“Every 1% unemployment goes up, 40,000 people die”. So said Ben Rickert in the movie The Big Short. Today is the 1 year old birthday of this blog (and incidentally my 19th) - as such, I thought I’d make it about the reason I started this blog i.e. my concern over the welfare consequences of economic crises. More than any other line in popular media, Rickert’s comment captures my sense for why business cycle fluctuations matter. But is it accurate? As it turns out, the book upon which the movie is based does not contain this claim, and the closest we get to a source is a 1981 book about corporate flight I couldn’t get my hands on. So I decided to figure out how mortality changes with unemployment myself.
A non-causal sketch
The simplest way to do so is just to look at the mortality rate for employed and unemployed individuals. A comprehensive meta-analysis of the effect of unemployment on mortality looked at 42 studies which covered1. It found that unemployment was associated with a 63% higher risk of death compared to being employed, having controlled for the necessary covariates. If we take the USA, the size of the civilian labour force based on the Bureau of Labor Statistics’s Current Population Survey was 164,455,000 people at the start of 2020, which we shall take to be the baseline2. In that case, a 1% rise in unemployment involves 1,644,550 people losing their jobs. The median age of the labour force was 41.9 years old in 20193. If we take the death rate for 35-44 year olds, this is given as 0.195% (3sf)4. So the excess deaths as a result of 1% higher unemployment would be 0.63 \times 0.00195 \times 1,644,550 = 2020. That means slightly over 2,000 people die as a result of a rise in unemployment by 1%. For context, US unemployment during the Great Recession peaked at 10% in October 2009, while it reached 14.8% during the COVID-19 recession in April 2020. In other words, the ballpark number of excess deaths from the recession alone were 20,200 and 29,900 deaths respectively. That is absolutely horrifying and should serve to provide some human context for the costs of unemployment.
To be clear, this isn’t an attempt to fully capture the harms of business cycle fluctuations or even just of unemployment. Nor is this even a particularly rigorous process of causal inference -it has obvious issues, and crucially I’ve accounted for no heterogeneity between demographic subgroups. But this is to make the hawks that worry too much about inflation a bit more uncomfortable - and to really hammer home what reducing actual people to mere unemployment statistics misses: the painful and unacceptable cost of lives lost. As James Tobin said, “there are worse things than inflation and we have them” - these are the worse things.
Roelfs et al. 2011↩︎
Murphy et al. 2021↩︎