Tuesday, March 31, 2009

Case-Shiller OFHEO faceoff

Two readers generously offered their valuable insights on the Case-Shiller OFHEO gap I noted earlier this morning.

Professor Nunes says that the CS index puts too much weight on particular regions. He has a paper showing how Case-Shiller and OFHEO agree much more region-by-region.

It looks like he has an important and very useful point. (I must also confess that he sent me the paper last fall, and I failed to remember his contribution).

Does the regional aggregation issue explain whether which series is right about January: OFHEO suggests we hit bottom and CS suggests that we're still going down? (Note: Professor Nunes does not like national (as opposed to regional) analysis of housing, so he may be understandably unwilling to answer). Note that OFHEO said that nearly all regions (the West being a notable exception) increased in January.

Sean MacLeod of Metacapital Management emails that "those in the mortgage finance market believe the OFHEO data to be c#$p." He prefers Case-Shiller and a third series, RadarLogic.

I am not sure CS is so much better. In particular, CS says that national housing prices almost doubled (relative to construction costs!) during the boom, which seems reasonable for a couple of regions, but a huge exaggeration for most of America. On the other hand, the OFHEO index was late in showing the bust (see the chart below, which measures both relative to the residential construction PPI). That's why I look at both, and support the Employment for Economists Act!

Many thanks to Mr. MacLeod and Professor Nunes! Housing prices are an incredibly important economic indicator these days, so their insights are worth many times more what I and the readers of this blog are paying for them!


João Marcus said...

I´m willing to take a shot! As you mentioned, the notable exception to a fall in prices in January was the WEST. That is due mostly to prices in California. Since the C-S index is heavily weighed with California prices, that should explain a good fraction of the difference with the OFHEO.

Milton Recht said...

If as CS speculates, house prices doubled relative to construction costs, would not vacant landowners who sold for residential development capture most of this excess gain? If not, then possibly there is an underestimation of the costs to make necessary lot upgrades to make the land capable of supporting a house structure and of the costs to acquire the necessary regulatory approvals. Over time, in regions with limited development potential with an undersupply of homes, either due to regulatory or topographical factors, the marginal cost of buildable land should rise.

Are there any indices of regional or national vacant land prices so regional growth rates in land prices versus home prices can be compared?

buermann said...

Are you still treating that month to month change in OFHEO data as statistically significant? So you are. Curious.

The difference in the OFHEO data and CS must be almost entirely explainable by the selection bias between Freddie/Fannie conforming loan requirements that create a massive selection bias in the OFHEO samples, zeroing out the impact of every non-conforming mortgage origination, which came to dominate the market after 2003.

João Marcus said...

But why does the "selection bias"! show up only (mostly) in the National data?
I fall back on the weighing scheme that CS puts higher weights on higher priced homes and these (for specific reasons) are found in the West (California) that has lots of representatives in CS

stilettoheels said...

The Commission to further the Employment for Economists Act should chart the RadarLogic (RL) and First American Loan Performance (FA) indices as well as Case-Shiller (CS) and FHFA.

RL and FA track the CS through the boom and then the bust. All 3 indices show that prices doubled from 2000 to the 2006 peak are now down about 30%. The FHFA index was the outlier index in the price rise to the peak.

My contribution to the Commission:

Equally weight RL, FA and CS at 30% and for good measure through in FHFA for 10% to ascertain the price bottom.

Karl Smith said...

I always suspected that the difference was primarily accounted for by the conforming limitations on the OFHEO data.

It we think that in some pockets housing has gotten way out of line with affordability these pockets will tend to contain a larger fraction of non-conforming loans.

In other words, OFHEO is biased towards areas in which conforming loans can still account for most of purchases - precisely the areas less effected by the bubble.

This dovetails with my sense that the primary driver of the bubble was the massive increase in non-conforming loans available to the market, not low short term interest rates or a decrease in the standards by Fannie / Freddie.

Karl Smith said...

I see Joao Marcus's objection. Good point.

Do you have any ready links to OFHEO vs Case-Shiller for metros?

João Marcus said...

Karl Smith
The CS data you will find at:

The OFHEO data at:

stilettoheels said...

Firstly, when the Wall Street securitization machine of residential mortgage backed securities went into high gear between 2004-2007, and secondly, when Fannie & Freddie were under strict regulatory scrutiny due to the accounting scandals, Frannie's market share of originations (re-fis and sales) dropped to 35%.

Assuming that originations were equally re-fi and sales, the Frannie data only captured about 20% of national sales for FHFA purposes.

Why would anyone (including Alan local froth Greenspan) rely on such suspect, spurious data is a conundrum. Fortunately, Bernanke officially now uses First American data. Better late than never, I suppose.

Boomer said...

Others here have covered the known differences in the two indices (conforming mortgages in OFHEO, regional weightings, etc.) so I'd like to provide a more personal example.

I live in Miami and have been a homeowner here since 1992. I live in a middle class neighborhood in the south part of the county. I'm about a mile and half from Biscayne Bay as the crow flies, but have no view unless I happen to be on the roof and even then, binoculars help a lot. Most of the houses in my area were built in the 1950s. In other words, its an established neighborhood on the "right" side of US 1 (US 1 is kind of a boundary in the south part of Miami; east of it, where I am, is much more desirable).

From 1992 to 2000 the value of my home rose modestly, say roughly 10-15% for the entire time. From 2000 to the peak (2006) the value quadrupled. The price has now fallen, but from 2000 I'm still sitting on at least a double and more likely a triple.

Neither CS or OFHEO comes anywhere close to approximating my experience. And I guess that is the point; real estate is always local. National indices will never capture what is really going on.

Personally, I think the housing bubble, as a national phenomenon, was vastly over rated. Most of the activity was concentrated in the sand states and that is where most of the foreclosures are concentrated. Having lived through the 80s in CA and Florida now, I've had a front row seat for volatile home prices. It just seems fairly normal to me.