Product Innovation & 
Mortgage Selection in the 
Subprime Era 
Souphala Chomsisengphet 
U.S. Department of the Treasury, The Office of the Comptroller of the Currency

Timothy Murphy 
Marquette University, Economics Department 
 
 &  
 
Anthony Pennington‐Cross* 
Marquette University, Finance Department 
 
October 2008 
 
 
 
 
*Contact author – Marquette University, College of Business Administration, Associate Professor, 
Department of Finance, Straz Hall room 328, PO Box 1881, Milwaukee, WI, 53201‐1881, 
Anthony.Pennington‐Cross@mu.edu, 414‐288‐1452  
 

Electronic copy available at: http://ssrn.com/abstract=1288726

Abstract
The introduction of subprime mortgage lending helped fill an unmet demand for mortgage credit
by those who did not qualify for prime credit. The rapid growth of the subprime market provides
indirect evidence that there were in the past (before subprime) and probably are again (after the
subprime meltdown) many households who cannot purchase a home due to the lack of available
mortgage products. Mortgage product innovation has made mortgage terms of art such as io
(interest only) or neg-am (negative amortization) part of everyday parlance in the real estate
business, but the last few years have also shown that very few institutional investors,
underwriters, originators, or servicers fully understood the risks associated with these types of
products. If these sophisticated market participants struggled with recent innovations it should
come as no surprise that borrowers also struggled to understand them and utilize them
effectively. Indeed, the great promise of subprime lending -- making homeownership feasible
and an affordable reality for most Americans -- has instead turned into a financial debacle and a
massive drag on the American economy.

Since the use of exotic or alternative mortgage products played a role in the subprime meltdown,
it is important to understand when these products make sense and how far the market deviated
from rationality. The research presented here begins this analysis using a county level empirical
examination of the mechanisms used to select different types of subprime mortgages from 2000
through July 2007. In particular we examine the use of adjustable rate loans, hybrid rate loans,
interest only loans, non-amortizing loans, and loans with balloon payments.

The empirical results indicate that the economic and financial incentives play a large part in
determining what types of loans people used to finance their home. We also examine how the
use of these types of mortgages changed over time and how much of those changes were driven
by economic and financial fundamentals. We find substantial evidence of the misallocation of
mortgage types over the 2004-2007 time period. This may help us to project what a sensible
subprime mortgage market should look like in the future and how far the mortgage market
deviated from that path.

Electronic copy available at: http://ssrn.com/abstract=1288726

I. Introduction

In the last 15 years, the subprime mortgage market has increased exponentially. At the

beginning of the 1990’s, few people had even heard of a subprime mortgage. Gramlich (2007)

indicates that in 1993 there were very few subprime mortgage originations but by 2005, $625

billion dollars of subprime mortgages were originated. By 2006 the subprime financial crisis

was just beginning as default rates began to increase rapidly, causing a series of subprime

originators and aggregators for Wall Street to fail. In 2007 and 2008 subprime lending was only

a fraction of its heyday (2004-2006). In an almost surreal and slow procession through the entire

financial system, subprime lending is now credited with destroying venerable financial

institutions such as Lehman Brothers (bankrupt), Merrill Lynch (purchased by Bank of America

or Wachovia), Washington Mutual (bankrupt), Fannie Mae & Freddie Mac (government

receivership) and AIG (bailed out by the federal government).

It is clear that the loans originated more recently performed much worse than loans

originated in the early 2000s. However, the cause of the increased default rates is of

considerable debate. Some stylized facts cannot be denied. For example, considerable fraud

aided the downfall of subprime lending while lenient and likely unethical businesses looked the

other way. Many homebuyers did not understand or did not care to understand the mortgage

products they were using. Very sophisticated financial institutions did not understand the

mortgage products they were underwriting, rating, and purchasing.

Our data indicate that the volume of subprime lending (this includes securitized, Alt A,

near prime, A-, Low Documentation, and deep into subprime market segments) increased by

over more than six times from 2000 through 2005, declined slightly in 2006, and retrenched
considerably in 2007. During the rapid growth of subprime mortgage lending, the introduction

of new or exotic mortgage products made it possible for many households to access the mortgage

credit market for both home purchases and mortgage refinances. Compared to traditional fixed-

rate mortgages these alternative mortgages enabled households to borrow more, amortize the

loans more slowly or not at all, and take on more interest rate risk (LaCour-Little and Yang

2008).

This paper examines the use of adjustable rate loans (loans that allow the interest rate

and payments to change over time), hybrid loans (loans that have a fixed interest rate for a short

time period then become adjustable), interest only loans (loans that allow only the interest to be

paid over a pre-specified time period and then usually become amortizing), loans with balloon

payments (loans that do not fully amortize and therefore have a lump sum payment due at the

end of the loan), and non-amortizing loans (loans that allow no or negative amortization) using a

county level data set representing subprime (or non-prime) lending. A reduced form model is

estimated that relates mortgage selection to borrower and local demographic characteristics, and

economic and financial conditions as suggested in the mortgage selection literature. One

question this approach can answer is whether the allocation of product types in subprime lending

followed well-established incentive structures found in prime lending. In addition, tests of

parameter stability and out of sample forecasts can reveal how much the allocation of mortgage

types changed over the boom and bust cycle.

II. Literature Review

Most of the extensive literature on mortgage selection examines the choice between

adjustable and fixed rate mortgages (for example, Dunn and Spatt, 1985; LeRoy, 1996; and

2 
 
Stanton and Wallace, 1999). In particular, Stanton and Wallace (1999) suggest that transaction

costs influence the choice of how many points to pay on a fixed rate mortgage. Brueckner

(1994) suggests that the greater the expected return on other investments the higher the optimal

loan-to-value ratio on a mortgage.

Campbell and Cocco (2003) explore the role of mortgage choice through the viewpoint of

understanding household risk. They find that an adjustable rate mortgage is more preferable in

an environment when the purchaser is less income risk-averse, has a low default cost, or a high

probability of moving. Szerb (1996) echoes these findings, indicating that larger shocks make it

more likely to use a fixed rate mortgage. Posey & Yavas (2001) also show that a separating

equilibrium exists where the more risky borrowers use adjustable rate loans.

Nichols, Pennington-Cross, and Yezer (2005) explain through a theoretical framework

that the mortgage market has been segmented into discrete risk classifications that limit the

mortgage supply of particular products. Empirical results indicate that credit scores, income,

non-real estate debt, and down payments are very important determinants of FHA (Federal

Housing Administration) and subprime market shares.

LaCour-Little and Yang (2008) and Chambers, Garriga, and Schlagenhauf (2007)

developed mortgage choice models that provide some of the first research focusing on mortgages

more typically used in subprime lending. Lacour-Little and Yang (2008) model which factors

influence the decision to defer costs to the future in exchange for lower initial payments that are

often offered through alternative mortgage products. Empirical evidence indicates that future

house prices, income, and risk preference play an important role in determining if the loan will

be traditional (fixed rate) or an alternative mortgage product. Chambers, Garriga, and

3 
 
Schlagenhauf (2007) indicate that younger and households with fewer assets can benefit from

mortgages with smaller payments.

