Amsterdam Business School
Faculty of Economics & Business
The Troubled Asset Relief Program (TARP)
A Program That Incentivizes Risk?
Master Thesis
Master of Science (Accountancy & Control)
Academic Year 2013 / 2014
Abstract:
The “Troubled Asset Relief Program” (TARP) was released as part of the “Emergency Economic Stabilization Act” in October 2008 to restore liquidity and stability to the financial system of the United States and therefore to prevent further damage to social welfare. As part of the program the US Department of the Treasury provided banks with cash. To protect taxpayers’ money banks participating in the program were subject to certain regulation. Part of this regulation was the prohibition of compensation structures that incentivize unnecessary and excessive risk-taking. This study compares the compensation structures of participating banks to the ones of non-participating banks before, during and after their participation in the program. It furthermore analyzes how the TARP affected the risk profiles of banks. It finds that the risk profiles of participating and non-participating banks aligned during the course of the program. However, this alignment was not due to a change in compensation structure.
Student
Fabian Klein 10604170
Date of Submission 23.06.2014
Supervisor / Co-Assessor Dr. B. Qin / Dr. A. Sikalidis
2
Content
1. Introduction ... 3
1.1 Background ... 3
1.2 Research questions ... 5
1.3 Motivation of this study ... 6
2. Literature review and hypothesis ... 7
2.1 The Troubled Asset Relief Program ... 7
2.1.1 Intentions and provisions of the TARP ... 7
2.1.2 TARP’s conflicting objectives ... 8
2.1.3 The effects of TARP on compensation structure ... 11
2.2 Compensation structure and risk-taking ... 15
2.3 Hypothesis regarding compensation structure and risk-taking ... 16
3. Research Methodology... 20
3.1 Sample selection and descriptive statistics... 20
3.2 Research design ... 25 4. Findings... 31 5. Conclusion ... 36 References... 40 Appendix A ... 43 Appendix B... 44
3
1. Introduction
1.1 BackgroundJust 5 years after the financial crisis broke out the Dow Jones reached its all-time high in the end of 2013 - financial markets are growing, banks pay high bonuses again (Murphy, 2012, p. 16, Appendix B) and there are voices predicting the next bubble to burst. Although the accounting profession and especially the financial sector are still under pressure from constantly changing regulations, the time seems to be right to examine the sustainable effectiveness of regulations implemented during the crisis.
Before, during and after the financial crisis a large number of articles have discussed and opined that standard executive payment arrangements induce motivation to excessively boost short-term results at the expense of long-term value (Bebchuk and Fried, 2010). Furthermore there can be recognized a growing public attention towards that topic. It is argued that certain compensation structures contributed to the crisis (Wall Street Journal, 2009, Murphy 2012, p. 1). Fahlenbrach and Stulz (2011) argue that there was no misalignment of interest between shareholders and managers but a rather short term focus which may have contributed to the crisis.
The recent financial crisis has led to widespread recognition of that fact and additionally led to the general opinion that payment arrangements that reward executives for short-term results can produce incentives to take excessive risks. Recognition of the significance of the problem has generated substantial interest in fixing it. Leading public officials, such as former1 Federal Reserve Chairman Ben Bernanke and Treasury Secretary Timothy Geithner, as well as top business leaders such as Goldman Sachs’s CEO Lloyd Blankfein, have emphasized the importance of avoiding such flawed structures (Bebchuk and Fried, 2010). In the aftermath and during the crisis there have been implemented regulations such as the “Troubled Asset Relief Program (TARP)”. The TARP eliminates incentives to take unnecessary and excessive risk in firms receiving TARP funds. Furthermore it can be expected that not only firms which receive funds from the TARP are mandated to change their incentive policies and that companies will change their policies “voluntarily” due to regrets or to pressure
4 from shareholders. Independent from the motivations a study of Deloitte (2010) has found that compensation arrangements have changed towards long term goals since the outbreak of the financial crisis. The presented study finds that risk profiles of fund-receiving companies and non-fund-receiving companies are aligned during the course of the program. However, this alignment is not due to a change in compensation structure since equity based compensation has rather increased than decreased since the implementation of the TARP.
Following the discussion above there are multiple interesting issues which should be investigated. Van der Stede (2011) examines the impact of the crisis on management accounting research and identifies new challenges and opportunities to conduct research. He states: “This [the crisis] opens up the opportunity to examine the effects of such shifts in pay (on managers’ propensities to take risks […]) in an “event-type” study research design.” (p. 17)
Prior literature has examined the impact of TARP on total compensation. Furthermore conflicting and overlapping objectives within the TARP were identified. This study examines the impact on TARP not only on total compensation but also on compensation structures and thereby helps to resolve the problem of the TARP’s overlapping effects. Furthermore, since the TARP does not influence compensation structures significantly, conclusions could be drawn about the effectiveness of incentivizing management to change compensation structures. The rest of this study is structured as follows: In the remaining of this chapter the research question as well as its motivation shall be outlined. In Chapter 2 the Troubled Asset Relief Program itself and related literature will be introduced. A detailed analysis of primary sources will be added to show which potential effects the TARP has on compensation structures. Chapter 2.2 reviews the relation between compensation structure and risk-taking. By combining the effects of TARP and the relation between compensation structure and risk-taking, 2 hypothesis are developed. Chapter 3 explains the research methodology. In chapter 4 the findings are presented. Finally, in chapter 5 these findings are interpreted and conclusions are drawn.
5
1.2 Research questions
Based on Van der Stede’s call this paper will investigate the effects of TARP in terms of changes in compensation structure and risk profile. It will examine the compensation structures and risk profiles of institutions participating in the program by comparing them to non-participating institutions. The comparison will be made over a time frame of several years. Therefore the observed time frame can be divided into three parts, as can be seen in figure 1.
