Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Complex Mortgages

Complex Mortgages Abstract Complex mortgages became a popular borrowing instrument during the bullish housing market of the early 2000s but vanished rapidly during the subsequent downturn. These non-traditional loans, including interest-only and negative-amortization mortgages, enable households to postpone loan repayment in contrast to fully amortizing traditional mortgages. Contrary to common perception, complex mortgages are used by households with high-income levels and prime credit scores, quite unlike the low-income population targeted by subprime mortgages. Nonetheless, we find that complex-mortgage borrowers become delinquent on their mortgages at rates twice as high as borrowers with plain-vanilla fixed-rate contracts even after controlling for household and loan characteristics. Our findings suggest a link between innovations in mortgage markets focused on prime borrowers and the financial crisis. 1. Introduction The residential mortgage market experienced a significant increase in product complexity in the early 2000s, followed by a rapid reversion back to simpler products after the 2007–2009 financial crisis. The set of complex products featured contracts with zero or negative amortization, short interest rate reset periods, and low introductory teaser interest rates. Unlike traditional fixed-rate (FRM) or adjustable-rate mortgages (ARM), complex mortgages (CM) are not fully amortizing and enable households to postpone loan repayment. The two main types of complex mortgages are interest-only (IO) and negative-amortization mortgages (NEGAM). Figure 1 shows the proportion of fixed-rate, adjustable-rate, interest-only, and negative-amortization mortgage products originated between 1998 and 2009, as reported by LPS Applied Analytics. The share of CM in the USA remained below 2% until the second half of 2003 before jumping to about 30% of mortgage originations just two years later. Figure 1. View largeDownload slide Composition of mortgage products. The figure depicts the composition between fixed-rate mortgages (FRM), adjustable-rate mortgages (ARM), interest-only mortgages (IO), and negative-amortization mortgages (NEGAM) originated over the period between 1998 and 2009. Figure 1. View largeDownload slide Composition of mortgage products. The figure depicts the composition between fixed-rate mortgages (FRM), adjustable-rate mortgages (ARM), interest-only mortgages (IO), and negative-amortization mortgages (NEGAM) originated over the period between 1998 and 2009. There are two contrasting views about the primary drivers of growth in CM products. On the one hand, CM might be pushed by predatory financial institutions to take advantage of naïve households who do not fully understand the contract terms (Gabaix and Laibson, 2006; Carlin, 2009; Carlin and Manso, 2011).1 On the other hand, CM might be taken out by sophisticated households to smooth consumption over their life cycle and to enhance the affordability of expensive homes. The low initial payments of CM are beneficial if households expect their income levels or housing prices to increase over time (Piskorski and Tchistyi, 2010; Cocco, 2013). CM might also be preferred by households who are less averse to defaulting on their mortgages in case of unfavorable house price shocks (Guiso et al., 2013; Garmaise, 2013). By minimizing the initial mortgage amortization and keeping a high mortgage balance, these households maximize the value of their default option. Understanding the clientele for CM and their default patterns is important in assessing the ex ante effectiveness of such financial products and in informing policy debates on financial regulation. Our article contributes to the debate by studying the characteristics of CM borrowers and their subsequent default behavior. We first analyze the borrowers’ choice of the mortgage contract type. Contrary to popular perception, CM are primarily used by households with high-income levels and prime credit scores. Campbell (2006) shows that better-educated and better-off borrowers are more likely to refinance their houses when interest rates are falling. In a similar vein, Bucks and Pence (2008) find that low-income and low-educated households are least knowledgeable about the terms of their mortgage contracts. Furthermore, households who can maintain a high FICO score show that they have the discipline and knowledge to plan their financial matters effectively. Therefore, the CM borrowers are likely more financially sophisticated than average households, in contrast to the subprime borrowers that have received much attention in recent studies.2 We find that both consumption smoothing and default costs are important determinants of selecting CM.3 Consistent with life-cycle consumption smoothing, geographic areas with a higher proportion of young households, higher past house price appreciation, and higher population growth have greater shares of CM. Furthermore, CM borrowers are also more likely to face lower costs of mortgage delinquency. For example, they are more likely to purchase investment properties and to reside in non-recourse states, in which lenders do not have access to the non-collateralized household assets in case of mortgage delinquency. On the other hand, CM borrowers are likely to stretch their income too much to purchase houses that they otherwise could not afford. Despite their high-income levels and prime credit scores, they have higher value-to-income ratios, and are more likely to provide incomplete documentation. We next study the default behavior of mortgage borrowers. The goal of our analysis is to tease out different factors contributing to mortgage delinquency. The contractual design of CM can affect the delinquency rate for several reasons. First, CM payments can change significantly over time, as low initial payments on deferred amortization contracts rise after amortization resets. Households who are already stretching to meet the initial contract obligations might have difficulty meeting the additional monthly payments, especially if they experience unfavorable income or expenditure shocks. Naïve households who are lured into CM to purchase expensive houses might be particularly prone to default. Second, the lack of amortization inevitably leads to higher loan-to-value ratios for any given path of house prices. Households are more likely to default on their mortgages when the current value of the house is lower than the remaining loan balance. Third, CM might also have different delinquency rates due to self-selection of borrowers. CM borrowers might be more risk seeking, might be more exposed to background risks, and might be more willing to default strategically, generating higher delinquency rates. To test these hypotheses, we evaluate default outcomes in the context of a multivariate hazard model. We find evidence of higher mortgage delinquencies associated with both payment resets and increased leverage inherent in CM contracts. Moreover, high value-to-income ratios, which measure the degree that households stretch their purchasing power, are a good predictor of future defaults. Therefore, innovations in mortgage contract design are important determinants of subsequent defaults, both due to the ex-ante enabling of larger house purchases and the ex-post cash flow patterns. However, even after controlling for leverage, payment resets, and other household and loan characteristics, we find significantly higher default rates among CM borrowers. The default probabilities on complex loans are around twice the default probabilities of similar households using FRM. Households that use CM are also significantly more likely to declare personal bankruptcy. This evidence suggests that households who select into CM exhibit higher risk exposures than households who select more traditional mortgages. To investigate why CM borrowers have higher propensities to default, we perform two additional tests. First, we interact the CM indicator with the loan-to-value ratio, which proxies for the value of the default option. We find that CM borrowers’ defaults are more sensitive to the loan-to-value ratio, suggesting that CM borrowers’ decisions depend more on whether defaulting is economically beneficial. In our second test, we interact mortgage contract type with income and credit scores, which proxy for financial savvy and liquidity constraints. While households with high income levels and high credit scores are less likely to default on average, delinquency rates are less affected by income and credit scores for borrowers with CM. Thus, delinquency rates are relatively high for the more savvy and less constrained households with complex contracts. Different types of mortgages exhibit different levels of complexity. FRMs feature constant monthly payments over the whole life of the loan and are therefore the least complex instruments. ARMs exhibit intermediate levels of complexity since the mortgage payments can adjust due to changes in short-term interest rates. CMs are more complex than ARMs, as they typically exhibit changes in mortgage payments due to both changes in interest rates and changes in the amortization schedules. Within CMs, negative-amortization loans are more complex than IO loans, as they feature the largest potential variation in mortgage payments. Consistent with these varying degrees of complexity, we find that our results are more pronounced for mortgages that are more complex. In summary, our findings suggest that complex borrowers are often sophisticated borrowers. Their active choices, both ex-ante for contract selection and ex-post for default decisions, are key drivers behind the mounting defaults of CM during the recent crisis. The expansion of credit and mortgage securitization has received much attention following the financial crisis of 2007–2009. Mian and Sufi (2009) show that the sharp increase in mortgage defaults in 2007 was significantly amplified in geographic areas with a high density of subprime loans that experienced an unprecedented growth in mortgage credit prior to 2007. Keys et al. (2010), Purnanandam (2011), Jiang et al. (2014a), and Griffin and Maturana (2016) focus on the role of mortgage securitization process, finding that lenders applied lower screening efforts on loans that had higher ex-ante probabilities of being securitized.4 Our article suggests a link between innovations in mortgage contract design and the subsequent default behavior of prime borrowers. A number of recent papers highlight the importance of middle-class borrowers for the mortgage crisis. Adelino et al. (2016) show that mortgage originations increased for borrowers across all income levels and FICO scores and that delinquencies during the crisis rose particularly sharply for middle- and high-income borrowers with prime credit. Ferreira and Gyourko (2015) further find that the jump in foreclosures among prime borrowers can be attributed to their negative equity. Our article peels back the lid on the mechanism behind the surge in prime borrower defaults by focusing on the role of contract design. As such, the paper contributes to this literature by suggesting an additional and important channel linking mortgage market innovations to the financial crisis of 2007–2009. Several papers have investigated the design of non-traditional mortgage contracts and their role in the recent crisis. Piskorski and Tchistyi (2010) demonstrate that households with sizeable income variability can benefit more from loans with flexible payment schedules and adjustable rates, whereas Piskorski and Tchistyi (2011) show that such loans are particularly beneficial to borrowers living in areas with high expected house price and income growth. Corbae and Quintin (2015) calibrate a model in which the presence of IO mortgages allows more households to afford larger houses and hence significantly increases the foreclosure rate during the crisis. Cocco (2013) shows that negative amortization contracts in the UK allowed households to smooth their life-cycle consumption of housing. CM households borrowed more relative to their current income and experienced higher ex-post income growth. Importantly, these benefits were concentrated in the period in which new regulations lessened the likelihood of predatory lending practices. Garmaise (2013) shows that offering CM contracts with greater flexibility resulted in an adversely selected pool of borrowers, which suggests an important role for self-selection into CM. Barlevy and Fisher (2011) focus on the effects of non-amortizing mortgages on house prices. They describe a rational expectations model in which both speculators and their lenders use IO mortgages when there is a bubble in-house prices. They provide evidence that IO mortgages were used extensively in cities where inelastic housing supply enables pronounced boom-bust cycles. Landier, Srear, and Thesmar (2015) provide evidence consistent with the use of CM for risk-shifting by a large mortgage originator, whose portfolio of mortgages retained for investment was hit by monetary policy tightening in 2004. In response, the originator began issuing deferred-amortization mortgages, which were designed to be more sensitive to real estate prices than standard contracts. In contrast to these papers, our paper focuses on the characteristics of CM borrowers and on the key determinants of their default behavior. The remainder of this article is structured as follows. Section 2 describes our data sources and reports summary statistics. In Section 3, we study the mortgage choice of households and describe the main features of mortgage contracts. In Section 4, we study the delinquency of different contract types. Section 5 offers concluding remarks. 2. Data Sources and Summary Statistics This section describes in detail the data sources and the main mortgage contracts in our sample. 2.1 Data Our study relies on several complementary data sources that cover various aspects of the housing market during the period between 2003 and 2009. In particular, the micro-level analysis of mortgage contract choice and performance relies heavily on the proprietary mortgage-level database offered by Lender Processing Services (LPS) Applied Analytics (formerly known as McDash Analytics). LPS collects data from some of the nation’s largest mortgage servicers that report contract and borrower details at the time of loan origination, as well as monthly information on mortgage performance. The LPS data coverage has grown steadily over time, with 9 out of 10 largest servicers reporting to the database by 2003. Our database covers about 10 million mortgages with a total loan value of more than $2 trillion originated between 2003 and 2007. We track the performance of all loans till the end of 2009. For the purposes of our study, the availability of granular information on mortgage contract terms is of particular importance. For each of the loans, LPS provides information on the loan interest rate, the amortization schedule, and the securitization status. For ARM, we know the rate at origination, the frequency of resets, the reference rate, and the associated contractual spread. For loans that do not amortize steadily over their term, we know the horizon of the IO period, whether negative amortization is allowed and if so, to what extent and over what period of time. This information allows us to categorize loan contracts. The LPS data also contains information on borrower and property characteristics at the time of origination. These include the appraised property value, the first-lien loan-to-value ratio (LTV), property type (single family or condominium), whether the property was to be occupied by the borrower, and the borrower’s creditworthiness as measured by their FICO (Fair Isaac Corporation) credit score. An important feature of the LPS database is that unlike some other data sources, it is not limited to a particular subset of the loan universe. The LPS data cover prime, subprime, and Alt-A loans,5 and include loans that are privately securitized, those that are sold to Government Sponsored Enterprises (GSEs), and loans that are held on banks’ balance sheets. Even though the coverage allows for a broad set of mortgage contracts, the data set is skewed in favor of securitized loans that are more likely to be serviced by large corporations reporting to LPS. Still, the large overall size of the dataset ensures that we have ample coverage of all contract types.6 We complement borrower information in LPS with household income data collected under the Home Mortgage Disclosure Act (HMDA). Doing so allows us to compute measures of loan affordability, such as the ratio of house value to income (VTI). We further augment the loan-level data with information on trends in local home prices. Quarterly data on home prices by metropolitan statistical area (MSA) are obtained from the Federal Housing Finance Agency (FHFA). We further utilize zip code level information from the 2000 US Census to control for broad demographic characteristics, such as education levels and age distributions. We also make use of the annual per capita income level and unemployment rate data at the MSA level from the Bureau of Economic Analysis (BEA). To determine whether lender recourse has an impact on mortgage choices and mortgage defaults, we follow Ghent and Kudlyak (2011) and classify US states into recourse and non-recourse categories. Whereas lender claims in non-recourse states are limited to the value of the collateral securing the loan, lenders in recourse states may be able to collect on debt not covered by the proceedings from a foreclosure sale by obtaining a deficiency judgment.7 For our delinquency hazard models we track three economic variables after the origination of the mortgage at an annual frequency. The increase in housing prices is defined as the cumulative change of home prices in the MSA since origination. The change in loan balance captures how aggressively households reduced their mortgage debt since origination. Finally, payment resets are defined as the increases in the minimum required mortgage payments since origination. Payment resets are driven by interest rate changes for ARMs and by both interest rate and amortization changes for CMs. The variable definitions and data sources are listed in Table I. Table I. Variable definitions and data sources This table reports the description of the variables used and the corresponding data sources. Variable Data Source Aggregation Description Loan amount LPS Individual First-lien loan amount House value LPS Individual Appraised home value at origination Income HMDA Individual Reported Income from loan application FICO LPS Individual FICO at origination LTV LPS Individual First-lien loan amount divided by appraised value of home VTI LPS Individual Appraisal value divided by income from loan application Interest Rate LPS Individual Average initial interest rate Low documentation LPS Individual Low or no documentation loan indicator variable Above conforming limit LPS Individual Indicator variable for conforming loan. Government securitized LPS Individual Securitization indicator variable after one year of loan life Private securitized LPS Individual Securitization indicator variable after one year of loan life Condo LPS Individual Condominium indicator variable Investment property LPS Individual Second home or investment property indicator variable Refinance LPS Individual Refinance indicator variable With prepayment penalty LPS Individual Indicator variable for prepayment penalty Prepayment penalty term LPS Individual Length in months of prepayment penalty College Census Zip (static) Proportion of 2000 population with college education or better Young Census Zip (static) Proportion of 2000 adult population between 20 and 40 years old House price change FHFA CBSA-Qtr Cumulative house price change in the 5 years prior to the loan origination Non-recourse Ghent and Kudlyak (2011) State/ Individual States where recourse in residential mortgages is limited by the value of the collateral. For California, only purchase loans are classified as non-recourse. BEA income BLS CBSA-Qtr Mean income level per capita Increase in house value FHFA Individual Cumulative house price appreciation at the MSA level since origination Increase in loan balance LPS Individual Percentage change in loan balance since origination Payments reset LPS Individual Percentage change in minimum monthly mortgage payment since origination (1% winsorization; correction of payments due to interest double-counting) Unemployment level BLS CBSA-Qtr Unemployment rate Income growth since origination BEA CBSA-Qtr Growth rate of per capita personal income Variable Data Source Aggregation Description Loan amount LPS Individual First-lien loan amount House value LPS Individual Appraised home value at origination Income HMDA Individual Reported Income from loan application FICO LPS Individual FICO at origination LTV LPS Individual First-lien loan amount divided by appraised value of home VTI LPS Individual Appraisal value divided by income from loan application Interest Rate LPS Individual Average initial interest rate Low documentation LPS Individual Low or no documentation loan indicator variable Above conforming limit LPS Individual Indicator variable for conforming loan. Government securitized LPS Individual Securitization indicator variable after one year of loan life Private securitized LPS Individual Securitization indicator variable after one year of loan life Condo LPS Individual Condominium indicator variable Investment property LPS Individual Second home or investment property indicator variable Refinance LPS Individual Refinance indicator variable With prepayment penalty LPS Individual Indicator variable for prepayment penalty Prepayment penalty term LPS Individual Length in months of prepayment penalty College Census Zip (static) Proportion of 2000 population with college education or better Young Census Zip (static) Proportion of 2000 adult population between 20 and 40 years old House price change FHFA CBSA-Qtr Cumulative house price change in the 5 years prior to the loan origination Non-recourse Ghent and Kudlyak (2011) State/ Individual States where recourse in residential mortgages is limited by the value of the collateral. For California, only purchase loans are classified as non-recourse. BEA income BLS CBSA-Qtr Mean income level per capita Increase in house value FHFA Individual Cumulative house price appreciation at the MSA level since origination Increase in loan balance LPS Individual Percentage change in loan balance since origination Payments reset LPS Individual Percentage change in minimum monthly mortgage payment since origination (1% winsorization; correction of payments due to interest double-counting) Unemployment level BLS CBSA-Qtr Unemployment rate Income growth since origination BEA CBSA-Qtr Growth rate of per capita personal income Table I. Variable definitions and data sources This table reports the description of the variables used and the corresponding data sources. Variable Data Source Aggregation Description Loan amount LPS Individual First-lien loan amount House value LPS Individual Appraised home value at origination Income HMDA Individual Reported Income from loan application FICO LPS Individual FICO at origination LTV LPS Individual First-lien loan amount divided by appraised value of home VTI LPS Individual Appraisal value divided by income from loan application Interest Rate LPS Individual Average initial interest rate Low documentation LPS Individual Low or no documentation loan indicator variable Above conforming limit LPS Individual Indicator variable for conforming loan. Government securitized LPS Individual Securitization indicator variable after one year of loan life Private securitized LPS Individual Securitization indicator variable after one year of loan life Condo LPS Individual Condominium indicator variable Investment property LPS Individual Second home or investment property indicator variable Refinance LPS Individual Refinance indicator variable With prepayment penalty LPS Individual Indicator variable for prepayment penalty Prepayment penalty term LPS Individual Length in months of prepayment penalty College Census Zip (static) Proportion of 2000 population with college education or better Young Census Zip (static) Proportion of 2000 adult population between 20 and 40 years old House price change FHFA CBSA-Qtr Cumulative house price change in the 5 years prior to the loan origination Non-recourse Ghent and Kudlyak (2011) State/ Individual States where recourse in residential mortgages is limited by the value of the collateral. For California, only purchase loans are classified as non-recourse. BEA income BLS CBSA-Qtr Mean income level per capita Increase in house value FHFA Individual Cumulative house price appreciation at the MSA level since origination Increase in loan balance LPS Individual Percentage change in loan balance since origination Payments reset LPS Individual Percentage change in minimum monthly mortgage payment since origination (1% winsorization; correction of payments due to interest double-counting) Unemployment level BLS CBSA-Qtr Unemployment rate Income growth since origination BEA CBSA-Qtr Growth rate of per capita personal income Variable Data Source Aggregation Description Loan amount LPS Individual First-lien loan amount House value LPS Individual Appraised home value at origination Income HMDA Individual Reported Income from loan application FICO LPS Individual FICO at origination LTV LPS Individual First-lien loan amount divided by appraised value of home VTI LPS Individual Appraisal value divided by income from loan application Interest Rate LPS Individual Average initial interest rate Low documentation LPS Individual Low or no documentation loan indicator variable Above conforming limit LPS Individual Indicator variable for conforming loan. Government securitized LPS Individual Securitization indicator variable after one year of loan life Private securitized LPS Individual Securitization indicator variable after one year of loan life Condo LPS Individual Condominium indicator variable Investment property LPS Individual Second home or investment property indicator variable Refinance LPS Individual Refinance indicator variable With prepayment penalty LPS Individual Indicator variable for prepayment penalty Prepayment penalty term LPS Individual Length in months of prepayment penalty College Census Zip (static) Proportion of 2000 population with college education or better Young Census Zip (static) Proportion of 2000 adult population between 20 and 40 years old House price change FHFA CBSA-Qtr Cumulative house price change in the 5 years prior to the loan origination Non-recourse Ghent and Kudlyak (2011) State/ Individual States where recourse in residential mortgages is limited by the value of the collateral. For California, only purchase loans are classified as non-recourse. BEA income BLS CBSA-Qtr Mean income level per capita Increase in house value FHFA Individual Cumulative house price appreciation at the MSA level since origination Increase in loan balance LPS Individual Percentage change in loan balance since origination Payments reset LPS Individual Percentage change in minimum monthly mortgage payment since origination (1% winsorization; correction of payments due to interest double-counting) Unemployment level BLS CBSA-Qtr Unemployment rate Income growth since origination BEA CBSA-Qtr Growth rate of per capita personal income 2.2 Mortgage Contract Design The menu of household mortgage choices was dominated for decades by fully amortizing long-term FRM and, to a lesser extent, by ARM that locked in the initial interest rate for the first years of the contract. In the early 2000s, CM that allow for the deferral of principal repayment became increasingly popular. FRM are level-payment fully amortizing loans with maturities between 10 and 30 years. Borrowers generally have the option to prepay the mortgage if they sell the property or if they refinance their loan when interest rates decline. ARM are fully-amortizing loans where the interest rate changes after an initial period according to a preselected interest rate index. These mortgages often exhibit caps and floors that prevent interest rates from changing too much over the lifetime of the loan. ARM interest rates are generally lower than those on FRMs due to the upward-sloping term structure of interest rates.8 CM include a variety of back-loaded mortgage contracts. Most CM feature adjustable interest rates and exhibit time-varying amortization schedules. The most popular contract is an IO mortgage that only requires borrowers to pay the mortgage interest over an initial time period lasting typically between five and ten years. Subsequently, the mortgage becomes a fully amortizing loan. The other popular type of a CM is a negative amortization mortgage, also known as an option ARM. These mortgages give borrowers the option to initially pay even less than the interest due. The difference between the interest due and the actual mortgage payment is added to the loan balance. These mortgages carry the risk of larger increases in mortgage payments, when the mortgage is recast to become a fully amortizing loan after 5–10 years or when the loan balance exceeds the initial balance at origination by more than a certain amount (typically 10–25%). An additional common feature of NEGAM is a low teaser interest rate of between 1% and 2% during the first 1–12 months. The minimum payment on a NEGAM contract is often set at the level sufficient to cover teaser interest rate charges, and is raised by up to 7.5% on each anniversary of the loan. Overall, one can trace a step function of complexity in mortgage contracts from a standard FRM to ARM, IO, and then NEGAM contracts. With each step, one of the key contract parameters—the interest rate or the amortization schedule—becomes variable, complicating the borrower’s ability to forecast future loan payments, but also expanding the array of borrower actions. 2.3 Geographic Distribution of Mortgages Figure 2 shows the concentration of CM in different counties across the USA in 2002, 2005, and 2008. Consistent with Figure 1, we find that CM were fairly uncommon in 2002. The distribution of CM looks dramatically different in 2005, when multiple counties in California, Colorado, Florida, and Nevada had CM shares in excess of 40%. In some zip codes in these states more than half of mortgage originations were complex loans. Although areas with larger shares of CM often enjoyed higher house price appreciation in the past, there are also exceptions. Numerous areas with high house price appreciation had few CM even at the peak of the housing boom. For example, CM contracts accounted for only about 5% of loans in the Albany, NY metropolitan area where house prices rose by more than 80% between 2001 and 2007. In contrast, CMs proved to be very popular in the Detroit MSA, where nominal house prices remained flat during this period. It is also worth noting that in some areas rapid price increases preceded the surge in CM contracts, whereas other areas had the opposite relationship. Figure 2. View largeDownload slide Geographic distribution of complex mortgages. These figures depict the geographic distribution of complex mortgages originated in 2002, 2005, and 2008. Panel A: Complex Mortgages in 2002; Panel B: Complex Mortgages in 2005; Panel C: Complex Mortgages in 2008. Figure 2. View largeDownload slide Geographic distribution of complex mortgages. These figures depict the geographic distribution of complex mortgages originated in 2002, 2005, and 2008. Panel A: Complex Mortgages in 2002; Panel B: Complex Mortgages in 2005; Panel C: Complex Mortgages in 2008. 