Piskorski and Tchistyi (2007) attempt to discover an optimal mortgage for both the lender

and borrower. Pareto efficiencies for mortgage choices can be gained through dependence on

expected income realization of the borrower and market interest rates. They make evident that

an ideal alternative mortgage contract between a lender and borrower can be either a

combination of an interest-only mortgage with a home equity line of credit or an option

adjustable rate mortgage.

Due to the recent growth and subsequent decline of the subprime mortgage market,

research on the subprime market segment is relatively limited. We extend the literature on

several fronts. First, instead of limiting the analysis to simply adjustable rate versus fixed rate

mortgages or alternative versus traditional mortgage products, this paper examines multiple types

of mortgage products. These include adjustable rate mortgages, hybrid mortgages, balloon

payment mortgages, interest only mortgages, and negative or no amortization mortgages.

Second, the stability of the results can be tested as the subprime market moved through a boom

and bust to test how far mortgage allocation deviated from normal practices during the boom

period. This may provide insight into one mechanism that contributed to the financial crisis.

III. Data Description

The data used to explore the use of different mortgage types is collected from the Loan

Performance (LP) Asset Backed Securities (ABS) data series covering January 2000 through

July 2007. The data includes loans that are securitized in the private label ABS market. It does

not include loans that were held in portfolio. Therefore, the results may only apply to the

4 
 
securities part of the market. LP reports that their data covers over 95% of the securities market.

As the subprime market has evolved, different market participants have attempted to re-label the

segment of the market they are involved in as non-prime, near prime, A-, or Alt-A. In this

research all these types of subprime loans are included under the label of subprime.

Appendix – County Level Distribution of Mortgage Types in 2007 provides maps at the

county level of the market share for various mortgage types over the time period January through

July 2007. They show that there is substantial geographic variation in the use of different types

of mortgages. For example, adjustable rate mortgages or ARMs and hybrid mortgages tend to be

used more often in Florida, Los Angeles, San Francisco, and Chicago areas. Interest only loans

are most popular along the California coast all the way from the San Francisco bay area down to

the border with Mexico. They are also moderately popular around the east coast cities of

Washington, D.C., New York and Boston. However, loans that have balloon payments are less

geographically concentrated. Non-amortizing loans are used most frequently along the east coast,

and Washington, DC and Florida in particular. In total, these maps indicate that many of the

product innovations have been concentrated in high cost locations that have seen house prices

increase substantially over the last 10 to 15 years. Therefore, affordability may be an important

driver of alternative mortgage product use. These geographic concentrations are also consistent

with the Posey & Yavas (2001) separating equilibrium where the higher risk loans/borrowers use

adjustable rate loans.

Table 1 provides a description of the variables and Table 2 provides summary statistics

for the variables used in the county level analysis. The statistics are reported on an un-weighted

basis. The average county has 45 percent of its loans with adjustable rates and the vast majority

of these adjustable rate loans are hybrids (98 percent). The average county has over 16 percent

5 
 
of loans with non or negative amortization features, 10 percent with balloon features, and 7

percent with interest only features.

The literature indicates that house prices can play an important role in mortgage

selection. Therefore, we include measures of house price appreciation using the Office of Federal

Housing Enterprise Oversight (OFHEO) repeat sales price index. The change in house prices

over the last year and the last five years are included as proxies for anticipated increases in house

prices. It is anticipated that rising prices will be associated with an increase in all types of

alternative mortgages. Since shocks to the housing market can also play a role in mortgage

selection, two measures of price volatility are included. The first measure is the standard

deviation in the house price index growth rate over the prior year and the second measure is the

volatile of individual house prices around the index itself (dispersion). In general, volatility in

the underlying asset’s value should lead to compensation in the underwriting process, such as by

removing interest rate risk, requiring larger down payments, or requiring larger borrower income

relative to debt.

If riskier households and riskier loans self select into loan types with lower initial

payments, then households living in locations with lower credit scores (FICO scores), smaller

down payments or higher loan to value ratios (LTVs), higher debt burdens (debt to income ratios

or D/I), and higher unemployment rates will all use alternative mortgages more frequently.

If affordability is a driving force behind the use of alternative mortgage products then

locations with higher house price to income ratios (HP/I) will be more likely to use alternative

products. Conversely lenders may require households to demonstrate some wealth to

compensate for taking on the additional risks. Wealth is proxied by the county per capita income

6 
 
times the state average household size. Households that are more likely to move should also

prefer lower initial payments available in alternative mortgage types. To proxy for mobility the

state median age is multiplied times the state average household size.

The spread between 10-year and 1-year treasury yields is included to measure the

steepness of the yield curve. Typically, steep yield curves are associated with rising short term

interest rates and a growing economy in the future due to an accommodative monetary policy. A

steep yield curve should increase the use of fixed rate loans to avoid large payment shocks but

improving economic conditions may encourage households to stretch payments in the short run

in the expectation of rising income in the future.

IV. Empirical Model

Our data set is organized as a panel data set where time is measured in years from 2000

through July of 2007 and the cross section is at the county level and includes only counties in

metropolitan areas. Therefore, the estimation must control for the lack of independence of

repeated observations of counties over a short time horizon. A robust Huber/White Sanwich

estimator of variance is used instead of the traditional approach. It adjusts the variance

covariance matrix to allow for correlation of error terms within each county. This procedure

does not affects only standard error estimates, not coefficient estimates.

Five separate models will be estimated for each mortgage selection or market share

analysis: 1. ARM share of all loans, 2. Hybrid share of ARM loans, 3. Non-Amortizing share of

all loans, 4. Balloon share of all loans, and 5, Interest Only share of all loans. Each separate

regression will be estimated using the same explanatory variables discussed above.

7 
 
 Market shares are bounded at zero and one. Therefore, a linear model may predict

market shares that are greater than 1 or less than 0. A logistic transformation of the market share

is used to limit the estimates between zero and one (log(sit / 1-sit), where s is the market share for

county i in time period t). Each market share does not represent the same number of mortgages

and the variance of the error term is expected to be

(1)

where nit is the total number of loans for county i in time period t. To reflect this non-constant

variance we weight each observation by 1 . Therefore, observations that have more

loan originations receive more weight; and the closer a county’s market share is to 50% the

more weight it receives in the regression.

This type of estimation approach is coined grouped logit by STATA because it conducts

a logistic transformation of the dependent variable and the dependent variable represents a group

of binary observations that are aggregated to the county level. To calculate the estimated market

shares the results must be transformed as follows:

(2)

where X represents a vector of explanatory variables and β represents a vector of corresponding

estimated coefficients in county i and year t.

