I’m wading into waters I don’t have complete mastery of here, but total factor productivity is a measure of economic growth attributable to innovation (“technology”, but technology here can be intangible such as process or organizational improvements.) If I remember my undergrad economics class correctly, this is really the only way an advanced economy can grow over the long term, although substantial underinvestment in physical and human capital by idiot politicians can certainly retard progress well below this upper speed limit. Collapse of the biophysical support system our human economy is embedded in is not considered, of course. Anyway, here is what U.S. TFP has been historically (Gemini, not fact checked):
| Historical Era | Average Annual TFP Growth | Key Economic Drivers |
|---|---|---|
| 1948 – 1973 (Post-War Golden Age) |
~1.9% – 2.1% | Post-war industrial scaling, interstate highway expansion, widespread electrification, and rapid technological diffusion across manufacturing. |
| 1974 – 1994 (Structural Slowdown) |
~0.5% – 0.6% | OPEC energy shocks, macroeconomic stagflation, structural reallocation from manufacturing toward services, and early-stage computing adjustment friction. |
| 1995 – 2004 (IT & Internet Boom) |
~1.4% – 1.7% | Widespread commercial adoption of enterprise software, personal computers, supply-chain logistics automation, and high-speed communications. |
| 2005 – Present (Secular Stagnation) |
~0.4% – 0.7% | Maturation of first-generation internet tech, structural expansion of low-productivity service sectors (healthcare/education), aging labor force, and slowing dynamism. |
Source: U.S. Bureau of Labor Statistics (Multifactor Productivity Series) & Federal Reserve Economic Data (FRED).
By the way, “dynamism” above refers to new business formation (startups) and competitive conditions, two things idiot politicians can also retard. (I use the word retard in its literal meaning of “slow down”.)
Okay, now, returning to the IMF World Economic Outlook, here is a paper they cite as a source of their assumptions on this:
and here’s an excerpt:
Acemoglu (2025) concludes that AI will lead to TFP gains of around 0.7 percent over the next 10 years, translating into an annual TFP growth increase of about 0.07 percentage points. This estimate hinges on 4.6 percent of tasks being replaced profitably by AI, and a 27 percent increase in productivity due to labor cost savings. 9 10
In contrast, Aghion and Bunel (2024) arrive at a much higher TFP growth estimate. Their estimate is based on three key parameters: the share of tasks exposed to AI in developed countries estimated at 60 percent; (ii) the share of exposed tasks for which it will be profitable to use AI estimated at 50 percent (Besiroglu and Hobbhahn, 2022); and (iii) labor saving costs induced by AI of 40 percent (Peng et al., 2023; Noy and Zhang, 2023; Brynjolfsson et al., 2023), which taken together imply a cumulative median 6.8 percent TFP growth over 10 years, or 0.7 per year (their range of estimates is 0.7 to 12.4 percent cumulative TFP growth over a ten-year horizon).
Considering this wide uncertainty, this paper uses two alternative benchmark assumptions regarding the AIdriven TFP impact, rooted in the simulations for the UK of Cazzaniga et al. (2024) and Rockall et al. (forthcoming), who find a TFP increase between 1.3 and 3.4 percent over a ten-year horizon. Rescaling these impacts by AI exposure and preparedness, using equation (1) delivers economy-wide values of 1.5 and 4.3 percent for the US. While the low benchmark is about double the impact predicted by Acemoglu (2025), it still represents a very modest TFP effect driven by AI. 11 The high benchmark is closer to (but still significantly smaller than) the proposed median TFP gain of Aghion and Bunel (2024). Figure 1 contains the values of TFP shocks under the low and high TFP growth scenarios across model regions and sectors. Higher overall TFP shocks are observed in regions with greater exposure and preparedness to AI, notably in the US and other AEs. Similarly, higher sectoral TFP shocks within each region are associated with greater sectoral exposure to AI, with the AI-intensive sector experiencing the largest impacts, followed by non-tradables and tradables.
Okay, let’s see if I am smart enough to make any sense of all the numbers above. The paper plays fast and loose with total growth as a percent over 10 years, and annual percent growth. This is a little hard for us mortals, and although I understand the equations involved, I am too lazy to do homework on a Sunday morning (this is also sometimes referred to as “labor productivity”, but it could also lead to plane crashes or collapsed bridges in a higher-stakes context). So I asked Gemini to do it of course – I always ask her to write the relevant equation, translate it to Python code, run the code, and show her work though, which I hope improves her reliability. Anyway, the ranges above translate to annual TFP growth increases over baseline of roughly 0.1-0.7% (first paragraph) or 0.1-0.4% (second paragraph). So adding this to our recent 0.4-0.7% rate would bring us to something in the range of 0.5-1.4% – Let’s just call it a 1% bump, and we are up to 1.1-1.7%, similar to the late 1990s “internet boom” but less than the pre-1973 “post-war golden age”.