2.4 Summary Statistics by Mortgage Type Table II reports statistics for our broad mortgage categories. Our data contain close to 10 million loan contracts originated between 2003 and 2007, of which 69% are FRMs, 13% are ARMs, and the remaining 18% are CMs.9 Table II. Summary statistics by mortgage type This table reports summary statistics for fixed-rate mortgages (FRM), adjustable-rate mortgages (ARM), complex mortgages (CM), and for different types of complex mortgages including interest-only mortgages (IO) and negative-amortization mortgages (NEGAM). All mortgages Complex mortgages FRM ARM CM IO NEGAM Number of observations 6,744,639 1,242,097 1,733,516 1,356,330 377,186 Loan-level variables at origination: Loan amount 180,938 223,838 334,277 328,600 354,691 House value 268,087 310,377 473,833 465,045 505,433 Income 88,329 100,218 133,481 131,066 142,168 Income with full documentation 86,049 95,967 117,904 117,195 121,20 FICO 710.14 681.69 713.18 714.89 707.04 FICO below 620 (in %) 9.85 23.18 5.97 6.65 3.54 First-lien loan-to-value ratio (LTV) 73.62 76.74 73.29 73.53 72.44 Value-to-income ratio (VTI) 3.50 3.54 4.16 4.15 4.23 Initial interest rate (in %) 6.15 6.15 5.03 5.92 1.86 Refinance (in %) 41.38 35.42 44.91 39.71 63.61 Condo (in %) 11.37 16.70 18.63 19.57 15.25 Investment property (in %) 9.05 9.76 11.50 11.63 11.03 Low documentation (in %) 11.13 11.96 24.94 19.99 42.72 Government securitized (in %) 79.04 40.28 25.56 31.06 5.81 Private securitized (in %) 15.05 42.10 53.15 51.94 57.48 With prepayment penalty (in %) 5.76 26.62 32.77 19.31 81.19 Above conforming limit (in %) 5.22 14.00 33.42 31.67 39.71 MSA-level variables at origination: BEA income 36,918 37,483 40,953 41,004 40,767 Proportion with college education (in %) 33.79 36.36 38.61 38.86 37.71 Proportion of young (in %) 40.24 41.21 41.06 41.16 40.68 House price change prior 5 years (in %) 49.94 55.53 74.03 71.71 82.30 Population Growth (in %) 1.10 1.11 1.10 1.14 0.96 Unemployment rate (in %) 5.03 5.20 4.79 4.75 4.97 Non-recourse mortgage (in %) 15.50 20.79 27.68 28.42 25.04 Dynamic variables after origination: Increase in housing values (in %) 9.41 9.49 1.70 1.55 2.35 Increase in loan balance (in %) −5.37 −4.34 −1.87 −2.69 1.55 Payment resets (in %) −0.44 0.10 1.59 0.35 6.82 All mortgages Complex mortgages FRM ARM CM IO NEGAM Number of observations 6,744,639 1,242,097 1,733,516 1,356,330 377,186 Loan-level variables at origination: Loan amount 180,938 223,838 334,277 328,600 354,691 House value 268,087 310,377 473,833 465,045 505,433 Income 88,329 100,218 133,481 131,066 142,168 Income with full documentation 86,049 95,967 117,904 117,195 121,20 FICO 710.14 681.69 713.18 714.89 707.04 FICO below 620 (in %) 9.85 23.18 5.97 6.65 3.54 First-lien loan-to-value ratio (LTV) 73.62 76.74 73.29 73.53 72.44 Value-to-income ratio (VTI) 3.50 3.54 4.16 4.15 4.23 Initial interest rate (in %) 6.15 6.15 5.03 5.92 1.86 Refinance (in %) 41.38 35.42 44.91 39.71 63.61 Condo (in %) 11.37 16.70 18.63 19.57 15.25 Investment property (in %) 9.05 9.76 11.50 11.63 11.03 Low documentation (in %) 11.13 11.96 24.94 19.99 42.72 Government securitized (in %) 79.04 40.28 25.56 31.06 5.81 Private securitized (in %) 15.05 42.10 53.15 51.94 57.48 With prepayment penalty (in %) 5.76 26.62 32.77 19.31 81.19 Above conforming limit (in %) 5.22 14.00 33.42 31.67 39.71 MSA-level variables at origination: BEA income 36,918 37,483 40,953 41,004 40,767 Proportion with college education (in %) 33.79 36.36 38.61 38.86 37.71 Proportion of young (in %) 40.24 41.21 41.06 41.16 40.68 House price change prior 5 years (in %) 49.94 55.53 74.03 71.71 82.30 Population Growth (in %) 1.10 1.11 1.10 1.14 0.96 Unemployment rate (in %) 5.03 5.20 4.79 4.75 4.97 Non-recourse mortgage (in %) 15.50 20.79 27.68 28.42 25.04 Dynamic variables after origination: Increase in housing values (in %) 9.41 9.49 1.70 1.55 2.35 Increase in loan balance (in %) −5.37 −4.34 −1.87 −2.69 1.55 Payment resets (in %) −0.44 0.10 1.59 0.35 6.82 Table II. Summary statistics by mortgage type This table reports summary statistics for fixed-rate mortgages (FRM), adjustable-rate mortgages (ARM), complex mortgages (CM), and for different types of complex mortgages including interest-only mortgages (IO) and negative-amortization mortgages (NEGAM). All mortgages Complex mortgages FRM ARM CM IO NEGAM Number of observations 6,744,639 1,242,097 1,733,516 1,356,330 377,186 Loan-level variables at origination: Loan amount 180,938 223,838 334,277 328,600 354,691 House value 268,087 310,377 473,833 465,045 505,433 Income 88,329 100,218 133,481 131,066 142,168 Income with full documentation 86,049 95,967 117,904 117,195 121,20 FICO 710.14 681.69 713.18 714.89 707.04 FICO below 620 (in %) 9.85 23.18 5.97 6.65 3.54 First-lien loan-to-value ratio (LTV) 73.62 76.74 73.29 73.53 72.44 Value-to-income ratio (VTI) 3.50 3.54 4.16 4.15 4.23 Initial interest rate (in %) 6.15 6.15 5.03 5.92 1.86 Refinance (in %) 41.38 35.42 44.91 39.71 63.61 Condo (in %) 11.37 16.70 18.63 19.57 15.25 Investment property (in %) 9.05 9.76 11.50 11.63 11.03 Low documentation (in %) 11.13 11.96 24.94 19.99 42.72 Government securitized (in %) 79.04 40.28 25.56 31.06 5.81 Private securitized (in %) 15.05 42.10 53.15 51.94 57.48 With prepayment penalty (in %) 5.76 26.62 32.77 19.31 81.19 Above conforming limit (in %) 5.22 14.00 33.42 31.67 39.71 MSA-level variables at origination: BEA income 36,918 37,483 40,953 41,004 40,767 Proportion with college education (in %) 33.79 36.36 38.61 38.86 37.71 Proportion of young (in %) 40.24 41.21 41.06 41.16 40.68 House price change prior 5 years (in %) 49.94 55.53 74.03 71.71 82.30 Population Growth (in %) 1.10 1.11 1.10 1.14 0.96 Unemployment rate (in %) 5.03 5.20 4.79 4.75 4.97 Non-recourse mortgage (in %) 15.50 20.79 27.68 28.42 25.04 Dynamic variables after origination: Increase in housing values (in %) 9.41 9.49 1.70 1.55 2.35 Increase in loan balance (in %) −5.37 −4.34 −1.87 −2.69 1.55 Payment resets (in %) −0.44 0.10 1.59 0.35 6.82 All mortgages Complex mortgages FRM ARM CM IO NEGAM Number of observations 6,744,639 1,242,097 1,733,516 1,356,330 377,186 Loan-level variables at origination: Loan amount 180,938 223,838 334,277 328,600 354,691 House value 268,087 310,377 473,833 465,045 505,433 Income 88,329 100,218 133,481 131,066 142,168 Income with full documentation 86,049 95,967 117,904 117,195 121,20 FICO 710.14 681.69 713.18 714.89 707.04 FICO below 620 (in %) 9.85 23.18 5.97 6.65 3.54 First-lien loan-to-value ratio (LTV) 73.62 76.74 73.29 73.53 72.44 Value-to-income ratio (VTI) 3.50 3.54 4.16 4.15 4.23 Initial interest rate (in %) 6.15 6.15 5.03 5.92 1.86 Refinance (in %) 41.38 35.42 44.91 39.71 63.61 Condo (in %) 11.37 16.70 18.63 19.57 15.25 Investment property (in %) 9.05 9.76 11.50 11.63 11.03 Low documentation (in %) 11.13 11.96 24.94 19.99 42.72 Government securitized (in %) 79.04 40.28 25.56 31.06 5.81 Private securitized (in %) 15.05 42.10 53.15 51.94 57.48 With prepayment penalty (in %) 5.76 26.62 32.77 19.31 81.19 Above conforming limit (in %) 5.22 14.00 33.42 31.67 39.71 MSA-level variables at origination: BEA income 36,918 37,483 40,953 41,004 40,767 Proportion with college education (in %) 33.79 36.36 38.61 38.86 37.71 Proportion of young (in %) 40.24 41.21 41.06 41.16 40.68 House price change prior 5 years (in %) 49.94 55.53 74.03 71.71 82.30 Population Growth (in %) 1.10 1.11 1.10 1.14 0.96 Unemployment rate (in %) 5.03 5.20 4.79 4.75 4.97 Non-recourse mortgage (in %) 15.50 20.79 27.68 28.42 25.04 Dynamic variables after origination: Increase in housing values (in %) 9.41 9.49 1.70 1.55 2.35 Increase in loan balance (in %) −5.37 −4.34 −1.87 −2.69 1.55 Payment resets (in %) −0.44 0.10 1.59 0.35 6.82 CM, on average, are associated with higher loan amounts relative to the traditional ARM and FRM mortgages, and are used to finance more expensive houses. For example, the average home value for CM is $473,833, whereas the average home values for FRMs and ARMs are $268,087 and $310,377, respectively. Counter to some of the commonly made assertions about CM, CM contracts are extended to borrowers with high-income levels and prime credit scores. Indeed, households who take out CM report significantly higher annual incomes ($133,481) than households borrowing through FRM ($88,329) or ARM ($100,218). This ranking persists even when the sample is restricted to loans underwritten on the basis of fully documented income.10 We also find that CM borrowers have credit scores that are better than those of ARM borrowers and similar to those of FRM borrowers. Whereas 23% of ARM borrowers have FICO credit scores below 620, the same can be said of only 10% of FRM and only 6% of CM borrowers. These results emphasize that the clientele for CM differs significantly from that for subprime loans. Nevertheless, the average ratio of house value to income (VTI)—an inverse measure of affordability—is considerably higher in CM contracts, suggesting that CM borrowers are purchasing more expensive houses relative to their income. Yet, higher spending on houses is not reflected in the loan-to-value (LTV) ratios, as all mortgage types have similar first-lien LTV values.11 Several other loan characteristics are different for CM. CM borrowers are more likely to live in a condominium and are more likely to use the property they are financing for investment purposes. We also find significant differences in the frequency of prepayment penalties across mortgage types. Unlike FRMs, a significant fraction of ARMs and CMs face penalties if the loans are prepaid within the first two or three years. CM have a slightly higher share of refinancings compared to new purchases. Since CM are particularly popular for expensive homes, they are also more likely to exceed the conforming loan limit (i.e., jumbo loans). Hence, although 79% of FRMs are securitized by government-sponsored enterprises (GSEs, such as Fannie Mae, Freddie Mac, and Ginnie Mae), only 26% of CMs go through the GSEs. Private securitization partially offsets the lack of GSE involvement in the ARM and CM markets. CM borrowers receive significantly lower initial interest rates than those with FRM or ARM loans. The mean initial interest rate of 5.03% on CM is significantly lower than the rates on FRMs and ARMs (both at about 6.15%). This result is primarily caused by negative amortization mortgages that charge, on average, an initial teaser interest rate of only 1.86%. Unfortunately, we do not observe the age and the education level of borrowers directly. However, we can compute the proportion of adults in zip codes between 20 and 40 years and the proportion of adults with a college education. We find that CM borrowers tend to live in zip codes with higher education levels. From a spatial standpoint, CMs are more common in cities that experienced high house price appreciation. The average 5-year cumulative price appreciation prior to origination amounted to 74% among complex borrowers, as compared with 50% among traditional FRM borrowers. Finally, the population growth rate and the unemployment rate at the time of origination, which capture macroeconomic conditions at the MSA level, are similar in areas with different mortgage compositions. CM are more likely to be originated in non-recourse states, where the lender cannot access assets of the defaulting households beyond the value of the collateral securing the loan. Whereas only 16% of FRMs are non-recourse, 28% of CMs are non-recourse. Since CM were originated relatively later during our sample period, the average house price appreciation after origination is lower for CMs than for FRMs and ARMs. Not surprisingly, borrowers of traditional mortgages reduce their loan balances more aggressively than CM borrowers. Required payments on CMs increase on average more than payments of FRMs and ARMs.12 To illustrate in more detail the distributions of payment resets for various types of mortgages, we compare the required payments over time to the required payments during the first year. Figure 3 shows that a significant fraction of NEGAM borrowers experience jumps in required payments during the first five years. On the other hand, ARM and IO borrowers face smaller changes in their required mortgage payments. Figure 3. View largeDownload slide Mortgage payments over time. These figures depict the cumulative distribution functions of the actual mortgage payments for adjustable- rate (ARM), interest-only (IO), and negative-amortization mortgages (NEGAM) after three and five years relative to the payments during the first year. Panel A: Third-year payment relative to first-year payment. Panel B: Fifth-year payment relative to first-year payment. Figure 3. View largeDownload slide Mortgage payments over time. These figures depict the cumulative distribution functions of the actual mortgage payments for adjustable- rate (ARM), interest-only (IO), and negative-amortization mortgages (NEGAM) after three and five years relative to the payments during the first year. Panel A: Third-year payment relative to first-year payment. Panel B: Fifth-year payment relative to first-year payment. By virtue of their amortization structure, CM largely maintain a high leverage ratio over time. In unreported results, we document that even after five years, only 22% of surviving CM paid down more than 5% of their initial balance, while about 17% increased their balance by 5% or more. This creates a sharp contrast with FRM and ARM borrowers who gradually pay down their loans. This dynamic deterioration in relative leverage ratios becomes particularly dramatic in the event of declining house prices, as experienced during the housing crisis of 2007–2009.13 The last two columns of Table II break out the key summary characteristics among the two CM types. Negative amortization loans, on average, appear to be used to finance more expensive homes and are associated with higher loan values. They also display the highest VTI ratios. As expected, negative amortization loans with their low teaser interest rates commonly carry prepayment penalties. Finally, IO contracts appear to have been subject to stricter underwriting criteria. Whereas only 20% of IOs were underwritten on the basis of less than full documentation, 43% of NEGAM loans were issued in this manner. CM contracts follow a distinct time trend, peaking between 2005 and 2007. To check whether the summary table that aggregates over all origination years (2003 through 2007 in our sample) obscures some important differences, we report in Figure 4 the means of key borrower and loan characteristics over time. We find that differences in these key characteristics across contract types persist across origination years. Income levels of CM borrowers are always higher than those of borrowers with amortizing mortgages. CM borrowers also purchase more expensive homes throughout the sample than other borrowers. The FICO scores of CM borrowers are always on par with those of FRM borrowers and higher than those of ARM borrowers, especially in years where ARMs include many subprime borrowers. CM borrowers are less likely to provide full documentation throughout the sample. CM borrowers take out larger loans relative to their income (VTI) in each of the origination years. The figure also highlights the effects of the near disappearance of the subprime market in early 2007. At that point, the FICO scores, reported income, and the VTI ratio of ARM borrowers all increase substantially. Figure 4. View largeDownload slide Time series of main characteristics. These figures depict the time series of the main characteristics of mortgage borrowers by origination year. Panel A: Income; Panel B: Home value; Panel C: FICO; Panel D: Low documentation; Panel E: Value-to-income ratio; Panel F: First-lien loan-to-value ratio. Figure 4. View largeDownload slide Time series of main characteristics. These figures depict the time series of the main characteristics of mortgage borrowers by origination year. Panel A: Income; Panel B: Home value; Panel C: FICO; Panel D: Low documentation; Panel E: Value-to-income ratio; Panel F: First-lien loan-to-value ratio. 3. Mortgage Choice There are two contrasting views about the appeal of CM products. On the one hand, CM could be predatory products that are pushed by financial institutions to take advantage of unsophisticated households who do not fully understand the contract terms. The low initial payments might obfuscate the long-term borrowing costs for naïve households (Gabaix and Laibson, 2006; Carlin, 2009; Carlin and Manso, 2011). On the other hand, CM might be taken out by sophisticated borrowers who benefit from the deferred amortization. They might take advantage of the payment profile to smooth life cycle consumption, and to afford more expensive houses. In particular, the low initial payments of CM can relax household liquidity and borrowing constraints and enable households to take larger exposures in housing assets (Amromin et al., 2007; Gerardi et al., 2010; Piskorski and Tchistyi, 2010; Cocco, 2013; Guiso, Sapienza, and Zingales, 2013). They might also maximize the value of the default option embedded in such contracts or reduce their tax burdens due to the deductibility of mortgage interest from taxable income. 3.1 Multinomial Logit Regressions of Contract Choice We estimate the likelihood of selection of a particular mortgage contract type (ARM or CM) relative to a baseline contract, which we take to be an FRM. These relative likelihoods are estimated as a function of loan- and borrower-level covariates, as well as MSA-level aggregates. Formally, we use maximum likelihood to estimate the following multinomial logit regressions: Prob(Yi=m)Prob(Yi=FRM)=eβmXi+FEiTime+FEiMSA+FEiLender, (1) where Prob(Yi=m)/Prob(Yi=FRM) is the probability of obtaining an ARM or CM relative to a FRM, X is a vector of mortgage-specific covariates, FETime are indicator variables for the origination quarters, FEMSA are MSA-indicator variables, and FELender are lender-specific indicator variables. To facilitate the interpretation of the economic significance of the results, we standardize the continuous variables by subtracting their mean and dividing by their standard deviation. Table III reports the estimated coefficients. All regressions include time-fixed effects and the standard errors are clustered by MSA. Since some of the MSA level variables are not available for the full sample, the corresponding specifications include fewer observations than the overall sample summarized in Table II. In addition, for computational reasons we only include the largest 50 lenders in the specification with lender-fixed effects. Table III. Multinomial logit regressions This table reports the coefficients of multinomial logit regressions for adjustable-rate mortgages (ARM) and complex mortgages (CM). The coefficients are measured relative to fixed-rate mortgages (FRM). The significance levels are abbreviated with asterisks: One and two asterisks denote significance at the 5% and 1% level, respectively. Individual-level MSA-level MSA Lender Covariates Covariates Fixed effects Fixed effects ARM CM ARM CM ARM CM ARM CM Log(Inc) 0.326** 0.632** 0.223** 0.482** 0.203** 0.420** 0.248** 0.459** (0.014) (0.022) (0.014) (0.015) (0.009) (0.013) (0.013) (0.015) FICO −0.512** −0.040** −0.511** −0.020* −0.519** −0.034** −0.455** 0.042** (0.011) (0.011) (0.011) (0.010) (0.011) (0.009) (0.010) (0.009) LTV 0.200** 0.320** 0.214** 0.345** 0.210** 0.350** 0.312** 0.433** (0.020) (0.027) (0.020) (0.029) (0.019) (0.031) (0.016) (0.028) VTI 0.301** 0.538** 0.186** 0.352** 0.143** 0.261** 0.233** 0.385** (0.023) (0.029) (0.020) (0.023) (0.013) (0.015) (0.023) (0.027) Low Doc 0.104* 0.787** 0.141** 0.822** 0.141** 0.815** 0.178** 0.528** (0.042) (0.050) (0.041) (0.053) (0.037) (0.046) (0.033) (0.039) Above limit 0.699** 1.274** 0.658** 1.170** 0.699** 1.137** 0.580** 1.132** (0.060) (0.084) (0.047) (0.062) (0.038) (0.039) (0.048) (0.067) Condo 0.585** 0.691** 0.424** 0.460** 0.373** 0.380** 0.490** 0.501** (0.050) (0.059) (0.051) (0.046) (0.030) (0.025) (0.047) (0.041) Investment 0.300** 0.229** 0.353** 0.208** 0.326** 0.166** 0.318** 0.343** (0.024) (0.041) (0.020) (0.031) (0.018) (0.030) (0.019) (0.029) Refinance −0.258** 0.224** −0.222** 0.287** −0.302** 0.089* −0.203** 0.044 (0.018) (0.032) (0.017) (0.045) (0.018) (0.039) (0.019) (0.047) College 0.114** 0.040* 0.127** 0.069** 0.125** 0.083** (0.012) (0.019) (0.009) (0.012) (0.012) (0.020) Young 0.091** 0.098** 0.093** 0.064** 0.081** 0.086** (0.016) (0.016) (0.010) (0.007) (0.014) (0.015) Price change 0.079** 0.364** 0.180** 0.176** 0.016 0.300** (0.027) (0.038) (0.030) (0.033) (0.026) (0.034) Pop growth 0.021 0.120** −0.024 −0.003 0.030 0.131** (0.026) (0.041) (0.015) (0.021) (0.025) (0.038) Log(BEA Inc) 0.100** 0.149** 0.035 0.712** 0.087** 0.139** (0.027) (0.041) (0.127) (0.155) (0.027) (0.039) Non-Recourse 0.344** 0.625** 0.250** 0.502** (0.048) (0.082) (0.047) (0.077) Year FE Yes Yes Yes Yes MSA FE No No Yes No Lender FE No No No Yes Observations 9, 720, 252 8, 915, 275 8, 915, 275 7, 538, 406 Individual-level MSA-level MSA Lender Covariates Covariates Fixed effects Fixed effects ARM CM ARM CM ARM CM ARM CM Log(Inc) 0.326** 0.632** 0.223** 0.482** 0.203** 0.420** 0.248** 0.459** (0.014) (0.022) (0.014) (0.015) (0.009) (0.013) (0.013) (0.015) FICO −0.512** −0.040** −0.511** −0.020* −0.519** −0.034** −0.455** 0.042** (0.011) (0.011) (0.011) (0.010) (0.011) (0.009) (0.010) (0.009) LTV 0.200** 0.320** 0.214** 0.345** 0.210** 0.350** 0.312** 0.433** (0.020) (0.027) (0.020) (0.029) (0.019) (0.031) (0.016) (0.028) VTI 0.301** 0.538** 0.186** 0.352** 0.143** 0.261** 0.233** 0.385** (0.023) (0.029) (0.020) (0.023) (0.013) (0.015) (0.023) (0.027) Low Doc 0.104* 0.787** 0.141** 0.822** 0.141** 0.815** 0.178** 0.528** (0.042) (0.050) (0.041) (0.053) (0.037) (0.046) (0.033) (0.039) Above limit 0.699** 1.274** 0.658** 1.170** 0.699** 1.137** 0.580** 1.132** (0.060) (0.084) (0.047) (0.062) (0.038) (0.039) (0.048) (0.067) Condo 0.585** 0.691** 0.424** 0.460** 0.373** 0.380** 0.490** 0.501** (0.050) (0.059) (0.051) (0.046) (0.030) (0.025) (0.047) (0.041) Investment 0.300** 0.229** 0.353** 0.208** 0.326** 0.166** 0.318** 0.343** (0.024) (0.041) (0.020) (0.031) (0.018) (0.030) (0.019) (0.029) Refinance −0.258** 0.224** −0.222** 0.287** −0.302** 0.089* −0.203** 0.044 (0.018) (0.032) (0.017) (0.045) (0.018) (0.039) (0.019) (0.047) College 0.114** 0.040* 0.127** 0.069** 0.125** 0.083** (0.012) (0.019) (0.009) (0.012) (0.012) (0.020) Young 0.091** 0.098** 0.093** 0.064** 0.081** 0.086** (0.016) (0.016) (0.010) (0.007) (0.014) (0.015) Price change 0.079** 0.364** 0.180** 0.176** 0.016 0.300** (0.027) (0.038) (0.030) (0.033) (0.026) (0.034) Pop growth 0.021 0.120** −0.024 −0.003 0.030 0.131** (0.026) (0.041) (0.015) (0.021) (0.025) (0.038) Log(BEA Inc) 0.100** 0.149** 0.035 0.712** 0.087** 0.139** (0.027) (0.041) (0.127) (0.155) (0.027) (0.039) Non-Recourse 0.344** 0.625** 0.250** 0.502** (0.048) (0.082) (0.047) (0.077) Year FE Yes Yes Yes Yes MSA FE No No Yes No Lender FE No No No Yes Observations 9, 720, 252 8, 915, 275 8, 915, 275 7, 538, 406 Table III. Multinomial logit regressions This table reports the coefficients of multinomial logit regressions for adjustable-rate mortgages (ARM) and complex mortgages (CM). The coefficients are measured relative to fixed-rate mortgages (FRM). The significance levels are abbreviated with asterisks: One and two asterisks denote significance at the 5% and 1% level, respectively. Individual-level MSA-level MSA Lender Covariates Covariates Fixed effects Fixed effects ARM CM ARM CM ARM CM ARM CM Log(Inc) 0.326** 0.632** 0.223** 0.482** 0.203** 0.420** 0.248** 0.459** (0.014) (0.022) (0.014) (0.015) (0.009) (0.013) (0.013) (0.015) FICO −0.512** −0.040** −0.511** −0.020* −0.519** −0.034** −0.455** 0.042** (0.011) (0.011) (0.011) (0.010) (0.011) (0.009) (0.010) (0.009) LTV 0.200** 0.320** 0.214** 0.345** 0.210** 0.350** 0.312** 0.433** (0.020) (0.027) (0.020) (0.029) (0.019) (0.031) (0.016) (0.028) VTI 0.301** 0.538** 0.186** 0.352** 0.143** 0.261** 0.233** 0.385** (0.023) (0.029) (0.020) (0.023) (0.013) (0.015) (0.023) (0.027) Low Doc 0.104* 0.787** 0.141** 0.822** 0.141** 0.815** 0.178** 0.528** (0.042) (0.050) (0.041) (0.053) (0.037) (0.046) (0.033) (0.039) Above limit 0.699** 1.274** 0.658** 1.170** 0.699** 1.137** 0.580** 1.132** (0.060) (0.084) (0.047) (0.062) (0.038) (0.039) (0.048) (0.067) Condo 0.585** 0.691** 0.424** 0.460** 0.373** 0.380** 0.490** 0.501** (0.050) (0.059) (0.051) (0.046) (0.030) (0.025) (0.047) (0.041) Investment 0.300** 0.229** 0.353** 0.208** 0.326** 0.166** 0.318** 0.343** (0.024) (0.041) (0.020) (0.031) (0.018) (0.030) (0.019) (0.029) Refinance −0.258** 0.224** −0.222** 0.287** −0.302** 0.089* −0.203** 0.044 (0.018) (0.032) (0.017) (0.045) (0.018) (0.039) (0.019) (0.047) College 0.114** 0.040* 0.127** 0.069** 0.125** 0.083** (0.012) (0.019) (0.009) (0.012) (0.012) (0.020) Young 0.091** 0.098** 0.093** 0.064** 0.081** 0.086** (0.016) (0.016) (0.010) (0.007) (0.014) (0.015) Price change 0.079** 0.364** 0.180** 0.176** 0.016 0.300** (0.027) (0.038) (0.030) (0.033) (0.026) (0.034) Pop growth 0.021 0.120** −0.024 −0.003 0.030 0.131** (0.026) (0.041) (0.015) (0.021) (0.025) (0.038) Log(BEA Inc) 0.100** 0.149** 0.035 0.712** 0.087** 0.139** (0.027) (0.041) (0.127) (0.155) (0.027) (0.039) Non-Recourse 0.344** 0.625** 0.250** 0.502** (0.048) (0.082) (0.047) (0.077) Year FE Yes Yes Yes Yes MSA FE No No Yes No Lender FE No No No Yes Observations 9, 720, 252 8, 915, 275 8, 915, 275 7, 538, 406 Individual-level MSA-level MSA Lender Covariates Covariates Fixed effects Fixed effects ARM CM ARM CM ARM CM ARM CM Log(Inc) 0.326** 0.632** 0.223** 0.482** 0.203** 0.420** 0.248** 0.459** (0.014) (0.022) (0.014) (0.015) (0.009) (0.013) (0.013) (0.015) FICO −0.512** −0.040** −0.511** −0.020* −0.519** −0.034** −0.455** 0.042** (0.011) (0.011) (0.011) (0.010) (0.011) (0.009) (0.010) (0.009) LTV 0.200** 0.320** 0.214** 0.345** 0.210** 0.350** 0.312** 0.433** (0.020) (0.027) (0.020) (0.029) (0.019) (0.031) (0.016) (0.028) VTI 0.301** 0.538** 0.186** 0.352** 0.143** 0.261** 0.233** 0.385** (0.023) (0.029) (0.020) (0.023) (0.013) (0.015) (0.023) (0.027) Low Doc 0.104* 0.787** 0.141** 0.822** 0.141** 0.815** 0.178** 0.528** (0.042) (0.050) (0.041) (0.053) (0.037) (0.046) (0.033) (0.039) Above limit 0.699** 1.274** 0.658** 1.170** 0.699** 1.137** 0.580** 1.132** (0.060) (0.084) (0.047) (0.062) (0.038) (0.039) (0.048) (0.067) Condo 0.585** 0.691** 0.424** 0.460** 0.373** 0.380** 0.490** 0.501** (0.050) (0.059) (0.051) (0.046) (0.030) (0.025) (0.047) (0.041) Investment 0.300** 0.229** 0.353** 0.208** 0.326** 0.166** 0.318** 0.343** (0.024) (0.041) (0.020) (0.031) (0.018) (0.030) (0.019) (0.029) Refinance −0.258** 0.224** −0.222** 0.287** −0.302** 0.089* −0.203** 0.044 (0.018) (0.032) (0.017) (0.045) (0.018) (0.039) (0.019) (0.047) College 0.114** 0.040* 0.127** 0.069** 0.125** 0.083** (0.012) (0.019) (0.009) (0.012) (0.012) (0.020) Young 0.091** 0.098** 0.093** 0.064** 0.081** 0.086** (0.016) (0.016) (0.010) (0.007) (0.014) (0.015) Price change 0.079** 0.364** 0.180** 0.176** 0.016 0.300** (0.027) (0.038) (0.030) (0.033) (0.026) (0.034) Pop growth 0.021 0.120** −0.024 −0.003 0.030 0.131** (0.026) (0.041) (0.015) (0.021) (0.025) (0.038) Log(BEA Inc) 0.100** 0.149** 0.035 0.712** 0.087** 0.139** (0.027) (0.041) (0.127) (0.155) (0.027) (0.039) Non-Recourse 0.344** 0.625** 0.250** 0.502** (0.048) (0.082) (0.047) (0.077) Year FE Yes Yes Yes Yes MSA FE No No Yes No Lender FE No No No Yes Observations 9, 720, 252 8, 915, 275 8, 915, 275 7, 538, 406 Because we find that these loans are not concentrated in low-income areas with poorly educated households, we find little support for the hypothesis that CM are pushed to naïve households. Instead, we find that households with higher income levels are significantly more likely to obtain a CM than to take out a more traditional FRM loan. The estimates imply that a one-standard-deviation change in log income raises the ratio between the probabilities of choosing a CM over an FRM contract almost twofold (⁠ e0.632=1.88 ⁠). Moreover, households with higher FICO scores are substantially more likely to choose a CM than to choose an ARM, although the results are mixed when we compare the propensity to choose a CM relative to a FRM.14 Areas with higher proportions of college graduates and with higher median incomes are also associated with a higher proportion of CM contracts. Overall, there is little evidence that a typical CM is taken out by poor households who are more prone to predatory lending. We find some evidence indicating that CM are “affordability products” for households who anticipate income growth or house price appreciation. The estimated coefficients on the loan-to-value (LTV) and the value-to-income (VTI) ratios are significantly higher for CM households, suggesting that these households are stretching their budget to afford more expensive homes. Another piece of evidence consistent with the idea of CM contracts as affordability product is that they are much more prevalent for mortgages above the GSE conforming loan limit. Such mortgages cannot be securitized by the GSEs and, consequently, result in higher interest rates (the so-called jumbo spread). This increases the relative appeal of payment-shrinking CM products. While we do not observe household expectations for their income and house price growth, we introduce several proxies for these expectations. Since young households anticipate a higher growth rate of their labor income than older households, we use the proportion of adults between 20 and 40 years to proxy for income expectations and find that CM contracts are more popular in areas with a larger proportion of younger households. To the extent that households might extrapolate past local experiences to build their expectations about future house price dynamics, we use the prior five years’ house price appreciation in the MSA to proxy for the expected future house price growth. Borrowers in geographic areas where appreciation was substantial might be more willing to accept non-amortizing loans if they expect the appreciation to continue in the future. In addition, the prior one-year population growth rate in the MSA captures primarily migration pressures. Geographic areas with significant population growth might be areas where households expect significant house price and income growth. We find that past house price appreciation and the local population growth significantly increase the propensity of obtaining a CM, which suggests that the expectations of continued growth are likely a driving force behind the popularity of CM. This evidence is also consistent with CM borrowers being subject to a behavioral bias of extrapolating past prices too much into the future, despite their higher education and income levels. Finally, we also find supporting evidence that CM are selected by a different type of households who might have different risk exposures and face different costs of mortgage delinquency. First, we observe that CM borrowers are much more likely to provide incomplete documentation for their loans. These households either are unable to provide documentation for their income due to unstable income streams, or are inflating their incomes to qualify for higher loan amounts. Going forward, they might continue to have more volatile incomes and hence higher default probabilities. Second, we find that CM mortgages are more likely to be used to finance investment properties. Owners of these properties do not lose their primary residence upon defaulting on their mortgages and hence have potentially lower costs of default. They might therefore have an incentive to pay down their mortgage balance relatively slowly to increase the default option value. Third, we also find that households in non-recourse states are significantly more likely to obtain a CM than households in recourse states. This might be caused by the higher option value of defaulting on non-recourse mortgages, when a delinquent household can default on the mortgage without worrying about lenders accessing their other assets. While it is possible that the positive association between CM contract choice and income reflects the propensity of CMs to be concentrated in high income MSAs, specifications that incorporate MSA-level controls and MSA-fixed effects preserve these relationships. Therefore, even within individual geographies, CM choice is favored by the relatively well-off. These MSA-fixed effects also control for other unobserved geographic differences in regulation, topography, and geography.15 It is also possible that the contract choices reflect the decision of lenders, who determine the menu of available contract options and possibly also steer the borrowers towards certain items on the menu. By including lender-fixed effects, we control for the fact that some lenders might offer only specific mortgage instruments and might target-specific clienteles. The coefficient on income remains economically and statistically significant after including lender-fixed effects.16 In summary, we find that CM borrowers are well-educated high-income households with prime credit scores. They are stretching their budget to purchase expensive houses, partly due to their expectation of higher future income or house price growth. They might also face lower costs of mortgage delinquency as they are more likely to provide incomplete documentation, to purchase investment properties, and to reside in non-recourse states. 3.2 Robustness Tests Table IV reports the coefficients of multinomial logit regressions that further differentiate between the two main types of complex contracts. The estimates are consistent with the univariate results in Table II. In particular, we see that NEGAM contracts are used by high-income borrowers to refinance their high-priced primary residences, often on the basis of only limited income and asset documentation. Table IV. Multinomial logit regressions for detailed classification This table reports the coefficients of multinomial logit regressions for adjustable-rate mortgages (ARM), interest-only mortgages (IO), and negative-amortization mortgages (NEGAM). The coefficients are measured relative to fixed-rate mortgages (FRM). The significance levels are abbreviated with asterisks: one and two asterisks denote significance at the 5% and 1% level, respectively. Individual-level covariates MSA-level covariates ARM IO NEGAM ARM IO NEGAM Log(Inc) 0.328** 0.582** 0.851** 0.226** 0.432** 0.714** (0.014) (0.021) (0.021) (0.015) (0.015) (0.018) FICO −0.513** −0.028* −0.089** −0.511** −0.015 −0.042** (0.011) (0.011) (0.016) (0.010) (0.010) (0.016) LTV 0.202** 0.284** 0.498** 0.216** 0.302** 0.558** (0.020) (0.026) (0.026) (0.020) (0.029) (0.026) VTI 0.302** 0.526** 0.602** 0.186** 0.345** 0.397** (0.023) (0.029) (0.031) (0.020) (0.022) (0.026) Low Doc 0.126** 0.534** 1.598** 0.163** 0.572** 1.636** (0.043) (0.047) (0.049) (0.042) (0.049) (0.053) Above limit 0.703** 1.271** 1.264** 0.661** 1.180** 1.096** (0.061) (0.080) (0.099) (0.047) (0.059) (0.072) Condo 0.582** 0.712** 0.575** 0.421** 0.486** 0.340** (0.051) (0.055) (0.089) (0.052) (0.043) (0.065) Investment 0.302** 0.207** 0.365** 0.355** 0.188** 0.351** (0.024) (0.045) (0.045) (0.020) (0.034) (0.043) Refinance −0.244** 0.021 1.070** −0.209** 0.084 1.207** (0.018) (0.032) (0.050) (0.017) (0.043) (0.057) College 0.112** 0.061** −0.066** (0.012) (0.020) (0.019) Young 0.091** 0.101** 0.066** (0.016) (0.016) (0.023) Price change 0.081** 0.319** 0.561** (0.027) (0.037) (0.050) Pop Growth 0.021 0.126** 0.072 (0.026) (0.041) (0.051) Log(BEA Inc) 0.100** 0.142** 0.192** (0.027) (0.039) (0.065) Non-recourse 0.344** 0.604** 0.825** (0.048) (0.081) (0.107) Year FE Yes Yes MSA FE No No Lender FE No No Observations 9,720,252 8,915,275 Individual-level covariates MSA-level covariates ARM IO NEGAM ARM IO NEGAM Log(Inc) 0.328** 0.582** 0.851** 0.226** 0.432** 0.714** (0.014) (0.021) (0.021) (0.015) (0.015) (0.018) FICO −0.513** −0.028* −0.089** −0.511** −0.015 −0.042** (0.011) (0.011) (0.016) (0.010) (0.010) (0.016) LTV 0.202** 0.284** 0.498** 0.216** 0.302** 0.558** (0.020) (0.026) (0.026) (0.020) (0.029) (0.026) VTI 0.302** 0.526** 0.602** 0.186** 0.345** 0.397** (0.023) (0.029) (0.031) (0.020) (0.022) (0.026) Low Doc 0.126** 0.534** 1.598** 0.163** 0.572** 1.636** (0.043) (0.047) (0.049) (0.042) (0.049) (0.053) Above limit 0.703** 1.271** 1.264** 0.661** 1.180** 1.096** (0.061) (0.080) (0.099) (0.047) (0.059) (0.072) Condo 0.582** 0.712** 0.575** 0.421** 0.486** 0.340** (0.051) (0.055) (0.089) (0.052) (0.043) (0.065) Investment 0.302** 0.207** 0.365** 0.355** 0.188** 0.351** (0.024) (0.045) (0.045) (0.020) (0.034) (0.043) Refinance −0.244** 0.021 1.070** −0.209** 0.084 1.207** (0.018) (0.032) (0.050) (0.017) (0.043) (0.057) College 0.112** 0.061** −0.066** (0.012) (0.020) (0.019) Young 0.091** 0.101** 0.066** (0.016) (0.016) (0.023) Price change 0.081** 0.319** 0.561** (0.027) (0.037) (0.050) Pop Growth 0.021 0.126** 0.072 (0.026) (0.041) (0.051) Log(BEA Inc) 0.100** 0.142** 0.192** (0.027) (0.039) (0.065) Non-recourse 0.344** 0.604** 0.825** (0.048) (0.081) (0.107) Year FE Yes Yes MSA FE No No Lender FE No No Observations 9,720,252 8,915,275 Table IV. Multinomial logit regressions for detailed classification This table reports the coefficients of multinomial logit regressions for adjustable-rate mortgages (ARM), interest-only mortgages (IO), and negative-amortization mortgages (NEGAM). The coefficients are measured relative to fixed-rate mortgages (FRM). The significance levels are abbreviated with asterisks: one and two asterisks denote significance at the 5% and 1% level, respectively. Individual-level covariates MSA-level covariates ARM IO NEGAM ARM IO NEGAM Log(Inc) 0.328** 0.582** 0.851** 0.226** 0.432** 0.714** (0.014) (0.021) (0.021) (0.015) (0.015) (0.018) FICO −0.513** −0.028* −0.089** −0.511** −0.015 −0.042** (0.011) (0.011) (0.016) (0.010) (0.010) (0.016) LTV 0.202** 0.284** 0.498** 0.216** 0.302** 0.558** (0.020) (0.026) (0.026) (0.020) (0.029) (0.026) VTI 0.302** 0.526** 0.602** 0.186** 0.345** 0.397** (0.023) (0.029) (0.031) (0.020) (0.022) (0.026) Low Doc 0.126** 0.534** 1.598** 0.163** 0.572** 1.636** (0.043) (0.047) (0.049) (0.042) (0.049) (0.053) Above limit 0.703** 1.271** 1.264** 0.661** 1.180** 1.096** (0.061) (0.080) (0.099) (0.047) (0.059) (0.072) Condo 0.582** 0.712** 0.575** 0.421** 0.486** 0.340** (0.051) (0.055) (0.089) (0.052) (0.043) (0.065) Investment 0.302** 0.207** 0.365** 0.355** 0.188** 0.351** (0.024) (0.045) (0.045) (0.020) (0.034) (0.043) Refinance −0.244** 0.021 1.070** −0.209** 0.084 1.207** (0.018) (0.032) (0.050) (0.017) (0.043) (0.057) College 0.112** 0.061** −0.066** (0.012) (0.020) (0.019) Young 0.091** 0.101** 0.066** (0.016) (0.016) (0.023) Price change 0.081** 0.319** 0.561** (0.027) (0.037) (0.050) Pop Growth 0.021 0.126** 0.072 (0.026) (0.041) (0.051) Log(BEA Inc) 0.100** 0.142** 0.192** (0.027) (0.039) (0.065) Non-recourse 0.344** 0.604** 0.825** (0.048) (0.081) (0.107) Year FE Yes Yes MSA FE No No Lender FE No No Observations 9,720,252 8,915,275 Individual-level covariates MSA-level covariates ARM IO NEGAM ARM IO NEGAM Log(Inc) 0.328** 0.582** 0.851** 0.226** 0.432** 0.714** (0.014) (0.021) (0.021) (0.015) (0.015) (0.018) FICO −0.513** −0.028* −0.089** −0.511** −0.015 −0.042** (0.011) (0.011) (0.016) (0.010) (0.010) (0.016) LTV 0.202** 0.284** 0.498** 0.216** 0.302** 0.558** (0.020) (0.026) (0.026) (0.020) (0.029) (0.026) VTI 0.302** 0.526** 0.602** 0.186** 0.345** 0.397** (0.023) (0.029) (0.031) (0.020) (0.022) (0.026) Low Doc 0.126** 0.534** 1.598** 0.163** 0.572** 1.636** (0.043) (0.047) (0.049) (0.042) (0.049) (0.053) Above limit 0.703** 1.271** 1.264** 0.661** 1.180** 1.096** (0.061) (0.080) (0.099) (0.047) (0.059) (0.072) Condo 0.582** 0.712** 0.575** 0.421** 0.486** 0.340** (0.051) (0.055) (0.089) (0.052) (0.043) (0.065) Investment 0.302** 0.207** 0.365** 0.355** 0.188** 0.351** (0.024) (0.045) (0.045) (0.020) (0.034) (0.043) Refinance −0.244** 0.021 1.070** −0.209** 0.084 1.207** (0.018) (0.032) (0.050) (0.017) (0.043) (0.057) College 0.112** 0.061** −0.066** (0.012) (0.020) (0.019) Young 0.091** 0.101** 0.066** (0.016) (0.016) (0.023) Price change 0.081** 0.319** 0.561** (0.027) (0.037) (0.050) Pop Growth 0.021 0.126** 0.072 (0.026) (0.041) (0.051) Log(BEA Inc) 0.100** 0.142** 0.192** (0.027) (0.039) (0.065) Non-recourse 0.344** 0.604** 0.825** (0.048) (0.081) (0.107) Year FE Yes Yes MSA FE No No Lender FE No No Observations 9,720,252 8,915,275 Our conclusion that borrowers with CM are relatively financially sophisticated is partially based on the fact that these borrowers report higher income levels. However, the income levels of low-documentation borrowers are not verified and might not be reliable. To investigate whether this biases our results, Table V presents the CM coefficients of the multinomial logit regressions for the subsample of households with full documentation loans.17 Overall, conditioning on full documentation loans does not affect our main results qualitatively. Table V. Multinomial