Figure 1: Research Questions
In order to investigate the TARP as a moderator of the relation between compensation structure and risk profile both periods, the one before (1) and the one after the implementation of TARP (2) have to be investigated. By doing this, it can be made sure that occurring differences result from the TARP. As stated above, during the crisis practitioners have recognized the threat of short term or wrong incentives. According to contingency theory I predict that institutions will adjust their incentives to their own best interest. Therefore it is interesting to compare voluntary to mandatory changes and their effects on risk profiles. Furthermore, since participating institutions were mandated to shift their compensation structure only as long as they participate in the program, the TARP is also an opportunity to examine the behavior of institutions after they resign from the program (3).
Time Enter the TARP Withdrawal from the TARP Compensation structure Risk profile Voluntary vs. Mandatory Change 1 2 3
6 The research questions investigated in this paper therefore can be stated as follows: 1. How was the relation of compensation structure and risk profile of institutions participating in the TARP compared to non-participating ones before they received funds from the TARP?
2. Which impact did the receiving of funds from the TARP have on the above relation?
3. How did institutions adjust their compensation structures and risk profiles after they fully paid back the funds?
1.3 Motivation of this study
Besides the interest of practitioners like Goldman Sachs’s CEO Lloyd Blankfein the crisis gives a lot of opportunities to improve existing or address new research questions. This paper examines the relation between compensation structure and risk-taking behavior in the context of regulation (namely the TARP). Although each of the 3 factors has been subject to research in an extensive way, the combination and relation between them have not been investigated yet. Van der Stede (2011) argues that there is the necessity to analyze the interplay of factors like incentives and risk management rather than to investigate both factors separately (p. 5). Furthermore he finds that the theoretical conclusions drawn from the crisis emphasizes the correctness of the “basics”. He argues that just relying on the “basics” is not sufficient to explain the crisis because problems with any given practice have to be investigated within its context. The regulatory context and the economic climate at a time form an “era” in which certain practices failed. The “basics” do not consider these factors and thus are not sufficient to explain recent events. He therefore calls to investigate phenomena across various fields of studies. This paper aims to answer this call by investigating the effects of the “Troubled Asset Relief Program (TARP)” in the context of an “incentive – risk taking” relationship.
As described in the literature review the TARP is a complex law and affects the banking sector in multiple ways. Black and Hazlewood (2011) identify two conflicting objectives within the TARP. On the one hand the TARP should eliminate excessive risk-taking by imposing risk averse payment structures; on the other hand it promotes risk by improving institutions financials with taxpayers’ money. Although Black and
7 Hazlewood (2011) identify both objectives, they only use the riskiness of loans to measure the risk profile. They do not measure the payment structure and therefore do not examine the first objective “eliminate excessive risk taking”. The research proposed will examine the risk profile as well as the payment structures of institutions participating in the TARP and compare them to these of non-participating institutions. By doing this, theories developed within the TARP related literature can be verified or falsified. Although the relation of compensation structure and risk-taking behavior has been examined, the research conducted might also add valuable knowledge to that field. If the TARP has a significant impact on the compensation structure relative to the peer group, a relative decrease in risk profile is expected. This would confirm the general opinion about compensation structures and risk-taking behavior.
2. Literature review and hypothesis
2.1 The Troubled Asset Relief Program
2.1.1 Intentions and provisions of the TARP
At first glance the combination of the two streams of research mentioned above seems to be problematic because they might initially lead to opposing predictions. TARP related literature finds that institutions participating in TARP increase their risk profile. The well founded general opinion in the other stream of research is that less incentivizing compensation structures, as mandated by the TARP, leads to less risky behavior. Duchin and Sosyura (2012) as well as Black and Hazlewood (2011) identify different overlapping effects of the TARP on risk taking behavior. These overlapping effects may result from the complexity of the law itself. The TARP therefore has to be analyzed in great detail. The TARP effects risk-taking behavior in two ways, through incentive structures as well as through the provision of taxpayers’ money. Because the proposed paper basically examines the relation between compensation structures and risk profiles, a short summary of that relation will be explained. Finally, the hypotheses development links certain issues of the TARP to findings within the “compensation – risk-taking” literature.
By the end of 2007 collateralized debt obligations based on mortgage loans were sold extensively in the booming market. When the underlying loans started to default in 2008 the whole financial industry was about to break down. As part of the
8 “Emergency Economic Stabilization Act” (EESA) the TARP was released in October 2008 as a direct reaction to the subprime crisis. The aim was to stabilize the financial market and thus to loosen bank’s lending behavior. Besides mortgage related assets the program allowed the United States Department of the Treasury to purchase or insure any troubled asset. The definition of troubled assets is held broadly so that the decision if an asset was troubled or not was made by the Department of the Treasury after consultation of representatives of the Federal Reserve System. The TARP consists of 13 separate programs. The “Capital Purchase Program (CPP)” allowed the Department to buy equity from banks. The CPP caused about 40% of the money spent and is the largest of the programs.2 This study examines the effects of the CPP, nevertheless CPP and TARP are used as synonyms in the remainder of this study. In order to participate in this program institutions had to be approved. As part of the tax payer protection, companies that took part in the CPP had to eliminate compensation structures that incentivize excessive risk-taking. The overall budget of the TARP was set to $700 billion and could be released in several steps.3 The monthly report to congress, created by the Department of the Treasury states that up until December 31, 2013 $422.2 billion were disbursed to financial institutions as well as other industries. In the same time $432.8 billion were recollected in cash. Although, as can be seen in Appendix A, not all obligations were disbursed it is likely that taxpayers will benefit from the TARP. Without going too deep into the data at this point, it is worth to be said that approximately 60% of the money was obliged and disbursed to the financial sector. It should be noted that the financial industry as a whole has already overpaid back its obligations.4
2.1.2 TARP’s conflicting objectives
Overall the literature about the TARP and its impact on markets and society is extensive. Nguyen and Enomoto (2009) find that volatility in stock index returns was reduced after October 2008. Veronesi and Zingales (2009) estimated the costs of TARP and find that bondholders were the greatest beneficiaries of the program. As the study
2 Internet source retrieved on 03 Jan. 2014: http://projects.propublica.org/bailout/programs 3 The Troubled Asset Relief Program: Report on Transactions Through December 31, 2008 (2009) Speech of Neel Kashkari (October 2008)
4 Monthly Report to Congress: December 2013 (from January 10, 2014) Troubled Asset Relief Program – U.S. Department of the Treasury (p. 2)
(retrieved 17th January, 2014 :
9 presented focuses rather on a micro than a macro level, the questions addressed and the methods used are too broad to examine the impact of TARP on the relation between compensation structure and risk-taking behavior. There are few authors who examine the impact of the TARP on risk-taking behavior. Duchin and Sosyura as well as Black and Hazlewood have done extensive work in that field during the last couple of years. Overall they find that institutions receiving funds from the TARP tended to take more risk but do not increase the volume of credit origination. With the aim to come up with possible explanations for their findings, they investigate multiple elements and mechanisms of the TARP. Hence, their most important ideas and findings should be presented as a basis for the hypothesis presented in this study.
