The Downfall of Freddie Mac
Changes in borrowing conditions between 1999 and 2013
Student: Sara Blokzijl Student number: 10070818
MSc Economics: Monetary Policy and Banking Faculty of Economics and Business
University of Amsterdam
First supervisor: Dr. K. Mavromatis Second supervisor: Dr. W.E. Romp
2
Abstract
The housing bubble that developed in the United States led to the financial crisis when the bubble burst in 2007. This thesis explores the changes in borrowing conditions between 1999 and 2013 for mortgages issued and purchased by Freddie Mac. The research focuses on averages, correlations and regressions to show the changes over time and especially around the financial crisis. The interest rate, credit score, LTV ratio and DTI ratio are the four most important variables to help answering the research question: To what extent did borrowing conditions change between 1999 and 2013 for mortgages sold to and provided by Freddie Mac? Borrowing conditions clearly loosened in the years before the crisis, and fluctuated heavily around the start of the financial crisis. To provide more insight in the years before the financial crisis, this thesis includes a chapter that focuses on the differences in borrowing conditions between prime borrowers and subprime borrowers. This chapter shows that borrowing conditions were not tighter before the financial crisis, and that the correlations between variables are different for subprime borrowers.
3
Contents
Abstract ... 2
1 Introduction ... 4
2 Literature review ... 6
2.1 The Great Recession ... 6
3 Data description and analysis ... 10
3.1 Averages ... 10
3.2 Correlations ... 13
4 Empirical strategy and results ... 16
4.1 The loan-to-value ratio ... 16
4.2 The credit score ... 18
4.3 Collinearity ... 19 4.4 Summary ... 19 5 Subprime borrowers ... 21 5.1 Averages ... 21 5.2 Correlations ... 22 5.3 Regressions ... 26 5.4 Summary ... 28 6 Conclusion ... 30 References ... 32
4
1 Introduction
In September 2006, Nouriel Roubini gave a now famous speech at the IMF. This professor at NYU quite accurately predicted the financial crisis that started the following year, currently known as the Great Recession.
“Once the bursting of the housing bubble occurs, you are going to have a fall in prices, a fall in demand for credit. Of course, at that point — after the crunch has occurred in terms of real economic activity in the housing sector falling — you are going to see some credit tightening because you are going to see increasing amounts of delinquencies, defaults, and foreclosures.” (Roubini, 2010)
Although a few economists predicted the recent financial crisis, most people ignored the warnings as they expected markets to be self-regulating, and did not expect the substantial effect that financial imbalances would have on the real economy (Roubini & Mihm, 2010).
Unfortunately, crises are not the exception rather than the rule. There is a basic pattern associated with financial crises, which will be explored later on. The Great Recession started with a bubble in the housing market. One seller and provider of mortgages was Freddie Mac, a government sponsored enterprise (GSE) created by the United States government. When financial innovation, deregulation and asymmetric information made mortgages increasingly attractive as an investment, a bubble started to develop.
This thesis explores the changes in borrowing conditions for mortgages sold to and provided by Freddie Mac, thereby focusing on the changes that can be observed during the bubble build-up and the differences before and after the financial crisis. The main question this thesis tries to answer is therefore:
“To what extent did borrowing conditions change between 1999 and 2013 for mortgages sold to and provided by Freddie Mac?”
In answering this question, I carefully look at the build-up towards the financial crisis, the contraction period during the start of the financial crisis, and the changes after the start of the financial crisis. As the share of subprime mortgages declined substantially after 2007, a separate chapter with features of subprime mortgages is included as well to show the differences in borrowing conditions for average mortgages and subprime mortgages.
5
This is economically relevant, because the outcomes can provide insight in the characteristics of the bubble and the effects of the financial crisis on the housing market. As the changes in borrowing conditions over time have not been explored explicitly for Freddie Mac, this thesis can make a contribution to the existing literature.
The data is explored using cross-sectional regressions for each year separately. The results are placed in a timeline to show annual changes in signs and size. I emphasize the differences before and after the financial crisis started. The dependent and independent variables vary, to get a clear view on the effects that the variables have on each other. A collinearity test is conducted to make sure that the estimates are not biased by the effects that the variables have on each other.
The results show signs of deterioration in borrowing conditions in the bubble years before the financial crisis, and a sharp increase in restrictions around the start of the financial crisis. The results for subprime mortgages differ substantially from the total dataset, because other factors play a role in determining borrowing conditions for these mortgages. The effects and significance of the variables change tremendously around the start of the financial crisis. It becomes clear that the borrowing conditions for subprime borrowers were not as tight as expected.
The rest of the thesis has the following order: First, the theoretical background is provided. Second, the data is described and explored with averages and correlations. Third, the results from the first set of regressions are discussed. Fourth, one chapter focuses solely on subprime mortgages, adding a separate data description, correlations and regression results. Lastly, the thesis is concluded.
6
2 Literature review
2.1 The Great Recession
The economy is characterized by a business cycle, consisting of booms and busts. Most crises begin when a boom turns into a bubble. Investors become overconfident and take on too much risk, often accompanied by heavy borrowing. In the recent financial crisis, real estate became the overvalued asset. As housing prices were expected to increase continuously, investments in real estate increased and credit supply grew excessively. Homeowners were able to expand their mortgage and therefore obtain cash relatively easy. This caused a drop in savings and growing debt obligations in the United States. Although this situation is unsustainable, the economy bloomed as the easy access to cash fostered economic growth (Roubini et al., 2010).
This bubble –as previous ones- benefitted from financial innovation, where investors focused on a new type of asset: mortgage backed securities (MBS). A MBS is a large amount of mortgages pooled and sold as one asset. The pooling of assets is also known as securitization. As the MBS consists of a variety of mortgages, it was considered to be safer than purchasing a single mortgage. When one mortgage defaults, it has little impact on the total value of the asset (Roubini et al., 2010).
Because the demand for MBS increased substantially, it became more profitable to increase the supply of mortgages. Borrowing conditions became less strict, partially due to the fact that the mortgages would be securitized and sold to investors. The risks of these mortgages would disappear from the balance sheets of the companies that created these assets. When the housing prices reached their peak in 2006 and mortgages had been extended to borrowers that were ineligible, the number of mortgage defaults started to increase.
It is not surprising that bubbles go hand in hand with financial innovation. The risks of new type of assets are usually hard to estimate, and the assets become a relatively attractive investment as the increasing interest shifts up the value (Roubini et al., 2010).
