Major challenges exist within our appraiser education given the dramatic changes in technology.  These challenges include appraisal standards, our basic and advanced education, and some claims by software providers.  A quick overview:

  1. Subjective acceptability in standards;
  2. Inferential assumption in curriculum;
  3. Regression coefficients all over the place.

 

  1. Subjective acceptability. Many consider appraisal to still be mostly “art.”  In the early days of valuation, this was probably true.  In fact, it can be argued that art is a part of science.  The problem lies in our standards which require subjective personal belief as proper justification.  The scope of work guidelines require an appraiser to produce credible work measured as acceptable when:
  • The work meets client’s expectations; and,
  • Other appraisers do it this way.

The key word is defined as “worthy of belief.”  Belief.  Our profession appears to be belief-basedAccuracy does not show up as a goal of the development standards.  Precision (sureness) is also missing from our standards.  In any case, measurable sureness is not required for an “opinion.”  Subjective belief is our mantra.

Today’s technology and complete data sets enable measureable trueness, sureness, and clarity of modeling decisions.  We must head in that direction.

  1. The Inferential Assumption. In the rush to bring “statistics” into the appraisal curriculum, it was somehow assumed by authorities that appraisers should use pretend populations (of competitive sales) then imagine that a judgement sample or convenience sample is close enough to randomness.  Next, sophisticated p-values and confidence things are calculated on the subjective, limited, data to prove that somehow the right model was selected.  This is all very sophisticated and very wrong.  (See the American Statistical Association position paper:  Statistical Significance and P-values).

The appraisal problem involves prediction, not characterizing a population from a sample.  Prediction involves data science, computation, and brain-machine optimization.  Data Science requires subject matter expertise.  It doesn’t require any inferential statistics.

  1. Regression. Regression coefficients are not equivalent to adjustment estimates.  (See the Appraisal of Real Estate, 14th edition, p. 400).  Yet some appraisal software developers continue to contend that their package provides wonderful results at the push of a button.

Beyond the realm of this short blog are two more issues related to regression.  The regression solution also commonly misuses the R2 calculationR2 is rarely an indication of the relevance of the regression model.  An even worse misunderstanding is assuming that the inferential statistical model also applies.  It seldom does.  In fact, for normal valuation, it never does.

In Evidence Based Appraisal© the principles are clean:  1) use the complete competitive market segment; 2) Apply classification and simple regression tools, and 3) create reproducible work files.