logit regressions for subsamples This table reports the coefficients of multinomial logit regressions of complex mortgages (CM) relative to fixed-rate mortgages (FRM) for the following subsamples: loans with full documentation; loans to borrowers with FICO scores below 620; loans originated to purchase a new house; loans used to finance an investment property; non-securitized loans; and loans originated in states other than California. The significance levels are abbreviated with asterisks: One and two asterisks denote significance at the 5% and 1% level, respectively. Full Subprime Purchases Invest Not Exclude Document Only Only Property Securitized California Log(Inc) 0.418** 0.366** 0.429** 0.408** 0.696** 0.456** (0.016) (0.023) (0.015) (0.018) (0.022) (0.016) FICO −0.146** 0.023 −0.086** −0.097** −0.086** −0.019 (0.010) (0.029) (0.011) (0.011) (0.017) (0.011) LTV 0.370** −0.037 0.162** 0.429** 0.381** 0.251** (0.029) (0.040) (0.034) (0.019) (0.026) (0.024) VTI 0.342** 0.167** 0.337** 0.347** 0.419** 0.357** (0.024) (0.025) (0.022) (0.041) (0.039) (0.031) Low Doc 0.126 0.580** −0.152** 2.143** 0.666** (0.077) (0.049) (0.033) (0.128) (0.045) Above limit 1.055** 1.775** 1.183** 0.973** 0.731** 1.034** (0.055) (0.115) (0.067) (0.052) (0.069) (0.044) Condo 0.450** 0.487** 0.459** 0.173** 0.350** 0.436** (0.037) (0.064) (0.044) (0.035) (0.053) (0.048) Investment 0.042 −0.140* 0.294** 0.033 0.289** (0.030) (0.068) (0.033) (0.042) (0.030) Refinance 0.093* 0.267** 0.030 0.423** 0.247** (0.038) (0.034) (0.033) (0.057) (0.040) College 0.051* −0.057* 0.040* 0.062** 0.062** 0.073** (0.022) (0.027) (0.020) (0.015) (0.024) (0.018) Young 0.085** 0.106** 0.118** 0.071** 0.041* 0.075** (0.015) (0.020) (0.019) (0.014) (0.020) (0.019) Price change 0.217** 0.273** 0.438** 0.367** 0.308** 0.280** (0.033) (0.042) (0.042) (0.034) (0.048) (0.056) Pop growth 0.127** 0.014 0.166** 0.166** 0.046 0.180** (0.038) (0.032) (0.047) (0.039) (0.051) (0.051) Log(BEA Inc) 0.114** 0.018 0.175** 0.124** 0.023 0.142** (0.038) (0.050) (0.041) (0.039) (0.047) (0.053) Non-recourse 0.626** 0.777** 0.712** 0.449** 0.858** 0.361** (0.060) (0.091) (0.099) (0.061) (0.097) (0.111) Origination year FE Yes Yes Yes Yes Yes Yes MSA FE No No No No No No Lender FE No No No No No No Observations 3,279,219 979,440 5,214,750 826,604 929,451 7,545,583 Full Subprime Purchases Invest Not Exclude Document Only Only Property Securitized California Log(Inc) 0.418** 0.366** 0.429** 0.408** 0.696** 0.456** (0.016) (0.023) (0.015) (0.018) (0.022) (0.016) FICO −0.146** 0.023 −0.086** −0.097** −0.086** −0.019 (0.010) (0.029) (0.011) (0.011) (0.017) (0.011) LTV 0.370** −0.037 0.162** 0.429** 0.381** 0.251** (0.029) (0.040) (0.034) (0.019) (0.026) (0.024) VTI 0.342** 0.167** 0.337** 0.347** 0.419** 0.357** (0.024) (0.025) (0.022) (0.041) (0.039) (0.031) Low Doc 0.126 0.580** −0.152** 2.143** 0.666** (0.077) (0.049) (0.033) (0.128) (0.045) Above limit 1.055** 1.775** 1.183** 0.973** 0.731** 1.034** (0.055) (0.115) (0.067) (0.052) (0.069) (0.044) Condo 0.450** 0.487** 0.459** 0.173** 0.350** 0.436** (0.037) (0.064) (0.044) (0.035) (0.053) (0.048) Investment 0.042 −0.140* 0.294** 0.033 0.289** (0.030) (0.068) (0.033) (0.042) (0.030) Refinance 0.093* 0.267** 0.030 0.423** 0.247** (0.038) (0.034) (0.033) (0.057) (0.040) College 0.051* −0.057* 0.040* 0.062** 0.062** 0.073** (0.022) (0.027) (0.020) (0.015) (0.024) (0.018) Young 0.085** 0.106** 0.118** 0.071** 0.041* 0.075** (0.015) (0.020) (0.019) (0.014) (0.020) (0.019) Price change 0.217** 0.273** 0.438** 0.367** 0.308** 0.280** (0.033) (0.042) (0.042) (0.034) (0.048) (0.056) Pop growth 0.127** 0.014 0.166** 0.166** 0.046 0.180** (0.038) (0.032) (0.047) (0.039) (0.051) (0.051) Log(BEA Inc) 0.114** 0.018 0.175** 0.124** 0.023 0.142** (0.038) (0.050) (0.041) (0.039) (0.047) (0.053) Non-recourse 0.626** 0.777** 0.712** 0.449** 0.858** 0.361** (0.060) (0.091) (0.099) (0.061) (0.097) (0.111) Origination year FE Yes Yes Yes Yes Yes Yes MSA FE No No No No No No Lender FE No No No No No No Observations 3,279,219 979,440 5,214,750 826,604 929,451 7,545,583 Table V. Multinomial logit regressions for subsamples This table reports the coefficients of multinomial logit regressions of complex mortgages (CM) relative to fixed-rate mortgages (FRM) for the following subsamples: loans with full documentation; loans to borrowers with FICO scores below 620; loans originated to purchase a new house; loans used to finance an investment property; non-securitized loans; and loans originated in states other than California. The significance levels are abbreviated with asterisks: One and two asterisks denote significance at the 5% and 1% level, respectively. Full Subprime Purchases Invest Not Exclude Document Only Only Property Securitized California Log(Inc) 0.418** 0.366** 0.429** 0.408** 0.696** 0.456** (0.016) (0.023) (0.015) (0.018) (0.022) (0.016) FICO −0.146** 0.023 −0.086** −0.097** −0.086** −0.019 (0.010) (0.029) (0.011) (0.011) (0.017) (0.011) LTV 0.370** −0.037 0.162** 0.429** 0.381** 0.251** (0.029) (0.040) (0.034) (0.019) (0.026) (0.024) VTI 0.342** 0.167** 0.337** 0.347** 0.419** 0.357** (0.024) (0.025) (0.022) (0.041) (0.039) (0.031) Low Doc 0.126 0.580** −0.152** 2.143** 0.666** (0.077) (0.049) (0.033) (0.128) (0.045) Above limit 1.055** 1.775** 1.183** 0.973** 0.731** 1.034** (0.055) (0.115) (0.067) (0.052) (0.069) (0.044) Condo 0.450** 0.487** 0.459** 0.173** 0.350** 0.436** (0.037) (0.064) (0.044) (0.035) (0.053) (0.048) Investment 0.042 −0.140* 0.294** 0.033 0.289** (0.030) (0.068) (0.033) (0.042) (0.030) Refinance 0.093* 0.267** 0.030 0.423** 0.247** (0.038) (0.034) (0.033) (0.057) (0.040) College 0.051* −0.057* 0.040* 0.062** 0.062** 0.073** (0.022) (0.027) (0.020) (0.015) (0.024) (0.018) Young 0.085** 0.106** 0.118** 0.071** 0.041* 0.075** (0.015) (0.020) (0.019) (0.014) (0.020) (0.019) Price change 0.217** 0.273** 0.438** 0.367** 0.308** 0.280** (0.033) (0.042) (0.042) (0.034) (0.048) (0.056) Pop growth 0.127** 0.014 0.166** 0.166** 0.046 0.180** (0.038) (0.032) (0.047) (0.039) (0.051) (0.051) Log(BEA Inc) 0.114** 0.018 0.175** 0.124** 0.023 0.142** (0.038) (0.050) (0.041) (0.039) (0.047) (0.053) Non-recourse 0.626** 0.777** 0.712** 0.449** 0.858** 0.361** (0.060) (0.091) (0.099) (0.061) (0.097) (0.111) Origination year FE Yes Yes Yes Yes Yes Yes MSA FE No No No No No No Lender FE No No No No No No Observations 3,279,219 979,440 5,214,750 826,604 929,451 7,545,583 Full Subprime Purchases Invest Not Exclude Document Only Only Property Securitized California Log(Inc) 0.418** 0.366** 0.429** 0.408** 0.696** 0.456** (0.016) (0.023) (0.015) (0.018) (0.022) (0.016) FICO −0.146** 0.023 −0.086** −0.097** −0.086** −0.019 (0.010) (0.029) (0.011) (0.011) (0.017) (0.011) LTV 0.370** −0.037 0.162** 0.429** 0.381** 0.251** (0.029) (0.040) (0.034) (0.019) (0.026) (0.024) VTI 0.342** 0.167** 0.337** 0.347** 0.419** 0.357** (0.024) (0.025) (0.022) (0.041) (0.039) (0.031) Low Doc 0.126 0.580** −0.152** 2.143** 0.666** (0.077) (0.049) (0.033) (0.128) (0.045) Above limit 1.055** 1.775** 1.183** 0.973** 0.731** 1.034** (0.055) (0.115) (0.067) (0.052) (0.069) (0.044) Condo 0.450** 0.487** 0.459** 0.173** 0.350** 0.436** (0.037) (0.064) (0.044) (0.035) (0.053) (0.048) Investment 0.042 −0.140* 0.294** 0.033 0.289** (0.030) (0.068) (0.033) (0.042) (0.030) Refinance 0.093* 0.267** 0.030 0.423** 0.247** (0.038) (0.034) (0.033) (0.057) (0.040) College 0.051* −0.057* 0.040* 0.062** 0.062** 0.073** (0.022) (0.027) (0.020) (0.015) (0.024) (0.018) Young 0.085** 0.106** 0.118** 0.071** 0.041* 0.075** (0.015) (0.020) (0.019) (0.014) (0.020) (0.019) Price change 0.217** 0.273** 0.438** 0.367** 0.308** 0.280** (0.033) (0.042) (0.042) (0.034) (0.048) (0.056) Pop growth 0.127** 0.014 0.166** 0.166** 0.046 0.180** (0.038) (0.032) (0.047) (0.039) (0.051) (0.051) Log(BEA Inc) 0.114** 0.018 0.175** 0.124** 0.023 0.142** (0.038) (0.050) (0.041) (0.039) (0.047) (0.053) Non-recourse 0.626** 0.777** 0.712** 0.449** 0.858** 0.361** (0.060) (0.091) (0.099) (0.061) (0.097) (0.111) Origination year FE Yes Yes Yes Yes Yes Yes MSA FE No No No No No No Lender FE No No No No No No Observations 3,279,219 979,440 5,214,750 826,604 929,451 7,545,583 The LPS database undersamples subprime loans and portfolio loans. Table V shows that the income level remains an important predictor of CM if we focus on subprime mortgages (i.e., mortgages with FICO scores below 620) or on loans that are not securitized. Table V also shows that our results remain materially unaffected if we only study purchase transactions or investment properties. Finally, we obtain similar results if we exclude all mortgages originated in the state of California, which accounts for around 15% of our observations but a greater proportion of the CM loans. In unreported robustness tests, we run separate multinomial logit models for each year and document that the determinants of mortgage choice are relatively stable over time. For example, the income level is positively related to the choice of CM for each year in our sample. 4. Mortgage Delinquencies In this section, we study the delinquency outcomes of different types of mortgages. A mortgage is considered delinquent if the borrower is at least 60 days late with a payment. 4.1 Summary of Mortgage Delinquency Figure 5 plots the distribution of mortgage delinquencies by contract type during the first five years after origination. In each month, we depict the proportion of remaining mortgages that become delinquent for the first time. Figure 5. View largeDownload slide Proportion of mortgage delinquencies by month after origination. The figure depicts the proportion of surviving loans that are delinquent by month after origination for fixed-rate (FRM), adjustable-rate (ARM), interest-only (IO), and negative-amortization mortgages (NEGAM) over the period between 2003 and 2009. Figure 5. View largeDownload slide Proportion of mortgage delinquencies by month after origination. The figure depicts the proportion of surviving loans that are delinquent by month after origination for fixed-rate (FRM), adjustable-rate (ARM), interest-only (IO), and negative-amortization mortgages (NEGAM) over the period between 2003 and 2009. We observe that IO and NEGAM have relatively low delinquency rates for the first months due to low initial mortgage payments. Mortgage delinquencies of IO (NEGAM) mortgages then steadily increase and reach peaks of about 1.3% (2.0%) of surviving loans at 27 (39) months after origination. These peaks occur three months after common reset intervals, since delinquency begins when a mortgage payment is at least 60 days late. We observe a similar peak for ARMs at the 27-month horizon. The delinquencies of FRMs are substantially lower than the delinquencies of CM, except for short horizons. Whereas ARMs have higher rates of delinquency at short horizons, CMs overtake them at longer horizons. It must be kept in mind that CM borrowers have higher delinquency propensities at longer horizons despite having better credit scores than ARM borrowers, as summarized in Table II. Moreover, the delinquency rate increases substantially even before the minimum loan payments are reset after two or three years, indicating that some CM borrowers are stretching their borrowing capacity beyond affordable levels. They do not even make the relatively low initial mortgage payments. 4.2 Hazard Model of Delinquency To investigate the determinants of mortgage delinquencies, we run the following Cox proportional hazard model: h(i,t)=h0(t;s,v)eβXi,t+FEtYear, (2) where the hazard rate h(t) is the estimated probability of first time 60-day delinquency at time t conditional on surviving to time t−, h0(t) is the baseline hazard rate, X is a vector of household-specific covariates, and FEtYear is an indicator variable for the calendar year to control for different vintage effects and macroeconomic conditions. We allow the baseline hazard to vary for each combination of the origination year v and the state of origination s or for each combination of the origination year v and the lender s.18 The loan sample is expanded to a loan-year level so that time-varying covariates can be included. Also, time is scaled so that the first observation date is the calendar year of origination (time 0), and subsequent calendar years are measured relative to the year of origination. Implicitly, loans of different vintages are compared with each other, so that the baseline hazard represents the probability of delinquency for a borrower with covariates of 0 at t years after origination. In some specifications, we separate CM into the two sub-types (IO and NEGAM). The continuous covariates are again standardized by subtracting the mean and dividing by the standard deviation. Table VI reports the estimated coefficients of the propensity of first time delinquency. In the first column, we only use borrower and loan characteristics at origination to estimate the delinquency hazards. In the second column, we include area-specific variables and time-varying loan and area characteristics. The third column incorporates controls for loan ownership to explore the impact of securitization. The fourth column replaces the year-state baseline with the year-lender baseline to control for lender-specific determinants of delinquency. The last column decomposes CMs into IOs and NEGAMs using year-state baselines. Table VI. Hazard model of mortgage delinquency This table reports the estimates from a Cox proportional hazard model for mortgage delinquency. The significance levels are abbreviated with asterisks: One and two asterisks denote significance at the 5% and 1% level, respectively. Individual-level MSA-level Securitization Lender-year Detailed CM Covariates Covariates Controls Baselines Contracts CM 0.723** 0.699** 0.557** 0.641** (0.011) (0.012) (0.012) (0.010) IO 0.659** (0.013) NEGAM 0.939** (0.020) ARM 0.444** 0.457** 0.303** 0.418** 0.461** (0.009) (0.010) (0.013) (0.008) (0.010) Log(Inc) −0.109** −0.058** −0.060** −0.032** −0.062** (0.008) (0.010) (0.009) (0.011) (0.010) FICO −0.672** −0.663** −0.634** −0.618** −0.664** (0.010) (0.012) (0.012) (0.012) (0.012) LTV 0.506** 0.492** 0.504** 0.458** 0.492** (0.012) (0.011) (0.010) (0.010) (0.011) VTI 0.045** 0.050** 0.051** 0.061** 0.050** (0.006) (0.006) (0.006) (0.006) (0.005) Low Doc 0.046** 0.042** 0.089** 0.016 0.028* (0.012) (0.012) (0.012) (0.011) (0.012) Above limit 0.187** 0.279** 0.129** 0.294** 0.271** (0.030) (0.018) (0.018) (0.022) (0.018) Condo −0.173** −0.092** −0.083** −0.024 −0.090** (0.029) (0.025) (0.024) (0.028) (0.025) Investment 0.348** 0.332** 0.294** 0.280** 0.329** (0.026) (0.027) (0.024) (0.031) (0.027) Refinance 0.089** 0.043** 0.015 0.003 0.034** (0.011) (0.012) (0.011) (0.009) (0.012) College −0.203** −0.195** −0.198** −0.201** (0.009) (0.009) (0.008) (0.009) Young 0.017* 0.016* −0.008 0.017* (0.007) (0.007) (0.007) (0.007) Log(BEA Inc) 0.038** 0.038** 0.041* 0.037** (0.014) (0.014) (0.016) (0.014) House price growth −0.414** −0.416** −0.450** −0.415** (0.020) (0.019) (0.015) (0.020) Loan balance growth 0.016 0.018 0.061** −0.002 (0.012) (0.011) (0.009) (0.012) Payment resets 0.031** 0.029** 0.051** 0.025** (0.002) (0.002) (0.002) (0.002) Change in unemployment 0.022** 0.023** 0.026** 0.022** (0.002) (0.002) (0.002) (0.002) Income growth −0.149** −0.144** −0.134** −0.151** (0.024) (0.024) (0.038) (0.024) Gov securitized −0.164** (0.015) Priv securitized 0.246** (0.010) State-year baselines Yes Yes Yes No Yes Lender-year baselines No No No Yes No Observations 29,690,919 23,151,373 23,151,373 23,151,373 23,151,373 Individual-level MSA-level Securitization Lender-year Detailed CM Covariates Covariates Controls Baselines Contracts CM 0.723** 0.699** 0.557** 0.641** (0.011) (0.012) (0.012) (0.010) IO 0.659** (0.013) NEGAM 0.939** (0.020) ARM 0.444** 0.457** 0.303** 0.418** 0.461** (0.009) (0.010) (0.013) (0.008) (0.010) Log(Inc) −0.109** −0.058** −0.060** −0.032** −0.062** (0.008) (0.010) (0.009) (0.011) (0.010) FICO −0.672** −0.663** −0.634** −0.618** −0.664** (0.010) (0.012) (0.012) (0.012) (0.012) LTV 0.506** 0.492** 0.504** 0.458** 0.492** (0.012) (0.011) (0.010) (0.010) (0.011) VTI 0.045** 0.050** 0.051** 0.061** 0.050** (0.006) (0.006) (0.006) (0.006) (0.005) Low Doc 0.046** 0.042** 0.089** 0.016 0.028* (0.012) (0.012) (0.012) (0.011) (0.012) Above limit 0.187** 0.279** 0.129** 0.294** 0.271** (0.030) (0.018) (0.018) (0.022) (0.018) Condo −0.173** −0.092** −0.083** −0.024 −0.090** (0.029) (0.025) (0.024) (0.028) (0.025) Investment 0.348** 0.332** 0.294** 0.280** 0.329** (0.026) (0.027) (0.024) (0.031) (0.027) Refinance 0.089** 0.043** 0.015 0.003 0.034** (0.011) (0.012) (0.011) (0.009) (0.012) College −0.203** −0.195** −0.198** −0.201** (0.009) (0.009) (0.008) (0.009) Young 0.017* 0.016* −0.008 0.017* (0.007) (0.007) (0.007) (0.007) Log(BEA Inc) 0.038** 0.038** 0.041* 0.037** (0.014) (0.014) (0.016) (0.014) House price growth −0.414** −0.416** −0.450** −0.415** (0.020) (0.019) (0.015) (0.020) Loan balance growth 0.016 0.018 0.061** −0.002 (0.012) (0.011) (0.009) (0.012) Payment resets 0.031** 0.029** 0.051** 0.025** (0.002) (0.002) (0.002) (0.002) Change in unemployment 0.022** 0.023** 0.026** 0.022** (0.002) (0.002) (0.002) (0.002) Income growth −0.149** −0.144** −0.134** −0.151** (0.024) (0.024) (0.038) (0.024) Gov securitized −0.164** (0.015) Priv securitized 0.246** (0.010) State-year baselines Yes Yes Yes No Yes Lender-year baselines No No No Yes No Observations 29,690,919 23,151,373 23,151,373 23,151,373 23,151,373 Table VI. Hazard model of mortgage delinquency This table reports the estimates from a Cox proportional hazard model for mortgage delinquency. The significance levels are abbreviated with asterisks: One and two asterisks denote significance at the 5% and 1% level, respectively. Individual-level MSA-level Securitization Lender-year Detailed CM Covariates Covariates Controls Baselines Contracts CM 0.723** 0.699** 0.557** 0.641** (0.011) (0.012) (0.012) (0.010) IO 0.659** (0.013) NEGAM 0.939** (0.020) ARM 0.444** 0.457** 0.303** 0.418** 0.461** (0.009) (0.010) (0.013) (0.008) (0.010) Log(Inc) −0.109** −0.058** −0.060** −0.032** −0.062** (0.008) (0.010) (0.009) (0.011) (0.010) FICO −0.672** −0.663** −0.634** −0.618** −0.664** (0.010) (0.012) (0.012) (0.012) (0.012) LTV 0.506** 0.492** 0.504** 0.458** 0.492** (0.012) (0.011) (0.010) (0.010) (0.011) VTI 0.045** 0.050** 0.051** 0.061** 0.050** (0.006) (0.006) (0.006) (0.006) (0.005) Low Doc 0.046** 0.042** 0.089** 0.016 0.028* (0.012) (0.012) (0.012) (0.011) (0.012) Above limit 0.187** 0.279** 0.129** 0.294** 0.271** (0.030) (0.018) (0.018) (0.022) (0.018) Condo −0.173** −0.092** −0.083** −0.024 −0.090** (0.029) (0.025) (0.024) (0.028) (0.025) Investment 0.348** 0.332** 0.294** 0.280** 0.329** (0.026) (0.027) (0.024) (0.031) (0.027) Refinance 0.089** 0.043** 0.015 0.003 0.034** (0.011) (0.012) (0.011) (0.009) (0.012) College −0.203** −0.195** −0.198** −0.201** (0.009) (0.009) (0.008) (0.009) Young 0.017* 0.016* −0.008 0.017* (0.007) (0.007) (0.007) (0.007) Log(BEA Inc) 0.038** 0.038** 0.041* 0.037** (0.014) (0.014) (0.016) (0.014) House price growth −0.414** −0.416** −0.450** −0.415** (0.020) (0.019) (0.015) (0.020) Loan balance growth 0.016 0.018 0.061** −0.002 (0.012) (0.011) (0.009) (0.012) Payment resets 0.031** 0.029** 0.051** 0.025** (0.002) (0.002) (0.002) (0.002) Change in unemployment 0.022** 0.023** 0.026** 0.022** (0.002) (0.002) (0.002) (0.002) Income growth −0.149** −0.144** −0.134** −0.151** (0.024) (0.024) (0.038) (0.024) Gov securitized −0.164** (0.015) Priv securitized 0.246** (0.010) State-year baselines Yes Yes Yes No Yes Lender-year baselines No No No Yes No Observations 29,690,919 23,151,373 23,151,373 23,151,373 23,151,373 Individual-level MSA-level Securitization Lender-year Detailed CM Covariates Covariates Controls Baselines Contracts CM 0.723** 0.699** 0.557** 0.641** (0.011) (0.012) (0.012) (0.010) IO 0.659** (0.013) NEGAM 0.939** (0.020) ARM 0.444** 0.457** 0.303** 0.418** 0.461** (0.009) (0.010) (0.013) (0.008) (0.010) Log(Inc) −0.109** −0.058** −0.060** −0.032** −0.062** (0.008) (0.010) (0.009) (0.011) (0.010) FICO −0.672** −0.663** −0.634** −0.618** −0.664** (0.010) (0.012) (0.012) (0.012) (0.012) LTV 0.506** 0.492** 0.504** 0.458** 0.492** (0.012) (0.011) (0.010) (0.010) (0.011) VTI 0.045** 0.050** 0.051** 0.061** 0.050** (0.006) (0.006) (0.006) (0.006) (0.005) Low Doc 0.046** 0.042** 0.089** 0.016 0.028* (0.012) (0.012) (0.012) (0.011) (0.012) Above limit 0.187** 0.279** 0.129** 0.294** 0.271** (0.030) (0.018) (0.018) (0.022) (0.018) Condo −0.173** −0.092** −0.083** −0.024 −0.090** (0.029) (0.025) (0.024) (0.028) (0.025) Investment 0.348** 0.332** 0.294** 0.280** 0.329** (0.026) (0.027) (0.024) (0.031) (0.027) Refinance 0.089** 0.043** 0.015 0.003 0.034** (0.011) (0.012) (0.011) (0.009) (0.012) College −0.203** −0.195** −0.198** −0.201** (0.009) (0.009) (0.008) (0.009) Young 0.017* 0.016* −0.008 0.017* (0.007) (0.007) (0.007) (0.007) Log(BEA Inc) 0.038** 0.038** 0.041* 0.037** (0.014) (0.014) (0.016) (0.014) House price growth −0.414** −0.416** −0.450** −0.415** (0.020) (0.019) (0.015) (0.020) Loan balance growth 0.016 0.018 0.061** −0.002 (0.012) (0.011) (0.009) (0.012) Payment resets 0.031** 0.029** 0.051** 0.025** (0.002) (0.002) (0.002) (0.002) Change in unemployment 0.022** 0.023** 0.026** 0.022** (0.002) (0.002) (0.002) (0.002) Income growth −0.149** −0.144** −0.134** −0.151** (0.024) (0.024) (0.038) (0.024) Gov securitized −0.164** (0.015) Priv securitized 0.246** (0.010) State-year baselines Yes Yes Yes No Yes Lender-year baselines No No No Yes No Observations 29,690,919 23,151,373 23,151,373 23,151,373 23,151,373 Our key finding is that CMs have significantly higher delinquency rates than FRMs in all specifications, notwithstanding a wide array of control variables. The effect is both economically and statistically significant. For example, in column 1, the coefficient of 0.723 for CM implies that the probability of delinquency for a borrower with a CM is about twice as high as for a fixed-rate borrower, holding all other characteristics fixed (⁠ e1×0.723/e0×0.723=2.06 ⁠). This impact of having a CM on mortgage delinquency is similar to a one-standard-deviation decrease in the FICO credit score, which is generally perceived to be a strong predictor of mortgage delinquency. ARMs also exhibit higher delinquency rates than FRMs. This result is primarily caused by the lower credit quality of ARM borrowers.19 The first set of additional explanatory variables in column 2 is related to shocks in cash flows. Of particular interest is the variable “Payment Resets,” defined as the increase in the minimum required mortgage payment since origination. Recall that payment resets are driven by interest rate changes for ARMs and by both interest rate and amortization changes for CMs. Consequently, CMs have larger resets than ARMs, as illustrated by the CDFs of payments over time in Figure 3. We find that payment resets increase the hazard rate of delinquency. For example, the quartile of CM borrowers with the highest resets after five years experience required average payment increases of 38% (which is 0.38/0.0567 = 6.7 standard deviations away from the mean). The estimates of the hazard model imply that such an increase in the required payment increases the delinquency rate by around 23% (⁠ e0.031×6.7−1=0.23 ⁠).20 Furthermore, we do not find a significant impact of an increase in the loan balance since origination on mortgage delinquency in the base case with state-year baselines.21 In sum, these results suggest that contract-driven resets and amortization changes partially contribute to the higher defaults of CMs. Other variables related to cash flow defaults include the income level and the FICO score, which partly reflect households’ financial conditions. Higher income and higher FICO households are less constrained and are indeed found to have lower delinquency rates. To gauge the impact of local macro-economic conditions on mortgage delinquency, we include the change in the unemployment rate in an MSA over the last year, and the income growth rate, defined as the growth rate of the mean income level at the MSA level since the mortgage was originated. Increases in local income growth rates and decreases in local unemployment rates significantly reduce mortgage delinquency. The second set of explanatory variables is related to the economic costs and benefits of default. Since households can always sell their house and pay off their mortgage in full when the remaining loan balance is low relative to the current house value, it is not surprising that higher LTV ratios at origination are associated with higher delinquency hazards. Both the LTV at origination and the subsequent change in house prices are significant drivers of mortgage delinquency. From column 2 of Table VI, a one-standard-deviation increase in LTV can increase delinquency by e0.492−1=64% and a one standard deviation drop in house prices can increase delinquency by e0.414−1=51% of the base default rate. Owners of investment properties might have lower non-monetary costs of defaulting than owners of owner-occupied homes. Indeed, our results indicate that delinquency rates of investment properties are significantly higher. Notably, the inclusion of these controls preserve the effect of contract choice, as the coefficient on the CM indicator variable remains practically unchanged. We control in column 3 for whether the mortgage was securitized by Government Sponsored Entities or by private parties. Since the impact of securitization has received significant attention in the literature, we want to ensure that the impact of CM is not subsumed by the lenders’ propensity to securitize. We find that CM are still associated with higher delinquency hazards after controlling for government and private securitization. Thus, the role of mortgage contract design is distinct from the well-documented impact of securitization. The fourth specification uses lender-year baseline hazards instead of state-year baselines, which accounts for the possibility that mortgages originated by different lenders exhibit different delinquency rates over time because individual lenders attract a particular borrower type that might focus on specific mortgage contracts. CM exhibit higher delinquency rates under all specifications. The last column of Table VI separates IO and NEGAM loans and indicates that the delinquency rates for negative amortization loans are larger than the delinquency rates for the more conservative IO loans. For example, an IO mortgage has a propensity to be delinquent that is around twice the propensity for a FRM. In contrast, a NEGAM has a propensity to default that is about 2.5 times higher than a FRM. Table VII shows that CM borrowers exhibit higher delinquency rates than borrowers of FRM for the subsamples of full documentation loans, for subprime loans, for purchase transactions, for investment properties, for non-securitized loans, and for loans not originated in California. Table VII. Hazard model of mortgage delinquency for subsamples This table reports the estimates from a Cox proportional hazard model for mortgage delinquency for the following subsamples: loans with full documentation; loans to borrowers with FICO scores below 620; loans originated to purchase a new house; loans used to finance an investment property; non-securitized loans; and loans originated in states other than California. The significance levels are abbreviated with asterisks: One and two asterisks denote significance at the 5% and 1% level, respectively. Full Subprime Purchases Invest Not Exclude Document Only Only Property Securitized California CM 0.561** 0.466** 0.779** 0.658** 0.368** 0.689** (0.012) (0.015) (0.015) (0.016) (0.022) (0.015) ARM 0.454** 0.348** 0.505** 0.342** 0.097** 0.434** (0.011) (0.008) (0.011) (0.017) (0.019) (0.009) Log(Inc) −0.112** −0.083** −0.085** 0.035** −0.042** −0.072** (0.010) (0.007) (0.010) (0.008) (0.009) (0.010) FICO −0.712** −0.339** −0.679** −0.643** −0.705** −0.683** (0.012) (0.007) (0.012) (0.015) (0.016) (0.012) LTV 0.467** 0.293** 0.433** 0.621** 0.483** 0.463** (0.012) (0.008) (0.008) (0.012) (0.012) (0.009) VTI 0.050** 0.086** 0.051** 0.069** 0.050** 0.078** (0.006) (0.007) (0.009) (0.011) (0.008) (0.007) Low Doc −0.021** 0.076** −0.078** 0.243** 0.042** (0.008) (0.014) (0.012) (0.017) (0.011) Above limit 0.259** 0.250** 0.236** 0.057* 0.331** 0.326** (0.019) (0.021) (0.023) (0.022) (0.023) (0.024) Condo −0.081** −0.071** −0.103** −0.172** −0.132** −0.105** (0.025) (0.016) (0.027) (0.028) (0.032) (0.031) Investment 0.331** 0.154** 0.281** 0.376** 0.407** (0.024) (0.015) (0.031) (0.034) (0.024) Refinance 0.018 −0.031** 0.233** 0.043* 0.064** (0.012) (0.011) (0.011) (0.019) (0.012) College −0.177** −0.091** −0.241** −0.216** −0.217** −0.186** (0.008) (0.004) (0.011) (0.011) (0.011) (0.008) Young 0.012 −0.006 0.026** 0.055** 0.002 0.013 (0.007) (0.005) (0.009) (0.010) (0.009) (0.008) Log(BEA Inc) 0.043** 0.035** 0.047** 0.025 0.007 0.067** (0.014) (0.007) (0.015) (0.015) (0.016) (0.014) House price growth −0.420** −0.353** −0.425** −0.380** −0.399** −0.416** (0.018) (0.019) (0.023) (0.022) (0.025) (0.027) Loan Balance Growth −0.043** −0.046** −0.034** 0.119** 0.026 0.036* (0.011) (0.009) (0.011) (0.013) (0.019) (0.015) Payment Resets 0.050** 0.027** 0.046** 0.070** 0.004 0.033** (0.002) (0.003) (0.003) (0.003) (0.004) (0.002) Change in Unemployment 0.023** 0.029** 0.023** 0.015** 0.015** 0.025** (0.002) (0.003) (0.002) (0.003) (0.004) (0.002) Income Growth −0.141** −0.059** −0.155** −0.146** −0.154** −0.127** (0.023) (0.020) (0.025) (0.033) (0.034) (0.022) Observations 8,595,153 1,751,099 13,672,439 2,259,834 2,043,822 19,681,780 Full Subprime Purchases Invest Not Exclude Document Only Only Property Securitized California CM 0.561** 0.466** 0.779** 0.658** 0.368** 0.689** (0.012) (0.015) (0.015) (0.016) (0.022) (0.015) ARM 0.454** 0.348** 0.505** 0.342** 0.097** 0.434** (0.011) (0.008) (0.011) (0.017) (0.019) (0.009) Log(Inc) −0.112** −0.083** −0.085** 0.035** −0.042** −0.072** (0.010) (0.007) (0.010) (0.008) (0.009) (0.010) FICO −0.712** −0.339** −0.679** −0.643** −0.705** −0.683** (0.012) (0.007) (0.012) (0.015) (0.016) (0.012) LTV 0.467** 0.293** 0.433** 0.621** 0.483** 0.463** (0.012) (0.008) (0.008) (0.012) (0.012) (0.009) VTI 0.050** 0.086** 0.051** 0.069** 0.050** 0.078** (0.006) (0.007) (0.009) (0.011) (0.008) (0.007) Low Doc −0.021** 0.076** −0.078** 0.243** 0.042** (0.008) (0.014) (0.012) (0.017) (0.011) Above limit 0.259** 0.250** 0.236** 0.057* 0.331** 0.326** (0.019) (0.021) (0.023) (0.022) (0.023) (0.024) Condo −0.081** −0.071** −0.103** −0.172** −0.132** −0.105** (0.025) (0.016) (0.027) (0.028) (0.032) (0.031) Investment 0.331** 0.154** 0.281** 0.376** 0.407** (0.024) (0.015) (0.031) (0.034) (0.024) Refinance 0.018 −0.031** 0.233** 0.043* 0.064** (0.012) (0.011) (0.011) (0.019) (0.012) College −0.177** −0.091** −0.241** −0.216** −0.217** −0.186** (0.008) (0.004) (0.011) (0.011) (0.011) (0.008) Young 0.012 −0.006 0.026** 0.055** 0.002 0.013 (0.007) (0.005) (0.009) (0.010) (0.009) (0.008) Log(BEA Inc) 0.043** 0.035** 0.047** 0.025 0.007 0.067** (0.014) (0.007) (0.015) (0.015) (0.016) (0.014) House price growth −0.420** −0.353** −0.425** −0.380** −0.399** −0.416** (0.018) (0.019) (0.023) (0.022) (0.025) (0.027) Loan Balance Growth −0.043** −0.046** −0.034** 0.119** 0.026 0.036* (0.011) (0.009) (0.011) (0.013) (0.019) (0.015) Payment Resets 0.050** 0.027** 0.046** 0.070** 0.004 0.033** (0.002) (0.003) (0.003) (0.003) (0.004) (0.002) Change in Unemployment 0.023** 0.029** 0.023** 0.015** 0.015** 0.025** (0.002) (0.003) (0.002) (0.003) (0.004) (0.002) Income Growth −0.141** −0.059** −0.155** −0.146** −0.154** −0.127** (0.023) (0.020) (0.025) (0.033) (0.034) (0.022) Observations 8,595,153 1,751,099 13,672,439 2,259,834 2,043,822 19,681,780 Table VII. Hazard model of mortgage delinquency for subsamples This table reports the estimates from a Cox proportional hazard model for mortgage delinquency for the following subsamples: loans with full documentation; loans to borrowers with FICO scores below 620; loans originated to purchase a new house; loans used to finance an investment property; non-securitized loans; and loans originated in states other than California. The significance levels are abbreviated with asterisks: One and two asterisks denote significance at the 5% and 1% level, respectively. Full Subprime Purchases Invest Not Exclude Document Only Only Property Securitized California CM 0.561** 0.466** 0.779** 0.658** 0.368** 0.689** (0.012) (0.015) (0.015) (0.016) (0.022) (0.015) ARM 0.454** 0.348** 0.505** 0.342** 0.097** 0.434** (0.011) (0.008) (0.011) (0.017) (0.019) (0.009) Log(Inc) −0.112** −0.083** −0.085** 0.035** −0.042** −0.072** (0.010) (0.007) (0.010) (0.008) (0.009) (0.010) FICO −0.712** −0.339** −0.679** −0.643** −0.705** −0.683** (0.012) (0.007) (0.012) (0.015) (0.016) (0.012) LTV 0.467** 0.293** 0.433** 0.621** 0.483** 0.463** (0.012) (0.008) (0.008) (0.012) (0.012) (0.009) VTI 0.050** 0.086** 0.051** 0.069** 0.050** 0.078** (0.006) (0.007) (0.009) (0.011) (0.008) (0.007) Low Doc −0.021** 0.076** −0.078** 0.243** 0.042** (0.008) (0.014) (0.012) (0.017) (0.011) Above limit 0.259** 0.250** 0.236** 0.057* 0.331** 0.326** (0.019) (0.021) (0.023) (0.022) (0.023) (0.024) Condo −0.081** −0.071** −0.103** −0.172** −0.132** −0.105** (0.025) (0.016) (0.027) (0.028) (0.032) (0.031) Investment 0.331** 0.154** 0.281** 0.376** 0.407** (0.024) (0.015) (0.031) (0.034) (0.024) Refinance 0.018 −0.031** 0.233** 0.043* 0.064** (0.012) (0.011) (0.011) (0.019) (0.012) College −0.177** −0.091** −0.241** −0.216** −0.217** −0.186** (0.008) (0.004) (0.011) (0.011) (0.011) (0.008) Young 0.012 −0.006 0.026** 0.055** 0.002 0.013 (0.007) (0.005) (0.009) (0.010) (0.009) (0.008) Log(BEA Inc) 0.043** 0.035** 0.047** 0.025 0.007 0.067** (0.014) (0.007) (0.015) (0.015) (0.016) (0.014) House price growth −0.420** −0.353** −0.425** −0.380** −0.399** −0.416** (0.018) (0.019) (0.023) (0.022) (0.025) (0.027) Loan Balance Growth −0.043** −0.046** −0.034** 0.119** 0.026 0.036* (0.011) (0.009) (0.011) (0.013) (0.019) (0.015) Payment Resets 0.050** 0.027** 0.046** 0.070** 0.004 0.033** (0.002) (0.003) (0.003) (0.003) (0.004) (0.002) Change in Unemployment 0.023** 0.029** 0.023** 0.015** 0.015** 0.025** (0.002) (0.003) (0.002) (0.003) (0.004) (0.002) Income Growth −0.141** −0.059** −0.155** −0.146** −0.154** −0.127** (0.023) (0.020) (0.025) (0.033) (0.034) (0.022) Observations 8,595,153 1,751,099 13,672,439 2,259,834 2,043,822 19,681,780 Full Subprime Purchases Invest Not Exclude Document Only Only Property Securitized California CM 0.561** 0.466** 0.779** 0.658** 0.368** 0.689** (0.012) (0.015) (0.015) (0.016) (0.022) (0.015) ARM 0.454** 0.348** 0.505** 0.342** 0.097** 0.434** (0.011) (0.008) (0.011) (0.017) (0.019) (0.009) Log(Inc) −0.112** −0.083** −0.085** 0.035** −0.042** −0.072** (0.010) (0.007) (0.010) (0.008) (0.009) (0.010) FICO −0.712** −0.339** −0.679** −0.643** −0.705** −0.683** (0.012) (0.007) (0.012) (0.015) (0.016) (0.012) LTV 0.467** 0.293** 0.433** 0.621** 0.483** 0.463** (0.012) (0.008) (0.008) (0.012) (0.012) (0.009) VTI 0.050** 0.086** 0.051** 0.069** 0.050** 0.078** (0.006) (0.007) (0.009) (0.011) (0.008) (0.007) Low Doc −0.021** 0.076** −0.078** 0.243** 0.042** (0.008) (0.014) (0.012) (0.017) (0.011) Above limit 0.259** 0.250** 0.236** 0.057* 0.331** 0.326** (0.019) (0.021) (0.023) (0.022) (0.023) (0.024) Condo −0.081** −0.071** −0.103** −0.172** −0.132** −0.105** (0.025) (0.016) (0.027) (0.028) (0.032) (0.031) Investment 0.331** 0.154** 0.281** 0.376** 0.407** (0.024) (0.015) (0.031) (0.034) (0.024) Refinance 0.018 −0.031** 0.233** 0.043* 0.064** (0.012) (0.011) (0.011) (0.019) (0.012) College −0.177** −0.091** −0.241** −0.216** −0.217** −0.186** (0.008) (0.004) (0.011) (0.011) (0.011) (0.008) Young 0.012 −0.006 0.026** 0.055** 0.002 0.013 (0.007) (0.005) (0.009) (0.010) (0.009) (0.008) Log(BEA Inc) 0.043** 0.035** 0.047** 0.025 0.007 0.067** (0.014) (0.007) (0.015) (0.015) (0.016) (0.014) House price growth −0.420** −0.353** −0.425** −0.380** −0.399** −0.416** (0.018) (0.019) (0.023) (0.022) (0.025) (0.027) Loan Balance Growth −0.043** −0.046** −0.034** 0.119** 0.026 0.036* (0.011) (0.009) (0.011) (0.013) (0.019) (0.015) Payment Resets 0.050** 0.027** 0.046** 0.070** 0.004 0.033** (0.002) (0.003) (0.003) (0.003) (0.004) (0.002) Change in Unemployment 0.023** 0.029** 0.023** 0.015** 0.015** 0.025** (0.002) (0.003) (0.002) (0.003) (0.004) (0.002) Income Growth −0.141** −0.059** −0.155** −0.146** −0.154** −0.127** (0.023) (0.020) (0.025) (0.033) (0.034) (0.022) Observations 8,595,153 1,751,099 13,672,439 2,259,834 2,043,822 19,681,780 In addition, we also run the hazard models separately for each annual origination cohort. The coefficients on CM are significantly positive for each individual origination cohort between 2003 and 2007. Furthermore, the remaining coefficients are generally consistent over the different cohorts. Overall, we find significantly higher default rates among CM borrowers, even after controlling for leverage, payment resets, and other household and loan characteristics. These results indicate that CM borrowers differ from other households in their risk taking and their willingness to default. 