2.2 Freddie Mac
One creator of MBS was Freddie Mac. Freddie Mac was created by Congress in 1970 and owned by the 12 Federal Home Loan Banks. The purpose of Freddie Mac was to buy mortgage loans from the savings and loan industry (S&L) and securitize them, but there were no restrictions on mortgage purchases from other originators. Limitations were placed on the size of the mortgage
7
that could be purchased by Freddie Mac, but this limit was above the median housing prices. Due to a combination of interest rate risk and deregulation of the industry, a lot of S&Ls became insolvent in the late 1980s and Freddie Mac became publicly traded (Acharya et al., 2011).
Freddie Mac is a GSE, and is engaged in residential mortgage securitization and residential mortgage investment. When it comes to securitization, mortgages are bought from originators and pooled into MBS. Investors can purchase the MBS and then have claims on the interest and principles repayments from the mortgages included in the MBS. Freddie Mac charges a guarantee fee to investors in exchange for guarantees against the risk of default by borrowers of the underlying mortgages. Freddie Mac also invests in mortgages and MBS, which is financed by the issuance of debt. Because the debt is guaranteed by the U.S. government, issuing debt is relatively cheap. This provides an incentive for Freddie Mac to leverage itself as much as possible, since issuing debt is cheaper than issuing equity and risk free because of the guarantee (Acharya et al., 2011).
When the first MBS were traded in 1971, the secondary market for mortgages was very illiquid. Banks wanted to get rid of the large amount of mortgage loans on the balance sheet in order to raise more capital for investments, while investors were not interested in purchasing these assets. The creation of MBS made mortgages more attractive for large investors, such as pension funds. The increase in demand would expand the secondary mortgage market and improve borrowing conditions for homeowners (Acharya et al., 2011).
The 1980s resembled a period of deregulation in the mortgage finance market and substantial growth for Freddie Mac. In 1992, the Federal Housing Enterprises Financial Safety and Soundness Act passed. This legislation required a capital buffer of 0.45% and 2.5% capital against its balance sheet assets, which consisted mostly of mortgages. Since these capital requirements are extremely light, the government should have expected it had to step in when economic circumstances deteriorated. In addition to this, banks had to meet higher capital requirements for regular mortgages than for MBS on their balance sheets. The easy transition into MBS caused an increasing leverage in the financial sector. Next to the weak capital requirements, the 1992 Act also specified a set of “mission goals” to increase the availability of mortgages for low- and moderate-income households. This gave Freddie Mac the possibility to take up more risk by purchasing lower-quality mortgages (Acharya et al., 2011).
8 2.3 The bubble and bust
The Federal Reserve decreased interest rates by 5.5% between 2001 and 2003 to foster economic growth after the internet bubble in 2000, which increased the demand for mortgages. A housing bubble started to develop, with the help of easy credit and weak regulation. Because house prices were expected to increase continuously, borrowers had the opportunity to expand their mortgage and effectively using their house as an ATM machine. At the end of 2005, the equity withdrawals peaked and borrowers could live well beyond their means. By 2006, the spread between safe Treasury bonds and high-risk junk bonds decreased to historic lows, reflecting the undervaluation of risk in the financial markets (Roubini et al., 2010).
Chomsisengphet & Pennington-Cross (2006) show the quick evolution of the subprime mortgage market. The large share of subprime borrowers involved in mortgages and the high levels of indebtness had created substantial risk. After the house prices reached their peak, the growing number of defaults under subprime borrowers caused the first bankruptcies among lenders specializing in subprime loans at the end of 2006 and throughout 2007. Mayer, Pence and Sherlund (2008) describe the burst in the subprime mortgage market.
Freddie Mac played a substantial role in the bubble build-up. In 2003, accounting scandals were found at Freddie Mac and three board members were replaced. In order to regain trust from the public, Freddie Mac had to improve its reputation. The new affordable housing mission became an important instrument to maintain the support from Congress. Since issuing and purchasing subprime mortgages became the easiest way to support affordable housing, the share of subprime mortgages associated with Freddie Mac increased substantially. Freddie Mac signaled that it was willing to guarantee any type of mortgage, which substantially increased the amount of subprime mortgages. The political support was crucial for Freddie Mac in maintaining the special status with lower interest rates, weaker regulation and excessive profits. Between 1998 and 2008, Freddie Mac spent $94.9 million on lobbying Congress. Between 2005 and 2007, it seems that around 40% of the loans that Freddie Mac added to the books consisted of junk loans (Wallison et al, 2008). Thompson (2009) also describes the political influences involved with the bubble build-up.
The accounting practices Freddie Mac adopted made it very hard to correctly assess the risk. Paul Krugman, a famous economist, wrote in July 2008 in the New York Times:
9
“…So whatever bad incentives the implicit federal guarantee creates have been offset by the fact that Fannie and Freddie were and are tightly regulated with regard to the risks they can take. You could say that the Fannie-Freddie experience shows that regulation works.” (Wallison et al, 2008)
Although Freddie Mac did take part in the subprime mortgage lending, Paul Krugman (2013) explains that Freddie Mac was a follower rather than a leader. The housing bubble and credit bubble would have been created without the help of Freddie Mac and affordable housing plans, and a lot of subprime lending was undertaken by private lenders. The importance of financial regulation for the shadow banking system is described by Adrian & Shin (2009).
Freddie Mac was taken over by the government on September 7, 2008. The leverage ratio had increased to forty to one and the debt had partially been used to purchase risky mortgages and MBS. The GSE was sustaining massive losses due to two causes. First of all, the losses on their investment portfolios had become too large, as they were filled with subprime assets. Investors started to panic, because they doubted whether the government would actually guarantee the loans extended by Freddie Mac. Second of all, the fees that Freddie Mac received for the securitization of mortgages turned out to be too low to cover the losses. When the safe borrowers started to default as well, the default rates easily exceeded the pre-calculated default rates and Freddie Mac could no longer cover the insurances (Roubini, 2010). The effectiveness of the takeover is described by Frame (2009).
10
3 Data description and analysis
The data used in this thesis is obtained from the Freddie Mac’s Single Family Loan-Level Dataset. This dataset contains all mortgages that were sold to Freddie Mac or issued in Freddie Mac Participation Certificates from the first quarter of 1999 until the third quarter of 2013. All mortgages are 30-year fixed rate mortgages and have verified or waived documentation to measure the borrowers’ creditworthiness. For the graphs and regressions in this thesis, the data is converged to annual data. The most important variables will be discussed below.
3.1 Averages
The interest rate used in the data is the original interest rate as indicated on the mortgage note. Figure 1 shows that the average interest rate has in general decreased over the years. The maximum interest rate increased between 2008 and 2009 while the mean interest rate declined, which might be an indication of increasing risk aversion from mortgage suppliers.
