Impact of China's Free Trade Zones on the Innovation Performance of Firms: Evidence from a Quasi-Natural Experiment

The study investigates whether and how China’s Free Trade Zone (FTZ) policy affects the innovation performance of firms. In the context of China’s economic transformation and push for higher-quality growth and openness, FTZs aim to reduce trade and financial frictions, improve the business environment, and catalyse innovation. While FTZs may lower transaction costs and stimulate R&D, they can also expose firms to intensified global competition. To clarify the net effect, the authors conduct a quasi-natural experiment leveraging the staggered rollout of 21 FTZs since 2013, matching FTZ establishment data with Chinese A-share listed firms’ patent applications (2012–2020). The paper contributes by: (1) grounding analysis in dynamic capability and competitive advantage theories to frame how FTZs can influence exploratory and exploitative innovation; (2) using a time-varying difference-in-differences (DID) approach with extensive robustness checks; (3) examining mechanisms via financing constraints and industrial agglomeration; and (4) addressing endogeneity with instrumental variables (IV) and propensity score matching (PSM), and exploring moderating contexts and firm heterogeneity. Three hypotheses are proposed: H1—FTZs improve the scale and quality of firms’ innovation; H2—FTZs enhance innovation by alleviating financing constraints; H3—FTZs enhance innovation by increasing industrial agglomeration.

The background review outlines China’s phased FTZ establishment since Shanghai (2013), expanding to 21 FTZs by 2020 across coastal and inland regions, aiming to liberalise trade and investment, improve financial services, and pilot regulatory innovations. FTZs are argued to: (i) expand trade and attract foreign entrants, increasing competition; (ii) concentrate capital and technology, improving the environment for high-tech firms; and (iii) provide policy flexibility to accelerate marketisation. The review identifies gaps: most prior research emphasises macro impacts rather than firm-level innovation outcomes; staggered FTZ rollouts complicate identification of dynamic effects; potential mediators such as industrial agglomeration and financing constraints are underexplored; and moderating factors (economic development, government support) and firm heterogeneity (size, ownership, industry, location) have not been fully analysed. The study addresses these by measuring innovation scale and quality via patent applications, using a time-varying DID, testing mediators (financing constraints, agglomeration), treating endogeneity with IV and PSM, and probing moderators and heterogeneity.

Design: Quasi-natural experiment exploiting staggered FTZ establishment across Chinese provinces/municipalities. Core identification uses a time-varying DID with two-way fixed effects and extensive controls. Robustness includes parallel trend and placebo tests, high-dimensional fixed effects, model substitutions, and alternative estimators. Mechanism tests assess mediation through financing constraints and industrial agglomeration. Endogeneity is additionally addressed via IV and PSM. Moderators and heterogeneity are examined. Data: Firm-level A-share listed companies, 2012–2020. Patent data from China National Intellectual Property Administration and CSMAR; financials from CSMAR and WIND; city/province data from China Statistical Yearbooks; FTZ establishment dates matched by the firm’s registered city-province. After exclusions (financial firms; ST/*ST; outliers; missing key variables) and 1% winsorization, the final panel has 10,805 firm-year observations. Cluster-robust standard errors at province and industry levels; Stata 17 used. Variables: Innovation outcomes—Patent (ln(total patent applications +1), innovation scale) and Invent (ln(invention patent applications +1), innovation quality). FTZ indicator equals 1 post-implementation in a firm’s province/municipality. Controls include ROA, growth, tangible asset ratio, leverage, cash flow, per capita disposable income (Perwage, log), local fiscal revenue (Revenue, log), FDI (log), and internet development (Telecom, log). Firm (Cid), year, industry (Ind), and province (Prov) fixed effects included. Model: Time-varying DID (two-way FE) following Callaway and Sant’Anna (2021): Innovation_{t+1} = β0 + β1 FTZ_t + Controls + Firm FE + Year FE + Industry FE + Province FE + ε. Lags account for patenting timelines. Robustness: (1) Event-study/parallel trend tests; (2) Placebo tests with 3,000 random re-assignments; (3) High-dimensional FE adding Year×Industry and Year×Province; (4) Excluding potential confounding from high-tech enterprise tax-incentive thresholds; (5) Alternative outcomes: R&D expenditure ratio (RD-ratio) and TFP (Levinsohn-Petrin); (6) Alternative estimators: Tobit, Poisson, and median regression. Mediation: Financing constraints measured by SA index (SA = −0.737×Size + 0.043×Size^2 − 0.04×Age; more negative implies tighter constraints). Stepwise regressions and Sobel tests assess mediation. Industrial agglomeration proxied by HHI (sum of squared industry shares); lower HHI indicates more competition/agglomeration. Mediation tested via regression and Sobel tests. Endogeneity: IV strategy uses industrial SO2 emissions and length of urban sewage pipes (Pipe) as instruments for FTZ selection (relevance via infrastructure/urban development; argued exogeneity to firm-level innovation). Two-stage least squares employed; overidentification and weak IV tests reported. PSM: nearest neighbour, caliper, kernel, and Mahalanobis matching to estimate ATT/ATU/ATE post balance checks (standardised bias <10%). Further analyses: Moderation by regional economic development (PerGDP) and government subsidies (Subsidy) via interaction terms with FTZ. Heterogeneity by firm size (LE vs SME by industry-median size), ownership (SOE vs non-SOE), industry (technology-intensive vs non-tech per GB/T 4754), and region (Eastern vs other).

