1
Was the CPP announcement informative for investors?
The impact of the CPP announcement on abnormal
returns of bank
1Mark de Vos
University of Amsterdam 31 January 2015
Abstract
This paper researched the abnormal returns of banks on the CPP announcement day. Results show that the information provided on the CPP announcement day caused investors to make informed decision on which banks would be able to benefit from the CPP. Worse performing banks, more in need of liquidity, which were still viable, had significantly higher abnormal returns. Further research also provides information about the information announced on the day preceding the CPP announcement. There is reason to believe that investors perceived this information as the savior or aid the whole financial system was waiting for.
Mark de Vos (0515272) Master thesis Finance
dr. R.E. Vlahu
Master of Business Economics: Finance
1 The author’s gratitude goes out to Frank de Vos for his aid on the econometrical part, to Bjorn Witlox for his tremendous support in retrieving all the data from the different databases, and to Razvan Vlahu for his guidance on the subject matter.
2
Contents
1. Introduction 3
2. Capital Purchase Program (CPP) 5
3. Related literature and hypotheses 10
4. Variables 13
4.1 Balance sheet variables 14
4.2 Performance ratios 17 4.3 Risk measures 18 5. Methodology 21 6. Results 26 7. Robustness Checks 34 8. Conclusion 38 References 41 Appendix 43
3
1. Introduction
The recent financial crisis in 2008 was in several ways different than other more recent crises. The economic downfall was the largest since the great depression, but more important the cause and resulting effects of the recent financial crisis were unseen before. The problems caused by restructured mortgages and shady off-balance sheet activities led to a new sort of crisis. In which, not the public (i.e. depositors) were in doubt about the solvability of banks or the banking system as a whole, but companies and the banking system itself. The public did not comprehend the financial difficulties the financial system was in and were therefore too late to act, while the companies were better safe than sorry by drawing down their existing credit lines (Ivashina and Scharfstein, 2010). However, it was the lack of trust that banks had among each other, which resulted in a lack of liquidity and the near demise of the financial system.
This new kind of financial crisis resulted in a number of unprecedented interventions by governments. These new interventions have led to many researches if they were effective or not. Fortunately, most interventions did what they were designed for (Wu, 2008).
However, the most expensive intervention, probably the most important, intervention of them all showed some conflicting research findings. The Emergency Economic Stabilization Act (EESA) or more commonly known as the Troubled Assets Relief Plan (TARP), which enabled an injection of $250 billion in the US financial system, through the Capital Purchase Program (CPP), was not continuously seen as beneficial for everyone. While, it eventually earned over $86 billion for the US economy (Veronesi and Zingales, 2010). At the initiation period a large portion of the public and media were against it. Even investors did not show confidence in the CPP in the first few quarters after its initiation, with CPP participating banks having lower stock returns than other banks (Ng, Vasvari and Wittenberg-Moerman, 2011).
The question arises: were the contents of the CPP well communicated? How else would there be such a discrepancy between the view of the media, public, and investors. Ng, Vasvari and Wittenberg-Moerman (2011) blames the media for creating the negative view of the CPP to the public and investors, but if the media and/or the investors and the public were well informed in the first place, this should not be possible. This paper investigates if the CPP announcement created abnormal returns for banks, thus leading to believe that investors did perceive the CPP to be positive. Also, which banks benefitted more from the CPP
4
announcement will be researched. This is in order to determine if investors were so well informed that they could even predict which banks would eventually receive aid from the CPP.
The abnormal returns of 424 US banks, of which most were bank holding companies, were researched. While, some banks did experienced abnormal returns, the mean of the abnormal returns of all the banks did not differ from zero. Future CPP participating banks did experienced a significantly higher abnormal return than non-CPP participating banks. The regression results show it was clear to the investors what banks would be aided by the CPP. Worse performing banks and banks in need for liquidity, which were still viable experienced significantly higher abnormal returns. The worse performing banks were banks, which had higher non-performing assets ratios and higher non-interest income ratios. The banks in need of liquidity had lower deposit ratios and lower liquid asset ratios. This showed exactly the problems of the instigation and of the resulting effects of the financial crisis. The instigation resulted in the forced devaluation banks had to do on their restructured loans and other crafty and unclear constructed products, which resulted in higher non-performing assets ratios and larger losses in the non-interest income. As well as, the public seeing the values of their houses decline resulting in a larger portion of mortgages being unpaid. The resulting effect of the financial crisis caused by a lack of mutual trust between banks, a shortage of liquidity, is showed by the importance the deposit ratios and the liquid assets ratios. Investors also knew that the CPP was not intended to aid non-viable banks, which is shown by banks with the worst performance ratios to experience significantly lower abnormal returns.
The duration of the event is also researched. In other words, was all the information about the CPP announced on 14 October, 2008 (the CPP announcement day) or was some of the information already disclosed? The results show that there was information provided on 13 October, 2008 on which investor acted. Although, the information on both days is not as clear as on only the CPP announcement day it seems that the information on 13 October, 2008 was perceived by investors as a savoir of the financial system. Either, by the way the CPP infusions would be distributed, or the fact that the decision to enact the CPP came closer.
The rest of the paper is structured as follows. Section 2 provides an explanation of all the events leading up to the CPP announcement as well as the results from the CPP. In section 3, the related research is discussed and the hypotheses are presented. Section 4
5
provides a detailed description of all the variables used in the regressions. The methodology and the data obtained are discussed in section 5. Section 6 shows the results of the regressions and provides an analysis on the obtained results. The robustness checks are done in section 7. Section 8 concludes.
2. Capital Purchase Program (CPP) 2
In September 2008 the financial crisis in the US reached its height with the near collapse of the entire US financial system after Lehman Brother’s bankruptcy and the government rescue of AIG. This instigated an acceleration in the decline of syndicated lending3. Lending volume
decreased in the fourth quarter of 2008 (2008Q4) with 47% in comparison to the previous quarter and 79% as to the peak of the credit boom in 2007Q2. The decrease in lending was across all types of loans (Ivashina and Scharfstein, 2010). However, while syndicated lending decreased, figure 1 shows that the reported commercial and industrial (C&I) loans actually increased by about $100 billion from September to mid-October 2008 (Chari, Christiano, and Kehoe, 2008). This increase was not caused by new C&I loans, but rather by the increase of drawdowns by corporate borrowers on their existing credit lines (prior arrangements to lend to corporations with pre-specified limits and rates) (Ivashina and Scharfstein, 2010).
Fig. 1. Commercial and Industrial (C&I) bank credit in the US (in billion USD). Figure taken from Ivashina and Scharfstein (2010, p. 326.)
These credit line drawdowns created a new kind of bank ‘run’, which happened at the height of the crisis. Unlike previously seen bank runs, brought about by uninsured depositors in search for their money, was this bank run started by short-term creditors, borrowers, and
2 CPP and TARP are used interchangeably, where CPP refers to only the CPP and TARP refers to both TARP and the containing CPP
6
counterparties who drew on their credit lines, due to their concerns about the solvency and liquidity of the banking sector (Ivashina and Scharfstein, 2010). “The market turmoil in the interbank market was not a liquidity problem of the kind that could be alleviated simply by central bank liquidity tools. Rather it was inherently a counterparty risk issue, which linked back to the underlying cause of the financial crisis.” (Taylor, 2009, p. 15).
