Census Tracts, Hedonic Regression and the CU Stew
- AVMs… The Not So Accurate Magic Bullet - November 22, 2022
- Visualizing Market Activity Using Graphs & Trendlines - November 18, 2022
- Diatribes Against Appraisers - October 31, 2022
Bet you didn’t know that the adjustment grid is a form of Hedonic Regression!
The new Collateral Underwriter electronic review process developed by Fannie Mae has many appraisers on edge. This will become the ‘ultimate authority’ or gold standard for reviewing appraisal reports as of January 26, 2015… at least as far as FNMA is concerned. Your reports will either ‘pass’ or ‘fail’, depending on many factors. Some of those factors are outside your immediate control.
“Big Data” is one giant pile of stuff that is being put into the CU pot, stirred together like a stew. Except there is no master chef involved that ‘we’ can interact with. Instead we have a bunch of secret sous chefs each contributing a chunk of meat, a bit of spice, some chopped carrots, and a few potatoes. None of them, or us, really knows the actual CU recipe, because part of what’s in the stew is a ‘model’ of something unknown. But some of that Big Data in the CU stew could be yours … or it might be data provided by your peer appraisers who work in your area – that your reports will be compared against. Not too tasty you say? Just add more pepper.
An aspect of this Big Data stew is Census Tract home price analysis, which is compared against your appraised property value. As an exercise, everyone reading this immediately write the neighborhood description using N, S, E & W directionals for the census tract in which your home residence is located. What? You don’t know the boundaries of your census tract? For shame! Some people using the CU stew might think you are deficient because you don’t know price trends in the exact census tract of the appraised property.
Then we have Hedonic Regression. It’s not a bad thing. But it’s becoming the buzz words of our appraisal adjustment process. It’s a ‘background component’ in the CU process, moving farther forward, faster than some might expect.
Bet you didn’t know that the adjustment grid is a form of Hedonic Regression! It’s a way a certain property’s components of value are itemized separately. By using Hedonic Regression, the individual value of the adjustable components can be calculated and plugged into the adjustment grid. In theory, this can lead to a more accurate property value.
The folks at Bradford Software were among the first to begin promoting use of regression techniques by appraisers. In other ways, the other appraisal software companies and some independent developers have been working on individual processes to make “Regression” more palatable and useful to appraisers. Bradford, and the independent developers, have either report software, or separate spreadsheets, that can help calculate property adjustable components, which in turn can lead to a more credible and supportable opinion of value for the appraised property.
The days of “I’ve been an appraiser for 27 years, so I know what this house is worth” are rapidly coming to an end. The Big Data CU stew is overtaking appraising like the snow avalanches that have closed State Highway 20 in north Washington State in the Cascade Mountain range, not far from where I sit in my cozy bathrobe and bunny slippers.
My observation in this process is that appraisers, as a group, are not statisticians by training and are somewhat scared of that term – even though ‘we’ deal with lots of statistics and data. Thus, appraisers don’t have a clear understanding of what “Regression” is, or does. As a result, ‘we’ have been reluctant to embrace this ‘actually old’ technology in modern appraisal reports. And ‘we’ certainly are skittish about FNMA’s soon to be released (to lenders only) Collateral Underwriter which will analyze reports using “Regression.”
Another perspective on this topic is from this blog: Does Fannie Mae support appraisers?.
This one is written by one of the regression spreadsheet developers, currently available to appraisers.
And for info on Hedonic Regression go to http://en.wikipedia.org/wiki/Hedonic_regression on Wiki.