The Value of a Statistical Life is not a good metric
TLDR: The VSL (Value of a Statistical Life) is often cited in Existential Risk Discussions. It’s not a good metric because it’s both used in specific circumstances and used inconsistently.
Often cited in discussions about existential risk is the US Government’s stated Value of a Statistical Life (VSL): somewhere in the ballpark of $10 million. For instance, Carl Shulman argues that even if longtermism doesn’t hold, reducing extinction of all humans alive today is still cost-effective when compared to this metric[1]. Will MacAskill recently argued in the Existential Risk symposium that this metric reflected ‘how much latent desire there is’ to reduce global catastrophic risk.
However, this number is mainly used to remove accountability for organisations in certain types of projects (although I’m not an expert on this).
For example, in infrastructure projects like roads, if it causes accidents, the government looks really bad for not prepping for it. This has no upper limit, so if the government wanted to do anything which adds risks to civilians, it would be impossible cost-wise. This number is like a theoretical cap for the government to say ‘Okay we’ve done our bit’.
The government does spend this amount in certain areas, but the value is much lower almost everywhere else. For instance, because plane crashes are a major deal for the public, the Department of Transport values human life at (and likely spends close to) $13.2 million per reduced death[2]. Meanwhile, the average fine for a workplace death is $12,000, more than 1000x less[3].
In short, this metric is fun to use—I’ve been guilty of it myself—but it’s not reflective at all of government spending and certainly not citizen attitudes towards the value of a life.
- ^
‘Carl Shulman on the Common-Sense Case for Existential Risk Work and Its Practical Implications’. n.d. 80,000 Hours. Accessed 18 March 2025. https://80000hours.org/podcast/episodes/carl-shulman-common-sense-case-existential-risks/.
- ^
‘Departmental Guidance on Valuation of a Statistical Life in Economic Analysis | US Department of Transportation’. n.d. Accessed 18 March 2025. https://www.transportation.gov/office-policy/transportation-policy/revised-departmental-guidance-on-valuation-of-a-statistical-life-in-economic-analysis.
- ^
‘Death on the Job: The Toll of Neglect, 2023 | AFL-CIO’. 2023. 25 April 2023. https://aflcio.org/reports/death-job-toll-neglect-2023
I think its an interesting question with many ways to estimate this value, looking at the market alone is one way but not definited. Perhaps some of them off the top of my head are
- WHO has used GDP per capita per year as a reasonable amount to spend to avert 1 DALY, so then perhaps 60x GDP per capita would work. In the USA this would be about 60 x 80k which is about 5 million for a life
- Look at what Government health systems are willing to spend on averting a DALY. In the UK NICE will spend something like 50kUS for a year of life, so that’s more like 3 million for a life
- You can ask people how much they think a life is worth to them. Not sure of the data there but GiveWell has done this pretty extensively.
- Then there are somewhat random US state department things like the 10 million cited above
Awkwardly many of these methods come up with different amounts in different countries depending on GDP and willingness to pay. This is important I think, as I don’t think we can just use huge numbers like 10 million from the Richest countries, when that’s not a practical reality for most of the world. A better approach might be to do something like divide by global GDP per capita (about 1/6th of the US) which would get us more n the 1.5-2 million realm per life at the upper end of estimations, and a lot lower (who knows) at the lower end....
https://forum.effectivealtruism.org/posts/ymEqipmiM3SLyQvaC/value-of-life-vsl-estimates-vs-community-perspective
This might be of interest—goes through the various substantive issues with VSL as a method
I’ve had drafts of this take lying around for years—really glad to see it out in the open!
I’d love to hear pushback from anyone who thinks it is still valuable.