Realist Support: Data Definitions/Logic
Real AVM
Real AVM provides the estimated current market value for most residential
properties.
RealAVM uses a patented hedonic-based model that applies multiple
methodologies to set each value. RealAVM uses multiple modeling approaches
because the nature of property data is different from county to county, market to
market. It utilizes property comparables, appraiser emulation, home price
indexes, and other statistical methods. MLS data and neighborhood trends are
also included in these algorithms. After all of the valuation methods are complete,
the RealAVM engine reconciles the values generated by the various methods to
achieve a final value.
One method RealAVM uses to determine value is “appraiser emulation,” which is
strongest in areas with significant in-depth property characteristics data. It also
tends to be the best method for valuing properties in the absence of supply-side
information. A typical model search pulls all properties near the subject that have
a full, arm’s length transaction and the same use code as the subject. The distance
used for the search will depend on the density of properties surrounding the
subject property. RealAVM brings data comparables current using proprietary
indices. The average number of sales pulled per property is around 90, but it is
not uncommon for RealAVM to include 300 comparables to assign an estimated
value to a property.
Valuation methods such as indexing do not require as much characteristic data
depth to produce valuation results. Indices examine individual properties on
which there have been multiple sales over time. Those repeat sales – and the
indices derived from the repeat sales – are a predictor of value changes over time
that can be applied to subject properties.
RealAVM also uses MLS data, where available and permitted, to refine
characteristic information available for the subject and comparable properties.
RealAVM also applies MLS data to model current market conditions and their
influence on actual sales value.
Tax assessed valuation methodologies can be performed with less rigorous access
to characteristic data, but as with all modeling, the less data available to a model,
the less likely the result is to be accurate.
RealAVM manages data variances by using multiple or hedonic modeling like that
described above. RealAVM is tested on a daily basis through an exhaustive, blind
testing process known as our GeoAVM™ testing process.
Property valuations are produced on 100% of the residential housing stock on a
rolling basis. There is a dedicated team that insures those processes run in a
timely manner, and that also tune and maintain RealAVM for maximum accuracy
and performance.
In addition to the estimated value and value-range for a home, Realist also shows
the Confidence Score and the Forecast Standard Deviation for the valuations.
- The Confidence Score is a statistically derived measure indicating the
likelihood that the subject, comparable and accompanying data are reliable
and useful. Confidence Scores range from 60% to 100%, and higher
Confidences Scores indicate more robust data and ensuing accuracy. - The Forecast Standard Deviation is also statistically derived measure
showing the high/low range within which RealAVM is predicted to be
accurate. Lower Standard Deviations are positive and derive from reliable
and robust data.- For example, a RealAVM of $100,000 with a Confidence Score of 80 and a
Forecast Standard Deviation of 82 can be said to have an 8o%
probability to have a value of $100,000, plus or minus 18%.
- For example, a RealAVM of $100,000 with a Confidence Score of 80 and a