Duchin and Sosyura (2012) investigate the effects of the TARP on banks by examining the institutions which applied for the CPP. They study the risk-taking behavior of banks that were approved for the program and compare it to banks whose application was denied. As expected they find that federal capital infusions improve the capitalization structure of banks. To examine the riskiness and volume of loans they spread their study over several types of loans and financial assets. Overall they find that banks shift their credit origination towards riskier loans. Furthermore the volume of credit origination does not change significantly relative to non-approved banks. Therefore the increased capitalization did not increase the lending but led to a shift towards riskier loan originations. Furthermore Duchin and Sosyura come up with 3 different explanations that may cause the increased risk.
The first hypothesis – government intervention – states that government imposes certain policies that lead to an increased risk. As mentioned above the TARP was a reaction to the subprime crisis, so the government could have imposed policies to invest into these high risk subprime mortgages. However, Duchin and Sosyura (2012) state that the policies imposed would rather reduce than increase the risk profile of participating banks (p. 4).
The second hypothesis – risk arbitrage – says that certain risky assets were underrated during the crisis. Additional funds from the government allowed banks to exploit that opportunity.
10 The third hypothesis – moral hazard – is based on the institution’s perception by others. As soon as the government buys equity of an institution, this institution has a better reputation in terms of credit worthiness. Firms like AIG or Citigroup were saved by the government multiple times. So other participants in the market might assume that the government will protect institutions in which it has invested and provide them with additional funds if needed. The overall findings of Duchin and Sosyura (2012) do not support the first two hypothesis but the third one. The examination of three different hypothesis shows that there are multiple ideas why the riskiness of loans would increase instead of decrease.
Bayazitova and Shivdasani (2012) find that there is a self-selection phenomenon. That means that banks participating in the CPP differ significantly from banks that do not participate. In this context they highlight the importance of executive compensation. Banks with an annual executive compensation exceeding $500.000 are less likely to participate in the CPP. Bayazitova and Shivdasani (2012) measure compensation as a whole and can therefore not examine the structure of compensation. The presented study will bring deeper insight into this topic and will mitigate bias that might result from self-selection.
Although Black and Hazlewood (2011) state that their findings support the moral hazard hypothesis, they also stress the importance of the imposed changes in compensation structure – which can be interpreted as a government intervention. In their study from 2011 they identify two conflicting objectives of the TARP which represent the tension between different approaches towards the financial crisis. One objective is to increase the bank lending. The other objective is to eliminate excessive risk-taking in order to protect the taxpayers’ money. Infusing money into the financial sector can improve the lending in two ways. First, institutions have more cash and can therefore increase investments; credit supply is increased. Secondly, the bank’s individual balance sheets are improved by additional equity or less troubled assets; therefore their credit-worthiness is improved. Still to increase lending, especially during a financial crisis, always imposes a certain risk. Black and Hazlewood (2011) argue that capital infused by the taxpayers induces the responsibility to increase lending. Therefore, providing governmental cash to banks implicitly increases risk-taking, which is contradictious to the other goal of eliminating excessive risk-taking
11 behavior. Black and Hazlewood (2011) then find that the risk of average loan originations at large TARP banks increased relative to non-TARP banks whereas the average risk at small TARP banks decreased relative to non-TARP banks. One explanation of the different behavior of large and small banks is the existence of two conflicting objectives within the TARP. Thus large TARP banks are under greater pressure (or feel more responsibility) to lend money instead of keeping it to polish their own balance sheets.
Overall it can be said that the TARP had multiple effects on the risk-taking behavior of financial institutions. Different mechanisms seem to enforce opposing or contradictious outcomes. Whereas government interventions should reduce risk-taking by imposing compensation structures that eliminate excessive risk, the provision of taxpayers’ money seems to outweigh this effect which leads to an overall increasing risk profile. This context raises the opportunity as well as the necessity to investigate the effects of compensation structure within the TARP.
2.1.3 The effects of TARP on compensation structure
The as part of the “Emergency Economic Stabilization Act” (EESA) the TARP is a code of approximately 450 pages divided in 46 sections from which 2 potentially affect executive compensation. Section 111 “Executive compensation and corporate governance” addresses executive compensation directly. Section 302 “Special rules for tax treatment of executive compensation of employers participating in the troubled assets relief program” amends the Internal Revenue Code (IRC) and thereby indirectly affects executive compensation. Both sections will be outlined and critically discussed in the following. Because the scientific literature about these effects is blurry, the summary is based on the code itself as well as governmental notices that provide guidance for application of the law. Due to the level difficulty of interpreting coded laws the findings are cross checked with secondary literature. (Shearman and Sterling, 2008)
Section 302 of the EESA amends the sections 162(m) and 280G of the Internal Revenue Code (IRC). Both amendments constraint the deductibility of executive remuneration from tax basis. The amendment to section 280G of the IRC disallows deductions for parachute payments for TARP participants. Section 162 of the IRC
12 limits the tax deductibility of executive compensation to $1 million for each covered employee in the taxable year. The definition of “taxable year” in combination with the TARP participation is a crucial aspect of this study and will be discussed in the methodology part. Notably, this limitation excludes “qualified performance-based compensation” which is compensation based upon the attainment of a preset objective performance measure. Furthermore it also excludes realized gains upon exercised stock options or stock appreciation rights (IRC 162(m)(4)(B)&(C); Shearman and Sterling, 2008). This preferred treatment of performance and equity based compensation gives incentives to pay executives in equity when their salary exceeds $1 million.
