IMPACT OF MORTGAGE UNDERWRITING ON SINGLE FAMILY HOME FORECLOSURES IN CHICAGO. A Doctoral Dissertation Research

Transcription

1 IMPACT OF MORTGAGE UNDERWRITING ON SINGLE FAMILY HOME FORECLOSURES IN CHICAGO A Doctoral Dissertation Research Submitted to the Faculty of Argosy University, Chicago College of Business In Partial Fulfillment of the Requirements for the Degree of Doctor of Business Administration by Charles F. Yeager July 2013

2 IMPACT OF MORTGAGE UNDERWRITING ON SINGLE FAMILY HOME FORECLOSURES IN CHICAGO Copyright 2013 Charles F. Yeager All rights reserved

3 IMPACT OF MORTGAGE UNDERWRITING ON SINGLE FAMILY HOME FORECLOSURES IN CHICAGO A Doctoral Dissertation Research Submitted to the Faculty of Argosy University, Chicago College of Business In Partial Fulfillment of the Requirements for the Degree of Doctor of Business Administration by Charles F. Yeager Argosy University July, 2013 Dissertation Committee Approval: Sema Barlas, Ph.D., Chair Date Elias Demetriades, Ph.D., Member Ayman Talib, DBA, Member Byron Coon, Ph.D., Program Chair

4 IMPACT OF MORTGAGE UNDERWRITING ON SINGLE FAMILY HOME FORECLOSURES IN CHICAGO Abstract of Doctoral Dissertation Research Submitted to the Faculty of Argosy University, Chicago College of Business In Partial Fulfillment of the Requirements for the Degree of Doctor of Business Administration by Charles F. Yeager Argosy University July 2013 Sema Barlas, Ph.D., Chair Elias Demetriades, Ph.D., Member Ayman Talib, DBA, Member Department: College of Business

5 v ABSTRACT This research examines the impact of mortgage underwriting on single family residential mortgage foreclosures in Chicago. The underwriting components investigated are disposable income (DI), loan-to-value (LTV) ratios, and debt-to-income (DTI) ratios. By utilizing pre-foreclosure filing records at the census tract level and single family home value estimates at the neighborhood level, a tight geographic scope was created for a unique perspective on underwriting components inter-relationships with foreclosures. The findings illustrate that DI, DTI ratios, and LTV ratios were significant factors. The single family home value change rates from origination until the half prior to the preforeclosure filings were correlated to foreclosure rates. These findings suggest that loan affordability, lack of equity, and strategic defaulting due to a sharp decline in home values were the main contributors to high foreclosure rates. It is recommended that mortgage underwriting components be adjusted to protect homeowners and lenders from excessive exposure to risk during turbulent market conditions.

6 vi ACKNOWLEDGEMENTS I would like to express my thanks to Dr. Sema Barlas, Dr. Elias Demetriades, and Dr. Ayman Talib. Their advice, encouragement, and analytical insight assisted in my completion of this dissertation. With their contributions of time and expertise, I was able to finish the long process of this research. In addition, I have a sincere appreciation to Patty Maier at Record Information Systems for collecting, organizing, and providing the foreclosure data.

7 vii DEDICATION I dedicate the following dissertation research to my loving and patient wife, Lucy.

8 viii TABLE OF CONTENTS Page TABLE OF TABLES... ix TABLE OF FIGURES...x CHAPTER ONE: INTRODUCTION...1 Definitions...3 Purpose of Study...5 CHAPTER TWO: LITERATURE REVIEW...7 Foreclosures...8 Debt-to-Income (DTI) Ratio...8 Loan-to-Value (LTV) Ratio...10 Disposable Income (DI)...13 Interest Rate...14 Chicago Foreclosure Studies...15 Adjustable Rate Mortgages (ARMs)...16 Credit Score...17 CHAPTER THREE: METHODOLOGY...20 Hypothesis...20 Data Description and Procedures...20 Limitations...24 CHAPTER FOUR: RESULTS...27 Description of Market Foreclosure Analysis Foreclosure Analysis Foreclosure Analysis...54 Collective Results...57 CHAPTER FIVE: DISCUSSION...63 REFERENCES...68

9 ix TABLE OF TABLES Table Page 1. Descriptive Statistics City of Chicago City of Chicago Foreclosure Counts and Rates Descriptive Statistics City of Chicago Foreclosure Data Regression Analysis Correlation Matrix Adjusted Rate Mortgage (ARM) Duration ARM Loan Duration ARM Foreclosures Regression Analysis Correlation Matrix Regression Analysis Correlation Matrix Utilizing 2000 to 2008 Changes in DI, DTI, and LTV Foreclosure and SFR Value Percent Change Correlation Comparison... 61

10 x TABLE OF FIGURES Figure Page Year Fixed Rate Mortgage Interest Rate HMDA Conventional Purchase and Refinance Loan Volume for the City of Chicago Chicago Single Value Home Values Cook County Unemployment Rate Chicago 2000 Single Family Residence (SFR) Foreclosure Origination Years Scatter Plot 2010 Disposable Income (DI) Selected Monthly Owner Cost (SMOC) Scatter Plot 2010 Debt-to-Income (DTI) Ratio SMOC Scatter Plot 2010 Loan-to-Value (LTV) Ratio Chicago 2010 Single Family Residence (SFR) Foreclosure Origination Years Scatter Plot 2010 Home Value Percent Change from Origination Scatter Plot 2010 Adjustable Rate Mortgage (ARM) Scatter Plot 2010 Interest Rate Premium Chicago 2008 Single Family Residence (SFR) Foreclosure Origination Years Scatter Plot 2008 Adjustable Rate Mortgage (ARM) Scatter Plot 2008 Disposable Income (DI) Selected Monthly Owner Cost (SMOC) Scatter Plot 2008 Debt-to-Income (DTI) Ratio SMOC Scatter Plot 2008 Loan-to-Value Ratio Scatter Plot 2008 Home Value Percent Change from Origination Scatter Plot 2008 Interest Rate Premium Scatter Plot 2000 Disposable Income (DI) Selected Monthly Owner Cost (SMOC) Scatter Plot 2000 Debt-to-Income (DTI) Ratio SMOC...58

11 xi Figure Page 22. Scatter Plot 2000 Loan-to-Value Ratio Scatter Plot 2000 Home Value Percent Change from Origination Scatter Plot 2000 Interest Rate Premium...59

12 1 CHAPTER ONE: INTRODUCTION During the middle of 2007, the U.S. witnessed a rapid increase in residential mortgage foreclosures because of an increase in high-risk lending during the preceding decade. In 2010, 2.9 million foreclosure filings occurred, which is a 142% increase from the 2006 levels of 1.2 million (RealtyTrac, 2007, 2011). The U.S. is still plagued with high foreclosure rates today. This dramatic increase in foreclosures indicates a demand to analyze underwriting standards. This study investigates the relationship of underwriting decisions and foreclosures. Preventing future waves of foreclosures is imperative to stabilizing the housing market and the economy. While there are external events such as economic downturns, unemployment, death, illness, and divorce that strongly relate to foreclosures, prudent underwriting can reduce the foreclosures. Beginning in 2006, a home appreciation bubble fueled by the availability of cheap and easy financing peaked, and home values began to depreciate in many regions of the U.S. This downturn became the source of the financial crises of 2009; at the same time, the crises accelerated the depreciation of home values, resulting in an unprecedented increase in foreclosures. High foreclosure rates left a major spreading and lasting negative effect on the economy. This led to a prolonged recession, decreased household wealth, and increased emotional and financial stress on American families. An in-depth research and analysis on the impact of underwriting decisions on foreclosures is thus essential to help prevent future housing collapses and its negative impact on the economy and the lives of families. This research analyzes 2010 foreclosure filings to assess differential contribution of the components of underwriting on high foreclosure rates across Chicago

13 2 communities. The focus of the paper is on three components of underwriting characteristics: disposable income (DI), loan to value (LTV) ratios, and housing debt to income (DTI) ratios. These three independent variables influencing the foreclosure rates are controllable at origination by prudent underwriting. This research analysis is different from the majority of prior research to date. The data utilized is based on public-record information of motion to foreclose filings and is analyzed at the census tract level, rather than using loan origination servicer data on a zip-code, county, national, or multiple metropolitan statistical area (MSA) comparison. The previous research focusing on the impact of underwriting on foreclosures at the census tract level is limited. The data used in prior studies have also limitations in capturing the actual dynamics in the market. For instance, the most commonly used data source, Loan Processing Services (LP), demonstrates a bias towards securitized mortgages. LP s data set over-weighs subprime loans and is not freely available to non-government researchers. Other sources are often proprietary company specific, which suffer from biases related to company underwriting standards as well as geographic location. This data source may come from failed banks, i.e. Indy Mac, or unnamed banks that may or may not have survived the mortgage collapse. In addition, various sources typically have only partial or no information about second liens. The source for foreclosure data in this study was obtained through Record Information Services [RIS] (2010), which circumvents many of the prior problems.