V. Results & Sensitivity Tests

Two factors complicate presenting results in a readily accessible form: 1) the logistic

transformation makes it difficult to interpret coefficients, and 2) because five different dependent

variables are used, specification tests generate more than two dozen sets of results. Tables 3-7

8 
 
provide the estimated coefficients and their level of significance for each of the five market

shares. Each table provides three different specifications that are designed to display coefficient

stability as additional variables are added.

In general the results are consistent with expectations and prior theoretical and empirical

research on mortgage selection. For example, in locations where house prices have been

increasing more people use adjustable rate mortgages. This is true whether measured over the

prior year or the prior five years. Locations with more wealth and lower credit scores are also

more likely to use adjustable rate loans. However, locations with smaller down payments use

more adjustable rate loans. These results are consistent with the theory that higher risk

borrowers self select in a separating equilibrium to adjustable rate loans. There is also evidence

that borrowers are more likely to use adjustable rate loans in locations where housing is more

expensive. A steep yield curve is positively associated with the use of adjustable rate loans,

indicating short run thinking and financing typical of subprime borrowing. However, not all

results meet expectations. For example, our proxy for mobility is negatively associated with

adjustable rate use. In addition, measures of risks in the housing market are mixed. In general

they show that individual house price volatility reduces adjustable rate use while metropolitan

area volatility tends to increase the use of adjustable rate loans.

Tables 4-7 also provide the point estimates for each explanatory variable for the other

four market share results. For the sake of brevity we will not review all these results but instead

examine sensitivity tests to illustrate the results. Five sensitivity tests are presented in the

Appendix: Sensitivity Tests. While holding all other variables constant and evaluating them at

their means, a single explanatory variable is varied +/- three standard deviations from its mean

and graphed against the predicted market share.

9 
 
 Four variables can have large impacts on the share of that use an adjustable rate loan –

income growth, credit scores, LTV, and wealth. The results indicate that more adjustable rate

loans will be used in locations displaying one or more of the following characteristics:

substantial wealth, increasing income, low credit scores, and small down payments. The

sensitivity tests for the use of hybrid loans show that the empirical results have not been able to

estimate economically meaningful results despite statistically significant coefficient point

estimates. Interest only loans are a larger part of the subprime market in locations where

borrowers have high credit scores, low down payments are made, and/or income is declining.

Again, many of the variables have large effects on the use of non-amortizing loans, but two

variables have an especially large effect. Non-amortizing loans are a much bigger part of the

subprime market in locations where borrowers have high credit scores and income is declining.

The primary drivers of balloon use are high debt to income ratios and good credit scores.

In summary, the sensitivity tests show strikingly consistent results. Good credit scores

are used to compensate for other risks associated with an alternative mortgage product. For

balloons the risk is driven by high payments. For non-amortizing loans and interest only loans

the risks are declining income and low down payments. However, for adjustable rate loans

higher wealth and house price appreciation are used to compensate for low credit scores, low

down payments, and declining income.

V. In Sample and Out of Sample Forecasts

Forecasts are generated to test the predictive power of the results and to test for any large

changes in how the market allocated mortgage types over time. To generate the market shares

we use the coefficients (see Appendix – Sample Specification Tests for coefficient estimates and

10 
 
precision of the results) to estimate the market share for each county in each year and then

calculate the average market share for the nation as a whole by weighting each observation by

the number of loans made in that location. The Appendix: In Sample Forecast shows that the

empirical estimates perform very well at a national level tracking the growing market shares for

non-amortizing loans and the rising use of adjustable rate loans through 2004 and the decline in

adjustable rate usage in 2005-2006.

Given the ability of the model to perform well in the sample, next we estimate the model

using only the years 2000, 2001, and 2003 and use the estimated coefficients to predict market

shares for 2004, 2005, 2006, and 2007. This break time period of 2003/2004 was chosen

because in 2004 the volume of subprime loans originated more than doubled and continued to

grow at a very rapid pace through 2005. Therefore, we are roughly trying to estimate a model

over a time period before subprime’s explosive growth and the subsequent increases in the use of

interest only and non-amortizing loans. The Appendix: Out of Sample Forecast 2003-2007

shows that under the earlier lending regime economic and financial conditions cannot explain the

increased use of interest only loans and adjustable rate loans and only a little over one-half of the

increase in use of non-amortizing loans. In contrast, the model predicts that more loans with

balloons would have been used. The balloon result may reflect the increased regulation of

balloons by state level anti-predatory lending laws that were being introduced over this time

period.

The extent of the deviation between the predicted product usage and actual usage also

varies by region of the country. For example, while interest only loans become almost 30

percent of the market for the nation as a whole, in the Pacific region they were almost 45 percent

of the market. By 2007 the use of interest only loans in the Pacific region was approximately

11 
 
four times higher than economic and financial conditions predict. In addition, non-amortizing

loans were being used in 2006 and 2007 approximately 20 percentage points more than

predicted. In contrast, the deviations are not nearly as large in the North East (Mid-Atlantic and

New England regions).

The primary finding is that the allocation of product types shifted dramatically in 2004

and 2005 and these shifts were not driven by economic or financial fundamentals but by other

factors in underwriting or borrower demand. This shift was largest in the Pacific region.

VI. Conclusion

Throughout the last seven years the subprime mortgage market became a $600 billion

industry that contributed trillions of dollars in the financial world through Mortgage Backed

Securities. Gaining an understanding of the mortgage products being purchased has become

increasingly important for policymakers in the midst of a dramatic subprime crisis and ever

increasing international financial crisis.

While there are many factors that have contributed to the subprime debacle, the

increasing use of exotic and new mortgage products helped to set the stage by increasing the

sensitivity of a cohort of loans and borrowers to contemporaneous economic and financial

conditions (Ho and Pennington-Cross, 2005). This paper examines the selection of the types of

loans that have been strongly associated with the rapid deterioration of subprime lending –

adjustable rate loans, hybrid rate loans, interest only loans, non-amortizing loans, and loans with

balloon payments.

The estimation results indicate that all of these loan types tend to respond to financial and

economic conditions as predicted by theory and prior empirical evidence in the prime market.

12 
 
Therefore, there is some evidence of a rational allocation of product types to locations and

households where they make the most sense. However, the results also indicate that model

predictions based on 2000-2002 underwriting standards and borrower preferences grossly

underestimate the rapid increased use of interest only loans and non-amortizing loans. This

deviation is even larger in the Pacific region. It is our interpretation that these deviations are not

caused by changes in the economic or financial conditions that traditionally determine mortgage

product use. Rather, the deviations are likely the result of loosened underwriting standards,

increased steering by originators and perhaps other members of the real estate industry, or a

change in borrower preferences in favor of more exotic loan types.. Whatever the root cause of

the dramatic shift toward exotic loans types, their over-use very likely contributed to the

subprime meltdown and ongoing financial crisis.

