Cross check of the S&Ls and subprime crises potential causes

3While the causes of the subprime crisis were already the subject of many debates both within the public opinion and the academic circles [1], it seems interesting to remind some facts concerning the origins of the S&Ls crisis and its development.

4Following the Great Depression of the 1930s, an important number of banking laws and regulations were introduced in the US in order to provide financial markets with liquidity, safety, efficiency, public convenience and stability (Hamlin and Hillyard, 1991). Within this new regulatory framework one could mention the Federal Home Loan Bank Board (FHLBB) created by the Congress in 1932 in order to sustain financial markets, to control thrifts institutions and to allow them to get to the secondary market for mortgages (Marcis, 1974; Hamlin and Hillyard, 1991; McCool, 2005), the Glass-Steagall Act of 1933 or the Federal Deposit and Insurance Company (FDIC) in 1933 with the main objective of providing liquidity by insuring commercial bank deposits. While until 1934 S&Ls managed to remain safe and to insure the necessary liquidity to the markets via their activity of receiving deposits and offering mortgages, i.e. a low risk activity after all, the creation of the Federal Savings and Loans Insurance Corporation (FSLIC) in 1934 changed the picture. Reinforced by the insurance of their deposits, S&Ls started to bear more risks. Meanwhile, they were very sensitive to interest rate fluctuations due to a maturity mismatch in their assets and liabilities; hence, an interest rate increase would dramatically impact their transformation risk. It was exactly what happened in 1965-1966 and 1969-1970 when S&Ls institutions faced a disintermediation crisis. Both periods were characterized by very high interest rates rendering thrifts less competitive. The economic shocks of the 1970s resulting in a huge inflation, high consumer indebtness ratios and important budget deficits, coupled to the fact that deposit rates were ceiling on market rates [2] and the introduction of the Bankruptcy Reform Act in 1978 [3] definitely eroded S&Ls competitiveness. This gap between market conditions and S&Ls capacities generated huge losses for S&Ls. 75 % of them went bankrupt as in addition to the fact that investors took out their money from thrifts, S&Ls had to face constant interest revenues but increasing interest costs (Caudill, Caudill and Gropper, 2001). Hence, between 1980 and 1982, the Congress adopted a set of deregulation laws with the aim of providing thrift institutions with more freedom and more autonomy and rendering them competitive again.

5Two major acts, i.e. the Depository Institutions Deregulation and Monetary Control Act (DIDMCA) in 1980 and the Garn-St Germain Depository Institutions Act (GSGDIA), signed in 1982, were aimed to fulfilling three objectives. First of all, thrifts were allowed to trespass interest rate ceilings on deposits (Hamlin and Hillyard, 1991; White, 1993; Walker, 1994; Reinstein and Steih, 1995) in order to remain competitive and attractive compared to commercial banks. Then, requirements became less demanding in order to avoid S&Ls institutions to be constrained with strict limits (Reinstein and Steih, 1995) and finally, the insurance level on S&Ls investments was extended (White, 1993; Walker, 1994; Reinstein and Steih, 1995). Both acts created new opportunities for thrift institutions which could grow rapidly but also engage in riskier activities led by moral hazard incentives. Therefore, new management and organizational problems appeared (White, 1993). By the late 1984, the government intervened in order to regulate the market by monitoring the development of inadequately capitalized thrifts, reducing direct equity investments and increasing requirements for professionals and particularly for accountants. Unfortunately, it was already too late (White, 1993).

6In 1986, the thrift crisis broke up with a decline of conservatism (Hamlin and Hillyard, 1991). One example is the Tax Reform Package which aimed at eliminating tax deduction for mortgage interests on investment properties and for losses on “passive” income (Hamlin and Hillyard, 1991; White, 1991). However, this act also had consequences on the evolution of the real estate investments and finally on the fluctuation of the real estate value; it contributed to its decline which generated losses for S&Ls, banks and insurance companies. As a consequence, between 1982 and 1987, the real estate value collapsed, with decreasing market values from 30 % to 70 % (Reinstein and Steih, 1995).

7The thrifts’ crisis contributed to a tremendous increase in bailout costs that climbed from $15 billion to $500 billion (Reinstein and Steih, 1995). In 1988, the FDIC experienced huge losses as insolvent debt started to represent a significant part of banks’ liabilities (Elliott, 1991). The Financial Institutions Regulatory Reform and Enforcement Act (FIRREA) of 1989 aiming at a new regulation of the thrift industry and at bailing out distressed institutions namely by attracting private investors willing to buy back loans purchased by the government, failed (Reinstein and Steih, 1995). The 1990 elections brought even more concerns as thrifts represented the main fund raiser of the campaign. In addition, as pointed out by Reinstein and Steih (1995), over the same time period, a lot of professionals, namely lawyers, were charged for malpractice in their relationships with banks and S&Ls. All in all, during the 1990s, the country accumulated $200 billion of losses (Reinstein and Steih, 1995).

8Hence, we can synthesize the S&Ls crisis in three major points that will allow us to design our comparison with the subprime crisis. For a brief summary, please refer to the Appendix.

9First of all, one should mention the impact of deregulation, the increase of the insurance levels and the decrease in interest rate to borrowers due to the high level of competition generated by S&Ls assets diversification practices. As previously seen, S&Ls were government-sponsored institutions, therefore strongly influenced by its different regulations on ceiling rates and tax reforms among others. Professional standards have also contributed to fuel the crisis. S&Ls problems appeared a few years after the introduction of the GAAP standards in the 1970s that allowed thrifts to recording in their balance sheets the book value of their assets instead of the market value. Finally, the increase in the insurance level played an important role in S&Ls’ risk exposure (Lee and Lynge, 1985; Elliott, 1991; Walker, 1994; Reinstein and Steih, 1995; Caudill, Caudill and Gropper, 2001). By this reform, the government took on the major part of the risk of financial institutions. Therefore, thrifts were managed to adopt risk-taking behaviours as they were insured by the government to a higher level than before the deregulation. Consequently, the financial system lost its stability (Kane, 1989; Thomson, 1986, 1989). Moreover, the fact that financial institutions operated in a flat-rate insurance system emphasized this risk oriented behaviour.

