Search results

The purpose is twofold. First, this study aims to establish that black box tree-based machine learning (ML) models have better predictive performance than a standard linear regression (LR) hedonic model for rent prediction. Second, it shows the added value of analyzing tree-based ML models with interpretable machine learning (IML) techniques.

Data on Belgian residential rental properties were collected. Tree-based ML models, random forest regression and eXtreme gradient boosting regression were applied to derive rent prediction models to compare predictive performance with a LR model. Interpretations of the tree-based models regarding important factors in predicting rent were made using SHapley Additive exPlanations (SHAP) feature importance (FI) plots and SHAP summary plots.

Results indicate that tree-based models perform better than a LR model for Belgian residential rent prediction. The SHAP FI plots agree that asking price, cadastral income, surface livable, number of bedrooms, number of bathrooms and variables measuring the proximity to points of interest are dominant predictors. The direction of relationships between rent and its factors is determined with SHAP summary plots. In addition to linear relationships, it emerges that nonlinear relationships exist.

Rent prediction using ML is relatively less studied than house price prediction. In addition, studying prediction models using IML techniques is relatively new in real estate economics. Moreover, to the best of the authors’ knowledge, this study is the first to derive insights of driving determinants of predicted rents from SHAP FI and SHAP summary plots.

This study aims to examine whether the methodology proposed by the European Supervisory Authorities (ESAs) within Delegated Regulation (European Union) 2017/653 for the calculation of market risk of certain packaged retail and insurance-based investment products (PRIIPs) is the most appropriate.

Risk models are put into effect to validate the appropriateness of the methodology announced by ESAs. ESAs have announced that the unit-linked (UL) products, labeled as Category II PRIIPs, will be subject to the Cornish–Fisher value-at-risk (CFVaR) methodology for their market risk assessment. We test CFVaR at 97.5% confidence level on 70 UL products, and we test Cornish–Fisher expected shortfall (CFES) at the same confidence level, which acts as a counter methodology for CFVaR.

The paper provides empirical insights about the Cornish-Fisher (CF) expansion being a method that incorporates the possibility of financial instability. When CFVaR by ESAs is calculated, it is shown that CF is in general a more robust risk model than the simpler historical ones. However, when CFES is applied, important points are derived. First, only in half of the occasions the CF expansion can be considered as a reliable method. Second, the CFES is a more coherent risk measure than CFVaR. We conclude that the CF expansion is unable to accurately estimate the market risk of UL products when excessive fat-tailed or non-symmetrical distributions are present. Hence, we suggest that a different methodology could also be considered by the regulatory bodies which will capture the excessive values of products in financial distress.

Literature, both theoretical and applied, regarding PRIIPs, is not extended. Although business and regulators research has begun to intensify in the last two years, to our knowledge this is one of the first studies that uses the CFES methodology for market risk assessment of Category II PRIIPs. In addition, we use a unique data set from a country in the headwinds of the recent financial crisis. This research contributes both to the academic and business community by enriching the existing literature and aiding risk managers in assessing the market risk of certain Category II PRIIPs. Considering the recent efforts of the regulatory authorities at the beginning of 2020 to implement certain amendments to the PRIIPs, we indicate relative risks related with the calculation of the market risk of the aforementioned products. Our findings could contribute to regulatory authorities’ persistent efforts in wrapping up this ongoing project.