13 
 
 References

Brueckner, J. K. (1995). "Borrower Mobility, Adverse Selection, and Mortgage Points." Journal
of Financial Intermediation, vol 3 issue 4, 416-441.

Buist, H., & Yang, T. (2000). "Housing Finance in a Stochastic Economy: Contract Pricing and
Choice." Real Estate Economics, vol 28 issue 1, 117-39

Campbell, J. Y., & Cocco, J. F. (2003). "Household Risk Management and Optimal Mortgage
Choice." The Quarterly Journal of Economics , vol 118 issue 4, 1449-1494.

Chambers, M., Garriga, C., & Schlagenhauf, D. (2007). "Accounting for Changes in the Home-
ownership Rate." Working Paper, Florida State University .

Dodd, R. (2007). "Subprime: Tentacles of a Crisis." Finance & Development, vol 44 issue 4, 15-
19.

Dunn, K., & Spatt, C. (1985). "An Analysis of Mortgage Contracting: Prepayment Penalties and
the Due-on Sales Clause." Journal of Finance , vol 40, 293-308.

Gramlich, E. M. (2007). "Booms and Busts: The Case of Subprime Mortgages." Economic
Review - Federal Reserve Bank of Kansas, 105-113.

Ho, G., & Pennington-Cross, A. (2007). "The Varying Effects of Predatory Lending Laws on
High-Cost Mortgage Applications." Federal Reserve Bank of St. Louis Review, 39-59.

Ho, G., & Pennington-Cross, A. (2006). "The Termination of Subprime Hybrid and Fixed Rate
Mortgages.” Federal Reserve Bank of St. Louis Working Paper, 2006-42.

LaCour-Little, M., & Yang, J. (2008). "Pay Me Now or Pay Me Later: Alternative Mortgage
Products and Housing Consumption." Working Paper, Presented at the American Real Estate
and Urban Economics Mid-Year Conference in 2007, 1-58.

Leroy, S. (1996). "Mortgage Valuation Under Optimal Repayment." Review of Financial

Studies, vol 9 issue 3, 817-844.

Nichols, J., Pennington-Cross, A., & Yezer, A. (2005). "Borrower Self-Selection, Underwriting
Costs, and Subprime Mortgage Credit Supply." The Journal of Real Estate Finance and
Economics , vol 30 issue 2, 197-219.

Piskorski, T., & Tchisty, A. (2006). "Optimal Mortgage Design." Working Paper, NYU .

14 
 
Posey, L., & Yavas, A. (2001). “Adjustable and Fixed Rate Mortgages as a Screening
Mechanism for Default Risk.” Journal of Urban Economics, vol 49 issue 1, 54-79.

Stanton, R., & Wallace, N. (1998). "Mortgage Choice: What's the Point?" Real Estate
Economics, vol 26 issue 2, 173-205.

Szerb, L. (1996). "The Borrower's Choice of Fixed and Adjustable Rate Mortgages in the
Presence of Nominal and Real Shocks." Real Estate Economics ,vol 24 issue 1, 43-55.

15 
 
Table 1: Mortgage Selection Variable Definitions and Sources
Variable  Definition  Source 
Market Share       
Adjustable Rate  Percentage of adjustable rate mortgages  LPABS 
Hybrid  Rate  Percentage of hybrid mortgages  LPABS 
Balloon Payment  Percentage of balloon mortgages  LPABS 
Interest Only  LPABS 
Payment  Percentage of interest only mortgages  
Non‐Amortizing  LPABS 
Payment  Percentage of non‐amortizing mortgages  
Explanatory       
%ΔHPI  Metropolitan area housing price index one year percent change    OFHEO 
5yr‐%ΔHPI  Metropolitan area housing price index five year percent change   OFHEO 
Svol  County one year standard deviation of the housing price index growth  OFHEO 
Vol  OFHEO calculated state parameter measures for volatility  OFHEO 
Unemp  County average yearly unemployment rate  BLS 
5yr‐%ΔI  County metropolitan five year income per capita growth rate  BLS 
HP/I  County average subprime house price divided by income per capita  LPABS/Census 
LTV  County average loan to assessed house value   LPABS 
D/I  County average amount of total debt as a ratio of pre‐taxed income   LPABS 
Wealth  County average income multiplied by state median age   Census 
Pop  County population per square mile   Census 
Mobility  State median age multiplied by state average household size  Census 
FICO  County average FICO score   LPABS 
Yield Curve  Difference of average 10 year treasury bill and 1 year treasury bill  SLFRB 
LPABS: Loan Performance Asset Backed Securities, OFHEO: Office of Federal Housing Enterprise
Oversight, SLFRB: St. Louis Federal Reserve Bank, BLS:.Bureau of Labor Statistics, Census: Census
Bureau.

16 
 
Table 2: Summary Statistics 
Variable  Mean  Standard Deviation
Market Share       
Adjustable Rate  0.45  0.14 
Hybrid  Rate  0.98  0.03 
Balloon Payment  0.10  0.08 
Interest Only Payment  0.07  0.10 
Non‐Amortizing 
Payment  0.16  0.15 
Explanatory       
%ΔHPI  5.88  4.79 
5yr‐%ΔHPI  29.73  18.83 
Svol  1.37  1.08 
Vol  1.30  0.19 
Unemp  4.91  1.52 
5yr‐%ΔI  21.29  6.05 
HP/I  4.34  1.76 
LTV  83.61  3.53 
D/I  36.43  4.48 
Wealth  10,811.78  2,034.16 
Pop  271.60  488.19 
Mobility  90.18  2.33 
FICO  632.81  24.23 
Yield Curve  0.68  0.78 

17 
 
Table 3: Adjustable Rate Results
   Specification 1     Specification 2     Specification 3 
Variable  Coefficient  t‐stat     Coefficient  t‐stat     Coefficient  t‐stat 
%ΔHPI  0.015**  15.60     0.018**  22.29     0.020**  22.94 
5yr‐HPI  0.005**  7.92     0.005**  7.88     0.006**  10.28 
Svol  0.020**  2.63     0.019**  2.81     0.027**  3.99 
Vol  ‐0.480**  ‐6.89     ‐0.439**  ‐5.86     ‐0.222**  ‐3.16 
Unemp  0.005  0.65     0.035**  4.04     0.028**  3.84 
5yr‐I  ‐0.026**  ‐12.48     ‐0.024**  ‐10.86     ‐0.022**  ‐8.65 
HP/I           0.031**  2.78     0.079**  4.73 
LTV           0.028**  4.43     0.045**  6.20 
D/I           ‐0.015*  ‐1.71     ‐0.011  ‐1.27 
Wealth           0.062**  6.87     0.087**  6.35 
Pop                    ‐0.007  ‐0.28 
Mobility                    ‐0.041**  ‐8.83 
FICO                    ‐0.005**  ‐3.61 
Yield                    0.088**  7.02 
Constant   0.826**   8.12      ‐2.104**   ‐3.67     2.395**  2.67 
Adjusted R‐squared     0.55        0.49        0.55 
**5 percent level of significance & *10 percent level of significance

