(This is Part 2 of a two-part series. See Part 1.)
It’s probably safe to assume that the early surge in Zillow’s traffic resulted in part from real estate agents and brokers who fear what Zillow may do to their business. Are “Zestimates” (Zillow’s estimate of what your property is worth) something brokers should fear?
The billion-dollar question for the real estate industry is whether computers can do a better job of accurately pricing property than experienced agents can. Like many other brokers, I decided to put Zillow to the test by evaluating the properties that I have owned in the past.
When it came to selecting comparable sales on my home in Austin, Texas, Zillow did not select a single comparable sale from within the subdivision where I live. While the price was off by about 15 percent, the comparables were so far away that they were useless. Furthermore, every single comparable sale Zillow selected had square footage that was only 50 percent of the size of my home.
After testing Zillow for Texas, I decided to see if there were any more accurate estimates in Southern California. When I priced the properties that I used to own in Beverly Hills, Bel Air and Brentwood, the Zillow algorithms gave them a premium value because the lots were larger than 2 acres. What Zilllow didn’t take into account was that each of these properties was not flat — i.e., the large lot size resulted from the property being on a ridge with a downslope. It also didn’t differentiate between those properties with views and those that lacked views.
These examples strike at the heart of trying to use an algorithm (mathematical formula) to establish value. The computer has no way to tell whether a house has a view, is on a downslope or flat lot, or is in good or poor condition. It can’t tell if a property has airplane noise or strange smells emanating from a landfill. Ultimately, pricing a property is more of an art form rather than a process that can be reduced to pure mathematics.
I was curious to see if any of the other tools did a better job than Zillow. I have been tracking values on my house in Los Angeles and was curious what it would be worth once we expand one bedroom and add another bath. I used Zillow and Moveup.com to determine which approach would give me the most accurate value. The Zillow algorithms allow me to program in a kitchen upgrade, increase the square footage, and add in a new bath. Zillow’s Zestimate told me that without the addition, the property was worth $448,000. By adding a single bath and expanding one of the existing bedrooms, the value jumped to $660,000. A 50 percent price increase seemed way out of line given the nature of the improvements. I then went to Moveup.com. Each of the 15 comparable sales was appropriate to the area. Moveup.com provided the sale date, the square footage, the price per square foot, bedroom-bath count, lot size, and whether the property had a pool. Even with this data plus a deep familiarity with the area, I still couldn’t nail down what the exact price should be.
The challenge was with the comparable sales that were available. Normally, I would do a price-per-square-foot calculation. The rule of thumb is that you should only use properties that are within 10 percent of the same size for both the improvements and the lot size. In California, where the improvements are worth little and the lot is worth a great deal, you can skew the results by using properties that have square footage that doesn’t fall into the appropriate categories. Even with 15 comparable sales, none of them fell into the 10 percent rule that I would normally use. The price per square foot ranged from $243 per foot on the low side to $626 on the high side. The high prices per square foot were for very small homes with the high lot values. (Smaller houses in areas where the land is valuable always sell for a higher price per square foot — larger houses always sell for less.) The low price per square foot was for homes that were twice the square footage of the smaller homes. Since our home would be in the middle of this range, the best comparable sales put the property value at $450 to $518 per square foot. On a 1,300-square-foot property, that’s the difference between $585,000 and $673,000. In truth, the only way to resolve where the property should be valued would be to personally visit the comparable sales or to hire an agent who works the area and knows property values.
The challenge with relying on computers to establish value is the difference between stagnant data that exists in a database and knowledge that relies on human experience and complex thought processes. In the book “Social Life of Information,” John Seely Brown and Paul Duguid make exactly this point. A computer relies on information. When it comes to real estate, this means the property’s features, including bedroom-bath count, lot size and floor plan. Even when two properties have identical floor plans, one may sell for more because of the beautiful landscaping, the privacy, or some other factor the computer cannot access. The value of these features is often more intuitive rather than quantitative. As such, computers may estimate values, but the estimates will continue to be flawed because there is no scientific way to value these other factors.
While Zillow has made a big splash, accuracy is still the name of the game when it comes to comparable sales. In those areas where the market is flattening or declining, overpricing can cost the seller thousands of dollars. Rather than bemoaning the fact that comparable sales are now available on the Web, agents and brokers must be proactive in helping sellers to understand that sites such as Zillow and Moveup.com are a starting place in determining what their properties are worth. They must also be prepared to educate sellers about the challenges of relying strictly on information rather than a human being’s wisdom of experience.
What’s your opinion? Send your Letter to the Editor to firstname.lastname@example.org.