14 3 Definitions In order to establish a common understanding of the terminology used in this study, the following terms are defined: Adjustable rate mortgage (ARM) is an interest rate of a home loan that is anticipated to change at a future point after the loan is originated. Typically, the interest rate will be a constant fixed rate up to a pre-determined inflection point. Common adjustment points are after 1-year, 2-year, 3-year, and 5-year periods of fixed interest rates. ARM rates after the initial fixed rate period adjusted every year, 6 months, or monthly. Debt-to-income (DTI) ratio is the monthly housing related debt payment divided by the gross household monthly income. For this study, this ratio is also known as top DTI ratio. The bottom DTI ratio, which is not analyzed in this study, would also include monthly consumer debt payments into the numerator of the equation. Default is synonymous with foreclosure. Disposable income (DI) is the net household monthly income minus housing related monthly expenses, minimum monthly consumer debt expenses, and additional expenses based on the total number of household dependents. The additional expenses are determined in this study by the multiplier of $100 per household person. Equity is the amount of value remaining from the difference of estimated home value (or actual based on purchase amount) and the total amounts of loan values. Fair Isaac Company (FICO) is the industry standard credit score utilized to estimate a borrower s credit worthiness based on previous payment history.

15 4 Foreclosure is the legal process in which a lender attempts to collect from the borrower because terms of the mortgage contract were not fulfilled. Often, repossession of the house and land occur. Typically, the pre-foreclosure lender legal filing is executed after 90 days (three consecutive missed payments) of delinquency. Fixed rate mortgage (FRM) is an interest rate that stays constant throughout the term of the home loan. Interest rate premium is the difference between the mortgage interest rate and the nation-wide average 30-year fixed interest rate of the origination year. London Interbank Offered Rate (LIBOR) is the interest rate determined by London Banks panel members based from an average of short-term bonds. Loan-to-value (LTV) ratio is the loan origination amount divided by the value of the home. Margin is the rate amount that an ARM is expected to change. Typically, the margin is an amount that consists of the LIBOR rate plus a pre-determined (margin) rate on the legal note of the mortgage contract. Strategic Default is a decision from a mortgage borrower to discontinue making monthly mortgage payments despite having the financial means to do so. This decision is typically made because the estimated value of the home is worth far less than the mortgage balance amount. Underwater is a situation when a borrower s loan amount owed is larger than the value of the home.

16 5 Purpose of Study The purpose of this research is to explore the relationship of underwriting decisions to Chicago community foreclosures. This research will contribute value to existing research by (1) utilizing information that is more recent, (2) analyzing foreclosure data from the seldom-used perspective of public record data, (3) taking advantage of Zillow.com s more precise neighborhood level historical single family home value estimates, (4) utilizing single family residence (SFR) housing unit totals instead of the commonly used loan origination totals, (5) concentrating the analysis on the uncommonly utilized census tract level differences in underwriting standards and foreclosures, (6) and investigating the effects of the less frequently analyzed estimated housing DTI ratio, DI and estimated LTV ratios. While aspects of this research have been analyzed before, no study has jointly analyzed foreclosures in this manner. This study aims to assess unique contributions and the relative importance of different underwriting tendencies in explaining foreclosure rates. This unique method of analysis is anticipated to prompt additional research. The high levels of residential foreclosures exposed a flaw in mortgage lenders risk management guidelines. DTI ratios, DI amounts, and LTV ratios have been tools utilized to manage risk. By analyzing these variables impact on foreclosures, the study reveals which components had a greater impact. In addition, the findings add value to existing research by illustrating if affordability of mortgage payments or if substantial negative equity making strategic default more advantageous were the culprit(s). Ultimately, implementing full documentation loans and requiring more equity in

17 6 refinance and purchase mortgages would have reduced the massive amounts of risk permitted by lenders. The foreclosure boom not only negatively impacts banks and homeowners facing default, the spillover effect hurts prudent lenders and creditworthy borrowers. Communities suffered from the loss of tax revenue, increased crime, and blighted neighborhoods. A more conservative estimate of future home price appreciation might have aided lenders in their risk-aversion policies. Demanding proof of income and assets from the borrower might have helped ensure that loans were affordable. It is important to analyze the impact of these mortgage risk variables to help ensure that catastrophic foreclosures do not occur in the future. This research consists of five chapters. Chapter Two reviews the relevant literature regarding foreclosures, debt-to-income (DTI) ration, loan-to-value (LTV), disposable income (DI), interest rate, Chicago foreclosure studies, adjustable rate mortgages (ARMs), and credit score. Chapter Three describes the research methodology, including the hypotheses, data description, procedures, and the limitations. Chapter Four discovers the results, including a description of the market, 2010 foreclosure analysis, 2008 foreclosure analysis, 2000 foreclosure analysis, and collective results. Chapter Five discusses the conclusions drawn from this study.

18 7 CHAPTER TWO: LITERATURE REVIEW Several researchers blamed the loose underwriting standards for the high rate of foreclosures (Johnson & Neave, 2008; Rose, 2008; Sandler, 2007); other researchers found that the decline in home price values was the actual culprit to the foreclosure boom (Amromin & Paulson, 2009; Gerardi, Shapiro, & Willen, 2009; Lang & Jagtiani, 2010). Subprime mortgages are the result of lax underwriting guidelines and have traditionally been more likely to default. Amromin and Paulson (2009) define subprime loans as borrowers with bad credit, which often utilize stated income/assets documentation, and have little or no equity at origination. To compensate for the higher risk, subprime loans are often more susceptible to unfavorable loan terms. Anti-predatory lending reform viewed the foreclosure problem as a result of excessively high interest rates: 3% above treasury yields (U.S. Government Accountability Office [GAO], 2009). In addition, the U.S. GAO report showed that a higher LTV resulted in higher levels of default. Another lax underwriting guideline allows the borrower or broker to submit untruthful income documentation. Stated income loans can distort DTI ratios and lead to higher risk of default. The Mortgage Asset Research Institute found that out of 100 stated income loans sampled, 90% had inflated incomes of 5% or more. Shockingly, nearly 60% had embellished their incomes by more than 50% (Sharick, Omba, Larson, & Croft, 2006). It is not surprising that stated income loans are also nicknamed liar loans. Thus, this research focuses on assessing the relative impact of disposable income (DI), loan-to-value (LTV) ratio, and housing debt-to-income (DTI) ratio on Chicago neighborhood foreclosures.

19 8 Foreclosures A foreclosure occurs when the borrower fails to meet their obligation to make their monthly mortgage payments, and the lender takes full ownership of the collateralized home. Typically, the lender files a lawsuit to achieve a judgment of foreclosure after a 90 day period of missed mortgage payments. The notice of intent to foreclose filing is considered the pre-foreclosure stage. From the judgment date, the borrower has a period of time to bring the loan to current status. If the loan is not remedied, a notice of sale is issued in which the home sale is conducted by a county sheriff or judge (Illinois Foreclosure Law, n.d.). A consented foreclosure can speed the process up, while the presence of a defense lawyer could prolong the process. For this research, the pre-foreclosure stage is analyzed because it allows for consistency in comparing the timing of the defaults. The term, default, in this study s context is synonymous to foreclosure; literature interpretations commonly utilize being 90 plus days delinquent on mortgage payments as the definition of default. Debt-to-Income (DTI) Ratio The main component of the liar loan is the over-stated income. The income is over-stated with the intent of lowering the overall DTI ratio below the underwriting maximum requirement. An overall DTI ratio (bottom or back ratio) of 50 or 55% is the most common threshold level for Alt-A and subprime lending. Alt-A stands for Alternative A lending. According to Danis and Pennington-Cross (2008), borrowers of Alt-A loans often have similar credit history as prime borrowers but usually provide no/limited income documentation or down payment. For conforming underwriting standards, household DTI ratio (top or front ratio) is 28% for conventional loans and 31%

20 9 for Federal Housing Authority (FHA) loans. The top ratio includes the sum of monthly payments of mortgage(s), taxes, homeowners insurance, and Homeowner Association (HOA) fees, if applicable, divided by gross income. The bottom ratio includes the monthly housing debt plus all debt payments (i.e. car loans, credit card) divided by monthly gross income. By permitting unverified income for lending decisions, lenders expose themselves to higher risk since income could easily be manipulated. Research has shown that stated income loans on average overstated the income by 20% (Blackburn & Vermilyea, 2010; Jiang, Nelson, & Vytlacil, 2010). By utilizing different data sources and methods of calculations, Blackburn and Vermilyea (2010) analyzed income obtained from the Home Mortgage Disclosure Act (HMDA) within specific metropolitan statistical areas (MSAs) and the American Housing Survey (AHS). Their research was limited to owner occupied one to four unit residential home purchases with conventional mortgages. The foreclosure rate utilized was provided by RealtyTrac on the MSA level. Blackburn and Vermilyea found that HMDA average MSA income levels were significantly higher than the AHS income levels in 2005 and The higher HMDA income total implies an exaggeration of true income in hopes of obtaining a home loan. Coinciding with this divergence in the mid-2000s, the use of low/no income-documented loans increased (Blackburn & Vermilyea, 2010). These findings indicate that high, real DTI ratios may be a significant source of high foreclosure rates. Jiang et al. (2010) utilized proprietary data from an anonymous large lending institution. When comparing full documented loans to low documented loans originated between , statistically significant higher delinquency rates were found amongst low documented loans. Delinquency means that the borrower is at least 30 days late on