The credit score summarizes the creditworthiness of the borrower. It is prepared by third parties and reflects the likelihood that the borrower will timely repay future obligations. The credit score can vary between 301 and 850. Although defining a subprime borrower depends on loan-to-value ratios as well, subprime borrowers are often defined as a borrower that has a credit score below 640. Figure 2 shows that average credit scores remained quite stable for the years in this dataset, except for the jump between 2007 and 2009. This shows the result of the financial crisis with respect to mortgage eligibility. For credit scores it is also very interesting to look at the minimum credit scores in the dataset.
0 2 4 6 8 10 12 Mean Min Max 0 200 400 600 800 1000 1999 2001 2003 2005 2007 2009 2011 2013 Mean Min Max
Figure 2: These graphs depict the minimum, the average, and the maximum interest rate in the sample between 1999 and 2013
Figure 1: These graphs depict the minimum, the average, and the maximum credit score in the sample between 1999 and 2013
11 Table 1: This table shows the share of subprime borrowers in the sample between 1999 and 2013
Until 2008, the dataset contained mortgages from borrowers that had credit scores below 400. Between 2008 and 2013, the minimum credit score has increased with almost 300 points.
We will also focus on the annual share of subprime borrowers in the dataset, borrowers with credit scores below 640. Table 1 shows how the financial crisis changed the borrowing conditions for mortgages issued by and purchased from Freddie Mac. 2008 becomes a turning point for the share of subprime borrowers. Where averages around 10% were repeated annually before 2008, this share declined 20 fold since the crisis.
The original loan-to-value (LTV) ratio is constructed by dividing the original mortgage loan by the purchase price of the house. The ratios in this dataset include values between 6% and 105%. Figure 3 shows that the average LTV ratio declined slightly between 2000 and 2004, and experienced a somewhat larger decline from72% in 2008 to 67% in 2009. Since 2009, the LTV ratio has been converging towards the pre-crisis LTV ratio averages, reaching 74% in 2013. The drop in LTV ratio averages around the financial crisis shows the caution with respect to credit availability from banks in providing mortgages. Other than that, the LTV ratio always fluctuated mildly but did not increase during the bubble, contradicting expectations. Since the minimum LTV ratio is 6 every year, it is not included in the graph.
Year 1999 2000 2001 2002 2003 2004 2005 2006 2007 Subprime share 9.93% 11.26% 10.46% 10.48% 6.75% 8.70% 8.07% 8.93% 9.80% Year 2008 2009 2010 2011 2012 2013 Subprime share 4.19% 0.56% 0.56% 0.47% 0.30% 0.39% 0 20 40 60 80 100 120 1999 2001 2003 2005 2007 2009 2011 2013 Mean Max
Figure 3: These graphs depict the average and the maximum value of the LTV ratio in the sample between 1999 and 2013
12 Table 2: This table depicts the share of mortgages in the sample with a DTI ratio above 50% between 1999 and 2013
Next to looking at LTV ratios, it is also interesting to look at debt-to-income (DTI) ratios. This ratio is constructed by dividing the sum of a borrower’s monthly debt payments by the total monthly income. The debt-to-income ratio provides a broader view of the financial situation of a borrower, as not only the current mortgage is taken into account but also the total indebtness. Figure 4 shows that there is a clear increase in the average DTI ratio between 2003 and 2008, while the DTI ratio experienced a substantial drop to pre-2003 values since the financial crisis. Because this increase is not visible when exploring the LTV ratio, the DTI ratio provides extra information of the total debt borrowers were taking on. The maximum DTI ratio has decreased substantially from 64 to 50 between 2012 and 2013. This seems like a large drop, but the share of borrowers with high DTI ratios has already been declining since the financial crisis. Since the minimum DTI in the dataset is 1 for every year, it is not included in the graph.
Table 2 presents the share of DTI ratios higher than 50% in the dataset. Although 50% is the maximum value in 2013, it was a common value before the financial crisis. The substantial decrease represents the change in the caution of mortgage suppliers since the financial crisis.
The house prices in this thesis have been constructed by dividing the original amount of the outstanding mortgage by the LTV ratio. Figure 5 shows that the house prices have increased enormously since 1999 until the financial crisis, reflecting the housing bubble. It may be
Year 1999 2000 2001 2002 2003 2004 2005 2006 2007 Share >50% 5.33% 6.05% 5.25% 7.09% 6.74% 9.18% 11.60% 13.21% 14.38% Year 2008 2009 2010 2011 2012 2013 Share >50% 15.69% 6.35% 1.76% 0.00% 0.00% 0.00% 0 100000 200000 300000 400000 500000 Mean
Figure 4: These graphs depict the average and maximum DTI ratio in the sample between 1999 and 2013
Figure 5: This graphs depicts the average house price in the sample between 1999 and 2013 0 10 20 30 40 50 60 70 1999 2001 2003 2005 2007 2009 2011 2013 Mean Max
13
surprising to see the large increase in average house prices between 2007 and 2009, as house prices are expected to decline because of the financial crisis. This can be explained by the fact that the borrowing conditions for mortgages in this sample tightened. As the share of mortgages of subprime borrowers decreased substantially, the dataset represents more creditworthy borrowers. Usually, these borrowers can afford more expensive houses. This is reflected by an increase in average house prices. The share of relatively cheap houses declined around the financial crisis.
3.2 Correlations
Figure 6 shows that the correlation between the credit score and the interest rate is negative, which is consistent with economic theory. Between 2003 and 2006 the correlation between credit scores and the interest rate was relatively low. This can be interpreted as the credit score being less of a factor in determining the interest rate. In 2008 a clear peak arises, resembling the increased importance of credit scores. After 2008, the correlation decreases again. This might also have to do with the fact that the share of subprime borrowers in the sample decreased.
The correlation between credit scores and house prices is positive, as borrowers with higher credit scores can afford more expensive houses. The correlation between credit scores and house prices increased between 2003 and 2007. This resembles the fact that a higher credit score had a higher impact on the size of the mortgage than since the financial crisis. With higher credit scores, borrowers could more easily obtain higher mortgages during that period..
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 Interest rate House price -0.3 -0.25 -0.2 -0.15 -0.1 -0.05 0 DTI LTV
Figure 6: These graphs depict the correlation between the credit score and the interest rate (blue line), and the correlation between the credit score and the house price (red line)
Figure 7: These graphs depict the correlation between the credit score and the DTI ratio (blue line), and the correlation between the credit score and the LTV ratio (red line)
14
Figure 7 shows that the correlation between credit scores the DTI ratio is negative, as borrowers with higher credit scores have a better financial situation. On the other hand, both a low DTI ratio and a low LTV ratio improve the credit score. Between 2003 and 2007, the correlation between credit scores and the DTI ratio was also relatively low, while the correlation between credit scores and LTV ratios continuously decreased. This resembles the decline in importance in the relationship between credit scores and indebtness. After 2007, the correlation with the LTV ratio continues to decline, while the correlation with the DTI ratio increases.
