Regional digital finance showed no stable resilience association in Chinese food firms

A public-data study of emerging listed food enterprises in China found no stable positive within-firm association between regional digital finance and an accounting-based proxy for food supply-chain resilience. The preferred two-way fixed-effects estimate was −0.005, with a standard error of 0.063 and a 95% confidence interval from −0.128 to 0.117. The analysis covered 732 firm-year observations from 119 companies between 2017 and 2023. Its authors stressed that the imprecise result did not establish practical equivalence to zero and could not rule out a meaningful positive association. The study therefore supports neither a conclusion that regional digital finance improves resilience nor one that it has no effect.

The analysis examined a deliberately bounded population. Emerging listed food enterprises were operationally defined as companies no more than 15 years past listing, a restriction intended to focus on listed businesses combining growth pressure, financing needs and developing operating routines rather than mature enterprises with established distribution, financing and management systems. The initial roster contained 188 A-share firms in Chinese food industries classified under CSRC codes C13, C14 and C15. After excluding firms with special-treatment status, missing profile or province information, unmatched firm-year records and listing ages above 15 years, 122 eligible companies remained. Public financial statements covering 2015–2023 provided 992 firm-year rows, from which 748 valid resilience observations for 119 firms were reconstructed. The common regression sample contained 732 observations from 2017 through 2023.

Food supply-chain resilience was not observed directly. Instead, the researchers constructed a composite accounting proxy from resistance, recovery and stability in revenue, operating cash flow and net margin. Resistance captured inverse adverse annual movement, recovery represented an accounting rebound following a prior decline, and stability measured inverse volatility over a rolling three-year period. An entropy-weighted combination integrated nine indicators, while separate resistance, recovery and stability measures were used in sensitivity tests. Financial records from 2015 and 2016 supplied the lagged and rolling inputs needed for the first eligible 2017 observations, and missing outcome components were excluded rather than imputed. This construction is transparent and reproducible, but it does not directly register supplier failures, logistics interruptions, cold-chain losses, stockouts, spoilage, food-safety incidents or traceability breakdowns. A recovery score may also reflect mean reversion or an ordinary business cycle rather than a rebound from a verified external disruption.

Regional digital finance was measured with the Peking University aggregate Digital Financial Inclusion Index and matched to each company’s registered province and reporting year. It represented the surrounding financial environment, not direct evidence that a company used digital payments, platform credit, digital procurement, inventory software or other enterprise-level services. This separation defines the study’s proposed regional-to-firm conversion gap: a digitally advanced province may make more financial options available, but a business must still translate those options into supplier settlement, inventory governance, quality assurance, traceability and disruption-response routines. The theoretical relationship is also not necessarily positive. Easier finance could help a company retain suppliers, maintain safety stock, fund inspections or support cold-chain investment. At the same time, it could finance more stock-keeping units, distributor expansion, marketing or larger inventory commitments, potentially increasing spoilage exposure and coordination complexity before operating routines can absorb the expansion.

The model sequence progressively separated broad regional differences from changes within individual firms. The digital-finance coefficient was 0.002 in pooled ordinary least squares and 0.002 in the random-effects model, 0.044 with firm fixed effects, −0.005 with firm and year fixed effects, and 0.059 after province-specific trends were added. This changing pattern did not support a stable positive within-firm relationship. Firm effects removed time-invariant differences such as business model, brand heritage, listing cohort, registered location and product category. Year effects absorbed common national patterns, including nationwide regulatory changes, pandemic disruption and general growth in digital finance. Province-trend specifications addressed differential regional trajectories but sacrificed identifying variation and produced wider intervals. Even so, the fixed-effects design is an associational test: it cannot fully remove time-varying regional confounding, reverse adaptation or geographic measurement error.