10Government actions can also be one explanation of the subprime crisis, almost starting with the floating FX regime installed by R. Nixon in 1971 and, more recently, the American economic policy aimed at maintaining US growth. Following the burst of the Internet bubble in 2000, the Federal Reserve engaged in a policy of short-term interest rate cut in order to sustain consumption and avoid the recession. This was particularly true after September 11^th, 2001. In addition, the Economic Growth and Tax Relief Reconciliation Act was signed on November 26^th, 2001 by the House of Representatives and the Senate. It implied, among others, a tax cut for individuals who get retired or contribute to a pension plan. The combination of these two policy measures contributed to a liquidity boom and led to huge investments, mainly made in the housing market with a deep restructuring of the mortgage market (Tully, 2007). The US government has been heavily criticized for the underestimation of the consequences of a decrease in real estate prices on the consumption hence leading to low interest rates. [4] Mortgages became cheap and very attractive for everybody, even for borrowers who did not meet bank requirements and had a poor credit history. Moreover, underwriting standards were lax too between 2004 and 2006 as indicated by an Office of the Comptroller of the Currency (OCC) survey on 78 of the largest American banks regarding their assessment of the conditions under which credit was extended; similarly, the percentage of loans with full documentation was in decline during the same time. This situation deeply affected the assessment of credit risk and, at the end, investors’ decisions (Tully, 2007; Crouhy, Jarrow and Turnbull, 2008; Thompson, 2008; Okongwu and Sabry, 2009). Finally, the implication of the government sponsored agencies (Fannie Mae and Freddie Mac) in the significant growth experienced by the mortgage market and mortgage securitization practices is already well known.

11In addition, the literature also agrees on the fact that the government interventions to avoid or limit the crash were each time implemented far too late. Both the safety-and-soundness regulation by the late 1984 and the government money injections of 2007 arrived once the market was already in distress. In addition, the Rule 157 of the Financial Accounting Standards Board effective by November 2007 imposed a stricter constraint on the level of bank reserves. It was followed by a stock market crash whereas default and bankruptcy rates increased. On the real estate market, the new trend for homeowners was to sell their houses as fast as they could (Tully, 2007).

12Second, in a context of increased flexibility that characterised both the 1980s and the beginning of the 21^st century, the mortgage market became more attractive for consumers and financial institutions but also more sensitive to economic fluctuations, namely interest rate policies. As mentioned previously, at the beginning of the 1980s, S&Ls experienced competitiveness problems as during periods of tight money interest rates were increasing while thrifts were locked in by rate ceilings. With less deposit inflows, they were therefore obliged to reduce their credit volume for mortgages. This in turn had a huge impact on the residential construction activity which is characterized by a “stop-and-go nature” (Marcis, 1974) and was followed by many S&Ls mergers and failures in the 1980s, dramatically decreasing the number of thrifts on the market (Reichert, 1991; Brewer III, 1995). The deregulation policy implemented between 1980 and 1982, was supposed to bring a solution to thrifts’ losses via the increase in interest rates caps. However, it made the situation even worse. S&Ls, with their huge losses, used this opportunity to diversify their financial activities and invest in riskier yield bonds.

13Between 2000 and 2006, the American real estate market prices experienced a huge increase and homeowners’ indebtness ratios exploded as credit was cheap and available to a broad range of debtors. This context created a strong reliance of borrowers on real estate prices and gave an illusion of low credit risk as there were a lot of potential opportunities for credit refinancing and property reselling. Therefore, when real estate prices started to decline while interest rates rose, default and foreclosure rates rapidly climbed (Okongwu and Sabry, 2009).

14Third, the broadening of financial activities and the financial innovation, made both S&Ls in the 1980s and banks over the recent years to engage in riskier and more opaque investments. The deregulation implemented at the beginning of the 1980s, allowed S&Ls to tremendously expand their activities. In this new context, they could provide, besides mortgage loans, commercial and consumer loans as they were also able to invest in junk bonds, classified as commercial loans. They could therefore diversify their portfolios. However, S&Ls were not experts in all these markets on which they had to deal with new objectives of profitability as the number of competitors increased (Hamlin and Hillyard, 1991). To achieve those new profitability objectives, S&Ls increased their investments in higher yield bonds, hence bearing higher risks, in a dangerous run for yield enhancement (Elliott, 1991; Brewer III, 1995; Reinstein and Steih, 1995; Cebenoyan, Cooperman and Register, 1995; Park and Peristiani, 1998). Moreover, S&Ls and other financial intermediaries started to develop innovative financial products like securitization (Reichert, 1991). One of the first examples was given by Citicorp in 1989 who issued a large volume of Mortgage Backed Securities in the secondary market (Horner, 1990). These new investment practices coupled to the possibility of using market rates as ceilings were abolished made S&Ls follow a policy of capital positions improvement through better margins, higher profits and increasing retained earnings, hence, higher risks. This was particularly true for thrifts with weak capital positions. Myers (1977), Kane (1989), Benston and Koehn (1989), Brewer III and Mondschean (1994), Brewer III (1995) among others show that the main part of the thrifts industry which failed after the deregulation was composed by thinly capitalized S&Ls that had nothing to loose as their financial situation was already damaged; if they could earn a higher profit, allowing them to benefit from a small change in their asset base which led to a large change in their net worth value, they would take this opportunity without too much care for risk. Thus, everything was put together to push them to adopt a moral hazard behaviour and to make risky investments which in turn increased the cost of capital and impacted the whole thrifts industry. Finally, as underlined by different authors (Ross, 1989; Pound and Zeckhauser, 1990; Reichert, 1991 or Cebenoyan, Cooperman and Register, 1995), depository institutions activity is rather opaque. First of all, when they make a decision, this decision remains very confidential and is not broadcasted to investors. Moreover, it was rather difficult for the investors to fully assimilate and understand all the new accounting principles, terms and practices used by S&Ls. Therefore, it was not easy for them to make the optimal investment decision (White, 1993).
Over the period preceding the subprime crisis financial markets were booming: more products, more investors and more trading opportunities. Okongwu and Sabry (2009) point out the increasing demand for innovative products, much riskier and with less and less fit with the characteristics of a traditional mortgage, i.e. interest-only, negative amortization loans, etc. Investors relied more and more on short-term funding. As market participants started to be more numerous, there was an increase in the systemic risk linked to the real estate value. All financial products had a link with the housing market hence, deeply impacted if the trend in the house’s price had reversed. The run for yield enhancement and wealth maximization was launched. In an environment of low interest rates, investors had plenty of opportunities to improve returns and, in their search for high yields, they increased risk significantly. A new model of credit intermediation was born: the original-to-distribute model. [5] In this context, borrowing cheap money became a real fashion and led to a leveraged-buyout boom. With the development of very sophisticated financial products that allowed increasing the quality of securities above the level fixed by the originator of the debt or asset via a securitization deal, risk management became the key issue and the crisis burst because financial institutions did not perform it correctly. As an example, instead of using the collateralized debt obligations to hedge credit risk, which was their original purpose, they were used to sell to different investors many derivative claims on the same asset. The systemic risk increased dramatically and created a litigation risk among all the parties of interest in the same pool. This was combined with a higher interest in short-term benefits compared to long term consequences (Crouhy, Jarrow and Turnbull, 2008; Thompson, 2008). Securitization, associated with the “originate-to-distribute” model created an agency problem: the responsibility of mortgage default was not beard by mortgage originators anymore and so, they were able to invest in riskier assets without any short term visible consequences (Crouhy, Jarrow and Turnbull, 2008). The inability to implement accurate valuation models led to risk mispricing. One major player on the market was represented by the rating agencies as main assessors of the debtors’ credit risk quality. Rating agencies had difficulties in updating the data they were using in their reports and took a long time to understand the consequences of a decline in the subprime market on the ratings of a monocline; when a monocline is downgraded, all the papers it had insured are also downgraded. This opaque process combined with a lack of reports on the implicit commitments to the market fund by banks, contributed to investors’ inability to estimate prices, to set valuation models and perform credit risk analysis. It finally affected their confidence (Tully, 2007; Crouhy, Jarrow and Turnbull, 2008; Thompson, 2008); investors started to require higher compensation as they felt bearing more risks. This compensation included higher required returns on stocks as well as higher interest payments on debt. Stocks became over priced; as always, when prices are very far from the average value on the long term, their decline has to be very significant. Hence, one cause of the subprime crisis burst can be found here (Tully, 2007). Finally, in presence of very sophisticated financial products investors became unable to clearly understand the characteristics of the assets they were manipulating and gathering information on some of those assets was very complicated. This led to confusion for all those who were supposed to make asset buying or selling decisions (Thompson, 2008).
The objective of our paper is to propose a comparative analysis of the financial mechanisms that governed the subprime and the S&Ls crises in order to stress their similarities and the differences in their consequences. By using an event study analysis on banks returns at the beginning of the 21^st century, i.e. when the main causes of the subprime crisis can be found, we will then try to identify some major macroeconomic and external events that may have impacted banks behaviour and hence, contribute to the beginning of the financial bubble. Our methodology goes in line with Cornett and Tehranian (1990), Fraser and Kolari (1990) and Graddy et al. (1994) on the S&Ls and banks returns in the 1980s. As such, we are able to compare the two crises and shed some light on the difference between their consequences.
Contrary to previous evidence showed by the above quoted studies, our results point out a certain reaction of bank returns to macroeconomic announcements, more specifically FED disclosures concerning the level of interest rates. As the results are statistically significant only for banks less involved in the securitization practices / less affected by the financial crisis ex-post, we may conclude that securitization plays a major role over the time period preceding the subprime crisis and helps differentiating the way the financial market evaluates the banks in our sample. It is one main difference between the two crises and may explain the impressive spread off of the contemporaneous crisis. Moreover, the impact on the returns is rather mitigated in terms of sign, which does not allow a clear cut conclusion and underlines the difficulty of assessing the effects of interest rate policies on bank returns when the regulatory framework is complex and financial activities are booming.
This paper is organized as follows. The next section describes our methodology, followed by the presentation of our data. In section three, we provide our results and analysis while the last section concludes.