4.3 Interaction Effects To study the motivations for delinquency by CM borrowers, we conduct two additional tests. Our first test investigates whether CM borrowers are more sensitive to measures that capture the economic benefits of the default option. Defaulting is generally only beneficial when the loan value exceeds the home value by a sufficient amount to cover various costs of default (e.g., decline in credit rating, moving costs). Furthermore, the economic benefit of default increases with the LTV ratio, as the moneyness of the option to default increases. Table VIII shows that the delinquency rate of complex borrowers is particularly sensitive to the LTV ratio as evidenced by a strong positive coefficient on the interaction term between CM and LTV. The coefficient of 0.109 for the interaction term in column 1 suggests that the deterioration (i.e., increase) in the hazard rate from a one standard deviation increase in the LTV ratio is about 10.9% larger for a CM mortgage than for an FRM mortgage.22 Table VIII. Hazard model of mortgage delinquency with interaction effects This table reports the estimates from a Cox proportional hazard model for mortgage delinquency, with interaction effects that capture the sensitivity of complex mortgage delinquencies to other loan and household characteristics. The table also includes unreported individual- and MSA-level covariates as in the second column of Table VI. The significance levels are abbreviated with asterisks: One and two asterisks denote significance at the 5% and 1% level, respectively. Interactions with CM Interactions with IO/NEGAM CM 0.697** (0.013) IO 0.661** (0.014) NEGAM 0.914** (0.020) LTV 0.472** 0.470** (0.010) (0.010) CM x LTV 0.109** (0.020) IO × LTV 0.087** (0.020) NEGAM × LTV 0.232** (0.022) Log(Inc) −0.078** −0.078** (0.010) (0.010) CM x Log(Inc) 0.083** (0.009) IO × Log(Inc) 0.061** (0.009) NEGAM × Log(Inc) 0.115** (0.011) FICO −0.680** −0.681** (0.010) (0.010) CM × FICO 0.078** (0.012) IO × FICO 0.057** (0.013) NEGAM × FICO 0.211** (0.009) ARM 0.453** 0.455** (0.010) (0.010) State-year baselines Yes Yes Additional individual and MSA covariates Yes Yes Observations 23, 151, 373 23, 151, 373 Interactions with CM Interactions with IO/NEGAM CM 0.697** (0.013) IO 0.661** (0.014) NEGAM 0.914** (0.020) LTV 0.472** 0.470** (0.010) (0.010) CM x LTV 0.109** (0.020) IO × LTV 0.087** (0.020) NEGAM × LTV 0.232** (0.022) Log(Inc) −0.078** −0.078** (0.010) (0.010) CM x Log(Inc) 0.083** (0.009) IO × Log(Inc) 0.061** (0.009) NEGAM × Log(Inc) 0.115** (0.011) FICO −0.680** −0.681** (0.010) (0.010) CM × FICO 0.078** (0.012) IO × FICO 0.057** (0.013) NEGAM × FICO 0.211** (0.009) ARM 0.453** 0.455** (0.010) (0.010) State-year baselines Yes Yes Additional individual and MSA covariates Yes Yes Observations 23, 151, 373 23, 151, 373 Table VIII. Hazard model of mortgage delinquency with interaction effects This table reports the estimates from a Cox proportional hazard model for mortgage delinquency, with interaction effects that capture the sensitivity of complex mortgage delinquencies to other loan and household characteristics. The table also includes unreported individual- and MSA-level covariates as in the second column of Table VI. The significance levels are abbreviated with asterisks: One and two asterisks denote significance at the 5% and 1% level, respectively. Interactions with CM Interactions with IO/NEGAM CM 0.697** (0.013) IO 0.661** (0.014) NEGAM 0.914** (0.020) LTV 0.472** 0.470** (0.010) (0.010) CM x LTV 0.109** (0.020) IO × LTV 0.087** (0.020) NEGAM × LTV 0.232** (0.022) Log(Inc) −0.078** −0.078** (0.010) (0.010) CM x Log(Inc) 0.083** (0.009) IO × Log(Inc) 0.061** (0.009) NEGAM × Log(Inc) 0.115** (0.011) FICO −0.680** −0.681** (0.010) (0.010) CM × FICO 0.078** (0.012) IO × FICO 0.057** (0.013) NEGAM × FICO 0.211** (0.009) ARM 0.453** 0.455** (0.010) (0.010) State-year baselines Yes Yes Additional individual and MSA covariates Yes Yes Observations 23, 151, 373 23, 151, 373 Interactions with CM Interactions with IO/NEGAM CM 0.697** (0.013) IO 0.661** (0.014) NEGAM 0.914** (0.020) LTV 0.472** 0.470** (0.010) (0.010) CM x LTV 0.109** (0.020) IO × LTV 0.087** (0.020) NEGAM × LTV 0.232** (0.022) Log(Inc) −0.078** −0.078** (0.010) (0.010) CM x Log(Inc) 0.083** (0.009) IO × Log(Inc) 0.061** (0.009) NEGAM × Log(Inc) 0.115** (0.011) FICO −0.680** −0.681** (0.010) (0.010) CM × FICO 0.078** (0.012) IO × FICO 0.057** (0.013) NEGAM × FICO 0.211** (0.009) ARM 0.453** 0.455** (0.010) (0.010) State-year baselines Yes Yes Additional individual and MSA covariates Yes Yes Observations 23, 151, 373 23, 151, 373 Our second test compares the delinquency rates of CM borrowers with different levels of financial sophistication. We use the income level and the FICO score as proxies of financial sophistication following Campbell (2006) and Bucks and Pence (2008). Table VIII shows that while CM borrowers on average default more than traditional mortgage borrowers, the difference in the delinquency rates between complex and traditional borrowers is particularly high for households with higher income levels and with higher FICO credit scores. The last column of Table VIII shows that these interaction effects are more pronounced for NEGAM than for IO contracts, as should be expected given the fact that NEGAM are more complex instruments than IO mortgages. 4.4 Personal Bankruptcy To further investigate the characteristics of complex borrowers, we investigate the determinants of personal bankruptcy for mortgage borrowers. Contrasting the determinants of personal bankruptcy with the determinants of mortgage delinquency gives us important insights into the motivation of the delinquency behavior. It is not necessary that households who default on their mortgages also declare bankruptcy. Nor is it necessary that households who declare bankruptcy default on their mortgages. For example, in our sample only 12% of households who are delinquent on their mortgage also declare bankruptcy.23 Table IX reports the estimates of the bankruptcy hazard model. CM borrowers, and especially NEGAM borrowers, have higher propensities to declare bankruptcy, indicating that these households are more financially distressed than traditional mortgage borrowers. Most coefficients have the same signs as in the delinquency regression of Table VI, since both bankruptcy and mortgage delinquency are distress events for households and should be driven by similar fundamentals. For example, higher income and higher FICO scores reduce the propensities of both mortgage delinquency and bankruptcy. It is interesting that some variables show up with different signs in the two specifications. For example, although households with investment properties have significantly higher mortgage delinquency rates, they are not more likely to file for personal bankruptcy. This evidence suggests that owners of investment properties might be more likely to walk away from the property when it is economical to do so. Similarly, loans with low documentation are typically more likely to be delinquent on their mortgage debt, but these households do not experience higher bankruptcy rates. Table IX. Hazard models of personal bankruptcy This table reports the estimates from a Cox proportional hazard model for personal bankruptcy. The significance levels are abbreviated with asterisks: One and two asterisks denote significance at the 5% and 1% level, respectively. Individual-level MSA-level Individual-level MSA-level covariates covariates covariates covariates CM 0.667** 0.619** (0.011) (0.013) IO 0.602** 0.581** (0.012) (0.014) NEGAM 0.956** 0.818** (0.027) (0.026) ARM 0.389** 0.405** 0.391** 0.409** (0.014) (0.017) (0.014) (0.017) Log(Inc) −0.156** −0.097** −0.159** −0.099** (0.010) (0.012) (0.010) (0.012) FICO −0.466** −0.462** −0.467** −0.463** (0.008) (0.009) (0.008) (0.009) LTV 0.583** 0.551** 0.579** 0.551** (0.011) (0.013) (0.011) (0.013) VTI −0.213** −0.183** −0.211** −0.181** (0.017) (0.017) (0.017) (0.017) Low Doc 0.000 −0.009 −0.017* −0.020* (0.008) (0.009) (0.008) (0.009) Above limit 0.174** 0.276** 0.162** 0.267** (0.026) (0.029) (0.027) (0.029) Condo −0.304** −0.162** −0.300** −0.160** (0.027) (0.024) (0.027) (0.024) Investment 0.018 −0.010 0.015 −0.011 (0.024) (0.023) (0.024) (0.023) Refinance 0.405** 0.363** 0.392** 0.355** (0.012) (0.013) (0.012) (0.013) College −0.201** −0.200** (0.009) (0.009) Young −0.063** −0.063** (0.010) (0.010) Log(BEA Inc) −0.018 −0.019 (0.017) (0.017) House Price Growth −0.329** −0.328** (0.022) (0.022) Loan balance growth 0.093** 0.082** (0.008) (0.011) Payment resets 0.039** 0.034** (0.002) (0.002) Change in unemployment 0.021** 0.021** (0.003) (0.003) Income growth −0.179** −0.181** (0.026) (0.027) Observations 30,920,127 23,978,775 30,920,127 23,978,775 Individual-level MSA-level Individual-level MSA-level covariates covariates covariates covariates CM 0.667** 0.619** (0.011) (0.013) IO 0.602** 0.581** (0.012) (0.014) NEGAM 0.956** 0.818** (0.027) (0.026) ARM 0.389** 0.405** 0.391** 0.409** (0.014) (0.017) (0.014) (0.017) Log(Inc) −0.156** −0.097** −0.159** −0.099** (0.010) (0.012) (0.010) (0.012) FICO −0.466** −0.462** −0.467** −0.463** (0.008) (0.009) (0.008) (0.009) LTV 0.583** 0.551** 0.579** 0.551** (0.011) (0.013) (0.011) (0.013) VTI −0.213** −0.183** −0.211** −0.181** (0.017) (0.017) (0.017) (0.017) Low Doc 0.000 −0.009 −0.017* −0.020* (0.008) (0.009) (0.008) (0.009) Above limit 0.174** 0.276** 0.162** 0.267** (0.026) (0.029) (0.027) (0.029) Condo −0.304** −0.162** −0.300** −0.160** (0.027) (0.024) (0.027) (0.024) Investment 0.018 −0.010 0.015 −0.011 (0.024) (0.023) (0.024) (0.023) Refinance 0.405** 0.363** 0.392** 0.355** (0.012) (0.013) (0.012) (0.013) College −0.201** −0.200** (0.009) (0.009) Young −0.063** −0.063** (0.010) (0.010) Log(BEA Inc) −0.018 −0.019 (0.017) (0.017) House Price Growth −0.329** −0.328** (0.022) (0.022) Loan balance growth 0.093** 0.082** (0.008) (0.011) Payment resets 0.039** 0.034** (0.002) (0.002) Change in unemployment 0.021** 0.021** (0.003) (0.003) Income growth −0.179** −0.181** (0.026) (0.027) Observations 30,920,127 23,978,775 30,920,127 23,978,775 Table IX. Hazard models of personal bankruptcy This table reports the estimates from a Cox proportional hazard model for personal bankruptcy. The significance levels are abbreviated with asterisks: One and two asterisks denote significance at the 5% and 1% level, respectively. Individual-level MSA-level Individual-level MSA-level covariates covariates covariates covariates CM 0.667** 0.619** (0.011) (0.013) IO 0.602** 0.581** (0.012) (0.014) NEGAM 0.956** 0.818** (0.027) (0.026) ARM 0.389** 0.405** 0.391** 0.409** (0.014) (0.017) (0.014) (0.017) Log(Inc) −0.156** −0.097** −0.159** −0.099** (0.010) (0.012) (0.010) (0.012) FICO −0.466** −0.462** −0.467** −0.463** (0.008) (0.009) (0.008) (0.009) LTV 0.583** 0.551** 0.579** 0.551** (0.011) (0.013) (0.011) (0.013) VTI −0.213** −0.183** −0.211** −0.181** (0.017) (0.017) (0.017) (0.017) Low Doc 0.000 −0.009 −0.017* −0.020* (0.008) (0.009) (0.008) (0.009) Above limit 0.174** 0.276** 0.162** 0.267** (0.026) (0.029) (0.027) (0.029) Condo −0.304** −0.162** −0.300** −0.160** (0.027) (0.024) (0.027) (0.024) Investment 0.018 −0.010 0.015 −0.011 (0.024) (0.023) (0.024) (0.023) Refinance 0.405** 0.363** 0.392** 0.355** (0.012) (0.013) (0.012) (0.013) College −0.201** −0.200** (0.009) (0.009) Young −0.063** −0.063** (0.010) (0.010) Log(BEA Inc) −0.018 −0.019 (0.017) (0.017) House Price Growth −0.329** −0.328** (0.022) (0.022) Loan balance growth 0.093** 0.082** (0.008) (0.011) Payment resets 0.039** 0.034** (0.002) (0.002) Change in unemployment 0.021** 0.021** (0.003) (0.003) Income growth −0.179** −0.181** (0.026) (0.027) Observations 30,920,127 23,978,775 30,920,127 23,978,775 Individual-level MSA-level Individual-level MSA-level covariates covariates covariates covariates CM 0.667** 0.619** (0.011) (0.013) IO 0.602** 0.581** (0.012) (0.014) NEGAM 0.956** 0.818** (0.027) (0.026) ARM 0.389** 0.405** 0.391** 0.409** (0.014) (0.017) (0.014) (0.017) Log(Inc) −0.156** −0.097** −0.159** −0.099** (0.010) (0.012) (0.010) (0.012) FICO −0.466** −0.462** −0.467** −0.463** (0.008) (0.009) (0.008) (0.009) LTV 0.583** 0.551** 0.579** 0.551** (0.011) (0.013) (0.011) (0.013) VTI −0.213** −0.183** −0.211** −0.181** (0.017) (0.017) (0.017) (0.017) Low Doc 0.000 −0.009 −0.017* −0.020* (0.008) (0.009) (0.008) (0.009) Above limit 0.174** 0.276** 0.162** 0.267** (0.026) (0.029) (0.027) (0.029) Condo −0.304** −0.162** −0.300** −0.160** (0.027) (0.024) (0.027) (0.024) Investment 0.018 −0.010 0.015 −0.011 (0.024) (0.023) (0.024) (0.023) Refinance 0.405** 0.363** 0.392** 0.355** (0.012) (0.013) (0.012) (0.013) College −0.201** −0.200** (0.009) (0.009) Young −0.063** −0.063** (0.010) (0.010) Log(BEA Inc) −0.018 −0.019 (0.017) (0.017) House Price Growth −0.329** −0.328** (0.022) (0.022) Loan balance growth 0.093** 0.082** (0.008) (0.011) Payment resets 0.039** 0.034** (0.002) (0.002) Change in unemployment 0.021** 0.021** (0.003) (0.003) Income growth −0.179** −0.181** (0.026) (0.027) Observations 30,920,127 23,978,775 30,920,127 23,978,775 5. Conclusions The recent housing crisis brought the extension of credit to subprime borrowers and agency problems inherent in mortgage securitization to the forefront of academic research. This article focuses on a different aspect of credit markets during this time—namely, the proliferation of non-amortizing mortgages. In addition to variable interest rates, such mortgages also feature changes in amortization schedules set off by a variety of triggers. These CM contracts became very popular during the mid-2000s and vanished almost completely after the housing crisis of 2007–2009. We find that CM are the contract of choice for households with high credit quality and high incomes, in contrast to the low-income and low-credit-score population targeted by subprime mortgages. Households use CM as affordability products to purchase houses that are expensive relative to their current incomes, partly due to their expectations of higher future income and house price growth. CM borrowers are more likely to provide incomplete documentation for their loans, to be owners of investment properties, and to reside in non-recourse states in which lenders do not have access to non-collateralized assets in the event of mortgage delinquency. Consistent with the notion that households who self-select into CM products are fundamentally different from traditional mortgage borrowers, we find that CM experienced substantially higher propensities of mortgage defaults and personal bankruptcies, controlling for contract features, a variety of borrower and loan characteristics, as well as macroeconomic shocks. Higher delinquency rates cannot be attributed solely to greater leverage of CM and the onset of amortization resets brought about by inability to refinance CM. Furthermore, the difference in the delinquency rates between complex and traditional borrowers increases with measures capturing the benefit of defaulting (like the LTV ratio) and measures of financial sophistication (like income or credit scores). Overall, both the characteristics of CM borrowers and their default behavior cast doubt on the popular perception that CM are pushed by predatory lenders to naïve households. Our findings suggest instead that CM are taken out by relatively sophisticated borrowers than borrowers with more traditional mortgage contracts. Our results also highlight the contribution of mortgage market innovations to the surge in defaults among prime credit score borrowers during the financial crisis. Footnotes * We thank the editor Amiyatosh Purnanandam, two anonymous reviewers, and Ethan Cohen-Cole, Yongheng Deng, Serdar Dinc, Andra Ghent, Craig Furfine, Stuart Gabriel, Wei Jiang, Pete Kyle, Debbie Lucas, Jaehoon Hahn, Lu Han, Jay Hartzell, Jingzhi Huang, Jeongmin Lee, Robert McDonald, Justin Murfin, Tomasz Piskorski, Wenlan Qian, Oleg Rytchkov, Amit Seru, John Shoven, Laura Starks, Amir Sufi, Sheridan Titman, Nancy Wallace, Michelle White, and seminar participants at the American Finance Association in Chicago, the China International Conference in Finance in Shanghai, the European Finance Association in Stockholm, the Financial Economics and Accounting Conference at the University of Maryland, the Korea America Finance Association International Conference in Seoul, the ShovenFest at Stanford University, the Society of Financial Studies Cavalcade at the University of Michigan, the Swiss Economists Abroad Conference in Zurich, Brigham Young University, the Federal Reserve Bank of Chicago, the Federal Reserve Bank of Dallas, the Hong Kong University of Science and Technology, Korea University, Nanyang Technological University of Singapore, the National University of Singapore, New York University, Renmin University, Rutgers University, the Shanghai Advanced Institute of Finance, the Singapore Management University, Tsinghua University, the University of California at Los Angeles, the University of California at San Diego, the University of Lausanne, the University of Texas at Austin, the University of Zurich, and Vanderbilt University for helpful comments and suggestions. Clemens Sialm is an independent contractor at AQR Capital Management and thanks the Stanford Institute for Economic Policy Research for financial support during his Sabbatical leave. 1 This characterization of complex mortgages also corresponds to a frequent portrayal of complex mortgages in the media. See, for example, the New York Times article, How Countrywide Covered the Cracks, by Gretchen Morgenson, October 16, 2010, at http://www.nytimes.com/2010/10/17/business/17trial.html. 2 See, for instance, Mian and Sufi (2009), Keys et al. (2010), and Jiang et al. (2014a), among others. 3 It should be noted that optimality of such contracts from the viewpoint of an individual household does not imply social optimality if these contracts are associated with higher default probabilities and if defaults generate negative externalities for neighboring properties. The latter effect is shown in Campbell et al. (2011), Mian et al. (2015), and Guren and McQuade (2018). 4 Additional papers on securitization and the expansion of credit include Bond et al. (2009), Keys et al. (2009), Loutskina and Strahan (2009), Mayer et al. (2009), Gerardi et al. (2010), Piskorski et al. (2010), An et al. (2011), Campbell et al. (2011), Demyanyk and Hemert (2011), Li et al. (2011), Agarwal et al. (2012), Goetzmann et al. (2012), Keys et al. (2012), Woodward and Hall (2012), Adelino et al. (2013), Agarwal et al. (2014), Jiang et al. (2014b), Rajan et al. (2015), Begley and Purnanandam (2017), and Melzer (2017). 5 Alt-A loans are a middle category of loans, more risky than prime and less risky than subprime. They are generally made to borrowers with good credit scores, but the loans have characteristics that make them ineligible to be sold to the GSEs (e.g., limited documentation of the income or assets of the borrower or higher loan-to-value ratios than those specified by GSE limits). 6 We exclude observations with missing home values, missing loan amounts, and with appraisal amounts above $10 million. We also exclude loans affected by the Hurricane Katrina due to unreliable home price index data. These restrictions reduce the number of observations by only 1.7%. 7 Ghent and Kudlyak (2011) classify the following states as non-recourse: Alaska, Arizona, California, Iowa, Minnesota, Montana, North Dakota, Oregon, Washington, and Wisconsin. There is some ambiguity with respect to the recourse status of California loans. Refinance loans in California are subject to recourse only if the lender chooses to pursue judicial foreclosure. Although we observe whether a loan is used for new purchase or refinancing, we cannot assess the credibility of the threat of lender recourse through judicial foreclosure. In this article, only new purchase loans in California are defined as non-recourse. The results are robust to categorizing all California loans as non-recourse. 8 Several papers study the tradeoff between FRMs and ARMs (e.g., Campbell and Cocco, 2003; Koijen et al., 2009; Fuster and Vickery, 2015). 9 Given the near disappearance of CM contracts following the onset of the crisis, we limit attention to pre-2008 originations. Each mortgage is tracked for a minimum of 24 months (end of 2009) to ensure sufficient data for hazard model estimation. 10 The income distribution for CM borrowers lies to the right of the distribution of borrowers with fully amortizing ARM and FRM contracts. This also holds true when loans underwritten on the basis of stated (undocumented) income are dropped from the sample. 11 LPS data is collected at the loan and not property level, which limits one’s ability to construct an accurate estimate of the total debt secured by the house. In particular, we are unable to account for second-lien mortgages loans (the so-called “piggyback loans”). 12 A small fraction of FRM loans show a change in recorded payments. These differences are largely due to small fluctuations in taxes and private mortgage insurance payments. 13 The higher long-term loan-to-value ratios of complex mortgages may have contributed to a further deterioration in housing markets, as suggested by the leverage effect of Stein (1995) and Lamont and Stein (1999). Additional papers that study the macro-economic aspects of housing prices include Lustig and Van Nieuwerburgh (2005), Ortalo-Magne and Rady (2006), Piazzesi et al. (2007), Brunnermeier and Julliard (2008), Van Nieuwerburgh and Weill (2010), Landvoigt et al. (2015), and Favilukis et al. (2017). 14 The coefficient on the FICO score variable is significantly positive for CM if we select ARMs as the baseline group or if we run a simple logit regression. 15 Our main results are not affected materially if we include state-fixed effects instead of the MSA-fixed effects. 16 In unreported results, we find that the results are not affected qualitatively if we include lender-year fixed effects to capture time-variation in lender behavior. 17 About half of our observations have a missing “Low Documentation” variable. Our base-case results in Table III include these households, setting the “Low Documentation” value to zero. Table V includes only the households for which we know explicitly that they submitted fully documented loan applications. 18 The results are not affected qualitatively if we use a common baseline hazard, origination year-specific baselines, or origination year and state-specific baselines. We also estimate hazard models where the baseline hazard depends on the mortgage type. The estimates on the control variables are consistent with the base-case results. However, we do not show the results with mortgage type-specific baselines because this specification does not allow us to concisely report the impact of the mortgage type. 19 If we split up subprime and non-subprime ARMs, then the hazard estimates on subprime ARMs are nearly four times larger than the coefficients on prime ARMs. 20 A recent paper by Fuster and Willen (2017) finds a substantial impact of downward payment resets on subsequent loan performance. Their sample focuses on hybrid ARM contracts with interest rate resets after 3, 5, 7, or 10 years and a 10-year interest-only period. Since the loans in their sample are originated in 2005–2006, the resulting rate resets 5 years later cut the payments by 50%, on average. 21 Note that the LTV is measured at the time of origination and does not capture changes in leverage since origination. The changes in the LTV are driven by changes in house prices (which significantly affect delinquencies) and by changes in loan balances (which have mixed effects on delinquencies). 22 The interpretation of interaction effects in non-linear models is subject to the well-known critique of Ai and Norton (2003). However, we make use of the specific functional form of the Cox proportional hazard model to argue that the reported coefficients have a direct and natural interpretation. To see this, let’s consider the example of the interaction term between the FICO score and the CM indicator. Taking logs of the hazard function and then differentiating with respect to FICO yields ∂log h(i,t)/∂FICO=βFICO+γ×CM ⁠. Since CM is a binary variable, γ shows the difference in relative changes in the hazard function in response to changes in the FICO score for different types of mortgages. 23 See Li, White, and Zhu (2011) for an insightful discussion of the relationship between bankruptcy laws and mortgage defaults. References Adelino M. , Gerardi K. , Willen P. ( 2013 ): Why don’t lenders renegotiate more home mortgages? Redefaults, self-cures, and securitization , Journal of Monetary Economics 60 , 835 – 853 . Google Scholar Crossref Search ADS Adelino M. , Schoar A. , Severino F. ( 2016 ): Loan originations and defaults in the mortgage crisis: the role of the middle class , Review of Financial Studies 29 , 1635 – 1670 . Google Scholar Crossref Search ADS Agarwal S. , Ambrose A. , Chomsisengphet S. , Sanders A. B. ( 2012 ): Thy neighbor’s mortgage: does living in a subprime neighborhood affect one’s probability of default?, Real Estate Economics 40 , 1 – 22 . Google Scholar Crossref Search ADS Agarwal S. , Amromin G. , Ben-David I. , Chomsisengphet S. , Evanoff D. D. ( 2014 ): Predatory lending and the subprime crisis , Journal of Financial Economics 113 , 29 – 52 . Google Scholar Crossref Search ADS Ai C. , Norton E. C. ( 2003 ): Interaction terms in logit and probit models , Economics Letters 80 , 123 – 129 . Google Scholar Crossref Search ADS Amromin G. , Huang J. , Sialm C. ( 2007 ): The tradeoff between mortgage prepayments and tax-deferred savings , Journal of Public Economics 91 , 2014 – 2040 . Google Scholar Crossref Search ADS An X. , Deng Y. , Gabriel S. A. ( 2011 ): Asymmetric information, adverse selection, and the pricing of CMBS , Journal of Financial Economics 100 , 304 – 325 . Google Scholar Crossref Search ADS Barlevy G. , Fisher J. ( 2011 ): Mortgage Choices and Housing Speculation, Federal Reserve Bank of Chicago. FRB of Chicago Working Paper No. 2010-12. Available at SSRN: https://ssrn.com/abstract=1713308. Begley T. , Purnanandam A. ( 2017 ): Design of financial securities: empirical evidence from private-label RMBS deals , Review of Financial Studies 30 , 120 – 161 . Google Scholar Crossref Search ADS Bond P. , Musto D. K. , Yilmaz B. ( 2009 ): Predatory mortgage lending , Journal of Financial Economics 94 , 412 – 427 . Google Scholar Crossref Search ADS Brunnermeier M. K. , Julliard C. ( 2008 ): Money illusion and housing frenzies , Review of Financial Studies 21 , 135 – 180 . Google Scholar Crossref Search ADS Bucks B. K. , Pence K. M. ( 2008 ): Do borrowers know their mortgage terms?, Journal of Urban Economics 64 , 218 – 233 . Google Scholar Crossref Search ADS Campbell J. Y. ( 2006 ): Household finance , Journal of Finance 61 , 1553 – 1604 . Google Scholar Crossref Search ADS Campbell J. Y. , Cocco J. F. ( 2003 ): Household risk management and optimal mortgage choice , Quarterly Journal of Economics 118 , 1449 – 1494 . Google Scholar Crossref Search ADS Campbell J. Y. , Giglio S. , Pathak P. ( 2011 ): Forced sales and house prices , American Economic Review 101 , 2108 – 2131 . Google Scholar Crossref Search ADS Carlin B. I. ( 2009 ): Strategic price complexity in retail financial markets , Journal of Financial Economics 91 , 278 – 287 . Google Scholar Crossref Search ADS Carlin B. I. , Manso G. ( 2011 ): Obfuscation, learning, and the evolution of investor sophistication , Review of Financial Studies 24 , 754 – 785 . Google Scholar Crossref Search ADS Cocco J. F. ( 2013 ): Evidence on the benefits of alternative mortgage products , Journal of Finance 68 , 1663 – 1690 . Google Scholar Crossref Search ADS Corbae D. , Quintin E. ( 2015 ): Leverage and the foreclosure crisis , Journal of Political Economy 123 , 1 – 65 . Google Scholar Crossref Search ADS Demyanyk Y. , Hemert O. V. ( 2011 ): Understanding the subprime mortgage crisis , Review of Financial Studies 24 , 1848 – 1880 . Google Scholar Crossref Search ADS Favilukis J. , Ludvigson S. C. , Van Niewerburgh S. ( 2017 ): The macroeconomic effects of housing wealth, housing finance, and limited risk sharing in general equilibrium , Journal of Political Economy 125 , 140 – 223 . Google Scholar Crossref Search ADS Ferreira F. , Gyourko J. ( 2015 ): A New Look at the U.S. Foreclosure Crisis: Panel Data Evidence of Prime and Subprime Lending, The University of Pennsylvania. NBER Working Paper No. w21261. Available at SSRN: https://ssrn.com/abstract=2618649. Fuster A. , Vickery J. ( 2015 ): Securitization and the fixed-rate mortgage , Review of Financial Studies 28 , 176 – 211 . Google Scholar Crossref Search ADS Fuster A. , Willen P. S. ( 2017 ): Payment size, negative equity, and mortgage default , American Economic Journal: Economic Policy 9 , 167 – 191 . Google Scholar Crossref Search ADS Gabaix X. , Laibson D. ( 2006 ): Shrouded attributes, consumer myopia, and information suppression in competitive markets , Quarterly Journal of Economics 121 , 461 – 504 . Google Scholar Crossref Search ADS Garmaise M. ( 2013 ): The attractions and perils of flexible mortgage lending , Review of Financial Studies 26 , 2548 – 2582 . Google Scholar Crossref Search ADS Gerardi K. S. , Rosen H. S. , Willen P. S. ( 2010 ): The impact of deregulation and financial innovation on consumers: the case of the mortgage market , Journal of Finance 65 , 333 – 360 . Google Scholar Crossref Search ADS Ghent A. C. , Kudlyak M. ( 2011 ): Recourse and residential mortgage default: evidence from US states , Review of Financial Studies 24 , 3139 – 3186 . Google Scholar Crossref Search ADS Goetzmann W. N. , Peng L. , Yen J. ( 2012 ): The subprime crisis and house price appreciation , Journal of Real Estate Finance and Economics 44 , 36 – 66 . Google Scholar Crossref Search ADS Griffin J. M. , Maturana G. ( 2016 ): Who facilitated misreporting in securitized loans , Review of Financial Studies 29 , 384 – 419 . Google Scholar Crossref Search ADS Guiso L. , Sapienza P. , Zingales L. ( 2013 ): The determinants of attitudes towards strategic default on mortgages , Journal of Finance 68 , 1473 – 1515 . Google Scholar Crossref Search ADS Guren A. , McQuade T. ( 2018 ): How do Foreclosures Exacerbate Housing Downturns? Boston University and Stanford University. Jiang W. , Nelson A. A. , Vytlacil E. ( 2014a ): Liar’s loan? Effects of origination channel and information falsification on mortgage delinquency , Review of Economics and Statistics 96 , 1 – 18 . Google Scholar Crossref Search ADS Jiang W. , Nelson A. A. , Vytlacil E. ( 2014b ): Securitization and loan performance: a contrast of ex ante and ex post relations in the mortgage market , Review of Financial Studies 27 , 454 – 483 . Google Scholar Crossref Search ADS Keys B. J. , Mukherjee T. , Seru A. , Vig V. ( 2009 ): Financial regulation and securitization: evidence from subprime loans , Journal of Monetary Economics 56 , 700 – 720 . Google Scholar Crossref Search ADS Keys B. J. , Mukherjee T. , Seru A. , Vig V. ( 2010 ): Did securitization lead to lax screeing? Evidence from subprime loans , Quarterly Journal of Economics 125 , 307 – 362 . Google Scholar Crossref Search ADS Keys B. J. , Seru A. , Vig V. ( 2012 ): Lender screening and the role of securitization: evidence from prime and subprime mortgage markets , Review of Financial Studies 25 , 2071 – 2108 . Google Scholar Crossref Search ADS Koijen R. S. J. , Van Hemert O. , Van Nieuwerburgh S. ( 2009 ): Mortgage timing , Journal of Financial Economics 93 , 292 – 324 . Google Scholar Crossref Search ADS Lamont O. , Stein J. C. ( 1999 ): Leverage and house-price dynamics in U.S. cities , RAND Journal of Economics 30 , 498 – 514 . Google Scholar Crossref Search ADS Landier A. , Srear D. , Thesmar D. ( 2015 ) The Risk-Shifting Hypothesis: Evidence from Subprime Originations , University of California at Berkeley, and HEC . Available at SSRN: https://ssrn.com/abstract=1786542. Landvoigt T. , Piazzesi M. , Schneider M. ( 2015 ): The housing market(s) of San Diego , American Economic Review 105 , 1371 – 1407 . Google Scholar Crossref Search ADS Li W. , White M. J. , Zhu N. ( 2011 ): Did bankruptcy reform cause mortgage defaults to rise?, American Economic Journal: Economic Policy 3 , 123 – 147 . Google Scholar Crossref Search ADS Loutskina E. , Strahan P. E. ( 2009 ): Securitization and the declining impact of bank financial condition on loan supply: evidence from mortgage originations , Journal of Finance 64 , 861 – 922 . Google Scholar Crossref Search ADS Lustig H. , Van Nieuwerburgh S. ( 2005 ): Housing collateral, consumption insurance and risk premia: an empirical perspective , Journal of Finance 60 , 1167 – 1219 . Google Scholar Crossref Search ADS Mayer C. , Pence K. , Sherlund S. ( 2009 ): The rise in mortgage defaults , Journal of Economic Perspectives 23 , 23 – 50 . Google Scholar Crossref Search ADS Melzer B. ( 2017 ): Mortgage debt overhang: reduced investment by homeowners at risk of default , Journal of Finance 72 , 575 – 612 . Google Scholar Crossref Search ADS Mian A. , Sufi A. ( 2009 ): The consequences of mortgage credit expansion: evidence from the U.S. mortgage default crisis , Quarterly Journal of Economics 124 , 1449 – 1496 . Google Scholar Crossref Search ADS Mian A. , Sufi A. , Trebbi F. ( 2015 ): Foreclosures, house prices, and the real economy , Journal of Finance 70 , 2587 – 2634 . Google Scholar Crossref Search ADS Ortalo-Magne F. , Rady S. ( 2006 ): Housing market dynamics: on the contribution of income shocks and credit constraints , Review of Economic Studies 73 , 459 – 485 . Google Scholar Crossref Search ADS Piazzesi M. , Schneider M. , Tuzel S. ( 2007 ): Housing, consumption, and asset pricing , Journal of Financial Economics 83 , 531 – 569 . Google Scholar Crossref Search ADS Piskorski T. , Seru A. , Vig V. ( 2010 ): Securitization and distressed loan renegotiation: evidence from the subprime mortgage crisis , Journal of Financial Economics 97 , 369 – 397 . Google Scholar Crossref Search ADS Piskorski T. , Tchistyi A. ( 2010 ): Optimal mortgage design , Review of Financial Studies 23 , 3098 – 3140 . Google Scholar Crossref Search ADS Piskorski T. , Tchistyi A. ( 2011 ): Stochastic house appreciation and optimal mortgage lending , Review of Financial Studies 24 , 1407 – 1446 . Google Scholar Crossref Search ADS Purnanandam A. ( 2011 ): Originate-to-distribute model and the subprime mortgage crisis , Review of Financial Studies 24 , 1881 – 1915 . Google Scholar Crossref Search ADS Rajan U. , Seru A. , Vig V. ( 2015 ): The failure of models that predict failure: distance, incentives, and defaults , Journal of Financial Economics 115 , 237 – 260 . Google Scholar Crossref Search ADS Stein J. C. ( 1995 ): Prices and trading volume in the housing market: a model with down-payment effects , Quarterly Journal of Economics 110 , 379 – 406 . Google Scholar Crossref Search ADS Van Nieuwerburgh S. , Weill P.-O. ( 2010 ): Why has house price dispersion gone up?, Review of Economic Studies 77 , 1567 – 1606 . Google Scholar Crossref Search ADS Woodward S. E. , Hall R. E. ( 2012 ): Diagnosing consumer confusion and sub-optimal shopping effort: theory and mortgage-market evidence , American Economic Review 102 , 3249 – 3276 . Google Scholar Crossref Search ADS © The Author(s) 2018. Published by Oxford University Press on behalf of the European Finance Association. All rights reserved. For permissions, please email: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Review of Finance Oxford University Press

Loading next page...