Table 4: Hybrid Rate Results

   Specification 1     Specification 2     Specification 3 
Variable  Coefficient  t‐stat     Coefficient  t‐stat     Coefficient  t‐stat 
%ΔHPI  ‐0.001  ‐0.490     0.003  1.640     0.003**  1.650 
5yr‐HPI  ‐0.004**  ‐3.920     0.001  0.640     ‐0.001  ‐1.280 
Svol  0.036  1.440     0.055**  3.460     0.028**  2.340 
Vol  0.126  0.710     0.090  0.680     0.051  0.330 
Unemp  0.016  1.070     0.029**  2.020     0.031**  2.810 
5yr‐I  0.007  0.880     0.001  0.200     ‐0.002  ‐0.420 
HP/I           ‐0.191**  ‐9.650     ‐0.066**  ‐4.970 
LTV           ‐0.019  ‐1.500     ‐0.012  ‐1.440 
D/I           0.159**  6.060     0.103**  4.490 
Wealth           ‐0.053**  ‐3.920     0.005  0.410 
Pop                    ‐0.062**  ‐2.020 
Mobility                    ‐0.019*  ‐1.740 
FICO                    ‐0.016**  ‐10.890 
Yield                    ‐0.172**  ‐8.090 
Constant   3.139**  9.580        0.107  0.100     13.123**  6.320 
Adjusted R‐
squared     0.04        0.29        0.42 
**5 percent level of significance & *10 percent level of significance

18 
 
Table 5: Interest Only Results
   Specification 1     Specification 2     Specification 3 
Variable  Coefficient  t‐stat     Coefficient  t‐stat     Coefficient  t‐stat 
%ΔHPI  0.010**  2.480     0.013**  3.720     0.029**  13.350 
5yr‐HPI  0.011**  7.080     0.004**  2.420     0.012**  8.950 
Svol  0.095**  5.820     0.060**  4.230     0.036**  2.470 
Vol  ‐0.605**  ‐2.630     ‐0.743**  ‐3.630     ‐0.196  ‐1.320 
Unemp  ‐0.115**  ‐5.370     ‐0.097**  ‐3.310     0.017  1.210 
5yr‐I  ‐0.049**  ‐6.630     ‐0.039**  ‐5.430     ‐0.029**  ‐5.340 
HP/I           0.220**  6.140     0.097**  4.700 
LTV           0.067**  2.680     0.073**  4.340 
D/I           ‐0.011  ‐0.370     ‐0.044**  ‐2.160 
Wealth           0.122**  4.170     0.083**  5.240 
Pop                    0.000  1.130 
Mobility                    ‐0.065**  ‐5.590 
FICO                    0.031**  15.450 
Yield                    ‐0.183**  ‐5.660 
Constant        0.110      0.370      ‐7.499**   ‐2.870     ‐22.074**  ‐9.510 
Adjusted R‐squared     0.36        0.49        0.73 
**5 percent level of significance & *10 percent level of significance

Table 6: No Amortization Results

   Specification 1     Specification 2     Specification 3 
Variable  Coefficient  t‐stat     Coefficient  t‐stat     Coefficient  t‐stat 
%ΔHPI  ‐0.002  ‐0.370     0.002  0.430     0.014**  11.930 
5yr‐HPI  0.017**  14.090     0.009**  6.110     0.014**  15.060 
Svol  0.059**  4.000     0.039**  3.260     ‐0.024**  ‐2.920 
Vol  ‐0.776**  ‐3.660     ‐0.759**  ‐4.220     ‐0.482**  ‐4.690 
Unemp  ‐0.154**  ‐6.120     ‐0.131**  ‐4.170     ‐0.006  ‐0.510 
5yr‐I  ‐0.052**  ‐9.560     ‐0.036**  ‐6.210     ‐0.023**  ‐8.600 
HP/I           0.171**  5.460     0.051**  3.350 
LTV           0.095**  5.180     0.055**  5.140 
D/I           0.130**  5.730     0.042**  3.110 
Wealth           0.092**  3.970     0.036**  3.820 
Pop                    0.000  0.110 
Mobility                    ‐0.047**  ‐6.900 
FICO                    0.028**  20.300 
Yield                    ‐0.473**  ‐17.840 
Constant   1.043**   3.700      ‐14.042**   ‐7.140     ‐21.288**  ‐18.300 
Adjusted R‐squared     0.48        0.60        0.85 
**5 percent level of significance & *10 percent level of significance

19 
 
Table 7: Balloon Results
   Specification 1     Specification 2     Specification 3 
Variable  Coefficient  t‐stat     Coefficient  t‐stat     Coefficient  t‐stat 
%ΔHPI  ‐0.027**  ‐8.260     ‐0.023**  ‐5.460     ‐0.016**  ‐7.060 
5yr‐HPI  0.011**  16.220     0.009**  7.920     0.008**  9.180 
Svol  ‐0.041**  ‐2.060     ‐0.025  ‐1.560     ‐0.082**  ‐7.000 
Vol  ‐0.357**  ‐2.960     ‐0.250**  ‐2.110     ‐0.264**  ‐2.440 
Unemp  ‐0.115**  ‐5.000     ‐0.102**  ‐4.510     ‐0.017  ‐1.570 
5yr‐I  ‐0.012**  ‐3.860     0.003  0.810     0.005**  2.050 
HP/I           ‐0.023  ‐1.200     ‐0.075**  ‐4.540 
LTV           0.062**  6.020     0.026**  3.450 
D/I           0.223**  17.780     0.132**  9.820 
Wealth           ‐0.027**  ‐2.820     ‐0.058**  ‐4.730 
Pop                    0.000**  ‐2.790 
Mobility                    0.006  0.890 
FICO                    0.014**  10.880 
Yield                    ‐0.545**  ‐22.850 
Constant   ‐0.790**   3.760      ‐14.667**   ‐13.71     ‐17.091**  ‐13.050 
Adjusted R‐squared     0.33        0.48        0.67 
**5 percent level of significance & *10 percent level of significance

20 
 
Appendix – County Level Distribution of Mortgage Types in 2007
ARM Percentage Breakdown

ar m_p 0%t o 25% >25%t o 50% >50%t o 75% >75%t o 100%

Hybrid Percentage Breakdown

hybr i d_p 0%t o 25% >25%t o 50% >50%t o 75% >75%t o 100%

21 
 
Interest Only Breakdown

i o_p 0%t o 25% >25%t o 50% >50%t o 75% >75%t o 100%

Non Amortization Breakdown

non_amor t _p 0%t o 25% >25%t o 50% >50%t o 75% >75%t o 100%

22 
 
Balloon Breakdown

bal l oon_p 0%t o 25% >25%t o 50% >50%t o 75% >75%t o 100%

23 
 
 Appendix – Sensitivity Tests

Adjustable Rate Sensitivity Tests

1 1
0.9 0.9
0.8 0.8
ARM Market Share

ARM Market Share
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
‐10 ‐7.5 ‐5 ‐2.5 0 2.5 5 7.5 10 12.5 15 17.5 20 2 3 4 5 6 7 8 9