Methodology

16The objective of our study is to assess the importance of government intervention, as one of the major common causes for both crises, on banks returns and, in a more general view, on investors’ behaviour. To do so, we follow the methodology proposed by Graddy et al. (1994) who in turn were inspired by Cornett and Tehranian (1990) and Fraser and Kolari (1990). In their paper, Graddy et al. (1994) analyse the effects of the implementation of the Garn-St Germain Depository Institutions Act on Savings and Loans institutions between August 11, 1981 and February 24, 1983. They collect the returns of 19 thrifts over the sample period along with 16 events, from the formulation of the Garn-St Germain Depository Institutions Act of 1982 until the Post-enactment of the Act, including the debate surrounding this reform. Their objective is to study the abnormal returns generated by these different announcements in order to understand the consequences of the Act on the behaviour of the S&Ls in their sample. To do so, Graddy et al. (1994) compare two portfolios, one composed by big size banks while the second one includes the S&Ls. They manage to detect a single significant, negative effect: it concerns only big banks returns on the date when the Senate Banking Committee approved the bill. Their explanation stresses the fact that the Garn-St Germain Act was perceived as eroding banks competitiveness. [6]

17We are going to conduct the same type of study for the subprime crisis following a twofold objective. First of all, we want to check whether government actions occurring at the beginning of the subprime crisis had an impact on banks returns. Second, our aim is to compare the obtained results with those of the above quoted study, in order to shed more light on the similarities and differences between the two crises.
We have chosen to study the period following the internet bubble burst as several events occurring over this time frame create the roots of the subprime crisis. We have selected 19 events, occurring between October 6, 2000 and April 19, 2002 corresponding to a time period of 80 weeks. They are all summarized in Table 1.

Table 1

Selected events

1 November 7, 2000 Announcement of the results of American presidential elections 2 November 15, 2000 The Federal Open Market Committee decided to maintain its target for the federal funds rate at 6.5% 3 December 6, 2000 Governor Edward M. Gramlich speech at the Community and Consumer Affairs Department Conference on Predatory Lending, Philadelphia, Pennsylvania 4 December 19, 2000 The Federal Open Market Committee decided to maintain its target for the federal funds rate at 6.5% 5 January 3, 2001 The Federal Open Market Committee decided to lower its target for the federal funds rate by 50 basis points to 6% 6 January 31, 2001 The Federal Open Market Committee decided to lower its target for the federal funds rate by 50 basis points to 5.5% 7 March 21, 2001 The Federal Open Market Committee decided to lower its target for the federal funds rate by 50 basis points to 5% 8 April 18, 2001 The Federal Open Market Committee decided to lower its target for the federal funds rate by 50 basis points to 4.5% 9 May 15, 2001 The Federal Open Market Committee decided to lower its target for the federal funds rate by 50 basis points to 4% 10 May 26, 2001 Bush tax cut through the Economic Growth and Tax Relief Reconciliation Act passed by House of Representatives and Senate 11 June 27, 2001 The Federal Open Market Committee decided to lower its target for the federal funds rate by 25 basis points to 3.75% 12 August 21, 2001 The Federal Open Market Committee decided to lower its target for the federal funds rate by 25 basis points to 3.5% 13 September 11, 2001 Terrorist attacks on the World Trade Center 14 September 17, 2001 The Federal Open Market Committee decided to lower its target for the federal funds rate by 50 basis points to 3% 15 October 2, 2001 The Federal Open Market Committee decided to lower its target for the federal funds rate by 50 basis points to 2.5% 16 October 16, 2001 John Reich, director of the Federal Deposit Insurance Corporation, warned of an increase in subprime lending and the risky securitization of mortgages 17 November 6, 2001 The Federal Open Market Committee decided to lower its target for the federal funds rate by 50 basis points to 2% 18 December 11, 2001 The Federal Open Market Committee decided to lower its target for the federal funds rate by 25 basis points to 1.75% 19 January 30, 2002 The Federal Open Market Committee decided to keep its target for the federal funds rate unchanged at 1.75%

