Implementing Automated Valuation Models (AVMs) in Absence of Reliable Transaction Data

Automated valuation models (AVMs) are predictive algorithms used to estimate market value based on property characteristics and market conditions. AVMs were first proposed over 100 years ago and have demonstrated an ability to achieve more efficient, accurate, equitable, and cost-effective valuations than alternative mass appraisal methodologies. Because AVMs are trained using transaction data, if data integrity is compromised (e.g. under-reported sales prices; missing, incomplete, inaccurate, or outdated fields), market value estimates will be inaccurate. Various constraints related to land cadasters, sales registries, data laws, and market transparency, have left governments outside of America, Canada, and western Europe at a disadvantage with regard to AVMs. This presentation will discuss methodologies and work flows to help overcome such constraints so the benefits of AVMs may be realized by more countries. Various research and country case studies (e.g. Moldova, China, Indonesia) as well as IAAO standards will be reviewed.