1
 
/lp/ou_press/complex-mortgages-SMd8XHxSgw

References (7)

Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of the European Finance Association. All rights reserved. For permissions, please email: journals.permissions@oup.com
ISSN
1572-3097
eISSN
1573-692X
DOI
10.1093/rof/rfy016
Publisher site
See Article on Publisher Site

Abstract

Abstract Complex mortgages became a popular borrowing instrument during the bullish housing market of the early 2000s but vanished rapidly during the subsequent downturn. These non-traditional loans, including interest-only and negative-amortization mortgages, enable households to postpone loan repayment in contrast to fully amortizing traditional mortgages. Contrary to common perception, complex mortgages are used by households with high-income levels and prime credit scores, quite unlike the low-income population targeted by subprime mortgages. Nonetheless, we find that complex-mortgage borrowers become delinquent on their mortgages at rates twice as high as borrowers with plain-vanilla fixed-rate contracts even after controlling for household and loan characteristics. Our findings suggest a link between innovations in mortgage markets focused on prime borrowers and the financial crisis. 1. Introduction The residential mortgage market experienced a significant increase in product complexity in the early 2000s, followed by a rapid reversion back to simpler products after the 2007–2009 financial crisis. The set of complex products featured contracts with zero or negative amortization, short interest rate reset periods, and low introductory teaser interest rates. Unlike traditional fixed-rate (FRM) or adjustable-rate mortgages (ARM), complex mortgages (CM) are not fully amortizing and enable households to postpone loan repayment. The two main types of complex mortgages are interest-only (IO) and negative-amortization mortgages (NEGAM). Figure 1 shows the proportion of fixed-rate, adjustable-rate, interest-only, and negative-amortization mortgage products originated between 1998 and 2009, as reported by LPS Applied Analytics. The share of CM in the USA remained below 2% until the second half of 2003 before jumping to about 30% of mortgage originations just two years later. Figure 1. View largeDownload slide Composition of mortgage products. The figure depicts the composition between fixed-rate mortgages (FRM), adjustable-rate mortgages (ARM), interest-only mortgages (IO), and negative-amortization mortgages (NEGAM) originated over the period between 1998 and 2009. Figure 1. View largeDownload slide Composition of mortgage products. The figure depicts the composition between fixed-rate mortgages (FRM), adjustable-rate mortgages (ARM), interest-only mortgages (IO), and negative-amortization mortgages (NEGAM) originated over the period between 1998 and 2009. There are two contrasting views about the primary drivers of growth in CM products. On the one hand, CM might be pushed by predatory financial institutions to take advantage of naïve households who do not fully understand the contract terms (Gabaix and Laibson, 2006; Carlin, 2009; Carlin and Manso, 2011).1 On the other hand, CM might be taken out by sophisticated households to smooth consumption over their life cycle and to enhance the affordability of expensive homes. The low initial payments of CM are beneficial if households expect their income levels or housing prices to increase over time (Piskorski and Tchistyi, 2010; Cocco, 2013). CM might also be preferred by households who are less averse to defaulting on their mortgages in case of unfavorable house price shocks (Guiso et al., 2013; Garmaise, 2013). By minimizing the initial mortgage amortization and keeping a high mortgage balance, these households maximize the value of their default option. Understanding the clientele for CM and their default patterns is important in assessing the ex ante effectiveness of such financial products and in informing policy debates on financial regulation. Our article contributes to the debate by studying the characteristics of CM borrowers and their subsequent default behavior. We first analyze the borrowers’ choice of the mortgage contract type. Contrary to popular perception, CM are primarily used by households with high-income levels and prime credit scores. Campbell (2006) shows that better-educated and better-off borrowers are more likely to refinance their houses when interest rates are falling. In a similar vein, Bucks and Pence (2008) find that low-income and low-educated households are least knowledgeable about the terms of their mortgage contracts. Furthermore, households who can maintain a high FICO score show that they have the discipline and knowledge to plan their financial matters effectively. Therefore, the CM borrowers are likely more financially sophisticated than average households, in contrast to the subprime borrowers that have received much attention in recent studies.2 We find that both consumption smoothing and default costs are important determinants of selecting CM.3 Consistent with life-cycle consumption smoothing, geographic areas with a higher proportion of young households, higher past house price appreciation, and higher population growth have greater shares of CM. Furthermore, CM borrowers are also more likely to face lower costs of mortgage delinquency. For example, they are more likely to purchase investment properties and to reside in non-recourse states, in which lenders do not have access to the non-collateralized household assets in case of mortgage delinquency. On the other hand, CM borrowers are likely to stretch their income too much to purchase houses that they otherwise could not afford. Despite their high-income levels and prime credit scores, they have higher value-to-income ratios, and are more likely to provide incomplete documentation. We next study the default behavior of mortgage borrowers. The goal of our analysis is to tease out different factors contributing to mortgage delinquency. The contractual design of CM can affect the delinquency rate for several reasons. First, CM payments can change significantly over time, as low initial payments on deferred amortization contracts rise after amortization resets. Households who are already stretching to meet the initial contract obligations might have difficulty meeting the additional monthly payments, especially if they experience unfavorable income or expenditure shocks. Naïve households who are lured into CM to purchase expensive houses might be particularly prone to default. Second, the lack of amortization inevitably leads to higher loan-to-value ratios for any given path of house prices. Households are more likely to default on their mortgages when the current value of the house is lower than the remaining loan balance. Third, CM might also have different delinquency rates due to self-selection of borrowers. CM borrowers might be more risk seeking, might be more exposed to background risks, and might be more willing to default strategically, generating higher delinquency rates. To test these hypotheses, we evaluate default outcomes in the context of a multivariate hazard model. We find evidence of higher mortgage delinquencies associated with both payment resets and increased leverage inherent in CM contracts. Moreover, high value-to-income ratios, which measure the degree that households stretch their purchasing power, are a good predictor of future defaults. Therefore, innovations in mortgage contract design are important determinants of subsequent defaults, both due to the ex-ante enabling of larger house purchases and the ex-post cash flow patterns. However, even after controlling for leverage, payment resets, and other household and loan characteristics, we find significantly higher default rates among CM borrowers. The default probabilities on complex loans are around twice the default probabilities of similar households using FRM. Households that use CM are also significantly more likely to declare personal bankruptcy. This evidence suggests that households who select into CM exhibit higher risk exposures than households who select more traditional mortgages. To investigate why CM borrowers have higher propensities to default, we perform two additional tests. First, we interact the CM indicator with the loan-to-value ratio, which proxies for the value of the default option. We find that CM borrowers’ defaults are more sensitive to the loan-to-value ratio, suggesting that CM borrowers’ decisions depend more on whether defaulting is economically beneficial. In our second test, we interact mortgage contract type with income and credit scores, which proxy for financial savvy and liquidity constraints. While households with high income levels and high credit scores are less likely to default on average, delinquency rates are less affected by income and credit scores for borrowers with CM. Thus, delinquency rates are relatively high for the more savvy and less constrained households with complex contracts. Different types of mortgages exhibit different levels of complexity. FRMs feature constant monthly payments over the whole life of the loan and are therefore the least complex instruments. ARMs exhibit intermediate levels of complexity since the mortgage payments can adjust due to changes in short-term interest rates. CMs are more complex than ARMs, as they typically exhibit changes in mortgage payments due to both changes in interest rates and changes in the amortization schedules. Within CMs, negative-amortization loans are more complex than IO loans, as they feature the largest potential variation in mortgage payments. Consistent with these varying degrees of complexity, we find that our results are more pronounced for mortgages that are more complex. In summary, our findings suggest that complex borrowers are often sophisticated borrowers. Their active choices, both ex-ante for contract selection and ex-post for default decisions, are key drivers behind the mounting defaults of CM during the recent crisis. The expansion of credit and mortgage securitization has received much attention following the financial crisis of 2007–2009. Mian and Sufi (2009) show that the sharp increase in mortgage defaults in 2007 was significantly amplified in geographic areas with a high density of subprime loans that experienced an unprecedented growth in mortgage credit prior to 2007. Keys et al. (2010), Purnanandam (2011), Jiang et al. (2014a), and Griffin and Maturana (2016) focus on the role of mortgage securitization process, finding that lenders applied lower screening efforts on loans that had higher ex-ante probabilities of being securitized.4 Our article suggests a link between innovations in mortgage contract design and the subsequent default behavior of prime borrowers. A number of recent papers highlight the importance of middle-class borrowers for the mortgage crisis. Adelino et al. (2016) show that mortgage originations increased for borrowers across all income levels and FICO scores and that delinquencies during the crisis rose particularly sharply for middle- and high-income borrowers with prime credit. Ferreira and Gyourko (2015) further find that the jump in foreclosures among prime borrowers can be attributed to their negative equity. Our article peels back the lid on the mechanism behind the surge in prime borrower defaults by focusing on the role of contract design. As such, the paper contributes to this literature by suggesting an additional and important channel linking mortgage market innovations to the financial crisis of 2007–2009. Several papers have investigated the design of non-traditional mortgage contracts and their role in the recent crisis. Piskorski and Tchistyi (2010) demonstrate that households with sizeable income variability can benefit more from loans with flexible payment schedules and adjustable rates, whereas Piskorski and Tchistyi (2011) show that such loans are particularly beneficial to borrowers living in areas with high expected house price and income growth. Corbae and Quintin (2015) calibrate a model in which the presence of IO mortgages allows more households to afford larger houses and hence significantly increases the foreclosure rate during the crisis. Cocco (2013) shows that negative amortization contracts in the UK allowed households to smooth their life-cycle consumption of housing. CM households borrowed more relative to their current income and experienced higher ex-post income growth. Importantly, these benefits were concentrated in the period in which new regulations lessened the likelihood of predatory lending practices. Garmaise (2013) shows that offering CM contracts with greater flexibility resulted in an adversely selected pool of borrowers, which suggests an important role for self-selection into CM. Barlevy and Fisher (2011) focus on the effects of non-amortizing mortgages on house prices. They describe a rational expectations model in which both speculators and their lenders use IO mortgages when there is a bubble in-house prices. They provide evidence that IO mortgages were used extensively in cities where inelastic housing supply enables pronounced boom-bust cycles. Landier, Srear, and Thesmar (2015) provide evidence consistent with the use of CM for risk-shifting by a large mortgage originator, whose portfolio of mortgages retained for investment was hit by monetary policy tightening in 2004. In response, the originator began issuing deferred-amortization mortgages, which were designed to be more sensitive to real estate prices than standard contracts. In contrast to these papers, our paper focuses on the characteristics of CM borrowers and on the key determinants of their default behavior. The remainder of this article is structured as follows. Section 2 describes our data sources and reports summary statistics. In Section 3, we study the mortgage choice of households and describe the main features of mortgage contracts. In Section 4, we study the delinquency of different contract types. Section 5 offers concluding remarks. 2. Data Sources and Summary Statistics This section describes in detail the data sources and the main mortgage contracts in our sample. 2.1 Data Our study relies on several complementary data sources that cover various aspects of the housing market during the period between 2003 and 2009. In particular, the micro-level analysis of mortgage contract choice and performance relies heavily on the proprietary mortgage-level database offered by Lender Processing Services (LPS) Applied Analytics (formerly known as McDash Analytics). LPS collects data from some of the nation’s largest mortgage servicers that report contract and borrower details at the time of loan origination, as well as monthly information on mortgage performance. The LPS data coverage has grown steadily over time, with 9 out of 10 largest servicers reporting to the database by 2003. Our database covers about 10 million mortgages with a total loan value of more than $2 trillion originated between 2003 and 2007. We track the performance of all loans till the end of 2009. For the purposes of our study, the availability of granular information on mortgage contract terms is of particular importance. For each of the loans, LPS provides information on the loan interest rate, the amortization schedule, and the securitization status. For ARM, we know the rate at origination, the frequency of resets, the reference rate, and the associated contractual spread. For loans that do not amortize steadily over their term, we know the horizon of the IO period, whether negative amortization is allowed and if so, to what extent and over what period of time. This information allows us to categorize loan contracts. The LPS data also contains information on borrower and property characteristics at the time of origination. These include the appraised property value, the first-lien loan-to-value ratio (LTV), property type (single family or condominium), whether the property was to be occupied by the borrower, and the borrower’s creditworthiness as measured by their FICO (Fair Isaac Corporation) credit score. An important feature of the LPS database is that unlike some other data sources, it is not limited to a particular subset of the loan universe. The LPS data cover prime, subprime, and Alt-A loans,5 and include loans that are privately securitized, those that are sold to Government Sponsored Enterprises (GSEs), and loans that are held on banks’ balance sheets. Even though the coverage allows for a broad set of mortgage contracts, the data set is skewed in favor of securitized loans that are more likely to be serviced by large corporations reporting to LPS. Still, the large overall size of the dataset ensures that we have ample coverage of all contract types.6 We complement borrower information in LPS with household income data collected under the Home Mortgage Disclosure Act (HMDA). Doing so allows us to compute measures of loan affordability, such as the ratio of house value to income (VTI). We further augment the loan-level data with information on trends in local home prices. Quarterly data on home prices by metropolitan statistical area (MSA) are obtained from the Federal Housing Finance Agency (FHFA). We further utilize zip code level information from the 2000 US Census to control for broad demographic characteristics, such as education levels and age distributions. We also make use of the annual per capita income level and unemployment rate data at the MSA level from the Bureau of Economic Analysis (BEA). To determine whether lender recourse has an impact on mortgage choices and mortgage defaults, we follow Ghent and Kudlyak (2011) and classify US states into recourse and non-recourse categories. Whereas lender claims in non-recourse states are limited to the value of the collateral securing the loan, lenders in recourse states may be able to collect on debt not covered by the proceedings from a foreclosure sale by obtaining a deficiency judgment.7 For our delinquency hazard models we track three economic variables after the origination of the mortgage at an annual frequency. The increase in housing prices is defined as the cumulative change of home prices in the MSA since origination. The change in loan balance captures how aggressively households reduced their mortgage debt since origination. Finally, payment resets are defined as the increases in the minimum required mortgage payments since origination. Payment resets are driven by interest rate changes for ARMs and by both interest rate and amortization changes for CMs. The variable definitions and data sources are listed in Table I. Table I. Variable definitions and data sources This table reports the description of the variables used and the corresponding data sources. Variable Data Source Aggregation Description Loan amount LPS Individual First-lien loan amount House value LPS Individual Appraised home value at origination Income HMDA Individual Reported Income from loan application FICO LPS Individual FICO at origination LTV LPS Individual First-lien loan amount divided by appraised value of home VTI LPS Individual Appraisal value divided by income from loan application Interest Rate LPS Individual Average initial interest rate Low documentation LPS Individual Low or no documentation loan indicator variable Above conforming limit LPS Individual Indicator variable for conforming loan. Government securitized LPS Individual Securitization indicator variable after one year of loan life Private securitized LPS Individual Securitization indicator variable after one year of loan life Condo LPS Individual Condominium indicator variable Investment property LPS Individual Second home or investment property indicator variable Refinance LPS Individual Refinance indicator variable With prepayment penalty LPS Individual Indicator variable for prepayment penalty Prepayment penalty term LPS Individual Length in months of prepayment penalty College Census Zip (static) Proportion of 2000 population with college education or better Young Census Zip (static) Proportion of 2000 adult population between 20 and 40 years old House price change FHFA CBSA-Qtr Cumulative house price change in the 5 years prior to the loan origination Non-recourse Ghent and Kudlyak (2011) State/ Individual States where recourse in residential mortgages is limited by the value of the collateral. For California, only purchase loans are classified as non-recourse. BEA income BLS CBSA-Qtr Mean income level per capita Increase in house value FHFA Individual Cumulative house price appreciation at the MSA level since origination Increase in loan balance LPS Individual Percentage change in loan balance since origination Payments reset LPS Individual Percentage change in minimum monthly mortgage payment since origination (1% winsorization; correction of payments due to interest double-counting) Unemployment level BLS CBSA-Qtr Unemployment rate Income growth since origination BEA CBSA-Qtr Growth rate of per capita personal income Variable Data Source Aggregation Description Loan amount LPS Individual First-lien loan amount House value LPS Individual Appraised home value at origination Income HMDA Individual Reported Income from loan application FICO LPS Individual FICO at origination LTV LPS Individual First-lien loan amount divided by appraised value of home VTI LPS Individual Appraisal value divided by income from loan application Interest Rate LPS Individual Average initial interest rate Low documentation LPS Individual Low or no documentation loan indicator variable Above conforming limit LPS Individual Indicator variable for conforming loan. Government securitized LPS Individual Securitization indicator variable after one year of loan life Private securitized LPS Individual Securitization indicator variable after one year of loan life Condo LPS Individual Condominium indicator variable Investment property LPS Individual Second home or investment property indicator variable Refinance LPS Individual Refinance indicator variable With prepayment penalty LPS Individual Indicator variable for prepayment penalty Prepayment penalty term LPS Individual Length in months of prepayment penalty College Census Zip (static) Proportion of 2000 population with college education or better Young Census Zip (static) Proportion of 2000 adult population between 20 and 40 years old House price change FHFA CBSA-Qtr Cumulative house price change in the 5 years prior to the loan origination Non-recourse Ghent and Kudlyak (2011) State/ Individual States where recourse in residential mortgages is limited by the value of the collateral. For California, only purchase loans are classified as non-recourse. BEA income BLS CBSA-Qtr Mean income level per capita Increase in house value FHFA Individual Cumulative house price appreciation at the MSA level since origination Increase in loan balance LPS Individual Percentage change in loan balance since origination Payments reset LPS Individual Percentage change in minimum monthly mortgage payment since origination (1% winsorization; correction of payments due to interest double-counting) Unemployment level BLS CBSA-Qtr Unemployment rate Income growth since origination BEA CBSA-Qtr Growth rate of per capita personal income Table I. Variable definitions and data sources This table reports the description of the variables used and the corresponding data sources. Variable Data Source Aggregation Description Loan amount LPS Individual First-lien loan amount House value LPS Individual Appraised home value at origination Income HMDA Individual Reported Income from loan application FICO LPS Individual FICO at origination LTV LPS Individual First-lien loan amount divided by appraised value of home VTI LPS Individual Appraisal value divided by income from loan application Interest Rate LPS Individual Average initial interest rate Low documentation LPS Individual Low or no documentation loan indicator variable Above conforming limit LPS Individual Indicator variable for conforming loan. Government securitized LPS Individual Securitization indicator variable after one year of loan life Private securitized LPS Individual Securitization indicator variable after one year of loan life Condo LPS Individual Condominium indicator variable Investment property LPS Individual Second home or investment property indicator variable Refinance LPS Individual Refinance indicator variable With prepayment penalty LPS Individual Indicator variable for prepayment penalty Prepayment penalty term LPS Individual Length in months of prepayment penalty College Census Zip (static) Proportion of 2000 population with college education or better Young Census Zip (static) Proportion of 2000 adult population between 20 and 40 years old House price change FHFA CBSA-Qtr Cumulative house price change in the 5 years prior to the loan origination Non-recourse Ghent and Kudlyak (2011) State/ Individual States where recourse in residential mortgages is limited by the value of the collateral. For California, only purchase loans are classified as non-recourse. BEA income BLS CBSA-Qtr Mean income level per capita Increase in house value FHFA Individual Cumulative house price appreciation at the MSA level since origination Increase in loan balance LPS Individual Percentage change in loan balance since origination Payments reset LPS Individual Percentage change in minimum monthly mortgage payment since origination (1% winsorization; correction of payments due to interest double-counting) Unemployment level BLS CBSA-Qtr Unemployment rate Income growth since origination BEA CBSA-Qtr Growth rate of per capita personal income Variable Data Source Aggregation Description Loan amount LPS Individual First-lien loan amount House value LPS Individual Appraised home value at origination Income HMDA Individual Reported Income from loan application FICO LPS Individual FICO at origination LTV LPS Individual First-lien loan amount divided by appraised value of home VTI LPS Individual Appraisal value divided by income from loan application Interest Rate LPS Individual Average initial interest rate Low documentation LPS Individual Low or no documentation loan indicator variable Above conforming limit LPS Individual Indicator variable for conforming loan. Government securitized LPS Individual Securitization indicator variable after one year of loan life Private securitized LPS Individual Securitization indicator variable after one year of loan life Condo LPS Individual Condominium indicator variable Investment property LPS Individual Second home or investment property indicator variable Refinance LPS Individual Refinance indicator variable With prepayment penalty LPS Individual Indicator variable for prepayment penalty Prepayment penalty term LPS Individual Length in months of prepayment penalty College Census Zip (static) Proportion of 2000 population with college education or better Young Census Zip (static) Proportion of 2000 adult population between 20 and 40 years old House price change FHFA CBSA-Qtr Cumulative house price change in the 5 years prior to the loan origination Non-recourse Ghent and Kudlyak (2011) State/ Individual States where recourse in residential mortgages is limited by the value of the collateral. For California, only purchase loans are classified as non-recourse. BEA income BLS CBSA-Qtr Mean income level per capita Increase in house value FHFA Individual Cumulative house price appreciation at the MSA level since origination Increase in loan balance LPS Individual Percentage change in loan balance since origination Payments reset LPS Individual Percentage change in minimum monthly mortgage payment since origination (1% winsorization; correction of payments due to interest double-counting) Unemployment level BLS CBSA-Qtr Unemployment rate Income growth since origination BEA CBSA-Qtr Growth rate of per capita personal income 2.2 Mortgage Contract Design The menu of household mortgage choices was dominated for decades by fully amortizing long-term FRM and, to a lesser extent, by ARM that locked in the initial interest rate for the first years of the contract. In the early 2000s, CM that allow for the deferral of principal repayment became increasingly popular. FRM are level-payment fully amortizing loans with maturities between 10 and 30 years. Borrowers generally have the option to prepay the mortgage if they sell the property or if they refinance their loan when interest rates decline. ARM are fully-amortizing loans where the interest rate changes after an initial period according to a preselected interest rate index. These mortgages often exhibit caps and floors that prevent interest rates from changing too much over the lifetime of the loan. ARM interest rates are generally lower than those on FRMs due to the upward-sloping term structure of interest rates.8 CM include a variety of back-loaded mortgage contracts. Most CM feature adjustable interest rates and exhibit time-varying amortization schedules. The most popular contract is an IO mortgage that only requires borrowers to pay the mortgage interest over an initial time period lasting typically between five and ten years. Subsequently, the mortgage becomes a fully amortizing loan. The other popular type of a CM is a negative amortization mortgage, also known as an option ARM. These mortgages give borrowers the option to initially pay even less than the interest due. The difference between the interest due and the actual mortgage payment is added to the loan balance. These mortgages carry the risk of larger increases in mortgage payments, when the mortgage is recast to become a fully amortizing loan after 5–10 years or when the loan balance exceeds the initial balance at origination by more than a certain amount (typically 10–25%). An additional common feature of NEGAM is a low teaser interest rate of between 1% and 2% during the first 1–12 months. The minimum payment on a NEGAM contract is often set at the level sufficient to cover teaser interest rate charges, and is raised by up to 7.5% on each anniversary of the loan. Overall, one can trace a step function of complexity in mortgage contracts from a standard FRM to ARM, IO, and then NEGAM contracts. With each step, one of the key contract parameters—the interest rate or the amortization schedule—becomes variable, complicating the borrower’s ability to forecast future loan payments, but also expanding the array of borrower actions. 2.3 Geographic Distribution of Mortgages Figure 2 shows the concentration of CM in different counties across the USA in 2002, 2005, and 2008. Consistent with Figure 1, we find that CM were fairly uncommon in 2002. The distribution of CM looks dramatically different in 2005, when multiple counties in California, Colorado, Florida, and Nevada had CM shares in excess of 40%. In some zip codes in these states more than half of mortgage originations were complex loans. Although areas with larger shares of CM often enjoyed higher house price appreciation in the past, there are also exceptions. Numerous areas with high house price appreciation had few CM even at the peak of the housing boom. For example, CM contracts accounted for only about 5% of loans in the Albany, NY metropolitan area where house prices rose by more than 80% between 2001 and 2007. In contrast, CMs proved to be very popular in the Detroit MSA, where nominal house prices remained flat during this period. It is also worth noting that in some areas rapid price increases preceded the surge in CM contracts, whereas other areas had the opposite relationship. Figure 2. View largeDownload slide Geographic distribution of complex mortgages. These figures depict the geographic distribution of complex mortgages originated in 2002, 2005, and 2008. Panel A: Complex Mortgages in 2002; Panel B: Complex Mortgages in 2005; Panel C: Complex Mortgages in 2008. Figure 2. View largeDownload slide Geographic distribution of complex mortgages. These figures depict the geographic distribution of complex mortgages originated in 2002, 2005, and 2008. Panel A: Complex Mortgages in 2002; Panel B: Complex Mortgages in 2005; Panel C: Complex Mortgages in 2008. 