The amendments to this section firstly reduce the threshold of deductible executive remuneration to $500,000. Secondly, the total remuneration qualifies for the threshold. Therefore performance and equity based compensation is taken into account when calculating the deductible amount.5 These two changes abolish the incentives initially given by section 162 IRC to pay executives high amounts of equity and performance based compensation. The magnitude of this change might be compromised since it only affects companies paying a yearly compensation that exceeds $500,000. Furthermore the maximum costs of this change is the reduction in the threshold multiplied by the applicable tax rate. The effect on compensation might also be reduced by the fact that the tax base cannot be reduced to less than 0. Financial institutions facing a financial crisis are likely to have operating losses and can therefore not benefit from a reduction in their tax base. Nevertheless, prior literature suggests that the amendments do have an effect. As mentioned in the previous chapter, Bayazitova and Shivdasani (2012) find that companies with executive remuneration exceeding $500,000 are less likely to join the TARP.
Both amendments, to section 162 and to 280G, are only applicable to employers who sold assets worth more than $300 million. Assets acquired by the Secretary through direct purchases are excluded from the threshold. Therefore, a company selling assets of 350 million to the Secretary through auctions will be subject to the amendments
5 For a covered employee who earned $1.2 million, from which 0.8 million was paid in stock
options and 0.4 million in cash, non-participating banks could deduct the whole amount, since the 0.4 million cash payment are below the threshold of 1 million and the 0.8 million equity based compensation are not taken into account. Under the amendments of the TARP, the deduction would be limited to 0,5 million, consisting of 0.4 million cash payment and 0.1 million equity based compensation.
13 whereas a company selling 150 million through direct purchases and 200 million through auctions will not be subject to the amendments. However, these companies will be subject to Section 111(b) and 111(c) which constrains executive compensation directly. Section 111 applies to any financial institution that participates in the TARP and furthermore distinguishes between direct purchases [Sec.111(b)] and auction purchases [Sec.111(c)]. In contrast to section 302 it does not incentivize companies through taxes but gives direct guidance for executive remuneration. Under subsection (c) a company is not allowed to close new employment contracts which include “golden parachutes”, if assets were purchased through an auction and exceed $300 million. Golden parachutes are defined in the IRC section 280G(e) as “any payment in the nature of compensation to (or for the benefit of) a SEO [Senior Executive Officer] made on account of an applicable severance from employment to the extent the aggregate present value of such payments equals or exceeds an amount equal to three times the SEO’s base amount.” Although the role of golden parachutes is limited within this study, the above quote raises the question who senior executive officers are, which is also important for the application of Section 111(b) – direct purchases. Senior executive officers are “named executive officers” as defined in the federal securities law (17 CFR 229.402) and additionally have to fulfill three conditions. They have to be employed by a participating institution while the treasury holds equity or debt of the company. Furthermore the principal executive officer (PEO) as well as the principal financial officer (PFO) are always subject to the law. Lastly, the three highest paid employees excluding PEO and PFO are also subject the regulations. The determination of the three highest paid employees also refers to the federal securities law.
In the case that the Secretary purchases assets directly, Section 111(b) applies. Under this subsection compensation contracts with SEOs have to include 3 distinct features. Similar to section 111(c) golden parachute payments are forbidden. Furthermore the contracts must contain a provisions for the recovery of any bonus or incentive payments made to the SEO which were based on any statements that are later proven to be wrong. This feature appears to be very similar to Section 304 of the Sarbanes-Oxley Act of 2002 (SOX). Indeed, questions about this issue were raised in several comment letters. According to an interim final rule published by the Department of the Treasury the section 304 of SOX and section 111(b)(2)(B) of EESA differ from each other in several ways. In contrast to the regulation under SOX, EESA is not limited
14 to publicly traded companies and is also applied to the five highest compensated executives / employees. In that sense the scope of EESA is wider, however it is only applied to banks participating in the program. More importantly, the provisions required under EESA must not exclusively include accounting restatements as triggering events. That means that any other event in the course of the business process can potentially trigger a clawback independently from its nature. Furthermore there is no limitation in the recovery period and it covers material inaccuracies relating to financial statements as well as non-financial metrics that were used to determine compensation. Section 111(b)(2)A is the most relevant paragraph for the development of this study. Under the circumstances mentioned above, compensation contracts have to include “[…] limits on compensation that exclude incentives for senior executive officers of a financial institution to take unnecessary and excessive risks that threaten the value of the financial institution during the period that the Secretary holds an equity or debt position in the financial institution;” The Department of Treasury issued “Notice 2008-PSSFI” to give additional guidance for the application of this rule. Promptly after entering the program, and no longer than 90 days afterwards, compensation committees are mandated to review compensation contracts of SEOs with a company’s risk manager. They have to ensure that SEO’s compensation does not encourage the SEO to take “unnecessary and excessive risks” (Notice 2008-PSSFI). Furthermore compensation committee and risk managers have to review the compensation contracts at least once a year. A certification of these processes by the compensation committee is required. As indicated by the quote from Notice 2008-PSSFI even in the notice there is no specific explanation what unnecessary or excessive risk means. The only thing explicitly mentioned in the notice is that companies have to eliminate elements of the contract that might lead to these risks.
