How does industrial policy experimentation influence innovation performance? A case of Made in China 2025

The paper investigates whether and how policy experimentation under Made in China 2025 (MIC 2025) affects firm innovation. Industrial policy is widely used to shape industrial structures, but debates persist over market versus government failure and the effectiveness of intervention. MIC 2025, launched in 2015, targets upgrading key technologies and sectors to move China up the manufacturing value chain. Unlike the broad, multi-level Five-Year Plans (FYP), MIC 2025 is implemented through place-based pilot cities coordinated by the Ministry of Industry and Information Technology (MIIT), integrating clear domains, goals, and instruments and involving central–local interactions. Prior evidence on MIC 2025’s innovation effects is mixed, with some finding increased R&D while others find limited effects or risks of overcapacity. The study reframes analysis by focusing on the place-based implementation—pilot cities—rather than sector designation alone. The research question is whether being located in MIC 2025 pilot cities improves firm innovation quantity and quality, and through which mechanisms. The main hypothesis (H1) posits that location in MIC 2025 pilot cities facilitates firm innovation by relaxing resource constraints and enhancing competitive incentives.

The literature distinguishes between industrial policy’s potential incentive effects versus risks of distortion and government failure. In China, many studies use Five-Year Plans to examine policy effects on firm innovation, yet FYPs are comprehensive and lack specific implementation instruments compared to MIC 2025’s targeted, longer-horizon program. MIC 2025 scholarship often treats it as sector-based, examining impacts on R&D, subsidies, and OFDI: some studies show positive R&D responses and welfare effects, while others argue subsidies may not increase innovation and could fuel overcapacity. The paper highlights that MIC 2025 differs via place-based policy experimentation through pilot cities, creating iterative public–private interactions and central–local coordination. Drawing on the resource-based view, the authors argue pilot cities provide resource incentives (tax relief, subsidies, financing) and competitive incentives (academic collaboration, talent), forming a better innovation ecosystem. The institutional background emphasizes MIIT’s staged approval of 30 pilot cities (2016–2017), principles of local priority with central guidance, one-city-one-case tailoring, and dynamic selection once local support systems were ready, making pilot city announcements a suitable quasi-natural experiment.

Design: Quasi-natural experiment using MIC 2025 pilot city status and a propensity score matching–difference-in-differences (PSM-DID) framework with firm and year fixed effects. Parallel trend tests and multiple robustness checks are conducted. Sample: Chinese A-share listed manufacturing firms in Shanghai and Shenzhen exchanges within MIC 2025 priority sectors, 2012–2022. After excluding missing data, PSM identifies controls from non-pilot cities. Final balanced panel: 201 treatment firms and 201 matched control firms over 11 years (4422 firm-year observations). Most pilot city selections occurred 2016–2017, enabling roughly four-year pre/post windows. Data sources: Patent application data from CNRDS (invention, utility model, design); financing and governance data from CSMAR; academic collaboration from Cninfo announcements (web scraping); pilot city status from MIIT. Non-discrete variables are winsorized at 1% and 99%. Measures: - Dependent variables: Innovation outputs via patent applications. Invention = log(1 + invention applications). Non-Invention = log(1 + utility model + design applications). Applications are used to capture dynamics given 2–4 year grant lags. - Treatment: Treat = 1 if firm located in MIC 2025 pilot city; Post = 1 after selection; main regressor Treat × Post. - Controls: Firm-level Size, Age, ROA, Lev, RD, Pee (fixed assets), Inst, HHI (industry concentration), KZ (financing constraint), CFC (free cash flow), Capex, NWC (operating capability), SOE; firm and year fixed effects. City-level controls used in robustness (GDP per capita, GDP growth, science and technology expenditure, population, R&D personnel ratio, infrastructure spending, patent applications/grants). - Mechanism (mediators): Tax (income tax rate), Subsidy (govt subsidy/revenue), Financing (bank loans/total assets), Collaboration (binary if firm announces collaboration with universities/research institutes), Talent (R&D staff share). PSM: Year-by-year one-to-one nearest-neighbor matching via logit of Treat on lagged firm covariates (Size, Age, ROA, Lev, RD, Pee, Inst, HHI, KZ, CFC, Capex, NWC, SOE) to construct comparable control firms. Propensity score distributions before/after matching confirm balance. Models: - Dynamic event-study DID to test parallel trends: interactions of Treat with leads/lags around selection year; expect no pre-trends and significant post effects. - Baseline time-varying DID: Innovation_it = α0 + α1(Treat × Post)_it + controls + firm FE + year FE + error. Robustness: - Placebo tests with random assignment and bootstrap. - Endogeneity checks: City-level logit predicting pilot city selection using pre-selection city fundamentals; additional tests for political connections to national leaders (workplace/birthplace) via t-tests. - Excluding other policies: Add indicators for central government priority industries (IMPIND) and high-tech enterprise status. - Alternative estimators/specifications: did_multiplegt (heterogeneous treatment effects), Poisson models for count data, excluding COVID-19 years (2012–2019). Further analyses: Mediation regressions to quantify indirect effects via Tax, Subsidy, Financing, Collaboration, Talent; heterogeneity by region (east/middle/west), industry (high-tech vs non-high-tech), ownership (SOE vs non-SOE), size; economic outcomes (Tobin’s Q and TFP_LP); geographic spillovers to bordering cities with leads (t, t+1, t+2, t+3).