Therefore, on September 19, 2008 the US Treasury proposed the Troubled Assets Relief Plan (TARP), which contained a plan for the government to purchase illiquid assets from financial institutions. A day later, on 20 September the US treasury announced a draft proposal in which up to $700 billion of ‘troubled assets’ would be purchased. Table 1 provides the timeline of all the important announcement and events surrounding TARP and CPP. TARP was initially rejected on 29 September, but after some modifications4 it was eventually passed on October 3, 2008. TARP originally envisioned the sale of ‘troubled assets’, such as, mortgage-backed securities and troubled mortgages to improve financial stability. However, the week after the approval of TARP the U.S. stock market had a negative return of 18%. This steep decline on the stock market was most probably caused by the lack of clarity of TARP. A survey by Securities Industry and Financial Markets Association (SIFMA) revealed that 94% of banks and securities firms found that TARP lacked clarity about its future operations (Taylor, 2009).
Table 1
Timeline of the events surrounding TARP and CPP.
September 19, 2008: The Bush administration began discussions with Congress for a plan to rescue banks and
financial institutions. In which, the government would buy up a large share of the US mortgage market.The Treasury secretary, Henry M. Paulson provided information that the amount could easily be $500 billion, while experts expected that the bill could exceed $1 trillion.
September 20, 2008: US Treasury announced the draft proposals of TARP to purchase up to US$700 billion of
‘troubled assets’.
September 29, 2008: US House of Representatives voted against an earlier version of TARP. After the vote,
the Dow Jones Index dropped over 777 points in a single day.
October 3, 2008: President Bush signs into law the Emergency Economic Stabilization Act (EESA) of 2008,
which enables the $700 billion of TARP.
October 13, 2008: US Treasury interim assistant secretary Neel Kashkari announces that the funds of EESA
would be used to purchase equity instead of previous plans to purchase troubled from financial institutions.
October 14, 2008: The US Treasury announces CPP, which allows US financial institutions to apply for capital
infusions by the US Treasury to be traded for preferred stock. Ten large financial institutions announced they will enter in the CPP for a total amount of $125 billion. The FDIC announces a new Temporary Liquidity Guarantee Program, which guarantees the senior debt and deposits of all FDIC-insured institutions through June 30, 2009.
The initial set up of TARP was eventually abandoned, on October 13, 2008, when the US treasury announced that it would invest in equity of financial institutions instead of the
4 The plan was adjusted to expand the bank deposit guarantees to $250,000 and to include tax breaks for an amount of $100 billion for businesses and alternative energy.
7
purchase of troubled assets, and led to creation of the CPP, on October 14, 2008. The aim of the CPP was to “strengthen the capital base of the financially sound banks” by providing the means for these banks for extra liquidity so that banks could “increase the flow of financing to US businesses and consumers and to support the US economy” (US Department of Treasury (Oct. 14, 2008)). The Treasury intended to use CPP funds to help temporarily unhealthy banks out of a period of financial distress by injecting capital into these banks to stimulate lending and restore the credit flow in the economy. The CPP contained $250 billion for purchases of preferred equity of US financial institutions. Of this $250 billion, $125 billion was reserved for the 10 largest financial institutions5. It was detrimental that these 10
largest financial institutions would participate voluntarily, therefore US Treasury Secretary Hank Paulson made an agreement with these financial institutions before announcing the CPP. In addition to the involuntary participation, the initial contracts also prevented these financial institutions from exiting the program for three years. The other $125 billion was made available for smaller banks, which could undergo a formal evaluation process in which the banks would be tested if they were viable enough to be selected for the government funds.
Under the CPP, the US Treasury would purchase non-voting senior preferred stock between 1%-3% of the risk-weighted assets up to $25 billion depending on the application of the qualifying financial institutions (QFIs) and the approval of the US Treasury. QFIs are bank, savings and loan, and other financial holding companies as well as insured depository institutions that are operating and established in the US and that are not controlled by a foreign company or bank. The dividend on the preferred equity was 5% for the first five years after which it raised to 9%. In addition, the US Treasury received ten-year life warrants to purchase common stock of 15% of the value of the CPP infusion. The strike price of these warrants was determined by average price of the stock during the last 20 working days before the money was invested. CPP infusions were in preferred stocks to be more attractive for banks, because they were non-dilutive to common shareholders. Until the preferred shares from the CCP infusions were repaid banks were restricted to increase dividend on common shares.
Figure 1 shows that the increase of C&I lending, created by the credit-line drawdowns of corporate borrowers, leveled off just after October 14, 2008. Figure 2 shows the Libor-OIS spread from September to mid-November 2008. Both figures show that CPP managed to
5 The list of 10 largest banks included Citigroup, JP Morgan, Bank of America, Merrill Lynch, Goldman Sachs,
8
Fig. 2. Libor-OIS spread surrounding TARP. Figure taken from Taylor (2009, p. 24). alleviate the ‘panic’ surrounding the financial markets. Li (2013) shows that the CPP managed to significantly boost loan supply and created safer banks during the 2008-2009 crisis, because one third of TARP funds was used for expanding loan supply and two third was used to strengthen their capital base. Also firms benefitted significantly from the CPP infusion in their banks. Especially, riskier and more bank-dependent firms as well as firms borrowing from smaller and less capitalized banks benefitted from the CPP infusions in their banks (Norden, Roosenboom and Wang, 2013). These firms benefitted more, because their banks were profitable, but had a shortage of liquidity. In general larger banks that posed greater systemic risk were approved for CPP injections. On the other hand, these banks had significantly stronger asset quality than the banks that were not approved for CPP injections. This implies that even though systemic riskier banks were aided this was not done on the level of troubled assets. (Bayazitova and Shivdasani, 2012). By providing liquidity for economically healthy banks TARP recipients ended up receiving competitive advantages and increased both their market share and market power relative to non-TARP recipients (Berger and Roman, Forthcoming). Non-TARP participants were either strong banks opting out of participating in CPP. Suggesting that better-performing banks perceived the CPP infusions to be relatively costly (Bayazitova and Shivdasani, 2012). Or weak banks, which were rejected from participating in the CPP due to the low probability of repayment of the government funds. Non-CPP banks experienced a decline in their capitalization when the CPP infusions were performed (Duchin and Sosyura, 2014). Leading to believe that either, investors could not distinguish between banks who opted out of CPP or, were rejected or investors believed participation in the CPP to be a profitable investment no matter what state the bank was in. All in all, TARP led to improvements in economic conditions in the local economic markets
9
in which a higher proportion of banks received TARP funds. It increased job creation, hiring establishments, and decreased personal bankruptcies (Berger and Roman, 2014).
In total CPP resulted in an efficiency gain of $86-$109 billion for the entire economy, however CPP ‘gifted’ even more to the ten largest banks, which received $130 billion. This difference was paid by the taxpayers, who lost $21-$44 billion due to the CPP. The three former investment banks (Goldman Sachs, Morgan Stanley and Merrill Lynch) and Citigroup were the largest beneficiaries of the intervention, while the loser was JP Morgan (Veronesi and Zingales, 2010). CPP was very successful in helping profitable smaller banks with liquidity shortage. However, larger banks used the CPP infusion to increase their portfolio risk. The average risk of loan originated after the CPP infusions at large CPP participating banks increased relative to non-CPP banks through 2009. On the same time, the average risk of the portfolios of small banks CPP banks decreased relative to non-CPP banks. Evidence suggests that the effect of CPP continued even after CPP funds were repaid (Black and Hazelwood, 2013). This can be explained by the timing of the CPP infusions. TARP funds were made available to increase bank stability and reduce incentives to take excessive risks. At the same time, the CPP funds were provided during a period with increased risk with the understanding that the funds would be used to expand lending (Black and Hazelwood, 2013).
