>but it seems almost childlike that the paper doesn't address the obvious confounding variable:

I thought that's the reason why they used "Evidence from Labor Market Shocks"? The idea is that when "Labor Market Shocks" (ie. mass layoffs) happen, the people who lose their jobs are somewhat random, so there isn't the confounding variable of low performers/sick people.

▲

AnimalMuppet 7 hours ago | parent | next [-]

I'm not sure that completely addresses the GP's point, because "mass layoffs" are still somewhat selective. You're more likely to lay off the people that you think aren't worth their pay, which would include those who are already suffering cognitive decline, those who are drinking enough that it's affecting their job performance, and so on.

	▲	littlexsparkee 7 hours ago \| parent [-]
		this paper uses a Bartik / shift-share instrument that looks at changes in sectors and the localized effects (exposure) due to concentration/reliance on that industry. It's an exogenous labor demand shock, not a mass layoff which frames the decision as made at the company level. Skills matter in this consideration but the point is to account for that statistically. https://en.wikipedia.org/wiki/Timothy_J._Bartik

▲

LeCompteSftware 8 hours ago | parent | prev [-]

It's not about "low performers / sick people," it's unemployment itself (especially sudden layoffs) making people more susceptible to substance abuse, regardless of their health when they're unemployed.

	▲	notahacker 7 hours ago \| parent [-]
		Whilst that's a perfectly valid hypothesis, they explicitly did test for relationships with rises in opioid misuse and didn't find evidence to support it. There isn't even consistent empirical data to show unemployment as a causal factor in aggregate increased drinking, never mind it being concentrated among recently laid off early onset dementia patients. The reality is unemployment influences some people to drink more and others to drink less, the circumstances of the unemployment and the demographics of the individual matter as do the patterns of the drinking, and even the statistical relationship between drinking and dementia is complex as statistically the lowest risk behaviour tends to be "moderate" rather than zero alcohol consumption