Zillow’s well-known Zestimates provide an ‘estimate of the market value of a property‘, in Zillow’s own words:
The Zestimate home valuation is Zillow’s estimated market value for a home, computed using a proprietary formula. It is a starting point in determining a home’s value and is not an official appraisal. The Zestimate is calculated from public and user-submitted data. Updating your home facts can help make your Zestimate more accurate.
That so-called proprietary formula is in fact fairly well-known standard econometrics, as Zillow combines information on properties (tax records have a wealth of information on the topics), information on neighborhoods, past transactions, and location-specific effects to provide a linear estimate. They also provide a confidence interval around these estimates.
Criticism of Zillow’s zestimates abounds. Part of the criticism is based on a misunderstanding of basic statistics. The zestimate is a forecast of the average price, based on minimizing the mean-squared error between the estimate and the actual transaction price. There will necessarily be prediction error. For instance, this website attacks the accuracy of zestimates based on 3 data points (!).
There are two more serious issues with the zestimates. The first issue is that estimating prices requires long time series — that’s how statistical estimates converge –, but also requires recent data — house prices can be very volatile in the short-run, e.g. Staten Island suddenly became a hot market experienced some very high upward volatility. There is a fundamental trade-off between how recent estimates are and how much precision one can expect from the zestimates.
A second issue with the zestimate is that, apart from a tax perspective, the ‘value’ of a house doesn’t really make sense/is not measurable. The buyer’s reservation price does, the asking price does, and the transaction price do make sense. There is no such thing as the value of a house. Certainly, the maximum price that a buyer is ready to pay for a property depends on:
- his time horizon, i.e. his ability to wait for more offers.
- the amount of friction on the market, i.e. the rate of arrival of offers.
- his specific preferences for amenities; single individuals won’t have the same valuation of a house than couples with kids.
- the characteristics of the mortgage that the buyer could get.
So at the end of the day there is no specific reason why even displaying the average or median transaction price would make sense from a buyer’s perspective, and it’s clear that Zillow is setting itself up for failure by providing an estimate that is independent of the buyer’s specific characteristics. The composition of the pool of buyers can change quite quickly over time.
Note: a few weeks after writing this blog post, I realized that the following paper in the Review of Economics and Statistics is an in-depth analysis of the points I mentioned.
Goetzmann, William, and Liang Peng. “Estimating house price indexes in the presence of seller reservation prices.” Review of Economics and statistics 88.1 (2006): 100-112.