As Josh noted, everyone’s talking about house prices; but they also seem to be talking a great deal about mortgages. The popular consensus appears to be that homebuyers, especially in hot housing markets, now make token down payments and can just scrape into their homes by resorting to interest-only mortgages; in this view, borrowers and lenders alike are vulnerable to any fall in house prices. In my prepared remarks I will address each of these issues.
An obvious measure of the vulnerability of a borrower to a decline in his house price is the ratio of the current mortgage balance to the current value of his home; this ratio is known as the current loan-to-value, or LTV, ratio. The top panel of your first exhibit compares the distribution of the estimated current LTV as of September 2003 (the black bars) with the distribution as of March 2005 (the red bars).
As shown by the leftmost red bar, 64 percent of borrowers currently have LTVs below 70 percent. Moving to the right, a further 18 percent currently have LTVs between 70 and 79 percent; 14 percent have LTVs between 80 and 89 percent; and only 4 percent have LTVs of 90 percent and above. Moreover, the recent rapid appreciation in house prices has actually outstripped mortgage debt growth, so that the average LTV has fallen over the past 18 months.
The middle left panel uses the same data to concentrate on the most leveraged borrowers. For a given drop in nominal house prices (the horizontal axis), the vertical axis gives the percent of borrowers who would have negative equity. As before, the black line reflects data from September 2003 while the red line reflects the most recent data. As shown by the red line, following a house-price drop of 10 percent, an estimated 4 percent of borrowers would have negative equity, while a drop of 20 percent would leave about 18 percent of borrowers in negative territory.
The middle right panel summarizes average LTVs at origination for homes purchased in 2004, by state, relative to price appreciation in that state over the previous four years. States with lower-than-average appreciation, such as Utah, Texas, and Oklahoma, are at the left, while states with faster-than-average appreciation, such as California, Massachusetts, and New York are at the right. As shown, LTV at origination in 2004 was actually lower in states with more appreciation. Thus, homebuyers in California and other high-appreciation states made larger down payments relative to the price of their house than homebuyers in low-appreciation states such as Oklahoma.
Increasing home equity, mainly driven by rising house prices, has supported mortgage credit quality in recent quarters. The bottom left panel plots delinquency rates for loans held on banks’ books (the black line) as well as for the broader MBA [Mortgage Bankers Association] measure (the red line). Both have fallen significantly in recent years.
The bottom right panel summarizes the vulnerability of borrowers to house- price shocks. The average LTV on mortgages has declined over the past 18 months, and most households currently have substantial equity in their homes. In the past few years, borrowers have benefited from rapidly rising house prices, which have kept mortgage delinquencies at extremely low levels. However, an estimated 4 percent of borrowers are highly leveraged and could lose all of their home equity if house prices were to fall 10 percent.
As I noted earlier, recent anecdotal reports have highlighted a second potential risk lurking within mortgage markets—the sudden popularity of interest-only mortgages. These are the subject of your next exhibit.
The statistics presented in exhibit 2 are taken exclusively from data on private- label—that is, non-GSE [government-sponsored enterprise]—residential mortgage- backed securities, or RMBS, pools. The overwhelming advantages of these pools for my purposes are that they are very transparent; that is, a great deal of information is available on the underlying mortgages and that they contain many heterodox mortgages, including interest-only mortgages. Many of the recent press articles citing the growth of novel mortgages are based on RMBS pool data and thus, because such mortgages are overrepresented in RMBS pools, have exaggerated, in my view, the role of heterodox mortgages.
Turning to the data, line 1 of the top left panel shows that the dollar value of RMBS pools has nearly doubled over the past two years. Moreover, as shown in line 2, RMBS pools backed by interest-only, or IO, mortgages have increased almost sixfold, and now amount to nearly $300 billion. However, total home mortgage debt, line 3, has also increased over the past two years. As shown in the memo line, interest-only RMBS pools now account for 3.6 percent of all home mortgage debt, up from less than 1 percent two years ago, but still a small share of all mortgages.
As their name implies, interest-only mortgages do not require the borrower to make principal payments, at least during an initial period. Borrowers willing to use IO mortgages could qualify for a larger mortgage and thus be able to buy a more expensive house. The top right panel gives some idea of the relationship between price appreciation and the popularity of IO mortgages. The vertical axis shows the fraction of IO mortgages used to purchase houses, by state, in 2004. Again, I should emphasize that the IO shares are calculated within the RMBS world, and so are probably overstated. The horizontal axis gives state-level house-price appreciation from 1999 through 2003. As you can see, IO mortgages are somewhat more popular in states that saw more appreciation, although, as seen in the inset box, the correlation is not particularly strong.
While the principal value of an IO loan doesn’t decline, if the initial down payment is large enough, the borrower may have a substantial equity cushion against price shocks. The middle panel reports the LTV at origination for IO loans made over the past three years. As shown, most IO mortgages had LTVs below 80 percent, although the trend over time has been away from the very lowest LTVs. That said, lenders have continued to make relatively few IO loans with LTVs above 80 percent. Those higher LTV loans now account for about 15 percent of all outstanding IOs.
