Data and Summary Statistics

We use loan-level data from Black Knight McDash (heretofore referred to as McDash). These data have been used to study the determinants of mortgage default (Elul et al., 2016), and the expansion of credit during the housing boom (Adelino, Schoar and Severino, 2016). These data are provided by the servicers of the loans, and the contributors include the majority of the top servicers. We focus on first mortgages that are originated or outstanding starting from 2005, since coverage of the McDash data was not as extensive prior to that date (particularly for subprime loans), and continuing through 2008.

The McDash data cover about two thirds percent of all mortgage originations in these years. We restrict attention owner-occupied homes and exclude multifamily properties. The McDash data set is divided into a “static” file, with values that do not change over time,

and a “dynamic” file. The static data set contains information obtained at the time of the original underwriting, such as the loan amount at origination, house value at the origination date, origination FICO score, documentation status (i.e., full-documentation versus low/no documentation of income and assets), the source of the loan (e.g., whether it was broker- originated), property location (zip code), type of loan (fixed-rate, ARM, prime, subprime, IO, Option-ARM, etc.), and whether there is a penalty for prepayment. The dynamic file is updated monthly. The most important dynamic variables for our analysis are the current principal balance and the investor type: private-securitized, Fannie Mae, Freddie Mac, portfolio, FHA. Because of the time it takes a loan to go through the securitization pipeline, many mortgages are initially recorded as portfolio loans when they first appear in the data set; therefore, we define the “investor type at origination” to be that reported at six months from loan origination. We focus attention on loans purchased by Fannie Mae and Freddie Mac (referred to as “GSE”) and Private securitized loans; the other investor types are combined into the a single residual category. Finally, we also merge in the house prices index3_{at the zip code level from Corelogic HPI}TM_{(heretofore referred to as Corelogic), and}

the housing supply measures from Saiz (2010).

Summary statistics for the key variables can be found in tables 1, 2 and 3. Table 1 summarizes the GSE and private-securitized share of the mortgage market and the high-risk (high LTV, low origination FICO) composition of their portfolios at the end of 2006 at the MSA-level. Table 2 summarizes the GSEs and private-securitized share of new originations and their high-risk activity share of originations in the first half of 2007 at the MSA-level. The GSEs increased their share of high-risk mortgage activity substantially in 2007. At the end of 2006, 10% of GSE-securitized loans (purchases and refinances) consisted of mortgage loans with LTV ratios of above 80% and 15% consisted of mortgage loans with origination FICO scores of less than 660. In 2007, 30% of new GSE-securitized loans had an LTV ratio of above 80% and 21% had origination FICO scores of less than 660.

Table 3 shows the GSEs and private-securitizied share of new originations and their high-risk activity share of originations separately for the first two quarters of 2007. The average GSE loan worsened over 2007 with the high-LTV and low- origination FICO share of originations increasing in the second quarter of 2007 relative to the first. A similar deterioration in mortgage quality did not occur on the private-label side with the high-LTV share remaining constant across the two quarters and the low origination FICO share of originations decreasing from 51% in the first quarter of 2007 to 41% in the second quarter.

2.2.1. Background on the Housing Market and GSE Activity in 2007

In this section, we give an overview of the market sentiment about house prices at the beginning of 2007. In particular, we want to establish that market participants were predicting a decline in house prices and that institutions that were active in mortgage markets were reacting to this outlook of the mortgage market and reducing risky activity. We hope to make the case that it is hard to explain GSE high-risk activity at this time and that new high-risk loans did not appear to be profitable ex-ante.

In October of 2006, Moody’s released a report stating that “The U.S. housing market downturn is in full swing.” They predicted that “100 of the nation’s 379 metro areas have a significant probability of experiencing price declines” with about 20 areas experiencing double-digit crashes in house prices. In March of 2007, the NYtimes reported that the Mortgage Bankers Association had shown a record number of homes entering the foreclosure process and stated that the “troubled housing market” was expected to “weaken further”. In March 2007, the Economist further reported that about 18% of homebuyers who took out mortgages in 2006 were in negative equity and would face the “greatest difficulties”.

Investors and participants in the housing market showed signs of reacting to this outlooks. Private securitizers started cutting back on high LTV lending in mid-2006. New Century Financial’s stock fell to half its value in March of 2007. In March 2007, housing starts had fallen 33% in two months and investors were reported to be “shunning subprime and all

mortgages that seemed risky.”

Furthermore, there is evidence that the agencies themselves shared this bleak outlook on the future of the housing market. As we mentioned earlier, in October 2006, Richard Syron, the then chairman and chief executive of Freddie Mac, October was “concerned that foreclosure and loss rates are going to increase” and expressed concern about “a bumpy landing at a national level” for the housing market. He also cautioned against going further out on the “risk curve” and “betting on a turnaround in pricing.” This makes it seem unlikely that the GSEs were planning to “gamble for resurrection.”

Despite this negative forecast of the housing market, the GSEs increased their market share of high LTV loans in 2007. At the same time, the private market for securitization withdrew from these riskier markets. Figure 8 plots the fraction of all high-LTV (LTV > 80) loan originations that were GSE and private-market originated (purchase and refinances). By the end of 2007, the GSEs grew their share of high-LTV originations to about 70%, up from 50% at the start of 2006. This ramp-up in high-risk activity was not simply the GSEs being passive and maintaining lending levels as the private-market cur back on lending. In figures 10 and 11, we plot the volume of new high-risk (LTV > 80) and lower-risk originations (LTV≤80). The GSEs increased their high-LTV originations three-fold while maintaining the level of lower-risk originations. Figure 9 plots the total share of high-risk originations of all GSE originations and this share almost doubled from the beginning of 2006 to the end of 2007. The GSEs increase high-risk activity seems to have been a very active undertaking at a time when they themselves seemed to have a bleak outlook of the housing market.

These high-risk mortgages appeared to perform badly ex-post for the GSEs. About 12% of high-LTV mortgages purchased by the GSEs experienced bad terminations from origination till date. This was an increase of over 50% from 2005 when the bad termination rate was about 7%. Moreover, this number is likely understated as it includes mortgages that were likely rescued from default by government policies such as the Home Affordable Refinance Program (HARP) and the Home Affordable Modification Program (HAMP). In figures 12

and 13, we plot the fraction of GSE and private-label mortgages that were terminated badly. The termination rate on high-LTV mortgages purchased by the GSEs was rising in 2007 as the GSEs expanded risky lending. At the same time, termination rates of private-label MBSes actually started falling in mid-2006.

In 2007, market participants seemed to believe that delinquencies were on the rise and house prices were going to decline. As we would expect, following this guidance, private securitizers retreated from the market. Surprisingly, at the same time GSEs increased their high-risk activity. In the next section, we provide evidence that this ramp-up in GSE high- risk activity could have partially been driven by a want to keep house prices high in areas where the GSEs had large outstanding mortgage exposure.