“I may be cool, but you can’t change the future” –Beavis & Butthead .
Zillow has recently re-announced it is forecasting the value of each property  out over the next year. It’s not a new tool for them, at least conceptually since the “What is a Zestimate Forecast?” page was last updated on October 3, 2012.
In a world with Big Data, it’s clearly inevitable to see an expansion of the capabilities of services from firms like Zillow and Trulia as their data set grows. Zillow’s Zestimate was a key web site feature at their launch (no listings!), but the company lit the real estate housing market industry on fire , establishing Zillow as a powerful brand that was here to stay, even if the Zestimate tool was problematic.
The challenges facing the Zillow Forecast tool
The Zestimates are still dependent on the quality of public record
Many markets (ie NYC), have quality-challenged public record. But as time passes, Zillow’s data set gets bigger and their logarithms get better and I have not doubt that the reliability will continue to improve.
If the Zestimate is wrong, the forecast will be wrong
Take a look at this chart on the highest price closed sale in Manhattan:
This is perhaps Manhattan’s most famous “trophy” sale of the past several years, 15 Central Park West . The property sold for $88M but the Zestimate at the time of sale indicated the value was $72M. However today the value is $11.9M and the forecast estimated an 8.6% increase next year to $12.9M.
The Zestimate Forecast projects the current Zestimate out over the next year using a bunch of indicators
-mortgage interest rate (local, but not much different than national)
-property tax rate(local)
-number of vacant homes(assumed local)
-percentage of loans that are subprime(assumed local)
-percentage of delinquent loans (assumed local)
-supply of homes for sale (local)
-change in household income (somewhat local, huge lag time)
-population growth (somewhat local, huge lag time)
-unemployment rate (somewhat local, lag time)
I feel that most of these indicators, when considered as a group, are important to consider won’t capture the nuance of next year’s view because they either lag or aren’t granular enough to be a key influence on value trends over a short period. I would think Zillow would add search patterns and other “Internety” things to leverage their proprietary data to help with accuracy. I’d also consider “new inventory”, not just total inventory (supply) to help catch the nuances of a tight time frame of forecasting.
The key national factor driving nearly all housing markets now – credit – is really hard to quantify.
Still, forecasts are the future (sorry) and kudos to Zillow for taking the first step, even though the results, like the early days of the Zestimate, are probably not very accurate.