Volume 1
Issue 1
First Quarter 1997

Mortgage Market Trends

National Mortgage Default Rates

and the Vintage Effect

1992 and 1993 Were Very Good Years

Research & Analysis
Office of Thrift Supervision
Washington, DC

Initial Issue of Mortgage Market Trends

This is the first issue in a series of reports that will examine developments in the residential mortgage market. Each issue will include a detailed analysis of one aspect of the market and a data appendix covering current trends in national and regional mortgage default rates and home price appreciation.

The reports will use data from a variety of sources, including the Mortgage Information Corporation’s (MIC) Loan Performance System, OTS’s TFR financial data, and the Office of Finance Housing Enterprise Oversight’s (OFHEO) House Price Index, among others.

The MIC mortgage delinquency data are drawn from the current loan portfolios of more than twenty participants. They include Government Sponsored Enterprises, banks, thrifts, and non-depository portfolio lenders, who, in aggregate, hold more than 21 million loans (with an a total value of $1.7 trillion). The loan data are quite detailed, with information on loan-to-value, size, location, product type, etc. These data are updated monthly.

Our TFR data are much less detailed. For example, we can track mortgage delinquency data only on the geographical basis of the headquarters of our reporting institutions. The MIC data, on the other hand, permit a geographical analysis by the location of the property securing the mortgage.

OFHEO bases its home price appreciation data on same-house resale data supplied primarily by Freddie Mac and Fannie Mae. These data are updated quarterly. The time series goes back to the early 1980s.

We welcome your comments, especially on how this report might be made more useful.

Mortgage Default Rates Steady in Fourth Quarter 1996

Less than 1% of the 21.3 million residential mortgages tracked by the Mortgage Information Corporation (MIC) were seriously delinquent (over 90 days past-due or in foreclosure) at the end of 1996. This represents a small rise in the delinquency rate from the end of the third quarter, but the rate remains below its level at the beginning of the year. As the trend line for the MIC data in Figure 1 shows, the large rise in the delinquencies that occurred during the last two quarters of 1995 and the first quarter of 1996 leveled off during the last three quarters of 1996.

Figure 1 also shows that the thrift industry has a higher (but parallel) rate of seriously delinquent mortgages than the MIC average¹. The higher delinquency rate for the thrift industry relative to the MIC database may be due to the mix of loans tracked by MIC, which is heavily weighted by the lower-risk conforming mortgages held by Fannie Mae and Freddie Mac.

Figure 1 National Mortgage Delinquency Trends
(Source: TFR, MIC)

The delinquency rate for the thrift industry continued its now three-quarter trend downward to its current level of 1.13%. This decline was preceded by a year-long rise during most of 1995 and the first quarter of 1996. The delinquency rate is now at its lowest level in the last four years, having dropped by more than 40% since the beginning of 1993.

Will the current historically low delinquency rate continue? This report explores several aspects of the pool of mortgages now held that suggest that delinquency rates will rise over the next few years, even without a downturn in the economy.

As shown in Figure 1, the MIC delinquency rates track the thrift experience well, except for one quarter (June 1995), where the sharp decline in the MIC delinquency rate was not matched by the thrift industry. Using MIC’s wealth of detail on 21.3 million loans that goes back five years, this report examines three factors that account for some of the movement in the delinquency rates: vintage, loan-to-value (LTV), and product type. We do so to provide information about the future performance of the mortgages held by thrifts. The MIC data suggest that the current low delinquency rates may not persist.

Vintage Effects

Home mortgages, like wine, vary in quality by year of origin. Delinquencies on home mortgages display a well-established distribution or "life-cycle." Newly underwritten mortgages tend to have few credit problems, but delinquencies and foreclosures rise sharply after two years, then decline after the fifth year. Data provided by Moody’s (see Figure 2) show that more than half of all defaults occur in years three through six for 30-year fixed-rate mortgages (FRM).² As mortgages season, though, loan-to-value ratios improve as loans amortize and home prices appreciate. This seasoning effect is typically accompanied by lower mortgage default rates, especially after ten years.

Figure 2: Default Distribution for 30-Year FRMs
Percent of the loan pool defaults occurring in each year

Although delinquency rates in the early years of an origination-year cohort tend to be quite low, a higher-than-normal rise in these delinquency rates can indicate a higher propensity to default over the cohort’s entire life. Such an increase can signal additional credit risk for a portfolio lender, especially during the cohort’s default-prone years. As Figure 3 illustrates, recently originated loans have a much higher delinquency rate than the 1992 and 1993 cohorts at the same point in their seasoning (24 months). This has raised concerns about the future performance of the 1994 and 1995 cohorts.

Figure 3: Seriously Delinquent (%) Loans after 24 Months Seasoning

Mortgage Origination Boom in 1992 and 1993

Loans originated in 1992 and 1993 are now entering their highest default-prone years. These loans are defaulting at historically low rates. Because of the mortgage origination boom of 1992 and 1993, these loans represent a large portion (38%) of the loans tracked by MIC and contribute substantially to the current low national default rate. As these well-performing cohorts leave their default-prone years, though, the average level of defaults should rise over the next few years, even without any deterioration in underlying economic conditions, should the higher propensity to default among the recently originated mortgages persist.

