Nonlinear coupling between digital economy development and regional income inequality in a reproducible multi-agent simulation framework

As a core driver of global economic growth, the digital economy reshapes regional economic landscapes and income‑distribution patterns. Existing work mostly estimates linear or single‑threshold relationships under static‑equilibrium assumptions, limiting the understanding of feedback between technology diffusion, capital reallocation and regional heterogeneity. This work adopts multi‑agent simulation to build an “economic‑geographical” dual‑gradient coupling model for interacting heterogeneous enterprises, residents and governmental agents on a spatial lattice. Simulations rely on a fully documented synthetic benchmark generated with fixed seed (1642026), without observational statistical, survey or satellite data; parameters are configured rather than econometrically calibrated. Therefore, outputs provide a reproducible simulation framework and mechanistic hypotheses rather than empirical estimates. Simulation results show that inequality variation is primarily explained by the technology‑diffusion gradient (52.8–66.4%), followed by capital‑flow (25.3–31.8%) and institutional gradients (8.3–16.4%). The technology‑capital contribution gap narrows from 41.1 pp in the Pearl River Delta to 22.0 pp in the Northwest at lower development levels. The DEI‑Theil link is monotonic with no interior turning point; thresholds 0.4, 0.6 and 0.8 are predefined model inputs, and the relationship disappears after first‑differencing (coefficient = 0.060, p = 0.262). Digital‑infrastructure investment stands out as the single policy tool reducing western inequality by 15.2% by 2040, while 0.5–50 km lattice‑resolution tests generate deviations below 0.054%. Owing to the synthetic benchmark, these results have no external validity.