Selected events

18The first major event that we selected is the announcement of the results of the American presidential elections; the electoral programme of the former candidate Bush stressed the importance of reducing taxes hence conveying a potential positive signal to the market and to investors’ community. Over the same time period under study, the Fed’s monetary policy mainly concerning the level of interest rates seemed to have a very strong impact on banks and investors’ behaviour (Tully, 2007; Crouhy, Jarrow and Turnbull, 2008; Daglish, 2009; Okongwu and Sabry, 2009). More specifically, at the beginning of the period under study, the interest rate level was maintained. This could also be considered as positive news as reflecting the stability of the economy and markets. However, soon afterwards, A. Greenspan’s monetary policy decisions subordinated to the aim of maintaining economic growth were largely criticized. One of the first to criticize Greenspan’s monetary policy was Governor Edward M. Gramlich, who tried to warn public opinion on the impact of deregulation on financial markets at the Federal Reserve Bank of Philadelphia’s Community and Consumer Affairs Department Conference on Predatory Lending. Governor Gramlich presented his arguments on the necessity of adopting new regulations as a higher-cost home loan in order to avoid market panic. Another announcement made by the director of the FDIC, John Reich, should have impacted the investors’ behaviour. He warned financial markets on the high probability of failure of the subprime lender Superior Bank, as a result of an increasing demand for subprime loans and risky securitization of mortgages. Both announcements should have had a negative impact on financial markets by raising doubts on the increase in the demand for subprime loans and market stability. Both critics followed the strong decline in interest rates levels over a single year. From January 3, 2001 to December 11, 2001, interest rates experienced a decrease by 475 basis points following the goal of sustaining economic growth. However, as stressed by Graddy et al. (1994) too, the implementation of such a policy may be difficult because of several events that may intervene during the process and perturb it. For example, such a policy may have unexpected effects like larger risk-taking behaviours. In addition, decreasing interest rates too much may also provide the opposite signal regarding the stability of the economy, i.e. deterioration of the economic environment. Therefore, at a certain point during the sample period we should observe a decline in bank returns.

19Finally, two other external events occurred during our sample period. The first one is the implementation of the Economic Growth and Tax Relief Reconciliation Act in May 2001 with one main objective that was to reduce taxes and encourage consumption and investment, on the real estate market among others. The second event is the terrorist attack on the WTC which impacted not only the United States but also the rest of the world. Its aftermath panic disturbed markets and customers as well as bankers.

20In line with Graddy et al. (1994), we will perform our analysis based on Stone’s (1974) two-factor model. More specifically, our equations will express a security’s yield, i.e. return of the stock, as a function of both equity and debt indices. Moreover, to take into account all the events we have selected, we are going to include dummies variables, one for each event, as follows:

21

22where R_jt represents the daily rate of return of our equally weighted portfolio j composed by banks, at time t, R_mt is the daily rate of return on the S&P 500 index at time t, R_jt stands for the Barclays debt index in day t and D_? proxies the ?^th event; this variable takes the value of 1 in the intervention window, otherwise, its value equals 0. In this setting, ?_? is the coefficient associated to each intervention date which allows observing the direct impact of the event on the portfolio’s return, while ?_? and ?_? are linked to the same intervention dates associated respectively to the equity and debt indices. The interaction of these coefficients allows understanding the structural changes in the return-generating process influenced by the intervention action. The null hypothesis under test for our portfolio of banks is that the selected events do not affect its returns, i.e. ?_? = ?_? = ?_? = 0.

23The debt index included in the model allows integrating information about wealth transfers between shareholders, depositors and borrowers, linked to unexpected moves in interest rates (Flannery and James, 1984; Kwan, 1991).

24In order to deepen our analysis, we decided to separate our general portfolio of banks into subgroups based on different criteria which will be detailed later. This is in line with Graddy et al. (1994) who tested the impact of several events on two separate groups of financial institutions: banks and S&Ls.

25The technique we use now in order to test the preceding equation is that of the Seemingly Unrelated Regression which is an extension of the linear regression model on a system of equations that allows error terms to be correlated.
Below is the system of equations that we tested. Two equations use the short-term index, one for the first group and the other for the second group. The two other equations follow the same principle but, this time, with the long-term index.

Data

26We used two international rankings in building our sample: the rank of the United States’ Largest Banks with respect to the value of their assets in dollars (Federal Reserve System, National Information Center) and the rank of the top 50 Bank Holding Companies based on the value of all their deposits in dollars (FDIC). We selected a sample composed by 25 banks among the most important in the world. All the banks in the sample are located in the United States and are all big size, internationally operating investment banks. Our sample contains the four major banks which hold more than half of U.S. banking assets, i.e. Bank of America, Citigroup, JP Morgan and Wells Fargo. [7] Unfortunately, due to data availability issues, we had to reduce the sample to 19 banks which are recorded in Table 2.

Table 2

Sample of banks in our portfolio

Bank of America Merrill Lynch JP Morgan Chase & Co. Fifth Third Bancorp Wells Fargo & Co. Westamerica Bancorp Citigroup Harris Cap Suntrust Banks Trustco Bank NY US Bancorp Marshall & Ilsley City National Huntington BCSH. Lehman Bros Morgan Stanley PNC Finl Capital One Finl Keycorp

Sample of banks in our portfolio

27Yap and Pierson (2009) published a ranking of the world’s biggest banks based on the level in dollars of their writedowns and credit losses. Our sample contains 8 banks over the first 20 banks in this ranking and 15 over 75. In our selection of financial institutions, we tried to integrate those which were the most impacted by the crisis. This should allow us to clearly see whether or not the events described in the previous section had a significant impact on those major players and whether these events could be one potential cause of the crisis.
All the daily data on banks stock prices as well as on both debt and equity indices were extracted from Datastream. They cover the time period going from October 6, 2000 to April 19, 2002. The equity index we have chosen is the Standard & Poor’s 500. For the debt index, we use two different holding-period return measures. The first one is a short-term period for which we are going to use the Barclays US Treasury Bills index. The second one is the long-term period for which we have chosen to use the Barclays Global Treasury Long index. The objective was to take into account the fact that there are different maturity structures for debt indices. All the computations are done on logarithmic returns.

Empirical evidence

28We started by performing a preliminary estimation with a standard OLS model. [8] We estimated a modified version of Equation 1 by including both short-term and long-term debt indices hence, testing the impact of events by accounting for the maturity structure of the debt. For our dummy variables, we used a two - day window in order to see whether the event had an impact before it was announced due to potential leaks of private information or whether it had an impact immediately after because of the delay investors need to get the information.

