It’s not me, I’m not biased!
Is it me, the model, or the data? A great problem with legacy appraisal is that bias is about personal bias, rather than analytical bias. USPAP (Uniform Standards of Professional Appraisal Practice) only addresses human bias. Similarly, The Appraisal of Real Estate by the Appraisal Institute only addresses analytical bias in the context of random sample statistics, and the assumption that multiple regression is used. Neither of these are practical for the usual nature of the appraisal problem.
Analytical bias
Three sources of analytical bias are worth noting. Reliable valuation assumes: 1) the right question is addressed; 2) the data is relevant; 3) the right analytic model (algorithm) is applied. Analytical bias is the focus of many of the posts on this blog, and it is an important topic in the TAAR (The Asset Analyst Report), available as a paid subscription. You may also refer to my peer-reviewed articles in The Appraisal Journal.
Intentional bias
Most appraisers who wish to “hit the number” know that the best way is not through large adjustments, but it is by selecting data biased toward the intentionally biased goal “value.” Intentional bias requires a set of skills which are convincing and credible – “worthy of belief.” Personal bias can rely on known analytical biases in model selection, data selection, “adjustments,” and even reconciliation/reporting. Where you find intentional bias, you will often find a client/system similarly motivated to “make the deal.”
Intentional bias can come from motives of money, security, or just plain seeking approval.
Unintentional bias
There are many sources. Unintentional bias can be from ignorance, bad training, or human response bias (psychology – having to do with how the human brain is wired).
Ignorance itself can come from different directions. It can be simply be from a fresh newbie, who has had little or no training. The training itself may be wrong or purposeless. Much of the “statistical” education for appraisers is itself at issue. Some is wrong. Some assumes random sampling inferential statistics (of little or no use for valuation). The lack of proper use of graphs is glaring.
Bad training itself comes from the passing of theoretical “knowledge.” This is basically stuff taught by those who do not actually apply modern analytics software, visual methods, or simple algorithm modeling. This is in addition to mis-intended statistical education, or reverence for traditional ‘judgment-based’ appraisal model. This outdated model blocks the light of science, given today’s technology. Science is the systematic study of the world through observation or experiment. Today’s technology enables evidence-based methods, replacing outdated ‘trust me’ methods.
In my opinion, there is a glaring lack of training in the proper use of judgment, given the computation, visualization, and dashboard analytics possible today.
Human response bias is a large topic of its own. There are two or three dozen forms of human bias, each of which arises unconsciously. Human bias can be extremely useful in a tribal primal setting. It can be quite damaging in a setting where ethical behavior is required. One example of this is called “anchoring.” Anchoring means the subconscious mind simply attaches to the first number brought to awareness. (Like a sale price . . .)
In my opinion, the appraisal profession will not advance until the psychological aspects and analytical aspects of bias are directly addressed. Admonitions to “be careful” will not do.
Neil Cahill
July 1, 2020 @ 7:35 am
here’s my standard “bias” provided by the ultimate end user(s) who will be rating my work and passing on audits and penalties based on how well I meet their bias expectations:
bracket everything- gla, age site size, etc (and put in redundant comments that the reason for something was because there was no alternative based on the market not providing a point with some element ) within acceptable location location location, AND both the sale prices (as a group) and the adjusted values (as a group) must bracket the estimate of value put forth; THAT bias set is like religion to the ultimate users who have cute names… Fannie this and Freddie that and their cousins. So if that’s bias, then they are providing the biasae for the bias-es. Those parameters dictate what comps are used because like a stupid time wasting soma puzzle, by process of analysis & elimination (and that takes lots of time and every analytical thingie you have) one ultimately ends up with a comp set that best fits, every time. its that simple. And I know few are doing it because it takes so much time and effort to arrive at the best fit comp set, and few are willing to go through the process plus they don’t have the skills or tools. after figgerin’ out the besty biasy comp set, you then comment on why the thing isnt perfect, and that’s as good as it gets. We play by the rules we are given, change the rules and you could have a different result, but then, the game would have been changed, with different bias-es.
Neil Cahill
July 1, 2020 @ 7:43 am
ps: if you do this , you wont be able to hit numbers, so you wont be able to satisfy lenders with values equaling or exceeding contract sale prices all the time, when the best fit comp set refuses (like a quadratic eq with “no solution”) to produce a “happy” number. For refi’s or servicing, ur set, because this is nearly bulletproof- and auditable.
Steven R. Smith, MSREA, MAI, SRA
July 1, 2020 @ 10:47 am
George, I started out in appraisal for a lender. My first day on the job, I was told to Bracket the Sales Price. Everything they did was for expediency, like their List of Adjustments, which was not market derived. But, it helped get the reports done fast, 3-per day I might add was the “quota”.
Now I do no lender work per se, well, I did one job this year so far, a $$$$ Spec house.
Most of my work is litigation, and we often/usually run into appraisals on the other side that were biased in favor of the client.
Sadly, even when we have better data as in more similar, more proximate, more recent; judges often split the difference in values.
Still, I market myself as a neutral and tell clients my value would be the same no matter which side hired me. We may not be the busiest, or the highest paid appraisers, but we are still working, one report at a time.
After doing our HABU analysis, we often go look at our data, before inspecting the subject. Assuming our selection criterion was appropriate; once we inspect out subject we have a pretty good felling about where it fits compared to the data set.
Fortunately, most of our work is local, so it is not like we are making two field trips to get this all done.
When I came into this field im my mind, I was coming into the most ethical aspect of the real estate world. Only to find out that an entire industry had been classically conditioned to be accomodating.
george dell
July 3, 2020 @ 2:03 pm
Yes, the next question:
Is the entire appraisal process dedicated to biased results?
In USPAP Standard 1, there are only two tests of an adequate scope of work.
> Do you do what your client expects?
> Do you do what ‘peers’ would do?
These assure a bias toward low-reliability, hi-credibility work!
These two “musts” alone assure that the profession resists modernization to greater value-added services and products.
tom molinari
December 2, 2020 @ 6:13 am
It never ceases to amaze me on how home buyers offer to pay the exact market value of a property based upon the subsequent appraisal by the lender’s appraiser. Simply amazing!