One Year House Price Appreciation Unemployment Rate

1
1
0.9 0.9
0.8 0.8
ARM Market Share

0.7 ARM Market Share 0.7

0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
1 2 3 4 5 6 7 8 9 ‐20 ‐15 ‐10 ‐5 0 5 10 15 20 25 30 35 40

House Price to Income Ratio Five Year Income Growth

1 1
0.9 0.9
0.8 0.8
ARM Market Share

ARM Market Share

0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
‐20 ‐10 0 10 20 30 40 50 60 70 80 0 0.4 0.8 1.2 1.6 2 2.4 2.8 3.2 3.6 4 4.4

Five Year House Appreciation One Year House Index Standard  Deviation

1 1
0.9 0.9
0.8 0.8
ARM Market Share

ARM Market Share

0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3
0.3
0.2
0.2
0.1
0.1
0
0
100
200
300
400
500
600
700
800
900
0

1000
1100
1200
1300
1400
1500
1600
1700

505 520 535 550 565 580 595 610 625 640 655 670 685 700 715

Population Per Square Mile Average FICO Score

1 1
0.9 0.9
0.8 0.8
ARM Market Share

ARM Market Share

0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
60 62 64 66 68 70 72 74 76 78 80 82 84 86 88 90 92 94 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50

Loan‐to‐value Ratio Average Debt‐to‐Income Ratio

1 1
0.9 0.9
0.8 0.8
ARM Market Share

ARM Market Share

0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
85 86 87 88 89 90 91 92 93 94 95 96 97 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2

Mobility Measurement (Household  Size x Median  Age) Volatility Measurement

1 1
0.9 0.9
0.8 0.8
ARM Market Share
ARM Market Share

0.7 0.7
0.6
0.6
0.5
0.5
0.4
0.4
0.3
0.3 0.2
0.2 0.1
0.1 0
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4

Slope of Yield Curve Wealth Proxy (Median  Age x Income)

24 
 
Hybrid Sensitivity Tests
1 1
0.9 0.9
0.8 0.8
Hybrid Market Share

Hybrid Market Share
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
‐10 ‐7.5 ‐5 ‐2.5 0 2.5 5 7.5 10 12.5 15 17.5 20 2 3 4 5 6 7 8 9

One Year House Price Appreciation Unemployment Rate

1 1
0.9 0.9
0.8 0.8
Hybrid Market Share

Hybrid Market Share
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
1 2 3 4 5 6 7 8 9 ‐20 ‐15 ‐10 ‐5 0 5 10 15 20 25 30 35 40

House Price to Income Ratio Five Year Income Growth

1 1
0.9 0.9
0.8 0.8
Hybrid Market Share

Hybrid Market Share

0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
‐20 ‐10 0 10 20 30 40 50 60 70 80 0 0.4 0.8 1.2 1.6 2 2.4 2.8 3.2 3.6 4 4.4

Five Year House Appreciation One Year House Index Standard  Deviation

1 1
0.9 0.9
0.8 0.8
ARM Market Share

Hybrid Market Share

0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3
0.3
0.2
0.2
0.1
0.1
0
0
100
200
300
400
500
600
700
800
900
0

1000
1100
1200
1300
1400
1500
1600
1700

505 520 535 550 565 580 595 610 625 640 655 670 685 700 715

Population Per Square Mile Average FICO Score

1 1
0.9 0.9
0.8 0.8
Hybrid Market Share

Hybrid Market Share

0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
60 62 64 66 68 70 72 74 76 78 80 82 84 86 88 90 92 94 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50

Loan‐to‐value Ratio Average Debt‐to‐Income Ratio

1 1
0.9 0.9
0.8 0.8
Hybrid Market Share

Hybrid Market Share

0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
85 86 87 88 89 90 91 92 93 94 95 96 97 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2

Mobility Measurement (Household  Size x Median  Age) Volatility Measurement

1 1
0.9 0.9
0.8 0.8
Hybrid Market Share
Hybrid Market Share

0.7 0.7
0.6
0.6
0.5
0.5
0.4
0.4
0.3
0.3 0.2
0.2 0.1
0.1 0
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4

Slope of Yield Curve Wealth Proxy (Median  Age x Income)

25 
 
Interest Only Sensitivity Tests
1 1
0.9 0.9
0.8 0.8
0.7 0.7
IO Market Share

IO Market Share
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
‐10 ‐7.5 ‐5 ‐2.5 0 2.5 5 7.5 10 12.5 15 17.5 20 2 3 4 5 6 7 8 9

One Year House Price Appreciation Unemployment Rate

1 1
0.9 0.9
0.8 0.8
0.7 0.7
IO Market Share

IO Market Share
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
1 2 3 4 5 6 7 8 9 ‐20 ‐15 ‐10 ‐5 0 5 10 15 20 25 30 35 40

House Price to Income Ratio Five Year Income Growth

1 1
0.9 0.9
0.8 0.8
0.7 0.7
IO Market Share

IO Market Share

0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
‐20 ‐10 0 10 20 30 40 50 60 70 80 0 0.4 0.8 1.2 1.6 2 2.4 2.8 3.2 3.6 4 4.4

Five Year House Appreciation One Year House Index Standard  Deviation

1 1
0.9 0.9
0.8 0.8
0.7
IO Market Share

0.7
IO Market Share

0.6 0.6
0.5 0.5
0.4
0.4
0.3
0.3
0.2
0.2
0.1
0.1
0
0
100
200
300
400
500
600
700
800
900
0

1000
1100
1200
1300
1400
1500
1600
1700

505 520 535 550 565 580 595 610 625 640 655 670 685 700 715

Population Per Square Mile Average FICO Score

1 1
0.9 0.9
0.8 0.8
0.7 0.7
IO Market Share

IO Market Share

0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
60 62 64 66 68 70 72 74 76 78 80 82 84 86 88 90 92 94 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50

Loan‐to‐value Ratio Average Debt‐to‐Income Ratio

1 1
0.9 0.9
0.8 0.8
0.7 0.7
IO Market Share

IO Market Share

0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
85 86 87 88 89 90 91 92 93 94 95 96 97 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2

Mobility Measurement (Household  Size x Median  Age) Volatility Measurement

1 1
0.9 0.9
0.8 0.8
IO Market Share

0.7 0.7
IO Market Share

0.6
0.6
0.5
0.5
0.4
0.4
0.3
0.3
0.2
0.2 0.1
0.1 0
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4

Slope of Yield Curve Wealth Proxy (Median  Age x Income)