29Our preliminary results show that none of the coefficients is significant. [9] All the p-values, except for the one of the equity index, were above the standard levels of confidence. Therefore we can state that the chosen events, occurring between 10/09/2000 and 04/19/2002, did not have any significant impact on the banks’ returns.

30We then perform an analysis of the different sub portfolios.

31The first criterion we used to create two new groups was the level of banks’ implication in securitization practices. As we have mentioned in the data description part, five banks in our sample, which will compose the first group, are largely involved in the securitization process [10] whereas the others, which will be included in the second group, are not such major players of this type of practice. Table 3 provides the repartition of our sample of banks among the two groups.

Table 3

Composition of the two sub portfolios of banks based on their involvement in the securitization practices

Portfolio 1 : strongly involved in securitization Portfolio 2 : weakly involved in securitization Bank of America Suntrust Banks JP Morgan Chase & Co US Bancorp Wells Fargo & Co City National Citigroup Lehman Bros Morgan Stanley PNC Finl Keycorp Merrill Lynch Fifth Third Bancorp. West America Bancorp Harris Cap Trustco Bk NY Marshall & Ilsley Huntington BCSH Capital One Finl

Composition of the two sub portfolios of banks based on their involvement in the securitization practices

32The results we got for the coefficients of the intervention dummy variables are presented in Table 4.

Table 4

Coefficients for the intervention dummy variables (D), decomposition based on securitization level

Intervention Coefficients Treasury Bill Index Model Treasury Bond Index Model Banks strongly involved in securitization Banks weakly involved in securitization Banks strongly involved in securitization Banks weakly involved in securitization ?1 0.000969 –0.000734 0.000906 –0.000791 (0.9887) (0.8568) (0.9894) (0.8449) ?2 –0.013290 –0.011267 –0.013219 –0.011272 (0.8218) (0.0012)*** (0.8226) (0.0014)*** ?3 0.000139 0.003295 –2.32E–05 0.003241 (0.9983) (0.3783) (0.9997) (0.3888) ?4 0.020119 0.018423 0.021051 0.018524 (0.7744) (0.0000)*** (0.7623) (0.0000)*** ?5 0.011661 0.004091 0.011444 0.004079 (0.8444) (0.2462) (0.8470) (0.2417) ?6 0.002306 –0.0021489 0.002682 –0.002020 (0.9689) (0.5355) (0.9638) (0.5652) ?7 –0.006004 –0.002036 –0.006260 –0.002112 (0.9194) (0.5595) (0.9158) (0.5443) ?8 –0.009160 –0.009523 –0.009014 –0.009455 (0.8940) (0.0182)** (0.8956) (0.0199)** ?9 –0.001388 0.000679 –0.002278 –5.76E–05 (0.9830) (0.8818) (0.9717) (0.9881) ?10 0.003472 –0.002437 0.003542 –0.002404 (0.9651) (0.6002) (0.9644) (0.6212) ?11 –0.003163 0.001300 –0.003370 0.001208 (0.9573) (0.7089) (0.9545) (0.7446) ?12 0.003526 0.000945 0.003544 0.000930 (0.9548) (0.7961) (0.9545) (0.7984) ?13 –0.001824 –0.002037 –0.001806 –0.002014 (0.9818) (0.6653) (0.9820) (0.6670) ?14 0.004659 0.003132 0.004600 0.003063 (0.9525) (0.5102) (0.9530) (0.5026) ?15 0.002473 –0.011404 –0.001598 –0.013372 (0.9755) (0.0162)** (0.9843) (0.0114)** ?16 0.004528 –4.42E–05 0.001185 –0.001455 (0.9483) (0.9917) (0.9865) (0.7378) ?17 0.008411 0.002070 0.008503 0.002103 (0.9155) (0.6559) (0.9145) (0.6507) ?18 0.002383 –0.002399 0.002622 –0.002232 (0.9799) (0.6660) (0.9779) (0.6874) ?19 –0.005717 –0.004526 –0.004742 –0.004280 (0.9235) (0.2042) (0.9405) (0.4678) The figures in parentheses are the probabilities associated to the t-statistic. ** Significant at 5%, *** Significant at 1 %. The Breush-Pagan tests of whether the residuals from the equations of the system are independent record p-values that are less than the critical value for a conventional risk level as low as 1 % (p-values = 0.0000), this is to say that they are not independent. Residuals are not Gaussian, i.e. p-values of the Jarque-Bera tests = 0.0000, no autocorrelation (Durbin-Watson statistics around 1.9).

Coefficients for the intervention dummy variables (D), decomposition based on securitization level

33The first point that has to be noticed when looking at Table 4 is the results are strictly similar regardless of the use of short or long term debt indices; the same coefficients are significant and their sign and amplitude are the same. Moreover, only the events linked to the interest rate policy announcements have significant impacts on bank returns. [11] Finally, compared to our expectations, only the first impact does not follow our intuition, i.e. negative coefficient while we were expecting the opposite. As suggested by one anonymous referee, we also applied a sequential elimination of regressors procedure as in Brüggeman and Lütkepohl (2001). More precisely, we estimate our model and delete sequentially those regressors with the smallest absolute values of their corresponding t-ratios until all remaining regressors are significant. A single regressor is eliminated at each step only. The use of this procedure leaves our conclusions unchanged; to ease the presentation, these results are only available upon request.

34We propose a separate analysis of the four events that show significant impacts on bank returns. First, the announcement of the FOMC to maintain the target rate negatively impacts banks return. This may still be the result of the internet bubble burst as this announcement occurred less than one year after the crisis. Hence, investors might still experience confidence problems regarding the stability of the financial environment. When the FOMC releases the same type of announcement one month later, bank returns answer positively; this new announcement might have signalled that the economy found its path towards stability. However, two weeks later, the FOMC begins its policy of decreasing the target rate. This started to negatively impact financial markets four months later. Bank returns negatively react to this announcement which is obviously interpreted as bad news. Based on previous and still fresh experience, investors might fear that the economy might plum back to instability, with an explosion of liquidity on the market. This trend continues up to a ninth consecutive drop which allows the target rate to reach a level as low as 2½ %. The impact on bank returns is negative and stronger than previously as the confidence crisis starts.

35Table 4 also shows that the four significant events seem to impact only the second group composed by the banks less involved in securitisation. Several hypotheses can be proposed in order to interpret these results. First of all, this result can be linked to the range of activities developed by each of the two categories of banks in the two sub portfolios. P2 group’s range of activities can be considered as less diversified; hence, a decrease in interest rates may affect them more than the other institutions as it directly impacts the main line of their business. For example, we can suppose that by not being too much involved in securitisation and therefore, without the possibility of transferring the risks associated to subprimes via these practices, those banks offered less subprime mortgages, initially conceived as low initial interest rate products. Therefore, P2 banks could be more impacted by a decrease in interest rates that they had to integrate in their other products compared to banks who were already dealing with mortgages at low interest rates.