The opinion of the public and media was not aligned with the results of the CPP. Often the CPP infusions were characterized as a ‘government bailout’ and a waste of taxpayers’ money used to safe relatively weaker banks. Especially, the government bailout term was popular, because the public knew that the Treasury Department would not let giant banks fail due to the associated systemic risk (Cornett, Li, and Tehranian, 2013; Li, 2013). To a certain point these accusations were correct because, even though the banks were not
entirely nationalized, the size of the CPP infusions were so large that they had an effect on the risk profile of these banks during the crisis. Furthermore, it increased the likelihood of these banks being bailed out when future losses occurred (Black and Hazelwood, 2013). It being less costly to aid these banks more than to lose the investment.
The distortion between the opinion of the public, the media and the actual intention of CPP was caused by a signaling problem. The signaling problem was probably predominantly caused by the lack of clarity of the evaluation process of TARP. There were only two
stipulations on whether banks could qualify for TARP funding: (1) banks were healthy as determined by their regulators, and (2) dividends paid on common stocks and compensation
10
packages for bank executives must satisfy certain conditions. These guidelines were neither very specific nor detailed. For outsiders, it was difficult, if not impossible, to tell if a bank could receive TARP funds based on these guidelines (Li, 2013). For instance, to avoid signaling that specific giant banks were weaker than others, the Treasury made capital injection into all of the top 10 largest U.S. banks under TARP (Li, 2013). On the other hand, the Treasury Department tried to signal the necessity and efficacy of CPP with a voluntary participation of a large targeted fraction of banks, which led to participation of healthy banks. The confidentiality of rejection was in order to prevent bank runs (Duchin and Sosyura, 2014). These different and sometimes contradictory signals led to the vast majority of articles in the Wall Street Journal about the CPP or its participant banks, after the CPP was initiated, to have a negative or pessimistic tone. This negative media coverage is associated with the significant undervaluation of the CPP banks during the CPP initiation period (Ng, Vasvari and Wittenberg-Moerman, 2011). The CPP bank portfolio underperformed the non-CPP bank portfolio during the CPP initiation period (2008Q4-2009Q1). In the post-CPP initiation period (2009Q1-2009Q4) the CPP bank portfolio was adjusted upwards (relative to the non-CPP bank portfolio). While non-CPP banks had a stronger performance compared to non-non-CPP banks both prior to and during the CPP initiation period (Ng, Vasvari and Wittenberg-Moerman, 2011). The results and perception of the CPP were not similar in the preceding quarters of its initiation for the banks, investors and the public and media. While banks in general profited from the CPP, investors did not acknowledge this profit until later. The media and public remained longer pessimistic about the CPP. Was the intervention poorly designed or was the view of investors and the public and media too pessimistic?
3. Related literature and hypotheses
Phillipon and Schnabl (2012) conclude that, when a bank is in need of liquidity, the optimal form of intervention is recapitalization, because an increase in debt would add to the debt overhang, thereby creating new problems. However, the incentive of bank restructuring measures must be to liquidate or restructure bad loans. This will significantly accelerate the recovery from a recessions caused by a banking crisis. Other interventions, such as liquidity support and guarantees on bank liabilities are only practical to prevent bank failures, because they enable zombie bank. These zombie banks will take higher risks by holding on to bad loans, hoping that those loans will repay with some small probability. The upside of the risk
11
is more beneficiary to them than the downside, because if the loans fail these banks receive more government aid and when these bad loans pay out the profit goes to the banks. Holding on to bad loans is apparently a value-destroying decision (Timotej and van Wijnbergen, 2014). More important, Timotej and van Wijnbergen (2014) show that bank recapitalizations have a large, positive, and significant effect on the probability of economic recovery and leads to substantially shorter periods of output loss than other ways of supporting banks in distress. Bank recapitalizations also increase lending (Calomiris et al., 2013). When during a banking crisis bank recapitalizations are done, firms more dependent on external finance grow faster (Leaven and Valencia, 2011). While most interventions have been used aggressively in economy crisis of advanced countries, actual bank restructuring is usually resorted to relatively late (Leaven and Valencia, 2013). Probably, because recapitalization are in general more expensive than other interventions (Leaven and Valencia, 2008). “Above all, speed appears of the essence. As soon as a large part of the financial system is deemed insolvent and has reached systemic crisis proportions, bank losses should be recognized, the scale of the problem should be established, and steps should be taken to ensure that financial institutions are adequately capitalized.” (Leaven and Valencia, 2012, p. 30). Furthermore, to minimize costs of the intervention the recapitalization should be an equity injection and in the form of preferred stock. Also, warrants must be used to make the offer unattractive for banks that are not in need of the capital (Phillipon and Schnabl, 2012).
The CPP program followed all these guidelines, which resulted in significant positive abnormal returns on October 14, 2008 not only for the initial 10 participants of the CCP, but also for the later CPP participants. For the initial 10 participants it was known that they would receive CPP infusions, but not for the later participants. It might be either, investors anticipated which banks would receiveCPP infusions or, that the CPP announcement had a positive effect on allequity valuations of banks by reducing systemic risk in the financial system (Bayazitova and Shivdasani, 2012). However, researching the medium-run Ng, Vasvari and Wittenberg-Moerman (2011) find that the CPP bank portfolio underperformed the non-CPP bank portfolio during the CPP initiation period (2008Q4-2009Q1). While Demirguc-Kunt et al. (2013) find that in 2008Q4 Tier 1 capital was affecting stock returns of banks particularly strong. Duchin and Sosyura (2014) find that non-CPP participants
experienced a decline in their capitalization at the time of the CPP infusions.
These conflicting research findings are reason to believe that the lack of clarity perceived by investors about the CPP program might be shown in the stock movement of the
12
banks. Fratzscher (2008) shows that independent of monetary policy and the use of actual interventions communication is an effective tool in influencing exchange rates over the medium-run. Through the signaling channel authorities can signal actual intervention or future policy changes to the public (i.e. investors) in the form of oral interventions. Fratzscher (2008) finds that oral interventions succeeded in influencing the exchange rate in the desired direction in over 75% of the researched cases and was even more successful in influences the exchange rate than actual interventions. Fratzscher (2008) researches the effect of oral
intervention on the exchange rate. However, extending this theory onto the effect of oral interventions on stock movements seems not far-fetched. For example, recently the power of oral intervention was show by the President of the European Central Bank (ECB) Mario Draghi when he pledged that he was “ready to do whatever it takes” to preserve to Euro after a steep rise of interest rates on government bond-yields in several European countries. Figure 3 shows the reaction of investors after the statement of Draghi, it also shows the initiation of the Outright Monetary Transactions (OMT) program, which objective was to purchase government bonds to lower bond yields of certain Euro member states in difficulty. Both interventions (oral and actual) were able to decrease bond-yields in Spain and Italy.