2.4 Summary Statistics by Mortgage Type Table II reports statistics for our broad mortgage categories. Our data contain close to 10 million loan contracts originated between 2003 and 2007, of which 69% are FRMs, 13% are ARMs, and the remaining 18% are CMs.9 Table II. Summary statistics by mortgage type This table reports summary statistics for fixed-rate mortgages (FRM), adjustable-rate mortgages (ARM), complex mortgages (CM), and for different types of complex mortgages including interest-only mortgages (IO) and negative-amortization mortgages (NEGAM). All mortgages Complex mortgages FRM ARM CM IO NEGAM Number of observations 6,744,639 1,242,097 1,733,516 1,356,330 377,186 Loan-level variables at origination: Loan amount 180,938 223,838 334,277 328,600 354,691 House value 268,087 310,377 473,833 465,045 505,433 Income 88,329 100,218 133,481 131,066 142,168 Income with full documentation 86,049 95,967 117,904 117,195 121,20 FICO 710.14 681.69 713.18 714.89 707.04 FICO below 620 (in %) 9.85 23.18 5.97 6.65 3.54 First-lien loan-to-value ratio (LTV) 73.62 76.74 73.29 73.53 72.44 Value-to-income ratio (VTI) 3.50 3.54 4.16 4.15 4.23 Initial interest rate (in %) 6.15 6.15 5.03 5.92 1.86 Refinance (in %) 41.38 35.42 44.91 39.71 63.61 Condo (in %) 11.37 16.70 18.63 19.57 15.25 Investment property (in %) 9.05 9.76 11.50 11.63 11.03 Low documentation (in %) 11.13 11.96 24.94 19.99 42.72 Government securitized (in %) 79.04 40.28 25.56 31.06 5.81 Private securitized (in %) 15.05 42.10 53.15 51.94 57.48 With prepayment penalty (in %) 5.76 26.62 32.77 19.31 81.19 Above conforming limit (in %) 5.22 14.00 33.42 31.67 39.71 MSA-level variables at origination: BEA income 36,918 37,483 40,953 41,004 40,767 Proportion with college education (in %) 33.79 36.36 38.61 38.86 37.71 Proportion of young (in %) 40.24 41.21 41.06 41.16 40.68 House price change prior 5 years (in %) 49.94 55.53 74.03 71.71 82.30 Population Growth (in %) 1.10 1.11 1.10 1.14 0.96 Unemployment rate (in %) 5.03 5.20 4.79 4.75 4.97 Non-recourse mortgage (in %) 15.50 20.79 27.68 28.42 25.04 Dynamic variables after origination: Increase in housing values (in %) 9.41 9.49 1.70 1.55 2.35 Increase in loan balance (in %) −5.37 −4.34 −1.87 −2.69 1.55 Payment resets (in %) −0.44 0.10 1.59 0.35 6.82 All mortgages Complex mortgages FRM ARM CM IO NEGAM Number of observations 6,744,639 1,242,097 1,733,516 1,356,330 377,186 Loan-level variables at origination: Loan amount 180,938 223,838 334,277 328,600 354,691 House value 268,087 310,377 473,833 465,045 505,433 Income 88,329 100,218 133,481 131,066 142,168 Income with full documentation 86,049 95,967 117,904 117,195 121,20 FICO 710.14 681.69 713.18 714.89 707.04 FICO below 620 (in %) 9.85 23.18 5.97 6.65 3.54 First-lien loan-to-value ratio (LTV) 73.62 76.74 73.29 73.53 72.44 Value-to-income ratio (VTI) 3.50 3.54 4.16 4.15 4.23 Initial interest rate (in %) 6.15 6.15 5.03 5.92 1.86 Refinance (in %) 41.38 35.42 44.91 39.71 63.61 Condo (in %) 11.37 16.70 18.63 19.57 15.25 Investment property (in %) 9.05 9.76 11.50 11.63 11.03 Low documentation (in %) 11.13 11.96 24.94 19.99 42.72 Government securitized (in %) 79.04 40.28 25.56 31.06 5.81 Private securitized (in %) 15.05 42.10 53.15 51.94 57.48 With prepayment penalty (in %) 5.76 26.62 32.77 19.31 81.19 Above conforming limit (in %) 5.22 14.00 33.42 31.67 39.71 MSA-level variables at origination: BEA income 36,918 37,483 40,953 41,004 40,767 Proportion with college education (in %) 33.79 36.36 38.61 38.86 37.71 Proportion of young (in %) 40.24 41.21 41.06 41.16 40.68 House price change prior 5 years (in %) 49.94 55.53 74.03 71.71 82.30 Population Growth (in %) 1.10 1.11 1.10 1.14 0.96 Unemployment rate (in %) 5.03 5.20 4.79 4.75 4.97 Non-recourse mortgage (in %) 15.50 20.79 27.68 28.42 25.04 Dynamic variables after origination: Increase in housing values (in %) 9.41 9.49 1.70 1.55 2.35 Increase in loan balance (in %) −5.37 −4.34 −1.87 −2.69 1.55 Payment resets (in %) −0.44 0.10 1.59 0.35 6.82 Table II. Summary statistics by mortgage type This table reports summary statistics for fixed-rate mortgages (FRM), adjustable-rate mortgages (ARM), complex mortgages (CM), and for different types of complex mortgages including interest-only mortgages (IO) and negative-amortization mortgages (NEGAM). All mortgages Complex mortgages FRM ARM CM IO NEGAM Number of observations 6,744,639 1,242,097 1,733,516 1,356,330 377,186 Loan-level variables at origination: Loan amount 180,938 223,838 334,277 328,600 354,691 House value 268,087 310,377 473,833 465,045 505,433 Income 88,329 100,218 133,481 131,066 142,168 Income with full documentation 86,049 95,967 117,904 117,195 121,20 FICO 710.14 681.69 713.18 714.89 707.04 FICO below 620 (in %) 9.85 23.18 5.97 6.65 3.54 First-lien loan-to-value ratio (LTV) 73.62 76.74 73.29 73.53 72.44 Value-to-income ratio (VTI) 3.50 3.54 4.16 4.15 4.23 Initial interest rate (in %) 6.15 6.15 5.03 5.92 1.86 Refinance (in %) 41.38 35.42 44.91 39.71 63.61 Condo (in %) 11.37 16.70 18.63 19.57 15.25 Investment property (in %) 9.05 9.76 11.50 11.63 11.03 Low documentation (in %) 11.13 11.96 24.94 19.99 42.72 Government securitized (in %) 79.04 40.28 25.56 31.06 5.81 Private securitized (in %) 15.05 42.10 53.15 51.94 57.48 With prepayment penalty (in %) 5.76 26.62 32.77 19.31 81.19 Above conforming limit (in %) 5.22 14.00 33.42 31.67 39.71 MSA-level variables at origination: BEA income 36,918 37,483 40,953 41,004 40,767 Proportion with college education (in %) 33.79 36.36 38.61 38.86 37.71 Proportion of young (in %) 40.24 41.21 41.06 41.16 40.68 House price change prior 5 years (in %) 49.94 55.53 74.03 71.71 82.30 Population Growth (in %) 1.10 1.11 1.10 1.14 0.96 Unemployment rate (in %) 5.03 5.20 4.79 4.75 4.97 Non-recourse mortgage (in %) 15.50 20.79 27.68 28.42 25.04 Dynamic variables after origination: Increase in housing values (in %) 9.41 9.49 1.70 1.55 2.35 Increase in loan balance (in %) −5.37 −4.34 −1.87 −2.69 1.55 Payment resets (in %) −0.44 0.10 1.59 0.35 6.82 All mortgages Complex mortgages FRM ARM CM IO NEGAM Number of observations 6,744,639 1,242,097 1,733,516 1,356,330 377,186 Loan-level variables at origination: Loan amount 180,938 223,838 334,277 328,600 354,691 House value 268,087 310,377 473,833 465,045 505,433 Income 88,329 100,218 133,481 131,066 142,168 Income with full documentation 86,049 95,967 117,904 117,195 121,20 FICO 710.14 681.69 713.18 714.89 707.04 FICO below 620 (in %) 9.85 23.18 5.97 6.65 3.54 First-lien loan-to-value ratio (LTV) 73.62 76.74 73.29 73.53 72.44 Value-to-income ratio (VTI) 3.50 3.54 4.16 4.15 4.23 Initial interest rate (in %) 6.15 6.15 5.03 5.92 1.86 Refinance (in %) 41.38 35.42 44.91 39.71 63.61 Condo (in %) 11.37 16.70 18.63 19.57 15.25 Investment property (in %) 9.05 9.76 11.50 11.63 11.03 Low documentation (in %) 11.13 11.96 24.94 19.99 42.72 Government securitized (in %) 79.04 40.28 25.56 31.06 5.81 Private securitized (in %) 15.05 42.10 53.15 51.94 57.48 With prepayment penalty (in %) 5.76 26.62 32.77 19.31 81.19 Above conforming limit (in %) 5.22 14.00 33.42 31.67 39.71 MSA-level variables at origination: BEA income 36,918 37,483 40,953 41,004 40,767 Proportion with college education (in %) 33.79 36.36 38.61 38.86 37.71 Proportion of young (in %) 40.24 41.21 41.06 41.16 40.68 House price change prior 5 years (in %) 49.94 55.53 74.03 71.71 82.30 Population Growth (in %) 1.10 1.11 1.10 1.14 0.96 Unemployment rate (in %) 5.03 5.20 4.79 4.75 4.97 Non-recourse mortgage (in %) 15.50 20.79 27.68 28.42 25.04 Dynamic variables after origination: Increase in housing values (in %) 9.41 9.49 1.70 1.55 2.35 Increase in loan balance (in %) −5.37 −4.34 −1.87 −2.69 1.55 Payment resets (in %) −0.44 0.10 1.59 0.35 6.82 CM, on average, are associated with higher loan amounts relative to the traditional ARM and FRM mortgages, and are used to finance more expensive houses. For example, the average home value for CM is $473,833, whereas the average home values for FRMs and ARMs are $268,087 and $310,377, respectively. Counter to some of the commonly made assertions about CM, CM contracts are extended to borrowers with high-income levels and prime credit scores. Indeed, households who take out CM report significantly higher annual incomes ($133,481) than households borrowing through FRM ($88,329) or ARM ($100,218). This ranking persists even when the sample is restricted to loans underwritten on the basis of fully documented income.10 We also find that CM borrowers have credit scores that are better than those of ARM borrowers and similar to those of FRM borrowers. Whereas 23% of ARM borrowers have FICO credit scores below 620, the same can be said of only 10% of FRM and only 6% of CM borrowers. These results emphasize that the clientele for CM differs significantly from that for subprime loans. Nevertheless, the average ratio of house value to income (VTI)—an inverse measure of affordability—is considerably higher in CM contracts, suggesting that CM borrowers are purchasing more expensive houses relative to their income. Yet, higher spending on houses is not reflected in the loan-to-value (LTV) ratios, as all mortgage types have similar first-lien LTV values.11 Several other loan characteristics are different for CM. CM borrowers are more likely to live in a condominium and are more likely to use the property they are financing for investment purposes. We also find significant differences in the frequency of prepayment penalties across mortgage types. Unlike FRMs, a significant fraction of ARMs and CMs face penalties if the loans are prepaid within the first two or three years. CM have a slightly higher share of refinancings compared to new purchases. Since CM are particularly popular for expensive homes, they are also more likely to exceed the conforming loan limit (i.e., jumbo loans). Hence, although 79% of FRMs are securitized by government-sponsored enterprises (GSEs, such as Fannie Mae, Freddie Mac, and Ginnie Mae), only 26% of CMs go through the GSEs. Private securitization partially offsets the lack of GSE involvement in the ARM and CM markets. CM borrowers receive significantly lower initial interest rates than those with FRM or ARM loans. The mean initial interest rate of 5.03% on CM is significantly lower than the rates on FRMs and ARMs (both at about 6.15%). This result is primarily caused by negative amortization mortgages that charge, on average, an initial teaser interest rate of only 1.86%. Unfortunately, we do not observe the age and the education level of borrowers directly. However, we can compute the proportion of adults in zip codes between 20 and 40 years and the proportion of adults with a college education. We find that CM borrowers tend to live in zip codes with higher education levels. From a spatial standpoint, CMs are more common in cities that experienced high house price appreciation. The average 5-year cumulative price appreciation prior to origination amounted to 74% among complex borrowers, as compared with 50% among traditional FRM borrowers. Finally, the population growth rate and the unemployment rate at the time of origination, which capture macroeconomic conditions at the MSA level, are similar in areas with different mortgage compositions. CM are more likely to be originated in non-recourse states, where the lender cannot access assets of the defaulting households beyond the value of the collateral securing the loan. Whereas only 16% of FRMs are non-recourse, 28% of CMs are non-recourse. Since CM were originated relatively later during our sample period, the average house price appreciation after origination is lower for CMs than for FRMs and ARMs. Not surprisingly, borrowers of traditional mortgages reduce their loan balances more aggressively than CM borrowers. Required payments on CMs increase on average more than payments of FRMs and ARMs.12 To illustrate in more detail the distributions of payment resets for various types of mortgages, we compare the required payments over time to the required payments during the first year. Figure 3 shows that a significant fraction of NEGAM borrowers experience jumps in required payments during the first five years. On the other hand, ARM and IO borrowers face smaller changes in their required mortgage payments. Figure 3. View largeDownload slide Mortgage payments over time. These figures depict the cumulative distribution functions of the actual mortgage payments for adjustable- rate (ARM), interest-only (IO), and negative-amortization mortgages (NEGAM) after three and five years relative to the payments during the first year. Panel A: Third-year payment relative to first-year payment. Panel B: Fifth-year payment relative to first-year payment. Figure 3. View largeDownload slide Mortgage payments over time. These figures depict the cumulative distribution functions of the actual mortgage payments for adjustable- rate (ARM), interest-only (IO), and negative-amortization mortgages (NEGAM) after three and five years relative to the payments during the first year. Panel A: Third-year payment relative to first-year payment. Panel B: Fifth-year payment relative to first-year payment. By virtue of their amortization structure, CM largely maintain a high leverage ratio over time. In unreported results, we document that even after five years, only 22% of surviving CM paid down more than 5% of their initial balance, while about 17% increased their balance by 5% or more. This creates a sharp contrast with FRM and ARM borrowers who gradually pay down their loans. This dynamic deterioration in relative leverage ratios becomes particularly dramatic in the event of declining house prices, as experienced during the housing crisis of 2007–2009.13 The last two columns of Table II break out the key summary characteristics among the two CM types. Negative amortization loans, on average, appear to be used to finance more expensive homes and are associated with higher loan values. They also display the highest VTI ratios. As expected, negative amortization loans with their low teaser interest rates commonly carry prepayment penalties. Finally, IO contracts appear to have been subject to stricter underwriting criteria. Whereas only 20% of IOs were underwritten on the basis of less than full documentation, 43% of NEGAM loans were issued in this manner. CM contracts follow a distinct time trend, peaking between 2005 and 2007. To check whether the summary table that aggregates over all origination years (2003 through 2007 in our sample) obscures some important differences, we report in Figure 4 the means of key borrower and loan characteristics over time. We find that differences in these key characteristics across contract types persist across origination years. Income levels of CM borrowers are always higher than those of borrowers with amortizing mortgages. CM borrowers also purchase more expensive homes throughout the sample than other borrowers. The FICO scores of CM borrowers are always on par with those of FRM borrowers and higher than those of ARM borrowers, especially in years where ARMs include many subprime borrowers. CM borrowers are less likely to provide full documentation throughout the sample. CM borrowers take out larger loans relative to their income (VTI) in each of the origination years. The figure also highlights the effects of the near disappearance of the subprime market in early 2007. At that point, the FICO scores, reported income, and the VTI ratio of ARM borrowers all increase substantially. Figure 4. View largeDownload slide Time series of main characteristics. These figures depict the time series of the main characteristics of mortgage borrowers by origination year. Panel A: Income; Panel B: Home value; Panel C: FICO; Panel D: Low documentation; Panel E: Value-to-income ratio; Panel F: First-lien loan-to-value ratio. Figure 4. View largeDownload slide Time series of main characteristics. These figures depict the time series of the main characteristics of mortgage borrowers by origination year. Panel A: Income; Panel B: Home value; Panel C: FICO; Panel D: Low documentation; Panel E: Value-to-income ratio; Panel F: First-lien loan-to-value ratio. 3. Mortgage Choice There are two contrasting views about the appeal of CM products. On the one hand, CM could be predatory products that are pushed by financial institutions to take advantage of unsophisticated households who do not fully understand the contract terms. The low initial payments might obfuscate the long-term borrowing costs for naïve households (Gabaix and Laibson, 2006; Carlin, 2009; Carlin and Manso, 2011). On the other hand, CM might be taken out by sophisticated borrowers who benefit from the deferred amortization. They might take advantage of the payment profile to smooth life cycle consumption, and to afford more expensive houses. In particular, the low initial payments of CM can relax household liquidity and borrowing constraints and enable households to take larger exposures in housing assets (Amromin et al., 2007; Gerardi et al., 2010; Piskorski and Tchistyi, 2010; Cocco, 2013; Guiso, Sapienza, and Zingales, 2013). They might also maximize the value of the default option embedded in such contracts or reduce their tax burdens due to the deductibility of mortgage interest from taxable income. 3.1 Multinomial Logit Regressions of Contract Choice We estimate the likelihood of selection of a particular mortgage contract type (ARM or CM) relative to a baseline contract, which we take to be an FRM. These relative likelihoods are estimated as a function of loan- and borrower-level covariates, as well as MSA-level aggregates. Formally, we use maximum likelihood to estimate the following multinomial logit regressions: Prob(Yi=m)Prob(Yi=FRM)=eβmXi+FEiTime+FEiMSA+FEiLender, (1) where Prob(Yi=m)/Prob(Yi=FRM) is the probability of obtaining an ARM or CM relative to a FRM, X is a vector of mortgage-specific covariates, FETime are indicator variables for the origination quarters, FEMSA are MSA-indicator variables, and FELender are lender-specific indicator variables. To facilitate the interpretation of the economic significance of the results, we standardize the continuous variables by subtracting their mean and dividing by their standard deviation. Table III reports the estimated coefficients. All regressions include time-fixed effects and the standard errors are clustered by MSA. Since some of the MSA level variables are not available for the full sample, the corresponding specifications include fewer observations than the overall sample summarized in Table II. In addition, for computational reasons we only include the largest 50 lenders in the specification with lender-fixed effects. Table III. Multinomial logit regressions This table reports the coefficients of multinomial logit regressions for adjustable-rate mortgages (ARM) and complex mortgages (CM). The coefficients are measured relative to fixed-rate mortgages (FRM). The significance levels are abbreviated with asterisks: One and two asterisks denote significance at the 5% and 1% level, respectively. Individual-level MSA-level MSA Lender Covariates Covariates Fixed effects Fixed effects ARM CM ARM CM ARM CM ARM CM Log(Inc) 0.326** 0.632** 0.223** 0.482** 0.203** 0.420** 0.248** 0.459** (0.014) (0.022) (0.014) (0.015) (0.009) (0.013) (0.013) (0.015) FICO −0.512** −0.040** −0.511** −0.020* −0.519** −0.034** −0.455** 0.042** (0.011) (0.011) (0.011) (0.010) (0.011) (0.009) (0.010) (0.009) LTV 0.200** 0.320** 0.214** 0.345** 0.210** 0.350** 0.312** 0.433** (0.020) (0.027) (0.020) (0.029) (0.019) (0.031) (0.016) (0.028) VTI 0.301** 0.538** 0.186** 0.352** 0.143** 0.261** 0.233** 0.385** (0.023) (0.029) (0.020) (0.023) (0.013) (0.015) (0.023) (0.027) Low Doc 0.104* 0.787** 0.141** 0.822** 0.141** 0.815** 0.178** 0.528** (0.042) (0.050) (0.041) (0.053) (0.037) (0.046) (0.033) (0.039) Above limit 0.699** 1.274** 0.658** 1.170** 0.699** 1.137** 0.580** 1.132** (0.060) (0.084) (0.047) (0.062) (0.038) (0.039) (0.048) (0.067) Condo 0.585** 0.691** 0.424** 0.460** 0.373** 0.380** 0.490** 0.501** (0.050) (0.059) (0.051) (0.046) (0.030) (0.025) (0.047) (0.041) Investment 0.300** 0.229** 0.353** 0.208** 0.326** 0.166** 0.318** 0.343** (0.024) (0.041) (0.020) (0.031) (0.018) (0.030) (0.019) (0.029) Refinance −0.258** 0.224** −0.222** 0.287** −0.302** 0.089* −0.203** 0.044 (0.018) (0.032) (0.017) (0.045) (0.018) (0.039) (0.019) (0.047) College 0.114** 0.040* 0.127** 0.069** 0.125** 0.083** (0.012) (0.019) (0.009) (0.012) (0.012) (0.020) Young 0.091** 0.098** 0.093** 0.064** 0.081** 0.086** (0.016) (0.016) (0.010) (0.007) (0.014) (0.015) Price change 0.079** 0.364** 0.180** 0.176** 0.016 0.300** (0.027) (0.038) (0.030) (0.033) (0.026) (0.034) Pop growth 0.021 0.120** −0.024 −0.003 0.030 0.131** (0.026) (0.041) (0.015) (0.021) (0.025) (0.038) Log(BEA Inc) 0.100** 0.149** 0.035 0.712** 0.087** 0.139** (0.027) (0.041) (0.127) (0.155) (0.027) (0.039) Non-Recourse 0.344** 0.625** 0.250** 0.502** (0.048) (0.082) (0.047) (0.077) Year FE Yes Yes Yes Yes MSA FE No No Yes No Lender FE No No No Yes Observations 9, 720, 252 8, 915, 275 8, 915, 275 7, 538, 406 Individual-level MSA-level MSA Lender Covariates Covariates Fixed effects Fixed effects ARM CM ARM CM ARM CM ARM CM Log(Inc) 0.326** 0.632** 0.223** 0.482** 0.203** 0.420** 0.248** 0.459** (0.014) (0.022) (0.014) (0.015) (0.009) (0.013) (0.013) (0.015) FICO −0.512** −0.040** −0.511** −0.020* −0.519** −0.034** −0.455** 0.042** (0.011) (0.011) (0.011) (0.010) (0.011) (0.009) (0.010) (0.009) LTV 0.200** 0.320** 0.214** 0.345** 0.210** 0.350** 0.312** 0.433** (0.020) (0.027) (0.020) (0.029) (0.019) (0.031) (0.016) (0.028) VTI 0.301** 0.538** 0.186** 0.352** 0.143** 0.261** 0.233** 0.385** (0.023) (0.029) (0.020) (0.023) (0.013) (0.015) (0.023) (0.027) Low Doc 0.104* 0.787** 0.141** 0.822** 0.141** 0.815** 0.178** 0.528** (0.042) (0.050) (0.041) (0.053) (0.037) (0.046) (0.033) (0.039) Above limit 0.699** 1.274** 0.658** 1.170** 0.699** 1.137** 0.580** 1.132** (0.060) (0.084) (0.047) (0.062) (0.038) (0.039) (0.048) (0.067) Condo 0.585** 0.691** 0.424** 0.460** 0.373** 0.380** 0.490** 0.501** (0.050) (0.059) (0.051) (0.046) (0.030) (0.025) (0.047) (0.041) Investment 0.300** 0.229** 0.353** 0.208** 0.326** 0.166** 0.318** 0.343** (0.024) (0.041) (0.020) (0.031) (0.018) (0.030) (0.019) (0.029) Refinance −0.258** 0.224** −0.222** 0.287** −0.302** 0.089* −0.203** 0.044 (0.018) (0.032) (0.017) (0.045) (0.018) (0.039) (0.019) (0.047) College 0.114** 0.040* 0.127** 0.069** 0.125** 0.083** (0.012) (0.019) (0.009) (0.012) (0.012) (0.020) Young 0.091** 0.098** 0.093** 0.064** 0.081** 0.086** (0.016) (0.016) (0.010) (0.007) (0.014) (0.015) Price change 0.079** 0.364** 0.180** 0.176** 0.016 0.300** (0.027) (0.038) (0.030) (0.033) (0.026) (0.034) Pop growth 0.021 0.120** −0.024 −0.003 0.030 0.131** (0.026) (0.041) (0.015) (0.021) (0.025) (0.038) Log(BEA Inc) 0.100** 0.149** 0.035 0.712** 0.087** 0.139** (0.027) (0.041) (0.127) (0.155) (0.027) (0.039) Non-Recourse 0.344** 0.625** 0.250** 0.502** (0.048) (0.082) (0.047) (0.077) Year FE Yes Yes Yes Yes MSA FE No No Yes No Lender FE No No No Yes Observations 9, 720, 252 8, 915, 275 8, 915, 275 7, 538, 406 Table III. Multinomial logit regressions This table reports the coefficients of multinomial logit regressions for adjustable-rate mortgages (ARM) and complex mortgages (CM). The coefficients are measured relative to fixed-rate mortgages (FRM). The significance levels are abbreviated with asterisks: One and two asterisks denote significance at the 5% and 1% level, respectively. Individual-level MSA-level MSA Lender Covariates Covariates Fixed effects Fixed effects ARM CM ARM CM ARM CM ARM CM Log(Inc) 0.326** 0.632** 0.223** 0.482** 0.203** 0.420** 0.248** 0.459** (0.014) (0.022) (0.014) (0.015) (0.009) (0.013) (0.013) (0.015) FICO −0.512** −0.040** −0.511** −0.020* −0.519** −0.034** −0.455** 0.042** (0.011) (0.011) (0.011) (0.010) (0.011) (0.009) (0.010) (0.009) LTV 0.200** 0.320** 0.214** 0.345** 0.210** 0.350** 0.312** 0.433** (0.020) (0.027) (0.020) (0.029) (0.019) (0.031) (0.016) (0.028) VTI 0.301** 0.538** 0.186** 0.352** 0.143** 0.261** 0.233** 0.385** (0.023) (0.029) (0.020) (0.023) (0.013) (0.015) (0.023) (0.027) Low Doc 0.104* 0.787** 0.141** 0.822** 0.141** 0.815** 0.178** 0.528** (0.042) (0.050) (0.041) (0.053) (0.037) (0.046) (0.033) (0.039) Above limit 0.699** 1.274** 0.658** 1.170** 0.699** 1.137** 0.580** 1.132** (0.060) (0.084) (0.047) (0.062) (0.038) (0.039) (0.048) (0.067) Condo 0.585** 0.691** 0.424** 0.460** 0.373** 0.380** 0.490** 0.501** (0.050) (0.059) (0.051) (0.046) (0.030) (0.025) (0.047) (0.041) Investment 0.300** 0.229** 0.353** 0.208** 0.326** 0.166** 0.318** 0.343** (0.024) (0.041) (0.020) (0.031) (0.018) (0.030) (0.019) (0.029) Refinance −0.258** 0.224** −0.222** 0.287** −0.302** 0.089* −0.203** 0.044 (0.018) (0.032) (0.017) (0.045) (0.018) (0.039) (0.019) (0.047) College 0.114** 0.040* 0.127** 0.069** 0.125** 0.083** (0.012) (0.019) (0.009) (0.012) (0.012) (0.020) Young 0.091** 0.098** 0.093** 0.064** 0.081** 0.086** (0.016) (0.016) (0.010) (0.007) (0.014) (0.015) Price change 0.079** 0.364** 0.180** 0.176** 0.016 0.300** (0.027) (0.038) (0.030) (0.033) (0.026) (0.034) Pop growth 0.021 0.120** −0.024 −0.003 0.030 0.131** (0.026) (0.041) (0.015) (0.021) (0.025) (0.038) Log(BEA Inc) 0.100** 0.149** 0.035 0.712** 0.087** 0.139** (0.027) (0.041) (0.127) (0.155) (0.027) (0.039) Non-Recourse 0.344** 0.625** 0.250** 0.502** (0.048) (0.082) (0.047) (0.077) Year FE Yes Yes Yes Yes MSA FE No No Yes No Lender FE No No No Yes Observations 9, 720, 252 8, 915, 275 8, 915, 275 7, 538, 406 Individual-level MSA-level MSA Lender Covariates Covariates Fixed effects Fixed effects ARM CM ARM CM ARM CM ARM CM Log(Inc) 0.326** 0.632** 0.223** 0.482** 0.203** 0.420** 0.248** 0.459** (0.014) (0.022) (0.014) (0.015) (0.009) (0.013) (0.013) (0.015) FICO −0.512** −0.040** −0.511** −0.020* −0.519** −0.034** −0.455** 0.042** (0.011) (0.011) (0.011) (0.010) (0.011) (0.009) (0.010) (0.009) LTV 0.200** 0.320** 0.214** 0.345** 0.210** 0.350** 0.312** 0.433** (0.020) (0.027) (0.020) (0.029) (0.019) (0.031) (0.016) (0.028) VTI 0.301** 0.538** 0.186** 0.352** 0.143** 0.261** 0.233** 0.385** (0.023) (0.029) (0.020) (0.023) (0.013) (0.015) (0.023) (0.027) Low Doc 0.104* 0.787** 0.141** 0.822** 0.141** 0.815** 0.178** 0.528** (0.042) (0.050) (0.041) (0.053) (0.037) (0.046) (0.033) (0.039) Above limit 0.699** 1.274** 0.658** 1.170** 0.699** 1.137** 0.580** 1.132** (0.060) (0.084) (0.047) (0.062) (0.038) (0.039) (0.048) (0.067) Condo 0.585** 0.691** 0.424** 0.460** 0.373** 0.380** 0.490** 0.501** (0.050) (0.059) (0.051) (0.046) (0.030) (0.025) (0.047) (0.041) Investment 0.300** 0.229** 0.353** 0.208** 0.326** 0.166** 0.318** 0.343** (0.024) (0.041) (0.020) (0.031) (0.018) (0.030) (0.019) (0.029) Refinance −0.258** 0.224** −0.222** 0.287** −0.302** 0.089* −0.203** 0.044 (0.018) (0.032) (0.017) (0.045) (0.018) (0.039) (0.019) (0.047) College 0.114** 0.040* 0.127** 0.069** 0.125** 0.083** (0.012) (0.019) (0.009) (0.012) (0.012) (0.020) Young 0.091** 0.098** 0.093** 0.064** 0.081** 0.086** (0.016) (0.016) (0.010) (0.007) (0.014) (0.015) Price change 0.079** 0.364** 0.180** 0.176** 0.016 0.300** (0.027) (0.038) (0.030) (0.033) (0.026) (0.034) Pop growth 0.021 0.120** −0.024 −0.003 0.030 0.131** (0.026) (0.041) (0.015) (0.021) (0.025) (0.038) Log(BEA Inc) 0.100** 0.149** 0.035 0.712** 0.087** 0.139** (0.027) (0.041) (0.127) (0.155) (0.027) (0.039) Non-Recourse 0.344** 0.625** 0.250** 0.502** (0.048) (0.082) (0.047) (0.077) Year FE Yes Yes Yes Yes MSA FE No No Yes No Lender FE No No No Yes Observations 9, 720, 252 8, 915, 275 8, 915, 275 7, 538, 406 Because we find that these loans are not concentrated in low-income areas with poorly educated households, we find little support for the hypothesis that CM are pushed to naïve households. Instead, we find that households with higher income levels are significantly more likely to obtain a CM than to take out a more traditional FRM loan. The estimates imply that a one-standard-deviation change in log income raises the ratio between the probabilities of choosing a CM over an FRM contract almost twofold (⁠ e0.632=1.88 ⁠). Moreover, households with higher FICO scores are substantially more likely to choose a CM than to choose an ARM, although the results are mixed when we compare the propensity to choose a CM relative to a FRM.14 Areas with higher proportions of college graduates and with higher median incomes are also associated with a higher proportion of CM contracts. Overall, there is little evidence that a typical CM is taken out by poor households who are more prone to predatory lending. We find some evidence indicating that CM are “affordability products” for households who anticipate income growth or house price appreciation. The estimated coefficients on the loan-to-value (LTV) and the value-to-income (VTI) ratios are significantly higher for CM households, suggesting that these households are stretching their budget to afford more expensive homes. Another piece of evidence consistent with the idea of CM contracts as affordability product is that they are much more prevalent for mortgages above the GSE conforming loan limit. Such mortgages cannot be securitized by the GSEs and, consequently, result in higher interest rates (the so-called jumbo spread). This increases the relative appeal of payment-shrinking CM products. While we do not observe household expectations for their income and house price growth, we introduce several proxies for these expectations. Since young households anticipate a higher growth rate of their labor income than older households, we use the proportion of adults between 20 and 40 years to proxy for income expectations and find that CM contracts are more popular in areas with a larger proportion of younger households. To the extent that households might extrapolate past local experiences to build their expectations about future house price dynamics, we use the prior five years’ house price appreciation in the MSA to proxy for the expected future house price growth. Borrowers in geographic areas where appreciation was substantial might be more willing to accept non-amortizing loans if they expect the appreciation to continue in the future. In addition, the prior one-year population growth rate in the MSA captures primarily migration pressures. Geographic areas with significant population growth might be areas where households expect significant house price and income growth. We find that past house price appreciation and the local population growth significantly increase the propensity of obtaining a CM, which suggests that the expectations of continued growth are likely a driving force behind the popularity of CM. This evidence is also consistent with CM borrowers being subject to a behavioral bias of extrapolating past prices too much into the future, despite their higher education and income levels. Finally, we also find supporting evidence that CM are selected by a different type of households who might have different risk exposures and face different costs of mortgage delinquency. First, we observe that CM borrowers are much more likely to provide incomplete documentation for their loans. These households either are unable to provide documentation for their income due to unstable income streams, or are inflating their incomes to qualify for higher loan amounts. Going forward, they might continue to have more volatile incomes and hence higher default probabilities. Second, we find that CM mortgages are more likely to be used to finance investment properties. Owners of these properties do not lose their primary residence upon defaulting on their mortgages and hence have potentially lower costs of default. They might therefore have an incentive to pay down their mortgage balance relatively slowly to increase the default option value. Third, we also find that households in non-recourse states are significantly more likely to obtain a CM than households in recourse states. This might be caused by the higher option value of defaulting on non-recourse mortgages, when a delinquent household can default on the mortgage without worrying about lenders accessing their other assets. While it is possible that the positive association between CM contract choice and income reflects the propensity of CMs to be concentrated in high income MSAs, specifications that incorporate MSA-level controls and MSA-fixed effects preserve these relationships. Therefore, even within individual geographies, CM choice is favored by the relatively well-off. These MSA-fixed effects also control for other unobserved geographic differences in regulation, topography, and geography.15 It is also possible that the contract choices reflect the decision of lenders, who determine the menu of available contract options and possibly also steer the borrowers towards certain items on the menu. By including lender-fixed effects, we control for the fact that some lenders might offer only specific mortgage instruments and might target-specific clienteles. The coefficient on income remains economically and statistically significant after including lender-fixed effects.16 In summary, we find that CM borrowers are well-educated high-income households with prime credit scores. They are stretching their budget to purchase expensive houses, partly due to their expectation of higher future income or house price growth. They might also face lower costs of mortgage delinquency as they are more likely to provide incomplete documentation, to purchase investment properties, and to reside in non-recourse states. 3.2 Robustness Tests Table IV reports the coefficients of multinomial logit regressions that further differentiate between the two main types of complex contracts. The estimates are consistent with the univariate results in Table II. In particular, we see that NEGAM contracts are used by high-income borrowers to refinance their high-priced primary residences, often on the basis of only limited income and asset documentation. Table IV. Multinomial logit regressions for detailed classification This table reports the coefficients of multinomial logit regressions for adjustable-rate mortgages (ARM), interest-only mortgages (IO), and negative-amortization mortgages (NEGAM). The coefficients are measured relative to fixed-rate mortgages (FRM). The significance levels are abbreviated with asterisks: one and two asterisks denote significance at the 5% and 1% level, respectively. Individual-level covariates MSA-level covariates ARM IO NEGAM ARM IO NEGAM Log(Inc) 0.328** 0.582** 0.851** 0.226** 0.432** 0.714** (0.014) (0.021) (0.021) (0.015) (0.015) (0.018) FICO −0.513** −0.028* −0.089** −0.511** −0.015 −0.042** (0.011) (0.011) (0.016) (0.010) (0.010) (0.016) LTV 0.202** 0.284** 0.498** 0.216** 0.302** 0.558** (0.020) (0.026) (0.026) (0.020) (0.029) (0.026) VTI 0.302** 0.526** 0.602** 0.186** 0.345** 0.397** (0.023) (0.029) (0.031) (0.020) (0.022) (0.026) Low Doc 0.126** 0.534** 1.598** 0.163** 0.572** 1.636** (0.043) (0.047) (0.049) (0.042) (0.049) (0.053) Above limit 0.703** 1.271** 1.264** 0.661** 1.180** 1.096** (0.061) (0.080) (0.099) (0.047) (0.059) (0.072) Condo 0.582** 0.712** 0.575** 0.421** 0.486** 0.340** (0.051) (0.055) (0.089) (0.052) (0.043) (0.065) Investment 0.302** 0.207** 0.365** 0.355** 0.188** 0.351** (0.024) (0.045) (0.045) (0.020) (0.034) (0.043) Refinance −0.244** 0.021 1.070** −0.209** 0.084 1.207** (0.018) (0.032) (0.050) (0.017) (0.043) (0.057) College 0.112** 0.061** −0.066** (0.012) (0.020) (0.019) Young 0.091** 0.101** 0.066** (0.016) (0.016) (0.023) Price change 0.081** 0.319** 0.561** (0.027) (0.037) (0.050) Pop Growth 0.021 0.126** 0.072 (0.026) (0.041) (0.051) Log(BEA Inc) 0.100** 0.142** 0.192** (0.027) (0.039) (0.065) Non-recourse 0.344** 0.604** 0.825** (0.048) (0.081) (0.107) Year FE Yes Yes MSA FE No No Lender FE No No Observations 9,720,252 8,915,275 Individual-level covariates MSA-level covariates ARM IO NEGAM ARM IO NEGAM Log(Inc) 0.328** 0.582** 0.851** 0.226** 0.432** 0.714** (0.014) (0.021) (0.021) (0.015) (0.015) (0.018) FICO −0.513** −0.028* −0.089** −0.511** −0.015 −0.042** (0.011) (0.011) (0.016) (0.010) (0.010) (0.016) LTV 0.202** 0.284** 0.498** 0.216** 0.302** 0.558** (0.020) (0.026) (0.026) (0.020) (0.029) (0.026) VTI 0.302** 0.526** 0.602** 0.186** 0.345** 0.397** (0.023) (0.029) (0.031) (0.020) (0.022) (0.026) Low Doc 0.126** 0.534** 1.598** 0.163** 0.572** 1.636** (0.043) (0.047) (0.049) (0.042) (0.049) (0.053) Above limit 0.703** 1.271** 1.264** 0.661** 1.180** 1.096** (0.061) (0.080) (0.099) (0.047) (0.059) (0.072) Condo 0.582** 0.712** 0.575** 0.421** 0.486** 0.340** (0.051) (0.055) (0.089) (0.052) (0.043) (0.065) Investment 0.302** 0.207** 0.365** 0.355** 0.188** 0.351** (0.024) (0.045) (0.045) (0.020) (0.034) (0.043) Refinance −0.244** 0.021 1.070** −0.209** 0.084 1.207** (0.018) (0.032) (0.050) (0.017) (0.043) (0.057) College 0.112** 0.061** −0.066** (0.012) (0.020) (0.019) Young 0.091** 0.101** 0.066** (0.016) (0.016) (0.023) Price change 0.081** 0.319** 0.561** (0.027) (0.037) (0.050) Pop Growth 0.021 0.126** 0.072 (0.026) (0.041) (0.051) Log(BEA Inc) 0.100** 0.142** 0.192** (0.027) (0.039) (0.065) Non-recourse 0.344** 0.604** 0.825** (0.048) (0.081) (0.107) Year FE Yes Yes MSA FE No No Lender FE No No Observations 9,720,252 8,915,275 Table IV. Multinomial logit regressions for detailed classification This table reports the coefficients of multinomial logit regressions for adjustable-rate mortgages (ARM), interest-only mortgages (IO), and negative-amortization mortgages (NEGAM). The coefficients are measured relative to fixed-rate mortgages (FRM). The significance levels are abbreviated with asterisks: one and two asterisks denote significance at the 5% and 1% level, respectively. Individual-level covariates MSA-level covariates ARM IO NEGAM ARM IO NEGAM Log(Inc) 0.328** 0.582** 0.851** 0.226** 0.432** 0.714** (0.014) (0.021) (0.021) (0.015) (0.015) (0.018) FICO −0.513** −0.028* −0.089** −0.511** −0.015 −0.042** (0.011) (0.011) (0.016) (0.010) (0.010) (0.016) LTV 0.202** 0.284** 0.498** 0.216** 0.302** 0.558** (0.020) (0.026) (0.026) (0.020) (0.029) (0.026) VTI 0.302** 0.526** 0.602** 0.186** 0.345** 0.397** (0.023) (0.029) (0.031) (0.020) (0.022) (0.026) Low Doc 0.126** 0.534** 1.598** 0.163** 0.572** 1.636** (0.043) (0.047) (0.049) (0.042) (0.049) (0.053) Above limit 0.703** 1.271** 1.264** 0.661** 1.180** 1.096** (0.061) (0.080) (0.099) (0.047) (0.059) (0.072) Condo 0.582** 0.712** 0.575** 0.421** 0.486** 0.340** (0.051) (0.055) (0.089) (0.052) (0.043) (0.065) Investment 0.302** 0.207** 0.365** 0.355** 0.188** 0.351** (0.024) (0.045) (0.045) (0.020) (0.034) (0.043) Refinance −0.244** 0.021 1.070** −0.209** 0.084 1.207** (0.018) (0.032) (0.050) (0.017) (0.043) (0.057) College 0.112** 0.061** −0.066** (0.012) (0.020) (0.019) Young 0.091** 0.101** 0.066** (0.016) (0.016) (0.023) Price change 0.081** 0.319** 0.561** (0.027) (0.037) (0.050) Pop Growth 0.021 0.126** 0.072 (0.026) (0.041) (0.051) Log(BEA Inc) 0.100** 0.142** 0.192** (0.027) (0.039) (0.065) Non-recourse 0.344** 0.604** 0.825** (0.048) (0.081) (0.107) Year FE Yes Yes MSA FE No No Lender FE No No Observations 9,720,252 8,915,275 Individual-level covariates MSA-level covariates ARM IO NEGAM ARM IO NEGAM Log(Inc) 0.328** 0.582** 0.851** 0.226** 0.432** 0.714** (0.014) (0.021) (0.021) (0.015) (0.015) (0.018) FICO −0.513** −0.028* −0.089** −0.511** −0.015 −0.042** (0.011) (0.011) (0.016) (0.010) (0.010) (0.016) LTV 0.202** 0.284** 0.498** 0.216** 0.302** 0.558** (0.020) (0.026) (0.026) (0.020) (0.029) (0.026) VTI 0.302** 0.526** 0.602** 0.186** 0.345** 0.397** (0.023) (0.029) (0.031) (0.020) (0.022) (0.026) Low Doc 0.126** 0.534** 1.598** 0.163** 0.572** 1.636** (0.043) (0.047) (0.049) (0.042) (0.049) (0.053) Above limit 0.703** 1.271** 1.264** 0.661** 1.180** 1.096** (0.061) (0.080) (0.099) (0.047) (0.059) (0.072) Condo 0.582** 0.712** 0.575** 0.421** 0.486** 0.340** (0.051) (0.055) (0.089) (0.052) (0.043) (0.065) Investment 0.302** 0.207** 0.365** 0.355** 0.188** 0.351** (0.024) (0.045) (0.045) (0.020) (0.034) (0.043) Refinance −0.244** 0.021 1.070** −0.209** 0.084 1.207** (0.018) (0.032) (0.050) (0.017) (0.043) (0.057) College 0.112** 0.061** −0.066** (0.012) (0.020) (0.019) Young 0.091** 0.101** 0.066** (0.016) (0.016) (0.023) Price change 0.081** 0.319** 0.561** (0.027) (0.037) (0.050) Pop Growth 0.021 0.126** 0.072 (0.026) (0.041) (0.051) Log(BEA Inc) 0.100** 0.142** 0.192** (0.027) (0.039) (0.065) Non-recourse 0.344** 0.604** 0.825** (0.048) (0.081) (0.107) Year FE Yes Yes MSA FE No No Lender FE No No Observations 9,720,252 8,915,275 Our conclusion that borrowers with CM are relatively financially sophisticated is partially based on the fact that these borrowers report higher income levels. However, the income levels of low-documentation borrowers are not verified and might not be reliable. To investigate whether this biases our results, Table V presents the CM coefficients of the multinomial logit regressions for the subsample of households with full documentation loans.17 Overall, conditioning