Hacker News Reader

Comment by keeda | Hacker News Reader

Labor productivity, even at a national level, possibly yes:

https://www.stlouisfed.org/on-the-economy/2025/nov/state-gen...

This is just ~2 - 4 years after ChatGPT launched, and despite very shallow adoption (only ~6% of all work hours.) As a sibling comment indicates, it took almost 2 decades for the Computer Revolution to be visible in national level statistics.

Also note this study was originally published in 2024, then revised in 2025, but this preliminary evidence has been around for a while, if people wanted to find it. It's even been posted to HN a couple of times, somehow it just doesn't get the attention you'd think it should get, even if it was just to poke holes in the conclusions.

View on HN

Adding to my other comment:

https://www.stlouisfed.org/on-the-economy/2026/jul/ai-produc...

From July 30:

> Since the release of ChatGPT in late 2022, artificial intelligence (AI) has been widely expected to raise productivity. But aggregate productivity data have so far offered a more muted signal: Utilization-adjusted total factor productivity grew only 0.07% over the four quarters ending with the first quarter of 2026

> When we feed these estimates into a standard aggregate production model, this suggests that generative AI may have increased labor productivity by up to 1.3% since the introduction of ChatGPT. This is consistent with recent estimates of aggregate labor productivity in the U.S. nonfarm business sector. For example, productivity increased at an average rate of 1.43% per year from 2015-2019, before the COVID-19 pandemic. By contrast, from the fourth quarter of 2022 through the second quarter of 2025, aggregate labor productivity increased by 2.16% on an annualized basis. Relative to its prepandemic trend, this corresponds to excess cumulative productivity growth of 1.89 percentage points since ChatGPT was publicly released

These long term data suggest that this "excess" remains bellow historical productivity growth

https://www.bls.gov/productivity/

https://www.bls.gov/productivity/images/pfei.png

The analysis bellow, more recent than the one you pointed to, is from May 2026, and an even stronger argument to support your position on the side of "computer technology investment caused a delayed excess growth in productivity". And as you can see at the end of my comment, they still write a very tentative conclusion.

https://www.frbsf.org/research-and-insights/publications/eco...

To be clear: I do not take a position. I think this is an open question, a very important one, and I am not fully convinced that the exponential growth in the investment on computer technology over the past 50 years has led to a corresponding gain in productivity, nor that it is entirely a drag and a mechanism for increasing firm size and driving asymmetric profitability concentrated in ever fewer firms as the increased concentration in the capitalization of American stock market index composition would indicate.

That said, the strongest case I have seen for the position that we are beginning to see these delayed gains is the letter I linked above, and it still takes care to conclude:

> As more data become available, it will be important to continue to monitor whether current patterns represent the early stages of a new era of booming productivity or merely a temporary uptick in an otherwise slow-growth environment.

> These long term data suggest that this "excess" remains bellow historical productivity growth.

That's partly because we are discussing productivity growth. Today's productivity growth is on top of the substantial productivity improvements that have been compounding due to past booms like the Computer and Internet one. So in relative terms the growth looks modest, but in absolute terms this is substantial.

Also that BLS chart is a bit unhelpful because it shows time periods covering multiple years and does not isolate the years after ChatGPT launched, which is what the St. Lous Fed looks at and finds interesting indications. Like currently productivity growth is 1.3 percentage points above what was forecasted just before ChatGPT was released. This discrepancy is not fully explained by other factors and lines up with other data sources related to the effects of AI.

The letter you linked is relevant, but it is trying to make a much broader point than I am. Note that:

1) it's asking whether we have entered a "high-growth regime" meaning a period of sustained productivity growth, and itself points out that it necessarily requires years to play out; and

2) its point of reference is the 90s when the computer revolution had truly kicked in after almost two decades of adoption starting in the mid/late-80's, during which any impact was famously hard to find: https://en.wikipedia.org/wiki/Productivity_paradox

So what is astounding is that the effects of the AI revolution may be visible in national-level economics data after only 2 - 4 years since the technology was introduced, and we're already wondering if we have shifted into a "high productivity growth era"!

I'm sorry I don't understand what you mean when you discriminate productivity growth from productivity improvement.

> So what is astounding is that the effects of the AI revolution may be visible in national-level economics data after only 2 - 4 years since the technology was introduced, and we're already wondering if we have shifted into a "high productivity growth era"!

I feel compelled to repeat the conclusion from the letter I linked. I used it as an example of how much of an open contention the productivity-from-computer-technology issue remains

I wasn't trying to differentiate between growth and improvement, I meant to use them interchangeably, apologies for the confusion. What I meant is the relative amounts in the BLS chart are prone to being misinterpreted because that growth is compounding and the time periods depicted don't correspond to the time period we are interested in.

I realize the findings from these studies are tentative; in such a short timeframe such conclusions have to be. But I don't really see much open contention regarding the key question here, which I think is "Has AI had an impact on national level labor statistics?"

E.g. the conclusion you quoted simply says that it is not clear if this is a temporary uptick or a sustained boom. But it agrees that there has been a significant positive impact on labor productivity already, even if the impact on TFP is more modest. Which is what the St. Louis study looking at survey data, and corroborated by various other data sources, finds too.

If the question is whether this is a sustained "productivity boom", I agree that we don't know that yet. But if the question is whether there has been any productivity impact at all, I would say there are multiple indications of that.

You are focusing too much on an informal letter. There are lots - I am talking dozens - of peer-reviewed papers discussing the so-called "productivity paradox"

There is a lot more to this discussion than one time series from the BLS, I have linked just 4 that are worth skimming elsewhere in this thread.

I am not trying to prove a point one way or the other. The topic interests me and there is a lot of analysis available on the problem of finding the productivity growth in economic data corresponding to the ever increasing investment in computer technology.

To me it is clear that it is an open problem, and the literature available is a great reflection on statistical methodology in economics.

> But if the question is whether there has been any productivity impact at all, I would say there are multiple indications of that.

That is indeed the question, and I have not found one paper that conclusively states that there is clear evidence or clear absence of a measured statistical effect.

Reply on news.ycombinator.com