21 10 their mortgage payment. Jiang et al. s (2010) findings also showed how this lender dramatically increased broker-originated loans as well as low documented loans from early 2004 through late Their composition changed from 73% to 94% and 39% to 75%, respectively. This study showed that low documented loans as well as broker originated loans tended to misrepresent income, among other variables. These liar loans understated DTI ratios and allowed for higher loan amounts, which increased the risk of delinquency. A study by Foote, Gerardi, Goette, and Willen (2009) utilized loan processing (LP) data to analyze DTI and its relationship to foreclosures. Their findings illustrated that a high DTI ratio at origination is not predictive of high default rates. They theorize that the common affordability measurement of DTI ratios is more meaningful throughout the loan and not just at origination. However, a flaw with this study is that LP possesses DTI ratios that may be falsified from the stated loans. In addition, only about half of the data included DTI ratios. Furthermore, it does not factor in the potential of second mortgages, which would raise DTI ratios. Foote et al. (2009) defended that a doubletrigger model, in which the interaction of income shock and falling home prices, were predictive of increased default. Loan-to-Value (LTV) Ratio LTV is another underwriting component that has typically been associated with mortgage risk. Danis and Pennington-Cross (2008) found that LTV ratios at origination are positively correlated with serious delinquencies. In this study, a subset of LP data, subprime single family residences (SFR) with 30 year fixed rate mortgages (FRM) was utilized. When analyzing the borrowers level of equity after the loan origination, a

22 11 monthly state level of home sale price values were utilized to estimate home values. Danis and Pennington-Gross also did not take into account second mortgages. Similar to most studies, this analysis used mortgage performance data sampled across the entire U.S. This broad analysis obtains general results that do not take advantage of geographically specific details. One disadvantage of using LP data is the limited amount of second lien data that can be obtained. To avoid this limitation, Elul, Souleles, Chomsisengphet, Glennon, & Hunt (2010) utilized credit bureau information combined with a LP dataset in their research. The credit bureau data permitted cumulative LTV (CLTV) estimates along with credit card utilization estimates. CLTV contains all mortgage liens on a home, and credit card utilization levels are a measure of consumer liquidity. This study helps to fill the void of the limited amount of CLTV research that is available. In addition, analyzing the credit card utilization levels is less commonly researched despite its important aspect for underwriting decisions. Elul et al. s (2010) LP data set consists of owner occupied non-multi unit fixed rate mortgages (FRMs) originated in 2005 and High CLTV ratios combined with low credit card utilization levels resulted in higher default risk than lower CLTV ratios with high credit card utilization levels. Interestingly, low CLTV ratios combined with high credit card utilization levels yielded even higher default risk. Inherently, one would assume that higher CLTV or higher credit card utilization would lead to a higher default risk. This situation implies that the borrower cannot find additional financing and that refinancing may not be an option because of the limited or negative home equity. This research indicates that CLTV, and evidently the absence of home equity, was a driving

23 12 force in borrower default and not their high consumer spending habits, which are inferred by high credit card utilization levels. In addition, Elul et al. (2010) found that low CLTV ratios combined with large increases of unemployment had little effect on the default rate, while high unemployment and high CLTV resulted in substantial default probability increases. This finding implies that areas with high employment and low CLTV ratios were better able to avoid default because they possibly were able to access a home equity line of credit to temporarily assist them through the employment transition. In addition, this suggests that borrowers with low CLTV ratios were more likely to have higher amounts of savings available because the borrower provided the ability to save with their higher available home equity. Underwriting standards and foreclosure rates vary based on geographic location; however, most previous research ignored geographic differences. Gerardi et al. s (2009) study is an exception in this regard. This study s analyses were based on the Warren Data set, which is concentrated in Massachusetts and contained home sales and foreclosures from In addition, this data set allowed analysis of all house liens (first and second mortgages), which represents homes CLTV. Their influential research found that home value depreciation was the leading factor for foreclosures despite the impact of deteriorating underwriting standards. However, the lax underwriting standards did fuel the housing value appreciation bubble (Gerardi et al., 2009). Gerardi et al. s research found that when the origination CLTV ratio is greater than 1.0, loans are six times more likely to default compared to when CLTV originates between 0.9 to 1.0 (2009). This study also confirmed that LTV ratios under 0.8 demonstrated the lowest risks.

24 13 Demyank, Koijen, and Van Hemert (2011) found that while underwriting standards had deteriorated since 2001, the diminished lending standards only lead to high foreclosure rates after rapid home price appreciation (HPA) halted. This supports Gerardi et al. s (2009) research, which suggests that perhaps the lax underwriting was not the foreclosure culprit, but the sudden decrease in HPA, and then the drop in home values hurt homeowners the most. Prior, borrowers had an exit strategy if they could not afford the home; they could either refinance or sell the home based on the increased home equity due to rapid home appreciation. However, if homes are affordable, then selling or refinancing is not a major concern. Disposable Income (DI) Disposable income (DI) is a dynamic variable because it relates to after-tax income earned, expenses paid, and it is rational to assume a strong correlation between DI and borrowers savings. Society has trended into an environment of spenders instead of savers. To some extent this consumerism is captured by DI values. Borrowers that have a low or negative DI value may exacerbate their payment shock risk by choosing not to opt for limited or any health, life, or disability insurance. Research demonstrated a form of DI can help predict insolvency. The gross annual debt payment to DI ratio was one of her key variables utilized. In addition, a study found that households at highest risk of insolvency, met one or more of the following conditions: Liquid Assets were less than onefourth of their Disposable Income, Annual Payments for Housing and Consumer Debt were larger than 35% of their Disposable Income, and Total Assets were less than Total Liabilities. (DeVaney, 1994, p.15) Hakim and Haddad (1999) performed a national study of conventional mortgages which was analyzed into six regions of the U.S. They found that borrowers who

25 14 defaulted tended to a relatively lower difference between monthly income and monthly fixed expenses. Their prime variable is similar to the established definition of DI, except they used gross income and did not take a deduction multiplier for dependents. Hakim and Haddad (1999) found that the number of dependents also increased the risk of default. Disposable Income (DI) has also been analyzed and broken into its subcomponents. Getter (2003) studied the impact of post credit approval in which he found that unexpected lower household income and wealth significantly increase the probability of delinquency. However, monthly debt payment obligation rises were not significantly associated with delinquency risk. In addition, household family size was a significant factor in the delinquency risk models. The increased debt payment load may be related to increased borrower confidence in payment ability. This aspect may be contradictory to the consumerism problems that were faced during the more recent financial crisis that started in the summer of Interest Rate The impact of interest rates on loan performance is another potentially important underwriting characteristic. Higher interest rates lead to higher monthly payments; the higher payments would in theory lead to higher income shock risk and hence, higher probably of default. Amromin and Paulson (2009) found that foreclosures are strongly associated with LTV ratios, FICO scores, and interest rates from the origination. While not available from the LP data utilized, Amromin and Paulson admit that CLTV would be a better measure of the borrower s leverage. In addition, their data found that subprime borrowers possessed a higher margin rate than prime borrowers. Margin rate is the

26 15 interest rate that is required to be paid on ARMs once the rate resets. Typically, the margin rate is an additional rate tied to the London Interbank Offered Rate (LIBOR). For prime loans, Amromin and Paulson found that a 1-percentage point rise in interest rates resulted in a 3-percentage point rise in the default rate. In Dom, Furlong, and Krainer s (2007) research, they find that interest rates are positively correlated with subprime mortgage delinquency rates. This study utilized LP data. In addition, they compared state level interest rate and delinquency performance. During 2005 to 2006, states with the highest home price appreciation (HPA) also had the lowest interest rates, while states with slow HPA had higher interest rates. For instance, California s mortgage subprime interest rates were 0.65% lower than Ohio s subprime interest rates. This finding implies that the lower interest rates fueled HPA and mortgage loan volume. The lower interest rates resulted from higher competition to lend based on the perception of a safer to lend environment. The rapidly increasing value of the collateral allowed for an easier exit of high risk borrowers if they could not meet the financial obligation of their mortgage payments. Once the housing bubble burst, the areas with the highest HPA were more exposed to foreclosures. Mayer s (2010) study supports this by claiming changes in interest rates drove home prices. For example, a decline in mortgage rates from 6% to 5% could reduce the cost of owning a home by up to 16%, leading to an increase in house prices of 13.6% (Mayer, 2010, p. 7). Chicago Foreclosure Studies Narrowing the geographic focus even more, Rose (2008) analyzed subprime foreclosures in the Chicago MSA. In this study, LP data was utilized and spanned from

