No Cheer For Housing Bulls From Goldman Which Goes Negative On House Prices

Goldman recently confirmed it has lost the magic touch when it joined the momentum brigade in anticipating a blow out 600,000 NFP number, revising its prior estimate by +100,000 on Thursday, even as the real NFP came out as a miserable dud 24 hours later. Which is why we urge readers to take the following note from Goldman's Sven Jari Stehn, even though conceptually we are in full agreement with its message, with a big grain of salt: "Despite normalization of valuations, we expect excess supply, high delinquencies and the fading boost from housing policies to push down house prices somewhat further in 2010 and 2011." And just like earlier we pointed out the discrepancy between the opinions of two BofA strategists on the EURUSD, and the huge implications from this divergence, so here we observe the inconsistency between Sven's bearish view on the oh so critical to the US economy housing segment, and David Kostin's hope for an S&P at 1,250 by the end of the year (and 1,300 by June 30).

From Goldman US Economics Analyst: 10/22 - House Prices Have Not Bottomed Yet June 4, 2010

US Economics Analyst

  • Following their sharp earlier decline, house prices have stabilized since early 2009 and valuations have returned to “normal” levels. But at the same time, temporary boosts from government housing policies are fading and the housing market remains plagued by excess supply and high—and apparently still rising—mortgage delinquencies.
  • To gauge what these opposing forces might imply for future house prices, we construct a model for a panel of 20 metro areas. Our results show that house price dynamics are explained by (1) price momentum, (2) price/rent valuations, (3) the change in mortgage delinquencies, (4) the mortgage rate, (5) excess supply and (6) temporary factors, including the government housing policies.
  • Given the excess supply in the housing market and rising delinquencies, our model suggests that the composite 20-city Case-Shiller index will fall by 3% over the next year and another 1% over the following year.
  • Our model projects the biggest price declines in Las Vegas, Seattle and Portland, due to high homeowner vacancy rates and/or rising mortgage delinquencies. Conversely, we expect modest house price gains in Cleveland, Minneapolis, San Diego and San Francisco.

House Prices Have Not Bottomed Yet

Following their earlier collapse, house prices appear caught in a cross current. On the one hand, there are indications that prices may have bottomed. While alternative house price indices differ in details, they generally show that house prices have stabilized since early 2009 (Exhibit 1). Second, measures of valuation appear to be back in “normal” territory (Exhibit 2). The Case-Shiller price/rent ratio—which stood nearly 25% above its long-run value in early 2006—is now broadly in line with its historical average. Housing affordability—measured as the percent of income spent on mortgage principal and interest—has also improved noticeably during this period.
Other indicators, however, point to further house price declines.  First, much of the stabilization of house prices since early 2009 appears due to government housing policies, including (1) the homebuyer tax credit, (2) the Fed’s purchase of mortgage-backed securities and (3) temporary mortgage modifications through the Obama administration’s Home Affordable Mortgage Program. We have estimated that these housing policies have temporarily boosted house prices by around 5%.  Second, the housing market remains plagued by enormous excess supply (Exhibit 3). Despite recent improvements, both the homeowner vacancy rate and the months’ supply of single-family homes for sale remain well above historical levels. Third, the mortgage market remains troubled. Mortgage delinquencies have continued to rise from their already elevated levels (Exhibit 3).

Given these cross currents, how should we expect house prices to develop over the next one or two years? Our working assumption has been for a renewed 5% drop in the national Case-Shiller index between end-2009 and end-2010, and we already saw a 1.3% decline in the first quarter.  In this comment we present results from a new house price model suggesting that the remaining decline could stretch out over a somewhat longer time period.  Specifically, the model points to declines of 3% over the next year and another 1% over the following year as excess supply and rising mortgage delinquencies take their toll.

Modeling House Prices

Our house price model is constructed as follows. First, we choose to model house prices at the metro area level. Most house price models have focused on forecasting aggregate prices at the national level, as more data with longer time series are available.   However, we believe that the housing market is characterized by sufficient regional variation to warrant a more disaggregated approach. Exhibit 4, for example, shows very different house price developments in the three largest US metropolitan areas.