on full documentation loans does not affect our main results qualitatively. Table V. Multinomial logit regressions for subsamples This table reports the coefficients of multinomial logit regressions of complex mortgages (CM) relative to fixed-rate mortgages (FRM) for the following subsamples: loans with full documentation; loans to borrowers with FICO scores below 620; loans originated to purchase a new house; loans used to finance an investment property; non-securitized loans; and loans originated in states other than California. The significance levels are abbreviated with asterisks: One and two asterisks denote significance at the 5% and 1% level, respectively. Full Subprime Purchases Invest Not Exclude Document Only Only Property Securitized California Log(Inc) 0.418** 0.366** 0.429** 0.408** 0.696** 0.456** (0.016) (0.023) (0.015) (0.018) (0.022) (0.016) FICO −0.146** 0.023 −0.086** −0.097** −0.086** −0.019 (0.010) (0.029) (0.011) (0.011) (0.017) (0.011) LTV 0.370** −0.037 0.162** 0.429** 0.381** 0.251** (0.029) (0.040) (0.034) (0.019) (0.026) (0.024) VTI 0.342** 0.167** 0.337** 0.347** 0.419** 0.357** (0.024) (0.025) (0.022) (0.041) (0.039) (0.031) Low Doc 0.126 0.580** −0.152** 2.143** 0.666** (0.077) (0.049) (0.033) (0.128) (0.045) Above limit 1.055** 1.775** 1.183** 0.973** 0.731** 1.034** (0.055) (0.115) (0.067) (0.052) (0.069) (0.044) Condo 0.450** 0.487** 0.459** 0.173** 0.350** 0.436** (0.037) (0.064) (0.044) (0.035) (0.053) (0.048) Investment 0.042 −0.140* 0.294** 0.033 0.289** (0.030) (0.068) (0.033) (0.042) (0.030) Refinance 0.093* 0.267** 0.030 0.423** 0.247** (0.038) (0.034) (0.033) (0.057) (0.040) College 0.051* −0.057* 0.040* 0.062** 0.062** 0.073** (0.022) (0.027) (0.020) (0.015) (0.024) (0.018) Young 0.085** 0.106** 0.118** 0.071** 0.041* 0.075** (0.015) (0.020) (0.019) (0.014) (0.020) (0.019) Price change 0.217** 0.273** 0.438** 0.367** 0.308** 0.280** (0.033) (0.042) (0.042) (0.034) (0.048) (0.056) Pop growth 0.127** 0.014 0.166** 0.166** 0.046 0.180** (0.038) (0.032) (0.047) (0.039) (0.051) (0.051) Log(BEA Inc) 0.114** 0.018 0.175** 0.124** 0.023 0.142** (0.038) (0.050) (0.041) (0.039) (0.047) (0.053) Non-recourse 0.626** 0.777** 0.712** 0.449** 0.858** 0.361** (0.060) (0.091) (0.099) (0.061) (0.097) (0.111) Origination year FE Yes Yes Yes Yes Yes Yes MSA FE No No No No No No Lender FE No No No No No No Observations 3,279,219 979,440 5,214,750 826,604 929,451 7,545,583 Full Subprime Purchases Invest Not Exclude Document Only Only Property Securitized California Log(Inc) 0.418** 0.366** 0.429** 0.408** 0.696** 0.456** (0.016) (0.023) (0.015) (0.018) (0.022) (0.016) FICO −0.146** 0.023 −0.086** −0.097** −0.086** −0.019 (0.010) (0.029) (0.011) (0.011) (0.017) (0.011) LTV 0.370** −0.037 0.162** 0.429** 0.381** 0.251** (0.029) (0.040) (0.034) (0.019) (0.026) (0.024) VTI 0.342** 0.167** 0.337** 0.347** 0.419** 0.357** (0.024) (0.025) (0.022) (0.041) (0.039) (0.031) Low Doc 0.126 0.580** −0.152** 2.143** 0.666** (0.077) (0.049) (0.033) (0.128) (0.045) Above limit 1.055** 1.775** 1.183** 0.973** 0.731** 1.034** (0.055) (0.115) (0.067) (0.052) (0.069) (0.044) Condo 0.450** 0.487** 0.459** 0.173** 0.350** 0.436** (0.037) (0.064) (0.044) (0.035) (0.053) (0.048) Investment 0.042 −0.140* 0.294** 0.033 0.289** (0.030) (0.068) (0.033) (0.042) (0.030) Refinance 0.093* 0.267** 0.030 0.423** 0.247** (0.038) (0.034) (0.033) (0.057) (0.040) College 0.051* −0.057* 0.040* 0.062** 0.062** 0.073** (0.022) (0.027) (0.020) (0.015) (0.024) (0.018) Young 0.085** 0.106** 0.118** 0.071** 0.041* 0.075** (0.015) (0.020) (0.019) (0.014) (0.020) (0.019) Price change 0.217** 0.273** 0.438** 0.367** 0.308** 0.280** (0.033) (0.042) (0.042) (0.034) (0.048) (0.056) Pop growth 0.127** 0.014 0.166** 0.166** 0.046 0.180** (0.038) (0.032) (0.047) (0.039) (0.051) (0.051) Log(BEA Inc) 0.114** 0.018 0.175** 0.124** 0.023 0.142** (0.038) (0.050) (0.041) (0.039) (0.047) (0.053) Non-recourse 0.626** 0.777** 0.712** 0.449** 0.858** 0.361** (0.060) (0.091) (0.099) (0.061) (0.097) (0.111) Origination year FE Yes Yes Yes Yes Yes Yes MSA FE No No No No No No Lender FE No No No No No No Observations 3,279,219 979,440 5,214,750 826,604 929,451 7,545,583 Table V. Multinomial logit regressions for subsamples This table reports the coefficients of multinomial logit regressions of complex mortgages (CM) relative to fixed-rate mortgages (FRM) for the following subsamples: loans with full documentation; loans to borrowers with FICO scores below 620; loans originated to purchase a new house; loans used to finance an investment property; non-securitized loans; and loans originated in states other than California. The significance levels are abbreviated with asterisks: One and two asterisks denote significance at the 5% and 1% level, respectively. Full Subprime Purchases Invest Not Exclude Document Only Only Property Securitized California Log(Inc) 0.418** 0.366** 0.429** 0.408** 0.696** 0.456** (0.016) (0.023) (0.015) (0.018) (0.022) (0.016) FICO −0.146** 0.023 −0.086** −0.097** −0.086** −0.019 (0.010) (0.029) (0.011) (0.011) (0.017) (0.011) LTV 0.370** −0.037 0.162** 0.429** 0.381** 0.251** (0.029) (0.040) (0.034) (0.019) (0.026) (0.024) VTI 0.342** 0.167** 0.337** 0.347** 0.419** 0.357** (0.024) (0.025) (0.022) (0.041) (0.039) (0.031) Low Doc 0.126 0.580** −0.152** 2.143** 0.666** (0.077) (0.049) (0.033) (0.128) (0.045) Above limit 1.055** 1.775** 1.183** 0.973** 0.731** 1.034** (0.055) (0.115) (0.067) (0.052) (0.069) (0.044) Condo 0.450** 0.487** 0.459** 0.173** 0.350** 0.436** (0.037) (0.064) (0.044) (0.035) (0.053) (0.048) Investment 0.042 −0.140* 0.294** 0.033 0.289** (0.030) (0.068) (0.033) (0.042) (0.030) Refinance 0.093* 0.267** 0.030 0.423** 0.247** (0.038) (0.034) (0.033) (0.057) (0.040) College 0.051* −0.057* 0.040* 0.062** 0.062** 0.073** (0.022) (0.027) (0.020) (0.015) (0.024) (0.018) Young 0.085** 0.106** 0.118** 0.071** 0.041* 0.075** (0.015) (0.020) (0.019) (0.014) (0.020) (0.019) Price change 0.217** 0.273** 0.438** 0.367** 0.308** 0.280** (0.033) (0.042) (0.042) (0.034) (0.048) (0.056) Pop growth 0.127** 0.014 0.166** 0.166** 0.046 0.180** (0.038) (0.032) (0.047) (0.039) (0.051) (0.051) Log(BEA Inc) 0.114** 0.018 0.175** 0.124** 0.023 0.142** (0.038) (0.050) (0.041) (0.039) (0.047) (0.053) Non-recourse 0.626** 0.777** 0.712** 0.449** 0.858** 0.361** (0.060) (0.091) (0.099) (0.061) (0.097) (0.111) Origination year FE Yes Yes Yes Yes Yes Yes MSA FE No No No No No No Lender FE No No No No No No Observations 3,279,219 979,440 5,214,750 826,604 929,451 7,545,583 Full Subprime Purchases Invest Not Exclude Document Only Only Property Securitized California Log(Inc) 0.418** 0.366** 0.429** 0.408** 0.696** 0.456** (0.016) (0.023) (0.015) (0.018) (0.022) (0.016) FICO −0.146** 0.023 −0.086** −0.097** −0.086** −0.019 (0.010) (0.029) (0.011) (0.011) (0.017) (0.011) LTV 0.370** −0.037 0.162** 0.429** 0.381** 0.251** (0.029) (0.040) (0.034) (0.019) (0.026) (0.024) VTI 0.342** 0.167** 0.337** 0.347** 0.419** 0.357** (0.024) (0.025) (0.022) (0.041) (0.039) (0.031) Low Doc 0.126 0.580** −0.152** 2.143** 0.666** (0.077) (0.049) (0.033) (0.128) (0.045) Above limit 1.055** 1.775** 1.183** 0.973** 0.731** 1.034** (0.055) (0.115) (0.067) (0.052) (0.069) (0.044) Condo 0.450** 0.487** 0.459** 0.173** 0.350** 0.436** (0.037) (0.064) (0.044) (0.035) (0.053) (0.048) Investment 0.042 −0.140* 0.294** 0.033 0.289** (0.030) (0.068) (0.033) (0.042) (0.030) Refinance 0.093* 0.267** 0.030 0.423** 0.247** (0.038) (0.034) (0.033) (0.057) (0.040) College 0.051* −0.057* 0.040* 0.062** 0.062** 0.073** (0.022) (0.027) (0.020) (0.015) (0.024) (0.018) Young 0.085** 0.106** 0.118** 0.071** 0.041* 0.075** (0.015) (0.020) (0.019) (0.014) (0.020) (0.019) Price change 0.217** 0.273** 0.438** 0.367** 0.308** 0.280** (0.033) (0.042) (0.042) (0.034) (0.048) (0.056) Pop growth 0.127** 0.014 0.166** 0.166** 0.046 0.180** (0.038) (0.032) (0.047) (0.039) (0.051) (0.051) Log(BEA Inc) 0.114** 0.018 0.175** 0.124** 0.023 0.142** (0.038) (0.050) (0.041) (0.039) (0.047) (0.053) Non-recourse 0.626** 0.777** 0.712** 0.449** 0.858** 0.361** (0.060) (0.091) (0.099) (0.061) (0.097) (0.111) Origination year FE Yes Yes Yes Yes Yes Yes MSA FE No No No No No No Lender FE No No No No No No Observations 3,279,219 979,440 5,214,750 826,604 929,451 7,545,583 The LPS database undersamples subprime loans and portfolio loans. Table V shows that the income level remains an important predictor of CM if we focus on subprime mortgages (i.e., mortgages with FICO scores below 620) or on loans that are not securitized. Table V also shows that our results remain materially unaffected if we only study purchase transactions or investment properties. Finally, we obtain similar results if we exclude all mortgages originated in the state of California, which accounts for around 15% of our observations but a greater proportion of the CM loans. In unreported robustness tests, we run separate multinomial logit models for each year and document that the determinants of mortgage choice are relatively stable over time. For example, the income level is positively related to the choice of CM for each year in our sample. 4. Mortgage Delinquencies In this section, we study the delinquency outcomes of different types of mortgages. A mortgage is considered delinquent if the borrower is at least 60 days late with a payment. 4.1 Summary of Mortgage Delinquency Figure 5 plots the distribution of mortgage delinquencies by contract type during the first five years after origination. In each month, we depict the proportion of remaining mortgages that become delinquent for the first time. Figure 5. View largeDownload slide Proportion of mortgage delinquencies by month after origination. The figure depicts the proportion of surviving loans that are delinquent by month after origination for fixed-rate (FRM), adjustable-rate (ARM), interest-only (IO), and negative-amortization mortgages (NEGAM) over the period between 2003 and 2009. Figure 5. View largeDownload slide Proportion of mortgage delinquencies by month after origination. The figure depicts the proportion of surviving loans that are delinquent by month after origination for fixed-rate (FRM), adjustable-rate (ARM), interest-only (IO), and negative-amortization mortgages (NEGAM) over the period between 2003 and 2009. We observe that IO and NEGAM have relatively low delinquency rates for the first months due to low initial mortgage payments. Mortgage delinquencies of IO (NEGAM) mortgages then steadily increase and reach peaks of about 1.3% (2.0%) of surviving loans at 27 (39) months after origination. These peaks occur three months after common reset intervals, since delinquency begins when a mortgage payment is at least 60 days late. We observe a similar peak for ARMs at the 27-month horizon. The delinquencies of FRMs are substantially lower than the delinquencies of CM, except for short horizons. Whereas ARMs have higher rates of delinquency at short horizons, CMs overtake them at longer horizons. It must be kept in mind that CM borrowers have higher delinquency propensities at longer horizons despite having better credit scores than ARM borrowers, as summarized in Table II. Moreover, the delinquency rate increases substantially even before the minimum loan payments are reset after two or three years, indicating that some CM borrowers are stretching their borrowing capacity beyond affordable levels. They do not even make the relatively low initial mortgage payments. 4.2 Hazard Model of Delinquency To investigate the determinants of mortgage delinquencies, we run the following Cox proportional hazard model: h(i,t)=h0(t;s,v)eβXi,t+FEtYear, (2) where the hazard rate h(t) is the estimated probability of first time 60-day delinquency at time t conditional on surviving to time t−, h0(t) is the baseline hazard rate, X is a vector of household-specific covariates, and FEtYear is an indicator variable for the calendar year to control for different vintage effects and macroeconomic conditions. We allow the baseline hazard to vary for each combination of the origination year v and the state of origination s or for each combination of the origination year v and the lender s.18 The loan sample is expanded to a loan-year level so that time-varying covariates can be included. Also, time is scaled so that the first observation date is the calendar year of origination (time 0), and subsequent calendar years are measured relative to the year of origination. Implicitly, loans of different vintages are compared with each other, so that the baseline hazard represents the probability of delinquency for a borrower with covariates of 0 at t years after origination. In some specifications, we separate CM into the two sub-types (IO and NEGAM). The continuous covariates are again standardized by subtracting the mean and dividing by the standard deviation. Table VI reports the estimated coefficients of the propensity of first time delinquency. In the first column, we only use borrower and loan characteristics at origination to estimate the delinquency hazards. In the second column, we include area-specific variables and time-varying loan and area characteristics. The third column incorporates controls for loan ownership to explore the impact of securitization. The fourth column replaces the year-state baseline with the year-lender baseline to control for lender-specific determinants of delinquency. The last column decomposes CMs into IOs and NEGAMs using year-state baselines. Table VI. Hazard model of mortgage delinquency This table reports the estimates from a Cox proportional hazard model for mortgage delinquency. The significance levels are abbreviated with asterisks: One and two asterisks denote significance at the 5% and 1% level, respectively. Individual-level MSA-level Securitization Lender-year Detailed CM Covariates Covariates Controls Baselines Contracts CM 0.723** 0.699** 0.557** 0.641** (0.011) (0.012) (0.012) (0.010) IO 0.659** (0.013) NEGAM 0.939** (0.020) ARM 0.444** 0.457** 0.303** 0.418** 0.461** (0.009) (0.010) (0.013) (0.008) (0.010) Log(Inc) −0.109** −0.058** −0.060** −0.032** −0.062** (0.008) (0.010) (0.009) (0.011) (0.010) FICO −0.672** −0.663** −0.634** −0.618** −0.664** (0.010) (0.012) (0.012) (0.012) (0.012) LTV 0.506** 0.492** 0.504** 0.458** 0.492** (0.012) (0.011) (0.010) (0.010) (0.011) VTI 0.045** 0.050** 0.051** 0.061** 0.050** (0.006) (0.006) (0.006) (0.006) (0.005) Low Doc 0.046** 0.042** 0.089** 0.016 0.028* (0.012) (0.012) (0.012) (0.011) (0.012) Above limit 0.187** 0.279** 0.129** 0.294** 0.271** (0.030) (0.018) (0.018) (0.022) (0.018) Condo −0.173** −0.092** −0.083** −0.024 −0.090** (0.029) (0.025) (0.024) (0.028) (0.025) Investment 0.348** 0.332** 0.294** 0.280** 0.329** (0.026) (0.027) (0.024) (0.031) (0.027) Refinance 0.089** 0.043** 0.015 0.003 0.034** (0.011) (0.012) (0.011) (0.009) (0.012) College −0.203** −0.195** −0.198** −0.201** (0.009) (0.009) (0.008) (0.009) Young 0.017* 0.016* −0.008 0.017* (0.007) (0.007) (0.007) (0.007) Log(BEA Inc) 0.038** 0.038** 0.041* 0.037** (0.014) (0.014) (0.016) (0.014) House price growth −0.414** −0.416** −0.450** −0.415** (0.020) (0.019) (0.015) (0.020) Loan balance growth 0.016 0.018 0.061** −0.002 (0.012) (0.011) (0.009) (0.012) Payment resets 0.031** 0.029** 0.051** 0.025** (0.002) (0.002) (0.002) (0.002) Change in unemployment 0.022** 0.023** 0.026** 0.022** (0.002) (0.002) (0.002) (0.002) Income growth −0.149** −0.144** −0.134** −0.151** (0.024) (0.024) (0.038) (0.024) Gov securitized −0.164** (0.015) Priv securitized 0.246** (0.010) State-year baselines Yes Yes Yes No Yes Lender-year baselines No No No Yes No Observations 29,690,919 23,151,373 23,151,373 23,151,373 23,151,373 Individual-level MSA-level Securitization Lender-year Detailed CM Covariates Covariates Controls Baselines Contracts CM 0.723** 0.699** 0.557** 0.641** (0.011) (0.012) (0.012) (0.010) IO 0.659** (0.013) NEGAM 0.939** (0.020) ARM 0.444** 0.457** 0.303** 0.418** 0.461** (0.009) (0.010) (0.013) (0.008) (0.010) Log(Inc) −0.109** −0.058** −0.060** −0.032** −0.062** (0.008) (0.010) (0.009) (0.011) (0.010) FICO −0.672** −0.663** −0.634** −0.618** −0.664** (0.010) (0.012) (0.012) (0.012) (0.012) LTV 0.506** 0.492** 0.504** 0.458** 0.492** (0.012) (0.011) (0.010) (0.010) (0.011) VTI 0.045** 0.050** 0.051** 0.061** 0.050** (0.006) (0.006) (0.006) (0.006) (0.005) Low Doc 0.046** 0.042** 0.089** 0.016 0.028* (0.012) (0.012) (0.012) (0.011) (0.012) Above limit 0.187** 0.279** 0.129** 0.294** 0.271** (0.030) (0.018) (0.018) (0.022) (0.018) Condo −0.173** −0.092** −0.083** −0.024 −0.090** (0.029) (0.025) (0.024) (0.028) (0.025) Investment 0.348** 0.332** 0.294** 0.280** 0.329** (0.026) (0.027) (0.024) (0.031) (0.027) Refinance 0.089** 0.043** 0.015 0.003 0.034** (0.011) (0.012) (0.011) (0.009) (0.012) College −0.203** −0.195** −0.198** −0.201** (0.009) (0.009) (0.008) (0.009) Young 0.017* 0.016* −0.008 0.017* (0.007) (0.007) (0.007) (0.007) Log(BEA Inc) 0.038** 0.038** 0.041* 0.037** (0.014) (0.014) (0.016) (0.014) House price growth −0.414** −0.416** −0.450** −0.415** (0.020) (0.019) (0.015) (0.020) Loan balance growth 0.016 0.018 0.061** −0.002 (0.012) (0.011) (0.009) (0.012) Payment resets 0.031** 0.029** 0.051** 0.025** (0.002) (0.002) (0.002) (0.002) Change in unemployment 0.022** 0.023** 0.026** 0.022** (0.002) (0.002) (0.002) (0.002) Income growth −0.149** −0.144** −0.134** −0.151** (0.024) (0.024) (0.038) (0.024) Gov securitized −0.164** (0.015) Priv securitized 0.246** (0.010) State-year baselines Yes Yes Yes No Yes Lender-year baselines No No No Yes No Observations 29,690,919 23,151,373 23,151,373 23,151,373 23,151,373 Table VI. Hazard model of mortgage delinquency This table reports the estimates from a Cox proportional hazard model for mortgage delinquency. The significance levels are abbreviated with asterisks: One and two asterisks denote significance at the 5% and 1% level, respectively. Individual-level MSA-level Securitization Lender-year Detailed CM Covariates Covariates Controls Baselines Contracts CM 0.723** 0.699** 0.557** 0.641** (0.011) (0.012) (0.012) (0.010) IO 0.659** (0.013) NEGAM 0.939** (0.020) ARM 0.444** 0.457** 0.303** 0.418** 0.461** (0.009) (0.010) (0.013) (0.008) (0.010) Log(Inc) −0.109** −0.058** −0.060** −0.032** −0.062** (0.008) (0.010) (0.009) (0.011) (0.010) FICO −0.672** −0.663** −0.634** −0.618** −0.664** (0.010) (0.012) (0.012) (0.012) (0.012) LTV 0.506** 0.492** 0.504** 0.458** 0.492** (0.012) (0.011) (0.010) (0.010) (0.011) VTI 0.045** 0.050** 0.051** 0.061** 0.050** (0.006) (0.006) (0.006) (0.006) (0.005) Low Doc 0.046** 0.042** 0.089** 0.016 0.028* (0.012) (0.012) (0.012) (0.011) (0.012) Above limit 0.187** 0.279** 0.129** 0.294** 0.271** (0.030) (0.018) (0.018) (0.022) (0.018) Condo −0.173** −0.092** −0.083** −0.024 −0.090** (0.029) (0.025) (0.024) (0.028) (0.025) Investment 0.348** 0.332** 0.294** 0.280** 0.329** (0.026) (0.027) (0.024) (0.031) (0.027) Refinance 0.089** 0.043** 0.015 0.003 0.034** (0.011) (0.012) (0.011) (0.009) (0.012) College −0.203** −0.195** −0.198** −0.201** (0.009) (0.009) (0.008) (0.009) Young 0.017* 0.016* −0.008 0.017* (0.007) (0.007) (0.007) (0.007) Log(BEA Inc) 0.038** 0.038** 0.041* 0.037** (0.014) (0.014) (0.016) (0.014) House price growth −0.414** −0.416** −0.450** −0.415** (0.020) (0.019) (0.015) (0.020) Loan balance growth 0.016 0.018 0.061** −0.002 (0.012) (0.011) (0.009) (0.012) Payment resets 0.031** 0.029** 0.051** 0.025** (0.002) (0.002) (0.002) (0.002) Change in unemployment 0.022** 0.023** 0.026** 0.022** (0.002) (0.002) (0.002) (0.002) Income growth −0.149** −0.144** −0.134** −0.151** (0.024) (0.024) (0.038) (0.024) Gov securitized −0.164** (0.015) Priv securitized 0.246** (0.010) State-year baselines Yes Yes Yes No Yes Lender-year baselines No No No Yes No Observations 29,690,919 23,151,373 23,151,373 23,151,373 23,151,373 Individual-level MSA-level Securitization Lender-year Detailed CM Covariates Covariates Controls Baselines Contracts CM 0.723** 0.699** 0.557** 0.641** (0.011) (0.012) (0.012) (0.010) IO 0.659** (0.013) NEGAM 0.939** (0.020) ARM 0.444** 0.457** 0.303** 0.418** 0.461** (0.009) (0.010) (0.013) (0.008) (0.010) Log(Inc) −0.109** −0.058** −0.060** −0.032** −0.062** (0.008) (0.010) (0.009) (0.011) (0.010) FICO −0.672** −0.663** −0.634** −0.618** −0.664** (0.010) (0.012) (0.012) (0.012) (0.012) LTV 0.506** 0.492** 0.504** 0.458** 0.492** (0.012) (0.011) (0.010) (0.010) (0.011) VTI 0.045** 0.050** 0.051** 0.061** 0.050** (0.006) (0.006) (0.006) (0.006) (0.005) Low Doc 0.046** 0.042** 0.089** 0.016 0.028* (0.012) (0.012) (0.012) (0.011) (0.012) Above limit 0.187** 0.279** 0.129** 0.294** 0.271** (0.030) (0.018) (0.018) (0.022) (0.018) Condo −0.173** −0.092** −0.083** −0.024 −0.090** (0.029) (0.025) (0.024) (0.028) (0.025) Investment 0.348** 0.332** 0.294** 0.280** 0.329** (0.026) (0.027) (0.024) (0.031) (0.027) Refinance 0.089** 0.043** 0.015 0.003 0.034** (0.011) (0.012) (0.011) (0.009) (0.012) College −0.203** −0.195** −0.198** −0.201** (0.009) (0.009) (0.008) (0.009) Young 0.017* 0.016* −0.008 0.017* (0.007) (0.007) (0.007) (0.007) Log(BEA Inc) 0.038** 0.038** 0.041* 0.037** (0.014) (0.014) (0.016) (0.014) House price growth −0.414** −0.416** −0.450** −0.415** (0.020) (0.019) (0.015) (0.020) Loan balance growth 0.016 0.018 0.061** −0.002 (0.012) (0.011) (0.009) (0.012) Payment resets 0.031** 0.029** 0.051** 0.025** (0.002) (0.002) (0.002) (0.002) Change in unemployment 0.022** 0.023** 0.026** 0.022** (0.002) (0.002) (0.002) (0.002) Income growth −0.149** −0.144** −0.134** −0.151** (0.024) (0.024) (0.038) (0.024) Gov securitized −0.164** (0.015) Priv securitized 0.246** (0.010) State-year baselines Yes Yes Yes No Yes Lender-year baselines No No No Yes No Observations 29,690,919 23,151,373 23,151,373 23,151,373 23,151,373 Our key finding is that CMs have significantly higher delinquency rates than FRMs in all specifications, notwithstanding a wide array of control variables. The effect is both economically and statistically significant. For example, in column 1, the coefficient of 0.723 for CM implies that the probability of delinquency for a borrower with a CM is about twice as high as for a fixed-rate borrower, holding all other characteristics fixed (⁠ e1×0.723/e0×0.723=2.06 ⁠). This impact of having a CM on mortgage delinquency is similar to a one-standard-deviation decrease in the FICO credit score, which is generally perceived to be a strong predictor of mortgage delinquency. ARMs also exhibit higher delinquency rates than FRMs. This result is primarily caused by the lower credit quality of ARM borrowers.19 The first set of additional explanatory variables in column 2 is related to shocks in cash flows. Of particular interest is the variable “Payment Resets,” defined as the increase in the minimum required mortgage payment since origination. Recall that payment resets are driven by interest rate changes for ARMs and by both interest rate and amortization changes for CMs. Consequently, CMs have larger resets than ARMs, as illustrated by the CDFs of payments over time in Figure 3. We find that payment resets increase the hazard rate of delinquency. For example, the quartile of CM borrowers with the highest resets after five years experience required average payment increases of 38% (which is 0.38/0.0567 = 6.7 standard deviations away from the mean). The estimates of the hazard model imply that such an increase in the required payment increases the delinquency rate by around 23% (⁠ e0.031×6.7−1=0.23 ⁠).20 Furthermore, we do not find a significant impact of an increase in the loan balance since origination on mortgage delinquency in the base case with state-year baselines.21 In sum, these results suggest that contract-driven resets and amortization changes partially contribute to the higher defaults of CMs. Other variables related to cash flow defaults include the income level and the FICO score, which partly reflect households’ financial conditions. Higher income and higher FICO households are less constrained and are indeed found to have lower delinquency rates. To gauge the impact of local macro-economic conditions on mortgage delinquency, we include the change in the unemployment rate in an MSA over the last year, and the income growth rate, defined as the growth rate of the mean income level at the MSA level since the mortgage was originated. Increases in local income growth rates and decreases in local unemployment rates significantly reduce mortgage delinquency. The second set of explanatory variables is related to the economic costs and benefits of default. Since households can always sell their house and pay off their mortgage in full when the remaining loan balance is low relative to the current house value, it is not surprising that higher LTV ratios at origination are associated with higher delinquency hazards. Both the LTV at origination and the subsequent change in house prices are significant drivers of mortgage delinquency. From column 2 of Table VI, a one-standard-deviation increase in LTV can increase delinquency by e0.492−1=64% and a one standard deviation drop in house prices can increase delinquency by e0.414−1=51% of the base default rate. Owners of investment properties might have lower non-monetary costs of defaulting than owners of owner-occupied homes. Indeed, our results indicate that delinquency rates of investment properties are significantly higher. Notably, the inclusion of these controls preserve the effect of contract choice, as the coefficient on the CM indicator variable remains practically unchanged. We control in column 3 for whether the mortgage was securitized by Government Sponsored Entities or by private parties. Since the impact of securitization has received significant attention in the literature, we want to ensure that the impact of CM is not subsumed by the lenders’ propensity to securitize. We find that CM are still associated with higher delinquency hazards after controlling for government and private securitization. Thus, the role of mortgage contract design is distinct from the well-documented impact of securitization. The fourth specification uses lender-year baseline hazards instead of state-year baselines, which accounts for the possibility that mortgages originated by different lenders exhibit different delinquency rates over time because individual lenders attract a particular borrower type that might focus on specific mortgage contracts. CM exhibit higher delinquency rates under all specifications. The last column of Table VI separates IO and NEGAM loans and indicates that the delinquency rates for negative amortization loans are larger than the delinquency rates for the more conservative IO loans. For example, an IO mortgage has a propensity to be delinquent that is around twice the propensity for a FRM. In contrast, a NEGAM has a propensity to default that is about 2.5 times higher than a FRM. Table VII shows that CM borrowers exhibit higher delinquency rates than borrowers of FRM for the subsamples of full documentation loans, for subprime loans, for purchase transactions, for investment properties, for non-securitized loans, and for loans not originated in California. Table VII. Hazard model of mortgage delinquency for subsamples This table reports the estimates from a Cox proportional hazard model for mortgage delinquency for the following subsamples: loans with full documentation; loans to borrowers with FICO scores below 620; loans originated to purchase a new house; loans used to finance an investment property; non-securitized loans; and loans originated in states other than California. The significance levels are abbreviated with asterisks: One and two asterisks denote significance at the 5% and 1% level, respectively. Full Subprime Purchases Invest Not Exclude Document Only Only Property Securitized California CM 0.561** 0.466** 0.779** 0.658** 0.368** 0.689** (0.012) (0.015) (0.015) (0.016) (0.022) (0.015) ARM 0.454** 0.348** 0.505** 0.342** 0.097** 0.434** (0.011) (0.008) (0.011) (0.017) (0.019) (0.009) Log(Inc) −0.112** −0.083** −0.085** 0.035** −0.042** −0.072** (0.010) (0.007) (0.010) (0.008) (0.009) (0.010) FICO −0.712** −0.339** −0.679** −0.643** −0.705** −0.683** (0.012) (0.007) (0.012) (0.015) (0.016) (0.012) LTV 0.467** 0.293** 0.433** 0.621** 0.483** 0.463** (0.012) (0.008) (0.008) (0.012) (0.012) (0.009) VTI 0.050** 0.086** 0.051** 0.069** 0.050** 0.078** (0.006) (0.007) (0.009) (0.011) (0.008) (0.007) Low Doc −0.021** 0.076** −0.078** 0.243** 0.042** (0.008) (0.014) (0.012) (0.017) (0.011) Above limit 0.259** 0.250** 0.236** 0.057* 0.331** 0.326** (0.019) (0.021) (0.023) (0.022) (0.023) (0.024) Condo −0.081** −0.071** −0.103** −0.172** −0.132** −0.105** (0.025) (0.016) (0.027) (0.028) (0.032) (0.031) Investment 0.331** 0.154** 0.281** 0.376** 0.407** (0.024) (0.015) (0.031) (0.034) (0.024) Refinance 0.018 −0.031** 0.233** 0.043* 0.064** (0.012) (0.011) (0.011) (0.019) (0.012) College −0.177** −0.091** −0.241** −0.216** −0.217** −0.186** (0.008) (0.004) (0.011) (0.011) (0.011) (0.008) Young 0.012 −0.006 0.026** 0.055** 0.002 0.013 (0.007) (0.005) (0.009) (0.010) (0.009) (0.008) Log(BEA Inc) 0.043** 0.035** 0.047** 0.025 0.007 0.067** (0.014) (0.007) (0.015) (0.015) (0.016) (0.014) House price growth −0.420** −0.353** −0.425** −0.380** −0.399** −0.416** (0.018) (0.019) (0.023) (0.022) (0.025) (0.027) Loan Balance Growth −0.043** −0.046** −0.034** 0.119** 0.026 0.036* (0.011) (0.009) (0.011) (0.013) (0.019) (0.015) Payment Resets 0.050** 0.027** 0.046** 0.070** 0.004 0.033** (0.002) (0.003) (0.003) (0.003) (0.004) (0.002) Change in Unemployment 0.023** 0.029** 0.023** 0.015** 0.015** 0.025** (0.002) (0.003) (0.002) (0.003) (0.004) (0.002) Income Growth −0.141** −0.059** −0.155** −0.146** −0.154** −0.127** (0.023) (0.020) (0.025) (0.033) (0.034) (0.022) Observations 8,595,153 1,751,099 13,672,439 2,259,834 2,043,822 19,681,780 Full Subprime Purchases Invest Not Exclude Document Only Only Property Securitized California CM 0.561** 0.466** 0.779** 0.658** 0.368** 0.689** (0.012) (0.015) (0.015) (0.016) (0.022) (0.015) ARM 0.454** 0.348** 0.505** 0.342** 0.097** 0.434** (0.011) (0.008) (0.011) (0.017) (0.019) (0.009) Log(Inc) −0.112** −0.083** −0.085** 0.035** −0.042** −0.072** (0.010) (0.007) (0.010) (0.008) (0.009) (0.010) FICO −0.712** −0.339** −0.679** −0.643** −0.705** −0.683** (0.012) (0.007) (0.012) (0.015) (0.016) (0.012) LTV 0.467** 0.293** 0.433** 0.621** 0.483** 0.463** (0.012) (0.008) (0.008) (0.012) (0.012) (0.009) VTI 0.050** 0.086** 0.051** 0.069** 0.050** 0.078** (0.006) (0.007) (0.009) (0.011) (0.008) (0.007) Low Doc −0.021** 0.076** −0.078** 0.243** 0.042** (0.008) (0.014) (0.012) (0.017) (0.011) Above limit 0.259** 0.250** 0.236** 0.057* 0.331** 0.326** (0.019) (0.021) (0.023) (0.022) (0.023) (0.024) Condo −0.081** −0.071** −0.103** −0.172** −0.132** −0.105** (0.025) (0.016) (0.027) (0.028) (0.032) (0.031) Investment 0.331** 0.154** 0.281** 0.376** 0.407** (0.024) (0.015) (0.031) (0.034) (0.024) Refinance 0.018 −0.031** 0.233** 0.043* 0.064** (0.012) (0.011) (0.011) (0.019) (0.012) College −0.177** −0.091** −0.241** −0.216** −0.217** −0.186** (0.008) (0.004) (0.011) (0.011) (0.011) (0.008) Young 0.012 −0.006 0.026** 0.055** 0.002 0.013 (0.007) (0.005) (0.009) (0.010) (0.009) (0.008) Log(BEA Inc) 0.043** 0.035** 0.047** 0.025 0.007 0.067** (0.014) (0.007) (0.015) (0.015) (0.016) (0.014) House price growth −0.420** −0.353** −0.425** −0.380** −0.399** −0.416** (0.018) (0.019) (0.023) (0.022) (0.025) (0.027) Loan Balance Growth −0.043** −0.046** −0.034** 0.119** 0.026 0.036* (0.011) (0.009) (0.011) (0.013) (0.019) (0.015) Payment Resets 0.050** 0.027** 0.046** 0.070** 0.004 0.033** (0.002) (0.003) (0.003) (0.003) (0.004) (0.002) Change in Unemployment 0.023** 0.029** 0.023** 0.015** 0.015** 0.025** (0.002) (0.003) (0.002) (0.003) (0.004) (0.002) Income Growth −0.141** −0.059** −0.155** −0.146** −0.154** −0.127** (0.023) (0.020) (0.025) (0.033) (0.034) (0.022) Observations 8,595,153 1,751,099 13,672,439 2,259,834 2,043,822 19,681,780 Table VII. Hazard model of mortgage delinquency for subsamples This table reports the estimates from a Cox proportional hazard model for mortgage delinquency for the following subsamples: loans with full documentation; loans to borrowers with FICO scores below 620; loans originated to purchase a new house; loans used to finance an investment property; non-securitized loans; and loans originated in states other than California. The significance levels are abbreviated with asterisks: One and two asterisks denote significance at the 5% and 1% level, respectively. Full Subprime Purchases Invest Not Exclude Document Only Only Property Securitized California CM 0.561** 0.466** 0.779** 0.658** 0.368** 0.689** (0.012) (0.015) (0.015) (0.016) (0.022) (0.015) ARM 0.454** 0.348** 0.505** 0.342** 0.097** 0.434** (0.011) (0.008) (0.011) (0.017) (0.019) (0.009) Log(Inc) −0.112** −0.083** −0.085** 0.035** −0.042** −0.072** (0.010) (0.007) (0.010) (0.008) (0.009) (0.010) FICO −0.712** −0.339** −0.679** −0.643** −0.705** −0.683** (0.012) (0.007) (0.012) (0.015) (0.016) (0.012) LTV 0.467** 0.293** 0.433** 0.621** 0.483** 0.463** (0.012) (0.008) (0.008) (0.012) (0.012) (0.009) VTI 0.050** 0.086** 0.051** 0.069** 0.050** 0.078** (0.006) (0.007) (0.009) (0.011) (0.008) (0.007) Low Doc −0.021** 0.076** −0.078** 0.243** 0.042** (0.008) (0.014) (0.012) (0.017) (0.011) Above limit 0.259** 0.250** 0.236** 0.057* 0.331** 0.326** (0.019) (0.021) (0.023) (0.022) (0.023) (0.024) Condo −0.081** −0.071** −0.103** −0.172** −0.132** −0.105** (0.025) (0.016) (0.027) (0.028) (0.032) (0.031) Investment 0.331** 0.154** 0.281** 0.376** 0.407** (0.024) (0.015) (0.031) (0.034) (0.024) Refinance 0.018 −0.031** 0.233** 0.043* 0.064** (0.012) (0.011) (0.011) (0.019) (0.012) College −0.177** −0.091** −0.241** −0.216** −0.217** −0.186** (0.008) (0.004) (0.011) (0.011) (0.011) (0.008) Young 0.012 −0.006 0.026** 0.055** 0.002 0.013 (0.007) (0.005) (0.009) (0.010) (0.009) (0.008) Log(BEA Inc) 0.043** 0.035** 0.047** 0.025 0.007 0.067** (0.014) (0.007) (0.015) (0.015) (0.016) (0.014) House price growth −0.420** −0.353** −0.425** −0.380** −0.399** −0.416** (0.018) (0.019) (0.023) (0.022) (0.025) (0.027) Loan Balance Growth −0.043** −0.046** −0.034** 0.119** 0.026 0.036* (0.011) (0.009) (0.011) (0.013) (0.019) (0.015) Payment Resets 0.050** 0.027** 0.046** 0.070** 0.004 0.033** (0.002) (0.003) (0.003) (0.003) (0.004) (0.002) Change in Unemployment 0.023** 0.029** 0.023** 0.015** 0.015** 0.025** (0.002) (0.003) (0.002) (0.003) (0.004) (0.002) Income Growth −0.141** −0.059** −0.155** −0.146** −0.154** −0.127** (0.023) (0.020) (0.025) (0.033) (0.034) (0.022) Observations 8,595,153 1,751,099 13,672,439 2,259,834 2,043,822 19,681,780 Full Subprime Purchases Invest Not Exclude Document Only Only Property Securitized California CM 0.561** 0.466** 0.779** 0.658** 0.368** 0.689** (0.012) (0.015) (0.015) (0.016) (0.022) (0.015) ARM 0.454** 0.348** 0.505** 0.342** 0.097** 0.434** (0.011) (0.008) (0.011) (0.017) (0.019) (0.009) Log(Inc) −0.112** −0.083** −0.085** 0.035** −0.042** −0.072** (0.010) (0.007) (0.010) (0.008) (0.009) (0.010) FICO −0.712** −0.339** −0.679** −0.643** −0.705** −0.683** (0.012) (0.007) (0.012) (0.015) (0.016) (0.012) LTV 0.467** 0.293** 0.433** 0.621** 0.483** 0.463** (0.012) (0.008) (0.008) (0.012) (0.012) (0.009) VTI 0.050** 0.086** 0.051** 0.069** 0.050** 0.078** (0.006) (0.007) (0.009) (0.011) (0.008) (0.007) Low Doc −0.021** 0.076** −0.078** 0.243** 0.042** (0.008) (0.014) (0.012) (0.017) (0.011) Above limit 0.259** 0.250** 0.236** 0.057* 0.331** 0.326** (0.019) (0.021) (0.023) (0.022) (0.023) (0.024) Condo −0.081** −0.071** −0.103** −0.172** −0.132** −0.105** (0.025) (0.016) (0.027) (0.028) (0.032) (0.031) Investment 0.331** 0.154** 0.281** 0.376** 0.407** (0.024) (0.015) (0.031) (0.034) (0.024) Refinance 0.018 −0.031** 0.233** 0.043* 0.064** (0.012) (0.011) (0.011) (0.019) (0.012) College −0.177** −0.091** −0.241** −0.216** −0.217** −0.186** (0.008) (0.004) (0.011) (0.011) (0.011) (0.008) Young 0.012 −0.006 0.026** 0.055** 0.002 0.013 (0.007) (0.005) (0.009) (0.010) (0.009) (0.008) Log(BEA Inc) 0.043** 0.035** 0.047** 0.025 0.007 0.067** (0.014) (0.007) (0.015) (0.015) (0.016) (0.014) House price growth −0.420** −0.353** −0.425** −0.380** −0.399** −0.416** (0.018) (0.019) (0.023) (0.022) (0.025) (0.027) Loan Balance Growth −0.043** −0.046** −0.034** 0.119** 0.026 0.036* (0.011) (0.009) (0.011) (0.013) (0.019) (0.015) Payment Resets 0.050** 0.027** 0.046** 0.070** 0.004 0.033** (0.002) (0.003) (0.003) (0.003) (0.004) (0.002) Change in Unemployment 0.023** 0.029** 0.023** 0.015** 0.015** 0.025** (0.002) (0.003) (0.002) (0.003) (0.004) (0.002) Income Growth −0.141** −0.059** −0.155** −0.146** −0.154** −0.127** (0.023) (0.020) (0.025) (0.033) (0.034) (0.022) Observations 8,595,153 1,751,099 13,672,439 2,259,834 2,043,822 19,681,780 In addition, we also run the hazard models separately for each annual origination cohort. The coefficients on CM are significantly positive for each individual origination cohort between 2003 and 2007. Furthermore, the remaining coefficients are generally consistent over the different cohorts. Overall, we find significantly higher default rates among CM borrowers, even after controlling for leverage, payment resets, and other household and loan characteristics. These results indicate that CM borrowers differ from other households in their risk taking and their willingness to default. 