Thread by @Noahpinion

1/I occasionally get neo-Luddites telling me that I'm like a climate denier because I say that automation isn't killing jobs or reducing pay. To them, the pernicious effects of automation are as clear and obvious and well-established as global warming.

Except they're wrong.

2/Neo-Luddites generally bring three types of "evidence" to support their case that automation is killing jobs.

The first is anecdotes.

"I know people who lost their jobs", "I've seen automation and it's scary", and so on.

These, of course, prove nothing.

3/The second kind of "evidence" Neo-Luddites bring is *forecasts*.

There are plenty of "studies" confidently telling us how many jobs will be eliminated by automation, or how many jobs are at risk from automation.

E.g.:
cnbc.com/2019/01/25/the…

4/These studies are all bogus. Basically what they do is go around and ask engineers whether they think this job or that job could be automated. That's a terrible methodology for making this sort of prediction.

bloomberg.com/opinion/articl…

5/First of all, engineers don't actually know what jobs they could automate. They might have an idea of what *tasks* they can automate, but tasks and jobs are different things. You can do different tasks but still have the same job, and even get a raise.

bloomberg.com/opinion/articl…

6/Second of all, these studies say nothing about overall labor market effects. Even if a certain type of job does vanish, other kinds of jobs get created, often as a result of the increased demand created by the increased output enabled by automation itself.

7/Our employment levels are near all-time highs, so apparently people who lose their jobs often get new jobs...but those studies about how many jobs will be lost to automation don't take this into account at all.

8/Third, predictions are hard, especially about the future.

Remember when we thought outsourcing would destroy white-collar jobs? Remember "The World is Flat"?

Well guess what. Didn't happen.

nytimes.com/2019/09/27/bus…

9/Finally, the third type of evidence Neo-Luddites bring are macro models.

For example: frbsf.org/economic-resea…

10/But it's pretty easy to make macro models in which the falling labor share -- or wage inequality, or wage stagnation, or whatever -- is caused by other stuff. Like market concentration, for example.

Don't take macro models as evidence!!

janeeckhout.com/wp-content/upl…

11/In the end, the Neo-Luddites have little evidence.

And they ignore or wave away the strong evidence against their hypothesis, such as the fact that studies find a positive correlation between IT investment and job growth at the company level.

epi.org/publication/th…

12/Or the productivity slowdown.

Slow productivity growth implies automation isn't proceeding as quickly lately as in the past.

But this just gets waved away and dismissed.

13/The Neo-Luddites are fired-up and aggressive and utterly convinced of the truth of their hypothesis, but they don't actually have the data to back it up. Instead they have to take it on faith.

Not a good way to do economic policy.

(end)