As a major economic engine of China, the Yangtze River Delta (YRD) region is pivotal in driving innovation across the scientific, technological, and digital economies. This study constructs the spatial associative network for digital economy innovation by treating 41 cities as nodes and applying a gravity model adjusted for institutional distance. Subsequently, the structural characteristics of the spatial associative network and their effects were empirically explored by using complex network analysis and regression models. The findings indicate that: (1) The linkages in digital economy innovation among cities are becoming increasingly closer, and the network structure exhibits an annual increasing trend in density, connectivity, and efficiency, while hierarchy decreases; (2) The examination of network node characteristics discloses that different cities possess diverse capabilities in terms of resource aggregation, regulation, and communication. The block model analysis further categorizes the cities into four functional groups. Among them, Block I (including cities like Shanghai, Nanjing, and Hangzhou) holds the “primary” status and acts as the “core city” for digital economy innovation; (3) The attributes of the spatial associative network have a remarkable effect on both the degree of digital economy innovation and the variations among cities.
Regional inequality is commonly measured with indices (e.g., the Gini and Theil indices) that depend only on the multiset of values and are therefore permutation-invariant: they cannot distinguish configurations differing solely in how the same values are ordered. This paper formalises that limitation as an information boundary and shows that the multifractal spectrum width (Δα) carries qualitatively distinct information. Four propositions are established: (1) with group membership fixed, the Gini, Theil, generalised entropy and Dagum Gini indices are permutation-invariant; (2) Δα is not, and can move in either direction under permutation; (3) Δα still varies appreciably when permutation-invariant indices are held constant, so the two classes are not informationally equivalent; and (4) controlled synthetic fields serve as a proof of concept linking Δα to multiscale heterogeneity. Permutation invariance secures the objectivity and comparability of inequality indices, but it also delineates their information boundary: Δα carries structural information that these indices cannot, in principle, capture, making the two complementary rather than substitutable. A finite-sample baseline and the Multifractal Magnitude Index (MMI) yield an operational diagnostic protocol; because Δα is defined only relative to the ordering rule and the MF-DFA parameterisation, all comparisons are made within a single parameterisation.