From the “Interim Final Rule”, another explanatory paper issued by the Treasury, it can be inferred that this section is held vaguely because different companies face different risks. This view of the relation between compensation and risk-taking implies that there are no effects that are applicable to all companies within the financial industry. Accordingly, there is no attempt to narrow down features that induce excessive risk-taking. However, literature suggest that there are generally applicable rules that could reduce excessive risk taking. Besides the length of the incentivized period, the portion of equity based compensation seems to play an important role when
15 it comes to risk-taking incentives. The relation between equity based compensation and risk-taking is outlined in the following chapter.
In February 2009 US President Obama strictly capped the annual pay for executives of companies receiving governmental funds to $500,000 and thereby reacted to public anger about the excessive payments in the financial sector during the crisis. According to several sources, participating companies can pay their executives more than $500,000 only by providing them with stocks that have to be hold until the bank has paid back governmental funds completely. Furthermore, bonus payments were limited to one third of their annual pay and the performance goals of these bonuses had to be long term goals. (Kuang and Qin, 2011; White House Announcement, 2009; CNN Money 2009; CNN Politics 2009; NBC 2009) However, these limits did not apply retroactively, so not to firms that were already participating in the program on February 4th, 2009. This announcement is contradictious to the incentives provided by the TARP initially. Where the amendments to 162 IRC reduce incentives to pay executives with shares, the announcement of February 2009 forced companies to use restricted shares to pay executives in excess of $500,000 annual pay.
In conclusion of this chapter we can see that regulators try to eliminate golden parachutes through direct enforcement or through excessive taxation. Furthermore companies participating in the TARP are incentivized to use less performance and equity based compensation but are also prohibited to pay more than $500,000 annual salary in cash, which induces the use of restricted stock options.
2.2 Compensation structure and risk-taking
Jensen and Meckling (1976) find that the separation of ownership and control induces agency problems. Through the neoclassical lens of agency theory the relation between shareholder (owning party) and executives (controlling party) is a principal-agent relation. Along with agency conflicts the misalignment of interests of executives and shareholders can lead to a decrease in shareholder value (Jensen and Meckling, 1976). In order to align executives’ interests with the ones of shareholders, incentivizing compensation structures can be implemented. By providing executives with stock or stock options (equity based compensation) executives have a greater interest to maximize firm value because a gain in shareholder value will also increase
16 executives’ wealth. The recognition of that fact led to an increase in equity based compensation over the last two decades (Appendix B). It is important to state that this expresses the relation between executives and the shareholders. The fact that bondholders are not included in the concept will lead to the motivation of executives to take excessive risk even when they are rewarded with stocks (instead of stock options). Murphy (2012) explains the relation between compensation structure and risk-taking by using call-options. Call-options on stock have an unlimited upside potential. As long as the market price of stock is higher than the strike price executives will benefit as soon as they exercises the option. According to Murphy (2012, p. 29) the downside potential is limited to zero because if the market price drops below the strike price, executives will not exercise the option. They will have no benefit, but they will lose no money either. Stock grants in leveraged firms can have similar effects. If a company has equity of $100 the downside potential will always be limited to $100. If a failed project causes losses of $300, $100 will be attributed to the equity holder. Debt holders have to bear for the remaining $200. An equity holder (in this case the executive who holds part of the equity) will therefore only take a possible loss of $100 into account because that is the most he / she can lose. Thus the equity holder participates only partially in the losses but fully in the gains. The net present value (NPV) of a project can be positive in leveraged companies although it would be negative in companies that are not leveraged. Therefore equity based compensation in the form of options or stocks will lead to increased risk-taking because the losses that could occur are limited whereas the executives will fully participate in gains.
There are several studies which confirm that idea and show that equity based compensation leads to increased risk taking. DeFusco et al. (1990) for example find that implied volatility as well as stock return variance increase after the implementation of equity based compensation plans. Chen et al. (2006) investigate the banking industry specifically and find that stock-option based compensation induces risk-taking in the banking sector.
2.3 Hypothesis regarding compensation structure and risk-taking
This study is concerned about the effects of the TARP which was initiated by the government in order to stabilize the financial markets. As prior literature has shown,
17 there are conflicting objectives within the TARP. Whereas incentives and restrictions regarding the compensation structure should reduce risk, empirical evidence shows that participating banks tend to engage in riskier assets. Therefore the effects of changes in the compensation structure are not effective or they are overlapped by other effects. In order to examine the effects in greater detail, compensation structure and risk profile will be examined separately in this study. The effects of the program will be identified by comparing companies that participated in the program to non-participating companies. Since the TARP is a temporary program, participating companies will enter and leave the program. Both events might have an effect on compensation structure as well as on the risk profile. Figure 2 shows which effect and event is cover by which hypothesis.
Figure 2 - Hypothesis overview
H1. Entry into the CPP
The examination of the TARP has shown that companies participating in the TARP have to review and adjust compensation contracts in order to eliminate elements that induce excessive and unnecessary risk-taking. Chapter 2.2 shows that risk-taking can be reduced by limiting equity based compensation. From chapters 2.1.3 and 2.2 it can be concluded that there are different aspects influencing the use of equity based compensation. Therefore it is hard to make predictions. The initial regulation in TARP reduced incentives to use equity as compensation. In addition, the call to reduce excessive and unnecessary risk-taking in combination with the findings of chapter 2.2 would also suggest a reduction of equity based compensation, because a reduction in equity based compensation will lead to a decrease in risk profile. On the other hand, the announcement released in February 2009 forces companies to use restricted stocks to
Time Enter the TARP Withdrawal from the TARP Compensation structure Risk profile Voluntary vs. Mandatory Change H1a H2a H1b H2b
18 pay executives in excess of $500,000. Since the announcement in February 2009 does not affect participants retroactively (CNN Money, 2009) and the majority of the sample entered into the TARP before that date, the initial regulation of TARP is expected to have greater impact. However, the contradictious regulation will be discussed in the findings and conclusion. For the purpose of formulating hypothesis 1 we assume that companies reduce their equity based compensation. On basis of Chapter 2.2, reduced equity based compensation will lead to a lower risk-profile. The hypotheses can be stated as follows:
H1a) After having received TARP funds the average percentage of equity based compensation in participating banks will decrease relative to the one of non-participating banks.