Especially when we explore the correlation with the DTI ratio, we yield some interesting results. The DTI ratio and the interest rate are positively correlated, in line with economic theory. The correlation with the interest rate was substantially lower between 2003 and 2007, reaching its low at 2005. While debt should have impact on determining the interest rate, and the interest rate should have impact in determining how much debt to take on, this relation seems to have become less important during the housing bubble. The DTI ratio and LTV ratio are positively correlated, as a higher LTV ratio directly influences the DTI ratio. The overall correlation between the DTI ratio and LTV ratio decreased.
The interest rate and LTV ratio are positively correlated, as relatively more debt is more risky. The correlation between the size of the mortgage and the interest rate is negative, which can be explained by the fact that borrowers that can obtain higher mortgages are more credit worthy and therefore less risky. The correlation changes over the years might seem quite surprising. While
-0.3 -0.2 -0.1 0 0.1 0.2 0.3 1999 2001 2003 2005 2007 2009 2011 2013 Mortgage LTV
Figure 8: These graphs depict the correlation between the DTI ratio and the size of the mortgage (blue line), the correlation between the DTI ratio and the LTV ratio (red line), and the correlation between the DTI ratio and the interest rate (green line)
Figure 9: These graphs depict the correlation between the interest rate and the size of the mortgage (blue line), and the correlation between the interest rate and the LTV ratio (red line)
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 1999 2001 2003 2005 2007 2009 2011 2013 Mortgage LTV Interest rate
15
one would expect a relatively low correlation between interest rates and the two variables before the financial crisis, the correlation decreased more since. It is however important to take into account the low interest rates due to cope with the financial crisis in the United States. This influences overall correlations between interest rates and other variables. Before the financial crisis, we do find a decrease in the correlation between interest rates and the LTV ratio. This resembles the relatively weaker relation between indebtness and interest rates.
16 4 Empirical strategy and results
This section discusses the results from the first regressions. The results are obtained from cross-sectional regressions placed in a timeline. The four variables used are the interest rate, the credit score, the LTV ratio and the DTI ratio. With a variety of regressions the relationships between these variables are explored.
4.1 The loan-to-value ratio
The second set of regressions is estimated using the LTV ratio as the dependent variable:
How much a borrower can borrow relative to the house price is expected to depend on the interest rate, the credit score and the indebtness. Therefore, the independent variables are all included in the regression to deal with omitted variable bias. All estimated results can be found in table 3 and are significant at a 1% significance level for the LTV ratio.
The interest rate has a positive effect on the LTV ratio, which is not completely in line with expectations. On the one hand, a higher interest rate increases the debt burden. However, a higher interest rate makes a borrower relatively more risky, which should restrict the amount someone would be able to borrow for a mortgage. On top of that, a higher interest rate makes it relatively less attractive to obtain a mortgage. This provides an incentive for borrowers to take on a smaller mortgage. The effect of the interest rate on the LTV ratio was not smaller during the bubble years, contrary to possible expectations. Since the start of the financial crisis, the influence of the interest rate has decreased. This might be due to the fact that the interest rates decreased as a result of monetary policy, while the LTV ratio remained quite stable. There is not a clear pattern for the effect of the interest rate on the LTV ratio.
The credit score has a negative effect on the LTV ratio, which is in line with economic theory. When a borrower is less creditworthy, the starting capital should be higher to deal with the risk properly. During the bubble years, the effect of the credit score on the LTV ratio has been declining. While one might expect a sudden increase at the start of the financial crisis, the effect became even smaller.
17 Table 3: This table depicts the regression results with the LTV ratio as the dependent variable and the credit score, interest rate and DTI ratio as explanatory variables between 1999 and 2013
The DTI ratio has a positive effect on the LTV ratio for all years, which might come as a surprise. It makes sense that the LTV ratio and the DTI ratio are positively correlated, because the mortgage debt contributes directly to the DTI ratio. However, it is strange that an increase in the DTI ratio increases the LTV ratio, because a borrower becomes relatively more risky when the DTI ratio is higher. This should decrease the eligibility for a high mortgage and increase the amount of down payment needed.
Dependent variable: LTV ratio
Year 1999 2000 2001 2002 2003 Credit score -0.056826*** -0.0510658*** -0.049395*** -0.0546193*** -0.0617817*** (-167.15) (-143.52) (-183.75) (-204.63) (-202.24) Interest rate 2.661776*** 2.667817*** 3.703166*** 3.517704*** 3.674005*** (79.99) (65.13) (104.8) (113.49) (93.54) DTI ratio 0.1314235*** 0.1577965*** 0.1295613*** 0.1488707*** 0.1260025*** (80.66) (86.84) (99.28) (117.99) (100.28) R2 0.0648 0.0648 0.0711 0.084 0.0751 AdjR2 0.0648 0.0648 0.0711 0.084 0.0751 N 709756 577664 946935 1028231 954089 ***=1%, **=5%, *=10% significance Year 2004 2005 2006 2007 2008 Credit score -0.0418528*** -0.0379431*** -0.0323953*** -0.0305583*** -0.020975*** (-137.63) (-127.6) (-105.62) (-96.74) (-54.63) Interest rate 4.194757*** 3.506034*** 4.126522*** 5.338425*** 5.043329*** (95.66) (77.76) (90.2) (125.21) (133.04) DTI ratio 0.0958602*** 0.1127434*** 0.1354999*** 0.1398327*** 0.1320767*** (71.87) (83.27) (94.21) (95.94) (87.78) R2 0.0468 0.0376 0.0401 0.0515 0.0492 AdjR2 0.0468 0.0376 0.0401 0.0515 0.0492 N 850180 950938 861944 832496 765395 ***=1%, **=5%, *=10% significance Year 2009 2010 2011 2012 2013 Credit score -0.0256803*** -0.0239158*** -0.0154047*** -0.0488978*** -0.023259*** (-54.98) (-45.52) (-24.94) (-81.6) (-38.42) Interest rate 4.164614*** 0.8480514*** 3.092898*** 2.723109*** 5.099663*** (87.56) (18.41) (58.93) (35.83) (95.82) DTI ratio 0.1286167*** 0.0567892*** 0.0344926*** 0.0442667*** 0.0300475*** (84.39) (28.43) (14.25) (19.65) (12.19) R2 0.0259 0.0073 0.0124 0.0189 0.0279 AdjR2 0.0259 0.0073 0.0124 0.0188 0.0279 N 896143 594661 429646 549343 468950 ***=1%, **=5%, *=10% significance
