Comparing apples and oranges.

Author:Mayer, Yanling

Comparing two different home-price indexes helps explain why different pictures of the recovery have emerged.

When the housing market hit bottom in early 2012, low-priced real estate-owned (REO) and foreclosed properties were being hotly pursued by homebuyers and investors alike. Interested investors included even some of the largest institutional investment funds, such as the Blackstone Group LP and BlackRock Inc., both based in New York. [paragraph] This period, marked by low interest rates and improving credit conditions, saw the housing market begin to stabilize with rising home prices. And this was happening even as the nation's unemployment rate remained relatively high. [paragraph] Home prices in some cities, such as Phoenix and Las Vegas, climbed so rapidly that pundits began to wonder whether the ongoing recovery would have long-term sustainability. [paragraph] Nationally, home prices have been rising since 2012 either modestly or rapidly, depending on which home-price index (HPI) you watch.

According to the S&P/Case-Shiller 20-City Composite Home Price Index, U.S. home prices were up 18 percent in the 20 months beginning January 2012, with the year-over-year accelerations reaching nearly 14 percent in recent months.

A number of other widely followed home-price indexes, including those from the Federal Housing Finance Agency (FHFA) and Irvine, California-based CoreLogic, have reported changes similar to those captured by Case-Shiller. However, a somewhat different picture of the housing recovery has emerged from the FNC Residential Price Index[TM] (RPI), published by mortgage technology company FNC Inc. The company is a provider of automated appraisal workflow systems for the mortgage industry.

The FNC RPI has shown the recovery appears to be taking place at a more moderate pace--only about half the pace observed by the S&P/Case-Shiller HPI and others. This article seeks to explain why.

Figures 1 and 2 compare the S&P/Case-Shiller HPI and FNC RPI in levels as well as year-over-year changes based on the same 20 metropolitan statistical areas (MSAs) covered by the Case-Shiller composite index.

Although the two indexes generally moved in tandem in the past, and substantial deviations from one another were mostly absent (with the exception of a short-lived period from May 2009 to May 2010), they have in recent months drifted increasingly apart--particularly as the housing recovery deepened.

Until now, more attention has focused on the methodological differences between existing home-price indexes. And for good reason: Despite their important technical differences, these indexes generally tracked closely with one another.

As far as the methodologies are concerned, the Case-Shiller, FHFA and CoreLogic indexes are all based on a repeat-sales method, whereas the FNC RPI uses a hedonic technique that constructs the index by considering the impact of various house attributes on prices. As apples and oranges of residential property value indicators, both approaches each have their own practical advantages--and limitations.

Historically, the reliance on repeat-sales methods--as opposed to a hedonic approach--is in large part due to a lack of detailed property information. Recently, enabled by extensive appraisal data on property characteristics acquired from appraisals ordered by loan originators, FNC developed and began to publish the industry's first hedonic index in 2010.

However, recent trending patterns exhibited by existing home-price indexes--which could lead to rather different policy implications for the housing market as well as the broader economy--are rather unsettling, particularly when one compares the Case-Shiller and FNC RPI.

Perhaps more puzzling is the fact that the two indexes rely primarily on the same data sources--public-record data collected from tax assessors and county recorders. FNC augments that data with data from appraisals, but the majority of the data is the same.

But looks can be deceiving, and the devil is in the details. There is certainly a good possibility that each index's data requirements and filtering process are producing very different final samples of properties...

To continue reading