Guest Contribution: “Can Expert Forecasters See Manufacturing Changes Coming?”
Today we are fortunate to present a guest post written by N. Kundan Kishor (University of Wisconsin-Milwaukee).
Lessons from three decades of long-term predictions, and what they mean for the AI productivity debate.
Artificial intelligence is expected to reshape the economy, but how much and how fast? Professional forecasters provide one useful benchmark because they are forced to translate broad narratives about technology into specific, quantitative predictions about growth, interest rates, and inflation. The median 10-year productivity forecast in the Survey of Professional Forecasters (SPF) rose from 1.3% in 2023 to 1.8% in early 2026. Notably, ChatGPT was launched in November 2022, yet the first SPF survey after its release (February 2023) saw predictions. below their productivity outlook reached 1.3%. The uptick didn’t start until a full year later. This is a meaningful review, but it is much less of a project for AI optimists. The nominal interest rate rose from around 1.0% to 1.7%. Long-term inflation expectations have slowed. So who is right: the cautious forecasters, or the optimists betting on the revolution?
The disagreement reflects a deeper problem: major technological changes are surprisingly difficult to detect in real time. Manufacturing data comes in late, is updated frequently, and struggles to capture intangible investments, organizational changes, and quality improvements associated with new technologies. One way to test whether forecasters are behind the curve today is to study how they performed during the last major productivity debate. SPF has collected long-term productivity expectations since the early 1990s, and the history, which includes the rise and fall of the Internet economy, is revealing.
The 1990s: data hid the revolution
From 1992 to 1998, the SPF held its 10-year yield forecast at 1.5%. The projected growth over the next decade was 2.1–2.7%. That seems like a big miss, until you check what the data showed at the time.
In February 1997, the BLS reported that 10-year productivity growth was 0.7%. The first estimate for growth in 1996 was 0.8%. After years of revisions (new hours data, reclassification of software as investment, 1999 NIPA benchmark), that 0.8% became 2.7% (Jorgenson, Ho, and Stiroh, JEP 2008). The CBO faced a similar problem, more than doubling its 10-year projection from 1.15% to 2.7% between 1997 and 2001, when updated data became available. SPF’s 1.5% was actually there and hope compared to published statistics.
But a deeper pattern persists throughout the full sample. Figure 1 plots the SPF forecast against the following updated facts and obtained results for the next 10 years calculated from the BLS nonfarm business labor production index (FRED OPHNFB series). Forecasters track the latest closely, and often miss turning points. They raised expectations to 2.5% in 2004, just as the boom was winding down and saw growth fall to 1.3%. They cut them to 1.3% in 2023, just as output is likely to grow again.
Productivity and actual measurement
The standard macroeconomic theory is clear: higher trend productivity raises the marginal product of money and therefore the equilibrium real interest rate. When forecasters use structural models, their productivity and expected actual rate should be consistent.
For most of the past three decades, they didn’t. Prior to 2010, the correlation between the SPF production forecast and its actual implied level (BOND10 minus CPI10) negative: -0.54. Forecasters who expected higher production at the same time expected lower real prices. This is a bad sign.
Something changed after the Great Financial Crisis. From 2011 onwards, the correlation turned positive at 0.61. The geospatial debate, the long zero-low limit, and the apparent focus on r-star seem to have revised the way forecasters connect growth and rates. The latest increase in productivity expectations from 1.3% to 1.8% was accompanied by an increase in the real rate from 1.0% to 1.7%, modest but theoretically correct.
Productivity and long-term inflation: a surprising anchoring
If productivity matters in real terms, does it matter what long-term inflation expectations matter? The theory offers competing predictions. Higher productivity can reduce inflation through lower unit costs, or it can be neutral if monetary policy adjusts to maintain its target.
The SPF data provides a straightforward answer: long-term expectations of inflation are essentially unchanged from the outlook for productivity. The correlation between PROD10 and CPI10 is just −0.19, effectively zero.
Figure 3 makes the point. While the productivity forecast has risen between 1.3% and 2.5% over three decades, the inflation forecast has remained remarkably stable at 2.2–2.5% since the mid-1990s. The disinflation of the early 1990s (CPI10 falling from 3.7% to around 2.5%) was a one-time adjustment as inflationary management gained credibility. Since then, even the inflation shock of 2022 has not moved the CPI10. It reached 2.5% and is already back to 2.3%.
This is arguably the strongest evidence for inflation expectations to hold in the SPF data. Forecasters treat the Fed’s long-term targets as fully credible and view productivity as neutral in long-term inflation. Regardless of what AI does in the real economy, the SPF consensus is that it will not change where inflation stabilizes.
What does this mean for the AI era
The resemblance to the 1990s is hard to miss. Big technology investment, quick returns, always modest legal figures. Just as the IT revolution was underway in 1995 but the SPF didn’t move until 2000, the AI moment has so far produced a slow and cautious response. If the impact of AI comes first on services that are difficult to measure, data may be more challenging than during the IT revolution.
But the 1990s also teach caution about optimists. The narrative of the “new economy” continued for years before the data finally confirmed it, and not all of those predictions survived contact with reality.
The SPF data reveals a structurally consistent forecast: productivity gains (1.8%), modestly higher real prices (1.7%), and fully concentrated inflation (2.3%). If AI delivers even half of what its proponents expect, production rates and real value are likely too low, and the policy implications are significant. Otherwise, the current consensus will appear to be a compromise.
The lesson from three decades of product forecasting is not that professional forecasters are poor at their jobs. That’s because structural breaks are really hard to spot in real time. The question is not whether AI will transform the economy. Whether the data will allow us to see it when it does.
Data: Survey of technical forecasters, Federal Reserve Bank of Philadelphia; Nonfarm Business Sector Labor Productivity (OPHNFB), Bureau of Labor Statistics via FRED.
A review of real-time data discussed in Jorgenson, Ho, and Stiroh (2008), “A Retrospective Look at the US Productivity Growth Resurgence,” Journal of Economic Perspectives, 22(1), 3–24.
This post was written by N. Kundan Kishor.