Baseline DID: FTZ significantly increases innovation scale and quality. In Table 4, FTZ coefficients remain positive and significant at 1% across specifications. Effects imply roughly 9.2% increase in total patent applications and 12.2% increase in invention patent applications compared to non-FTZ firms. Local fiscal revenue and internet development also associate positively with innovation. Robustness: Event-study shows pre-treatment parallel trends with insignificant coefficients; post-FTZ coefficients become significantly positive. Placebo tests (3,000 random assignments) yield coefficient distributions centered near zero and below baseline estimates, supporting identification. High-dimensional FE (Year×Industry; Year×Province) and iterative FE keep FTZ positive and significant (e.g., Patent 0.113**, Invent 0.140***). Excluding firms at patent thresholds for high-tech tax incentives leaves results intact (Patent 0.089***, Invent 0.112***). Alternative outcomes confirm positive effects (RD-ratio 0.171*, TFP 0.046***). Alternative models (Tobit, Poisson, median) consistently show positive, significant FTZ effects on Patent and Invent. Mediation: Financing constraints—FTZ reduces SA-based financing constraints (FC) (−0.048***). Including FC, FTZ remains positive (Patent 0.087***; Invent 0.119***), while FC negatively relates to innovation (Patent −0.100***; Invent −0.050***). Sobel tests confirm partial mediation (Patent 0.066**; Invent 0.133***). Industrial agglomeration—FTZ lowers HHI (−0.034*, indicating more agglomeration/competition) and HHI negatively associates with Patent and Invent (−0.030*; −0.022*), with Sobel tests indicating partial mediation (Patent 0.071**; Invent 0.135***). Endogeneity corrections: IV 2SLS—First stage: SO2 negatively (−) and Pipe positively (+) predict FTZ at 1% significance; overidentification p=0.36 (not rejected), weak-IV F=129.19. Second stage: FTZ remains positive and significant for Patent (1.173***) and Invent (1.324***). PSM—Across nearest neighbour, caliper, kernel, and Mahalanobis, ATT on Patent and Invent is positive and significant (mostly 1%); ATE and ATU also positive. Moderation: Interactions FTZ×PerGDP and FTZ×Subsidy are positive and significant (often at 1%), indicating stronger FTZ effects in more developed areas and where government subsidies are higher; effects are larger for invention patents, suggesting higher-quality innovation gains in developed/subsidised contexts. Heterogeneity: Size—Effects for large enterprises exceed SMEs; quality impacts exceed scale (e.g., LE Invent 0.132*** vs SME Invent 0.087***). Ownership—Both SOEs and non-SOEs benefit, but effects are more pronounced for non-SOEs (e.g., non-SOE Invent 0.120***). Industry—Significant positive effects for technology-intensive firms; effects for non-tech firms are not significant. Region—Stronger effects in Eastern region (Patent 0.170***; Invent 0.105***) than elsewhere (Patent 0.099**; Invent 0.106**). Overall, results support H1–H3.

The findings show that FTZs enhance both the scale and quality of firm innovation, aligning with dynamic capability and competitive advantage theories: increased openness and competition in FTZs pressure firms to invest in R&D and to leverage external knowledge. Mechanism analyses indicate that FTZs alleviate financing constraints—expanding and diversifying capital access—which enables sustained R&D investment, and that FTZs promote industrial agglomeration—fostering synergies, peer effects, and knowledge spillovers—thereby raising innovation productivity. Moderation and heterogeneity results demonstrate that FTZ impacts are context-dependent: economically advanced regions and higher subsidy environments amplify innovation quality improvements; large firms and non-SOEs are better positioned to translate FTZ advantages into R&D outputs; technology-intensive sectors benefit most; and Eastern coastal regions show the strongest gains due to superior infrastructure, market depth, and talent pools. These results corroborate the causal interpretation after extensive robustness, IV, and PSM analyses, and inform where FTZ policy can most effectively catalyse innovation.

The study provides firm-level causal evidence that China’s FTZ policy significantly improves innovation outputs, increasing both total and invention patent applications among listed firms. Mechanism tests confirm partial mediation via reduced financing constraints and increased industrial agglomeration. The innovation-promoting effects are stronger in more developed and subsidised regions, for large and non-state firms, in technology-intensive industries, and in Eastern China. Policy implications include: (1) continuing and scaling FTZ institutional reforms nationwide; (2) prioritising high-end, intelligent industrial systems and support for high-tech enterprises to deepen innovation ecosystems; and (3) targeting subsidies and recognition mechanisms to reward innovative firms while assisting SMEs to overcome financing barriers. Future research should incorporate novel innovation measures beyond patents, track longer-term impacts as FTZ coverage expands, and examine links between FTZ policy and broader sustainable development goals.

The paper acknowledges potential endogeneity due to non-random FTZ placement (e.g., economic and locational advantages) and addresses it with time-varying DID, clustered standard errors, IVs (SO2, sewage pipe length), and PSM. Measurement challenges arise from staggered FTZ rollouts and the lag between R&D and patenting. Some robustness analyses rely on subsamples because of limited disclosure of R&D ratios and TFP, reducing sample sizes. The study focuses on Chinese A-share listed firms, which may limit generalisability to private or unlisted firms. Although multiple robustness checks mitigate concerns, residual unobserved factors cannot be fully ruled out.