An Analysis of the Accuracy of Automated Valuation Model and Transparency Management Plan

This study provided measures to improve the accuracy and reliability of automated valuation models (AVMs), which have been gradually expanding their scope of use. The US market, which introduced AVMs earlier than other markets, has improved the requirements for AVM fairness, accuracy and reliability through the Dodd-Frank Act, while the South Korean AVM market mostly remains unregulated, leaving AVM firms to regulate themselves. When valuations of domestic AVM firms which are possible to be massively obtained were analyzed by mainstream statistical standards such as median absolute value percentage error and percentage predicted error@10%, levels of accuracy were different in regional types and the same result applied to apartments and officetels. This suggests that even for the same AVM company, the level of accuracy could be different from regional types and that it is necessary to present accuracy standards for each region. In addition, when the accuracy of AVM is stable and validated, it could be used as a means to prevent non-performing loans and Jeonse fraud due to the gap between market and appraised values. Moreover, the scope of the AVM could be expanded to monitoring collateral values in the secondary market of real estate-backed securities and inspecting investment portfolios. To achieve this, it is necessary to establish accuracy standards for them and introduce an third-party, non-profit and independent verification agency.