DataQuick announced substantial enhancements to CMV-Portfolio, the company’s automated property valuation model (AVM). The enhancements are desiged to improve valuation accuracy and increase by up to 20 percent the likelihood that an automated property valuation will fall within 10 percent of the actual sales price.
CMV-Portfolio consists of four unique submodels that use different methodologies and sources of data to estimate the value of a property. Estimates are combined using a weighted average determined by property-specific accuracy measures of each submodel. More accurate submodels receive more weight when the values are combined. This approach provides more accurate and stable estimates than any of the individual submodel estimates alone by being less susceptible to extreme over- or under-valuations. It also significantly increases the number of properties that can be accurately valued and allows mortgage professionals to customize valuations based on what is most accurate for a specific market or property.
“Our analytics team has boosted the accuracy in each of these models in order to provide the most precise property valuations possible,” said John Walsh, president of DataQuick. “AVMs offer a cost effective valuation service through the life of a mortgage, going beyond the loan origination and underwriting process to provide valuation intelligence including monthly property value and mark to market value even after the loan is on the books. What sets CMV-Portfolio apart is its stable estimate based on multiple data methodologies and an interpretable confidence score which can be easily embedded into risk models.”
The enhanced submodels included in the CMV-Portfolio AVM include:
►Appraisal Emulation Model – Uses a comparable sales approach to estimate the market value of a property based on the value of similar properties in a specific area
►Inflated Sale Price Model – Estimates the market value by multiplying the most recent sale price of the property by how much the DataQuick Neighborhood Housing Price Index has changed since the most recent sale
►Tax Assessed Value Model – Estimates the ratio of the assessed value of a property to the sales price of sample properties in the vicinity to determine the market value
►Hedonic Model – Uses a robust set of property characteristics to construct a local property value regression model using nearby properties that have sold recently