H1 b) After having received TARP funds the average risk-profile of participating banks will decrease relative to the one of non-participating banks.
H2. Exit out of the CPP
Executives of participating banks might feel the urge to make up the salary they missed being under the forces of TARP. An indication for that can be found in the study of Bayazitova and Shivdasani (2012), who mention that the restrictions imposed on compensation structure had the positive side effect that banks paid back their money as quickly as possible in order to increase their compensation again. Since institutions were able to pay back the funds, they are expected to be in a good financial position. Thus, an increase in compensation can be realized through a change in compensation structure. The growing financial markets and salaries mentioned in the introduction are also an indication for this relation. In addition, TARP constraints incentives to use equity and performance based compensation. After leaving the program banks might take advantage of these incentives again. Therefore the “make-up hypothesis” can be stated as:
H2 a) After having paid back the TARP funds the average percentage of equity based compensation in participating banks will increase compared non-participating banks.
19 In opposition to some voices in the public media, neither shareholders nor managers of banks have benefited from the crisis (Fahlenbrach and Stulz, 2011). That fact and statements like “[a]n individual’s performance should be evaluated over time so as to avoid excessive risk-taking”6 (Financial Times, 2009) from Goldman Sachs’ CEO Lloyd Blankfein show that high executives within the banking industry learned from the financial crisis. Therefore one could argue that executives will implement stricter compensation plans which are closer to the ones imposed by the government. Although the “lessons learned” are not restricted to companies participating in the TARP there could be significant differences in ways banks learned from the crisis.
Banks participating in the TARP were mandated to change their compensation structures to eliminate unnecessary and excessive risk. The adjusted compensation plans were reviewed by risk managers. Since executives try to maximize their own compensation they can be expected to change compensation structures in a way that reduces risk when in the same time maximizing the overall compensation within the limits of the TARP. As a consequence, participating banks will have a relatively high compensation relative to their risk profile. It is questionable when this effect is measurable. Although compensation structures have to be reviewed immediately after entering the TARP, companies might take some time to figure out how to improve the compensation / risk ratio. Therefore this effect might increase during the time of their participation. Maximizing the compensation / risk ratio is favorable for managers for several reasons. Kuang and Qin (2014) show that compensation structures are incorporated into credit ratings. Less risky incentives might therefore lead to a better rating and real economic benefits (e.g. lower interest). This will induce a change in the relation between compensation structure and risk-taking. Following hypothesis can be stated.
H2 b) After having paid back the TARP funds the risk-profiles of participating banks stay significantly lower than the one of non-participating banks.
6 Financial Times from Feb. 8th 2009, accessed on Jan. 31st 2014 under
20
3. Research Methodology
3.1 Sample selection and descriptive statistics
Data of U.S. banks from 2006 to 2012 was retrieved from the databases EXECUCOMP (compensation), COMPUSTAT (annual financial data) and CRSP (daily stock prices). Data sets from EXECUCOMP and CRSP had to be adjusted manually in order to have one entry per year per company. As described in the literature review only CFO, CEO, and the 3 highest paid employees were initially subject to the TARP. The data retrieved from EXECUCOMP was adjusted accordingly: Compensation of employees who were not one of the 5 highest paid employees were excluded. Additionally the amounts and dates of the cash flows between the secretary and participants were hand collected from ‘ProPublica’.7 ProPublica is an independent non-profit organization that produces investigative journalism in the public interest. It has won 2 Pulitzer prices for national reporting. One of the articles awarded was related to management behavior during the financial crisis. Therefore the organization is expected to be trustworthy. The annual financial reports of 2009 of Bank of America and JP Morgen were examined to see if ProPublica reports the correct data – it does. The website lists all companies participating in the CPP, their entry dates as well as all cash flows between the Secretary and the individual bank.
After merging the data bases and deleting observations with insufficient information, 113 banks remained for which compensation data, financial data and daily stock prices were available; 67 of these banks were participating in the TARP. Especially in times of crises risk-profile and compensation structure might fluctuate over time. In order to eliminate timely effects, banks participating in the program are compared to a peer group, consisting of banks which are not participating in the program. Bayazitova and Shivdasani (2012) find that banks participating in the TARP have specific criteria that differ from banks that do not participate in the program. When selecting banks for the peer group, these criteria were taken into account to avoid potential selection bias.
21 Propensity score matching was used to form pairs of participating and non-participating banks with approximately the same characteristics. These characteristics were based on Bayazitova and Shivadansi (2012). The authors state that banks first had to apply for participation in the TARP and then had to be approved by the Secretary. Therefore both decisions, to apply and to approve, potentially preselects participants. As the name “Emergency Economic Stabilization Act” already states, the act was supposed to stabilize financial markets. Furthermore, the notion “Emergency” indicates that this stabilization had to be done quickly and therefore efficiently in regards to time. In order to increase efficiency the relatively bigger market players who were in greater trouble were addressed first. This is supported Bayazitova and Shivadansi (2012) who find that capital infusions were directed towards large banks with greater derivative exposure. To protect tax payers no funds were issued to banks with distress loan portfolios. Furthermore they find that companies with less stable funding mixes were more likely to accept capital infusion, and healthier banks were more likely to reject them. The company size is measured by the natural logarithm of its assets. Great derivative exposures are measured by the risk adjusted capital ratio tier 1. The funding mix is expressed as whole sale debt ratio, which are non-deposit liabilities scaled by assets. Finally the healthiness of a bank is measured by its return on assets; ROA is the earnings before interest and tax scaled by total assets. Due to limited data availability distress loan portfolios could not be measured without losing a significant amount of data. Since the characteristics described above change over time, a specific year had to be used to match pairs of companies. Within the sample, the earliest date of entering the TARP was 28-Oct-2008 (JP Morgan Chase / Bank of America) and the latest date was 10-Apr-2009 (City National Corp). Therefore the financial data retrieved for 2008 was the closest to the date of the participation decision and was used to match pairs. The matching at a tolerance level of 0.3 resulted in 38 pairs, so 76 companies and 510 year observations.
