Our Hedgeye Housing Compendium table (below) aspires to present the state of the housing market in a visually-friendly format that takes about 30 seconds to consume.
Today's Focus: December CoreLogic Home Price Report
The CoreLogic HPI series has essentially become a satirical version of its former self.
We’ve profiled the emergent problems with the HPI series in recent months but to quickly review:
The Pattern: Over the past year the prevailing pattern in the home price series has been one in which the estimate for the latest month reflects a material acceleration in home price growth – an estimate which then gets revised down significantly in the subsequent release. This serial overestimate-subsequent large scale negative revision pattern has characterized every month since early 2015.
This pattern was conspicuous again in the reported December data released this morning. We compare the latest month’s estimates and last month’s November estimate in the chart below.
The CoreLogic data used to provide the most accurate, real-time read on home prices. Released on the 1st week of the month for the period 2-months prior (i.e. data released today was for the month of December), the data led the Case-Shiller HPI series by a full 2-months. Inclusive of the front-month projection (i.e. today’s official release for December also includes a projection for January) which, historically, was highly accurate, the CoreLogic data was almost 3-months more current than other HPI series and only 1-week lagged to the latest month.
The estimation error manifest following the (specious) decision early last year to base the forward projection on a new econometric model rather than the direct tabulation of MLS data that drove the preliminary estimates historically.
Whether the methodological change – which effectively made the front-month projection unusably inaccurate - is also responsible for the serial overestimation in the monthly estimate is unclear but it’s probably not incidental. We have yet to receive a satisfactory response from the company.
The Problem: Given the strong tethering of housing related equities to the slope of home price growth, the revisions are not inconsequential as it completely distorts the read-through. Whereas the original estimates over the last year reflected conspicuous acceleration and a positive read-through for housing related equities, the revision to flat-to-modest HPI carries a less bullish read-through for the complex. In fact, the collective revision to the Apr-July period shows HPI decelerating modestly and completely reverses the HPI-Equity Performance conclusion.
The Penalty Box: We’ve decided to stop using both the front-month projection and the initial estimate for the latest month. In other words, we’re going to disregard both the January projection and the initial December estimate in this morning’s release as a distorted and potentially completely inaccurate reflection of 2nd derivative price trends and view the revised November data as the latest month.
As a reminder, we expect HPI to follow the slope of demand growth on a 9-12-month lag and begin to flatline and decelerate as we move through 2016. Decelerating price growth will serve as a modest-to-moderate headwind for the group and the builders in particular.
For reference, here's the same chart from one month earlier ...
... and two months before that ...
and the month before that ...
CoreLogic HPI incorporates more than 30 years worth of repeat sales transactions, representing more than 55 million observations sourced from CoreLogic's property information database. The CoreLogic HPI provides a multi-tier market evaluation based on price, time between sales, property type, loan type (conforming vs. nonconforming), and distressed sales. The CoreLogic HPI is a repeat-sales index that tracks increases and decreases in sales prices for the same homes over time, which provides a more accurate constant-quality view of pricing trends than basing analysis on all home sales. The CoreLogic HPI covers 6,208 ZIP codes (58 percent of total U.S. population), 572 Core Based Statistical Areas (85 percent of total U.S. population) and 1,027 counties (82 percent of total U.S. population) located in all 50 states and the District of Columbia."
Joshua Steiner, CFA
Christian B. Drake