Mortality Cost of Taxation

Here are some interesting questions:

1) If Japan raises $1 billion by increasing taxes on investment income, how many people will die as a result?

2) If Kenya raises $1 billion with a bundle of sales taxes, how many people will die as a result?

3) If the USA raises $1 billion with a tariff on Chinese goods, how many people will die as a result?

(If it is not obvious to you that these numbers are all more than zero, consider the basic facts that poverty kills people, and taxes take money from people. I like telling people that cost-benefit analysis is how to do "primum non nocere" for public health policy.)

In a competent civilization, there would be an academic sub-discipline devoted to these Mortality Cost of Taxation (MCT) questions. University professors would research MCT methodology and develop MCT best practices, and there would be a neutral MCT agency in all G20 countries that estimated the mortality effect of every tax policy under consideration. Then people would estimate how many lives would be saved by spending the tax money, and compare. (In reality, they would use a metric like the DALY, to account for both quality and quantity of life lost, but I am focusing on deaths in this discussion to keep things simple and vivid.)

In the world we live in, this academic discipline only exists in my head (as far as I know, and I would love to learn otherwise), and I suck at it. The best answer I can give is:

1) Ask a taxation or tariff economist how much the policy will impact GDP.

2) Divide that number by the country's GNI per capita.

3) Divide that number by 100.

So, in Japan (GNI/cap $40k), every $4 million in economic impact will kill someone, and in Kenya (GNI/cap $2k), every $200,000 in economic impact will kill someone.

My answer assumes that all taxation is basically the same, and assumes that the MCT is roughly twice the Value of Statistical Life (VSL). This is the best I can do, and it probably isn't wrong by more than an order of magnitude. 

VSL estimates come from looking at the wage differences between similar jobs with different occupational fatality rates, for example lumberjacks and gardeners. In the USA, for every $13 million or so in increased wages among we-assume-otherwise-identical jobs, one person dies on the job. In Kenya, many more people will die for $13M in increased wages.

If everyone responded to the tax increase by working a more dangerous job, then the VSL would be the correct estimate of the MCT. But this is obviously impossible at a societal level; the composition of jobs is roughly fixed and the tax change does not alter occupational safety standards.

If everyone was perfectly rational and made consistent tradeoffs between money and mortality risk, then the VSL would be the true estimate of the effects of people reducing their consumption in response to a tax increase. But we know that the all-cause mortality rate among people with dangerous professions is higher, so they cannot be turning their extra income into lower risk via consumption at the VSL rate.

For this reason, and for various other handwavey reasons I can't formally justify (e.g. budget constraints, short-run elasticities being lower, behavioral-economics effects, and vaguely gesturing at the literature on how people respond to income shocks), I assume that broad-based income loss from taxation or tariffs turns into mortality at roughly twice the VSL rate. Although I still use the VSL in all my work, because that is the current methodological standard and I can't cite anything better.

If you think you can do better, or know someone who has done better, or would like to collaborate to find ways to do better, please let me know.

PS Obviously the story becomes different with more targeted taxes. Taxing rich people causes less mortality (assuming that the tax incidence actually falls on them). Taxing harmful substances or things with negative externalities could have zero or even negative MCT, if it doesn't move much activity into a violent black market. I would assume that taxing positional goods causes roughly zero MCT, but there may be weird side effects I am not considering.

PPS It's a weird feeling to be so bad at something, and also the best in the world, because nobody else is even trying. I have looked, and can't find any discussion of this. Maybe I'm just missing it. But in all the discussion of VSLs and how to use them and how to derive them, nobody seems to justify the methodology by pointing to the harm of taxation, or to make the obvious point that taxation kills people so it would be nice to know if your policy is saving more lives than it kills.