27 to the second quarter of This study s sample found that from there were twice as many adjustable rate mortgages (ARMs) to fixed rate mortgages (FRMs) in Chicago. This characteristic is unique compared to the nationwide sample, which found a ratio of about 50%. An ARM is a mortgage in which the interest rate changes during the life of the mortgage. This product is inherently more risky for the borrower and typically, the interest rate adjusts every six months, monthly, or annually after the initial fixed rate period. FRMs interest rate stays the same for the life of the mortgage. During the period analyzed by Rose (2008), Chicago MSA s LTV ratio was about 5% lower and their FICO (Fair Isaac and Company) credit score was about 24 points lower than the national average. All other components appear similar. One important finding was that [l]ow- or no-documentation is associated with a greater probability of foreclosure for refinance FRMs and ARMS (Rose, 2008, p. 24). This finding implies that the true DTI ratio is an important indicator to forecast future foreclosures for the Chicago area. While the true DTI ratio is unknown, census tract level owner-occupied household income values provide a consistent estimation based on nearby tendencies. Adjustable Rate Mortgages (ARMs) One could theorize that ARMs higher interest rate adjustments lead to payment shock to the borrower, which increases foreclosure likelihood. However, Bhutta, Dokko, and Shan s (2010) research found that the interest rate changed upward in less than 10% of the mortgages. Their data consisted of a large LP sample data set of first lien purchase mortgages at 100% CLTV originated in 2006 in Arizona, California, Florida, and Nevada. In this study, the median loan age was just 18 months before termination of the

28 17 loan. Bhutta et al. (2010) concluded that interest rate changes were likely not a significant factor relating to loan defaults. The primary findings illustrated that the median borrower strategically defaulted once their home equity fell to -62%. In addition, 80% of the defaults resulted from negative equity combined with income shocks. County level unemployment change, interest rate changes, and ZIP code level credit card delinquency trends were proxy for income shocks. Studies can further analyze ARMs impact by comparing subprime and prime mortgage patterns. Amromin and Paulson s (2009) research found that subprime ARMs and fixed rate mortgages (FRMs) are correlated with each other and have the same default rate. Through their use of LP data, their findings showed that 62-73% of subprime mortgages were originated utilizing ARMs from ; the preponderance of those ARMs were due to reset in three years or less. Conversely, 23-26% of prime loans were ARMs; less than half of those ARMs reset in three years or less. This illustrates that subprime loans were more dependent in refinancing and were impacted more by declines in home values. The margin rates for subprime mortgages from illustrated a significant relationship with defaults. Subprime margin rates were higher than prime margin rates, and as a result, could provide a higher mortgage payment shock if the ARM resets. Credit Score Another underwriting determinant for mortgage risk is credit. Underwriters utilize a three digit number to assist in evaluating a borrower s credit worthiness. This three digit number is often referred to as a Fair Isaac Company (FICO) score. The FICO score was popularized in the mid-1990s. The credit rating is the result of an

29 18 accumulation of payment history, amounts, and results. Credit score is typically derived from the three main national credit agencies: TransUnion, Equifax, and Experian. Since these three agencies calculate payment history and risk differently, lenders use the middle credit score. In addition, some financial lenders and judgments do not report to all of the credit reporting bureaus. The credit score ranges from about mid 300s mid 800s with a higher score indicating a lower risk of the borrower defaulting. Borrowers credit is worsened by becoming 30 days delinquent on debt, becoming bankrupt, having judgments, having too much revolving unsecured debt, or even from having a large amount of recent credit inquires. Several researchers found that a low credit score is a significant component for high default areas (Ben-Shahar, 2008; Grover, Smith, & Todd, 2008; Okah & Orr, 2010). The first signal to lenders of default risk is obtained through a borrower s credit score. Sub-screening options of various LTV ratios and interest rates become available based upon the credit score. Ben-Shahar (2008) illustrated that mortgage default is interrelated to credit score, LTV ratio, and interest rate. The loan characteristic categories develop as tools for resale of the mortgage as investors price the risk and return in part on the raw underwriting data. Credit score data results in Grover et al. (2008) was found to be the strongest single variable to determine foreclosures. Their study consisted of 2002 foreclosure sheriff sale data from two Minnesota counties. The data was geo-coded and analyzed at the census tract level. The credit data was obtained through PCI Corporation and CRA Wiz, which was derived from a contact on the Board of Governors with the Federal Reserve. Bhardwaj and Sengupta (2011) concluded that the value of credit scoring was

30 19 overly relied upon, which assisted in justifying the acceptance of other riskier origination components of the loan despite the average increase in overall credit scores.

31 20 CHAPTER THREE: METHODOLOGY Hypothesis This research theorized that at the time of loan origination, lenders were able to calculate the impact of interest rates and LTV ratios. However, the inaccurate estimation of the true housing DTI, by permitting stated-income loans, resulted in higher foreclosure rates in Chicago communities. Prolonged low interest rates and the high HPA permitted the most dangerous underwriting standard, lending to borrowers that could not afford the loan. These hypotheses are theorized despite strong volatile home value changes and high unemployment changes. Hypothesis 1: Disposable Income (DI) and Debt-to-Income (DTI) ratios at loan origination are the most important factors in Chicago foreclosures. Hypothesis 2: Loan-to-Value (LTV) ratios at loan origination influence the foreclosure rates in Chicago. Data Description and Procedures The source for foreclosures in this study was a data set provided by Record Information Services, Inc. [RIS] (2011). RIS is a company that specializes in consolidating public record data for the six county Chicago regions. As a testament to their quality of data, the Cook County Assessor s Offices utilize their organized foreclosure data set. This data set represents all the 2010, 2008, and 2000 housing perforeclosures (notice of intent to foreclose filings) in Chicago as it is derived from the Cook County public record data. RIS foreclosure data was also utilized in research from Hartley (2010), Immergluck and Smith (2005), and Malkova (2008).

32 21 The census tract housing totals and median owner occupied household incomes were derived from the 2000 U.S. Decennial Census Bureau. The income utilized is based off the loan origination year values, as adjusted by the Chicago MSA consumer price index (CPI). The mortgage contract interest rate, mortgage amount, loan duration, and foreclosure census tract location were derived from RIS data. The mortgage rate and amount is based from loan origination data. For this study, foreclosure rate was determined by SFR foreclosures divided by 1- unit residential homes in each census tract. This foreclosure rate calculation was also utilized by Roger and Winter s research (2009). By concentrating on just single family and not two-four unit housing foreclosures, it was anticipated that investor related foreclosure filings would be reduced. Alternative measures not utilized were total foreclosures divided by total housing units or owner occupied housing units. Foreclosures divided by existing homes with mortgage(s) was not an available option, but it was used by researchers that possess access to LP data set. The SFR foreclosure duplicates and perceived second mortgages, with less than $50,000 origination mortgage amounts, were eliminated from the data set. In addition, the property type was re-checked via the Cook County property search website (Cook County Assessor s Office, n.d.) for all filings with high LTV ratios above 120%. Multiunit properties typically have higher values than SFRs because of increased functionality and overall square footage. Foreclosures with high LTV ratios or unit/apartment number in their address were inspected and eliminated if the property type was not a SFR. Properties with unit or apartment numbers have a higher potential of being a condominium or an apartment.

33 22 Surprisingly, disposable income (DI) was a scarcely researched underwriting tool. This underwriting component was more commonly utilized in subprime lending and was not a Fannie Mae, Freddie Mac, or FHA lending requirement. DI is also known as cash flow, and it is commonly utilized in commercial property underwriting. In this analysis, DI was calculated by the following formula. Estimates are based off 2000 U.S. Census Bureau (2000a, 2000b, 2000c) at the census tract level and were adjusted by the consumer price index for Chicago s MSA (U.S. Department of Labor, 2012). Results were similar to the foreclosure based mortgage payment data. DI = I (1 t) SMOC c f I = monthly owner occupied household income as of the origination date t = tax rate, estimate of 25% SMOC = selected monthly owner occupied costs from the 2000 U.S. Census data and adjusted by Chicago MSA CPI, which includes: payment interest taxes, insurance payments, and energy costs. SMOC was also utilized in Ergungor (2007) research. c = monthly consumer debt estimate of $700 f = census tract size family size multiplied by $100 cost estimate per person The foreclosure based DI calculations did not use SMOC but instead used principle + interest + taxes and insurance (PITI) estimates. These calculations were averaged for census tracts that possessed at least five filings per census tracts. However, the DI SMOC were utilized in the results section because of their similarity to foreclosure based DI. The DI SMOC were also analyzed because this data provided more extensive information for census tracts that had zero or few foreclosures. The alternative SMOC calculation is as follows: Housing Cost = mortgage payment + property tax payment + insurance payment