Fig. 3. Announcement of ECB President Draghi
This gives the incentive to research the effect of the oral intervention of the CPP, the CPP announcement. Is the short-run stock movement on the CPP announcement day
(October 14, 2008) comparable to the medium-run CPP initiation period (2008Q4-2009Q1), and the long-run? Was it clear to investors, which banks would be able to benefit from the CPP? First we formulate our hypotheses:
Hypothesis 1: Did the CPP announcement have a significant impact on bank's market
13
As in Acharya (1988), Peterson (1989) and more recently Yorulmazer (2009), and
Bayazitova and Shivdasani (2012) the impact of the CPP announcement on bank’s market valuation is measured using the abnormal returns of the banks on the CPP announcement day. A more detailed description of the research method will be introduced in the methodology section. Bayazitova and Shivdasani (2012) already researched the returns on the CPP announcement day and found these to be abnormal and positive, which we will accept them to be. The abnormal returns are also used for the second part of the research, which is formulated in the second hypothesis:
Hypothesis 2: Which banks did benefit more from the CPP announcement?
The bank’s balance sheet fundamentals, performing ratios, and risk measures are regressed on the abnormal returns. In order to research if the oral intervention (CPP announcement) was as clear as it should be. If investors were able to determine, which banks would be able to participate in the CPP and benefit from it. Bayazitova and Shivdasani (2012) find that not only the initial 10 recipients of the CCP infusions, but also the later CPP recipients experienced a significant abnormal return, while it was not yet known that these financial institutions would receive CPP infusions. One explanation could be that the oral intervention was clear. It could also be that the oral intervention was not entirely clear, but that the CPP announcement improvedequity valuations of all banks by lowering systemic risk in the financial system. As Bayazitova and Shivdasani (2012, p. 378) put it “a government intervention can increase the market’s expectation of future interventions and regulatory seizures, potentially impeding financial sector recovery”
4. Variables
To measure the effect of the CPP announcement on bank’s valuation, fundamentals of the bank’s balance sheet, performing ratios, and risk measures are regressed on the abnormal returns of the banks on the CPP announcement day. First, the fundamentals of the bank’s balance sheet used for the regression, are explained. After which, the performing ratios will be discussed. Finally, we will look at different risk measurements to determine banks’ individual risk and systemic risk and which measurement was most applicable to determine abnormal returns of the banks at the CPP announcement.
14
4.1 Balance sheet variables
Capital is expected to be an important factor in explaining the abnormal returns of the banks on the CPP announcement day. However, it depends on the level of capital what the
correlation with abnormal returns are expected to be. Additional capital or government support results in different actions by banks with different capitalization levels. Banks, which are poorly capitalized and receive additional capital, still tend to choose a very risky
portfolio, because the upside of the risk is still more valuable to them then the downside. Only high returns will bring their capitalization levels back to normal. Banks, which have a slightly below average capitalization return to being safe due to the addition of the additional capital. Therefore, they tend to reduce the risk of their portfolio, because the high returns are not detrimental anymore. Banks so well capitalized that insolvency is remote will perceive the additional capital as a cheap source of funding, which can be used to add on more risk to benefit from the upside. This model from Calem and Rob (1999) sees the relationship between bank capital and risk as U-shaped. Figure 4 draws a crudely picture of the Calem and Rob (1999) model. The model is not perfectly U-shaped as banks with low capitalization only survival strategy is to take on more risk and thus take the most risk as possible, whereas well capitalized banks only selectively add on more risk with the additional capital. This additional risk rises as the capitalization improves, but not as steep a left or low capitalization side of the U-shaped model.
Fig 4. U-shaped model crudely drawn from the Calem and Rob (1999) paper
Berger et al. (2014) research banking crisis in Germany from 1999-2009 and find the effect from the Calem and Rob (1999) model, that when capital support is distributed
correctly it succeeds in reducing bank risk taking. They suggest that the disciplinary effect of capital support may have caused a decline in lending, but the higher capital may have helped these banks to attract more deposits. However, CPP funds were available for all banks which
0 50
100 U-shaped model
Risk related to capitalization
15
were viable. Thus, certain banks, which were in the right side of the U-shaped model, could also apply for CPP funds. This effect is shown by Duchin and Sosyura (2014). They find that, CPP participation did not lead to an increase in credit origination, but it did result in a shift in credit origination towards riskier mortgages. However, the increased risk taking was not followed by superior risk-adjusted returns. Cornett et al. (2013) explain that healthy banks with temporary financial problems are more likely to use CPP funds as a cheap source of financing. Black and Hazelwood (2013) examine all banks and find that large banks converted their CPP funds it into riskier loans (greater loan infusion, higher risk), while medium and small CPP banks did need the capital (greater loan infusion, lower risk). The same effects as the U-shaped model as described by Calem and Rob (1999). CPP rejected all non-viable banks for the CPP funds, therefore banks in the left side of the U-shape should be not included in the CPP. Thus, the level of capital is not expected to correlate linear with abnormal returns. Capital is measured using the Tier 1 ratio (regulatory capital/risk-weighted assets), which is the most common metric to measure a banks’ capitalization. The tier 1 ratio is expected to capture the U-shaped model of Calem and Rob (1999), thus that banks with a low tier 1 ratio will not benefit from the CPP, because they are not allowed to enter the program. Banks with an intermediate tier 1 ratio will benefit more from the CPP
announcement, because this includes the banks in need for additional capital, which were still viable. Banks with a higher tier 1 ratio should benefit less from the CPP announcement, because they will use the additional capital to add risk. This is measured by the dummy variables High Tier 1 ratio and Low Tier 1 ratio, which is one when the tier 1 ratio of the bank is in the 5th quintile or in the 1st quintile. The second capital ratio used is the tangible common equity ratio (tangible common equity/tangible assets). Tangible common equity is measured by deducting the intangible assets, goodwill, and preferred stocks from the stockholders’ equity.CPP infusions were in the form of preferred stock. Therefore, if the valuation of the bank decreased the common equity of the bank would absorb the losses, but any decrease in valuation after that would lead to losses in the value of the government’s preferred equity claim. Thus, the higher the common equity a bank has the less chance the government would seize a bank if losses became too large. This could be the explanation why even after the CPP infusions bank valuations continued to decline, because ongoing losses diminished the common equity capital of banks (Bayazitova and Shivdasani, 2012). The higher the tangible equity ratio the less chance it would be seized by the government if losses in valuation continued, therefore (in combination with other independent variables) it would lead to higher abnormal returns.
16
Cornett et al. (2013) states that CPP participating banks with less capital, less
liquidity, and weaker loan portfolios were the first banks not being able to pay back the CPP funds. Demirguc-Kunt et al. (2013) show that banks with more liquidity after the fall of Lehman have a significantly lower stock market performance. Therefore, the liquid assets ratio (liquid assets/total assets) is researched, because a bank with more liquid assets has less need for extra capital, thus will benefit less from the CPP announcement. To determine the effect of the loan portfolio the total loans ratio (total loans/total assets) is researched6. In
general, a higher ratio of total loans on total assets would predict less liquidity and therefore would lead to higher abnormal returns. However, this variable might be affected by the quality of the loans. The last variable used from the asset side of the balance sheet is total assets (ln(total assets)). On the day of the CPP announcement, it was also announced that the 10 largest banks were obligated to participate in CPP. Also, in general larger banks that posed greater systemic risk were approved for CPP injections (Bayazitova and Shivdasani, 2012). Thus, it is expected that size will have a positive correlation with abnormal returns.
