There is no single comprehensive source for securitization rates of various products and yield spreads. We obtain data on a monthly and quarterly basis for the time period 1986 to 2006 from the following sources: the Federal Housing Finance Board (“FHFB”), which includes the Monthly Interest Rate Survey (“MIRS”), the Federal Reserve, Inside Mortgage Finance, Freddie Mac, eMBS, LoanPerformance, and Bloomberg. The specific variables used are discussed below.
The yield spread. One of the key variables in our studies is the yield spread (“YSP”) on the primary market interest rates over an appropriate proxy for a market benchmark interest rate.136 For non-conforming mortgage products, we use the weighted-average coupon rates on
135 We understand from LoanPerformance that based on their data, no subprime loans originated after September 2007 were securitized. Origination activity for subprime mortgages in recent months has been minimal.
136 As a proxy for risk-free rates, we use 10-year constant maturity Treasury rates. Naranjo and Toevs (2002) also used this series as their proxy for risk-free rates. As our data are of a monthly frequency, we use average monthly series from the Federal Reserve.
primary market mortgages from LoanPerformance’s servicing database,137 broken down into the industry categories of subprime and jumbo loans (Tests 1 and 2, see Figure VI.2 below).
For conventional conforming mortgages, we obtain the primary market rates from FHFB (Tests 3 and 4). For the mortgage studies, the yield spreads for these tests are calculated by
subtracting average monthly 10-year constant maturity U.S. Treasury rates from these coupon rates.
For Tests 5 and 6, we define YSP as the spread between the weighted-average coupon rates of jumbo and conforming loans for both adjustable rate loans (Test 5) and fixed-rate loans (Test 6) from FHFB.
For auto loans, we use the average interest rate on new car loans at auto finance
companies as reported by the Federal Reserve and subtract the average 3-year constant maturity Treasury rates on a monthly basis, also from the Federal Reserve. For credit cards, we use the average credit card rates at commercial banks and subtract the average quarterly Bank Prime Loan Rate. Both series are from the Federal Reserve.
The securitization rate. One of the major factors in our models is the securitization rate (“SEC”), which is defined as the amount of primary market consumer loans passed on to the secondary market for securitization divided by the total amount of primary market consumer loans originated in the same period. To allow for easier interpretation of the results, we apply a logarithmic transformation to the SEC variable in all our studies.
For the non-conforming mortgage market (Tests 1 and 2), we use monthly issuance data from LoanPerformance’s securitization database and monthly originations data from
LoanPerformance’s servicing database. LoanPerformance recognizes that their data do not capture the whole U.S. mortgage market and recommend scaling their numbers using estimates of the total market size. For the subprime loans market, we scale up the issuance and
originations numbers using estimates from the Federal Reserve.138 The Federal Reserve has
137 Weighted-average coupon rates are calculated as coupon rates for individual loans weighted by the size of the loan.
138 Similar data on total U.S. market issuance and originations are also available from Inside Mortgage Finance (“IMF”). However, IMF changed its definition of subprime loans to include Alt-A and Home Equity Loans in 2001. As our study period for the subprime mortgage loans extends beyond 2001, we could not use IMF data for this market.
estimates of the total debt outstanding for mortgage-backed securities and primary market mortgages on a quarterly basis, so we linearly interpolate the monthly amount outstanding figures between any two quarters. We then divide the total amount outstanding figures by the LoanPerformance amount outstanding figures to come up with a scaling factor. Finally, we multiply each month’s scaling factor by the amount of issuance or originations in the respective LoanPerformance database each month.
For the jumbo loan market, we scale up the issuance and originations numbers by those published by Inside Mortgage Finance (“IMF”). For the period of our research, IMF provides total issuance and originations only on a quarterly basis. To be able to scale monthly
LoanPerformance data, we assume an equal amount of the issuance/origination ratio in each month in a given quarter and divide it by the LoanPerformance issuance/origination figures to get a scaled monthly series.139
The relationship between YSP and SEC for jumbo mortgages appears to experience a regime change. To examine the impact of this change, we examine a subperiod of the available data from the beginning of 2003 through June 2006. The existence of the regime change is clear from Figure VI.1 below, which shows the relationship between YSP and SEC over the period of January 1999 to June 2006.
139 We are able to obtain reasonable estimates of the jumbo loan market securitization only from January 1999 to June 2006.
Figure VI.1
1999M01 1999M09 2000M05 2001M01 2001M09 2002M05 2003M01 2003M09 2004M05 2005M01 2005M09 2006M05 Year - Month
YSP = Weighted Average Coupon on Jumbo Loans (from LoanPerformance) - 10-Year Constant Maturity Treasury
SEC = Log-Transformed Dollar Volume of Jumbo Securities Issuance divided by Dollar Volume of Jumbo Mortgage Originations (from LoanPerformance, scaled by Inside Mortgage Finance numbers)
Minimum: -0.380
In the studies that involve conforming mortgage loans (Tests 3 - 6), we use Fannie Mae and Freddie Mac mortgage issuance, as reported by Inside Mortgage Finance and eMBS Inc.
The data for the auto loans and credit cards studies (Tests 7 and 8) are obtained from the Federal Reserve. Securitization for the auto loan market is defined as the ratio of the dollar value of securitized motor vehicle loans to the dollar value of owned motor vehicle loans.
Securitization for the credit card market is calculated as the ratio of securitized revolving credit loans to owned revolving credit loan assets.
In our study of the subprime market (Test 1), we also consider two additional variables that may have an effect on yield spreads: prepayment risk (“PRS”) and default risk (“DFL”).
Prepayment risk. Prepayment risk is proxied by subtracting a 1-year constant maturity Treasury rate from a 10-year constant maturity treasury rate.140 This variable is intended to measure the relative attractiveness of refinancing a mortgage at a lower rate.
Default rates. Default risk is modeled using the foreclosure rate for subprime mortgages from LoanPerformance.
The graphs and descriptive statistics of the data used in all of our other studies are presented in Appendix VI.A. Figure VI.2 below summarizes the key variables (SEC and YSP) used in our studies.
Figure VI.2
Test # Market Period
Number of
Observations YSP SEC
1 Subprime
140 Data are from the Federal Reserve.
3 Conventional
C. Statistical Analysis: Theoretical Framework and Estimation