Impact of demonstration zone policy on agricultural science and technology innovation: evidence from China

The study addresses whether the National Modern Agricultural Demonstration Zone (NMADZ) promotes agricultural science and technology innovation (ATI) in China and through what mechanisms. Against the backdrop of rising risks and uncertainties and the strategic importance of agriculture for national development, China relies on science and technology innovation to transition from a large agricultural country to an agricultural power. The NMADZ policy, launched by the Ministry of Agriculture in three batches (2010, 2012, 2015), aims to modernize agriculture through policy guidance and financial support. Key research questions are: Can NMADZs promote ATI, and what are the channels? The paper contributes by: (1) using precise district/county-level data covering all three batches of NMADZs; (2) applying a difference-in-differences (DID) framework rather than mostly qualitative prior approaches; (3) focusing on ATI outcomes rather than predominantly ecological or income effects studied before; and (4) testing mechanisms and heterogeneity to provide practical guidance for enhancing ATI.

Prior studies on NMADZs have largely examined ecological/environmental effects, agricultural modernization, income, and employment outcomes. Theoretical perspectives relevant to ATI include technology diffusion (regional policies and demonstration effects accelerate adoption), industrial agglomeration (clustering enhances efficiency, structure, and innovation), and resource-based views emphasizing capital, finance, and human capital as core inputs to innovation. Policies such as preferential taxation, finance, land, and flexible mechanisms can attract firms and research institutions, building innovation systems and infrastructure, and fostering collaboration among enterprises, universities, and research institutes. However, fewer studies rigorously evaluate NMADZ effects on ATI per se, leaving a gap this study addresses.

Design: Quasi-natural experiment using multi-period difference-in-differences (DID) on panel data from 696 Chinese regions (districts/counties and prefecture-level cities matched to NMADZ locations), 2007–2017. Treated units are regions approved as NMADZs in three batches (2010: 52 regions; 2012: 101; 2015: 157), with treatment indicator switching to 1 from the approval year onward; remaining regions are controls. Data sources: China Science and Technology Statistical Yearbook, National Bureau of Statistics, China Rural Statistical Yearbook, CSMAR, China Banking and Insurance Regulatory Commission. Missing region-year data within series were imputed via linear interpolation. Models: Baseline DID model ATI_it = β0 + β1*DID_it + β2*Controls_it + γ_i + μ_t + ε_it, with region and year fixed effects. Mechanism (mediation) tests follow Wen and Ye (2014): (i) mediator_it = α0 + α1*DID_it + α2*Controls_it + γ_i + μ_t + ε_it for mediators Finance (financial resource allocation), Government (financial support to agriculture), and Human (human capital); (ii) ATI_it = γ0 + γ1*DID_it + γ2*Mediator_it + γ3*Controls_it + γ_i + μ_t + ε_it. Event-study specification tests parallel trends with leads/lags. Variables: - Dependent variable (ATI): Composite index of agricultural science and technology innovation measured via entropy method, integrating four dimensions: environment (rural disposable income; rural fixed asset investment), innovation support (average higher-education students; number of cell phones), inputs (R&D expenditure and R&D FTE scaled by total agricultural output/regional GDP; total power of agricultural machinery), and outputs (green patent applications per 10,000 people; value added to agricultural output). - Treatment (DID): Indicator equals 1 for treated regions from approval year onward (per batch), 0 otherwise. - Mediators: Finance (log stock of financial institutions), Government (ratio of fiscal expenditure on agriculture/forestry/water to agricultural GDP), Human (share of population with general undergraduate degree or above; used as proxy for average years of education). - Controls: GGDP (primary industry share of GDP), Power (log total power of agricultural machinery), Fixed (fixed asset investment/GDP), Load (sum of resident savings and financial institution year-end loan balance/GDP), Budget (general fiscal budget revenue and expenditure/GDP). Identification and validation: Region and year fixed effects to control for time-invariant heterogeneity and common shocks; parallel trend tested via event-study; robustness checks include: alternative ATI proxy (agricultural machinery power per cultivated land area), excluding municipalities/provincial capitals, excluding regions overlapping with national agricultural science and technology parks, placebo timing (setting 2010 as false treatment for later batches), PSM-DID (kernel matching on controls), and Callaway–Sant’Anna (CSDID) estimator to address staggered adoption and heterogeneous treatment timing.

