Guiding Monetary Policy during a Productivity Boom: Evidence from the 1990s

Monetary policymakers are sometimes accused of being too quick to tighten monetary policy out of concern that a strong labor market or rapid economic growth will cause higher inflation. An often-cited exception occurred in 1996 when Federal Reserve Chairman Alan Greenspan resisted calls to raise the Federal Open Market Committee (FOMC) policy rate despite strong economic growth and evidence of a tightening labor market.

Greenspan perceived that advances in information technology had increased the productive capacity of the economy and thus would prevent inflation from rising even as the pace of economic activity quickened. Greenspan’s hunch was borne out as inflation remained stable and revised data later revealed that aggregate productivity growth had in fact risen significantly.Initial reported estimates indicated that the four-quarter percent change in labor productivity (output per hour) had slowed from 3.4% in the third quarter of 1995 to 0.4% in the third quarter of 1996, but revised data showed that productivity growth had increased from 1.6% in the third quarter of 1995 to 2.5% in the third quarter of 1996. In a conventional neoclassical growth framework, productivity is the sum of (1) capital deepening (fixed investment), (2) labor quality (skills and knowledge) and (3) total factor productivity, which can be thought of as nonlabor productivity gains that are due to new innovations or methods of production or a reconfiguring of supply chains.

AI and Productivity Growth

Greenspan’s prescience about technological progress and productivity growth, and the implications for monetary policy in the 1990s, may be relevant today. The recent development and rapid deployment of generative artificial intelligence (AI) and its associated innovations are widely expected to increase the nation’s productive capacity and potentially put downward pressure on inflation.

Although the effects of AI on aggregate productivity growth are unclear as yet, firms are increasingly discussing AI on earnings calls and expressing optimism that AI will generate large improvements in the productivity of their operations.See Serdar Ozkan and Nicholas Sullivan’s Aug. 26, 2025, On the Economy blog post, “Is AI Contributing to Rising Unemployment? Evidence from Occupational Variation.” Also, see the March 25, 2026, CFO Survey, “How Might AI Change the Workplace? Evidence from Corporate Executives.” Further, surveys indicate that AI tools are being adopted rapidly in the workplace. Across firms and industries, productivity appears to have increased in occupations in which AI use has become more prevalent.See the following On the Economy blog posts by Alexander Bick and his co-authors: “The State of Generative AI Adoption in 2025,” Nov. 13, 2025, and “Measuring AI Adoption among Firms: How You Ask Matters,” June 1, 2026.

The prospect of an AI-driven acceleration in aggregate productivity today suggests that a review of the 1990s experience could provide policy insights for other times. Accordingly, this blog post reviews some key facts and features of the episode.

Economic Conditions Leading Up to 1996

The first figure plots the consumer price index (CPI) inflation rate (headline and core) and the FOMC’s policy rate—the federal funds rate target—for 1990-2000.The level of inflation as measured by the personal consumption expenditures (PCE) price index was somewhat lower than CPI inflation but followed a similar trend. The FOMC focused primarily on the CPI measure in the 1990s. The FOMC formally announced a switch of emphasis to the PCE price index in a footnote embedded in the Feb.17, 2000, Monetary Policy Report to the Congress (PDF). The FOMC cut the policy rate (dotted line) in a series of steps from 8.25% in 1990 to 3% in late 1992, where it remained until early 1994. Meanwhile, headline inflation (solid line) declined from over 6% in 1990 to around 3% in late 1991, while core inflation (dashed line) fell from over 5% to just over 4%.

In February 1994, the FOMC began a series of rate hikes that boosted the policy rate to 6% by early 1995. The hikes were a preemptive move to head off a possible increase in inflation. The policy was evidently successful in that inflation remained stable around 3%. The FOMC subsequently lowered the policy rate in three steps to 5.25% in late 1995 and early 1996 in response to forecasts of a sharp slowing of growth in real gross domestic product.

On the real side of the economy, as shown in the second figure, the unemployment rate rose sharply during the 1990-91 recession and for several more quarters, reaching a peak of over 7.5% in 1992. The rate then fell steadily to around 5.5% in 1995, where it remained until mid-1996. With a 5.5% rate of unemployment, the labor market was widely considered as putting neither upward nor downward pressure on inflation.For example, at the July 2, 1996, FOMC meeting, Federal Reserve Board staff reported estimates of the nonaccelerating inflation rate of unemployment (PDF) (NAIRU) between 5.5% and 6%.

The Decision to Keep Rates Steady amid Falling Unemployment

By the middle of 1996, the unemployment rate was again falling, which was widely seen as evidence the labor market was overheating and risking higher inflation. By the second half of 1996, a “substantial minority” of FOMC members favored tighter monetary policy.See former Federal Reserve Chairman Ben Bernanke’s 2022 book, 21st Century Monetary Policy: The Federal Reserve from the Great Inflation to COVID-19, page 67. However, Greenspan convinced the FOMC to keep the policy rate at 5.25%. In Greenspan’s view, greater use of information-processing technology in the workplace was having two effects:

• Boosting productivity growth and hence the economy’s potential growth rate
• Discouraging workers from demanding outsized wage increases for fear of losing their jobs to automation. Contrary to Fed staff estimates, Greenspan believed the unemployment rate could fall farther without generating higher inflation.For discussion, see Bernanke (2022), pages 65-7, or former Federal Reserve Gov. Laurence Meyer’s 2004 book, A Term at the Fed: An Insider’s View, pages 125-29.