Power and equivalence diagnostics showed why the near-zero point estimate could not be treated as proof of no relationship. The standardized digital-finance coefficient was −0.046, while the approximate 80% minimum detectable effect was 1.575 standard deviations of the resilience proxy. Equivalence tests using bounds of 0.05 and 0.10 SCR standard deviations were not significant, with p-values of 0.497 and 0.462, so the researchers did not establish practical equivalence. The confidence interval remained compatible with both modest adverse and modest beneficial associations. The short panel also left relatively little independent variation in province-year indicators after persistent firm differences and common annual patterns were removed. Raw regional digital finance and provincial household consumption expenditure were strongly correlated at 0.823, although their residualized correlation fell to 0.090 after firm and year means were removed, making conditional estimates difficult to estimate precisely.

Neither proposed indirect mechanism was established by the available proxy-based evidence. The association between regional digital finance and the entropy-weighted operational-capability measure was −0.038, with a 95% confidence interval from −0.119 to 0.043. Operational capability’s association with the resilience proxy was 0.009, with an interval from −0.064 to 0.082. The operational measure combined inventory turnover, receivables turnover, asset turnover and the inverse cash-conversion cycle. It was treated as an outcome-oriented proxy for operating efficiency rather than a direct measure of enterprise-resource-planning deployment, supplier-platform integration, replenishment routines, cold-chain monitoring, traceability or disruption protocols. Equal-weight variants were similarly imprecise. A single-indicator test found a positive association for asset turnover, but the study did not treat that result as validation of the complete pathway. Tests of whether regional consumption capacity moderated the digital-finance relationship were also inconclusive: the linear interaction was −0.016, with a 95% interval from −0.052 to 0.019, and the quadratic interaction was 0.006, with an interval from −0.004 to 0.016.

Robustness checks did not reveal a consistent positive pattern across credible definitions and samples. The digital-finance estimate remained close to zero or changed sign when researchers used an equal-weight resilience outcome, isolated its resistance, recovery or stability components, substituted alternative digital-finance dimensions, changed covariance estimators, excluded municipalities, restricted the sample to balanced panels or shifted exposure by one or two years. Province-clustered standard errors yielded an estimate of −0.005 with a 95% interval from −0.118 to 0.108, while Driscoll–Kraay inference produced −0.005 and an interval from −0.131 to 0.121. One-year and two-year lag estimates were −0.084 and 0.060, respectively, and both were imprecise. Four theory-led interactions tested whether baseline size, maturity, operational capability or inventory pressure changed the association. None remained statistically significant after Benjamini–Hochberg or Holm adjustment, making the subgroup analysis exploratory rather than evidence that particular firms respond differently.

The study’s main contribution is diagnostic: it identifies where contextual financial development may fail to become an observable firm-level resilience outcome and specifies the operational evidence needed to examine that conversion more directly. Its findings remain constrained by the annual accounting proxy, registered-province matching, short 2017–2023 estimation period, limited statistical power and sample of recently listed companies. Registered location may not represent the places where a company produces, stores goods, obtains inputs or sells products, and such measurement error could attenuate estimates toward zero. The analysis also cannot observe direct digital-finance adoption or responses to identified supply-chain shocks. Future research would need information on enterprise adoption, plant and warehouse locations, supplier networks, cold-chain performance, product-level demand, traceability and disruption-specific operating responses, along with designs that can better separate contextual exposure from firm-level conversion. Until such evidence is available, the most defensible conclusion is that this panel did not demonstrate a stable benefit from regional digital finance, while also lacking the precision to conclude that meaningful resilience associations are absent.

Subject of Research: Associations between regional digital finance and accounting-based supply-chain resilience among emerging listed food enterprises in China

Article Title: Regional digital finance and food supply-chain management resilience: evidence on the regional-to-firm conversion gap in China’s emerging listed food enterprises

Article References: Zhao, J. (2026). Regional digital finance and food supply-chain management resilience: evidence on the regional-to-firm conversion gap in China’s emerging listed food enterprises. Frontiers in Sustainable Food Systems, 10, Article 1948192. https://doi.org/10.3389/fsufs.2026.1948192

Image Credits: AI Generated

DOI: 10.3389/fsufs.2026.1948192

Keywords: regional digital finance, food supply chains, supply-chain resilience, China, listed food enterprises, fixed-effects analysis, operational capability, regional consumption, statistical power, Regional, digital, finance