Governance Surprise, Attention Allocation, and Bureaucratic Responsiveness Heterogeneity in Digital Government

Why do bureaus operating within the same digital government infrastructure respond so differently when governance conditions shift unexpectedly? This paper introduces governance surprise—defined as the Kullback-Leibler (KL) divergence between the current distribution of citizen complaints and the bureau's seven-day prior expectation—as a formal measure of demand-side uncertainty. Drawing on bounded rationality theory and attention-based accounts of organizational behavior, we argue that cross-bureau responsiveness heterogeneity under environmental shock reflects differences in organizational attention capacity. Using 1,899 bureau-day observations from the 12345 Government Service Hotline in Sanya, China, we estimate a Bayesian hierarchical linear model. Results confirm three findings: (1) governance surprise is positively associated with response time on average; (2) bureau-level surprise coefficients are highly heterogeneous, revealing agile adapters and rigid respondents; and (3) baseline efficiency is negatively correlated with shock resilience (r =  − 0.56, p < 0.05)—an efficiency-resilience asynchrony that challenges current digital government performance frameworks.