22
Table 1 - Logistic Regression of Matching Pairs8
A logistic regression was run to assess whether the participation decision within the sample followed the pattern predicted by Bayazitova and Shivadansi (2012). The measures increased the model’s predictability from 50% to 69.7%. The prefixes of all coefficients confirm prior finding. There is a significant positive relation between company size and likelihood of participation. Although not significant a positive prefix indicates that funds were directed towards banks with greater derivative exposure. Furthermore, companies which had less stable funding mixes and which were in less healthy conditions were more likely to accept capital infusions as indicated by negative prefixes.
Since the matching was only based on parts of the sample, the whole sample was tested to check if the matching was successful. A paired sampled t-test as well as a related-samples Wilcoxon test were run before and after the matching in order to test for differences in mean and media. Tables 2 and 3 shows the results.
8A logistic regression with PARTICIPANT (0 for non-participants and 1 for participants) as dependent variable was run for the year 2008 since this is the year of decision if a company participates in the TARP or not. Independent variables were calculated as described in the text;*** represents significance at 1%; ** represents significance at 5%; * represents significance at 10%
Measure Variable df S.E. Prior
Findings
Size ln_assets 0.461 ** 1 0.224
a
Stability Of
Funding Mix FundingMix -3.872 1 2.991
a
Derivative Exposure RiskAdjusted-CapitalRatioTier1 0.134 1 0.096
a
Healthiness ROA_EBIT -73.385 *** 1 28.251a
Constant -3.958 * 1 2.278 Coefficient23
Table 2 - Paired sample t-test pre-matching9
Table 3 - Paired sample t-test post-matching10
The test that was performed before propensity score matching was used shows that there are significant differences in mean and median for the risk adjusted capital ratio tier 1, the funding mix, and the assets. After the matching companies participating in the program are still significantly bigger. The difference in derivative exposure as well as the effect of stable funding mixes could be mitigated. However in the matched sample, the return on assets of participating banks is significantly lower than the one of non-participants. These results are confirmed by the non-parametric Wilcoxon test. Alternatively the matching was run at a tolerance level of 10%, resulting in at sample size of 33 pairs, but with the same significant differences in company size and return on assets. However, by using these variables as control variables in the final model the bias is mitigated. The relatively low number of observations results from technical issues. In order to run the paired sample t-tests the observations for one pair and year were combined. Therefore the number of observations was reduced by approximately 50%, but each observations contains information about participants and non-participants.
The descriptive statistics for the regression model are listed below. The basic data was tested for normal distribution. Based on the measured skewness and kurtosis the data was transformed to adjust them as close as possible to normal distribution. Natural
9 *** Wilcoxon test yield significant changes on a 1% level. With variables as described in the text 10 *** Wilcoxon test yield significant changes on a 1% level. With variables as described in the text
Significance Participating Non-participating (2-tailed) Risk-Adjusted Capital
Ratio T1 284 11.0426 11.8823 - 0.840 0.005***
FundingMix 284 0.1926 0.1746 0.018 0.039***
ROA_EBIT 277 0.0185 0.0172 0.001 0.206
ln_assets 267 9.7373 9.0127 0.725 0.000***
Paired Samples t-test
N Mean Difference
Significance Participating Non-participating (2-tailed) Risk-Adjusted Capital Ratio T1 239 11.9011 11.9745 - 0.073 0.789 FundingMix 240 0.1811 0.1700 0.011 0.232 ROA_EBIT 237 0.0155 0.0181 - 0.003 0.007*** ln_assets 240 9.5808 8.9989 0.582 0.000*** N Mean
Paired Samples t-test
24 logarithm and its reflection were used widely. Additionally, ‘income_diversification’, ‘ln_Equity_ratio’, and ‘ln_CaptRatioTier1’ were winsorized at a 1% level, which means that all values in the 1st percentile substituted by the smallest value of the 2nd percentile. Accordingly values in the 100th were substituted by the biggest value of the 99th percentile. In order to run paired sample t-test as additional analysis, this data was split in two parts, participants and non-participants. Afterwards it was merged again by combining observations which had the same pair number and year. Although the skewness and kurtosis are in acceptable range for almost all variables, related-samples Wilcoxon tests were run in addition to paired samples t-tests to verify the results.
