Residential attractiveness and housing prices in belgian municipalities: A composite indicator approach

Details

Supervisors

Vande Kerckhove, Corentin ; Mané, Braima

Faculty

Louvain School of Management

Degree label

Master [120] : ingénieur de gestion, à finalité spécialisée

Abstract

enThis thesis investigates the relationship between residential attractiveness and housing prices across Belgian municipalities from both explanatory and predictive perspectives. Residential attractiveness is defined as the capacity of a municipality to attract and retain residents through territorial characteristics likely to influence residential choices. To operationalise this concept, a Residential Attractiveness Index (RAI) is constructed for 565 Belgian municipalities over the 2015–2024 period. The index combines 21 indicators grouped into eight dimensions: Accessibility, Demographics, Economy, Governance, Urbanism, Housing, Infrastructure, and Social Conditions. The empirical analysis combines econometric and machine learning methods. Cross-sectional relationships are estimated using between estimators with spatial standard errors, while temporal dynamics are analysed through two-way fixed effects models. Predictive performance is assessed using expanding-window cross-validation and hold-out forecasting at horizons of 0, 1, and 3 years with Ridge regression, Random Forest, and XGBoost. The results highlight a strong explanatory but limited predictive role for residential attractiveness. Cross-sectionally, municipalities with higher RAI scores consistently exhibit higher housing prices, even after controlling for regional and provincial fixed effects. However, within municipalities, year-on-year changes in the RAI are not significantly associated with housing price adjustments. From a predictive perspective, Ridge regression consistently outperforms the non-linear models, while the composite RAI provides only limited forecasting gains above price persistence alone. Models based on the 21 individual indicators systematically outperform the composite-based specifications, suggesting that aggregation into a single score removes part of the predictive information contained in the underlying variables. Predictive performance also deteriorates substantially as the forecasting horizon increases. Overall, the findings suggest that residential attractiveness is more strongly associated with persistent spatial differences in Belgian housing prices rather than with explaining or predicting their short-term evolution. This thesis contributes to the literature by constructing a municipal-level composite index of residential attractiveness in Belgium, by systematically comparing its explanatory and predictive roles, and by showing that disaggregated indicators outperform the composite in forecasting tasks.