Next, we turn our focus to the liability side of the balance sheet. Banks with more deposit financing decreased their syndicated lending with a lower amount than other banks (Ivashina and Scharfstein, 2010). Cornett et al. (2013) witness the same effect of deposit financing. The loss of liquidity either from low amount of liquid assets, decreasing core deposits, or drawdowns of loan commitments diminish the bank’s ability to continue lending. Thus, TARP injections allow this group to continue successful lending. Demirguc-Kunt et al. (2013) find a significant positive correlation between deposits and stock market return during the crisis. Therefore, we investigate the effect of the deposits ratio (total deposits/total liabilities) on the abnormal returns. Again, as liquidity we expect for lower levels of deposits to result in higher abnormal returns, thus have a negative correlation. Finally, we look at the non-deposit, short term funding of banks, the wholesale ratio (non-deposit, short term liabilities/total liabilities). Bayazitova and Shivdasani (2012) find that banks with higher wholesale funding were more likely to participate in CPP. Yorulmazer (2009) investigates the bank run on Northern Rock and find that banks that relied more on wholesale funding were affected more by the run on Northern Rock and the following bailout. Thus, wholesale is expected to correlate positively with abnormal returns. These banks are more dependent on
6 Unfortunately no extra information is provided about the contents of the loan portfolios and therefore are not included as independent variables
17
the solvability of the financial system as most of their financing comes from other banks in comparison to other banks, which finance themselves with deposits.
4.2 Performing ratios
The performing ratios of banks are indicators about the viability of these banks. First, we investigate the income ratios, which determine a bank’s ability to make profit and determines where these profits come from. Return on assets (ROA) (total income/total assets) determines the ability of a bank to earn profits relatively to its size. Therefore, better performing banks have a higher ROA and will have less need for CPP injections, thus the correlation with abnormal return is expected to be negative. Furthermore, we define were the income of the bank comes from with the total non-interest income ratio (non-interest income/total income) of the banks. A higher percentage of non-interest income is explained by a higher
dependency on non-traditional income earning products (i.e. restructured loans, derivatives, etc.), which were during the financial crisis in general riskier. Also, most losses experienced by banks in the period before the CPP were in the non-interest income. Therefore, a higher non-interest income is expected correlate positively with abnormal returns as these banks needed extra capital to cover risks from these non-traditional products. Next, we focus on ratios which indicate the asset quality and potential losses of the banks, which in turn predict the future viability of these banks. Especially, when there profit is not sufficient enough to cover the risks of low asset quality and potential future losses. These ratios all have the same effect; the larger these ratios are the riskier the portfolios of these banks are, and the more need these banks have for the CPP infusions to cover these risks. Thus, these ratios are
assumed to have a positive correlation with abnormal returns. However, riskier portfolios of a bank could indicate lower future viability of these banks, therefore the higher the chance that they will be turned down for the CPP infusions. If the requirements for banks to participate in the CPP were clear for investors then investors would know that the higher these ratios were, the smaller the chance that these banks would be able to participate in the CPP. Thus clarity about the requirements of the CPP could be indicated by a lack of correlation, because the higher the ratio the more need for additional capital, but also the higher the chance for decline of participation. Lack of clarity produces a positive correlation indicating that CPP will cause a decrease of systemic risk in the financial system or a ‘bailout’ of the entire banking system. These performing ratios are: non-performing assets ratio (non-performing assets/total loans), loan loss provisions ratio (loan loss provisions/total loans), andnet charge-offs ratio (net
18
charge-offs/total loans). However, net charge-offs are seen as losses and/or expenses and their values are therefore in general negative. Thus, the expected correlation of net-charge offs with abnormal returns is negative.
4.3 Risk measures
Risk measures are often researched in recent time. Still, many researched show different kind of risk measures to be of importance. Zhou (2010) concludes that the systemic impact of a bank is not correlated with size measures. Therefore, ‘too big to fail’ is not valid. Although the too big to fail measure was true from 1994-1999. Acharya and Yorulmazer (2007) their main result is that small banks have incentives to not differentiate themselves from the big banks. If a small bank fails alone it will not be saved or aided by the government, while if a large and or systemic bank fails it will be saved more often due to the effects on the financial system. However, if the small banks fail together with the large banks, as in the recent crisis, the government will aid both. This problem is called the too-many-to-fail problem, which is different from the too-big-to-fail and it affects small banks more. The problem with current financial regulations, such as Basel I and Basel II, is that they are designed to limit the risk of each financial institution. However, the rationale of limiting the risk of each financial
institution is to prevent systemic risk, while systemic risk is not controlled for by these regulations (Acharya et al., 2010). As Wagner (2010, pp. 373-374) explains “Bank assets carry idiosyncratic risk. Thus, diversifying into the assets of the other bank reduces the likelihood of a bank’s portfolio value dropping below its liabilities. This lowers a bank’s probability of failure, but also makes the banks more similar to each other. Therefore, increasing the total systemic risk i.e., when two undiversified banks fail simultaneously, diversification has no effect. While, if one bank fails and other survives by a margin, diversification could let them fail together”. Thus, the design of bank closure policy and capital adequacy requirements, which are based only on individual bank risk, could be less preferable in a financial system with multiple banks (Acharya, 2009).
These shortcomings of regulation on systemic risk is shown by the difference between a bank failure occurring in a well-capitalized system, which has no or little effect on the economy, while the failures of Bear Stearns and Lehman Brothers almost caused a collapse of the financial system (Acharya et al., 2010). The Basel Committee on Banking Supervision (2009) has recognized that under the standards previously to the financial crisis some banks
19
were able to show strong capitalization while holding a limited amount of tangible common equity. As previously explained, this is the component of capital of banks, which shows the ability to absorb losses (Demirguc-Kunt et al., 2013). Investors might have been informed about the inadequacy of the capital adequacy requirements during the financial crisis.
Leverage ratio, which is an easier measure to indicate the banks free capital to absorb losses, was a better guidance in determining stock movements than the risk-adjusted Basel ratio during the financial crisis. Also, Tier 1 capital was seen as an important notion of capital especially for larger banks (Demirguc-Kunt et al., 2013). This leads to believe that it was not only hard to determine the best measure for risk and systemic risk, it was also not clear which measure of risk was perceived best by investors. The addition of several risk measures will be done in order to test, which risk measure was perceived best during the financial.
First, the risk measures used to determine if banks were viable enough to participate in the CPP are investigated. These measures are called CAMELS, where every letter stands for a risk measure. CAMELS stands for: Capital Adequacy (equity capital/gross total assets) it measures the extent in which a bank can absorb potential losses and increase lending and
commitments (negative correlation with abnormal returns), Asset Quality (non-performing assets/total loans) to account for the overall condition of a bank’s portfolio (positive correlation with abnormal returns), Management Quality is the age of the bank (negative correlation with abnormal returns), Earnings is return on equity (ROE), because banks that are more profitable are in better positions to lend and improve local economic conditions (negative correlation with abnormal returns), Liquidity (cash/total deposits) the ratio of free cash in relation to the deposits outstanding (negative correlation with abnormal returns), and Sensitivity to Market Risk (total loans/total deposits) the relation between outstanding long-term assets and long-term outstanding liabilities (positive correlation with abnormal returns). Next, we add the risk-adjusted Basel ratio (Tier 1 capital + Tier 2 capital)/(risk-weighted assets), and the leverage ratio (regulatory capital/total assets). Demirguc-Kunt et al. (2013) find that for larger banks the leverage ratio was seen as the best proxy for investors to determine if banks were ‘healthy’. Both ratios are indicators for the available capital a bank has, thus the correlation with abnormal returns is expected to be negative.
