U.S.-China trade conflicts and R&D investment: evidence from the BIS entity lists

The study investigates how U.S. export controls deployed during the U.S.-China trade conflicts affect Chinese firms' innovation, focusing on R&D investment. Motivated by limited evidence on micro-level impacts in China relative to extensive U.S.-focused research, the authors treat BIS export control lists as exogenous shocks to Chinese firms. They hypothesize that inclusion on BIS Entity List or Military End-User List increases firms' R&D intensity, potentially via government subsidies, inventory adjustments due to supply disruptions, and heightened firm risk-taking. The work is important for understanding innovation dynamics in developing economies under trade frictions and for informing policy responses.

Prior literature documents macroeconomic and price impacts of tariffs and trade wars on the U.S. economy (Amiti et al., 2019; Fajgelbaum et al., 2020; Cavallo et al., 2021) and the costs borne by consumers and firms. Micro-level evidence on export controls’ impacts on Chinese firms is scarce, partly due to data limitations (e.g., lag in CIFD updates). The innovation literature shows importing advanced intermediates boosts productivity and innovation (Grossman & Helpman, 1991; Kasahara & Rodrigue, 2008; Halpern et al., 2015; Chen et al., 2017), but import liberalization can dampen domestic innovation incentives (Liu & Qiu, 2016; Autor et al., 2020). Models suggest for laggards, innovation incentives depend on distance to frontier (Aghion et al., 2005; Acemoglu et al., 2006). Trade protection can sometimes spur indigenous innovation (Rajeswari, 2014; Miyagiwa et al., 2016; Jabbour et al., 2019). Research on export controls often centers on regime design and domestic effects (Berman & Garson, 1967; Hosoe, 2021; Shin & Balistreri, 2022). This paper adds firm-level causal evidence from the sanctioned country, examining mechanisms including government subsidies, inventory behavior, and risk-taking.

Design: A staggered Difference-in-Differences (DID) framework treats BIS export control listings as quasi-natural experiments. The key treatment indicator (Treat) equals 1 from the year a listed company (or its parent/subsidiaries/affiliates within identified groups) is first sanctioned via BIS Entity List (EL) or Military End-User List (MEU), and 0 otherwise. The outcome is firms' R&D intensity, defined as R&D expenditures scaled by total assets, front-loaded by one year (RD_{i,t+1}). Data: Chinese A-share listed manufacturing firms (Shanghai and Shenzhen) from 2013–2022. Financials and R&D from CSMAR; BIS EL and MEU from Supplement No.4 to Part 744 EAR (BIS website). After excluding ST/ST* firms, serious missing data, and winsorizing continuous variables at 1%, the panel comprises 1,934 firms and 13,781 observations; due to the one-year lead outcome, regressions use 11,948 observations. Authors identify 215 sanctioned listed companies by matching BIS-listed entities (including groups and tracing parent-subsidiary structures) to A-share listings; within the final manufacturing sample, 97 firms are sanctioned (about 5%). Empirical specification: RD_{i,t+1} = β0 + β1 Treat_{i,t} + X_{i,t}γ + φ_i + μ_t + Industry×Year FE + ε_{i,t}. Controls X include firm Size (log assets), Age, ROA, Growth (revenue growth), Lev, Cash flow/Assets, Tangible assets ratio, Top1 (largest shareholder’s stake), Tobin’s Q, CEO duality, and SOE status. Standard errors are clustered at industry level; robustness also clusters at firm level. Dynamic effects are assessed via event-study dummies D_{j} for years relative to first listing to test parallel trends. Mechanism tests: Three channels are examined using analogous DID with industry×year FE: (1) Government subsidies (log subsidies); (2) Non-finished goods inventory (NFGI; log of inventory of non-finished goods); (3) Risk-taking measured by industry-adjusted stock return volatility. Robustness: (a) Expanded sanction lists incorporating U.S. DoD (CMC/NDAA) and Treasury NS-CMIC lists merged with BIS EL/MEU; (b) Placebo tests by randomly assigning treatment 500 times (Chetty et al., 2009); (c) Bacon decomposition to assess two-way FE bias in staggered DID and Callaway & Sant’Anna (2020) estimator for ATT; (d) Excluding tariff effects using WTO HS6 tariffs mapped to industries and a Tariff dummy; (e) Alternative variables: dependent variable LnRD; alternative size proxy (log operating revenue). Heterogeneity analyses split by ownership (SOE vs POE), executives’ foreign experience, and industrial policy support (industries flagged as encouraged in provincial Five-Year Plans). Further analysis examines innovation outputs (two-year lead): total patent applications and invention patent applications.

