AI’s Wage Impact: Key Elasticity Thresholds Revealed in New Study
A groundbreaking study published in the Journal of Industrial Technological Economics has uncovered critical thresholds in the relationship between artificial intelligence (AI) and labor markets, offering new clarity on when automation begins to suppress wages rather than enhance them. The research, conducted by Hao Lixiao and Lv Rongjie of the School of Economics and Management at Hebei University of Technology, challenges conventional narratives about technological progress and employment by introducing a nuanced framework that differentiates between two distinct forms of AI: labor-replacing AI and capital-enhancing AI.
For decades, economists and policymakers have debated whether technological innovation ultimately benefits or harms workers. The prevailing historical view has been optimistic—each industrial revolution, from steam power to digital computing, has disrupted certain jobs but ultimately created more opportunities and raised living standards. However, the rapid advancement of artificial intelligence in the 21st century has reignited concerns about a new era of “technological unemployment,” where machines not only replace manual labor but also encroach on cognitive and professional tasks once thought to be immune.
The new study moves beyond this binary debate by developing a sophisticated economic model that accounts for the dual nature of AI in production systems. Unlike previous research that typically treats AI as a form of capital, Hao and Lv argue that AI can function in two fundamentally different ways: as a direct substitute for human labor—such as robotic process automation in accounting or AI-driven customer service chatbots—or as a productivity-enhancing capital input, like machine learning algorithms that optimize supply chains or predictive maintenance systems in manufacturing.
This distinction is crucial. When AI acts as labor, it competes directly with human workers, potentially driving down wages through substitution. But when AI functions as capital, it can complement human labor, increasing overall productivity and, under the right conditions, boosting wages. The outcome depends not on AI itself, but on the underlying economic structure—specifically, the elasticity of substitution between different production factors.
The researchers introduce a multi-factor, two-level Constant Elasticity of Substitution (CES) production function to model this complexity. In their framework, the economy is divided into two broad aggregates: “broad labor,” which includes both human workers and labor-type AI, and “broad capital,” which encompasses traditional capital and capital-type AI. The interaction between these aggregates, and the substitutability within them, determines how AI adoption affects wages.
One of the study’s most significant contributions is its identification of specific elasticity thresholds that determine whether AI adoption leads to wage gains or losses. Using the Hauthakker model—a method that treats production as a collection of heterogeneous firms or “cells” with varying input requirements—the authors analyze labor market data from China and the United States to calculate the critical elasticity values at which AI begins to exert downward pressure on wages.
Their findings reveal that in China, when the elasticity of substitution between human labor and AI exceeds 1.8649, further AI adoption leads to a decline in average wages. In the United States, the threshold is slightly higher, at 1.900. These numbers represent a tipping point: below them, the productivity gains from AI outweigh the displacement effects; above them, the substitution effect dominates, leading to what the authors term “technology replacing people”—a phenomenon where automation reduces labor demand and suppresses compensation.
Even more striking is the discovery of an absolute threshold beyond which wage suppression becomes inevitable, regardless of current wage levels. In China, if the elasticity exceeds 2.35, AI adoption will always lead to lower wages. In the U.S., that threshold is 2.73. These values suggest that there is a point of no return in the automation process, where the structural shift in production is so profound that it cannot be offset by productivity gains or market adjustments.
What drives these differences between the two economies? The study points to three key factors: industrial structure, education systems, and overall skill levels in the workforce. China’s economy remains heavily oriented toward labor-intensive, low-end manufacturing, where AI can easily replicate routine tasks. In contrast, the U.S. economy is more concentrated in high-value, knowledge-intensive sectors, where AI often complements rather than replaces human expertise.
Moreover, the U.S. education system tends to emphasize critical thinking, creativity, and adaptability—skills that are harder to automate. American workers are more likely to possess general cognitive abilities that allow them to work alongside AI tools rather than be replaced by them. In China, the educational focus has historically been on vocational training and specialized skills, which, while effective for rapid industrialization, may leave workers more vulnerable to automation in the face of disruptive technologies.
The data supports this interpretation. According to the Organisation for Economic Co-operation and Development (OECD), high-skilled workers make up about 65% of the U.S. labor force, compared to a much lower proportion in China. AI tends to substitute for low- and medium-skilled labor while complementing high-skilled workers, reinforcing existing inequalities and creating a polarized labor market.
The implications of these findings are profound for both policymakers and business leaders. The study suggests that blanket promotion of AI adoption may be counterproductive, especially in economies with large populations of low-skilled workers. Instead, governments should adopt a differentiated strategy based on the type of AI and the structure of their industries.
