Abstract:

Where students live shapes whether they can study, sleep, and participate in university life. Most institutional housing surveys, however, reduce this complexity to single campus-wide averages that treat all students as facing the same problem at different intensities. A student living close to campus but under severe financial pressure occupies a fundamentally different situation from one whose rent is affordable but whose flat is noisy, poorly heated, and badly maintained — yet both might give the same overall satisfaction rating of 3 out of 5. They would appear identical in a campus-wide average but need entirely different kinds of institutional support. This paper argues that a profile-based approach is both more accurate and more actionable than conventional satisfaction measurement.

We apply Bernoulli latent class analysis (LCA) to a bilingual survey of 158 students at the University of Koblenz, a German university city whose river-and-bridge geography creates real spatial heterogeneity in accessibility. The survey covered three multi-select questions, what students valued when choosing accommodation, what they currently appreciate about their housing situation, and what challenges they regularly face, producing 40 response options. After removing one option endorsed by fewer than 3% of respondents, 39 binary indicators were retained for analysis. Models with two to six latent classes were estimated using an expectation-maximisation algorithm with 150 random initialisations across multiple random seeds. A four-class solution was selected on the basis of interpretability and entropy-based class separation.

The four profiles are: students constrained primarily by commuting distance and privacy deficits (26%, n=41); students living close to campus but under sustained financial pressure (20%, n=32); students whose rent is manageable but whose everyday living conditions are persistently frustrating due to noise, poor maintenance, and inadequate facilities (25%, n=40); and students in genuinely study-supportive environments who report good internet, thermal comfort, and quiet (29%, n=45). The profiles differ significantly in willingness to remain in current accommodation (χ², p=0.005) and willingness to recommend it to peers (p=0.002), with the study-supportive group showing the highest endorsement rates and the commuting-and-privacy-constrained group the lowest.

The findings carry differentiated implications for student services, university administrators, housing providers, and students. For student services, the profiles provide a diagnostic framework beyond satisfaction scores, identifying which combination of problems a student is likely facing and enabling more targeted outreach. For housing providers and municipalities, the profiles describe the specific quality features students identify as most important — not an abstract index, but concrete combinations of noise levels, internet reliability, thermal comfort, room size, and proximity to campus. For students arriving in a new city without local knowledge, the profiles offer a vocabulary for the trade-offs they are likely to encounter. A profile-based approach surfaces the specific combinations of problems that different groups of students are actually living with, giving institutions something far more actionable than a single satisfaction score.

Keywords:

Student housing, higher education, latent class analysis, need segmentation, student support, housing policy.