The demand for purpose-built student accommodation (PBSA) has attracted significant attention over the last decade, particularly in developing countries, due to the inability of higher education institutions (HEIs) to provide adequate hostels for their growing student population. This paper examines the investment performance of this real estate asset class in the context of holding period returns (HPRs) and return dispersion across three PBSA market segments over 2014–2023. This study adopted a quantitative research methodology using a descriptive case study strategy. This approach incorporated personal observation and baseline data collection from individual PBSA properties. Data on annual rent and outgoings were collected from 73 PBSA properties, totalling 3058 rooms and 5889 bed spaces within the studentified neighbourhood of Calabar. The study assessed investment performance using mean annual HPRs, and examined return dispersion using within-year property standard deviation (SD), individual property temporal SD, and category-portfolio temporal SD. Mean annual HPRs were broadly comparable across the three PBSA segments, ranging from 9.46% for middle-end PBSA to 9.98% for low-end PBSA and 9.97% for high-tech/premium PBSA. However, the segments differed more clearly in return dispersion. The mean within-year property SD was 6.27% for high-tech/premium, 6.34% for middle-end, and 8.23% for low-end PBSA. Individual property and category-level temporal variation also differed across the segments, indicating that similar average returns can coexist with different patterns of return variability across properties and over time. PBSA investment appraisers should consider return dispersion alongside average HPR. The findings provide investors, financiers, and valuers with segment-specific evidence to compare PBSA opportunities and assess property-level variability. While earlier studies have largely relied on aggregated property data to assess investment performance, this study extends PBSA investment research through longitudinal, property-level evidence from an emerging market. It distinguishes return dispersion across PBSA properties and over time, providing a basis for understanding PBSA investment performance in the emerging PBSA property market.
Mass appraisal of real estate is a central challenge for municipal tax administration, which must estimate assessed values fairly. Real estate market is strongly related to location, which is difficult to measure. Spatial autocorrelation in property prices violates ordinary least squares assumptions and potentially biases inference. This study investigates Moran’s Eigenvector Spatial Filtering (ESF) and its random-effects specification (RE-ESF) for land price modeling, comparing them with a hedonic ordinary least squares model and three classical spatial specifications (SAR, SEM, and SARAR). The analysis uses 2635 urban land records from Fortaleza, Ceará, Brazil, the country’s fourth most populous city, with a real estate stock exceeding 850,000 properties, and a row-standardized 30-nearest-neighbor weights matrix. All compared models share the same explanatory variables and are validated out of sample on the same random 80/20 partitions, compared in paired tests; residual autocorrelation is assessed against the Cliff–Ord null hypothesis, the appropriate reference for regression residuals. The filtered models remove the residual autocorrelation left by the unfiltered model and improve out-of-sample accuracy by a small but consistent margin: 0.022 in R2 for ESF in 18 of 20 partitions, with the spatial error model as the closest classical competitor in R2, though it is the worst of the four unfiltered specifications in mean absolute percentage error. The conclusions hold for neighborhood sizes from 10 to 40 and under a distance band. By removing the spatially systematic component of assessment error, spatial filtering improves the geographic dimension of assessment equity and thereby supports municipal fiscal management.
The period 2018–2026 offers the first opportunity to observe a complete, closed mortgage interest rate cycle in the Czech Republic: rates fell to a historic trough of 1.98% (March 2021), rose to 5.37% (September 2023) and stabilized around 4.5% from 2025. This paper examines whether the relationship between mortgage rates and housing prices is reversible over such a cycle. The analysis draws on the EVAL software (version 2.0), which has continuously collected and evaluated real estate advertising across all Czech municipalities since 2007, combined with Czech National Bank and Czech Statistical Office data; methods include phase-specific logarithmic regressions, Chow tests, asymmetric difference regressions, cross-correlation analysis and a composite affordability indicator, the Mortgage Payment Burden (MPB). The point estimates are asymmetric: rate increases are followed by lower listing volumes within four months and lower prices within seven, whereas rate cuts show no detectable direct effect; the difference between the two responses is, however, not statistically significant at the available sample size, so the asymmetry is suggestive rather than conclusive. The price–rate trajectory forms an open hysteresis loop—at an identical rate of 4.5%, the median price was 35.5% higher after the cycle than during tightening. Adjustment ran through quantities rather than prices; the correction was deepest in cheap peripheral regions, and by April 2026 the national MPB reached 41.2% of the average wage, its historical maximum—a result that survives plausible alternative wage-growth assumptions. The evidence is consistent with monetary policy acting as an effective brake but a weak accelerator of the housing market, although the reduced-form design cannot separate the effect of rates from that of the concurrent macroeconomic shocks.
Real estate valuation increasingly requires methodologies that combine predictive accuracy, economic interpretability, and model transparency. This study proposes a Generalized Explainable Hybrid Econometric–Machine Learning Framework integrating econometric modelling, machine learning, and Explainable Artificial Intelligence (XAI) within a unified analytical architecture. The framework is organized around a generalized conceptual valuation equation and implemented through a workflow comprising data quality assessment, econometric benchmarking, hyperparameter optimization, nonlinear machine-learning modelling, five-fold cross-validation, and SHAP-based explainability. The methodology is demonstrated using an internally constructed Azerbaijan commercial real estate dataset and independently assessed using the publicly available UCI residential benchmark dataset through an identical Python-based computational pipeline. Under the complete income-capitalization benchmark, the Artificial Neural Network achieved the highest predictive performance for the Azerbaijan dataset, whereas Random Forest performed best for the external benchmark. A leakage-reduced robustness analysis further showed that the framework retained meaningful predictive capability after excluding Net Operating Income and the Capitalization Rate, with XGBoost providing the strongest performance under the reduced specification. SHAP analysis identified economically meaningful valuation drivers, highlighting income generation in the commercial market and transport accessibility in the residential benchmark. The proposed framework provides a transparent, explainable, and transferable decision-support methodology for modern real estate valuation.