- Baseline effect: NMADZs significantly increase ATI. DID coefficient ≈ 0.036 (Table 3, model 2) with region and year fixed effects and controls, significant at 1%, implying treated regions' ATI increased by about 3.6% relative to controls. Without controls, coefficient ≈ 0.057 (1% sig.). - Parallel trends and dynamics: Pre-trends largely insignificant; some anticipatory effects likely due to application processes. Post-treatment effects positive and grow, with rapid gains in years 4–5, indicating lagged policy impact. - Robustness: Effects remain significant when (i) using alternative ATI proxy (machinery power per cultivated land; DID significant, Table 4 col. 1), (ii) excluding centrally administered municipalities and provincial capitals (coef ≈ 0.009, 1% sig., Table 4 col. 2), (iii) excluding regions overlapping with national agricultural S&T parks (coef ≈ 0.016, 1% sig., Table 4 col. 3), (iv) placebo timing (Treat*2010 not significant, Table 4 col. 4), (v) PSM-DID (coef ≈ 0.045, 1% sig., Table 4 col. 5), and (vi) CSDID (coef ≈ 0.010, 1% sig., Table 4 col. 6). - Mechanisms (mediation): NMADZs increase mediators and mediators raise ATI. Effects on mediators: Finance α1 ≈ 0.160 (sig.), Government α1 ≈ 0.359 (sig.), Human α1 ≈ 0.413 (sig.). In ATI equations with mediators: Finance → ATI coef ≈ 0.007 (1% sig.); Government → ATI coef ≈ 0.040 (1% sig.); Human → ATI coef ≈ 0.001 (ns). Sobel tests for mediation all p < 0.01, indicating partial mediation via financial resource allocation, financial support to agriculture, and human capital. - Heterogeneity: Effects are stronger in regions with (i) higher fiscal autonomy (coef high group ≈ 0.021 vs low ≈ 0.004; difference significant), (ii) western region (coef ≈ 0.032 vs eastern ≈ 0.014; larger and significantly different), and (iii) better transportation infrastructure (high-infrastructure coef ≈ 0.012, 1% sig.; low-infrastructure group not significant, −0.003).

Findings confirm that the NMADZ policy effectively promotes agricultural science and technology innovation, addressing the central research question (Hypothesis 1). The demonstrated channels—enhanced financial resource allocation, increased fiscal support to agriculture, and improved human capital (Hypothesis 2)—align with theories of technology diffusion, agglomeration economies, and resource-based views. The stronger effects in regions with higher fiscal autonomy, in western China, and with superior transportation infrastructure (Hypothesis 3) underscore the importance of local fiscal capacity, development stage, and connectivity for realizing policy benefits. The observed lag in effects suggests time is needed for policy-driven investments, institutional improvements, and factor reallocation to translate into measurable innovation outcomes. These results extend prior qualitative findings by providing causal, region-level evidence and highlight how regional context conditions policy effectiveness, offering guidance for targeted implementation and scaling.

This study provides causal evidence that China's National Modern Agricultural Demonstration Zones significantly enhance agricultural science and technology innovation, with effects that are robust across multiple checks and that operate partly through financial resource allocation, fiscal support to agriculture, and human capital. The impacts are more pronounced where local fiscal autonomy is higher, in western regions, and where transportation infrastructure is better. Policy recommendations include: (1) leveraging local fiscal autonomy to tailor support, expand financial inputs, and cultivate professional farmers to amplify NMADZ impacts; (2) prioritizing NMADZ construction and financial institution support in western regions to accelerate modernization; and (3) investing in rural transportation infrastructure to reduce factor flow costs, enhance agglomeration, and facilitate knowledge and technology exchange. The findings contribute actionable insights for China and similar economies seeking to boost agricultural innovation through place-based policies.

- Anticipatory effects: Event-study indicates some anticipatory responses prior to approval, potentially biasing short-run pre-trend assessments. - Policy overlap: Some regions also hosted national agricultural science and technology parks; although analyses exclude these to isolate NMADZ effects, residual overlap cannot be entirely ruled out. - Data imputation: Missing region-year values were linearly interpolated, which may attenuate variability and affect precision. - Selection bias: Self-selection into NMADZs (application-based) is mitigated via PSM-DID, but unobserved selection cannot be completely excluded. - Scope and period: Results pertain to China (2007–2017) and may not generalize beyond this context or capture longer-run effects beyond the study window. - Sample adjustments: Placebo tests omit the first batch due to small sample size, which may limit inference about earliest adopters.