34 23 Mortgage payment = p [i (1+i) n ]/ [(1+i) n -1] p = principal, original loan amount i = contract interest rate per month n = number of compounding periods Property tax payment = [[v(a)(s) e](t)] / 12 v = Estimated property value, based off Zillow, Inc. neighborhood singlefamily estimate at loan origination a = assessment level, 16% s = estimated state equalizer e = homeowner exemption t = estimated tax rate Insurance payment = based of the following staggered payment amount based off loan value estimate at origination: $45,000 $100,000 = $25 $100,001 $250,000 = $50 $250,001 - $450,000 = $75 $450,001 - $750,000 = $100 $750,001 - $1,000,000 = $125 $1,000,000+ = $150 Home value estimates were derived through the publically available Zillow website (2012). This data set provides uniquely specialized data, which segments the value results not just at the MSA or zip code level, but at the smaller, more homogeneous, geographic level of Chicago neighborhoods. To further improve accuracy

35 24 of the data, this historical home value data was made available at the single family residence (SFR) level. At the SFR level, Zillow provided information on 47 Chicago zip codes and 140 neighborhoods. In about 10% of the data, neighborhood value was unavailable; zip code data was instead utilized as a close alternative for the data points. Other research has noted Zillow s close relationship to another home value data source of Case-Schiller (Bontas, 2011; Guerrieri, Hartley, & Hurst, 2010; Mian, Sufi, & Trebbi, 2011). While some researchers have taken advantage of Zillow s city or zip code free data set, other researchers have not taken advantage of the Zillow s neighborhood level values. This is perhaps because the majority of studies are not focused on a similar geographic scope of just one large city. In addition, other researchers often overlooked the value of geo-coding their loan or foreclosure data at the neighborhood level. Limitations At the time of this research, the 2010 U.S. Census data had not yet been compiled at the census tract geographic level. This lack of available information forced this research to utilize CPI adjustments from the 2000 U.S. Census data to obtain estimates of income level changes. In addition, while Chicago is a well-developed urban environment, 2000 one unit housing totals were utilized instead of the preferred 2010 totals. This may have caused some inaccuracies with the foreclosure rate totals for each census tract. Follow-up research is warranted utilizing the soon to be released 2010 Census data at the census tract level. The new income information could help control for unemployment variations that may be more dramatic in certain neighborhoods compared to others.

36 25 The Home Mortgage Disclosure Act (HMDA) data also possesses limitations. The HMDA data was utilized to estimate LTV ratios for the census tracts that possessed fewer than five foreclosures. The HMDA data may have encompassed values that were higher than the single family home value for areas with high amounts of multi-unit (2-4) residential home purchases and refinances. However, there should be a downward LTV ratio tendency in the census tracts that HMDA data was utilized as a proxy because loan data was incorporated instead of foreclosure loan data. Foreclosed home LTV ratios would likely be higher than the average home loans. Similar to most foreclosure research, this research failed to encompass the impact of potential second mortgage data. The absence of this data could cause an underestimation of area LTV ratios and DTI ratios. In the 2000s, second mortgages, as well as first mortgages that go up to 100% of the value of the home, were more common than in the past on originated loans. The foreclosure record data was illustrated to have great analytical value at the census tract level. However, a limitation with the foreclosure data was that just 2000, 2008, and 2010 years were utilized. Perhaps, if more years of foreclosure data were utilized, a more comprehensive analysis could be ascertained. In addition, a deep analysis of multi-unit and condominium unit foreclosures could have provided an interesting comparative analysis. Time and financial constraints limited the study to just three years of data and single family foreclosure data. In the mortgage underwriting process, the credit score is a key variable. Unfortunately, this study did not have access to credit score data. Credit score data is available at zip code aggregate level for a steep price through credit agencies such as

37 26 Trans Union, Equifax, and Experian. The Loan Processing (LP) Services data set provides credit score data, however, this study lacked the financing or connections to incorporate the credit score data. This study chose not to utilize credit score data obtained through and publically available through a local advocacy group. Unfortunately, that data would have contamination problems. The credit score data represented Chicago area zip code credit score values from June 30, Ideally, credit scores from before the foreclosure boom and not during the heightened foreclosure boom would have assisted this analysis and avoided contamination concerns. Despite the absence of credit score data, this research was effective in evaluating the impact of the other main underwriting variables.

38 27 CHAPTER FOUR: RESULTS Chapter Four examines the underwriting components impact on single family mortgage foreclosures. This chapter is organized into five sections. Section One discusses the descriptive market and correlations among variables. Section Two analyzes 2010 foreclosure filing data. Section Three analyzes 2008 foreclosure filing data. Section Four analyzes 2000 foreclosure filings. Finally, Section Five discusses the collective impact of these findings on the hypotheses. Description of Market The single family residential (SFR) mortgage market represents about $11 trillion (Frame & White, 2010). In 2010 alone, nearly 2.9 million U.S. homes received notice of default (pre-foreclosure filings) according to RealtyTrac (2011). The high foreclosure amount represents a large economic loss to our society. To gain insight from the foreclosure filing data, first the descriptive statistics from the City of Chicago were examined (see Table 1). An interesting change from 2000 to 2010 in Chicago was that the selected monthly owner cost (SMOC) increased 60%, while the median household (HH) owner-occupied (OO) income increased by just half as much (U.S. Census Bureau, 2000a, 2000b; U.S. Census Bureau, 2010a, 2010b). The city wide disposable income (DI) calculated from the SMOC illustrated a 24.2% increase from $1,273 to $1,582 a month from This was a positive sign for the City of Chicago residential owners. However, this economic progress was not demonstrated in the debt to income (DTI) ratio. In 2000, the DTI ratio at the origination of the mortgage was 27% and escalated to 33.2% in 2010.

39 28 Table 1 Descriptive Statistics City of Chicago Description Percent Change Population* 2,695,598 2,896, % Single family Units** 355, , % Selected Monthly Owner Costs (SMOC)** Median Household (HH) Owner Occupied (OO) Income** $1,945 $1, % $70,330 $54, % DTI SMOC 33.2% 27.0% 23.00% DI SMOC $1,582 $1, % Average HH size of OO Units** % Poverty Percent** 20.9% 19.6% 6.60% Unemployment Percent 10.8% 4.8% % Single family Home Value** $161,892 $130, % 1-4 Unit Residential Home Value*** $189,817 $151, % 1-4 Unit Residential New Mortgage Amount Average $220,374 $133, % 1-4 Residential Units LTV 116.1% 88.3% 31.52% * Source: 2010 and 2000 Decennial U.S. Census ** Source: 2010 American Community Survey and 2000 Decennial Census *** Source: Zillow, Inc., 2012 Source: U.S. Department of Labor: Bureau of Labor Statistics, Cook County, IL, 2012 Source: 2010 and 2000 HMDA data, Note: The average Chicago new mortgage amounts are likely weighted differently than the average Chicago home value.

40 29 This change illustrated that underwriting standards became lax over this period. Conventional loan underwriting requires a limit of the front-end DTI ratio at 28%, while the Federal Housing Administration (FHA) limits the front-end DTI ratio at 31%. The rise in loan amounts, which is propelled by home price increases, made owning a home less affordable and in turn led to the increase in DTI limits. The rise in the City of Chicago DTI ratios occurred despite the downward movement of the national 30 year fixed conventional interest rates. In 2000, the average national average conventional mortgage rate was 8.05%, while in 2010 it shrank to 4.69% (see Figure 1). Figure year fixed rate mortgage interest rate. The loan fees (points) also reduced from 1.0 to 0.7 (Freddie Mac, 2012). In 2007, lending underwriting standards became stricter. The loan volume decreased because of stricter lending standards. The refinance market slowed because numerous borrowers

41 30 were underwater. Purchases slowed because borrowers were fearful of the crashing home values (see Figure 2). The peak in loan volume in 2006 was reflective of the lax underwriting standards. Figure 2. HMDA conventional purchase and refinance loan volume for the city of Chicago. The rise of speculative purchases also occurred because of buyers faith that home values always rise and at a high rate. From , a single family home investor in Chicago could make a rate of return of 14.1% a year on home value appreciation. This rate of return far exceeded the annualized rate of return from the Standard and Poor (S&P) 500 of 1.5% for the same seven year period. The loan volume peak corresponded with the top of single family home value bubble (see Figures 2 and 3). The economic recession demonstrated a 123.8% increase in unemployment from 2000 to 2010 in Cook County, Illinois. The average unemployment rate jumped from 4.8% in 2000 to 10.8% in 2010.

42 31 Figure 3. Chicago single family home values. The unemployment rate changed dramatically over the last decade (see Figure 4). Figure 4. Cook County unemployment rate.

