In the new decade a lot is expected to change in the way real estate is valued. DNB (2019) recently published a report where an increasing number of stakeholders express concerns about the quality and independence of the current valuation practice. Advances made in terms of data quality, computational power and econometric modeling provide opportunities to improve estimations based on historical evidence. But experts also know that a lot of progress still has to be made before full automation can be achieved. Furthermore, with the decision of the ECB to ban fully automated valuations for real estate mortgages (Tweede Kamer, 2020), the need arises for hybrid approaches where man and machines work in conjunction, each capitalizing their own skills. This article investigates the implementation of data-driven methodologies in the current (commercial) residential valuation practices from a valuer’s perspective and discusses findings from an experiment where model estimates are compared to manual valuations to analyze when and why the two might differ.
Authors: Bas Hilgers, Jurre Brantsma & Jacques Boeve