36Another hypothesis which can be linked to the first one is investors’ moral hazard behaviour. Banks in the second group experience a competitiveness problem compared to banks involved in innovative and profitable investments provided by the securitization techniques. This situation is even more exacerbated in a context of tremendous development of government insurance. The fact that the government sponsored agencies guarantees provided an incentive for banks, as well as for investors, to get involved in riskier investments having a “last resort insurer” that would cover for their losses. At this point, as we have seen in the literature review, credit risk completely disappears. The only remaining incentive is the yield enhancement which leads to moral hazard behaviour. Therefore, with the expansion of securitization, banks heavily involved in such practices (P1) seem to attract the majority of capital flows, i.e. get all the benefits and put banks in the second group (P2) in a bad position. Hence, banks in the latter group become more sensitive to market moves. As such, they are more affected than those which are protected (at least at the beginning) by securitization.

37Another advantage of being involved in securitization is the complexity of this financial innovation which rendered it very opaque for investors who tried to price assets. Banks involved in these practices benefited from this opacity at least at the very beginning. Investors could only see the positive features of securitization and fully trust banks in explaining these new products to them. It was again beneficial for banks in the first group.

38Finally, borrowers’ behaviour can also be an explanation of our results as it directly impacts banks activities. One illustration is the evolution of the demand for subprime mortgages. Thanks to the decrease in interest rates, borrowing became a way of living. In addition, financial innovation contributed to the development of risky loans as banks involved in securitization practices held a guarantee in case of borrower’s default. On the contrary, banks not involved in securitization did not benefit from the possibility to transform below investment grade assets into AAA. This made banks in our second group (P2) less attractive compared to those in the first sub group (P1) and may explain the negative reaction of P2 banks to the FOMC decision to decrease the interest rate. However, it does not explain the lack of reaction for P1.

39Following Graddy et al. (1994) we then perform Wald coefficient tests to check several hypotheses that help getting a clearer picture of our conclusions. The first hypothesis, that we call H_1A, is provided below:

40

41This test assesses the joint impact of the intervention intercepts. The chi-square statistics that we obtain for the short and long term debt index models are 44.79 and 47.85 respectively. For 39 degrees of freedom and a level of significance of 5 %, its critical value is 53.10. Therefore, we cannot reject the null hypothesis. The chosen events, all together, did not have any impact on banks’ returns in line with Graddy et al. (1994) conclusions.

42The next hypothesis we test allows us to see whether the risk intervention regressors had an impact on the independent variable which is the group’s return H_2A as follows:

43

44For this test, we obtain a chi-square statistic of 29.48 for short term debt index model. For a critical value of 97.07, corresponding to 76 degrees of freedom and a 5 % level of significance, we cannot reject the null hypothesis that the coefficients for the risk interaction terms are jointly equal to zero. We get to the same conclusion for the long term index model with a chi-square statistic of 29.57 which is below the critical value 92.52 for 72 degrees of freedom and a level of significance of 5 %. Put it differently, structural changes associated with the intervention dates did not have any impact on banks’ return which is again in line with the results reported by Graddy et al. (1994) on the banks versus S&Ls portfolios.
We then wanted to monitor the event coefficients for the two subgroups; to do so, we formulate our hypothesis as follows:

Table 5

Chi-square statistics of each coefficient ? associated with the intervention date (Hypothesis 2A)

? ?²(ST) ?²(LT) 1 0.033309 0.038960 2 10.50041*** 10.24664*** 3 0.779897 0.746435 4 16.62694*** 20.41415*** 5 1.359220 1.386092 6 0.390667 0.337970 7 0.344483 0.372338 8 5.588130* 5.433235* 9 0.023423 0.001418 10 0.281120 0.250694 11 0.145696 0.112873 12 0.068328 0.066963 13 0.187252 0.185196 14 0.433957 0.449862 15 5.838602* 6.448051** 16 0.004442 0.113832 17 0.204405 0.211344 18 0.189326 0.164868 19 1.613473 0.528549 * Significant at 10%,** Significant at 5%, *** Significant at 1%.

Chi-square statistics of each coefficient ? associated with the intervention date (Hypothesis 2A)

45The next step was to jointly test the intervention and risk coefficients. To run this test, we propose the following hypothesis:

Table 6

Chi-square statistics of each equality ? = ? = ?, associated with the intervention date (Hypothesis 2B)

? ?²(ST) ?²(LT) 1 0.074168 0.070056 2 14.30780** 14.37426** 3 1.689992 1.529515 4 30.77452*** 30.90979*** 5 4.956683 5.000538 6 3.357626 3.422891 7 9.167381 9.309009 8 9.029933 9.053286 9 0.499833 0.502585 10 1.021017 1.018254 11 0.152596 0.157235 12 0.385000 0.380355 13 0.766584 0.766884 14 0.485475 / 15 6.892603 6.952656 16 2.448869 2.522404 17 0.337564 0.348235 18 0.328742 0.326883 19 5.437713 5.470737 * Significant at 10%,** Significant at 5%, *** Significant at 1%.

Chi-square statistics of each equality ? = ? = ?, associated with the intervention date (Hypothesis 2B)

46The three last hypotheses assess whether there is a difference between the two subgroups regarding the impact of the different events in terms of wealth and risk. The first one focuses on the effect of the nineteen interventions on the banks abnormal returns.

47

48The second hypothesis focuses on the market risk.

49

50Finally, the third hypothesis concerns the sensitivity of the abnormal returns with respect to changes in interests.

51

Table 7

Chi-square statistics of each equality of coefficient (?1 = ?2; ?1 = ?2; ?1 = ?2) (Hypothesis 3)

Null Hypothesis Short term debt index model Long term index model H3A 0.014080 0.012210 H3B 0.002057 0.065145 H3C 0.000949 0.001067

Chi-square statistics of each equality of coefficient (?1 = ?2; ?1 = ?2; ?1 = ?2) (Hypothesis 3)

52The null hypothesis of the equality between coefficients cannot be rejected for the three tests because the chi-square statistics are far below the critical value of 3.50 associated to a 5 % level of significance. Differently said, the two groups had the same reaction in front of each announcement. Changes in the market risk impact both groups in the same way.

53In order to check for the robustness of our results we performed the same analysis by including two-period leads and lags on both the market and debt indices. All our conclusions remain unchanged. [12] We also conducted our analysis with two-period leads and lags considering now a permanent change after the intervention date. The results are roughly similar.

54The second criterion we used separate our portfolio into two was the level of banks’ losses following the financial crisis. The reason underlying this choice is due to the fact that we wanted to assess how far the events we have chosen have impacted banks and whether or not they have had a larger effect on banks experiencing higher losses ex-post than the others. As we have mentioned in the data description part, eight banks in our sample, which will compose our first subgroup, were among the first twenty banks which lost huge amounts whereas the others, which will be included in the second subgroup, experienced more limited losses. Table 8 provides the distribution of the banks in our sample among the two subgroups.

