Tradable vs. Nontradable: How Structural Shifts Are Redefining U.S. Productivity Paths

An IMF, Denison University, and Stanford University study finds the U.S. economy increasingly dominated by low-productivity nontradable sectors, widening the gap with faster-growing tradables and fueling inequality. It warns that while new technologies like AI could spark a productivity revival, their impact will hinge on broad adoption across both sectors.

The International Monetary Fund, in collaboration with Denison University and Stanford University, has released a comprehensive working paper, authored by Belinda Azenui, Sandile Hlatshwayo, and Nobel laureate Michael Spence. This research examines more than two decades of profound changes in the U.S. economy, revealing how a widening productivity gap between tradable and non-tradable sectors has reshaped employment patterns, wage dynamics, and the nation’s economic resilience. By 2023, non-tradable service sectors, such as healthcare, government, real estate, and hospitality, employed 77 percent of the workforce and accounted for 68 percent of real value added. In contrast, the tradable sector, comprising manufacturing, agriculture, mining, and competitive services like finance and professional consulting, posted far stronger productivity growth, nearly 3 percent annually from 1998 to 2023, compared to just 0.7 percent in nontradables, a divergence with far-reaching implications for overall growth and inequality.

From Hyperglobalization to Post-GFC Reshaping

The paper traces these dynamics to the hyperglobalization era before the Global Financial Crisis (GFC), when global value chains, offshoring, and labor-saving manufacturing technologies steadily eroded U.S. manufacturing employment even as output climbed. Gains in tradable services, including finance and consulting, offset only part of the job losses, resulting in flat overall tradable employment. After the GFC, however, the trend shifted. Manufacturing employment stabilized and even edged higher, while tradable services expanded more robustly. Despite these shifts, productivity growth, which had surged in the late 1990s and early 2000s, slowed sharply across most industries in the post-GFC decade. The slowdown was broad-based, affecting both low- and high-productivity sectors, and reflected a combination of underinvestment, reduced incentives for innovation, and labor reallocation toward slower-growing areas.

Pandemic Stress Test and the Power of Telework

The COVID-19 pandemic exposed vulnerabilities in the nontradable sector. Employment in hospitality, healthcare, and other in-person industries plummeted, with nontradable employment falling 7 percent in 2020 compared to a 4 percent drop in tradables. Tradable industries, especially knowledge-based services like computer systems design and finance, proved far more resilient, aided by their ability to operate remotely and tap into global markets. Using U.S. Bureau of Labor Statistics data, the authors demonstrate a strong correlation between a sector’s teleworkability and its post-pandemic performance in both employment and value added. Sectors with high remote work potential weathered the crisis with less disruption and in some cases, emerged stronger. Nontradable industries with low telework potential, such as arts, entertainment, and recreation, suffered deeper and more prolonged setbacks.

Breaking Down Productivity Drivers

A central feature of the study is its Divisia index decomposition of labor productivity growth into three components: within-sector gains, reallocation effects, and terms-of-trade changes. The analysis reveals that most aggregate productivity gains since 1998 stemmed from within-sector improvements, especially in the tradable economy, driven by technological advances and capital investment. Reallocation effects, by contrast, have generally been negative, reflecting the movement of labor from higher- to lower-productivity sectors. The terms-of-trade effect has often worked against tradables because prices in high-productivity industries like manufacturing and information have risen more slowly than the economy-wide average, reducing their weight in total productivity. Meanwhile, in nontradables such as healthcare and education, prices have climbed faster despite stagnant productivity, boosting their economic share without increasing output per worker.

Wages, Productivity, and the Shifting Balance

The relationship between wages and productivity has also undergone a marked shift. In the late 1990s, tradable sectors enjoyed higher wage convergence with productivity than nontradables, but by 2023, the reverse was true. This reversal partly reflects exceptional productivity gains in sectors like computer systems design, where wages have not kept pace, and long-run stagnation in manufacturing wage convergence, which fell from 71 percent in 1998 to 38 percent in 2023. In the nontradable sector, labor-intensive industries such as education, healthcare, and construction have relatively high wage convergence, while real estate sits at the bottom, with workers capturing only a fraction of their productivity value. These trends suggest that productivity growth alone does not guarantee proportional wage gains, particularly in highly automated or globally competitive sectors.

Digital Technology’s Potential to Close the Gap

The authors devote substantial attention to the transformative promise of digital technologies, particularly generative artificial intelligence. Large language models (LLMs) and AI-enhanced robotics could drive major productivity surges in knowledge-intensive industries, both tradable and nontradable. These tools, acting as “digital assistants,” have shown the ability to boost output while narrowing performance gaps between new and experienced workers, as in one case study where AI support increased customer service productivity by 14 percent, with the largest gains among less experienced staff. If such technologies spread widely, they could help reverse the chronic productivity stagnation in large nontradable sectors. However, the authors caution that benefits may remain uneven if adoption is concentrated in high-skill, high-productivity industries, potentially widening inequality and leaving the structural divide intact.

The paper closes on a note of cautious optimism. Geopolitical fragmentation, supply chain rewiring, and industrial policy shifts will likely alter the tradable–nontradable balance, potentially giving U.S. manufacturing a short-term lift but increasing long-term costs. Inflationary pressures could persist if supply-side constraints from demographics, trade restrictions, and global diversification remain unresolved. Yet the authors see a reasonable chance that new technologies, in digital, biomedical, and energy sectors, could by the end of the decade deliver a broad-based productivity revival. The decisive factor will be whether these innovations penetrate the large, labor-intensive nontradable industries that have long lagged behind. Ultimately, the trajectory of the U.S. economy will depend on how effectively policy, industry strategy, and technological diffusion align to bridge the widening gap between its two economic worlds.