As Greenspan had predicted, inflation remained tame. As shown in the first figure, headline CPI inflation hovered around 3% before ticking up in late 1996 when energy prices rose, but core inflation continued to drift lower throughout 1996 and 1997. Meanwhile, the unemployment rate continued to fall, reaching 4.7% in December 1997.

Monetary Policy Remained Relatively Tight from 1996 to 1998

Although Greenspan deflected calls for a higher policy rate, the FOMC did not reduce the rate as inflation fell. Consequently, the ex post real rate rose to around 3% in 1997. At that level, the real rate was comparable to the peak reached during the period of preemptive tightening in 1994-95, and was considerably higher than it is today.Estimates of the neutral real policy rate today are generally lower than estimates for the 1990s. Thus, a comparison of ex post real rates in different decades does not necessarily indicate differences in the degree of monetary restraint.

The committee subsequently raised its policy rate by 25 basis points in March 1997, citing “persisting strength in demand, which is progressively increasing the risk of inflation imbalances developing in the economy that would eventually undermine the long expansion.” The FOMC maintained this stance until well into 1998, when it lowered the policy rate in response to the Asian debt crisis, Russia’s debt default and collapse of the hedge fund Long-Term Capital Management.

In policy discussions in 1999-2000, Greenspan and other FOMC members increasingly expressed concerns about inflationary pressures stemming from spending driven by wealth gains associated with higher equity prices of technology companies, a phenomenon that has also been noted recently in the context of AI-driven gains in equity prices.On the productivity-driven boost in equity prices and aggregate demand in the late-1990s period, see Richard G. Anderson and Kevin L. Kliesen’s 2012 article, “How Does the FOMC Learn About Economic Revolutions? Evidence from the New Economy Era, 1994-2001,” in Business Economics. On the AI-driven boost in equity prices, see St. Louis Fed President Alberto Musalem’s May 28, 2026, presentation “How Should the Promise of Higher Productivity Growth Change the Reality of Monetary Policy Today?” at the Reykjavik Economic Conference 2026.

Thus, while Greenspan’s prescient hunches about productivity growth and inflation likely prevented increases in the FOMC’s policy rate in 1996, policy remained relatively tight in real terms, and the FOMC did not move to ease policy until the period of global financial distress in late 1998.

The 1990s experience demonstrated how widespread adoption of a transformative technology can boost productivity growth and aggregate supply, and improve trade-offs between employment and inflation. AI is a new technology that could have similarly transformative effects, though as always, differences in underlying conditions and other forces can limit comparability across periods.

Notes

1. Initial reported estimates indicated that the four-quarter percent change in labor productivity (output per hour) had slowed from 3.4% in the third quarter of 1995 to 0.4% in the third quarter of 1996, but revised data showed that productivity growth had increased from 1.6% in the third quarter of 1995 to 2.5% in the third quarter of 1996. In a conventional neoclassical growth framework, productivity is the sum of (1) capital deepening (fixed investment), (2) labor quality (skills and knowledge) and (3) total factor productivity, which can be thought of as nonlabor productivity gains that are due to new innovations or methods of production or a reconfiguring of supply chains.
2. See Serdar Ozkan and Nicholas Sullivan’s Aug. 26, 2025, On the Economy blog post, “Is AI Contributing to Rising Unemployment? Evidence from Occupational Variation.” Also, see the March 25, 2026, CFO Survey, “How Might AI Change the Workplace? Evidence from Corporate Executives.”
3. See the following On the Economy blog posts by Alexander Bick and his co-authors: “The State of Generative AI Adoption in 2025,” Nov. 13, 2025, and “Measuring AI Adoption among Firms: How You Ask Matters,” June 1, 2026.
4. The level of inflation as measured by the personal consumption expenditures (PCE) price index was somewhat lower than CPI inflation but followed a similar trend. The FOMC focused primarily on the CPI measure in the 1990s. The FOMC formally announced a switch of emphasis to the PCE price index in a footnote embedded in the Feb.17, 2000, Monetary Policy Report to the Congress (PDF).
5. For example, at the July 2, 1996, FOMC meeting, Federal Reserve Board staff reported estimates of the nonaccelerating inflation rate of unemployment (PDF) (NAIRU) between 5.5% and 6%.
6. See former Federal Reserve Chairman Ben Bernanke’s 2022 book, 21st Century Monetary Policy: The Federal Reserve from the Great Inflation to COVID-19, page 67.
7. For discussion, see Bernanke (2022), pages 65-7, or former Federal Reserve Gov. Laurence Meyer’s 2004 book, A Term at the Fed: An Insider’s View, pages 125-29.
8. Estimates of the neutral real policy rate today are generally lower than estimates for the 1990s. Thus, a comparison of ex post real rates in different decades does not necessarily indicate differences in the degree of monetary restraint.
9. On the productivity-driven boost in equity prices and aggregate demand in the late-1990s period, see Richard G. Anderson and Kevin L. Kliesen’s 2012 article, “How Does the FOMC Learn About Economic Revolutions? Evidence from the New Economy Era, 1994-2001,” in Business Economics. On the AI-driven boost in equity prices, see St. Louis Fed President Alberto Musalem’s May 28, 2026, presentation “How Should the Promise of Higher Productivity Growth Change the Reality of Monetary Policy Today?” at the Reykjavik Economic Conference 2026.