Discussion on Thomas Philippon's paper on TFP growth being linear

post by Arjun Yadav · 2022-04-30T13:17:07.878Z · EA · GW · 4 comments

Note: This post was written quite quickly and I'm not well versed in this subject matter. 

Thomas' paper here and Dylan Matthews' excellent write-up on it here.

I would love to spark some discussion on this: total factor productivity growth being linear in many developed countries, not exponential, could potentially be very scary.

Of course, as Dylan mentioned, TFP has issues. I believe the main critique is that, due to its simplicity, it can sometimes remain the same even after changes in technology and productivity.


Comments sorted by top scores.

comment by MichaelStJules · 2022-05-01T08:12:32.178Z · EA(p) · GW(p)

I think TFP should have a constant upper bound due to physical limits, but maybe we're unlikely to get anywhere near it in practice; I wouldn't know.

Separately, capital and labour growth are limited by exploring and exploiting space at a cubic rate, bounded by the speed of light in all directions.

So, growth is bounded above by a cubic function, assuming our current understanding of physics.

Replies from: Arjun Yadav
comment by Arjun Yadav · 2022-05-01T11:52:42.313Z · EA(p) · GW(p)

Good point! Though I think it shouldn't be difficult to figure out a lower upper bound, maybe an economist is working on that right now, depending on how actively researched this domain is.

comment by Lukas_Finnveden · 2022-05-09T18:45:56.877Z · EA(p) · GW(p)

There's an excellent critique of that paper on LW: https://www.lesswrong.com/posts/yWCszqSCzoWTZCacN/report-likelihood-ratios [LW · GW]

The conclusion is that exponentials look better for longer-run trends, if you do fair comparisons. And that linear being a better fit than exponentials in recent data is more about the error-model than the growth-model, so it shouldn't be a big update against exponential growth.

Replies from: Arjun Yadav
comment by Arjun Yadav · 2022-05-10T13:26:04.988Z · EA(p) · GW(p)

Great post! I was mainly concerned with the -values heading haha. I wonder if Thomas Philippon will follow up on all of the attention his paper received.