4.3 Interaction Effects To study the motivations for delinquency by CM borrowers, we conduct two additional tests. Our first test investigates whether CM borrowers are more sensitive to measures that capture the economic benefits of the default option. Defaulting is generally only beneficial when the loan value exceeds the home value by a sufficient amount to cover various costs of default (e.g., decline in credit rating, moving costs). Furthermore, the economic benefit of default increases with the LTV ratio, as the moneyness of the option to default increases. Table VIII shows that the delinquency rate of complex borrowers is particularly sensitive to the LTV ratio as evidenced by a strong positive coefficient on the interaction term between CM and LTV. The coefficient of 0.109 for the interaction term in column 1 suggests that the deterioration (i.e., increase) in the hazard rate from a one standard deviation increase in the LTV ratio is about 10.9% larger for a CM mortgage than for an FRM mortgage.22 Table VIII. Hazard model of mortgage delinquency with interaction effects This table reports the estimates from a Cox proportional hazard model for mortgage delinquency, with interaction effects that capture the sensitivity of complex mortgage delinquencies to other loan and household characteristics. The table also includes unreported individual- and MSA-level covariates as in the second column of Table VI. The significance levels are abbreviated with asterisks: One and two asterisks denote significance at the 5% and 1% level, respectively. Interactions with CM Interactions with IO/NEGAM CM 0.697** (0.013) IO 0.661** (0.014) NEGAM 0.914** (0.020) LTV 0.472** 0.470** (0.010) (0.010) CM x LTV 0.109** (0.020) IO × LTV 0.087** (0.020) NEGAM × LTV 0.232** (0.022) Log(Inc) −0.078** −0.078** (0.010) (0.010) CM x Log(Inc) 0.083** (0.009) IO × Log(Inc) 0.061** (0.009) NEGAM × Log(Inc) 0.115** (0.011) FICO −0.680** −0.681** (0.010) (0.010) CM × FICO 0.078** (0.012) IO × FICO 0.057** (0.013) NEGAM × FICO 0.211** (0.009) ARM 0.453** 0.455** (0.010) (0.010) State-year baselines Yes Yes Additional individual and MSA covariates Yes Yes Observations 23, 151, 373 23, 151, 373 Interactions with CM Interactions with IO/NEGAM CM 0.697** (0.013) IO 0.661** (0.014) NEGAM 0.914** (0.020) LTV 0.472** 0.470** (0.010) (0.010) CM x LTV 0.109** (0.020) IO × LTV 0.087** (0.020) NEGAM × LTV 0.232** (0.022) Log(Inc) −0.078** −0.078** (0.010) (0.010) CM x Log(Inc) 0.083** (0.009) IO × Log(Inc) 0.061** (0.009) NEGAM × Log(Inc) 0.115** (0.011) FICO −0.680** −0.681** (0.010) (0.010) CM × FICO 0.078** (0.012) IO × FICO 0.057** (0.013) NEGAM × FICO 0.211** (0.009) ARM 0.453** 0.455** (0.010) (0.010) State-year baselines Yes Yes Additional individual and MSA covariates Yes Yes Observations 23, 151, 373 23, 151, 373 Table VIII. Hazard model of mortgage delinquency with interaction effects This table reports the estimates from a Cox proportional hazard model for mortgage delinquency, with interaction effects that capture the sensitivity of complex mortgage delinquencies to other loan and household characteristics. The table also includes unreported individual- and MSA-level covariates as in the second column of Table VI. The significance levels are abbreviated with asterisks: One and two asterisks denote significance at the 5% and 1% level, respectively. Interactions with CM Interactions with IO/NEGAM CM 0.697** (0.013) IO 0.661** (0.014) NEGAM 0.914** (0.020) LTV 0.472** 0.470** (0.010) (0.010) CM x LTV 0.109** (0.020) IO × LTV 0.087** (0.020) NEGAM × LTV 0.232** (0.022) Log(Inc) −0.078** −0.078** (0.010) (0.010) CM x Log(Inc) 0.083** (0.009) IO × Log(Inc) 0.061** (0.009) NEGAM × Log(Inc) 0.115** (0.011) FICO −0.680** −0.681** (0.010) (0.010) CM × FICO 0.078** (0.012) IO × FICO 0.057** (0.013) NEGAM × FICO 0.211** (0.009) ARM 0.453** 0.455** (0.010) (0.010) State-year baselines Yes Yes Additional individual and MSA covariates Yes Yes Observations 23, 151, 373 23, 151, 373 Interactions with CM Interactions with IO/NEGAM CM 0.697** (0.013) IO 0.661** (0.014) NEGAM 0.914** (0.020) LTV 0.472** 0.470** (0.010) (0.010) CM x LTV 0.109** (0.020) IO × LTV 0.087** (0.020) NEGAM × LTV 0.232** (0.022) Log(Inc) −0.078** −0.078** (0.010) (0.010) CM x Log(Inc) 0.083** (0.009) IO × Log(Inc) 0.061** (0.009) NEGAM × Log(Inc) 0.115** (0.011) FICO −0.680** −0.681** (0.010) (0.010) CM × FICO 0.078** (0.012) IO × FICO 0.057** (0.013) NEGAM × FICO 0.211** (0.009) ARM 0.453** 0.455** (0.010) (0.010) State-year baselines Yes Yes Additional individual and MSA covariates Yes Yes Observations 23, 151, 373 23, 151, 373 Our second test compares the delinquency rates of CM borrowers with different levels of financial sophistication. We use the income level and the FICO score as proxies of financial sophistication following Campbell (2006) and Bucks and Pence (2008). Table VIII shows that while CM borrowers on average default more than traditional mortgage borrowers, the difference in the delinquency rates between complex and traditional borrowers is particularly high for households with higher income levels and with higher FICO credit scores. The last column of Table VIII shows that these interaction effects are more pronounced for NEGAM than for IO contracts, as should be expected given the fact that NEGAM are more complex instruments than IO mortgages. 4.4 Personal Bankruptcy To further investigate the characteristics of complex borrowers, we investigate the determinants of personal bankruptcy for mortgage borrowers. Contrasting the determinants of personal bankruptcy with the determinants of mortgage delinquency gives us important insights into the motivation of the delinquency behavior. It is not necessary that households who default on their mortgages also declare bankruptcy. Nor is it necessary that households who declare bankruptcy default on their mortgages. For example, in our sample only 12% of households who are delinquent on their mortgage also declare bankruptcy.23 Table IX reports the estimates of the bankruptcy hazard model. CM borrowers, and especially NEGAM borrowers, have higher propensities to declare bankruptcy, indicating that these households are more financially distressed than traditional mortgage borrowers. Most coefficients have the same signs as in the delinquency regression of Table VI, since both bankruptcy and mortgage delinquency are distress events for households and should be driven by similar fundamentals. For example, higher income and higher FICO scores reduce the propensities of both mortgage delinquency and bankruptcy. It is interesting that some variables show up with different signs in the two specifications. For example, although households with investment properties have significantly higher mortgage delinquency rates, they are not more likely to file for personal bankruptcy. This evidence suggests that owners of investment properties might be more likely to walk away from the property when it is economical to do so. Similarly, loans with low documentation are typically more likely to be delinquent on their mortgage debt, but these households do not experience higher bankruptcy rates. Table IX. Hazard models of personal bankruptcy This table reports the estimates from a Cox proportional hazard model for personal bankruptcy. The significance levels are abbreviated with asterisks: One and two asterisks denote significance at the 5% and 1% level, respectively. Individual-level MSA-level Individual-level MSA-level covariates covariates covariates covariates CM 0.667** 0.619** (0.011) (0.013) IO 0.602** 0.581** (0.012) (0.014) NEGAM 0.956** 0.818** (0.027) (0.026) ARM 0.389** 0.405** 0.391** 0.409** (0.014) (0.017) (0.014) (0.017) Log(Inc) −0.156** −0.097** −0.159** −0.099** (0.010) (0.012) (0.010) (0.012) FICO −0.466** −0.462** −0.467** −0.463** (0.008) (0.009) (0.008) (0.009) LTV 0.583** 0.551** 0.579** 0.551** (0.011) (0.013) (0.011) (0.013) VTI −0.213** −0.183** −0.211** −0.181** (0.017) (0.017) (0.017) (0.017) Low Doc 0.000 −0.009 −0.017* −0.020* (0.008) (0.009) (0.008) (0.009) Above limit 0.174** 0.276** 0.162** 0.267** (0.026) (0.029) (0.027) (0.029) Condo −0.304** −0.162** −0.300** −0.160** (0.027) (0.024) (0.027) (0.024) Investment 0.018 −0.010 0.015 −0.011 (0.024) (0.023) (0.024) (0.023) Refinance 0.405** 0.363** 0.392** 0.355** (0.012) (0.013) (0.012) (0.013) College −0.201** −0.200** (0.009) (0.009) Young −0.063** −0.063** (0.010) (0.010) Log(BEA Inc) −0.018 −0.019 (0.017) (0.017) House Price Growth −0.329** −0.328** (0.022) (0.022) Loan balance growth 0.093** 0.082** (0.008) (0.011) Payment resets 0.039** 0.034** (0.002) (0.002) Change in unemployment 0.021** 0.021** (0.003) (0.003) Income growth −0.179** −0.181** (0.026) (0.027) Observations 30,920,127 23,978,775 30,920,127 23,978,775 Individual-level MSA-level Individual-level MSA-level covariates covariates covariates covariates CM 0.667** 0.619** (0.011) (0.013) IO 0.602** 0.581** (0.012) (0.014) NEGAM 0.956** 0.818** (0.027) (0.026) ARM 0.389** 0.405** 0.391** 0.409** (0.014) (0.017) (0.014) (0.017) Log(Inc) −0.156** −0.097** −0.159** −0.099** (0.010) (0.012) (0.010) (0.012) FICO −0.466** −0.462** −0.467** −0.463** (0.008) (0.009) (0.008) (0.009) LTV 0.583** 0.551** 0.579** 0.551** (0.011) (0.013) (0.011) (0.013) VTI −0.213** −0.183** −0.211** −0.181** (0.017) (0.017) (0.017) (0.017) Low Doc 0.000 −0.009 −0.017* −0.020* (0.008) (0.009) (0.008) (0.009) Above limit 0.174** 0.276** 0.162** 0.267** (0.026) (0.029) (0.027) (0.029) Condo −0.304** −0.162** −0.300** −0.160** (0.027) (0.024) (0.027) (0.024) Investment 0.018 −0.010 0.015 −0.011 (0.024) (0.023) (0.024) (0.023) Refinance 0.405** 0.363** 0.392** 0.355** (0.012) (0.013) (0.012) (0.013) College −0.201** −0.200** (0.009) (0.009) Young −0.063** −0.063** (0.010) (0.010) Log(BEA Inc) −0.018 −0.019 (0.017) (0.017) House Price Growth −0.329** −0.328** (0.022) (0.022) Loan balance growth 0.093** 0.082** (0.008) (0.011) Payment resets 0.039** 0.034** (0.002) (0.002) Change in unemployment 0.021** 0.021** (0.003) (0.003) Income growth −0.179** −0.181** (0.026) (0.027) Observations 30,920,127 23,978,775 30,920,127 23,978,775 Table IX. Hazard models of personal bankruptcy This table reports the estimates from a Cox proportional hazard model for personal bankruptcy. The significance levels are abbreviated with asterisks: One and two asterisks denote significance at the 5% and 1% level, respectively. Individual-level MSA-level Individual-level MSA-level covariates covariates covariates covariates CM 0.667** 0.619** (0.011) (0.013) IO 0.602** 0.581** (0.012) (0.014) NEGAM 0.956** 0.818** (0.027) (0.026) ARM 0.389** 0.405** 0.391** 0.409** (0.014) (0.017) (0.014) (0.017) Log(Inc) −0.156** −0.097** −0.159** −0.099** (0.010) (0.012) (0.010) (0.012) FICO −0.466** −0.462** −0.467** −0.463** (0.008) (0.009) (0.008) (0.009) LTV 0.583** 0.551** 0.579** 0.551** (0.011) (0.013) (0.011) (0.013) VTI −0.213** −0.183** −0.211** −0.181** (0.017) (0.017) (0.017) (0.017) Low Doc 0.000 −0.009 −0.017* −0.020* (0.008) (0.009) (0.008) (0.009) Above limit 0.174** 0.276** 0.162** 0.267** (0.026) (0.029) (0.027) (0.029) Condo −0.304** −0.162** −0.300** −0.160** (0.027) (0.024) (0.027) (0.024) Investment 0.018 −0.010 0.015 −0.011 (0.024) (0.023) (0.024) (0.023) Refinance 0.405** 0.363** 0.392** 0.355** (0.012) (0.013) (0.012) (0.013) College −0.201** −0.200** (0.009) (0.009) Young −0.063** −0.063** (0.010) (0.010) Log(BEA Inc) −0.018 −0.019 (0.017) (0.017) House Price Growth −0.329** −0.328** (0.022) (0.022) Loan balance growth 0.093** 0.082** (0.008) (0.011) Payment resets 0.039** 0.034** (0.002) (0.002) Change in unemployment 0.021** 0.021** (0.003) (0.003) Income growth −0.179** −0.181** (0.026) (0.027) Observations 30,920,127 23,978,775 30,920,127 23,978,775 Individual-level MSA-level Individual-level MSA-level covariates covariates covariates covariates CM 0.667** 0.619** (0.011) (0.013) IO 0.602** 0.581** (0.012) (0.014) NEGAM 0.956** 0.818** (0.027) (0.026) ARM 0.389** 0.405** 0.391** 0.409** (0.014) (0.017) (0.014) (0.017) Log(Inc) −0.156** −0.097** −0.159** −0.099** (0.010) (0.012) (0.010) (0.012) FICO −0.466** −0.462** −0.467** −0.463** (0.008) (0.009) (0.008) (0.009) LTV 0.583** 0.551** 0.579** 0.551** (0.011) (0.013) (0.011) (0.013) VTI −0.213** −0.183** −0.211** −0.181** (0.017) (0.017) (0.017) (0.017) Low Doc 0.000 −0.009 −0.017* −0.020* (0.008) (0.009) (0.008) (0.009) Above limit 0.174** 0.276** 0.162** 0.267** (0.026) (0.029) (0.027) (0.029) Condo −0.304** −0.162** −0.300** −0.160** (0.027) (0.024) (0.027) (0.024) Investment 0.018 −0.010 0.015 −0.011 (0.024) (0.023) (0.024) (0.023) Refinance 0.405** 0.363** 0.392** 0.355** (0.012) (0.013) (0.012) (0.013) College −0.201** −0.200** (0.009) (0.009) Young −0.063** −0.063** (0.010) (0.010) Log(BEA Inc) −0.018 −0.019 (0.017) (0.017) House Price Growth −0.329** −0.328** (0.022) (0.022) Loan balance growth 0.093** 0.082** (0.008) (0.011) Payment resets 0.039** 0.034** (0.002) (0.002) Change in unemployment 0.021** 0.021** (0.003) (0.003) Income growth −0.179** −0.181** (0.026) (0.027) Observations 30,920,127 23,978,775 30,920,127 23,978,775 5. Conclusions The recent housing crisis brought the extension of credit to subprime borrowers and agency problems inherent in mortgage securitization to the forefront of academic research. This article focuses on a different aspect of credit markets during this time—namely, the proliferation of non-amortizing mortgages. In addition to variable interest rates, such mortgages also feature changes in amortization schedules set off by a variety of triggers. These CM contracts became very popular during the mid-2000s and vanished almost completely after the housing crisis of 2007–2009. We find that CM are the contract of choice for households with high credit quality and high incomes, in contrast to the low-income and low-credit-score population targeted by subprime mortgages. Households use CM as affordability products to purchase houses that are expensive relative to their current incomes, partly due to their expectations of higher future income and house price growth. CM borrowers are more likely to provide incomplete documentation for their loans, to be owners of investment properties, and to reside in non-recourse states in which lenders do not have access to non-collateralized assets in the event of mortgage delinquency. Consistent with the notion that households who self-select into CM products are fundamentally different from traditional mortgage borrowers, we find that CM experienced substantially higher propensities of mortgage defaults and personal bankruptcies, controlling for contract features, a variety of borrower and loan characteristics, as well as macroeconomic shocks. Higher delinquency rates cannot be attributed solely to greater leverage of CM and the onset of amortization resets brought about by inability to refinance CM. Furthermore, the difference in the delinquency rates between complex and traditional borrowers increases with measures capturing the benefit of defaulting (like the LTV ratio) and measures of financial sophistication (like income or credit scores). Overall, both the characteristics of CM borrowers and their default behavior cast doubt on the popular perception that CM are pushed by predatory lenders to naïve households. Our findings suggest instead that CM are taken out by relatively sophisticated borrowers than borrowers with more traditional mortgage contracts. Our results also highlight the contribution of mortgage market innovations to the surge in defaults among prime credit score borrowers during the financial crisis. Footnotes * We thank the editor Amiyatosh Purnanandam, two anonymous reviewers, and Ethan Cohen-Cole, Yongheng Deng, Serdar Dinc, Andra Ghent, Craig Furfine, Stuart Gabriel, Wei Jiang, Pete Kyle, Debbie Lucas, Jaehoon Hahn, Lu Han, Jay Hartzell, Jingzhi Huang, Jeongmin Lee, Robert McDonald, Justin Murfin, Tomasz Piskorski, Wenlan Qian, Oleg Rytchkov, Amit Seru, John Shoven, Laura Starks, Amir Sufi, Sheridan Titman, Nancy Wallace, Michelle White, and seminar participants at the American Finance Association in Chicago, the China International Conference in Finance in Shanghai, the European Finance Association in Stockholm, the Financial Economics and Accounting Conference at the University of Maryland, the Korea America Finance Association International Conference in Seoul, the ShovenFest at Stanford University, the Society of Financial Studies Cavalcade at the University of Michigan, the Swiss Economists Abroad Conference in Zurich, Brigham Young University, the Federal Reserve Bank of Chicago, the Federal Reserve Bank of Dallas, the Hong Kong University of Science and Technology, Korea University, Nanyang Technological University of Singapore, the National University of Singapore, New York University, Renmin University, Rutgers University, the Shanghai Advanced Institute of Finance, the Singapore Management University, Tsinghua University, the University of California at Los Angeles, the University of California at San Diego, the University of Lausanne, the University of Texas at Austin, the University of Zurich, and Vanderbilt University for helpful comments and suggestions. Clemens Sialm is an independent contractor at AQR Capital Management and thanks the Stanford Institute for Economic Policy Research for financial support during his Sabbatical leave. 1 This characterization of complex mortgages also corresponds to a frequent portrayal of complex mortgages in the media. See, for example, the New York Times article, How Countrywide Covered the Cracks, by Gretchen Morgenson, October 16, 2010, at http://www.nytimes.com/2010/10/17/business/17trial.html. 2 See, for instance, Mian and Sufi (2009), Keys et al. (2010), and Jiang et al. (2014a), among others. 3 It should be noted that optimality of such contracts from the viewpoint of an individual household does not imply social optimality if these contracts are associated with higher default probabilities and if defaults generate negative externalities for neighboring properties. The latter effect is shown in Campbell et al. (2011), Mian et al. (2015), and Guren and McQuade (2018). 4 Additional papers on securitization and the expansion of credit include Bond et al. (2009), Keys et al. (2009), Loutskina and Strahan (2009), Mayer et al. (2009), Gerardi et al. (2010), Piskorski et al. (2010), An et al. (2011), Campbell et al. (2011), Demyanyk and Hemert (2011), Li et al. (2011), Agarwal et al. (2012), Goetzmann et al. (2012), Keys et al. (2012), Woodward and Hall (2012), Adelino et al. (2013), Agarwal et al. (2014), Jiang et al. (2014b), Rajan et al. (2015), Begley and Purnanandam (2017), and Melzer (2017). 5 Alt-A loans are a middle category of loans, more risky than prime and less risky than subprime. They are generally made to borrowers with good credit scores, but the loans have characteristics that make them ineligible to be sold to the GSEs (e.g., limited documentation of the income or assets of the borrower or higher loan-to-value ratios than those specified by GSE limits). 6 We exclude observations with missing home values, missing loan amounts, and with appraisal amounts above $10 million. We also exclude loans affected by the Hurricane Katrina due to unreliable home price index data. These restrictions reduce the number of observations by only 1.7%. 7 Ghent and Kudlyak (2011) classify the following states as non-recourse: Alaska, Arizona, California, Iowa, Minnesota, Montana, North Dakota, Oregon, Washington, and Wisconsin. There is some ambiguity with respect to the recourse status of California loans. Refinance loans in California are subject to recourse only if the lender chooses to pursue judicial foreclosure. Although we observe whether a loan is used for new purchase or refinancing, we cannot assess the credibility of the threat of lender recourse through judicial foreclosure. In this article, only new purchase loans in California are defined as non-recourse. The results are robust to categorizing all California loans as non-recourse. 8 Several papers study the tradeoff between FRMs and ARMs (e.g., Campbell and Cocco, 2003; Koijen et al., 2009; Fuster and Vickery, 2015). 9 Given the near disappearance of CM contracts following the onset of the crisis, we limit attention to pre-2008 originations. Each mortgage is tracked for a minimum of 24 months (end of 2009) to ensure sufficient data for hazard model estimation. 10 The income distribution for CM borrowers lies to the right of the distribution of borrowers with fully amortizing ARM and FRM contracts. This also holds true when loans underwritten on the basis of stated (undocumented) income are dropped from the sample. 11 LPS data is collected at the loan and not property level, which limits one’s ability to construct an accurate estimate of the total debt secured by the house. In particular, we are unable to account for second-lien mortgages loans (the so-called “piggyback loans”). 12 A small fraction of FRM loans show a change in recorded payments. These differences are largely due to small fluctuations in taxes and private mortgage insurance payments. 13 The higher long-term loan-to-value ratios of complex mortgages may have contributed to a further deterioration in housing markets, as suggested by the leverage effect of Stein (1995) and Lamont and Stein (1999). Additional papers that study the macro-economic aspects of housing prices include Lustig and Van Nieuwerburgh (2005), Ortalo-Magne and Rady (2006), Piazzesi et al. (2007), Brunnermeier and Julliard (2008), Van Nieuwerburgh and Weill (2010), Landvoigt et al. (2015), and Favilukis et al. (2017). 14 The coefficient on the FICO score variable is significantly positive for CM if we select ARMs as the baseline group or if we run a simple logit regression. 15 Our main results are not affected materially if we include state-fixed effects instead of the MSA-fixed effects. 16 In unreported results, we find that the results are not affected qualitatively if we include lender-year fixed effects to capture time-variation in lender behavior. 17 About half of our observations have a missing “Low Documentation” variable. Our base-case results in Table III include these households, setting the “Low Documentation” value to zero. Table V includes only the households for which we know explicitly that they submitted fully documented loan applications. 18 The results are not affected qualitatively if we use a common baseline hazard, origination year-specific baselines, or origination year and state-specific baselines. We also estimate hazard models where the baseline hazard depends on the mortgage type. The estimates on the control variables are consistent with the base-case results. However, we do not show the results with mortgage type-specific baselines because this specification does not allow us to concisely report the impact of the mortgage type. 19 If we split up subprime and non-subprime ARMs, then the hazard estimates on subprime ARMs are nearly four times larger than the coefficients on prime ARMs. 20 A recent paper by Fuster and Willen (2017) finds a substantial impact of downward payment resets on subsequent loan performance. Their sample focuses on hybrid ARM contracts with interest rate resets after 3, 5, 7, or 10 years and a 10-year interest-only period. Since the loans in their sample are originated in 2005–2006, the resulting rate resets 5 years later cut the payments by 50%, on average. 21 Note that the LTV is measured at the time of origination and does not capture changes in leverage since origination. The changes in the LTV are driven by changes in house prices (which significantly affect delinquencies) and by changes in loan balances (which have mixed effects on delinquencies). 22 The interpretation of interaction effects in non-linear models is subject to the well-known critique of Ai and Norton (2003). However, we make use of the specific functional form of the Cox proportional hazard model to argue that the reported coefficients have a direct and natural interpretation. To see this, let’s consider the example of the interaction term between the FICO score and the CM indicator. Taking logs of the hazard function and then differentiating with respect to FICO yields ∂log h(i,t)/∂FICO=βFICO+γ×CM ⁠. Since CM is a binary variable, γ shows the difference in relative changes in the hazard function in response to changes in the FICO score for different types of mortgages. 23 See Li, White, and Zhu (2011) for an insightful discussion of the relationship between bankruptcy laws and mortgage defaults. References Adelino M. , Gerardi K. , Willen P. ( 2013 ): Why don’t lenders renegotiate more home mortgages? Redefaults, self-cures, and securitization , Journal of Monetary Economics 60 , 835 – 853 . Google Scholar Crossref Search ADS Adelino M. , Schoar A. , Severino F. ( 2016 ): Loan originations and defaults in the mortgage crisis: the role of the middle class , Review of Financial Studies 29 , 1635 – 1670 . Google Scholar Crossref Search ADS Agarwal S. , Ambrose A. , Chomsisengphet S. , Sanders A. B. ( 2012 ): Thy neighbor’s mortgage: does living in a subprime neighborhood affect one’s probability of default?, Real Estate Economics 40 , 1 – 22 . Google Scholar Crossref Search ADS Agarwal S. , Amromin G. , Ben-David I. , Chomsisengphet S. , Evanoff D. D. ( 2014 ): Predatory lending and the subprime crisis , Journal of Financial Economics 113 , 29 – 52 . Google Scholar Crossref Search ADS Ai C. , Norton E. C. ( 2003 ): Interaction terms in logit and probit models , Economics Letters 80 , 123 – 129 . Google Scholar Crossref Search ADS Amromin G. , Huang J. , Sialm C. ( 2007 ): The tradeoff between mortgage prepayments and tax-deferred savings , Journal of Public Economics 91 , 2014 – 2040 . Google Scholar Crossref Search ADS An X. , Deng Y. , Gabriel S. A. ( 2011 ): Asymmetric information, adverse selection, and the pricing of CMBS , Journal of Financial Economics 100 , 304 – 325 . Google Scholar Crossref Search ADS Barlevy G. , Fisher J. ( 2011 ): Mortgage Choices and Housing Speculation, Federal Reserve Bank of Chicago. FRB of Chicago Working Paper No. 2010-12. Available at SSRN: https://ssrn.com/abstract=1713308. Begley T. , Purnanandam A. ( 2017 ): Design of financial securities: empirical evidence from private-label RMBS deals , Review of Financial Studies 30 , 120 – 161 . Google Scholar Crossref Search ADS Bond P. , Musto D. K. , Yilmaz B. ( 2009 ): Predatory mortgage lending , Journal of Financial Economics 94 , 412 – 427 . Google Scholar Crossref Search ADS Brunnermeier M. K. , Julliard C. ( 2008 ): Money illusion and housing frenzies , Review of Financial Studies 21 , 135 – 180 . Google Scholar Crossref Search ADS Bucks B. K. , Pence K. M. ( 2008 ): Do borrowers know their mortgage terms?, Journal of Urban Economics 64 , 218 – 233 . Google Scholar Crossref Search ADS Campbell J. Y. ( 2006 ): Household finance , Journal of Finance 61 , 1553 – 1604 . Google Scholar Crossref Search ADS Campbell J. Y. , Cocco J. F. ( 2003 ): Household risk management and optimal mortgage choice , Quarterly Journal of Economics 118 , 1449 – 1494 . Google Scholar Crossref Search ADS Campbell J. Y. , Giglio S. , Pathak P. ( 2011 ): Forced sales and house prices , American Economic Review 101 , 2108 – 2131 . Google Scholar Crossref Search ADS Carlin B. I. ( 2009 ): Strategic price complexity in retail financial markets , Journal of Financial Economics 91 , 278 – 287 . Google Scholar Crossref Search ADS Carlin B. I. , Manso G. ( 2011 ): Obfuscation, learning, and the evolution of investor sophistication , Review of Financial Studies 24 , 754 – 785 . Google Scholar Crossref Search ADS Cocco J. F. ( 2013 ): Evidence on the benefits of alternative mortgage products , Journal of Finance 68 , 1663 – 1690 . Google Scholar Crossref Search ADS Corbae D. , Quintin E. ( 2015 ): Leverage and the foreclosure crisis , Journal of Political Economy 123 , 1 – 65 . Google Scholar Crossref Search ADS Demyanyk Y. , Hemert O. V. ( 2011 ): Understanding the subprime mortgage crisis , Review of Financial Studies 24 , 1848 – 1880 . Google Scholar Crossref Search ADS Favilukis J. , Ludvigson S. C. , Van Niewerburgh S. ( 2017 ): The macroeconomic effects of housing wealth, housing finance, and limited risk sharing in general equilibrium , Journal of Political Economy 125 , 140 – 223 . Google Scholar Crossref Search ADS Ferreira F. , Gyourko J. ( 2015 ): A New Look at the U.S. Foreclosure Crisis: Panel Data Evidence of Prime and Subprime Lending, The University of Pennsylvania. NBER Working Paper No. w21261. Available at SSRN: https://ssrn.com/abstract=2618649. Fuster A. , Vickery J. ( 2015 ): Securitization and the fixed-rate mortgage , Review of Financial Studies 28 , 176 – 211 . Google Scholar Crossref Search ADS Fuster A. , Willen P. S. ( 2017 ): Payment size, negative equity, and mortgage default , American Economic Journal: Economic Policy 9 , 167 – 191 . Google Scholar Crossref Search ADS Gabaix X. , Laibson D. ( 2006 ): Shrouded attributes, consumer myopia, and information suppression in competitive markets , Quarterly Journal of Economics 121 , 461 – 504 . Google Scholar Crossref Search ADS Garmaise M. ( 2013 ): The attractions and perils of flexible mortgage lending , Review of Financial Studies 26 , 2548 – 2582 . Google Scholar Crossref Search ADS Gerardi K. S. , Rosen H. S. , Willen P. S. ( 2010 ): The impact of deregulation and financial innovation on consumers: the case of the mortgage market , Journal of Finance 65 , 333 – 360 . Google Scholar Crossref Search ADS Ghent A. C. , Kudlyak M. ( 2011 ): Recourse and residential mortgage default: evidence from US states , Review of Financial Studies 24 , 3139 – 3186 . Google Scholar Crossref Search ADS Goetzmann W. N. , Peng L. , Yen J. ( 2012 ): The subprime crisis and house price appreciation , Journal of Real Estate Finance and Economics 44 , 36 – 66 . Google Scholar Crossref Search ADS Griffin J. M. , Maturana G. ( 2016 ): Who facilitated misreporting in securitized loans , Review of Financial Studies 29 , 384 – 419 . Google Scholar Crossref Search ADS Guiso L. , Sapienza P. , Zingales L. ( 2013 ): The determinants of attitudes towards strategic default on mortgages , Journal of Finance 68 , 1473 – 1515 . Google Scholar Crossref Search ADS Guren A. , McQuade T. ( 2018 ): How do Foreclosures Exacerbate Housing Downturns? Boston University and Stanford University. Jiang W. , Nelson A. A. , Vytlacil E. ( 2014a ): Liar’s loan? Effects of origination channel and information falsification on mortgage delinquency , Review of Economics and Statistics 96 , 1 – 18 . Google Scholar Crossref Search ADS Jiang W. , Nelson A. A. , Vytlacil E. ( 2014b ): Securitization and loan performance: a contrast of ex ante and ex post relations in the mortgage market , Review of Financial Studies 27 , 454 – 483 . Google Scholar Crossref Search ADS Keys B. J. , Mukherjee T. , Seru A. , Vig V. ( 2009 ): Financial regulation and securitization: evidence from subprime loans , Journal of Monetary Economics 56 , 700 – 720 . Google Scholar Crossref Search ADS Keys B. J. , Mukherjee T. , Seru A. , Vig V. ( 2010 ): Did securitization lead to lax screeing? Evidence from subprime loans , Quarterly Journal of Economics 125 , 307 – 362 . Google Scholar Crossref Search ADS Keys B. J. , Seru A. , Vig V. ( 2012 ): Lender screening and the role of securitization: evidence from prime and subprime mortgage markets , Review of Financial Studies 25 , 2071 – 2108 . Google Scholar Crossref Search ADS Koijen R. S. J. , Van Hemert O. , Van Nieuwerburgh S. ( 2009 ): Mortgage timing , Journal of Financial Economics 93 , 292 – 324 . Google Scholar Crossref Search ADS Lamont O. , Stein J. C. ( 1999 ): Leverage and house-price dynamics in U.S. cities , RAND Journal of Economics 30 , 498 – 514 . Google Scholar Crossref Search ADS Landier A. , Srear D. , Thesmar D. ( 2015 ) The Risk-Shifting Hypothesis: Evidence from Subprime Originations , University of California at Berkeley, and HEC . Available at SSRN: https://ssrn.com/abstract=1786542. Landvoigt T. , Piazzesi M. , Schneider M. ( 2015 ): The housing market(s) of San Diego , American Economic Review 105 , 1371 – 1407 . Google Scholar Crossref Search ADS Li W. , White M. J. , Zhu N. ( 2011 ): Did bankruptcy reform cause mortgage defaults to rise?, American Economic Journal: Economic Policy 3 , 123 – 147 . Google Scholar Crossref Search ADS Loutskina E. , Strahan P. E. ( 2009 ): Securitization and the declining impact of bank financial condition on loan supply: evidence from mortgage originations , Journal of Finance 64 , 861 – 922 . Google Scholar Crossref Search ADS Lustig H. , Van Nieuwerburgh S. ( 2005 ): Housing collateral, consumption insurance and risk premia: an empirical perspective , Journal of Finance 60 , 1167 – 1219 . Google Scholar Crossref Search ADS Mayer C. , Pence K. , Sherlund S. ( 2009 ): The rise in mortgage defaults , Journal of Economic Perspectives 23 , 23 – 50 . Google Scholar Crossref Search ADS Melzer B. ( 2017 ): Mortgage debt overhang: reduced investment by homeowners at risk of default , Journal of Finance 72 , 575 – 612 . Google Scholar Crossref Search ADS Mian A. , Sufi A. ( 2009 ): The consequences of mortgage credit expansion: evidence from the U.S. mortgage default crisis , Quarterly Journal of Economics 124 , 1449 – 1496 . Google Scholar Crossref Search ADS Mian A. , Sufi A. , Trebbi F. ( 2015 ): Foreclosures, house prices, and the real economy , Journal of Finance 70 , 2587 – 2634 . Google Scholar Crossref Search ADS Ortalo-Magne F. , Rady S. ( 2006 ): Housing market dynamics: on the contribution of income shocks and credit constraints , Review of Economic Studies 73 , 459 – 485 . Google Scholar Crossref Search ADS Piazzesi M. , Schneider M. , Tuzel S. ( 2007 ): Housing, consumption, and asset pricing , Journal of Financial Economics 83 , 531 – 569 . Google Scholar Crossref Search ADS Piskorski T. , Seru A. , Vig V. ( 2010 ): Securitization and distressed loan renegotiation: evidence from the subprime mortgage crisis , Journal of Financial Economics 97 , 369 – 397 . Google Scholar Crossref Search ADS Piskorski T. , Tchistyi A. ( 2010 ): Optimal mortgage design , Review of Financial Studies 23 , 3098 – 3140 . Google Scholar Crossref Search ADS Piskorski T. , Tchistyi A. ( 2011 ): Stochastic house appreciation and optimal mortgage lending , Review of Financial Studies 24 , 1407 – 1446 . Google Scholar Crossref Search ADS Purnanandam A. ( 2011 ): Originate-to-distribute model and the subprime mortgage crisis , Review of Financial Studies 24 , 1881 – 1915 . Google Scholar Crossref Search ADS Rajan U. , Seru A. , Vig V. ( 2015 ): The failure of models that predict failure: distance, incentives, and defaults , Journal of Financial Economics 115 , 237 – 260 . Google Scholar Crossref Search ADS Stein J. C. ( 1995 ): Prices and trading volume in the housing market: a model with down-payment effects , Quarterly Journal of Economics 110 , 379 – 406 . Google Scholar Crossref Search ADS Van Nieuwerburgh S. , Weill P.-O. ( 2010 ): Why has house price dispersion gone up?, Review of Economic Studies 77 , 1567 – 1606 . Google Scholar Crossref Search ADS Woodward S. E. , Hall R. E. ( 2012 ): Diagnosing consumer confusion and sub-optimal shopping effort: theory and mortgage-market evidence , American Economic Review 102 , 3249 – 3276 . Google Scholar Crossref Search ADS © The Author(s) 2018. Published by Oxford University Press on behalf of the European Finance Association. All rights reserved. For permissions, please email: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Journal

Review of FinanceOxford University Press

Published: Oct 1, 2018

There are no references for this article.