Thread by @SteveCicala

There's an important paper on automation and inequality out today in AEJ:Macro.

It's elegant, it has kickstarted research on automation, and yet it's also under-appreciated.

So here's an appreciation thread:

Before Hemous & Olsen, we thought about skill-biased technological change (SBTC) as something that enhanced the productivity (and increased the wages) of high-skill workers.

Productivity growth among a subset of workers raises inequality.

This could explain a rising wage premium for high-skill workers, but there was a huge hole:

How to explain falling low-skill wage levels?

Another big gap was that technological innovation was a bit too Deus ex machina--computers arrived, labor or capital supplies shifted, etc.

Why did those things happen?

Hemous & Olsen propose an elegant explanation: a thermostat model, where there's a negative feedback loop between demand for low-skill (LS) labor and automation.

In the paper they show how this model can rationalize striking patterns in the labor share, wage premia, etc. in the US economy over the second half of the 20th century.

Prior models could explain parts of these dynamics, but would have to assume exogenous processes or insert little hacks to fit with the full picture.

Here's how the model works:

When wages are low, firms focus their resources on inventing new stuff. Those products require LS labor, and the growth of new products increases LS labor demand. LS wages rise.

As LS labor becomes more expensive, it becomes worthwhile to invest more resources in automating existing production (reducing the now-hefty wage bill) instead of inventing new products, which rely on that expensive LS labor.

How much LS work gets automated? It can't go too far because LS wages would fall so much that it is no longer to be profitable to automate.

So automation cools off the LS labor market, and LS wages feed back into decisions to automate.

That means LS wages may decrease temporarily, but not permanently.

Since LS wages grow in the long-run, automation becomes an important source of economic growth over time.

Pretty cool, right?

If you think, "Hey, that sounds familiar" you're probably right: in the past few years there's been an explosion in research on automation.

They often build on the work horse that Hemous & Olsen built in 2013.

Good to see it in print at last!