Regression models for platform-based administrative engagement with clustered administrative log data

Abstract Administrative workload in universities is difficult to observe empirically and is often measured using self-reported surveys or time-use diaries, which are subject to well-known limitations. This paper studies platform-based administrative engagement of university faculty using administrative log data generated by a digital academic management system. Using login and logout records from the Esse3 platform of an Italian public university over four academic years, we reconstruct session-level digital traces and aggregate them at the faculty member–academic year level. The final analytical dataset consists of 3,280 faculty-year observations referring to 877 faculty members. Rather than aiming to capture the full scope of administrative workload, the analysis focuses on observable digital interactions and examines how platform-based engagement varies with individual characteristics, academic rank, disciplinary field, and teaching-related administrative intensity. We estimate regression models with standard errors clustered at the department level and complement conventional inference with wild cluster bootstrap procedures to account for the limited number of higher-level clusters. The results suggest the presence of meaningful heterogeneity in platform-based administrative engagement across academic positions and teaching-related administrative intensity. Unconditional gender differences are not robust once individual characteristics, academic rank, year effects, and conservative clustered inference are taken into account. Extended specifications are interpreted cautiously, especially when explanatory variables are mechanically related to total platform time. From a methodological perspective, the paper illustrates how administrative log data can be used to study clustered behavioral outcomes in organizational settings, while also highlighting the interpretative limits of digital traces when offline activities are not observed.