43 32 An increase in unemployment together with a decrease in home values is considered a double-trigger event that results in increase in defaults (Foote et al., 2009). As Figures 3 and 4 indicate, high levels of unemployment rates existed at the same time home values plummeted. The peak of the housing bubble also represented the lowest unemployment levels. The average single family residence (SFR) value peaked September 2006 at $252,100 (Zillow, 2012). This peak was a 93.5% increase from the 2000 average value of $130,292. The 2010 home values represented a -35.6% drop from the bubble bursting point. In order to estimate the City of Chicago s loan to value (LTV) ratio, the average of the Home Mortgage Disclosure Act (HMDA) residential 1-4 unit purchases and refinances were divided by the Chicago home values via Zillow (see Table 1). The result illustrated 2010 Chicago loan origination LTV ratio 27.8% higher than 2000 s LTV ratio. The foreclosures rose over the observed years from (see Table 2). From , Chicago foreclosure rates were relatively stable; starting in 2006 foreclosures escalated at a higher rate. While this study focuses on single family residential (SFR) foreclosures, it is interesting to review how 2-4 multi-unit residences and condominiums performed throughout the foreclosure crisis. The 2-4 unit multi-unit foreclosures rose from just 714 to 7,325 in 2008, over 10 times as much as in Condominium foreclosures jumped from just 171 in 2000 to 6,245 in 2010, a 3,552% gain. Another perspective is to review the total of SFR, 2-4 units, and condominium foreclosures divided by the total owner-occupied (OO) housing units with a mortgage. This measure resulted in doubling the rate of foreclosure from 3.1% in 2000 to 6.74% in

44 The SFR foreclosures rates increased from 2000, 2008 to 2010 were 1.4%, 2.8% and 3.0%, respectively. Table 2 City of Chicago Foreclosure Counts and Rates Type of Housing Single family Foreclosures *,** 10,594 9,929 4,621 Condominium Foreclosures *** 6,245 4, Multi-Unit Residences Foreclosures **, 5,822 7, Total Foreclosures *,**, 22,661 22,012 5,506 Single family Units* 355, , , Unit Multi Family Housing Units* 379, , ,983 Owner-Occupied Housing Units,* with a Mortgage 363, , ,801 Total Housing Units* 1,197,741 1,188,767 1,152,871 Single Family Foreclosure Rate 2.98% 2.84% 1.42% 2-4 Unit Multi-Unit Foreclosure Rate Total Foreclosures to OO Housing Units with a Mortgage Ratio 1.54% 1.94% 0.19% 6.24% 6.74% 3.05% Total Foreclosure Rate 1.89% 1.85% 0.48% Foreclosure data was obtained from Record Information Services, and reflects pre-foreclosure filing data available at the Cook County, IL Chancery Department. * 2010 Housing totals are based from the 2010 and 2000 Decennial U.S. Census, 2008 is estimated assuming consistent change. ** Data excludes SFR with values under $45,000, construction, or vacant land. *** U.S. Census data is unclear on condominium amount totals. Data excludes construction properties, vacant land, construction, commercial, and industrial properties. Data utilizes 2010 American Community Survey 5-year to estimate 2010 and 2008 figures.

45 34 The descriptive statistics of the foreclosure data provided a good comparison view of how the underwriting components changed for 2000, 2008 and 2010 (see Table 3). Table 3 Descriptive statistics City of Chicago Foreclosure Data Foreclosure Data 2010 st.dev st.dev * st.dev. DTI ARM adj. 41.2% % % LTV** 88.1% % % DI ARM adj. $ $ $ Interest Rate Premium*** 0.82% % % Interest Rate 7.02% % % ARM Percent 29.4% % n/a n/a Home Value Percent Change from Origination Mortgage Origination Amount -13.2% % % $205, ,281 $205, ,181 $84,166 39,922 Loan Duration Home Value at Origination** Mortgage Payment ARM adj. $238, ,376 $229, ,153 $94,887 44,988 $1, $1, $ *Disposable Income, DTI, and mortgage payment for 2000 data are not ARM adjusted, because ARM data was not available. ** Neighborhood home values were derived from Zillow, Inc., *** Interest rate premiums were estimated using Freddie Mac 30-year fixed mortgage rates, The DTI ratio ranged from 27.9% in 2000 to 45.0% in Overall, 2008 foreclosures appeared to be the riskiest, as the Disposable Income was $300 lower than 2010 and

46 , and the loans consisted of a higher percentage of ARMs, with 47.5%. Interestingly, in 2000 the highest interest rate premium at 2.47% was above the Freddie Mac average rate for the year of the mortgage originations. The interest rates and premiums decreased to 1.59% in 2008 and 0.82% in Interestingly, the LTV ratio of foreclosed data was slightly higher in 2000 (92.2%) compared to 2008 (90.3%) and 2010 (88.1%) despite similar standard deviations (see Table 3). This may be attributed to the potentially inflated estimate of 18.2% of the 2000 foreclosure loans that originated prior to 1996 (see Figure 5). Zillow s home value estimate data was not available prior to 1996, thus because of this, from , 2.5% home value growth was utilized as an estimated yearly reduction in value. Perhaps, this blanket estimation was too drastic and underestimated the value of houses for these older loans, which may have resulted in a slightly higher LTV ratio. Figure 5. Chicago 2000 single family residence (SFR) foreclosure origination years.

47 36 Taking this into account, s LTV ratio of the foreclosed data average was around 90%, which is higher than the conventional mortgage norm of 80%. When loans originate at higher than 80% LTV, higher interest rates and/or private mortgage insurance is required to compensate the lender for the higher risk. The result is a higher mortgage payment. With large increases in home values from (see Figure 3), many borrowers sought out refinances from , and 2006 (see Figure 2). During the refinance surging periods, borrowers took advantage of refinancing to qualify for 80% LTV (with lower payments), additional financing to pay consumer debt, and/or to perform home improvements. In addition, low interest rates compared to historic rates (see Figure 1) also provided financial savings for many borrowers to refinance. Home value depreciation is known to be a factor in the high rates of foreclosure. However, the home value change from loan origination to foreclosure filing was an appreciation of 14.5% for 2008 foreclosures and 12.3% for The average 2008 foreclosure, on the surface, appeared to be not impacted from the change in home values. This was despite that 2008 accumulated twice as many foreclosures compared to Perhaps the home value decline, anticipation of continued home value decline, or the restrictions of credit to refinance or purchase caused the foreclosure increases in Rising unemployment and an economic slow-down due to the 2008 financial collapse likely contributed to the spike in foreclosures. In addition, while the average foreclosure in 2008 experienced home value appreciation since the loan s inception, tighter lending standards and a sharply slowing realtor market created a difficult environment to refinance or sell a home. Employment security fears along with concerns that home values would drop even further created panic for home buying.

48 37 The -13.2% home value depreciation from loan origination to the 2010 foreclosure data illustrated a more ideal environment for borrowers to walk away from their payment obligation. This combined with very high prolonged unemployment rates and high and increasing inventory levels of foreclosed properties created several neighborhoods of blight and poverty. By 2010, borrowers were more likely to have known or lived near a borrower that had been foreclosed upon. Perhaps the foreclosure contagion effect that researchers Goodstein, Hanouna, Ramirez, and Stahel (2011) found truly made it easier for borrowers to resend their fiduciary mortgage payment obligations Foreclosure Analysis Each underwriting component, disposable income (DI), debt to income (DTI) ratios, and loan to value (LTV) ratios possess a unique interaction with foreclosures (see Figures 6-8). Lending underwriting policies represent a preventive step to ensure that the borrowers are capable of making their financial obligations and that the risk of the loan creation is low. A regression analysis of 2010 foreclosures illustrates that DI, DTI ratios, and LTV ratios were significant (p < 0.01, see Table 4) contributors to foreclosure rates. Table Regression Analysis Coefficients a Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta (Constant) DI DTI LTV a. Dependent Variable: Foreclosure Rate 2010 Adjusted R 2 = F = df = 737

49 38 Figure 6. Scatter Plot 2010 Disposable income (DI) and selected monthly income cost (SMOC). Figure 7. Scatter Plot Debt-to-income ratio and SMOC.

50 39 Figure 8. Scatter Plot 2010 Loan-to-value ratio. A decrease in DI by -7.7% increases foreclosures by 1%. Much of the foreclosed household income is based off owner-occupied household income levels (see Figure 9). This period represents a period of low unemployment levels and high mortgage amounts. Cook County, Illinois, unemployment rate was 5.0% for at the time of most mortgage originations, compared to 2010 s 10.8% levels (see Figure 4). However, higher unemployment levels were not alone in stimulating the increase in foreclosures. Real DI was distorted on loan applications by the commonly utilized stated-income loans. This increased the potential for borrowers to default. Areas of Chicago that had low DI levels were more likely to have lower savings and were more susceptible to income shock. Nevertheless, the significant impact of DI at the origin of the mortgage on the foreclosure rates in 2010, despite the changing market conditions, indicated that DI is an important indicator of mortgage risk.

51 40 Figure 9. Chicago 2010 SFR foreclosure origination years. DTI exhibited a negative relationship to foreclosures as a result of multicollinearity to DI (r = ). The collinearity statistics showed DTI and DI variance inflation factors (VIF) (1.53, 1.61) were within acceptable norm levels. The recommended cutoff point is above 10 for VIF (Stevens, 2001). The main components of monthly income and monthly housing expense were represented in both variables. The scatter plots (see Figures 6, 7) illustrate DI possesses a stronger forecasting impact compared to DTI ratio. The DTI ratio is a measure of risk and indicates if a borrower is living within their means. DTI ratio, as it relates to affordability is relative to borrowers income levels. Therefore, the DTI ratio can be a misleading underwriting tool. For example, the following scenario illustrates an interesting depiction between two DTI ratios with different interpretations.

