Using Object-Level Ratio Studies for the Incorporation of Validation Models in Single Family Residential Real Estate Assessment

Research output: Contribution to journalArticlepeer-review

Abstract

Purpose – Recent developments in the Dutch real estate assessment market have led to the emergence of ‘so-called’ validation models. These validation models run alongside already existing mass appraisal (initial) models. So far, there is no consensus on how these validation models can and should be used to improve the quality control for initial model assessments.
Methodology – this paper proposes a new methodology to compare the results of two automated valuation models by identifying nine possible situations that can occur between the results of the models. To test the proposed methodology, the automated valuation models are created using different statistical approaches. The models and the methodology are applied to two different Dutch municipalities.
Findings – the principal finding emanating from this study is that the proposed methodology does indeed allow for an in-depth analysis of the synergy between two separate automated valuation models, permitting object-level analysis to become a viability for properties with and without observed sales prices
Originality – this paper is the first paper that looks at how two separate models can be used alongside one another instead of looking at which automated valuation model outperforms the other with respect to accuracy and uniformity. Furthermore, this paper calls for a ‘re-operationalization’ of market value with respect to ratio studies.
Original languageEnglish
JournalJournal of Property Tax Assessment and Administration
Publication statusAccepted/In press - 4 Feb 2022

Keywords

  • Automated Valuation Models
  • Validation models
  • Gradient Boosting Machines
  • Model-to-Model validation

Fingerprint

Dive into the research topics of 'Using Object-Level Ratio Studies for the Incorporation of Validation Models in Single Family Residential Real Estate Assessment'. Together they form a unique fingerprint.

Cite this