For firms with high concentrations of advanced human capital—such as tech companies, research institutions, or financial services—both labor-type and capital-type AI should be encouraged. In these settings, AI can amplify human capabilities, leading to innovation and higher productivity. However, for firms dominated by routine, manual, or procedural work, the introduction of labor-type AI should be approached with caution. The study recommends prioritizing capital-type AI that enhances existing processes without directly displacing workers.
Another key policy implication is the urgent need to reform education and training systems. As AI continues to evolve, the ability to adapt, learn new skills, and work collaboratively with intelligent machines will become increasingly valuable. Universities and vocational schools must shift from narrow, task-specific training to broader, interdisciplinary curricula that foster creativity, problem-solving, and digital literacy.
The study also highlights the importance of lifelong learning and retraining programs. Workers in industries most vulnerable to automation—such as manufacturing, transportation, and administrative support—should have access to affordable, high-quality training that equips them with skills relevant to the AI-driven economy. This is not just a matter of economic efficiency but of social equity. Without proactive intervention, the benefits of AI could be concentrated among a small elite, while large segments of the population face stagnant wages or job displacement.
The research further challenges the assumption that technological progress is inherently beneficial for labor. While past innovations have generally led to higher wages over time, AI may represent a departure from this trend if not managed carefully. The reason lies in the nature of AI as a general-purpose technology with unprecedented flexibility. Unlike earlier machines that were designed for specific tasks, AI systems can be rapidly reconfigured to perform a wide range of functions, making them more versatile and, therefore, more disruptive.
This versatility increases the elasticity of substitution between labor and AI, pushing economies closer to or beyond the critical thresholds identified in the study. In other words, the very adaptability that makes AI powerful also makes it more likely to displace workers unless countervailing forces—such as institutional support, education reform, and strategic investment—are in place.
The study’s methodology is also noteworthy for its attempt to overcome common pitfalls in empirical research on technology and labor. Many previous studies rely on statistical correlations that can be confounded by measurement error, endogeneity, or differences across countries and time periods. By using the Hauthakker model, Hao and Lv derive a general expression for the impact of AI on wages that is independent of specific data assumptions, lending greater credibility to their conclusions.
Their approach also allows for cross-country comparison under a unified theoretical framework, enabling a more rigorous analysis of why AI affects labor markets differently in China and the U.S. Rather than attributing differences to cultural or institutional factors alone, the study grounds its analysis in measurable economic parameters—elasticities, factor shares, and productivity distributions.
Despite its strengths, the study is not without limitations. The authors acknowledge that their model assumes competitive markets and does not account for market power, unionization, or government regulation, all of which can influence wage outcomes. Additionally, the data used—particularly for China—is based on aggregate industry-level statistics, which may mask important variations within sectors.
Nevertheless, the research provides a robust foundation for future inquiry. It opens up new avenues for exploring how different types of AI interact with labor across various economic contexts. It also underscores the need for interdisciplinary collaboration between economists, computer scientists, educators, and policymakers to ensure that the AI revolution benefits society as a whole.
Looking ahead, the findings suggest that the future of work will not be determined solely by technological advances, but by how societies choose to govern and integrate those advances. AI does not have to be a force of job destruction; it can be a tool for human augmentation. But realizing that potential requires deliberate policy choices, strategic investment in human capital, and a commitment to inclusive growth.
The study by Hao Lixiao and Lv Rongjie serves as a timely warning and a call to action. As AI continues to reshape the global economy, understanding the precise conditions under which it enhances or undermines labor is essential. The thresholds they identify—1.8649 and 1.900 for conditional wage suppression, and 2.35 and 2.73 for inevitable suppression—are not just academic figures. They are early indicators of a structural shift that could redefine the relationship between humans and machines in the workplace.
Policymakers would do well to treat these numbers as red lines—boundaries that should not be crossed without adequate safeguards. Beyond that, the study reminds us that technology is not destiny. The impact of AI on wages depends not on the machines themselves, but on the institutions, incentives, and investments that shape how they are deployed. In the race between man and machine, the outcome is still very much in human hands.
AI’s Wage Impact: Key Elasticity Thresholds Revealed in New Study
Hao Lixiao, Lv Rongjie, School of Economics and Management, Hebei University of Technology
Journal of Industrial Technological Economics, DOI: 10.3969 / j. issn.1004－910X.2021.11.018