Over the past several years there has been extensive discussion regarding the need for more support for the quantitative analysis in appraisals. Continuing education providers now offer more classes than ever before to address this need. Many focus on new methods for adjustment support and statistical analysis. The available technological tools to perform quantitative analysis have also increased exponentially.
The historical shortcomings of the appraisal industry as a whole have been widely chronicled. As with many other historical contexts, when the pendulum begins to swing, it often swings too far. In this case, the pendulum consists of “big data” on one side of the scale and the appraiser’s analysis on the other side of the scale. The problem? The pendulum is now off balance.
There are issues with over-reliance on statistics. One of the primary lessons in statistics is the big difference between correlation and causation. The data and tools that are available today offer better documentation of correlated factors than we’ve had in the past. However, a correlation between two factors is not the same as a cause and effect relationship between those two factors.
My favorite example of this is the correlation between ice cream and murder. As a researcher once noted, when sales of ice cream increase, so do the number of murders. While the correlation is statistically valid, it would be absurd to suggest a cause and effect relationship exists.
There is no substitute for an educated human mind to know the difference between correlation and causation. This is particularly true in the world of real estate appraisal, where the number of variables is usually high, and the sample size is usually low.
For example, in a recent study of the relationship between oversized lots and sale price I noticed a trend. In nearly every pool of data, the correlated difference exceeded the causal difference by a significant margin. In one instance the correlated difference was $5/sq. ft., whereas a paired sales and sensitivity analysis resulted in a supported adjustment at $3/sq. ft. based on the same set of data. Why? In this case, homes with oversized lots tended to have superior outdoor amenities such as larger patios, outdoor kitchens, or larger pools. These factors were correlated to differences in lot size, but not caused by it. Had I relied upon the “big data” approach for adjustment support I would have over adjusted for differences in lot size.
It is the ability to interpret data, understand relationships, and explain the psychology of buyers and sellers that make the appraiser indispensable. Correlation and causation issues are just one of a myriad of factors in valuation that are best served by the human touch. Those of us that care about the valuation industry, the mortgage industry, and the impact of a healthy real estate sector need to take these factors to heart. How should we define a ‘supported’ adjustment? How reliable is an ‘appraiser-assisted’ AVM? What scope of work can best limit costs without sacrificing reliability? As we continue to seek answers to these questions we will move closer to better balancing that pendulum.