Table 8

Composition of the two portfolios based on the consequences of the crisis

Portfolio 1 : most impacted by the crisis Portfolio 2 : less impacted by the crisis Bank of America Suntrust Banks JP Morgan Chase & Co US Bancorp Wells Fargo & Co PNC Finl Citigroup Keycorp Morgan Stanley Fifth Third Bancorp. Merrill Lynch West America Bancorp City National Harris Cap Lehman Bros Trustco Bk NY Marshall & Ilsley Huntington BCSH Capital One Finl

Composition of the two portfolios based on the consequences of the crisis

55The results we got for the coefficients of the intervention dummy variables are synthesized in Table 9 and are strictly similar to those reported for the previous portfolio decomposition. We also tested the seven hypotheses described above and their conclusions do not change. Regarding the four events found as significant, they are the same than for the first two sub groups. They include the two annoucements for the maintenance of the target rate and the fourth and the ninth decreases of this target rate.

Table 9

Coefficients for the intervention dummy variables (D), decomposition based on the impact of the crisis

Intervention Coefficients Treasury Bill Index Model Treasury Bond Index Model Banks most impacted by the crisis Banks less impacted by the crisis Banks most impacted by the crisis Banks less impacted by the crisis ?1 3.14E–06 –0.000567 5.80E–05 –0.000545 (0.9999) (0.8973) (0.9989) (0.9005) ?2 –0.010671 –0.012645 –0.010506 –0.012571 (0.7734) (0.0008)*** (0.7766) (0.0009)*** ?3 0.002215 0.002661 0.002085 0.002585 (0.9559) (0.5100) (0.9584) (0.5229) ?4 0.017052 0.019586 0.017921 0.019956 (0.6990) (0.0001)*** (0.6818) (0.0000)*** ?5 0.011509 0.002131 0.011371 0.002111 (0.7578) (0.5762) (0.7601) (0.5731) ?6 0.001423 –0.002720 0.001687 –0.002569 (0.9695) (0.4677) (0.9638) (0.4970) ?7 –0.002570 –0.003442 –0.002772 –0.003554 (0.9450) (0.3613) (0.9406) (0.3429) ?8 –0.008489 –0.010102 –0.008439 –0.010062 (0.8441) (0.0204)** (0.8449) (0.0213)** ?9 –0.001005 0.001171 –0.001977 0.000241 (0.9804) (0.8136) (0.9609) (0.9536) ?10 0.001046 –0.002283 0.001119 –0.002232 (0.9833) (0.6493) (0.9821) (0.6704) ?11 –0.002825 0.002260 –0.0029986 0.002152 (0.9393) (0.5481) (0.9359) (0.5908) ?12 0.002161 0.001230 0.002164 0.001216 (0.9559) (0.7556) (0.9557) (0.7563) ?13 –0.001354 –0.002453 –0.001329 –0.002418 (0.9785) (0.6297) (0.9789) (0.6311) ?14 0.002824 0.004100 0.002742 0.003995 (0.9542) (0.4258) (0.9554) (0.4164) ?15 –0.002372 –0.011716 –0.005112 –0.013721 (0.9626) (0.0221)** (0.9200) (0.0163)** ?16 0.002883 –0.000137 0.000526 –0.001675 (0.9476) (0.9760) (0.9904) (0.7209) ?17 0.008105 0.000559 0.008160 0.000592 (0.8707) (0.9114) (0.8697) (0.9057) ?18 0.000491 –0.002374 0.000747 –0.002142 (0.9934) (0.6925) (0.9900) (0.7196) ?19 –0.003991 –0.005591 –0.002714 –0.004765 (0.9150) (0.1467) (0.9460) (0.4562) ** Significant at 5 %, *** Significant at 1 %; p-values for the Breush-Pagan tests = 0.0000, JB p-value = 0.0000, no autocorrelation.

Coefficients for the intervention dummy variables (D), decomposition based on the impact of the crisis

56In addition, in this case, we can observe that only banks having recorded less important losses seem to be impacted by the four announcements of the FOMC compared to banks heavily touched by the financial crisis.

57Some of the explanations proposed for the previous decomposition of our general portfolio of banks can also be used here, like the limited range of activities. Banks with huge losses generally show larger size compared to the one composing our second sub group. Therefore, smaller banks can be more impacted by a decrease in interest rates because of their more restrained business activities or size. They can absorb less than larger banks which can benefit from more opportunities, i.e. securitization for example. Indeed, one significant point to notice is the similarity between the sub group of banks weakly involved in securitization and the sub group of banks with less important losses. Only three banks, Merrill Lynch, City National and Lehman Bros, have been taken off from the first sub group. This means that the majority of the new P2 group is composed by banks less involved in securitization practices. This is the reason why we can also use the conclusions drawn previously, namely those based on the use of securitization mechanisms.

58In our approach, we use the OLS/SUR estimator on an equally weighted portfolio. We though implicitly assume that all the coefficients are the same for all assets in this portfolio. In order to lessen this assumption and cross check the robustness of our results to this hypothesis, we also perform alternative estimations, namely standard panel ones. [13] The results that we obtain are consistent with those obtained previously. To save space, we only provide the coefficients of the intervention dummy variables on the two sub portfolios created on the securitization criteria (Table 10). The results for the second group of sub portfolios are strictly similar, and can be provided upon request.

