Backwash or Spread? Absorptive Capacity and the Nonlinear Mechanism of Regional Convergence in China

Abstract

Despite prolonged government interventions, China's regional economy continues to exhibit a "convergence puzzle," characterized by persistent club solidification. Existing literature relies on the assumption of homogeneous spatial spillovers, overlooking how local absorptive capacity conditions the direction and size of external benefits. This paper incorporates endogenous technology diffusion into a neoclassical growth framework and estimates the moderating role of tech density in a Spatial Durbin specification with interaction terms, using Chinese provincial panel data and an identification-robust empirical design. The results indicate: (1) Distribution dynamics over 1992–2024 confirm pronounced club solidification, with retention probabilities of 0.933 for the low-income club and 0.990 for the high-income club. (2) On the 2011–2019 benchmark window, fixed-effects estimates of the moderation coefficient are attenuated toward zero, a pattern consistent with measurement error in the interaction term; an instrumental-variable strategy based on 2010 baseline tech density recovers a moderation coefficient of 0.048 (cluster-robust standard error 0.011), with a Kleibergen–Paap first-stage F of 30.5, an Anderson–Rubin weak-instrument-robust p-value of 0.0007, and wild cluster bootstrap p-values between 0.018 and 0.057 across weight schemes. (3) A Hansen-type grid search finds no sharp threshold: the residual sum of squares varies by only 3.3% across 21 candidate split points, and the regime-specific coefficients are statistically indistinguishable from each other. The moderation is a continuous and era-dependent mechanism: windows extending back to the 1990s yield a significantly negative interaction, while the positive moderation concentrates in the digital-era 2010s. Policy should therefore aim at strengthening absorptive capacity rather than crossing any single numerical threshold.