52 41 Borrower A DTI = 55% Household income of $150,000/ year Borrower B DTI = 40% Household income $40,000/ year Disposable Income (DI) formula = Monthly income x (1- tax rate) (DTI ratio x monthly income) est. consumer debt payment (family size estimate x $100) DI = ($150,000/ 12) x (1 25%) (0.55 x ($150,000/12)) $700 (3 x $100) = DI = $1,500 DI = ($40,000/12) x (1 25%) (0.40 x ($40,000/12)) $700 (3 x $100) = DI = $167 On the surface, the 40% DTI ratio of borrower B appears to be a much safer investment compared to the 55% DTI ratio of borrower A. However, the additional information of income makes a significant difference as borrower A possesses a higher cash flow savings cushion. In addition, if both had the average 2010 foreclosed data LTV of 88.1%, and both were to purchase loans, it would indicate borrower A would have a much higher savings committed as down-payment compared to borrower B. As the loan duration continues, borrower A has the potential to amass a 10-month debt payment savings reserve, compared to less than a 4-month reserve of borrower B. This assumes 4.5-year duration and no income generation increase. When LTV ratio is raised by 1.7%, foreclosure rates are impacted by a 1% increase, which indicates that LTV is a more significant predictor of foreclosure rates than DI. The LTV ratio does not just reflect higher risk, but it also represents insight to borrowers available savings. While it is possible for borrowers to have large savings and a 100% LTV ratio, it is unlikely that borrowers would have a low LTV ratio with a small amount in savings. Alternatively, with interest rates being at historic lows,

53 42 sophisticated borrowers could have leverage investments with the anticipation of obtaining a higher net return. The lower LTV ratios also represent a higher level of skin in the game that could turn into realized financial losses if the borrower were to default. With home values declining from 2007, it is possible that speculating investors decided to stop making payments and allowed their properties to foreclose. If the loans were originated with a high LTV ratio, then refinancing or selling the home becomes not probable, unless additional savings are utilized at a loss. When loans occur with a small or no down-payment, investors are encouraged to exit their commitments when current and future short-term outlooks on returns are negative. Other variables that were not used but were viewed to better understand the underwriting components relationships can be viewed in Table 5 in a correlation matrix. The variables interest rate premium, ARM, and house price appreciation from origination were excluded from the regression analysis because they did not add substantial value to the existing adjusted R 2 of Home price appreciation from the origination amount until the second half of 2009 impacted foreclosures (r = , p < 0.01). Interest rate premiums were analyzed and at first glance appeared to be a factor (r = 0.134, p < 0.01). The percentage of ARM foreclosures in census tracts was found uncorrelated and not impacting predicting foreclosures (r = , p > 0.1). The home price appreciation from origination, ARMs, and interest rate premiums were reviewed through correlation and scatter plots (see Figures 10, 11, 12). The 2010 ARMs had a 64.5% (s = 0.18) probability of adjusting before preforeclosure filings from the lenders (see Tables 5, 6, 7). This indicated that these ARM foreclosures were negatively impacted by the rate adjustment.

54 43 Table Correlation Matrix Foreclosure Rate DI SMOC DTI SMOC LTV DI DTI Interest Rate Interest Rate Premium Foreclosure Rate **.081*.267** -.256**.104*.127**.134** DI SMOC -.360** ** -.214**.723** -.178** -.352** -.360** DTI SMOC.081* -.571** **.487** LTV.267** -.214** ** -.124* DI -.256**.723** -.607**.128** ** DTI.104* -.178**.487** -.124* -.763** ** -.267** Interest Rate.127** -.352** ** 1.946** Interest Rate Premium.134** -.360** **.946** 1 ARM Pct **.296** -.102* Org. Mtg. Amt **.520** **.677** -.528** -.521** Pct. Chg. from Org. HH Income OO CPI Adj 1999 Pct. Below Poverty Level -.116** -.383**.158**.084*.936** -.384** **.240** * ** -.340** -.213** ** -.439**.425**.253** -.438**.418** * Correlation is significant at the 0.05 level (2-tailed). ** Correlation is significant at the 0.01 level (2-tailed). ARM Pct **.296** -.102* ** -.181**.162** -.121* Org. Mtg. Amt **.520** **.677** -.528** -.521**.351** **.675** -.441** Pct. Chg. from Org **.084* **.331**.253** -.181** -.324** ** HH Income OO CPI Adj **.936** -.307** -.098*.609** ** -.438**.162**.675** ** Pct. Below Poverty Level.158** -.38**.240** ** **.418** -.121* -.44**.164** -.46** 1

55 44 Figure 10. Scatter Plot 2010 Home value percent change from origination. Figure 11. Scatter Plot 2010 Adjustable rate mortgage (ARM).

56 45 Figure 12. Scatter Plot 2010 Interest rate premium. While this adjustment rate was high in comparison to the overall loans, only 18.9% (s = 0.31) of all foreclosed loans possessed ARMs that adjusted. This is relatively similar to the findings of Bhutta et al. (2010) in which the interest rate change impacted less than 10% of their extensive LP data set that included current and default loans. The overall DI and DTI from the 2010 foreclosure ARM data decreased by -$73 a month and increased just a 1.5%, respectively. This assumption assumed a 2.5% margin increase adjustment. The vast majority of the 2010 foreclosed ARMs had durations between 3-5 years. These loans were likely underwater and were not able to be refinanced out of the loan as they may had anticipated when the loan was first originated. These estimates assume that the ARM reset terms are similar to the convenient sample of 200 ARMs from Cook County, Illinois, Chancery Division residential mortgages foreclosures (Tables 6, 7, 8).

57 46 Table 6 Adjustable Rate Mortgage (ARM) Duration Total Reviewed Adjusted Before Default Percent Defaulted After Rate Adjustment ARMs Duration 6-month or less % 1-year % 2-year % 3-year % 5-year % 7-year % 10-year % 15-year % Total % Note: Duration is defined as time between loan origination and pre-foreclosure filing date Table 7 ARM Loan Duration Probability of ARM Rate Adjustment 2010 ARM Loan Duration Distribution 2008 ARM Loan Duration Distribution Loan Duration Duration 0 > yr 0.5 year 0.0% 0.2% 0.2% Duration > 0.5yr 1 yr 12.0% 0.5% 5.0% Duration > 1yr 2 yr 17.0% 2.2% 31.3% Duration > 2yr 3yr 42.5% 9.0% 38.0% Duration > 3yr 5yr 59.0% 61.4% 21.3% Duration > 5yr 7yr 89.5% 21.9% 1.8% Duration > 7yr 10yr 96.0% 2.7% 1.5% Duration > 10yr 15yr 99.0% 1.3% 0.6% Duration > 15yr 100.0% 0.1% 0.1% Unknown % 0.0%

58 47 Table 8 ARM Foreclosures 2010 ARM Foreclosures 2008 ARM Foreclosures Probability of ARM Rate Adjustment 64.5% 38.5% Standard Deviation Probability of All Loans Rate Adjustment 18.9% 18.3% Standard Deviation Home price appreciation from loan origination was a post underwriting variable and was considered a covariance factor and not a dependent variable. Interest rate premiums were reflective, to the most part, of the combination of the DI, DTI, and LTV risk factors. Interest rate premiums cause collinearity of the combined variables and drastically reduced the adjusted R 2 to The ARM variable set was eliminated because it also drastically reduced the adjusted R 2. ARMs were shown to have no correlation to foreclosure rates (see Table 5) Foreclosure Analysis Foreclosures in 2008 possessed similar results as 2010; however, 2008 borrowers had higher risk characteristics. The underwriting components DI, DTI ratio, and LTV ratio were all significant (p < 0.01), but with a higher adjusted R 2 of This illustrated that in 2008, these components were even stronger predictors of foreclosures than in Similar to 2010, an 8.0% decrease in the DI or a 2.9% increase in a borrower s LTV ratio would lead to a 1% increase in foreclosures (see Table 9). DTI exhibited a negative relationship to foreclosures as a result of multicollinearity to DI (r = ). The collinearity statistics showed DTI and DI variance inflation factors (VIF) (1.61, 1.91) were within acceptable norm levels.

59 48 Table Regression Analysis Coefficients a Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta (Constant) DI DTI LTV a. Dependent Variable: Foreclosure Rate 2008 Adjusted R 2 = F = df = 732 One interesting difference was that the percent change of home values from the origination demonstrated a significant positive correlation to foreclosures in 2008 (r = 0.343, p < 0.01). This illustrated that areas which possessed home value bubbles from the fourth quarter in 2005 until the second half of 2007 were areas that were more probable to yield foreclosures (see Figure 3). The 2008 foreclosure loan origination dates (see Figure 13) align similarly to this home bubble. This is in contrast to 2010 foreclosures that had a negative correlation (r = , p < 0.01) because of home value decreases. The foreclosures in 2008 illustrated a higher percentage of ARMs than in 2010, 45.5% compared to 29.4%. However, ARMs were revealed to be unrelated to foreclosure rates (r = , p > 0.10) [see Table 10, Figure 14]. From these ARMs, it was estimated that only 38.5% (s = 0.20) actually adjusted before foreclosure filing procedures started. The overall impact resulted in only 18.3% (s = 0.24) of all foreclosed loans that adjusted their rate before the foreclosure filing. Similar to 2010, 2008 ARM

60 49 Figure 13. Chicago 2008 single family residence (SFR) foreclosure origination years. Figure 14. Scatter Plot 2008 Scatter plot Adjustable rate mortgage (ARM).