18 4.2 The credit score
Dependent variable: Credit score
Year 1999 2000 2001 2002 2003 DTI ratio -0.605816*** -0.6028211*** -0.6915596*** -0.752896*** -0.7819178*** (-107.39) (-89.23) (-137.49) (-160.23) (-187.18) R2 0.016 0.0136 0.0196 0.0244 0.0354 AdjR2 0.016 0.0136 0.0196 0.0244 0.0354 N 709756 577664 946935 1028231 954089 ***=1%, **=5%, *=10% significance Year 2004 2005 2006 2007 2008 DTI ratio -0.6068141*** -0.6232328*** -0.6576007*** -0.7124011*** -0.7436824*** (-127.43) (-132.73) (-129.49) (-139.55) (-163.05) R2 0.0187 0.0182 0.0191 0.0229 0.0336 AdjR2 0.0187 0.0182 0.0191 0.0229 0.0336 N 850180 950938 861944 832496 765395 ***=1%, **=5%, *=10% significance Year 2009 2010 2011 2012 2013 DTI ratio -0.6566584*** -0.6821173*** -0.6460362*** -0.6090822*** -0.6731761*** (-190.92) (-138.65) (-107.86) (-119.85) (-112.78) R2 0.0391 0.0313 0.0264 0.0255 0.0264 AdjR2 0.0391 0.0313 0.0264 0.0255 0.0264 N 896143 594661 429646 549343 468950 ***=1%, **=5%, *=10% significance
Table 4: This table depicts the regression results with the credit score as the dependent variable and the DTI ratio as the explanatory variable
The credit score is determined by different factors, such as credit history and current debt. It is therefore interesting to see whether the effect of the DTI ratio on the credit score changed over the years in the dataset.
Because the interest rate on the mortgage and the LTV ratio of the mortgage do not directly influence the credit score included in the data, they are excluded from the regression.
The DTI ratio has a negative effect on the credit score, which is consistent with economic theory. A higher level of indebtness decreases the creditworthiness of a borrower. The effect of the DTI on the credit score has not changed tremendously over the years. While there is a slight decline during the bubble years, this difference is not as large as one would expect. The start of the financial crisis did not cause large changes either.
19 Year 1999 2000 2001 2002 2003 2004 2005 2006 Credit score 1.02 1.04 1.06 1.07 1.06 1.04 1.05 1.05 Interest rate 1.02 1.03 1.05 1.06 1.02 1.02 1.03 1.03 DTI 1.02 1.01 1.02 1.03 1.04 1.02 1.02 1.02 Year 2007 2008 2009 2010 2011 2012 2013 Credit score 1.07 1.11 1.08 1.07 1.06 1.06 1.07 Interest rate 1.05 1.09 1.04 1.05 1.04 1.04 1.05 DTI 1.02 1.04 1.04 1.04 1.03 1.03 1.03
Table 5: This table depicts the VIF outcomes for the regression in table 3
This might partially have to do with the fact that the share of subprime borrowers decreased substantially in the dataset since the start of the financial crisis.
4.3 Collinearity
The variables in this dataset have significant effects on each other. The credit score and the DTI ratio are both explanatory variables in section 4.1, while the DTI ratio turns out to have a significant effect on the credit score in section 4.2. The significant effect found in section 4.2 could have biased the estimated results in section 4.1 due to collinearity. When there is collinearity, one explanatory variable can have an effect on the dependent variable through another variable, and therefore cause wrongly estimated results. Collinearity always exists to some extent, but it creates more damage as the degree of collinearity increases. The degree of collinearity is tested with the variance inflation factor (VIF). A VIF higher than 10 implies severe collinearity between variables. Table 5 shows the outcomes for the collinearity tests and does not indicate severe collinearity.
4.4 Summary 4.4.1 The bubble
Some of the results showed characteristics of the bubble build-up between 2002 and 2006. The regressions with the interest rate as the dependent variable showed the most prominent results. The effect of the credit score and the DTI ratio on the interest rate clearly declined, which resembles the fact that these factors had a smaller impact on the interest rate during the bubble years. This change in impact reflects an increasing risk appetite in the financial markets, as it became relatively cheap for less creditworthy borrowers to close a mortgage. In 2005, the equity withdrawals from mortgages peaked, with people living well beyond their means. This peak becomes visible in the data when we take a look at the low effect of the LTV ratio on the interest
20
rate. This is also the year that the effect of the DTI ratio on the interest rate became negative, which is a sign of the undervaluation of risk.
4.4.2 The financial crisis
This market behavior changed during 2007, when borrowers started to default and house prices started to decrease. As the panic started to spread through the markets, the financial markets became more risk averse. The data provides proof of the reaction to the credit crunch in 2008. Both the credit score and the LTV ratio had the largest impact on the interest rate that year, and the DTI ratio started to have a positive effect again.
21
5 Subprime borrowers
This chapter only focuses on subprime borrowers, defined as borrowers with a credit score below 640. First, the averages of the variables will be explored and compared with the averages of the total dataset discussed in the data description. Second, the same will be done for the correlations between the variables. Lastly, the regressions discussed in the results chapter will be repeated for subprime borrowers and compared. The purpose of this chapter is to emphasize the difference between the borrowing conditions for subprime borrowers and all borrowers. Subprime borrowers should have tighter restrictions and higher borrowing costs when foreclosing a mortgage. The share of subprime borrowers in the data declined after the financial crisis, and some of the results from the regressions become insignificant from 2009 on. Therefore, no conclusions will be drawn from the averages and correlations since the financial crisis, as the outcomes sometimes seem quite unlikely.
5.1 Averages
Figure 10 shows that the annual average interest rate on subprime mortgages is higher than the total average interest rate, although the difference is not that clear. The spread between the interest rates was relatively small until the start of the financial crisis. This might be a sign of risk being underpriced, as the interest rate on subprime mortgages should in theory be significantly higher due to the larger risk.
The share of subprime borrowers in the dataset has changed tremendously over the years. Before the financial crisis, the share of subprime borrowers was around 10% of all mortgages.