- Innovation effects: Treat × Post increases invention patents and reduces non-invention patents, indicating improved innovation quality. Baseline DID coefficients: Invention 0.395 (p < 0.01); Non-Invention −0.325 (p < 0.01). Event-study shows no significant pre-trends; significant and persistent post-policy effects (e.g., Treat × Current and Treat × After1..After3 significant with expected signs). - Mechanisms: MIC 2025 pilot city location lowers effective tax rates and increases subsidies, bank financing, academic collaborations, and R&D talent share. Mediation shares (relative reduction in Treat × Post coefficient on Invention): Tax ≈ 1.266%, Subsidy ≈ 4.051%, Financing ≈ 5.570%, Collaboration ≈ 5.823%, Talent ≈ 3.291%; total ≈ 20% mediated, with financing and collaboration most important. - Robustness: Results survive placebo tests, city-level endogeneity checks (city fundamentals insignificant; politically connected cities less likely to be selected), inclusion of other policy controls (IMPIND, High-Tech), alternative DID estimator (DIDM), Poisson models, and exclusion of COVID-19 years. - Heterogeneity: Stronger innovation gains in western China pilot cities (e.g., Treat × Post ≈ 0.760, significant) than eastern or middle regions; SOEs benefit more than non-SOEs; effects somewhat stronger in high-tech industries (difference not statistically significant); size-based differences not significant. - Economic outcomes: Treat × Post positively affects firm performance: Tobin’s Q ≈ 0.203 (p < 0.01) and TFP_LP ≈ 0.110 (p < 0.01); inclusion of invention output also positively linked to TQ and TFP. - Spatial spillovers: No significant spillover to bordering cities at t or t+1; positive spillovers emerge at t+2 and t+3 (e.g., Border × Post significant for Invention at t+2 and t+3).

The findings support the hypothesis that place-based policy experimentation under MIC 2025 enhances firm innovation quality by shifting innovation toward higher-value invention patents. By focusing on pilot cities, the study shows that implementation design—central–local coordination and local embedded autonomy—matters for translating sectoral priorities into concrete firm-level innovation outcomes. Resource incentives (tax relief, subsidies, easier bank financing) and competitive incentives (academic collaboration, talent attraction) alleviate information and financing frictions and bolster innovation capabilities. The stronger effects for SOEs and western regions suggest institutional capacity and regional baseline conditions shape policy effectiveness. Positive effects on Tobin’s Q and TFP indicate that innovation improvements translate into economic value and productivity. The evidence advances debates over industrial policy by moving beyond whether to adopt such policies to how to implement them effectively through place-based experimentation integrating sectoral and regional instruments.

The study contributes by (1) demonstrating that MIC 2025’s sector-based goals are effectively implemented through place-based pilot city experimentation that fosters central–local interactions and (2) unpacking the mechanisms—tax and subsidy support, financing facilitation, academic collaborations, and talent incentives—through which pilot city designation translates into higher-quality firm innovation and improved economic performance. Policy implications include integrating sectoral priorities with localized support systems, iterative public–private engagement, and talent and collaboration infrastructures to build robust innovation ecosystems. Future research should deepen analysis of local implementation via city-level policy documents, examine policy learning and diffusion across pilot cities, extend samples to unlisted and smaller firms, and evaluate MIC 2025’s impacts on green and digital innovation and sector-specific and inter-sector spillovers.

- Data constraints: Firm-level analysis focuses on listed manufacturing firms and headquarters’ patent applications; city-level policy implementation data are limited. - External validity: Results may differ for unlisted or smaller firms; sample restricted to ten MIC 2025 priority sectors. - Outcome scope: Innovation measured via patent applications; additional innovation types (green, digital) and other performance metrics warrant study. - Sectoral granularity: Heterogeneous effects across specific manufacturing subsectors and inter-sector spillovers remain to be fully identified.