Remote Appraisals of Homes Could Reduce Racial Bias

[Ron Brady]

In today's New York Times: Remote Appraisals of Homes Could Reduce Racial Bias - The New York Times (nytimes.com)

My initial take is that this is a sound move by Fannie and Freddie to address a real issue.  If it helps increase equal opportunity and reduce the negative effects on homeowners of racial bias, great! If it can also speed up the appraisal process, that's an added bonus.

I am always pleased when moves are made in field of real estate to increase equality of opportunity.

[Chris Mason]

Context is for kings. :)

This is the big study that convinced the always skeptical powers-that-be that it wasn't just anecdotes: https://www.freddiemac.com/fma...

Their data set was 12,000,000 appraisals, all for purchase transactions, between 2015 and through the end of 2020. So that's not an anecdote. And these aren't refis, where values are all over the place to begin with (I had a fourplex go from $800k to $1.1m two weeks apart, a couple months ago... without any change in who was hosting it or what coffee table books were out). So that's a big plus for the study, it's not based on refis. On top of 12m data points.

They measured for gaps between contract price and appraised value. Most purchase transaction appraisals hit the contract price exactly. So you can measure % that come in low, and look for patterns. Keep in mind on a purchase transaction, you generally have a seller seeking top dollar, and a buyer seeking the lowest price. The interaction of able/willing buyers and able/willing sellers is the very definition of "market value" via the mechanism of "price discovery." Where they intersect SHOULD be the appraised value, MOST of the time. And when it doesn't, it's reasonable to conclude that one of three things is true.

1) The value is there, but there's not data to support it. Appraisers must primarily used closed comparable sales. If 123 Banana St is the only house with a giant unicorn statue on the roof, it's totally plausible to me that the value could be there, but not in a way that appraisers can see/document. OK, so be it, that's reasonable and fair. That means you as the buyer are going to pay cash for, essentially the unicorn statue component of the house, and finance the rest as normal. If the appraisal comes in $20k short on the unicorn statue house, that's how you'd interpret that: you're financing the house-house parts as normal, and paying cash for the unique unicorn statue. 

2) The buyer or seller is making an error. No one cares when an appraisal comes in high, so let's focus on it coming in low. That means the appraiser is saying "hey, buyer, this house isn't worth that price, you've messed up, so we're going to reign this in." This could be a version of (1) above too, note. 

3) Something else is funky. Someone's bias is creeping in, or something else is going on.

Let's take a look. 

12,000,000 data points later...

Houston, we have a problem. Appraisals come in short in Latino and Black neighborhoods at 2x the frequency of white neighborhoods. 

They go a lot into hypothesis about the methods and the "how" it comes out. A few things with really no results:

- Comp reconciliation. There might be on average $400 or $500 of value opinion difference here. When we're talking about $300k and $700k houses, that's nothing. So disregard.

- Comp variance. The average here is about $5k, but it goes in opposite directions compared to white, for Latino and Black neighborhoods. That's interesting. Put a pin in that one for future research. I wonder if, back in the redlining days (and/or today?), Black neighborhoods were built more cookie-cutter copy/paste, and Latino area home construction was more diverse in terms of features (varying bathroom count, for example, instead of all the houses getting the exact same bathroom count).

- Applicant race. Appraisal gaps are more likely, but not a huge swing. As a matter of course, the appraiser doesn't know the race of the person looking to buy the house. But appraisers do know the demographics of the neighborhood, so I'm guessing this is just indirectly measuring that.

- Are Latino and Black homebuyers pressured to overpay for the house, maybe "realtors are evil" or something? Results trivial, not statistically significant. 0.05%. So like $250 on a $500k house (yup, $250, there isn't a 0 or a k missing). This could be captured by the comp reconciliation thing above, two different measures of the same thing. This could be the impact of 1 in 500 realtors being evil, sure, and when you average it against the other 499, that's where your 0.05% comes from.

- The one big thing that jumped out at me was this:

Context: Appraisers generally want to hit value on purchase transactions. That means they aren't going to get a pair of realtors and a loan officer calling them to yell at them and threaten to withhold business and saying they're the bastard child of a prostitute, and other mean things. No one in the history of the world has ever called an appraiser to say "thanks" for an appraisal that hits value, but all us commissioned people are up in arms when a low appraisal threatens one of our "deals," for context. So that's what they WANT to do, so they can move onto the next appraisal order, the next appraisal fee. OK, but how badly do they want it? How much will they work to produce that outcome? 

Two interpretations here. 

Racism is creeping in. Appraisers will to great lengths, finding comps up to 1.4 miles away, to justify the price, when the house being purchased is in a white neighborhood. But when it's a Latino or Black neighborhood, they get lazy in their search for comps, and on average stop only 0.75 miles out (appraisers have to drive out to those comps for the picture, driving 0.75 miles obviously takes less time than driving 1.4 miles - like they will apparently do in a white neighborhood). So this isn't burning crosses and explicitly racist agenda racism, it's "laziness and indifference" racism. The appraisers work less hard if the subject property is in a Black or Latino neighborhood. Well, if this is the conclusion you reach, then the move to automated and computerized and remote and desktop, etc, might make sense. If the lazy and indifferent appraiser doesn't have to drive to the subject property or to the comps for that picture, then the "comp distance" race gap, a proxy measure for "how hard to they have to work to hit value," may shrink. This is a very imperfect solution, treating the symptom and not the sickness. But if I'm a homebuyer and I just want my piece of the American dram (incidentally, my personal [white guy btw] first home purchase as in a neighborhood of color and, yup, appraisal came in short), and this is my only option, maybe I'll take it.

I can offer a critique of my own conclusion, however. It may be the case that higher income white people buy bigger houses on bigger parcels of land. If your lot is 5000 sq ft, that probably means all the houses around you ALSO have 5000 sq ft lots. Which means you need to drive physically farther to find the best comps. Going 75 houses down might just be 1.5 miles. Imagine when you highlight 20 rows in Excel and make ALL of them just a smidgen bigger, suddenly you have to do a lot more scrolling to get from row 1 to row 20. By contrast, if these Latino and Black dominant neighborhoods were all built with smaller yards back in the redlining days (and/or today?), that may mean that going 75 houses down is only 0.75 miles. 

So, yeah, I'm sure no one's mind was blown by this post. Hopefully it was informative. I'm reading the research as it comes out. Two things are clear to me:

- Yes, there's a gap. And it's a problem. (12 million data points from purchase transactions, I'll remind everyone again)

- No, there aren't going to be any "push button, problem solved" simple solutions. 

FYI: I also read the articles and "research" done by AVM companies themselves. Their methods suck and it's fake research. I'm not saying AVMs are good or bad, but I am saying you can't rely on the AVM companies evaluating their own products, any more than you can rely on Toyota to fairly review Toyota cars.