Systemic risk is the firm’s overall contribution to system wide failure (Acharya et al., 2012). To determine if systemic risk was an important factor in explaining the abnormal returns at the CPP announcement, at least more important than the other, more frequently used, risk measures, SRISK is introduced. SRISK measures the capital shortfall (i.e. the
20
capital that a firm is expected to lose) if another financial crisis would occur. The measure is lagged, because it measures what the capital shortfall is at the measured period if in the previous period a financial crisis occurred.
SRISKi,t = Et −1
(
Capital Shortfalli│Crisis)SRISK can be used as an alternative to the much criticized Basel risk weights, because expected capital shortfall captures the effects of size, leverage, and interconnectedness of a bank in a single measure. Firms with systemically riskier assets and higher leverage will have a larger expected capital shortfall, which results in a higher SRISK measure and therefore, banks must hold higher amounts of capital (Acharya et al., 2012). However, the SRISK measure provides more information than the Basel risk-weights, because the expected capital shortfall also measures the co-movement of the financial firm’s assets with the rest of the financial system in a crisis (Acharya et al., 2012). Acharya et al. (2010) show that the SRISK measure had the ability to forecast the outcome of the stress test (SCAP) and the equity performance during the financial crisis, which we hoping to see as well in our research. The higher the SRISK measure the more systemic risk a bank has thus the lower the chance it has to participate in the CPP. What should lead to lower abnormal returns. However, if controlled for institutional risk (i.e. Basel risk measures) the SRISK should have positive correlation with abnormal returns. Bayazitova and Shivdasani (2012) find that in general larger banks that posed greater systemic risk were approved for CPP injections. Table 2 provides a
Table 2
Balance sheet variables Performance ratio Risk measures Sign of cor.
Tier 1 ratio Not linear
High Tier 1 ratio ROA Capital Adequacy
Negative Low Tier 1 ratio Net Charge-Offs ratio Management Quality
Liquid Assets ratio Earnings
Total Deposits ratio Liquidity
risk-adjusted Basel ratio leverage ratio
SRISK Tangible common equity ratio Non-interest Income ratio Asset Quality
Positive Total Loans ratio Non-performing Assets ratio Sensitivity to Market Risk
Total assets Loan Loss Provisions ratio Wholesale ratio
21
summarization of the expected signs of the correlation of the independent variable with abnormal returns.
5. Methodology
In order to research the effect of the CPP announcement we have to take into account the unusual state the US economy was in that period in time. So many government programs and market events occurred around the same time period, which makes any improvement in the short-run overall stability of the financial system is hard to document (Berger and Roman, 2014). Therefore we research what investors perceived to be the consequents of the CPP announcement to banks. Even correctly measuring the investors’ reaction to the CPP announcement is not easy. Again, the number government programs and market events occurring around the same time brings the need for prudence. Normally using a longer event horizon will present more data and clearer results, however it will include more noise of other events.
First of all, we need to make sure that the CPP announcement day is free of noise when determining the event horizon. Veronesi and Zingales (2010) researched the other initiatives that were also announced on October 14, 2008, such as the bank debt guarantee program and the extension of FDIC insurance. They estimate the effect of these other announcement to be relatively modest. Therefore we can conclude that the CPP
announcement day is unaffected. However, does the CPP announcement day contain all the information about the view of investors of the effect of CPP? Table 1 shows that TARP and CPP weren’t enacted immediately. It took some time create a government enacted measure, which was preferable to all parties. So investors were informed before the CPP
announcement day that a government plan containing $700 billion was about to be enacted to aid the banks. However, looking closely at the time line of events preceding the CPP
announcement no event discusses the way the funds would be used and the exact amount of funds as eventually enacted by CPP, other than the announcement of Neel Kashkari on October 13, 2008. The announcement on October 13, 2008 provided investors with the information that the funds would be used to purchase equity instead of buying up troubled assets. Thus, we conclude that most if not all information about the view of investors of the
22
effect of CPP is provided by the banks’ stock movement on the CPP announcement day7. Expected is that, given the rationality in the marketplace, the effect of the CPP announcement will be reflected immediately in stock prices (MacKinlay, 1997).
Now that we have set our event horizon, we continue to determine the data which should be researched. The event horizon is very short, therefore the data of the banks should be as up-to-date as possible. From the COMPUSTAT financials quarterly database data is collected of the third quarter of 2008 (2008Q3), which was the quarter preceding the CPP. Unfortunately, this database only provides data of bank holding companies (BHC). While, CPP was intended to aid any qualified financial institutions (QFIs). QFIs are banks or other financial companies (i.e. insurance companies), which were established and operating in the US, and were not controlled by a foreign company or bank (Li, 2013). No other database available provided either quarterly data or the correct variables in order to add more QFIs to the list obtained from COMPUSTAT. In order to complete the list of the banks we added manually the missing banks from the list of 19 largest banks8, which were first to obtain CPP funds and of which some were obligated to participate or asked to participate (Duchin and Sosyura, 2014). After deletion of banks without enough or the essential data and making sure they were all compliant to be a QFI a list of 424 banks was obtained.
Table 3 provides data about the banks and the variables discussed in the previous section, where Table 3a contains the balance sheet variables. It shows that the mean and median of the Tier 1 ratio of the banks was above 10%, which was well within regulatory requirements. The tangible common equity ratio is on average 3% lower than the Tier 1 ratio and its
minimum and maximum are further apart. From the other balance sheet variables it becomes clear that the distribution of banks is skewed. This could be expected, because there are only a few really large banks and many smaller banks. The large difference between the mean and median shows that in fact there is a larger presence of smaller banks in the dataset. The range of banks included in the dataset is large, with total assets ranging from just over $100 million to more than $2 trillion. The negative mean of net income shows the financial problems the banks were in, while the median of net income still remains positive. Again, showing evidence of a skewed distribution caused by large losses from larger banks, as shown by the
7 In the robustness section we will combine the banks’ stock movement of both days (October 13 and 14, 2008) to test if information was lost by only using the CPP announcement day
8 The list of 19 largest banks included Citigroup, JP Morgan, Bank of America, Merrill Lynch, Goldman Sachs,
Morgan Stanley, State Street, Bank of New York Mellon, Wells Fargo, Wachovia, KeyCorp, Fifth Third Bancorp, Regions Corp., BB&T, Capital One, SunTrust, U.S. Bancorp, American Express, and PNC Financial Services
23
minimum of net income. Table 3b contains the performance ratios. ROA shows the same financial difficulties as net income. The non-interest income has high range, probably caused by the investment banks in the dataset. The other performing ratios are relatively low,
however the maximum of non-performing assets of over 11% shows the potential financial difficulties of certain banks. As mentioned before, net charge-offs are investments or assets a bank sees as lost, therefore these values tend to be negative. The risk measures ratios mean in table 3c are all above regulatory requirements. However, the troubled banks are shown in the minimum of these ratios. Leverage ratio and Tangible common equity ratio seem to be very comparable. The SRISK measure shows the capital needed by banks if another systemic crisis occurs.