0 2 4 6 8 10 1999 2001 2003 2005 2007 2009 2011 2013 Subprime Total 595 600 605 610 615 620 625 630 635 1999 2001 2003 2005 2007 2009 2011 2013 Credit score
Figure 10: These graphs depict the average interest rate for subprime mortgages (blue line) and the total sample (red line) between 1999 and 2013
Figure 11: This graph depicts the average credit score for subprime mortgages in the sample between 1999 and 2013
22
Since the financial crisis, the share of subprime borrowers represents around 0.5% of the mortgages. Figure 11 shows how the average credit score for subprime borrowers increased since the financial crisis as well. 2008 represents the turning point once again.
Figure 12 shows that the average LTV ratio for subprime borrowers is higher than the average LTV ratio until 2008, which is quite unexpected. As subprime borrowers are relatively more risky, they should have a lower LTV ratio to compensate for the risk. The fact that subprime borrowers have a higher LTV ratio on average makes the borrowers even more risky.
Figure 13 shows that the average subprime DTI ratio is above the average DTI ratio for all years. A higher DTI ratio makes the subprime borrower more risky, although it is straightforward that subprime borrowers usually have a lower and more uncertain income
5.2 Correlations
While the correlation between the credit score and interest rate is negative for both the subprime data and the total data, the differences are substantial according to figure 14. The subprime correlation between the credit score and the interest rate is smaller, which means that the relation between the credit score and the interest rate is less clear for subprime mortgages. The correlation becomes smaller during the bubble years.
Figure 15 shows that the correlation between the credit score and house prices is positive for both subprime mortgages and the total mortgages, but much smaller for subprime mortgages. The correlation for subprime mortgages is relatively more constant over the years, and there is no sign of a less clear relation between the two variables during the bubble years.
0 20 40 60 80 100 1999 2001 2003 2005 2007 2009 2011 2013 Subprime Total 0 10 20 30 40 50 1999 2001 2003 2005 2007 2009 2011 2013 Subprime Total
Figure 12: These graphs depict the average LTV ratio for subprime mortgages (blue line) and the total sample (red line) between 1999 and 2013
Figure 13: These graphs depict the average DTI ratio for subprime mortgages (blue line) and the total sample (red line) between 1999 and 2013
23
Figure 16 shows that the correlation between the credit score and the LTV ratio becomes positive for subprime mortgages, while the correlation is negative for the total dataset. For subprime borrowers, an improvement in the credit score could mean that a borrower would be able to take on a higher mortgage relative to the price of the house. During the bubble years, the correlation became even larger. This means that relatively more creditworthy borrowers could more easily obtain higher LTV ratios and thus take on more risk, since it is unlikely that a higher LTV ratio has a positive effect on the credit score.
Figure 17 shows that the correlation between the credit score and the DTI ratio also becomes positive for subprime mortgages, while the correlation for the total dataset is negative. On the one hand, a higher DTI ratio should deteriorate the credit score.
-0.3 -0.25 -0.2 -0.15 -0.1 -0.05 0 0.05 1999 2001 2003 2005 2007 2009 2011 2013 Subprime Total -0.02 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 1999 2001 2003 2005 2007 2009 2011 2013 Subprime Total -0.3 -0.25 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 1999 2001 2003 2005 2007 2009 2011 2013 Subprime Total Figure 14: These graphs depict the correlation between the credit score and the interest rate for subprime mortgages (blue line) and the total sample (red line) between 1999 and 2013
Figure 15: These graphs depict the correlation between the credit score and house price for subprime mortgages (blue line) and the total sample (red line) between 1999 and 2013
Figure 16: These graphs depict the correlation between the credit score and LTV ratio for subprime mortgages (blue line) and the total sample (red line) between 1999 and 2013
24
On the other hand, an improvement in the credit score for subprime borrowers increases the possibility to take on more debt relative to income. The effect of the credit score on the DTI ratio seems to be dominating the correlation. The correlation becomes larger during the bubble years, which means that an increase in the credit score had a larger positive impact on the DTI ratio. It cannot be the case that an increase in the DTI ratio had a smaller impact on the credit score, as the rules for the credit score determination did not change over the period.
Figure 18 shows that the correlation between the DTI ratio and the size of the mortgage is positive for the subprime borrowers as well, which is in line with expectations. The correlation is larger for subprime borrowers, because these borrowers usually have a smaller income and cannot easily obtain other types of debt. A mortgage therefore has a relatively larger impact on the DTI ratio. On top of that, a higher DTI ratio could have a larger effect on the size of the mortgage that can be taken on for subprime borrowers. During the bubble years, the correlation increased relatively more than for the total dataset.
Figure 19 shows that the correlation between the DTI ratio and the LTV ratio is lower for subprime mortgages than for the total dataset. Both correlations are positive, as we will not interpret the correlations for subprime mortgages after 2008. Apparently the DTI ratio has a smaller impact on the LTV ratio for subprime mortgages, since it is unlikely that the LTV ratio has a smaller impact on the DTI ratio for subprime mortgages
-0.25 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 1999 2001 2003 2005 2007 2009 2011 2013 Subprime Total 0 0.05 0.1 0.15 0.2 0.25 1999 2001 2003 2005 2007 2009 2011 2013 Subprime Total
Figure 17: These graphs depict the correlation between the credit score and the DTI ratio for subprime mortgages (blue line) and the total sample (red line) between 1999 and 2013
Figure 18: These graphs depict the correlation between the DTI ratio and the size of the mortgage for subprime mortgages (blue line) and the total sample (red line) between 1999 and 2013
25
Figure 20 shows that the correlation between the DTI ratio and the interest rate fluctuates heavily for subprime mortgages as well. While the correlation is positive in the first couple of years, the correlation drops enormously between 2002 and 2004. Between 2004 and 2007, the correlation even becomes negative. On the one hand, this might seem absurd as a higher DTI ratio implies more risk and borrowers with higher DTI ratios should therefore be charged with higher interest rates. On the other hand, a lower interest rate could have a positive effect on the amount of debt a borrower is willing to take on. It seems like the interest rate effect on the DTI ratio is dominating the correlation. This is a sign of underpriced risk in the financial markets, because the effect of the DTI ratio on the interest rate should be substantial, especially for subprime borrowers.