Table 10

Coefficients for the intervention dummy variables (D) [14], panel estimations

Intervention Coefficients Treasury Bill Index Model Treasury Bond Index Model Banks strongly involved in securitization Banks weakly involved in securitization Banks strongly involved in securitization Banks weakly involved in securitization ?1 0.003266 0.000502 0.000183 0.000177 (0.9687) (0.9108) (0.9982) (0.9673) ?2 –0.011816 –0.010373 –0.014062 –0.010061 (0.8554) (0.0030)*** (0.8504) (0.0119)** ?3 –0.001280 0.000580 –0.000981 0.000389 (0.9845) (0.8704) (0.9891) (0.9193) ?4 0.020481 0.018311 0.021133 0.016608 (0.8207) (0.0002)*** (0.7693) (0.0000)*** ?5 0.008690 0.000770 0.009107 0.000876 (0.9003) (0.8366) (0.8838) (0.7932) ?6 0.002476 –0.002280 0.003515 –0.000455 (0.9673) (0.4837) (0.9602) (0.9040) ?7 –0.006292 –0.002915 –0.009794 –0.005216 (0.9225) (0.4024) (0.8787) (0.1299) ?8 –0.009938 –0.008871 –0.008796 –0.008150 (0.8882) (0.0198)** (0.9109) (0.0533)* ?9 –0.006492 0.007023 0.001066 –0.000582 (0.9610) (0.3259) (0.9886) (0.8849) ?10 0.003265 –0.001624 0.001837 –0.000925 (0.9678) (0.7090) (0.9885) (0.8920) ?11 –0.001334 0.001829 –0.002084 –0.000420 (0.9841) (0.6123) (0.9848) (0.9432) ?12 0.004619 0.001733 0.003981 0.001362 (0.9456) (0.6344) (0.9512) (0.6964) ?13 –0.002018 –0.003128 –0.001953 –0.002100 (0.9814) (0.5024) (0.9808) (0.6303) ?14 0.005190 0.006343 0.004450 0.003285 (0.9650) (0.3196) (0.9553) (0.4404) ?15 0.003440 –0.010494 –0.006641 –0.020148 (0.9671) (0.0194)** (0.9564) (0.0020)*** ?16 0.009495 0.001665 –0.009486 –0.007468 (0.9122) (0.7194) (0.9225) (0.1533) ?17 0.007769 0.001089 0.008798 0.001928 (0.9236) (0.8029) (0.9148) (0.6623) ?18 0.001950 –0.002706 0.002253 –0.001204 (0.9842) (0.6093) (0.9816) (0.8187) ?19 –0.005785 –0.004572 –0.006107 –0.002111 (0.9272) (0.1802) (0.9632) (0.7664) The figures in parentheses are the probabilities associated to the t-statistic. ** Significant at 5 %, *** Significant at 1%, * Significant at 10%, p-values for the Breush-Pagan tests = 0.0000, JB p-value = 0.0000, no autocorrelation.

Coefficients for the intervention dummy variables (D) [14], panel estimations

59Our results show that the American interest rate policy at the beginning of the 21^st century designed to sustain growth and increase the demand for credit by deeply decreasing the target rate [15] had some impact on bank returns. The impact is significant only for banks less involved in securitization practices / less affected by the financial crisis afterwards and is mainly negative. We may advance two potential explanations. First of all, these results point out that this category of banks, e.g. less active in designing and offering asset-backed securities products, did not seem to benefit, at least at the very beginning, from the successive interest rate cuts. Unable/less able to take advantage of this low interest rate environment because of a less diversified offer of innovative practices and financial products and hence, a lower ability to transfer and reallocate risks, these banks are in a sense punished by the market and record negative significant returns. Securitization practices appear thus as a way of avoiding such a discount. Second, the negative market reaction may also suggest that investors’ fears regarding a potential explosion of the liquidity on the market following an unlimited development of risky practices favoured by a low rate context offset the positive potential effects that the boom of activities in the financial sector may have on bank returns. The experience of the Internet bubble burst and of its consequences seems to fuel market investors’ reluctance concerning the potential benefits of a cheap money environment and a run for yield enhancement behaviour. It is hence difficult to put forward a clear cut conclusion as about the exact impact of monetary policy decisions on bank returns in a complex regulatory context and a highly competitive and rapidly changing financial environment.

60By consequent, securitization appears as an attractive practice, almost a guarantee against credit and default risks. As such, banks involved in securitization benefit from a competitive advantage mainly through the level of their interest rates. This safety illusion pushed investors to be more and more involved in the process and so, almost totally neglect the risk of lending money to borrowers with poor credit standards. As S&Ls did in the 1980s – following the deregulation and the new opportunities to expand their activities in commercial loans – at the end of the 1990s, banks took advantage of the numerous opportunities provided by securitization techniques to earn higher returns “without” apparently increasing the risk. Their objective of yield enhancement made them deaf to alarm signals like those of Governor Edward M. Gramlich at the Community and Consumer conference or of director John Reich of the FDIC. Unfortunately, securitization allowed international financial markets to be more interconnected than ever, involving investors sometimes very far from the original source of the credit; hence, they were unable to understand and manage all the risks associated to the underlying assets. This was the main channel through which the crisis spread up to the global financial community. Those ties did not exist during the S&Ls crisis which may explain its regional consequences that mainly affected the American financial market.

61To conclude, the two crises have, at their beginning, the same origins, i.e. a real estate and a credit bubble, but their consequences strongly differ. One of the main reasons is the securitization process which allowed the globalization of financial products and flows but also of investor behaviours across countries. Therefore, when the panic starts, it affects the whole community of investors who possess the same assets, with reduced diversification at the international level.

62Despite their different consequences, one has to mention another common point between the two crises, namely the opacity surrounding some very innovative financial products and practices. During the S&Ls crisis, investors had to deal with the opacity of depository institutions’ managerial decisions and the newly introduced accounting standards while during the subprime crisis, the opacity characterized the securitization practices. The final result in both crises was investors’ riskier behaviour.

63There are at least two arguments that support our event study analysis and the relevance of our results in the specific context of the subprime crisis. [16] First of all, our results stress the role played by government decisions in terms of interest rate policy, namely the low level of interest rates. This interest rate policy was particularly important during the period under analysis, as it ended up with the FED rate being at levels well below what experience during the previous two decades of good macroeconomic performance would have predicted. The particular importance of the interest rate policy for banks’ behaviour under the period under study is closely linked to the housing and mortgage markets. Taylor (2007) performs simulations that allow him conclude that starting with mid 2000, a higher federal funds rate path than the one actually observed would have avoided much of the housing boom. There is also evidence that the high housing inflation led to a marked reduction in delinquency and foreclosure rates. Before the reversal, many mortgages and mortgage backed securities were issued with ratings reflecting these unusual low delinquency rates. Therefore, risk assessment was particularly difficult in such a context, in which mortgages were also packaged into securities that combined different risk profiles. Said it differently, more than never during the subprime crisis persistently low real interest rates fuelled a boom in asset prices and securitized credit and led banks to take on increasing risk and leverage.
The effect of monetary policy decisions on bank returns depends on banks’ characteristics. However, as suggested by Altunbas et al. (2007), securitization has changed the bank lending channel mechanism as it also probably altered the bank characteristics that are usually set forward in order to identify shifts in loan supply. The size indicator matters less as securitization can considerably reduce the amount of loans on banks’ balance sheets, the short-term inflows generated by the sale of securitized products modify liquidity ratios and the regulatory requirements for capital may also be reduced while the standard asset-to-capital ratio becomes a poor indicator of the relevant capital constraints faced by banks. All in all, securitization may insure banks with additional flexibility to face changes in market conditions associated with monetary policy decisions. Our results seem to be in line with these ideas, as the involvement in securitization practices seems to be a discriminant criterion within our portfolio of banks. Moreover, the effects of the monetary policy decisions, namely interest rate levels, affect only the banks less involved in those practices, hence less flexible in responding to these decisions.