Finally, anecdotes often emphasize that IO loans are extended to borrowers with lower credit quality. The bottom panel shows the distribution of credit quality, measured by FICO scores, among IO borrowers. As a rough approximation, most lenders define prime quality borrowers as those with FICO scores of 660 or above. The leftmost set of bars thus represents subprime borrowers. As shown, between 8 and 10 percent of IO loans have been extended to these borrowers. Moving to the right, the next two sets of bars show that the great majority of IO borrowers had solid credit scores between 660 and 779. As shown by the rightmost set of bars, about 10 percent of IO borrowers had credit scores above 780. On the whole, therefore, the credit quality of borrowers using interest-only loans does not appear particularly risky.
One might wonder if financial institutions and investors have, in the face of the continuing housing boom, dropped their defenses against the mortgage losses that would accompany a house-price bust. The top left panel of your next exhibit lists the main institutions exposed to residential mortgage credit risk and the main types of mortgages held by these institutions. The housing GSEs, line 1, almost exclusively hold or guarantee conforming mortgages with fixed rates. Private mortgage insurers, line 2, insure the component of mortgage principal that exceeds 80 percent of the property’s value and so are effectively exposed to the credit risk associated with high-LTV loans. RMBS pools (line 3), as well as banks and thrifts (line 4), hold a wide variety of different mortgage types, including traditional fixed- rate mortgages as well as variable-rate and junior liens.
The top right panel emphasizes that the housing GSEs hold very little credit risk. As shown on line 1, the average LTV at origination of GSE-guaranteed mortgages was 70 percent; based on regional house-price appreciation, the estimated current LTV of these mortgages (line 2) has fallen to 57 percent. The average credit score of the underlying borrowers (line 3) is also solidly in the “prime” category. Obviously, these average values mask some variation in the borrower population, which no doubt contains some higher-risk borrowers. However, as shown on line 4, 19 percent of the mortgages guaranteed by the GSEs carry some form of credit enhancement. If one of these mortgages defaults, the GSE receives a payment from the insurer, usually a private mortgage insurance, or PMI, company.
The middle left panel examines the health of the PMI industry. As shown by the black line (left axis), the ratio of total insured mortgages to capital, the risk- capital ratio, has declined steadily over the past 10 years, indicating that PMI companies have historically high capital on hand relative to the risks they insure. The red line (right axis) shows net underwriting income—that is, income from premiums less losses and expenses, relative to capital. After suffering large losses in the late 1980s, PMI companies have consistently recorded positive underwriting income. In sum, PMI companies appear to have built up a historically large cushion to absorb the losses that might be associated with a widespread drop in real estate values.
The middle right panel analyzes the risks posed to investors in RMBS pools. These pools contain some of the riskier outstanding mortgages. However, they are structured so that investors can choose their risk exposure. Further, RMBS pools are exceptionally transparent, so investors have extensive information on each mortgage in the pool. In principle, investors should have understood, and appropriately priced, the risks inherent in these mortgages. In practice, however, investors price these mortgages using loss models, which are estimated using relatively little data from major house-price declines.
The bottom two panels discuss the exposure of the 8,900 banks and thrifts in the United States to residential mortgage credit risk. The panel to the left divides institutions into quartiles by the fraction of their portfolios accounted for by residential mortgage assets, defined as whole loans, home equity lines of credit, non- GSE RMBS pools, and residual tranches on securitized mortgages. As shown, for institutions in the bottom quartile, mortgages account for less than 5 percent of total assets. This fraction rises by quartile until, for institutions at the top, mortgages account for more than 40 percent of assets. Residential mortgage credit risk is more concentrated at these institutions than at the institutions in the lower quartiles.
The panel to the right shows the average size and capital-to-asset ratio of institutions in each of the quartiles. Reading down the first column, which gives average institution size, one can see that smaller institutions are concentrated in the first two quartiles, which have relatively little mortgage exposure. Reading down the second column, which gives average tier 1 capital ratios, institutions in all quartiles are extremely well capitalized. Thus, institutions with large amounts of mortgage credit risk on their portfolios are well positioned to handle severe losses.
To sum up, neither borrowers nor lenders appear particularly shaky. Indeed, the evidence points in the opposite direction: borrowers have large equity cushions, interest-only mortgages are not an especially sinister development, and financial institutions are quite healthy. Nonetheless, even the most sanguine analyst quails when contemplating a historically unprecedented drop in nationwide nominal house prices. Such a drop will obviously hurt both borrowers and lenders and will also no doubt expose weaknesses that will only be obvious in hindsight. Thus, perhaps it would be best simply to venture the judgment that the national mortgage system might bend, but will likely not break, in the face of a large drop in house prices. That concludes my prepared remarks.