Friday, 1 June 2018

Value of a New Zealander's life

Ages back, I'd run the usual income elasticity of the value of a statistical life drill to get a reckon on what Viscusi might say about the value of a statistical life in New Zealand. It looked like about $5m was the right figure.

But, time passes. And Viscusi last year put up an actual figure. The updated figure for America is now $9.6m (where the figure I keep in my head is $7m) and the value for New Zealand is now $6.885m (where the figure I'd kept in my head was $5m). They also use a better income elasticity measure. I'd been using the 0.5 figure that had come out of American work; they use 1.0 because it's more elastic in the international sample.

And so let us all update. VSL for New Zealand is about $6.9m. Conduct yourself accordingly. You're worth it.

And thanks to the student in my public economics course at Vic who wanted the Viscusi work I'd been talking about. That prompted me to find the much fresher estimates. I'll have to let the kids all know in next week's lecture that they're all more valuable than I'd previously thought.

The international VSL figures range from $45,000 to $18.3 million. Remember that when folks push for international harmonisation of safety rules and the like.

The Ministry of Transport survey-derived, inflation-adjusted VSL measures are getting increasingly out of date. I think it's pretty second-order whether the VSL figure used is $4.2m or the new figure since the main gains are using a standard number across policy areas, but it's probably getting to be time to update it.

Update: Ryan Murphy points me to some robustness worries on the VSL figures. Ionnidis, Stanley and Doucouliagos hit on power problems across a range of economic studies, and hit VSL here:
To understand the practical consequences of accounting for power in economic research better, it might be instructive to consider some specific areas. For example, of the 1,474 reported estimates of the employment elasticity of a US minimum wage increase (Doucouliagos and Stanley, 2009), 96% are underpowered and the median power is 8.5%. The weighted average elasticity of the 60 adequately powered estimates is -0.0113, less than one-tenth (6.5%) of the reported average (-0.19) across all of these 1,474 estimates. As a second example, consider the 39 estimates of the value of a statistical life (Doucouliagos et al., 2012), 74% are underpowered. The WAAP estimate of the 10 adequately powered studies is $1.47 million compared to the simple average of $9.5 million across all 39. Of the 110 reported price elasticities of residential water demand (Dalhuisen et al., 2003), 84% are underpowered. The weighted average elasticity of the 10 estimates that have adequate power is -0.1025, while the average across all 110 is -0.378. This means that for these three research areas, 94%, 84.5% and 72.9%, or more, of the average reported effects are likely to be biased.
Their critique is not particular to the areas here quoted. They looked at over 6700 studies across a range of areas and found substantial problems with underpowered studies.

Update 2: Another excellent econometrician, very well versed in metastudy work and who I trust, warns me not to rely on the power estimates in the above-cited metastudy. So I can go back to relying more heavily on the Viscusi numbers.

No comments:

Post a Comment