Digital economy and urban innovation level: A quasi-natural experiment from the strategy of "Digital China"

China’s digital economy expanded from 11 to 45.5 trillion yuan and from 20.9% to 39.8% of GDP, becoming a key driver of quality growth. National planning emphasizes building "Digital China" to integrate digital and real economies and to power an innovation-driven development model. While digitalization offers efficiency, structural upgrading, and inclusive growth, it also raises challenges such as the digital divide and data privacy/security. Given technological innovation’s central role in high-quality urban development, this study asks whether and how the digital economy—operationalized via local implementations of the "Digital China" strategy—drives urban innovation, and through what mechanisms. The paper hypothesizes: (H1) the digital economy has a driving effect on urban innovation; (H2) the digital economy promotes urban innovation through industrial structure upgrading.

Existing work links the digital economy to growth, efficiency, and structural upgrading but faces measurement disagreements and potential endogeneity. Measures range across digital access, infrastructure, platform construction, application, and proxies like internet development and digital finance, often producing mixed or dimension-specific effects on innovation. Prior studies frequently examine informatization or internet infrastructure (e.g., "Broadband China") rather than holistic digital economy strategies, and operate at macro, industry, or firm levels, noting spillovers, efficiency gains, and green innovation effects. However, bi-directional causality (innovative regions may advance digitalization), measurement error, and omitted variables remain concerns. Policy-evaluation-based causal evidence on the broader digital economy’s impact on urban innovation is scarce. This study fills the gap by exploiting staggered local digital economy strategic plans as a quasi-natural experiment to identify causal effects and mechanisms.

Design: A multi-temporal difference-in-differences (DID) framework exploits the staggered rollout of provincial-level digital economy strategic plans (framed as local implementations of the national "Digital China" strategy) as a quasi-natural experiment. The treatment indicator (DID) equals 1 for cities in provinces that issued a major digital economy plan, from the year after release onward; 0 otherwise. Municipalities are excluded. Sample: 286 prefecture-level cities in mainland China, 2008–2018, yielding 3,123 observations after cleaning. Treatment and control: 137 cities in 11 treated provinces (e.g., Jilin, Zhejiang, Anhui, Fujian, Sichuan, Guizhou, Hunan, Guangxi, Guangdong, Gansu, Shaanxi) vs. 149 control cities. Outcome: Urban innovation measured primarily by patent applications per 10,000 people; robustness outcomes include an urban innovation index, granted patents per capita, and above-scale new product sales rate. Model: Innovation_it = a + β·DID_it + q·Controls_it + μ_i + λ_t + ε_it, with city and year fixed effects. Parallel trends are tested using event-time leads/lags (Before, After, and Current dummies relative to policy year). Controls include economic development and fiscal capacity (GDP, fiscal revenue/expenditure), industrial structure (secondary industry share), investment (fixed asset investment, FDI actually used), population (year-end population), and S&T expenditure; most in logs except shares. Mechanism: Mediation via industrial structure upgrading (ins), proxied by tertiary-to-secondary output ratio; alternative proxies include tertiary value-added share of GDP and a structural advancement index (1*primary + 2*secondary + 3*tertiary shares). Robustness: (i) sample adjustments excluding early movers (Fujian; Fujian + Hunan), and limiting pre-2016 to avoid overlap with national informatization initiatives; (ii) alternative dependent variables; (iii) per-capita control specifications; (iv) placebo with 500 random reassignments of treatment to assess spurious significance. Heterogeneity: By innovation type (substantive/invention vs. non-invention patents) and by region (eastern/central vs. western).

- Baseline DID effect: Implementation of the digital economy strategy increases patent applications per 10,000 people by about 2.32 (Column (5), SE=0.950, p<0.05). The effect remains significant with extensive controls and fixed effects. - Controls: GDP and population positively relate to innovation; a higher secondary industry share is negatively associated. Fiscal revenue negatively associates; city science and technology expenditure is positively associated. FDI shows a negative and significant association in the fully controlled model (coef ≈ -3.281, p<0.01). - Parallel trends: No significant pre-trends between treated and control cities; innovation rises significantly in treated cities post-policy, consistent with a causal effect. - Robustness: Results hold when excluding Fujian or Fujian+Hunan, restricting to pre-2016, replacing outcome measures (urban innovation index; granted patents per capita), and altering control specifications. Using the new product sales revenue share as the outcome indicates the policy raises the share by approximately 4.25 percentage points. - Placebo: In 500 random reassignments, only ~1% produce t-values exceeding the true estimate, indicating low risk of false positives. - Heterogeneity: The digital economy significantly promotes substantive (invention) innovation (coef ≈ 1.255, p<0.05), with no significant effect on non-invention (utility/design) patents. Regionally, significant positive effects are observed in eastern and central regions (coef ≈ 2.751, p<0.05), but not in western regions. - Mechanism (industrial upgrading): The policy increases industrial structure upgrading (ins) (coef ≈ 0.025, p<0.05). Including ins in the innovation regression shows ins positively promotes innovation (coef ≈ 3.237, p<0.01), and the policy effect remains positive, supporting mediation via industrial upgrading. Alternative ins measures corroborate the mechanism.

The findings support the hypothesis that the digital economy, as advanced by local "Digital China" strategies, causally boosts urban innovation. The effect operates not only directly—through improved information flows, reduced transaction costs, and diffusion of general-purpose technologies (AI, big data, cloud, blockchain)—but also indirectly through industrial structure upgrading. The stronger impact on invention patents indicates an enhancement in innovation quality, aligning with the policy goal of high-quality growth. Regional heterogeneity suggests that absorptive capacity and existing development foundations in eastern/central regions enable greater returns to digital economy policies than in western regions. Overall, the results validate policy-led digitalization as a lever for city-level innovation upgrading, addressing measurement and endogeneity concerns via a quasi-experimental approach.

Using a multi-temporal DID on 286 Chinese cities (2008–2018), the study shows that local implementations of the "Digital China" strategy significantly increase urban innovation, particularly high-quality (invention) innovation, with stronger effects in eastern and central regions. Industrial structure upgrading is a key pathway, as digital industrialization and industrial digitization foster shifts toward higher-value activities that stimulate innovation. Policy recommendations include accelerating implementation of digital economy strategies; prioritizing investments in core digital technologies and infrastructure; strengthening IP protection, financing, and talent systems; and deepening integration of digital and real economies to catalyze industrial upgrading. Future research should employ finer-grained microdata, develop theoretical models linking digital economy to innovation, and broaden innovation measures to capture diverse outcomes.

Although endogeneity is mitigated via staggered policy timing and extensive robustness checks, potential overlaps with concurrent national informatization policies may remain. Measurement choices for digital economy exposure (provincial-level policy timing) may mask within-province variation. The sample period ends in 2018, limiting coverage of subsequent policy waves. Innovation measures rely primarily on patent-based and product indicators, which may not capture all innovation dimensions. City-level data constraints and partial interpolation may introduce measurement error. The authors note data access limitations and suggest future work with micro-level enterprise data and richer innovation metrics.