A Twitter thread by Noah Smith.

1/OK, let us talk about robots and jobs, once again!

Andrew Yang is a good guy, and I like a number of his ideas (including UBI and VAT). But on the threat of automation, the math just doesn't check out. Here is why:

https://www.bloomberg.com/opinion/articles/2019-11-20/robots-replacing-human-workers-isn-t-supported-by-data …

2/First piece of evidence that robots aren't currently taking human jobs overall: Employment levels are high right now. Prime-age employment rates are at or above their peaks in the mid-2000s and late-80s booms.

3/But what about wages??

Well, real median weekly earnings are going up, for the first time in a long time.

4/Note that this is true for real *hourly* earnings of production and nonsupervisory workers as well. And it's true across the wage distribution, unlike in earlier expansions.

5/Piece of evidence #3: Although many automation alarmists will cite that 2017 Acemoglu and Restrepo paper about how industrial robots destroyed jobs, that same paper found that IT investment overall had the OPPOSITE effect. It created jobs!

https://www.epi.org/publication/the-zombie-robot-argument-lurches-on-there-is-no-evidence-that-automation-leads-to-joblessness-or-inequality/ …

6/Piece of evidence #4: All of the countries with lots of industrial robots have very strong and competitive manufacturing sectors. The U.S., with far fewer industrial robots, lags behind.

Competitiveness is important. Robots help.

7/OK, but now let us turn to piece of evidence #5 - the slam dunk.

Labor productivity growth has slowed to a crawl since the financial crisis.

8/This directly contradicts the thesis that automation is accelerating.

Paul Krugman pointed this out the other day.

9/Andrew Yang replied with a thesis that he has been repeating a number of times since the summer: The idea that productivity is slow because automation is pushing humans into low-productivity jobs.

But the math for this thesis just doesn't work out.

10/Here is the math for why "humans being pushed into low-productivity jobs by automation" can't explain slowing aggregate labor productivity growth.

Anyone who wants to talk about automation and the labor market needs to understand this math!

11/Now, note that this story MIGHT ACTUALLY WORK for the 1980s!

Back then we had accelerating labor productivity and falling wages. Maybe it was partly because of the rise of machine tools in manufacturing?

Could be an interesting argument.

https://www.epi.org/productivity-pay-gap/ …

12/But even if automation clobbered manufacturing workers back in the 80s, it ain't doing so today.

13/Now, it's important to note, even though @AndrewYang isn't right about the present, he might be right about the past, and - more importantly - he might be right about the FUTURE.

Machine tools might have taken our manufacturing jobs, and AI algos might take our service jobs!!

14/So it's fine to propose UBI as an insurance policy against the POSSIBLE FUTURE AI takeover of service jobs.

Nothing wrong with that!!

15/BUT, the misplaced alarmism about robots is also fueling an anti-technology narrative in the U.S.

When most people hear "robots are taking our jobs", their first instinct is not "let's eliminate jobs", it's "let's protect our jobs".

And that means attacking technology.

16/"Automation is taking our jobs SO THEREFORE we need UBI" is a two-step reasoning process.

Most people are not as far-thinking or calm as @AndrewYang, and will only heed the first of those two steps.

Their reaction will be: SMASH THE LOOMS!

https://www.cleveland.com/pdextra/2013/01/original_luddites_were_weavers.html …

17/You can already see this disastrous idea in the Neo-Luddite proposals of Bill DeBlasio.

https://www.wired.com/story/why-american-workers-need-to-be-protected-from-automation/ …

18/@AndrewYang's policy ideas are good, but the justification from them is grist for the Neo-Luddites' mills.

And the math shows the robot takeover just isn't happening (yet).

So, let's all calm down. We've got time to think this one through.


https://www.bloomberg.com/opinion/articles/2019-11-20/robots-replacing-human-workers-isn-t-supported-by-data …