“Hard” and “soft” signals: Bounded rationality and weighing performance versus compliance in multi-level governance

Principal-agent models often simplify how principals integrate heterogeneous signals from multiple subnational agents. We theorize that, under information overload, hard performance signals may condition the relevance of softer rhetorical alignment signals. Using a panel of annual Government Work Reports from 267 Chinese prefecture-level cities, linked to subsequent earmarked transfers over the 2005–2009 study window, we compare a simple additive account with the observable implications of a lexicographical heuristic. Textual policy compliance has no independent association with transfer allocation. In continuous interaction models, however, its association with transfers becomes positive within empirically favourable regions of economic vitality and social stability. This conditional pattern is consistent with the proposed informational hierarchy, although percentile-based dichotomizations are sensitive to the selected cut-off and do not identify a unique institutional threshold or a literal two-stage decision process. The findings extend bounded-rationality perspectives on multi-signal evaluation in China’s province – prefecture setting.