Appraisal standards and regulations are mostly designed to discourage personal bias.  But what about traditional appraisal practice?

To understand this question, we must separate personal bias from analytic bias.

Personal bias can be open, declared prejudice – or it may grow from basic human instinct – towards those who look like us.  This basic tribal response comes from our lizard brain (limbic system).

Analytic bias is different.  It is the result of wrong data selection (picking comps), or wrong adjustments.  Analytic bias can be from incompetent model selection, or a vehicle for conscious personal bias.

What is important is that USPAP (Uniform Standards of Professional Appraisal Practice) says nothing about analytic bias.  Only that an appraiser should be independent, impartial, and objective.  USPAP only addresses personal bias.  A lot of “shoulds.”

Tradition, habit, and groupthink in appraisal practice: “find a set of comparable sales … as similar as possible to the subject property.”  Usually this means three to six recent sales.  These three to six ‘comps’ may or may not reflect the actual number of competing properties as of the date of value.

But what if there are more than six, or less than three?

If there are fewer than three competitive sales, we just use those.  But.  We have to expand our data set to other sales which do not compete directly.  No matter what additional sales we pick, there will be some resulting selection bias.  Each and every sale we use is not exactly at “market price.”  Each has some built-in variation.  That individual variation causes some bias.  Bias from selection.

If there are more than six actual directly competitive sales, (say 15)

  • If I pick less than 15, say 3, the same concept happens again. The 3 will have selection bias.  
  • If I pick more than 15, say 45 (to pack my regression), there will be more variance (or more uncertainty. This is because the additional 30 are not directly competitive, and vary randomly.

This is called the “bias-variance tradeoff.”  Too little data begets bias.  Too much data begets variation, lower reliability, and more risk.  [Wrong data always causes bias and can conceal intentional bias.]

Here we have a conflict between USPAP and traditional appraisal practice.

  • Practice and tradition make us use three to six comps.
  • USPAP says “analyze such sales data as are available”, and “all information necessary.” [SR1-4]
  • The data science ideal data set size is the full Competitive Market Segment (CMS)©, optimizing the tradeoff between bias and variance. An ideal point in the amount of data.

Given the above reasoning, traditional appraisal practice always enables bias – and causes analytic bias:

  • Less-than-optimal data selection causes analytic bias.
  • More-than-optimal data selection causes variation (lower reliability and credibility).

Neither bias nor variation are good results for valuation.  The ideal must be the CMS.

So, does traditional appraisal practice enable bias?  Yes.  It is systemic.  Systemic, analytic, and enabling.

It causes analytic bias.  AND it enables personal bias.  The personal bias may be intentional.  Or the personal bias may be unconscious.

In any case, analytical bias enables personal bias.  Today’s appraisal practice, appraiser education, review practices, laws, regulations, education are out of date, and need to reflect today’s data realities.