The OTS does not collect the origination year for mortgages held by thrifts. However, 1992 and 1993 were boom years for the thrift industry’s mortgage origination business. Total thrift originations during 1992-1993 ($302 billion) were 38% higher than the total in 1994-1995 ($219 billion). Although thrifts sold many of these mortgages in the secondary market, it is likely that the current thrift mortgage portfolios reflect this origination pattern. Thus, the thrift and MIC origination data should correspond.

At origination, borrowers provide a down payment and meet other underwriting criteria such as debt service load. Over time, the down payment can be eroded through home price depreciation, and personal factors can alter a borrower’s financial status. Higher default rates in the early years of a mortgage cohort can result from changes in down payment and underwriting requirements, changes in the level of debt service load, and from rapid shifts in underlying economic conditions, such as employment or home prices. In the next section, we explore some of these factors, specifically the composition (LTV, product type, and documentation) of the cohorts after 24 months of seasoning, to better understand the causes of this vintage effect.

Changes in the product mix

Loan-to-value (LTV) ratio explains more of the variation in default rates than any other factor. Changes in LTV composition might explain why 1992 and 1993 cohorts are outperforming loans originated in the other years. Table 1 shows the LTV composition of the cohorts after 24 months of seasoning.

Table 1: LTV Portfolio Composition (Percentage)
(Source: MIC)

The data show that the percentage of high LTV loans (over 90%) increased dramatically from about 7% in 1991-1993 to 19% in 1994 and 26% in 1995. This rise in LTV might be due to changes in underwriting standards, as firms might have taken on more risk to gain more return. However, the reported increase might also be due to changes in the composition of the loan portfolios tracked by MIC, as MIC has added participants over time. The more recently added participants have been mostly portfolio lenders.

Table 2: Percentage of Seriously Delinquent Loans
after 24 Months of Seasoning
(Source: MIC)

Table 2 shows the percentages of seriously delinquent loans by LTV for the five cohort years after 24 months of seasoning. Delinquency rates generally rise with LTV. Mortgages in the 96-105% LTV category have delinquency rates at least sixteen times higher than mortgages in the 60% and below LTV category.

The delinquency rates for the 1992 and 1993 cohorts outperform the other years across all LTV ranges. This suggests that the superior performance of the 1992-1993 cohorts (or the relative poor performance of the other years) is due to other factors, such as product type and documentation.

Table 3 shows the product type and documentation composition of the cohorts after 24 months of seasoning.

Table 3: Cohort Portfolio Composition
(Source: MIC)

The most striking differences in the cohort compositions are found in the mix of fifteen and thirty year fixed-rate mortgages, with higher levels of fifteen year fixed-rate mortgages in the 1992 and 1993 cohorts, relative to the other cohorts, as well as the high level of low-documentation loans in the 1991 cohort.

How these changes in composition affect the performance of the portfolio can be inferred on the basis of the risk factors that the national rating agencies use in assessing securitized mortgage loan portfolios. Table 4 shows the various risk factors of a mortgage pool that are then multiplied to determine a composite risk measure.

Table 4: Multiplicative Risk Factors

The 0.65 risk factor for fifteen year mortgages suggests that the shift towards fifteen year fixed-rate mortgages in the 1992 and 1993 cohorts would contribute much to the lower default rates of those cohorts. The higher risk factor for low and no documentation loans also suggests that the higher level of such loans in the 1991 cohort might partially account for that year’s later poor performance.

Declining interest rates provide an incentive to refinance. Because of the rapid and steep decline in mortgage rates in 1992 and 1993, many borrowers refinanced their loans. Many also opted for shorter term loans. Table 5 shows the composition of cohort years by purpose of loan -- purchase vs. refinancing -- which confirms the role of refinancing during the 1992-1993 period.

Table 5: Cohort Composition by Purpose
(Source: MIC)

 

Refinancing at a lower interest rate increases the wealth of the borrower, as the present value of the borrower’s liability falls. A non-cash out refinancing also lowers the borrower’s monthly payment. Risk factors provided by Moody’s and Duff & Phelps indicate that mortgages with interest rates below 8.00% perform better than those with higher rates. Lower rates, with its lower debt service load, as well as the self-selection indicated by refinancing borrowers (borrowers with good payment histories are more likely to refinance), have contributed to the superior performance of the 1992 and 1993 cohorts.

Conclusion

The recent average default experience of more than $1.7 trillion in home mortgages tracked by MIC has been heavily influenced by the 1992 and 1993 cohorts. These cohorts are dominated by mortgages with shorter terms and lower rates due to the two-year refinancing boom. As a result, these two cohorts are performing much better than any prior or subsequent cohort. They are now entering their peak default years and thus have contributed much to the overall decline in default rates. However, recent data suggest we may have seen a low point in average default rates. We are likely to see a gradual increase in the default rates as the more recent and poorer-performing cohorts displace the 1992 and 1993 loans in the prime default years.

Footnotes

¹ The thrift data include all OTS reporters except one, whose primary business involves acquiring seriously delinquent loans. Payment aging is based on the payment cycle, rather than the number of calendar days past due. For example, a loan is not considered thirty-days past due until the second payment has been missed.

² Moody's Investor Service, "Moody's Approach to Rating Residential Mortgage Pass-Through Securities," November 8, 1996.

Data Appendix (Graphic Intensive)

Top