Network data envelopment analysis in the insurance industry: a literature review

Abstract

Data envelopment analysis (DEA) and, more specifically, network DEA (NDEA) have been extensively employed in recent decades to assess the efficiency and performance of insurance companies. Yet, despite this growing body of work, there is still no coherent and structured overview of which network structures have been adopted, how insurance processes have been modeled using specific performance indicators, and where the main gaps in the existing literature lie. The purpose of this study is to provide a comprehensive and systematic mapping of NDEA applications in the insurance sector. Based on a structured search of the Web of Science database, we identify 83 studies that employ network DEA models to evaluate the performance of insurance-related decision-making units (DMUs). Our analysis proceeds in two main stages. First, we conduct a structure-based review in which the network structures employed in the literature, including two-stage, series, and mixed structures, are classified within static and dynamic frameworks. For each study, we document the families of inputs, outputs, intermediate products, and, where applicable, intertemporal link variables (carry-overs) that connect different periods. Second, we carry out a feature-based review that classifies the studies according to geographical region, line of insurance business, type of methodological innovation in the network model, returns-to-scale (RTS) assumptions, model orientation, presence or absence of explicit uncertainty modelling, and theoretical- or application-based research orientation. The results show that static two-stage network structures overwhelmingly dominate the existing literature, whereas dynamic network formulations and variables related to risk, investment activities, and regulatory requirements remain relatively underexplored. The findings also reveal an imbalance between theoretical developments and empirical applications, highlighting the need for greater alignment between methodological advances and the operational characteristics of insurance processes. On this basis, we identify several research gaps concerning both the choice of network structures and the design of performance indicators, thereby offering a structured research agenda for future theoretical developments and empirical applications of NDEA in the insurance industry.