Thursday, August 18, 2016

Housing: Part 173 - Attrition at Fannie Mae

I posted this graph the other day of the weighted average FICO score on Fannie Mae's books over time.  Their average FICO score has generally been rising since the 1990s.  It has especially risen since conservatorship.  The federal government has taken near complete control of the mortgage market and basically locked out the bottom half of the market from mortgage credit.  This is the single most important economic development of the past 8 years - possibly more important than the NGDP collapse of late 2008.  It has resulted in a loss of wealth targeted at middle class and lower middle class households across the country.  Most losses in home equity at the bottom end of the housing market came after this development.  Conservatorship and QE mark a turning point where the federal government shifted to a policy of economic stability instead of a policy of instability.  While the public discussion centers around bailouts, this is probably the more important story.  Beginning in 2008, the federal government engaged in a massive (and necessary) liquidity injection, but it was targeted at the least credit constrained households.  Through regulatory threats aimed at the banks and government ownership of the GSEs, this liquidity injection was specifically withheld from credit constrained households.

After September 2008, while unemployment raced to 10%, those households were kept out of the recovery.  So while home prices at the top half of the income distribution began to stabilize, home prices at the bottom half continued to collapse as they had been since late 2007.

Here are price changes over time in LA.  LA is probably the most extreme example, but the pattern is pretty common across cities.  High priced homes were the first to decline in value and the first to recover.   Most of the lost value in the top quintile of zip codes within MSAs happened before September 2008.  Most of the lost value in the bottom quintile of zip codes within MSAs happened after September 2008.

The next graph compares the top quintiles in each of the top 20 MSAs, by price, to the bottom quintiles.  (In the Contagion cities, most of the extra rise in the bottom quintile is from Riverside and, to a lesser extent, Miami.  Price changes were relatively similar across Phoenix and Tampa during the boom.)  You can see in this graph that prices in the less credit constrained zip codes of the Closed Access and Contagion cities topped out first by early 2006.  Low priced zip codes held on fairly well until mid 2007 when the private securitization market collapsed.  The private securitization market was highly focused on the Closed Access cities.  Its collapse meant demand for low priced homes in Closed Access cities collapsed, which also meant that the pressure pushing low income households out of the Closed Access cities and into the Contagion cities also collapsed.

So, those cities saw sharp price decreases in 2007 and early 2008, focused on low priced homes.  The other cities had less difference between different quintiles and also had less of a price collapse in general during that time.  The non-metropolitan parts of the country generally looked like the Open Access cities.  The GSEs were capital constrained because of their mounting losses.  That constraint was relieved in September 2008 when the federal government took over.  But, the new very tight lending standards that had been put in place because of capital constraints remained in place.

By early 2009, the Federal Reserve was purchasing mortgage backed securities to introduce liquidity into those markets, and we can see the effect Fed efforts had.  For zip codes that the federal government was supporting, prices generally stabilized.  But, prices at the low end continued to collapse for three more years in all types of cities.  Notice that there wasn't much price decline at all in the Open Access cities until after conservatorship and QE, and it was focused on low priced zip codes, which fell on average by 15% after the end of 2008, compared to the top quintile, which fell by 6% on average in those cities (Dallas, Houston, Atlanta).

In the Contagion cities, prices in the lowest quintiles fell by 38% after 2008 while the top quintiles fell by 15%.  This, despite the fact that by the end of 2008, any excess valuation gains that had gone to low priced zip codes in some cities had been reversed.

Consider the scale of that economic dislocation - homes in the lowest priced zip codes of the Contagion cities took a 38% hit to market values that was unnecessary and unrelated to any sort of bubble pricing.  By then, their price levels weren't relatively any higher than the "Other" category of large MSAs, which includes everything from Seattle to Detroit.  We whacked nearly 40% off the values of the homes owned by the most credit constrained households as a policy choice.

Those houses have valuations that have now moved back above the valuations of the "Other" MSAs, without a functionally expanding mortgage market, if there is any doubt about that.

And, if there is any doubt about the source of pressure on home values in this country, notice how the highest priced zip codes in the country, by a long-shot - the high priced zip codes of the Closed Access cities - are now more than 20% above their peak "bubble" prices.

Sorry, I get going...

The topic of this post was supposed to be the more minor issue of GSE underwriting.  Notice in the first graph that while the average FICO score on new business was higher in 2003-2007 than it had been in 2000-2001, the rise of the average FICO on the book of business was even stronger than we might expect.  In 2004-2007, the average FICO on new business was slightly lower than the average FICO on the book of mortgages.  This should cause the average FICO on the book of business to decline.  But, it continued to rise or remain stable. That means there was attrition of low FICO mortgages out of the GSEs during the housing boom.

This is what Foote, Gerardi, Goette, and Willen and Foote, Loewenstein, and Willen have found.  Data which only tracks originations misses some of the churn within the existing pools of mortgages.  So, to get an idea of how the GSEs were shifting, we need to not only consider their originations, but also the changing shape of their existing book of business.

Using Fannie Mae's reported book of business and new business, by FICO score, I have estimated the proportion of mortgages that were lost from their existing book of business each year.  Some of this attrition would simply represent mortgages that were refinanced as Fannie Mae mortgages. But, we know that, since the average FICO score on the book of business was rising faster than we would expect from the FICO scores on new business, some of the low FICO attrition was moving out of Fannie's books.

So, at the back end, there was a shift of mortgages out of Fannie Mae that were either being refinanced in private securitizations or were households who were selling their homes.  That is what Foote et. al. find in the first link above, that subprime mortgages were not particularly associated with new ownership.  Many of those mortgages were refinanced conventional mortgages.

Attrition was highest among the lower FICO scores during the private securitization boom.*  This has reversed since the bust because it is easier for high FICO score households to refinance.  In recent years there has been some loosening of underwriting, so low FICO score originations are slowly rising again.  This could be partially due to the slow recovery of equity that is finally available to those households simply through inflation.  There could be a significant amount of latent demand in those neighborhoods as equity recovers and those homeowners are able to use leverage to adjust their housing consumption upwards after a decade of deprivation.

* Attrition was very high in 2003 because low interest rates at the time created a rush of refinancing.  The <620 category is very noisy, because it represents a very small portion of Fannie mortgages, and the measure uses Fannie's reported proportions, which are rounded to the nearest percentile.

1 comment:

  1. Amazing post.

    I sometimes wonder if all financial policy-making inevitably migrates to a place that represents is some types of class and industry bias.

    It is like asking farmers to devise agriculture policy and then to expect an even-handed result.