• Baseline DID: Inclusion on BIS EL/MEU increases firms’ next-year R&D intensity by 0.274 log points, implying a 16.58% rise (e^0.274−1), significant at 5% with industry×year FE; effect robust when clustering by firm/industry.
• Event study: No pre-trend differences; significant positive jump from treatment year onward.
• Placebo: 500 random assignments yield coefficients centered near 0 (mean ≈ 0.044) and p-values larger than actual, supporting causal interpretation.
• Mechanisms: Export controls increase: • Government subsidies: coefficient 0.249 (log), ≈ +24.9%, p<0.10. • Non-finished goods inventory (NFGI): 0.076 (log), p<0.10. • Risk-taking: 0.005 (stock return volatility), p<0.01.
• Robustness: • Expanded lists (adding DoD/Treasury sanctions): treatment effect 0.274 (p<0.05); mechanisms replicate (0.249; 0.076; 0.005). • Bacon decomposition: “bad” weight (treated later vs earlier) ≈ 0.4%; CS estimator ATT = 0.409 (p<0.05). • Excluding tariff effects: results remain positive and significant controlling for Tariff dummy. • Alternative outcome (LnRD) and alternative size proxy both yield consistent positive effects.
• Heterogeneity: • SOEs: Treat = 0.171 (p<0.05); POEs not significant. • Executives with foreign experience: Treat = 0.428 (p<0.01); without foreign experience: not significant. • With industrial policy support: Treat = 0.274 (p<0.05); without support: not significant.
• Innovation outputs: Two-year lead outcomes show insignificant or negative coefficients: Total patents (−0.427, n.s.) and invention patents (−0.045, n.s.), indicating limited or adverse short-run effects on patenting despite higher R&D inputs.

Findings support the hypothesis that U.S. export controls act as an external shock that induces Chinese listed manufacturing firms to expand R&D spending, primarily through greater government support, inventory adjustments to mitigate supply chain risks, and higher risk tolerance. The absence of pre-trends strengthens causal claims. The stronger response among SOEs, firms with executives possessing foreign experience, and firms in industrial-policy-supported industries suggests institutional and managerial capacity conditions the R&D response. Despite higher inputs, patent outputs do not immediately improve, highlighting frictions in translating spending into measurable innovation, possibly due to technological blockages and adjustment lags. The results contribute micro-level causal evidence on how trade restrictions can inadvertently foster indigenous R&D effort in developing economies while underscoring the role of domestic policy support and corporate governance in shaping responses.

The paper provides firm-level causal evidence that BIS export controls during the U.S.-China trade conflicts increased Chinese listed manufacturing firms’ R&D intensity by about 16.6% in the subsequent year. This effect operates via elevated government subsidies, inventory strategies, and increased firm risk-taking, and is more pronounced in SOEs, firms led by executives with foreign experience, and sectors supported by industrial policy. However, enhanced R&D efforts have not translated into higher short-run patent outputs, pointing to low input–output conversion efficiency amid technological constraints. The study broadens understanding of trade conflict impacts on innovation in transition economies and suggests that while external constraints can catalyze R&D investment, complementary policies are needed to improve the productivity of R&D. Future research should incorporate richer local policy measures and longer horizons to capture output realization and explore unlisted firms and broader sectors.

The analysis focuses on A-share listed manufacturing firms, limiting generalizability to unlisted or non-manufacturing firms. The study period is relatively short relative to innovation gestation, which may understate long-run innovation outputs. Treatment identification relies on public sanction lists and matching to corporate structures, which may omit indirect exposures. Mechanism measures (e.g., subsidies, inventory categories, stock volatility) capture proxied channels and may be subject to measurement error. Tariff controls are addressed, but other concurrent policies or shocks may remain imperfectly captured despite high-dimensional fixed effects.