Second, we decide to focus on the Case-Shiller house price index. Of the three most prominent house price indices, the Federal Housing Finance Agency (FHFA) index is least desirable as it covers only transactions involving agency-backed mortgages and our previous statistical work has shown that the Case-Shiller index is the better index at the regional level, containing more useful information for future house price appreciation.  While the Loan Performance house price index (excluding distressed sales) is desirable in principle, too short a history is available at the metro level to build a panel model.

We therefore construct a quarterly model of the Case-Shiller house price indexes for a panel of the 20 largest US metropolitan areas for the period spanning from 1997Q1 until 2010Q1. Whenever possible we use data at the metro area level; when insufficient data are available we either proxy the metro-area variable with the corresponding state data (for existing home sales and mortgage delinquencies) or use national data when no state-level data are available (for months’ supply of houses for sale and the mortgage rate).

Our model explains current house price appreciation by past price changes and a number of lagged explanatory variables.   Although this approach does not allow for rich quarter-to-quarter dynamics, it permits us to forecast future house prices with current values of the explanatory variables without the need to project data at the metro-area level. We run separate models to project house prices four and eight quarters ahead. To allow for structural differences in house price dynamics across metro areas, we include fixed effects in our panel.

In selecting our specification we aim for a model that describes house prices well both before and after the collapse of the bubble. When estimated for the full sample until 2010, our model does a decent job at capturing the turning point in 2006. However, the model is less successful at predicting the 2006 house price decline when estimated with data through 2005.

Key House Price Determinants

We identify six house price determinants (Exhibit 5):

1. Persistence. Lagged house price appreciation is statistically significant with a sizable coefficient, confirming the existence of short-term momentum in house prices. All else equal, a 1% price decrease over four quarters is typically followed by another ½% fall one year later.

2. Price/rent valuation. We find a strongly negative effect from “overvaluation” on future house prices. All else equal, a 1 percentage point increase in the price/rent ratio lowers house prices by 0.2% after four quarters and by a full percentage point eight quarters later.

3. Excess supply. A one-percentage point increase in the homeowner vacancy rate lowers house prices by 1.8% four quarters later (and 5.4% after eight quarters), while a one-point increase in the months’ supply of homes for sale lowers house prices by 1.4% four quarters later. A higher volume of existing home sales raises prices, as excess supply is reduced.
4. Mortgage delinquencies. Rising delinquencies have a negative effect, lowering house prices by 3.2% after four quarters and 5% after eight quarters for a one percentage point increase in the delinquency rate.

5. Mortgage rates. Higher borrowing costs also have significantly negative effects on house prices, lowering prices by 1.7% after four quarters for every 100 basis points of nominal mortgage rate increases.

6. Temporary factors. To control for the effects of the housing components of the fiscal stimulus bill, we include dummy variables for the period from 2009Q2 to 2010Q1, which suggest that housing policies—including the homebuyer tax credit—have provided substantial support to house prices during this period (details not shown).

Prices Have Not Bottomed Yet

Given the excess supply in the housing market and rising delinquencies, our model suggests that the composite 20-city Case-Shiller index will fall by about 3% over the next year and another 1% over the following year. This projection is weaker than the current consensus forecast of a 0.4% drop in the national Case-Shiller index in 2010 followed by a  1.6% increase in 2011.

We predict the largest house price declines for Las Vegas, Seattle and Portland (Exhibit 6).  While high home vacancy rates and steeply rising delinquencies are expected to push down prices in all three areas, some interesting differences emerge. Price declines in Las Vegas are projected to be front loaded, as negative price momentum and excess supply lead to near-term price declines, before valuation undershoots sufficiently to push up prices. For Seattle and Portland, the model projects back-loaded price declines as house prices currently look overvalued.

The model projects the largest house price appreciation in Cleveland, Minneapolis, San Diego and San Francisco. None of these areas suffers from sharply rising delinquency rates or high vacancy rates (except Cleveland). In addition, house prices in Cleveland appear undervalued and San Diego/San Francisco benefit from positive price momentum.

Our conclusion: Despite normalization of valuations, we expect excess supply, high delinquencies and the fading boost from housing policies to push down house prices somewhat further in 2010 and 2011.