26 
 
Non Amortization Sensitivity Tests
1 1
0.9 0.9

Non‐Amort Market Share
0.8 0.8
ARM Market Share

0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
‐10 ‐7.5 ‐5 ‐2.5 0 2.5 5 7.5 10 12.5 15 17.5 20 2 3 4 5 6 7 8 9

One Year House Price Appreciation Unemployment Rate

1 1
0.9 0.9
Non‐Amort Market Share

Non‐Amort Market Share
0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
1 2 3 4 5 6 7 8 9 ‐20 ‐15 ‐10 ‐5 0 5 10 15 20 25 30 35 40

House Price to Income Ratio Five Year Income Growth

1 1
0.9 0.9
Non‐Amort Market Share

0.8 0.8
0.7 0.7
Non‐Amort Market Share

0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
‐20 ‐10 0 10 20 30 40 50 60 70 80 0 0.4 0.8 1.2 1.6 2 2.4 2.8 3.2 3.6 4 4.4

Five Year House Appreciation One Year House Index Standard  Deviation

1 1
0.9 0.9
Non‐Amort  Market Share

Non‐Amort Market Share

0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3
0.3
0.2
0.2
0.1
0.1
0
0
1000
1100
1200
1300
1400
1500
1600
1700
100
200
300
400
500
600
700
800
900
0

505 520 535 550 565 580 595 610 625 640 655 670 685 700 715

Population Per Square Mile Average FICO Score

1 1
0.9 0.9
Non‐Amort Market Share

Non‐Amort Market Share

0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
60 62 64 66 68 70 72 74 76 78 80 82 84 86 88 90 92 94 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50

Loan‐to‐value Ratio Average Debt‐to‐Income Ratio

1 1
0.9 0.9
Non‐Amort Market Share

Non‐Amort Market Share

0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
85 86 87 88 89 90 91 92 93 94 95 96 97 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2

Mobility Measurement (Household  Size x Median  Age) Volatility Measurement

1 1
0.9
Non‐Amort Market Share

0.9
Non‐Amort Market Share

0.8 0.8
0.7 0.7
0.6
0.6
0.5
0.5
0.4
0.4 0.3
0.3 0.2
0.2 0.1
0.1 0
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4

Slope of Yield Curve Wealth Proxy (Median  Age x Income)

27 
 
Balloon Sensitivity Tests
1 1
0.9 0.9
0.8 0.8
Balloon Market Share

Balloon Market Share
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
‐10 ‐7.5 ‐5 ‐2.5 0 2.5 5 7.5 10 12.5 15 17.5 20 2 3 4 5 6 7 8 9

One Year House Price Appreciation Unemployment Rate

1 1
0.9 0.9
0.8 0.8
Balloon Market Share

Balloon Market Share
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
1 2 3 4 5 6 7 8 9 ‐20 ‐15 ‐10 ‐5 0 5 10 15 20 25 30 35 40

House Price to Income Ratio Five Year Income Growth

1 1
0.9 0.9
0.8 0.8
Balloon Market Share

0.7 0.7
Balloon Market Share

0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
‐20 ‐10 0 10 20 30 40 50 60 70 80 0 0.4 0.8 1.2 1.6 2 2.4 2.8 3.2 3.6 4 4.4

Five Year House Appreciation One Year House Index Standard  Deviation

1 1
0.9 0.9
0.8 0.8
Balloon Market Share

Balloon Market Share

0.7 0.7
0.6 0.6
0.5 0.5
0.4
0.4
0.3
0.3
0.2
0.2
0.1
0.1
0
0
100
200
300
400
500
600
700
800
900
0

1000
1100
1200
1300
1400
1500
1600
1700

505 520 535 550 565 580 595 610 625 640 655 670 685 700 715

Population Per Square Mile Average FICO Score

1 1
0.9 0.9
0.8 0.8
Balloon Market Share

Balloon Market Share

0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
60 62 64 66 68 70 72 74 76 78 80 82 84 86 88 90 92 94 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50

Loan‐to‐value Ratio Average Debt‐to‐Income Ratio

1 1
0.9 0.9
0.8 0.8
Balloon Market Share

Balloon Market Share

0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
85 86 87 88 89 90 91 92 93 94 95 96 97 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2

Mobility Measurement (Household  Size x Median  Age) Volatility Measurement

1 1
0.9 0.9
0.8 0.8
Balloon Market Share
Balloon Market Share

0.7 0.7
0.6
0.6
0.5
0.5
0.4
0.4
0.3
0.3 0.2
0.2 0.1
0.1 0
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4

Slope of Yield Curve Wealth Proxy (Median  Age x Income)

28 
 
 Appendix – In Sample Forecast
Actual = , Predicted = ------

ARM Market Share Hybrid Market Share
1 1
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
2000 2001 2002 2003 2004 2005 2006 2007 2000 2001 2002 2003 2004 2005 2006 2007

Non‐Amortizing Market Share Interest Only Market Share
1 1
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
2000 2001 2002 2003 2004 2005 2006 2007 2000 2001 2002 2003 2004 2005 2006 2007

Balloon Market Share
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
2000 2001 2002 2003 2004 2005 2006 2007

29 
 
 Appendix – Out of Sample Forecast 2003-2007
Actual = , Predicted = ------

ARM Market Share Hybrid Market Share
1 1
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
2000 2001 2002 2003 2004 2005 2006 2007 2000 2001 2002 2003 2004 2005 2006 2007

Non‐Amortizing Market Share Interest Only Market Share
1 1
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
2000 2001 2002 2003 2004 2005 2006 2007 2000 2001 2002 2003 2004 2005 2006 2007

Balloon Market Share
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
2000 2001 2002 2003 2004 2005 2006 2007

30 
 
 Appendix – Out of Sample Forecast 2003-2007 North East
New England and Middle Atlantic Census Divisions
Actual = , Predicted = ------

ARM Market Share Hybrid Market Share
1 1
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
2000 2001 2002 2003 2004 2005 2006 2007 2000 2001 2002 2003 2004 2005 2006 2007

Non Amortizing Market Share Interest Only Market Share
1 1
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
2000 2001 2002 2003 2004 2005 2006 2007 2000 2001 2002 2003 2004 2005 2006 2007

Balloon Market Share
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
2000 2001 2002 2003 2004 2005 2006 2007

31 
 
 Appendix – Out of Sample Forecast 2003-2007 West Coast
Pacific Census Division
Actual = , Predicted = ------

ARM Market Share Hybrid Market Share
1 1
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
2000 2001 2002 2003 2004 2005 2006 2007 2000 2001 2002 2003 2004 2005 2006 2007

Non Amortizing Market Share Interest Only Market Share
1 1
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
2000 2001 2002 2003 2004 2005 2006 2007 2000 2001 2002 2003 2004 2005 2006 2007

Balloon Market Share
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
2000 2001 2002 2003 2004 2005 2006 2007

