圣路易斯联储分析：过去20年行业劳动生产率繁荣与较低的生产者价格通胀相关

KEY TAKEAWAYS

Over the past 20 years, U.S. industries experiencing unusually strong growth in labor productivity have tended to see their producer price inflation decline relative to that of other industries.
This is consistent with the idea that gains in labor productivity can help ease the cost pressures that businesses face and ultimately slow increases on the prices they charge.
Industry-level producer price inflation tends to fall after a productivity boom begins, with the effects building over the following year. Producer prices dip 3.7 percentage points below the cross-industry average four quarters into a boom and stay negative for roughly six quarters after its onset.
The effect of AI-driven productivity growth on prices could appear first in industries that are quick to adopt the technology. How such productivity growth might affect aggregate inflation is less clear.

Recent advances in artificial intelligence (AI) have raised hopes that the U.S. economy may be entering a new era of rapid productivity growth. If AI delivers the productivity gains that many anticipate, firms may be able to produce more with the same amount of inputs, lowering production costs. Some of these cost savings should be passed through to consumers, potentially slowing price growth. But is that what typically happens when productivity booms?

Because economywide productivity growth and inflation are jointly shaped by monetary policy, aggregate demand and global commodity prices, we focus on cross-industry evidence to investigate this question. Specifically, we identify productivity booms within industries over the past 20 years and ask whether prices in industries with such episodes rose more slowly than those in other industries.

Measuring Productivity Booms and Busts

Our analysis uses quarterly labor productivityLabor productivity is an imperfect measure of productivity. It can reflect technological progress, capital deepening, changes in labor composition or other changes in production. Its impact on prices is therefore expected to be more muted than an increase in total factor productivity. data from the Federal Reserve Bank of Chicago that covers 85 industries and the period from 2006 through mid-2025.See Bart Hobijn, Martí Mestieri, Nicolas Werquin and Jing Zhang’s January 2025 Federal Reserve Bank of Chicago Economic Perspectives article, “Quarterly Industry-Level Labor Productivity Data for the U.S.” We then adjust this measure of productivity growth to account for persistent differences across industries and common changes in productivity across industries over time; that is, we strip out industry and time fixed effects, keeping residual productivity growth for each industry in each quarter. Next, for each industry-quarter, we cumulate the residualized productivity growth over that quarter and the subsequent three quarters, calling the industry-quarter a boom if the cumulated residualized productivity growth lands in the top decile of all observations and a bust if it lands in the bottom decile. Said another way, we define a boom using a four-quarter period in which an industry falls in the top 10% of labor productivity growth.

Since most of the industries in our sample produce intermediate inputs used by other firms, we match the residualized productivity growth to the U.S. Bureau of Labor Statistics’ industry producer price index (PPI). We compute the year-on-year industry-level inflation rate and subtract the cross-industry average. What remains is relative price inflation, i.e., a measure of whether an industry’s producer prices are rising faster or slower than the average inflation rate. This allows us to abstract from overall inflation trends, such as the 2021-22 inflation surge. In all, 164 productivity booms and 174 busts survived the match with price data. Productivity booms and busts are widespread across different industries and time periods.

The Inflation Response

The first figure below traces average producer price inflation from eight quarters before a productivity episode begins through 12 quarters after its start. To avoid counting the same episode multiple times, we marked the first quarter in which an industry enters the top or bottom decile of productivity growth as the start of the boom or bust, respectively. If an industry remained in the top decile for consecutive quarters, we counted it as a single episode. Thus, quarter 0 in the figure below represents the onset of a new boom or bust, not a quarter in an episode already underway.

As the figure illustrates, with the onset of a boom, producer price inflation dips 3.7 percentage points below the cross-industry average at four quarters into the episode and stays negative for roughly a year and a half. In the quarters before a boom begins, producer price inflation runs above the cross-industry average. Two potential explanations fit this pattern. One is that some productivity booms are preceded by strong demand, which initially puts upward pressure on prices. Another is that adopting a new technology may require first installing capacity, which raises input costs before any efficiency gain shows up.

Busts are nearly a mirror image. Producer price inflation peaks at 4.0 percentage points above the cross-industry average by the third quarter after the episode’s onset, with inflation pressures lasting seven quarters. Some of this delay in booms and busts is mechanical; year-on-year inflation in a given quarter reflects price changes over the prior four quarters, so a response that begins at the onset of a boom or bust shows up gradually.

The next figure takes a different cut. We sort every industry-quarter by productivity growth into 50 equal-sized bins and plot the average year-on-year inflation rate against average productivity growth within each bin. The resulting relationship is close to linear and clearly negative, with a slope of -0.22. This implies that an industry-quarter experiencing a 1 percentage point gain in productivity growth (after accounting for persistent differences across industries and common changes in productivity across industries over time) sees yearly producer price inflation fall by 0.22 percentage points relative to the average cross-industry inflation rate.

Some productivity episodes coincide with periods of technological change in industries. For example, labor productivity grew 6.5% above its average rate in computer and electronic products from 2006 to 2008, around the time of the arrival of the modern smartphone; 4.3% in data processing from 2012 to 2013, as cloud computing went mainstream; 8.3% in mining from 2014 to 2015, as shale drilling efficiency rose sharply; and 1.7% in healthcare from 2007 to 2008, as electronic health record adoption accelerated. Relative producer price inflation fell in each case, ranging from 1.2 percentage points to 6.7 percentage points below the cross-industry average.

What’s Next?

Periods of unusually strong labor productivity growth historically have been associated with slower producer price inflation at the industry level, with the effect building over the year after a productivity boom begins. This is consistent with the idea that productivity gains can ease firms’ cost pressures and ultimately put downward pressure on prices. The effects of AI on the economy are likely to unfold over a much longer horizon than the productivity episodes in our sample. Adoption of the technology may occur at different times across industries, generating relative price declines first among industries in which AI creates the largest productivity gains. Such movements could therefore become visible in industry-level prices well before they are apparent in aggregate inflation. Whether AI can ultimately keep aggregate inflation low will depend in part on how large and widespread AI-related productivity gains become.

Notes

Labor productivity is an imperfect measure of productivity. It can reflect technological progress, capital deepening, changes in labor composition or other changes in production. Its impact on prices is therefore expected to be more muted than an increase in total factor productivity.
See Bart Hobijn, Martí Mestieri, Nicolas Werquin and Jing Zhang’s January 2025 Federal Reserve Bank of Chicago Economic Perspectives article, “Quarterly Industry-Level Labor Productivity Data for the U.S.”