In Depth Analysis: CalculatedRisk Newsletter on Real Estate (Ad Free) Read it here.

Tuesday, August 27, 2013

Lawler: Measuring Changes in Home Prices is Difficult!

by Calculated Risk on 8/27/2013 04:25:00 PM

From housing economist Tom Lawler:

Below is a table showing the latest YOY % changes in various entities’ home price indexes for the metro areas covered in today’s Case Shiller home price report, as well as “national” HPIs. The YOY changes are for the period ended in June, with Case-Shiller, FHFA expanded, and Zillow effectively being quarterly averages; CoreLogic a weighted 3-month moving average; and LPS and FNC being “June.” Zillow does not include foreclosure resales in constructing its HVI, while LPS’ HPI “reflects” non-distressed sales, as it “takes into account” price discounts for REO and short sales. FNC’s RPI is a “hedonic” price index, as is (pretty much) Zillow’s HVI. CoreLogic produces HPIs for all of the metro areas below, but only releases to the public the ones shown below. Note that in some cases (e.g., NY), Case-Shiller’s “metro area” definition differs from the other entities (it is broader).

The biggest “outliers” are the “hedonic” home price indexes, which attempt to take into account the characteristics of homes sold. FNC’s RPI for many metro areas seem “most strange” – the RPIs for DC and Portland showed virtually no change from a year ago, which just doesn’t seem right. And FNC’s “national” (100 metro area) RPI gain of 3.7% over the last year seems way too low.

Even if all of the metro/regional/state HPIs from these entities “matched up,” the various entities’ “national” HPIs would show different growth rates. Case-Shiller constructs a “national” HPI from Census division HPIs with weights based on each division’s share of the market value of the housing stock from Census 2000. FHFA constructs a “national” HPI using weights based on state shares of the housing stock in units using the latest available ACS data. CoreLogic does not “build up” a national HPI from regional HPIs, but instead aggregates all transactions across the county. Zillow’s HVI is a “median” home-value measure, I believe using all homes for which it has a “Zestimate.” I’m not rightly sure how LPS or FNC construct their “national” HPIs.

The method used to construct a “national” HPI from state/regional HPIs can have a material impact. As I noted last week, if CoreLogic used Census 2000 market value weight applied to its state HPIs, its “national” HPI for June would have been up 9.4% YOY, compared to the reported 11.9%. And if the FHFA had applied Census 2000 market value weights to its state HPIs, its “national” HPI for June would have bee up 8.1%, instead of the reported 7.6%. The Case Shiller “national” HPI for “June” (Q2) was up 10.1% YOY. Case-Shiller doesn’t produce state HPIs and doesn’t release its Census division HPIs to the public.

YOY % Change, Various Home Price Indexes ("June"/Q2 2013)
 Case-ShillerFHFA ExpandedCoreLogicLPSFNCZillow
AZ-Phoenix19.8%21.2%17.1%16.6%27.5%22.0%
CA-Los Angeles19.9%18.3%20.7%18.8%7.8%19.7%
CA-San Diego19.3%17.5%19.9%18.0%9.1%21.8%
CA-San Francisco24.5%19.7% 23.9%10.2%26.3%
CO-Denver9.4%10.3%10.5%9.8%8.8%12.7%
DC-Washington5.7%9.9%9.4%7.7%0.8%7.0%
FL-Miami14.8%13.4%13.4%14.3%5.8%12.5%
FL-Tampa11.1%11.7%9.4%11.0%5.9%10.6%
GA-Atlanta19.0%18.6%16.1%13.0%2.7%6.5%
IL-Chicago7.3%6.7%4.6%6.8%-1.0%1.1%
MA-Boston6.7%7.5% 6.4%3.0%6.7%
MI-Detroit16.4%15.2% 11.3%6.3%15.1%
MN-Minneapolis11.5%10.2%9.1%8.4%4.4%11.7%
NC-Charlotte7.8%8.5% 6.3%5.3%1.9%
NV-Las Vegas24.9%24.5% 27.0%18.6%29.5%
NY-New York3.3%2.6%7.2%4.2%0.2%1.8%
OH-Cleveland3.5%3.1% 3.5%0.9%3.2%
OR-Portland11.8%13.8%14.6%10.3%0.9%13.3%
TX-Dallas8.0%9.0%9.9%6.5%7.4%6.0%
WA-Seattle11.8%14.9%15.1%11.9%2.9%13.0%
“National”10.1%7.6%11.9%8.4%3.7%5.6%