Median home prices again stir debate

Too often, the median home sales price is used to judge actual price movement in the housing market. For December, 2008, real estate research firm MDA DataQuick reported a 47% drop from peak value in the actual (unadjusted) median sales price in deeds recorded on all home sales recorded in southern California. In contrast, Zillow.com recently reported a smaller 32% drop in the median home value for Los Angeles and Orange counties, an estimate obtained by artificially adjusting the median sales price to comparable property values.

f.t. observation:

Unfortunately, neither the actual or adjusted median value represents any single property, whether sold or listed to be sold. Brokers seeking the actual value of a specific property would do well to remember that there is no such animal as a “median priced home”; you simply can not find it.

The median price report gives an acutely misleading picture of price changes for any one property or tier of property values. Median price is a statistical point which fails to work in real estate analysis of any one property or tier of properties.

The best way to initially evaluate a property and set its price is to study comparable property values in the same demographic location (same house, same tract). Other ways to set the ceiling price of a property include cost per square foot (replacement cost) and income analysis methods. The application of a “mathematical abstraction,” as in a median price or a comparison of median prices over time, will never produce a useful result when applied to an asset as unique and variable as a parcel of real estate. [For additional insight on how a generally derived median price is an inferior pricing tool, please consider first tuesday’s recent case decision article on REO resales as disqualified appraisal “comps”]

The median value also gives an erroneous representation of the market. The rate and extent of changes in property value prices varies dramatically between low-tier, mid-range, and high-end properties. While values for low-tier properties generally change quickly and dramatically, high-end properties are slower to react to the rise and fall of the general market. Different tiers sometimes even move in opposite directions, and when upward or downward price movement occurs each tier experiences a different percentage of price change than the others.

A median price is nothing more than a line bisecting the total sales prices; one half above, and one half below. It never represents any individual tier of property. The formula of median home price must be put in the trash bin of bad ideas for setting values—just ask any real estate appraiser or economics student.

  • Upon our inquiry, Andrew LePage of MDA Dataquick offered the following response to zillow.com’s press release:

In our monthly housing market reports for the news media we use the median price paid (the median sale price), not a median estimated home value. The latter is simply a statistical construct. It is a guess: sometimes educated, sometimes not.

There is no such guesswork, no secret sauce, involved with generating our median sale price data – the sort of data that are often a key ingredient of the statistical models others use to estimate home values.

Behind our median sale price data are real transactions with a buyer, seller, a final sale price and an address, all documented in public property records. The median sale price has done a good job over time of tracking the ups and downs of home values. Like the statistically based valuation models out there, it is not perfect. But the median is very transparent and easy for people to understand:  the point where half of the homes sold for more and half for less. My colleague John Karevoll, who’s been issuing housing market reports to the news media for two decades, says that in the past he’s offered to use a home price index for media reports, but the median sale price was usually preferred.

John and I have spoken and written extensively about how changes in the median sale price reflect not just changes in home values but changes in the types of homes selling. Through our sales, price, foreclosure and lending data combined we have told the tale of a California housing market where the more affordable inland markets, slammed by foreclosures and price declines, have come to represent a greater share of all sales. We’ve pointed out continuously that this trend has helped tug down the median sale price and, statewide, probably accounts for close to half of the decline in the median price. Meanwhile, sales in many higher-cost coastal neighborhoods remain very sluggish. In these areas, where recent sales are scarce, it is more difficult to get a strong reading on current home values with any of the available tools, whether it’s the median sale price, one of the statistically based “automated valuation models”  or a home price index. In addition, recent sales remain a crucial element of all of these tools aimed at explaining home value trends, and so they all are impacted by the heavy concentration of foreclosure properties in today’s sales. As we pointed out recently, more than half of all homes that resold statewide in December had been foreclosed on in the prior 12 months.

By the way, DataQuick has its own home valuation technology and a home price index. So far we have not employed those tools for our routine monthly housing market reports for the news media.

  • Readers who want to understand the “big picture” of the disparity between low, middle, and high tier sales fluctuations will find the January 2009 Standard & Poor’s/Case-Schiller home price index review a helpful source of information and price comparisons on a national scale. The principles demonstrated also apply in practice to California property prices. Chart 9, and the text which follows, provides a good visual reference of the convergence and divergence of the separate tiers. (http://www2.standardandpoors.com/spf/pdf/index/Case-Shiller_Housing_Whitepaper_YearinReview.pdf)

http://www.latimes.com/business/la-fi-zillow3-2009feb03,0,6862780.story

www.dataquick.com

www.zillow.com