A Markowitz Optimization Approach for Automating the Italian Research Quality Monitoring and Evaluation

This paper presents a selection-supporting framework which can help research institutions to optimally and automatically pool the research products for research quality assessment programs, with a specific focus on the Italian evaluation process (VQR). After providing a mathematical description of the VQR exercise at the institutional level, we formulate the robust optimization problem which yields the optimal pair of research products and associated authors. We show how such a formulation quickly becomes unpractical, due to combinatorial issues, and propose a novel Markowitz-based alternative approach, which preserves computational feasibility and effectiveness. We strive to propose a reusable framework, not too tightly connected to the ruleset of the current VQR session (2020-2024). Finally, we validate the proposed framework on a synthetic set of parameters, which mimics a medium-sized research institution, with the aim of checking the computational feasibility of the proposed Markowitz-based variants.