Valuation of the collateral pool of mortgage-backed securities and automated valuation models

This study examines securitization and characteristics of secondary mortgage markets and analyzes prepayment and default risks for the valuation of mortgage-backed securities produced through securitization. 'Refinancing incentive' resulting from falling mortgage interest rates and 'negative equity' which occurs when the value of property falls below the outstanding mortgage balance are shown as the main drivers of the prepayment and default risks respectively. However, risk analysis made with Competing Risks Cox Regression Analysis and two machine learning methods (Multinomial Logistic Regression and Random Forest) indicate that mortgage borrowers are heterogenous in their repayment behavior, refinancing incentive and negative equity are not sufficient in explaining mortgage termination risks. Therefore, characteristics of mortgage loans and borrowers, and local economic factors should be considered in mortgage risk analysis. Afterwards, considering the negative impacts of default risk on mortgage-backed securities, mortgage markets and country economies, the role of automated valuation models is examined in valuation of properties which serve as the collateral of the mortgages and therefore of the mortgage-backed securities, quality control of valuations and monitoring the values until the maturity date. Instead of the conventional property valuation implementations, automated valuation models developed with statistical models and tests, and alternative valuation approaches developed with the support of these models offer fast, reliable, and cost-effective solutions. It is understood that automated valuation models already being used in various countries for wide range of purposes including revaluation activities and production of real estate price indices and portfolio valuation will continue to keep by increasing their importance in the functioning of secondary mortgage markets. It is obvious that new models with more accurate predictions will be developed by improving data quality and transparency simultaneously with technological developments and these models will become the key to the transformation of valuation profession to a consultancy.