61 50 adjustments changed the foreclosure data set by a decrease of the overall average DI by $68 and an increase of DTI of 1.5%. These estimates assumed that the ARM reset terms are similar to the convenient sample of 200 ARMs from Cook County, Illinois, Chancery Division residential mortgages foreclosures (see Tables 6, 7, 8). To add a visual of the 2008 results, DI, DTI, and LTV ratio are illustrated in scatter plots (see Figures 15, 16, 17). These scatter plot diagrams illustrate a relationship between these variables with foreclosure rates. Other variables not utilized in the regression analysis of interest were also plotted. Home value percent change from origination (see Figure 18) and interest rate premium (see Figure 19) both showed a strong positive association with the 2008 foreclosures. Interest rate premium essentially is the result of assigned underwriting risk. The home value percent change from origination was related to the strong association of home price bubble areas that burst. Figure 15. Scatter plot 2008 Disposable income (DI) and selected monthly owner cost (SMOC).

62 51 Table Correlation Matrix Foreclosure Rate DI SMOC DTI SMOC LTV DI DTI Interest Rate Interest Rate Premium Foreclosure Rate ** ** -.269**.141**.371**.361** DI SMOC -.422** ** -.402**.575** -.211** -.257** -.265** DTI SMOC ** ** -.615**.518** LTV.404** -.402** -.123** DI -.269**.575** -.615** ** DTI.141** -.211**.518** ** ** -.214** Interest Rate.371** -.257** ** 1.973** Interest Rate Premium.361** -.265** **.973** 1 ARM Pct **.146** Org. Mtg. Amt **.490** ** -.337**.619** -.443** -.440** Pct. Chg. from Org. HH Income OO CPI Adj Pct. Below Poverty Level.322** -.472**.343** -.129**.939** -.412** **.236**.177** -.518**.101*.162**.424** -.335** -.244** ** -.342**.417**.216** -.346**.407** * Correlation is significant at the 0.05 level (2-tailed). ** Correlation is significant at the 0.01 level (2-tailed). ARM Pct **.146** ** -.246** ** Org. Mtg. Amt **.490** ** -.337**.619** -.443** -.440**.165** **.663** -.380** Pct. Chg. from Org..322** -.129** **.162** -.244**.313**.216** -.246** -.310** **.209** Income OO CPI Adj **.939** -.295** -.518**.424** ** -.346** ** -.185** ** Pct. Below Poverty Level.343** -.412**.236**.101* -.335** **.407** -.140** -.380**.209** -.446** 1

63 52 Figure 16. Scatter plot 2008 Debt-to-income (DTI) ratio and selected monthly owner cost (SMOC). Figure 17. Scatter plot 2008 Loan-to-value (LTV) ratios.

64 53 Figure 18. Scatter plot 2008 Home value percent change from origination. Figure 19. Scatter plot 2008 Interest rate premium.

65 Foreclosure Analysis The 2000 foreclosures represented a period of time of relative normalcy in regards to economic conditions and the housing market. Compared to the inflated home values of 2006, the stable home prices yielded lower mortgage payments and lower DTI ratios. This occurred despite much higher interest rate premiums from the 2000 s foreclosed homes. The lower DTI ratios (x = 27.9%) and tighter dispersion (s = 0.13) may have contributed to the low relationship to foreclosures in 2000 (r = 0.20, p > 0.10). In addition, in a normal economic and underwriting environment, foreclosures are more likely to be related to random shock triggers, such as job loss, health problems, and divorce. The VIF of DTI and DI illustrated an acceptable collinearity amount (1.55, 1.72). The regression analysis (see Table 11) indicated DI and LTV were strong contributions to foreclosures (p < 0.01). A -8.6% decline in DI and a 1.8% increase in LTV at origination led to a 1% increase in 2000 foreclosure rates. The 2000 correlation matrix (see Table 12) illustrates that DI, DTI ratio, and LTV ratios had a relationship with 2000 foreclosure rates. Table Regression Analysis Coefficients a Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta (Constant) DI DTI LTV a. Dependent Variable: Foreclosure Rate 2000 Adjusted R 2 = F = df = 718

66 55 Table Correlation Matrix Foreclosure Rate DI SMOC DTI SMOC LTV DI DTI Interest Rate Interest Rate Premium Foreclosure Rate ** ** -.269**.141**.371**.361** DI SMOC -.422** ** -.402**.575** -.211** -.257** -.265** DTI SMOC ** ** -.615**.518** LTV.404** -.402** -.123** DI -.269**.575** -.615** ** DTI.141** -.211**.518** ** ** -.214** Interest Rate.371** -.257** ** 1.973** Interest Rate Premium.361** -.265** **.973** 1 ARM Pct **.146** Org. Mtg. Amt **.490** ** -.337**.619** -.443** -.440** Pct. Chg. from Org. HH Income OO CPI Adj Pct. Below Poverty Level.322** -.472**.343** -.129**.939** -.412** **.236**.177** -.518**.101*.162**.424** -.335** -.244** ** -.342**.417**.216** -.346**.407** * Correlation is significant at the 0.05 level (2-tailed). ** Correlation is significant at the 0.01 level (2-tailed). ARM Pct **.146** ** -.246** ** Org. Mtg. Amt **.490** ** -.337**.619** -.443** -.440**.165** **.663** -.380** Pct. Chg. from Org..322** -.129** **.162** -.244**.313**.216** -.246** -.310** **.209** Income OO CPI Adj **.939** -.295** -.518**.424** ** -.346** ** -.185** ** Pct. Below Poverty Level.343** -.412**.236**.101* -.335** **.407** -.140** -.380**.209** -.446** 1

67 56 While stated income loans were available from , the stated income loans were more commonly utilized from Haughowout, Lee, Tracy, & VanDerKlaauw s (2011) calculations through Loan Performance (LP) data showed an increasing trend of low document loans of 55.56% of Alt-A loans and 21.76% subprime loans in 2001 compared to 76.56% and 38.00% in 2006, respectively. In addition, Agarwal, Ambrose, Chomsisengphet, and Sanders (2012) found that in their 2000 data, 75% were full documentation loans compared to just 41% in full documentation loans in These higher risk loans were dependent on the honesty of the borrower and broker. With this in mind, it is not surprising that the DTI ratio was a significant variable in the 2010 and 2008 regression analysis and was not in This implies that in 2010 and 2008, owner occupied census tract household income was an excellent estimation contributing to DTI estimates. The 2000 foreclosure data DTI ratios were not a reliable indicator because stated loans were less utilized and because comparatively lower home values provided much lower mortgage payments. The average DTI ratios corresponded with the findings that limited document loans were 19.8% more likely to default than full document loans (Danis & Pennington-Cross, 2008). The 2000 results are illustrated in scatter plots for the DI, DTI ratio, and LTV ratios (see Figures 20, 21, 22). The strongest of these 2000 foreclosures to variable relationship were the DI and LTV ratios. Two additional variables not utilized in the regression analysis were also plotted. The home value percentage from origination (see Figure 23) and the interest rate premium (see Figure 24) scatter plots demonstrated strong relationships with foreclosures. Interestingly, unlike the positive home value percentage change from origination charges 2010 and 2008, 2000 conveyed a negative relationship.

68 57 This illustrated that prior to 2000, home values increased at a normal pace, and that areas with high home values since origination had low foreclosure rates because those areas likely had higher equity built up. For 2008 and 2010 foreclosures, the opposite was true because of the housing price bubble. Figure 20. Scatter plot 2000 Disposable income (DI) and selected monthly owner cost (SMOC). Collective Results The foreclosures filing years of 2000, 2008, and 2010 illustrate three different time periods of foreclosure tendencies. In 2000, foreclosure filings represented a more normal dispersion of foreclosures. It is implied that many of the default reasons had less to do with the DTI ratio, which corresponds to less stated income loans. While 2008 and 2010 foreclosures were closely related to DI, LTV, and DTI causes, they were also related to the housing value market movements. For 2008, foreclosures were strongly related to areas that demonstrated depreciated home price values.

69 58 Figure 21. Scatter plot 2000 Debt-to-income (DTI) ratio and selected monthly owner cost (SMOC). Figure 22. Scatter plot 2000 Loan-to-value ratio (LTV).

70 59 Figure 23. Scatter plot 2000 Home value percent change from origination. Figure 24. Scatter plot 2000 Interest rate premium.