More research on speeding and road deaths
Thanks to reader Christoper Short for alerting me to this NBER Working Paper by Orley Ashenfelter and Michael Greenstone which gives more information on speeding and road deaths. Most of the article concerns statistical pitfalls in estimation, but I’ll focus on the bottom line which is their estimate that the implicit value of the time saved per life lost as a result of relaxation of rural speed limits in the US in the 1980s was in the range $1.2 million to $3.2 million in 1997 values.
This sounds like a lot, but it’s lower than most estimates of the value people place on risks to life (for example, looking at stated willingness to accept risk, wage premiums for risky occupations, or costs of medical procedures that are generally accepted as cost-effective). I discuss this a bit more here and you can find more discussion by searching the site for “speeding”.
A couple of other points are worth noting.
First, while “number of deaths” is taken as a proxy for the danger of accidents there are numerous injuries (some severely disabling) for each death – the cost needs to take these into account. This probably doesn’t have a big effect on comparisons with estimates based on wage premiums, where injuries are treated in the same way), but is important relative to measures derived from stated preferences or from comparison with medical procedures.
Second, Ashenfelter and Greenstone value time saved at the average pretax wage, which is an upper bound estimate. Most studies treat time saved on non-work travel as being less valuable than paid labour time.
To sum up, the value of time saved by the increase in speed limits is, on these estimates, lower than the value of lives lost, implying that the policy reduces social welfare.
As Ashenfelter and Greenstone note, the outcome appears to reflect the political preferences of Americans, as mediated by the political system there. I have no problem in putting forward the hypothesis that this is an instance of government failure, either because participants in the process are not behaving rationally or because the process is failing to take proper account of preferences. I’m pleased to say that the Australian process seems to produce more rational outcomes.
fn1. This reminds me that I’m still hoping to resurrect the excellent comments threads on some of these posts