32 
 
 Appendix – Sample Specification Tests

Sample Specification Tests: Adjustable Rate Results

   2000‐2002 Sample     2003‐2007 Sample 
Variable  Coefficient  t‐stat     Coefficient  t‐stat 
%ΔHPI  0.013**  3.100     0.019**  22.570 
Unemp  0.026**  2.890     0.030**  3.920 
HP/I  0.101**  6.270     0.096**  5.240 
5yr‐I  ‐0.012**  ‐5.030     ‐0.021**  ‐6.320 
5yr‐HPI  0.007**  4.830     0.006**  8.740 
Svol  ‐0.027**  ‐2.090     0.024**  3.420 
Pop  ‐0.036**  ‐2.880     0.000  ‐0.010 
FICO  ‐0.010**  ‐8.970     ‐0.006**  ‐4.130 
LTV  0.022**  3.650     0.051**  5.240 
D/I  0.043**  5.820     ‐0.054**  ‐4.340 
Mobility  ‐0.029**  ‐5.240     ‐0.046**  ‐7.500 
Vol  0.093  1.130     ‐0.237**  ‐2.750 
Yield  0.173**  8.130     0.101**  6.440 
Wealth  0.076**  5.100     0.088**  6.360 
Constant  3.623**  4.780     4.678**  3.760 
Adjusted R‐squared     0.50        0.47 
**5 percent level of significance & *10 percent level of significance
 

Sample Specification Tests: Hybrid Rate Results

   2000‐2002 Sample     2003‐2007 Sample 
Variable  Coefficient  t‐stat     Coefficient  t‐stat 
%ΔHPI  ‐0.019*  ‐1.820     ‐0.002  ‐1.240 
Unemp  0.050**  2.570     0.026**  2.640 
HP/I  ‐0.021  ‐0.610     ‐0.063**  ‐4.070 
5yr‐I  0.011  1.610     0.017**  4.150 
5yr‐HPI  ‐0.020**  ‐6.690     ‐0.006**  ‐6.190 
Svol  0.033  0.570     0.014  1.410 
Pop  0.004  0.110     ‐0.053**  ‐2.090 
FICO  ‐0.016**  ‐5.440     ‐0.018**  ‐11.530 
LTV  ‐0.007  ‐0.480     ‐0.037**  ‐3.390 
D/I  0.094**  6.850     0.127**  6.050 
Mobility  ‐0.004  ‐0.350     0.001  0.130 
Vol  0.470**  2.750     ‐0.068  ‐0.370 
Yield  ‐0.102*  ‐1.820     ‐0.259**  ‐7.670 
Wealth  0.053**  2.480     0.023**  ‐2.140 
Constant  10.088**  6.510     14.400**  9.320 
Adjusted R‐squared     0.48        0.54 
**5 percent level of significance & *10 percent level of significance
 

33 
 
Sample Specification Tests: Non Amortizing Results
   2000‐2002 Sample     2003‐2007 Sample 
Variable  Coefficient  t‐stat     Coefficient  t‐stat 
%ΔHPI  ‐0.004  ‐0.560     0.016**  12.760 
Unemp  0.043  1.810     0.008  0.700 
HP/I  0.000  0.010     0.047**  2.700 
5yr‐I  ‐0.016**  ‐2.640     ‐0.026**  ‐6.690 
5yr‐HPI  0.013**  5.290     0.014**  12.470 
Svol  ‐0.079**  ‐2.740     ‐0.029**  ‐3.990 
Pop  0.003  0.080     0.013  0.430 
FICO  0.014**  6.230     0.033**  23.060 
LTV  0.024**  2.170     0.074**  5.510 
D/I  ‐0.012  ‐0.970     0.073**  4.010 
Mobility  0.003  0.360     ‐0.050**  ‐5.530 
Vol  ‐1.020**  ‐4.010     ‐0.328**  ‐2.540 
Yield  ‐0.311**  ‐6.120     ‐0.382**  ‐13.760 
Wealth  0.016  0.800     0.040**  3.890 
Constant  ‐11.602**  ‐7.910     ‐26.980**  ‐18.140 
Adjusted R‐squared     0.34        0.88 
**5 percent level of significance & *10 percent level of significance

Sample Specification Tests: Interest Only Results

   2000‐2002 Sample     2003‐2007 Sample 
Variable  Coefficient  t‐stat     Coefficient  t‐stat 
%ΔHPI  0.019  1.350     0.028**  16.300 
Unemp  ‐0.038  ‐1.370     0.031**  2.600 
HP/I  0.224**  5.500     0.111**  5.330 
5yr‐I  ‐0.035**  ‐4.140     ‐0.022**  ‐4.130 
5yr‐HPI  0.020**  5.030     0.011**  8.360 
Svol  ‐0.061  ‐1.480     0.032**  2.620 
Pop  0.102**  2.360     0.064  1.490 
FICO  0.024**  7.580     0.030**  15.970 
LTV  0.061**  2.630     0.072**  4.480 
D/I  ‐0.112**  ‐2.610     ‐0.045  ‐1.990 
Mobility  ‐0.069**  ‐4.910     ‐0.071**  ‐6.030 
Vol  0.231  0.810     ‐0.049  ‐0.290 
Yield  0.677**  7.850     ‐0.156**  ‐4.860 
Wealth  0.081**  2.490     0.080**  4.990 
Constant  ‐17.014**  ‐6.820     ‐21.037**  ‐9.550 
Adjusted R‐squared     0.73        0.75 
**5 percent level of significance & *10 percent level of significance
 

34 
 
Sample Specification Tests: Balloon Results
   2000‐2002 Sample     2003‐2007 Sample 
Variable  Coefficient  t‐stat     Coefficient  t‐stat 
%ΔHPI  ‐0.004  ‐0.690     ‐0.011**  ‐5.040 
Unemp  0.050**  2.050     ‐0.021  ‐1.640 
HP/I  ‐0.023  ‐0.850     ‐0.117**  ‐6.720 
5yr‐I  ‐0.011*  ‐1.910     ‐0.012**  ‐3.510 
5yr‐HPI  0.011**  4.490     0.011**  10.040 
Svol  ‐0.078**  ‐3.130     ‐0.076**  ‐7.030 
Pop  ‐0.015  ‐0.480     ‐0.044**  ‐4.560 
FICO  0.012**  5.440     0.018**  11.960 
LTV  0.030**  2.840     0.033**  3.580 
D/I  ‐0.004  ‐0.380     0.212**  11.860 
Mobility  0.010  1.100     ‐0.001  ‐0.110 
Vol  ‐1.008**  ‐4.060     ‐0.157  ‐1.360 
Yield  ‐0.383**  ‐7.590     ‐0.487**  ‐15.830 
Wealth  0.000  0.550     0.000**  ‐4.230 
Constant  ‐11.805**  ‐7.730     ‐22.659**  ‐12.360 
Adjusted R‐squared     0.28        0.75 
**5 percent level of significance & *10 percent level of significance
 

 
 

35