ROA is the return on assets measured by the net income/total assets. Non-interest income ratio is non-interest income/total income. Non-performing assets ratio, Loan loss provisions ratio, and net charge-offs ratio are all divided by total loans. Capital adequacy is equity capital/total assets. Risk adjusted Basel ratio is (tier 1 + tier 2 capital)/risk-weighted assets. Leverage ratio is regulatory capital/total assets. SRISK is the amount of capital needed if another systemic crisis occurs.
Table 3a N Mean Median Minimum Maximum
Tier 1 ratio 386 10.68% 10.03% 5.46% 25.65%
Tangible Common Equity ratio 402 7.92% 6.99% -0.4% 29.6% Total Assets (in millions of $) 424 33522.23 1448.06 117.86 2251469.00
Liquid Assets 423 8936.43 253.58 .00 836076.00 Total Loans 423 13722.09 1013.50 58.45 920260.00 Total Deposits 424 13898.12 977.14 .00 969783.00 Total Liabilities 424 31108.38 1287.23 98.73 2105626.00 Wholesale 422 11759.28 105.46 .00 872656.00 Net Income 424 -58.76 1.05 -23698.00 1637.00
Table 3b N Mean Median Minimum Maximum
ROA 424 -0.02% 0.11% -3.89% 0.65%
Non-interest Income ratio 423 19.61% 12.51% -139.0% 3386.1% Non-performing Assets ratio 407 1.93% 1.47% 0.00% 11.84% Loan Loss Provisions ratio 422 0.26% 0.14% -0.01% 3.10%
Net Charge-Offs ratio 401 -0.16% -0.08% -3.05% 0.03%
Table 3c N Mean Median Minimum Maximum
Leverage ratio 424 7.97% 7.02% -0.39% 29.56%
Capital Adequacy ratio 420 9.58% 8.66% 4.1% 29.4%
Risk Adjusted Basel ratio 387 13.55% 12.10% 8.92% 50.96% SRISK (in millions of $) 53 12214.13 141.00 -14832.00 135566.00
24
To answer both hypotheses the abnormal returns of the banks must be calculated. This is done in similar fashion as Acharya (1988) and Peterson (1989), but Yorulmazer (2009) will be used as guideline. We start with the standard market model. An OLS regression, which is
= + +
controlled for autocorrelation and heteroskedasticity, is used to estimate and . For this regression we estimate with T = 254 trading days or T = 365 (normal) days and the window begins 385 days before the event (October 14, 2008). To estimate the Nasdaq index is used, because the majority of the banks in the dataset are listed on the Nasdaq (363 out of the 424 banks total). When rewritten and are used to estimate the abnormal returns. are the daily returns of the banks on the CPP announcement day.
= − +
To test whether the abnormal returns are significantly different from zero the standardized abnormal returns are calculated.
= ~ ( − 2)
is the sample standard deviation of the abnormal returns, which are also obtained using the same OLS regression as used for the estimation of and . Finally the formulas for the cumulative abnormal returns for a period of ≤ ≤ are presented, which will be used in the robustness section.
( , ) =
!
Again, to test whether the abnormal returns are significantly different from zero the standardized abnormal returns are calculated.
= ( , ) ~ ( − 2)( , )
Table 4 provides data containing the raw returns and abnormal returns of the banks. Unexpected, is the low mean of the abnormal returns and that only the mean of the raw returns differs significantly from zero. This result does not coincide with the results of
25
Bayazitova and Shivdasani (2012), who find much larger abnormal returns. However, when we look at the differences between the means of the abnormal returns of the later CPP recipients and non-CPP recipients, we find that later CPP recipients have a significant larger mean of abnormal returns than non-CPP recipients. This result does coincide with the results of Bayazitova and Shivdasani (2012). Out of 424 banks 82 banks experienced a significant positive abnormal return and 94 banks had a significant negative abnormal return.
1 of which 60 with 1% significance level and 22 with 5% significance level. 2 of which 48 with 1% significance level and 46 with 5% significance level. *,**,*** denotes significance respectively at a 10%, 5%, and 1% level, with H0: mean = 0. A,B,C denotes significance respectively at a 10%, 5%, and 1% level, with H0: mean1= mean2.
Figure 5a and 5b show the distribution of the raw returns and the abnormal returns. These figures show that both distributions are somewhat skewed towards the negative (abnormal) returns with some large outliers on the positive side. Making it possible to formulate a preliminary answer to our first hypotheses: did the CPP announcement have a significant impact on bank's market valuation? Overall the CPP announcement does not seem to have a (large) effect on bank valuations and the effect seen is somewhat skewed to the negative returns. However, there are several significant large positive outliers showing that there were banks, which experienced a large valuation increase due to the CPP
announcement.
Fig 5a, b. Distribution of the raw returns and abnormal returns
Table 4 N Mean Std. Deviation # sig. positive # sig. negative Raw Returns 424 .0280*** .0968 Abnormal Returns 424 .0023 .0989 821 942
Abnormal Returns CPP recipients 225 .0093B .1093 49 45 Abnormal Returns non-CPP recipients 199 -.0055 .0853 33 49
26
6. Results
In this section we will start to answer the hypotheses: did the CPP announcement have a significant impact on bank's market valuation and which banks did benefit more from the CPP announcement? Also the more subjective question if the CPP announcement was clear enough for investors to make informed decisions will be investigated. In section 4 many variables are introduced and explained. While, it is clear what effect these variables should have on abnormal returns (see table 2), it remains unclear which is the best set of independent variable to determine the abnormal returns. Many of the variables have an overlapping
prediction value and therefore will correlate with each other. This might result in
multicollinearity or other unwanted effects. Therefore, the independent variables are first examined if they act as predicted in single regressions on abnormal returns. After which, multiple regressions are done containing all the independent variables of the previously created groups; balance sheet variables, performance ratios, and risk measures. In these multiple regressions unwanted effects will occur, which will later be dismissed in order to create the best model with the independent variables which explain the most variance. This model should also be able to tell best what investors believed to be the effect of the CPP announcement for banks. If not, it explains that the CPP announcement was not seen as informative as it should be.
In order to answer these questions we must analyze the regression results of table 5. Like in the previous section we will discuss the different types of independent variables separately and in order. In table 5a the balance sheet variables are regressed one by one on the dependent variable abnormal returns, after which the multiple regression is shown. The high and low tier 1 ratio variables are created to capture the effect of the Calem and Rob (1999) model. These are dummy variables, which are 1 for banks with a tier 1 ratio in the upper 80% (5th quintile) or lower 20% (1st quintile), and zero otherwise. The betas from the independent variables are small in absolute number, because they show what an increase in percentage points of a certain ratio will do to the abnormal returns (which are also in
percentages). The variables in the simple regressions show small effects and only the deposits ratio and low tier 1 ratio variables are significant. The independent variables in the multiple regression have a larger effect on abnormal returns and the sign of the betas sometimes change.