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 1999 2001 2003 2005 2007 2009 2011 2013 Subprime Total
Figure 21: These graphs depict the correlation between the interest rate and the LTV ratio for subprime mortgages (blue line) and the total sample (red line) between 1999 and 2013
-0.1 -0.05 0 0.05 0.1 0.15 0.2 1999 2001 2003 2005 2007 2009 2011 2013 Subprime Total -0.04 -0.02 0 0.02 0.04 0.06 0.08 0.1 0.12 1999 2001 2003 2005 2007 2009 2011 2013 Subprime Total
Figure 19: These graphs depict the correlation between the DTI ratio and the LTV ratio for subprime mortgages (blue line) and the total sample (red line) between 1999 and 2013
Figure 20: These graphs depict the correlation between the DTI ratio and the interest rate for subprime mortgages (blue line) and the total sample (red line) between 1999 and 2013
26
Figure 21 shows that the correlation between the interest rate and the LTV ratio is positive for subprime mortgages and the total dataset, but fluctuates more for the subprime data. The correlation decreases more for the subprime data and becomes lower than the total correlation during the bubble years. Apparently, the interest rate is less determined by the LTV ratio for subprime mortgages, and the influence becomes even smaller during the bubble years. As the bubble burst sooner for subprime mortgages than for prime mortgages, the correlation increases sooner and more prominent. The sudden increase reflects the increase in the awareness of the risk associated with subprime borrowers.
5.3 Regressions
5.3.1 The loan-to-value ratio
The LTV ratio is the dependent variable in the first set of subprime regressions. Table 6 shows that the effect of the credit score on the LTV ratio for subprime mortgages is significantly positive for all years, while the effect was significantly negative for the total dataset. Subprime borrowers can take on more risk when there is an improvement in the credit score by obtaining a higher LTV ratio. The relation between the LTV ratio and credit score is here driven by different factors. For the total dataset, borrowers with higher credit scores usually do not need such a high mortgage compared to the value of the house, as they can afford a higher down payment. For subprime borrowers, the possibility to take on a higher mortgage is often restricted because of the credit score. This might explain the difference in results.
The effect of the interest rate on the LTV ratio is significantly positive for subprime mortgages as well as for the total dataset. From 2003 until 2005 the effect was lower for subprime mortgages, but after the effect started to increase until it doubled in 2007 and remained that high. It might be the case during the booming years that a higher interest rate limited the possibility to borrow more, as a higher interest rate increases the riskiness of the loan. It is interesting to see however that the effect of the interest rate on the LTV ratio increased since the start of the financial crisis. This result might be affected by the fact that interest rates were lowered due to monetary easing, while LTV ratios became lower for subprime borrowers.
27 Table 6: This table depicts the regression results for subprime mortgages with the LTV ratio as the dependent variable and the credit score, interest rate and DTI ratio as explanatory variables between 1999 and 2013
The effect of the DTI ratio on the LTV ratio is statistically significantly positive for subprime mortgages and the total dataset, although the effect is clearly smaller for all years. While it came as a surprise that the DTI ratio had a positive influence on the LTV ratio for the total dataset, it makes sense that the effect is smaller for subprime mortgages. As these borrowers are already relatively risky and a higher DTI ratio increases the riskiness of the borrower, the positive effect of the DTI ratio on the LTV ratio should be smaller.
5.3.2 The credit score
Table 7 shows that the effect of the DTI ratio on the credit score is statistically significantly positive for subprime mortgages for all years except 2001. This result is contrary to the negative effect of the DTI ratio on the credit score for the total dataset. A higher DTI ratio automatically deteriorates the credit score, as it is one of the determinants of the credit score. The effect is however very small compared to the results from the total dataset.
Dependent variable: LTV ratio
Year 1999 2000 2001 2002 2003 2004 Credit score 0.0283266*** 0.0416153*** 0.03886*** 0.0392632*** 0.0187691*** 0.0457158*** (14.26) (21.65) (25.64) (26.61) (8) (19.16) Interest rate 1.980895*** 2.656517*** 3.874868*** 3.59454*** 3.397628*** 3.417571*** (22.02) (28.48) (52.07) (50.21) (27.98) (26.23) DTI 0.0495649*** 0.0893366*** 0.0731243*** 0.0739556*** 0.0683284*** 0.035791*** (10.13) (16.71) (19.12) (19.65) (13.94) (7.45) R2 0.0112 0.0238 0.0358 0.0317 0.016 0.0138 AdjR2 0.0112 0.0237 0.0357 0.0317 0.0159 0.0138 N 70465 65033 99081 107723 64448 73994 ***=1%, **=5%, *=10% significance Year 2005 2006 2007 2008 2009 2010 Credit score 0.0615196*** 0.0922065*** 0.1310022*** 0.1212361*** 0.0593678*** 0.0461431 (23.48) (33.24) (49.27) (30.42) (3.81) (1.51) Interest rate 2.37678*** 4.707437*** 5.483676*** 9.014622*** 7.869451*** 8.684734*** (17.26) (33.53) (44.55) (59.28) (21.53) (16.26) DTI 0.0377226*** 0.0756718*** 0.067044*** 0.0497457*** 0.0654765*** 0.0556587** (7.68) (14.96) (13.81) (6.96) (3.71) (2.16) R2 0.0109 0.0287 0.0475 0.1165 0.0892 0.0763 AdjR2 0.0108 0.0286 0.0474 0.1164 0.0887 0.0755 N 76718 76942 81581 32071 4987 3307 ***=1%, **=5%, *=10% significance
28 Table 7: This table depicts the regression results for subprime mortgages with the credit score as the dependent variable and the DTI ratio as the explanatory variable between 1999 and 2013
Year 1999 2000 2001 2002 2003 2004 2005 2006 Credit score 1 1 1.01 1.01 1.01 1.01 1.01 1.02 Interest rate 1.01 1.01 1.01 1.02 1.01 1.01 1.01 1.02 DTI 1 1 1 1.01 1 1 1 1 Year 2007 2008 2009 2010 2011 2012 2013 Credit score 1.03 1.01 1.01 1 1 1 1 Interest rate 1.03 1.01 1.01 1 1 1 1 DTI 1 1 1 1 1 1 1
Table 8: This table depicts the VIF outcomes for the regression in table 6
The logical explanation for a positive result stems from the fact that a higher credit score can cause higher DTI ratios, and that this effect dominates the regression the other way around.
5.3.3 Collinearity
Because the DTI ratio has a significant effect on the credit score for some years in the subprime mortgage dataset, it is important to test for collinearity again. Table 8 shows that the VIF outcomes are far below 10 for the subprime mortgage regressions as well, and that there is therefore no sign of severe collinearity.
