Trade friction drives AI adoption and better innovation outcomes

A new analysis indicates that the Sino-U.S. trade friction may be doing more than raising costs for Chinese manufacturers. It may also be pushing exposed firms to improve the quality of their innovation by adopting artificial intelligence (AI).

The study, titled "Does Sino–U.S. Trade Friction Promote Corporate Innovation Quality? The Mediating Role of Artificial Intelligence," was published in the MDPI journal Systems. Using the U.S. Section 301 investigation as a quasi-natural experiment, the research finds that trade friction significantly improves corporate innovation quality among Chinese listed manufacturing firms, with AI adoption serving as a partial channel through which that effect occurs.

Trade pressure shifts firms from patent quantity to innovation quality

Traditional trade theory suggests that tariffs and policy uncertainty raise export costs, worsen financing conditions and reduce firms' ability to invest in research and development. This view has shaped much of the literature on trade shocks, especially where firms face higher compliance costs and weaker access to international markets.

However, the study points to a competing mechanism. Under the escaping-competition effect, firms facing tougher market conditions may increase innovation to protect market share, differentiate products and build stronger technological capabilities. In this view, trade pressure does not only impose costs. It can also push firms to upgrade.

The researchers argue that the key distinction lies in innovation quality. Earlier studies often measured innovation through R&D spending, patent counts or other quantity-based indicators. Those measures can be misleading in China because patent activity may be affected by subsidies, incentives and strategic behavior that encourage low-productivity innovation. A rise in patent numbers does not necessarily mean that firms are producing more valuable technologies.

To avoid that limitation, the study measures corporate innovation quality through patent breadth. This captures the diversity of technological claims across patent subclasses, with broader patents indicating greater complexity, wider application scope and stronger protection. The researchers also test alternative measures, including the share of invention patent applications and forward citations of invention patents, and find that the core result remains robust.

The study uses data on Chinese A-share listed manufacturing firms from 2014 to 2022, covering 14,722 firm-year observations and 2074 unique enterprises. It builds a firm-level exposure measure that combines industry-level tariff changes caused by the Section 301 tariffs with firm-level export dependence before the trade shock. This allows the analysis to capture both the intensity of tariff exposure and the degree to which each firm relied on exports.

The baseline results show that exposure to Sino-U.S. trade friction had a positive and statistically significant effect on corporate innovation quality. The estimated effect implies that a one-standard-deviation increase in trade friction exposure raised innovation quality by about 4.35% relative to the sample mean. The authors describe this as economically meaningful, especially against the backdrop of China's manufacturing value-added growth during the sample period.

The finding does not mean tariffs are broadly beneficial. It suggests that firms under pressure may respond strategically when they have the capacity and incentives to upgrade. Trade friction can force firms to move beyond cost-based competition and seek more durable advantages through better-quality innovation.

AI adoption becomes a strategic response to uncertainty

The researchers argue that firms facing trade pressure are more likely to adopt AI as a forward-looking response to uncertainty, rising costs and intensified competition. AI can help firms process information, improve decision-making, optimize operations, reduce risk and support technological search.

To test this mechanism, the study measures AI adoption through textual analysis of corporate annual reports. The researchers use AI-related keywords appearing in the management discussion and analysis section, including terms linked to machine learning, intelligent computing, computer vision, big data analytics, natural language processing, intelligent manufacturing and other digital technologies. They construct two measures: the ratio of AI-related terms to total text and the raw count of AI-related terms.

The mediation analysis shows that trade friction significantly increases AI adoption, and that AI adoption is positively associated with innovation quality. When AI adoption is added to the model, the direct effect of trade friction remains positive but becomes smaller, indicating partial mediation. Bootstrap tests further confirm that the indirect effect through AI adoption is statistically significant. This means AI does not fully explain the link between trade friction and innovation quality, but it forms an important part of the pathway. Firms exposed to external pressure appear to use AI not only as a digital tool, but as part of a broader organizational response to uncertainty.

The study frames this through socio-technical systems theory, which views organizational change as the joint development of technical systems and social structures. AI adoption can reshape production, decision-making and innovation processes. It can help firms identify technological opportunities, draw on cross-domain knowledge and improve the efficiency of innovation activities.

Innovation quality depends not only on spending more money on research. It also depends on how firms organize knowledge, allocate resources, manage uncertainty and convert technological capabilities into stronger outputs. AI can strengthen those processes by improving data analysis, information processing and organizational coordination.

Trade friction is often viewed mainly as a source of disruption. However, the study suggests that, for some firms, geopolitical pressure may accelerate digital transformation and innovation upgrading. This dynamic may be especially important in manufacturing, where firms face pressure to reduce dependence on vulnerable markets and build more resilient technological capabilities.

Digital leadership and governance determine who benefits most

The study finds that the innovation-enhancing effect of trade friction is not evenly distributed. It is stronger among firms with a higher proportion of executives who have information technology experience and among firms with stronger corporate governance.

The role of executive IT background is significant because AI adoption is not automatic. Firms need leaders who understand digital technologies, can evaluate AI opportunities and can guide organizational changes needed for implementation. Without that expertise, firms may struggle to turn external pressure into effective innovation strategies.

The results show that the positive effect of trade friction on innovation quality is statistically significant only among firms with above-median levels of executive IT experience. This suggests that digital leadership acts as a key enabling condition. Firms with stronger technological understanding at the top are better positioned to adopt AI and use it to improve innovation outcomes.

Corporate governance also matters. The study uses the proportion of independent directors as a proxy for governance quality. Firms with stronger governance show a larger and more statistically significant innovation response to trade friction. This suggests that effective oversight helps firms allocate resources more strategically, reduce wasteful behavior and respond more efficiently to external shocks.

Together, these findings show that external pressure alone is not sufficient. Trade friction may create incentives to innovate, but firms need internal capabilities to capture the benefits. Digital competence, management quality and governance structures shape whether the pressure leads to meaningful innovation upgrading or simply higher costs.

The study's robustness checks strengthen the conclusion. The researchers test the parallel trends assumption behind their difference-in-differences design and find no evidence of significant pre-treatment differences. A placebo test using 2000 random simulations shows that the main result is unlikely to be driven by chance. The findings also hold when alternative measures of trade friction and innovation quality are used.

The study further controls for other policy initiatives that could affect innovation, including the China-Europe Railway Express, National AI Innovation Pilot Zone policies and the Made in China 2025 strategy. The positive effect of trade friction remains significant after accounting for these possible confounding policies. The result also holds when excluding firms with exceptionally high innovation quality and firms headquartered in China's four centrally administered municipalities.

If firms under trade pressure are using AI to improve innovation quality, governments seeking high-quality industrial development should focus not only on protecting firms from shocks, but also on helping them build digital capabilities. That could include support for AI infrastructure, R&D incentives, data governance systems and training programs for digital management talent.

For corporate managers, the findings point to AI as a strategic buffer against external uncertainty. Firms exposed to trade tensions may need to move faster in integrating AI into operations, product development and innovation management. However, adoption must be substantive, not rhetorical. The study acknowledges that its AI measure is based on corporate disclosure, which may not fully capture the depth or effectiveness of AI implementation.

The researchers also note that the study focuses on tariff sanctions under the U.S. Section 301 investigation and does not examine non-tariff restrictions such as entity list designations, technical standards or export controls. Those tools are increasingly central to Sino-U.S. technological competition and could affect firms differently.