27
The tier 1 ratio has no or little effect on abnormal returns. The effect increases somewhat in the multiple regression. The high tier 1 ratio and the low tier 1 ratio have the opposite effect as expected. The theory predicts that well and underfinanced banks receiving additional capital are likely to take on more risk. Thus, those banks should experience a negative effect from the CPP announcement on abnormal returns. However, the opposite proves to be the case especially in the multiple regression when combined with the tier 1 variable. The high tier 1 ratio has no significance, but the low tier 1 ratio is significant in both the single and the multiple regression. The result from the multiple regression shows the effect of an inverted U-shape, meaning that banks either well or undercapitalized profited more from the CPP announcement than other banks. Again, the lack of significance of all three variables makes it hard to conclude if this effect was the case. Also the tangible common equity has the opposite sign as expected, while also not being significant. The wholesale ratio does not behave as the theory predicts in the simple regression, but this changes in the multiple regression. The liquid assets ratio and the deposits ratio are, besides the low tier 1 ratio, the only variables which are significant in the multiple regression. The liquid assets ratio shows that banks with more liquid assets had significantly lower abnormal returns, as expected, because these banks did not need the liquidity of CPP. Banks with more deposits also experienced significant lower abnormal returns, because these banks still had the ability to continue lending. That is why these banks were seen as viable and in less need for additional capital, and therefore did not profit from the CPP announcement. Both effect show that an increase of one percentage point in these ratios led to lower abnormal returns of 0.15%, which is pretty substantial regarding table 4. A bank with a tier 1 ratio in the 1st quintile, so in the lowest 20%, experienced 0.36% higher abnormal returns.
Table 5b shows the simple and multiple regressions of the performance ratios on abnormal returns. In the simple regression all independent variables except non-interest income show the signs which were expected from the literature. Banks with lower
performance ratios are generally more in need of additional capital, which could be provided by CPP. On the other hand, higher return on assets (ROA) depicts a better performing bank, which is in less need for additional capital, because it has its own means to accumulate capital. The larger values of the betas are explained by the smaller values of the ratios. The value of the non-interest income ratio beta changes in the sign expected from the literature in the multiple regression and becomes slightly significant. However, some of the performing ratios behave differently in the multiple regression probably because these ratios are very
28
This table presents an OLS regression with abnormal returns as dependent variable. Tier 1 ratio is regulatory capital/risk-weighted assets. High tier 1 ratio and Low tier 1 ratio are dummy variables, which are 1 when a bank’s tier 1 ratio is in the 5th quintile or 1st quintile. Tangible common equity ratio is tangible common equity/tangible assets. Liquid assets ratio and total loans ratio are divided by total assets. Total assets is the natural logarithm of total assets. Deposits ratio and wholesale ratio (non-deposits, short term liabilities) are divided by total liabilities. Standard errors are in parentheses.*,**,*** denotes significance respectively at a 10%, 5%, and 1% level.
Dependent variable: Abnormal returns
Table 5a (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Constant .006 .001 -.003 .004 .012 .006 -.012 .060** .001 .163
(.020) (.006) (.006) (.011) (.009) (.026) (.022) (.029) (.007) (.103)
Tier 1 ratio .000 .004
(.002) (.005)
High tier 1 ratio .006 .012
(.013) (.023)
Low tier 1 ratio .028** .036**
(.013) (.018)
Tangible Common Equity ratio -.022 -.212
(.130) (.210)
Liquid Assets ratio -.045 -.168**
(.040) (.073)
Total Loans ratio -.005 -.062
(.036) (.078) Total Assets .002 -.003 (.003) (.004) Deposits ratio -.074** -.125** (.036) (.054) Wholesale ratio .015 -.013 (.041) (.054) N 386 386 386 402 423 423 424 424 422 368 R-squared .000 .001 .013 .000 .003 .000 .001 .010 .000 .049
29
similar, this causes a multicollinearity problem. Multicollinearity causes one of the similar independent variables to overestimate their effect on the dependent variable, while the other variable counter-effects this overestimation. Sometimes causes a variable to change sign or become small in value. Overall the performance ratios, both in the simple as in the multiple regression, do not provide a lot of explanatory power on abnormal returns (in its current form).
This table presents an OLS regression with abnormal returns as dependent variable. ROA is the return on assets measured by the net income/total assets. interest income ratio is non-interest income/total income. Non-performing assets ratio, Loan loss provisions ratio, and net charge-offs ratio are all divided by total loans. Standard errors are in parentheses.*,**,*** denotes significance respectively at a 10%, 5%, and 1% level.
The results of the simple and multiple regressions of risk measures on abnormal returns are shown in table 5c. Excluding liquidity all the risk measure variables show the signs expected from the literature. However, management quality, risk adjusted Basel ratio, and SRISK show really small to no correlation with abnormal returns. The correlation of SRISK with abnormal returns seems small, however the SRISK value is in millions. In other words, an increase in SRISK of a million in USD results in 0.35% increase in abnormal returns on the CPP announcement day, however the effect is not significant. Only asset quality and sensitivity to market risk show some significance. Interesting to see is that the significance of the asset quality variable increases in the multiple regression, while the significance of the non-performing assets ratio, which is the same variable decreased in the multiple regression. The effect Demirguc-Kunt et al. (2013) find in their paper can also been
Dependent variable: Abnormal returns
Table 5b (1) (2) (3) (4) (5) (6)
Constant .002 .003 -.007 -.002 -.003 -.011
(.005) (.005) (.007) (.006) (.006) (.008)
ROA -.652 -1.502
(1.045) (1.550)
Non-interest Income ratio -.002 .039*
(.003) (.020)
Non-performing Assets ratio .490* .491
(.291) (.345)
Loan Loss Provisions ratio 1.887 -1.369
(1.356) (2.912)
Net Charge-Offs ratio -3.103* -1.568
(1.838) (3.261)
N 424 423 407 422 401 392
30
This table presents an OLS regression with abnormal returns as dependent variable. Capital adequacy is equity capital/total assets. Asset quality is non-performing assets/total loans. Management quality is the age of the bank in years. Earnings is net income/shareholders’ equity. Liquidity is cash/total assets. Sensitivity to market risk is total loans/total deposits. Risk adjusted Basel ratio is (tier 1 + tier 2 capital)/ risk-weighted assets. Leverage ratio is regulatory capital/total assets. SRISK is the amount of capital needed if another systemic crisis occurs. Standard errors are in parentheses.*,**,*** denotes significance respectively at a 10%, 5%, and 1% level.
Dependent variable: Abnormal returns
Table 5c (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Constant .011 -.007 .004 .002 .001 -.034* -.002 .003 .043* -.072* (.014) (.007) (.011) (.005) (.005) (.019) (.016) (.011) (.018) (.038) Capital Adequacy -.089 -.383 (.133) (.271) Asset Quality .490* .662** (.291) (.318) Management Quality -.000044 (.000) Earnings -.038 .033 (.077) (.083) Liquidity .055 .236 (.062) (.208)
Sensitivity to Market Risk .036** .054*
(.017) (.030)
Risk Adjusted Basel ratio .000 .002
(.001) (.002) Leverage ratio -.008 .134 (.128) (.320) SRISK .00000035 (.000) N 420 407 134 424 421 423 387 424 53 379 R-squared .001 .007 .001 .001 .002 .010 .000 .000 .009 .027