5.4 Summary
The subprime mortgage data differs substantially from the total data used in this thesis. With respect to the averages, it is surprising to see that the average interest rate is not significantly higher for subprime mortgages, while the DTI ratio and the LTV ratio are. Since subprime
Dependent variable: Credit score
Year 1999 2000 2001 2002 2003 2004 DTI ratio 0.0797066*** 0.023534** -0.0096531 0.0492836*** 0.0412168*** 0.0209927*** (8.6) (2.16) (-1.2) (6.32) (5) (2.83) R2 0.001 0.0001 0 0.0004 0.0004 0.0001 AdjR2 0.001 0.0001 0 0.0004 0.0004 0.0001 N 70465 65033 99081 107723 64448 73994 ***=1%, **=5%, *=10% significance Year 2005 2006 2007 2008 2009 2010 DTI ratio 0.0644052*** 0.1251606*** 0.1060905*** 0.1036128*** 0.03694** 0.0054709 (9.47) (18.93) (16.4) (10.32) (2.3) (0.37) R2 0.0012 0.0046 0.0033 0.0033 0.0011 0 AdjR2 0.0012 0.0046 0.0033 0.0033 0.0009 -0.0003 N 76718 76942 81581 32071 4987 3307 ***=1%, **=5%, *=10% significance
29
borrowers should already be associated with more risk, it is quite surprising that these borrowers have higher DTI ratios and LTV ratios on average, and are not charged with substantially higher interest rates.
The most striking findings with respect to the correlations are the correlations between the credit score and the interest rate, the credit score and the LTV ratio, the credit score and the DTI ratio, and the DTI ratio and the interest rate. The differences in signs of the correlation that occur with the correlation between the credit score and the LTV ratio and the credit score and the DTI ratio show the differences in behavior for subprime mortgages. These signs are positive instead of negative because of a decrease in riskiness when the credit score increases, and therefore represent an increase in the eligibility of mortgage conditions.
The effect of the credit score on the LTV ratio becomes positive, and the effects of the interest rate and DTI ratio become smaller for subprime mortgages. These results show that eligibility plays a large role in the estimates for subprime mortgages, since the positive credit score effect on the LTV ratio stems from the possibilities created by a higher credit score. The effects from the other variables on the LTV ratio are smaller for subprime mortgages, because subprime borrowers are already relatively more risky. The effect of the DTI ratio on the credit score also seems to be dominated by eligibility, since a higher DTI score should always have a negative effect on the credit score.
30
6 Conclusion
This thesis explored the borrowing conditions for 11.5 million mortgages sold to and provided by Freddie Mac between 1999 and 2013. First, we focused on the differences in borrowing conditions over the years and especially around the start of the financial crisis. Second, we explored differences between the total dataset and subprime borrowers. This meant to gain insight in the bubble years before the financial crisis. The research question this thesis tried to answer is:
“To what extent did borrowing conditions change between 1999 and 2013 for mortgages sold to and provided by Freddie Mac?”
The mortgages in the Freddie Mac Single Family Loan-Level Dataset changed tremendously between 1999 and 2013, and thereby match the literature to some extent. The total data shows clear signs of deterioration in mortgage eligibility conditions during the bubble years. The effects of the explanatory variables on the interest rate declined during these years. During the start of the financial crisis, mortgage eligibility conditions tightened substantially. The financial crisis caused changes in averages, correlations and regression results. The share of subprime mortgages decreased tremendously during this period. After the start of the financial crisis, the mortgages that were issued and purchased by Freddie Mac were better than before the financial crisis. This shows the tighter regulation associated with mortgages.
The subprime mortgage chapter especially provides insight in the bubble years and the deterioration in mortgage eligibility conditions. The results that were found in that chapter during those years can be related to the political pressure that was associated with these mortgages and the need for an increase in mortgage supply. Besides the changes in borrowing conditions over the years for these mortgages, it is also interesting to explore the differences between regular mortgages and subprime mortgages. The differences in borrowing conditions between prime borrowers and subprime borrowers does not become clear when we look at the interest rate, DTI ratio and LTV ratio averages over the years, which is quite unexpected. While it makes sense that the correlation between the credit score and the DTI ratio and the correlation between the credit score and the LTV ratio is negative for regular mortgages, it also makes sense that this correlation is positive for subprime mortgages. The regression results show smaller effects from the DTI ratio and the interest rate on the LTV ratio, which can be attributed to the overall higher riskiness
31
associated with subprime borrowers. The different effect of the credit score on the LTV ratio for subprime mortgages clearly shows the difference in the relation between variables for regular mortgages and subprime mortgages. Since the share of subprime mortgages declined since 2008, this thesis cannot provide a clear overview of the differences between subprime mortgages before and since the financial crisis.
Although the exact role that political pressure and the distribution of MBS played remains unknown, Freddie Mac contributed to the housing bubble in the United States with the help of government guarantees.
For further research it might be interesting to divide the mortgages in more different groups of borrowers, instead of just subprime borrowers and the total dataset. This could provide more insight in the risks associated with different credit scores. It would also be possible to construct monthly data and to explore the relation between the share of subprime borrowers in the data and the value of the stock. Since the government implicitly backed Freddie Mac, the correlation between these two variables should be lower than for private mortgage issuers. Since not a lot of research has been done solely for Freddie Mac, there are still a lot of features associated with government guarantees that could be explored.
32
References
Acharya, V., Richardson, M., Van Nieuwerburgh, S., White, L.J., (2011). Guaranteed to Fail. Fannie, Freddie, and the Debacle of Mortgage Finance. NYU Stern.
Adrian, T.,& Shin, H.S. (2009). The Shadow Banking System: Implications for Financial Regulation. Federal Reserve Bank of New York. Staff Report No. 382.
Chomsisengphet, S., & Pennington-Cross, A., (2006). The Evolution of the Subprime Mortgage Market. Federal Reserve Bank of St. Louis. Review January/February 2006, 88(1), pp. 31-56.
Frame, W.S. (2009). The 2008 Federal Intervention to Stabilize Fannie Mae and Freddie Mac.
Federal Reserve Bank of Atlanta. Working Paper No. 2009-13.
Freddie Mac (2015). Single Family Loan-Level Dataset General User Guide.
Krugman, P (2013). End This Depression Now! New York, NY: W.W. Norton & Company, Inc.
Mayer, C., Pence, K., & Sherlund, S.M., (2008). The Rise in Mortgage Defaults. The Journal of Economic Perspective. Winter 2009, 23(1), pp.27-50.
Roubini, N., (2010). EconoMonitor Flashback: Roubini’s IMF Speech – September 7, 2006.
Roubini, N., & Mihm, S., (2010). Crisis Economics: A Crash Course in the Future of Finance. New York, NY: The Penguin Press.
Thompson, H (2009). The Political Origins of the Financial Crisis: The Domestic and
International Politics of Fannie Mae and Freddie Mac. The Political Quarterly. January-March 2009, 80(1), pp. 17-24.
Wallison, P.J., & Calomiris, C.W., (2008). The Last Trillion-Dollar Commitment: The Destruction of Fannie Mae and Freddie Mac. American Enterprise Institute for Public
