| 1.
| Design an appropriate testing methodology that takes into account the lender’s geographical footprint, types of AVMs used, lending program and other considerations.
|
| 2.
| Generate a sample of residential sales transactions. Boxwood employs its in-house national database of residential sales to pull a requisite number of records including a sufficient number of out-of-sample sales transactions. Samples are randomly selected, represent arms-length transactions only and include single-family, condominiums and cooperative sales unless otherwise directed by our client. We control the sample to provide adequate resolution for established geographical areas, pricing tiers and property types.
|
| 3.
| Create a benchmark price date for the transactions as a reference point for the vendor(s) when they receive the sample.
|
| 4.
| Promptly obtain the AVM estimates back from the vendor(s) after they run their values. Calculate various measures of "fit" for the AVM values including: overall coverage (i.e., "hit rate"); centrality; accuracy; risk ("underexposure" and "overexposure"); outlier frequency; and confidence score reliability for the various geographical areas, pricing tiers and property types.
|
| 5.
| Draft a validation report that presents the resulting data in a clear and understandable manner for clients, while preserving a high degree of statistical analysis for use/interpretation by regulatory authorities. Discuss the study’s results with the client and submit actionable recommendations related to AVM use and policy in a final report.
|
| 6.
| If requested, support the client in discussions of the study results with internal risk teams or outside regulatory agencies.
|
