Tuesday, December 16, 2008

The ‘Top 5′ Foreclosure Myth

From the Housing Wire:

Sometimes you just have to say enough is enough: and I feel compelled to weigh in on a false logic that has been gaining ground for the past few months — the idea that says “if you remove the worst 5 states from the nation’s housing data, things don’t look so bad.”

Foreclosure share versus origination share for each U.S. state. Negative values mean foreclosure share is below origination share, and vice versa. (Source: HousingWire)

Mark Perry, an economics professor at the
Univ. of Michigan, is the latest to echo this nonsense, getting linkage on Bill Coppedge’s Mortgage News Clips for suggesting that if we removed AZ, CA, FL, NV and MI from the mix, Nov. 2008 foreclosure activity actually fell 1.10 percent versus year ago numbers, using RealtyTrac data.

Of course, Perry isn’t alone; economists at the Mortgage Bankers Association, the National Association of Realtors, and elsewhere have all made similar claims in discussing their data. Since I’m seeing otherwise brilliant academics make this mistake, and repeatedly so, there are two very important things to keep in mind here. The first is pretty simple to understand:

AZ, CA, FL, NV, MI account for 40.3 percent of all mortgage originations, according to the most recent HDMA data.

Which is another way of saying that any attempt to repackage this crisis by suggesting it is limited to just five states — even were that the truth — conveniently ignores the fact that a disproportionate amount of our nation’s housing stock lies in these five states. Add in New York, Illinois and Texas, and you’re well over 50 percent of the nation’s mortgage market.

The second point here is related to the above, but more technical, and one that I’d have expected an economics professor to remember: the “top 5″ comparison currently being made is a wildly incorrect and arbitrary calculation, statistically speaking. RealtyTrac attempts to normalize its data using population, so as to enable cross-state comparisons of foreclosure activity — in other words 10 foreclosures in a state with 10 people is a very different thing than 10 foreclosures in a state with 1,000 people. And it then ranks states based on its normalized “foreclosure rate per household,” which is what Perry and others use to identify the “top 5″ states.

Lopping off those “top 5″ by foreclosure rate, and then comparing the remaining data by raw foreclosure totals (and not by rate), is the statistical equivalent of using oranges to compare apples. Such a comparison cuts out a whopping 58.5 percent of all foreclosure activity observed nationwide during November — which means that you aren’t just trimming the tail end of some distribution of foreclosure activity to make a comparison, but that you’re forced to ignore the entire distribution itself. (Further, to create a true trimmed mean under such scenario, we’d need to lop off enough states to equal the “lower 58.5 percent” of all foreclosure activity, which isn’t even possible to do.)

Further yet, the “lower 5 states” by normalized foreclosure rate in November represent just 0.1% of all foreclosures. The bottom line here is this: by selecting on one set of criteria and comparing data on another, it’s pretty easy to make wildly inaccurate conclusions.

The real “top 5″
The problem here, as many have surmised, is that a few states dominate the national mortgage landscape. And to do a true comparison of the worst and best states for foreclosures, it follows that you need to have some way to control for where the mortgages are — otherwise there is little chance of getting to true trending in the market.

For this purpose, available 2006 HMDA data makes for an ideal (if wholly underutilized) normalization tool: the data is relatively recent, and roughly captures mortgage share by state since the vast majority of current loans were originated in that timeframe. By comparing foreclosure activity to mortgage activity, it’s much easier to get a read on those particular states where foreclosures are really a problem.

To put this into plain English: If a given state owns a 5 percent share of mortgage originations in the HMDA data, seeing it own a 5 percent share of foreclosures shouldn’t be surprising, all else being equal. So a state like California, which owns a whopping 23.9 percent share in the HMDA data, would be expected to own a similar share of the nation’s foreclosures. By comparing the differences between foreclosure share and origination share, it’s easier to see which states are truly facing distress relative to others.

In doing this simple analysis, however, the results are telling: California — despite its massive raw number of foreclosures — is actually faring far better than other states, relative to mortgages originated. Specifically, California ranks only as the 38th worst state for foreclosures in HousingWire’s analysis, meaning that there are 37 other states in which current foreclosure share is outpacing share of originations: the Golden State reported a 23.38 percent share of foreclosure activity during Nov., according to RealtyTrac, relative to its 23.90 percent share of originations in the HDMA dataset.

Other states are far more problematic, however. Florida, for example, represents 9.2 percent of originations in the HDMA data, but a whopping 19.01 percent of foreclosure activity nationwide; and Michigan, which had a 5.64 percent share of foreclosure activity in Nov., represents just 1.9 percent of originations. These two spreads between foreclosure share and origination share are the two worst in the nation.

Using these spreads to identify a set of “top 5″ states for foreclosures, we arrive at the following: Florida (+9.81), Michigan (+3.74), Nevada (+3.70), Ohio (+3.18), and Arizona (+1.48). Positive spread values mean that a state’s share of foreclosures is proportionately greater than its share of originations; negative spread values mean that a state’s share of foreclosures is less than its share of originations.

The nation’s best states, then, include New York (-3.49), Washington (-1.9), Maryland (-1.86), New Jersey (-1.54) and Massachusetts (-1.52). Surprised?

Relying on RealtyTrac’s data again, and excluding the real top 5 and the bottom 5 states from the data, what we find is that foreclosures are up 21.9 percent from one year ago. Even an improperly trimmed mean — excluding only the top 5, as many academics and economists have been doing as of late — shows that foreclosures are up 17.1 percent year-over-year. The takeaway here is that anyone suggesting that foreclosures are moderating outside of problem states probably needs to take a more refined look at what we define as “problem states.”

I’ve yet to see any major economists take a swing at rationalizing the number of foreclosures against mortgages in a given market; while the 2006 HMDA data is at best a proxy for origination share, it’s not likely that you’d see California shoot up the rankings were a different set of data used for comparison.

We’ve got a housing mess that demands our best analytical attention and best economic thought; so long as those who are supposed to be the best and brightest on this topic refuse to use the tools readily available to them to understand the problem, we’re likely to continue to wonder why the solutions put into place aren’t solving anything.

See the full Excel spreadsheet used for our analysis.

No comments: