doesn't really matter what is chosen, what matters is making it easy to compare the risks of things relative to each other.  The Health Department says we're at High risk... but High compared to what?


ripper234 said...

Interesting, posted to Quora.

BTW, you should really join the site, it's the best site in the world (perhaps bitcointalk comes second :)).

If you do, please follow me at

AlexKG said...

I was listening to a Radiolab about pain recently, and apparently there have been efforts to create a standard unit of pain, as well. Strikes me as similar.

Anonymous said...

What about just simple percentage odds?

"We estimate that for this week the average person outside at dusk will have a X% chance of infection per minute, compared to the normal Y% per minute."

The thing about risk is that different events will cost a different amount for each individual.

Anonymous said...

This is the folly that got us into the financial crisis: the illusion that risk is a scalar, one-dimensional thing.

I have bad news for you: it isn't.

Theo said...

Risk is the combination of the probability that an unwanted event occurs and the severity of that event.

The severity seems to be high (people can die) but is the same for all locations, so it's a question of occurrence. Frequency of bites per outside hour is a one-dimensional thing. You could even compare this to the frequency of having an accident per car hour. While there are probably somewhat suitable statistics for car accidents, there are surely none for mosquito bites. And without probabilities for the base events, the overall frequency remains unknown.

Roger Browne said...

The standard unit of risk is the micromort, which represents a one-in-a-million increment in the risk of death.

When faced with an economic choice, people are willing to pay (on average) somewhere around $50 for each micromort reduction in risk.

KLP said...

@Roger: The micromort only measures risk of death where the consequence, death, is constant and only the probability of that outcome varies. But when death isn't the outcome, that doesn't mean that there isn't an unfortunate consequence. Really, the micromort is just a probability unit, rather than a risk unit.

So, a universal risk unit would have to somehow include and balance differently unfortunate consequences. Death could have a coefficient of 1, long-term/permanent coma/brain-damage 0.9, total body paralysis 0.8, etc. I realize that there's some subjectivity there, but with some crowd-sourcing we might arrive at some acceptable coefficients. So, for a given activity or change in conditions, the universal risk unit would measure the sum of the (change in) probabilities of any of those consequences each times their respective coefficients.

Sound good?