Table 4- Descriptive statistics regression model11
11 With LN_STD as the natural logarithm of the standard deviation of daily stock prices,
Avg_Rel_Equity_COMP as average relative equity compensation of the 5 highest paid employees scalded by average total compensation of these employees, LN_AddUp_equity_absolute as natural logarithm of absolute equity compensation of the 5 highest paid employees, PARTICIPANT, ENTER, ENTER_PARTI, EXIT, and EXIT_PARTI, as dummy variables as described above, ln_assets as natural logarithm of assets, ln_fundingmix as natural logarithm of the wholesale debt ratio which is non-deposit liabilities scaled by total assets, income_diversification was calculated by dividing total non-interest income by the sum of interest income and non-interest income (Chen, 2006), ln_Equity_ratio as the natural logarithm of total equity divided by total assets,
ln_CaptRatioTier1 as the natural logarithm of the risk adjusted capital ratio tier 1, R_ROA_EBIT as the reflected logarithm of the earnings before interest and tax divided by total assets, and
LN_TotalComp_Addup as the natural logarithm of the total compensation of the 5 highest paid employees. Variables N Mean Std. Deviation Minimum 25th Percentile Median 75th Percentile Maximum
Statistic Statistic Statistic Statistic Statistic Statistic Statistic Statistic
LN_STD 446 0.62 0.71 -1.12 0.14 0.65 1.09 2.20 Avg_Rel_Equity_COMP 510 0.27 0.23 0.00 0.07 0.24 0.41 0.91 LN_AddUp_equity_absolute 510 5.77 3.03 0.00 4.93 6.73 7.67 11.09 PARTICIPANT 510 0.49 0.50 0.00 0.00 0.00 1.00 1.00 ENTER 510 0.71 0.46 0.00 0.00 1.00 1.00 1.00 ENTER_PARTI 510 0.35 0.48 0.00 0.00 0.00 1.00 1.00 EXIT 510 0.22 0.42 0.00 0.00 0.00 0.00 1.00 EXIT_PARTI 510 0.11 0.31 0.00 0.00 0.00 0.00 1.00 ln_assets 510 9.32 1.38 6.85 8.41 9.09 9.75 14.67 ln_fundingmix 510 0.16 0.08 0.02 0.09 0.15 0.20 0.45 income_diversification 509 0.28 0.16 -0.75 0.18 0.27 0.36 0.82 ln_Equity_ratio 510 -2.33 0.38 -5.14 -2.47 -2.31 -2.13 -1.11 ln_CaptRatioTier1 509 2.44 0.31 -0.62 2.28 2.45 2.60 3.45 R_ROA_EBIT 505 0.39 0.01 0.37 0.39 0.39 0.40 0.44 LN_TotalComp_Addup 510 8.29 0.86 6.64 7.67 8.14 8.69 11.77 Descriptive statistics Dependent variables Independent variables Controls
25
3.2 Research design
The research questions addressed was investigated through a quantitative analysis in an “event-type” style as recommended by Van der Stede (2011). The procedure differs from a classical event study because the association before, between and after two events are compared. The basic model is a multiple linear regression which was used to perform a difference-in-differences analysis to examine the effects of the TARP on compensation structure and risk profile. To gain a more detailed insight paired sample t-tests were run to compare the mean of the discussed variables in each period. The results of these t-tests were confirmed by related samples Wilcoxon signed rank tests for non-parametric data. Figure 3 illustrates the analysis performed.
Figure 3 - Test design
The difference-in-difference analysis allows cross sectional analysis between participating and non-participating banks as well as time series analysis before and after each event. Furthermore the interaction variable combines cross sectional and time series analysis and may yield meaningful results. The disadvantage of relying solely on a difference-in-differences analysis is that it compares participants to non-participants on the basis of at least 2 periods. Therefore there will always be a certain bias when interpreting cross sectional data. This can be mitigated by using paired sample t-test for each period. However, since these tests do not control for other effects, they are only used additionally to confirm results.
Time Enter the TARP Withdrawal from the TARP Participants Non-Participants Compensation structure And Risk profile TARP Post-TARP Pre-TARP 2 3 1 Difference in mean and median t-test Cross sectional and time series
testing (H1)
Cross sectional and time series
testing (H2)
t-test t-test
Difference in differences
26 As described in the literature review equity based compensation can incentivize excessive risk-taking. Therefore compensation structure should be measured by incentives. There are basically two ways to do that. Murphy (2012) defines his measure for executives’ incentives as “effective ownership percentage”. This measure goes back to Jensen’s and Murphy’s (1990b) ‘Pay-Performance Sensivity’. Both measures are based in agency theory which claims that the alignment of interests of shareholders and executives decreases with a decrease in the portion of equity owned by executives (Murphy, 2012). The other way to measure incentives is to measure the portion of equity based compensation on the whole compensation. In this study, compensation structure is measured by equity based compensation relative to total compensation and by the natural logarithm of the total equity based compensation. Equity based compensation is the sum of stocks and stock options granted. The value of stock options was estimated on basis of a Black-and-Scholes scheme. Adding bonus payments, salary, and total other compensation to that results in total compensation. The reason not to follow Jensen’s and Murphy’s idea is that the TARP does not give any incentives to change the pay-performance sensivity but to change the compensation structure of senior executive officers.
Black and Hazlewood use loan portfolios of banks to estimate the risk profile. This measure excludes the perception of the market to a relatively large extent. This procedure has the advantage that bias from market driven effects are excluded. However, besides compensation structure there are other non-market driven factors influencing risk-profile. Nonetheless, Kuang and Qin show that compensation structures affect companies’ ratings. Therefore, a change in compensation structure will not only be reflected in loan portfolios but also in market measures. Based on that it can be concluded that market measures are as good as non-market measures to examine the effect of compensation on the risk profile. The risk profile was therefore measured by the volatility of daily stock prices.
In a traditional difference-in-differences analysis there are 2 indicator variables implemented. In the presented study two events were investigated separately. Therefore one additional indicator was added. The indicator “PARTICIPANT” indicates if a company participates in the program (1) or not (0). So independently from the year the observation was made, ‘PARTICIPANT’ will turn 1 or 0 for each company. To observe
27 the change before and after an event, there needs to be implemented one additional time-variable for each event. Time-variables indicate if an observation was made before or after an event. Since this study examines 2 events, there are 2 time-variables implemented. The indicator ’ENTER’ is 0 for all observations made before the entry into the program and turns 1 for observations made after a company enter the TARP. Similar to that ‘EXIT’ is 0 for all observations made before a company leaves the program and 1 for observations made after leaving the program. In summary, all observations are categorized by 3 variables. The time-variables can be combined to show if an observation was made before, during or after the company participated in the program. For entries where both, ‘ENTER’ and ‘EXIT’ are 0, the observation was made before the company entered the program. If the company enters the program, ‘ENTER’ turns 1 but ‘EXIT’ stays 0 as long as the company has not left the program. Therefore the combination ‘ENTER = 1 & EXIT = 0’ indicates observations made during participation in the program. Accordingly, leaving the program will turn ‘EXIT’ to 1; an observations with both time-variables being 1 was made after a company left the program. For the purpose of completeness it should be mentioned that the combination of two binary variables can result in 4 possible combinations. The combination ‘ENTER = 0 and EXIT = 1’ is only theoretically possible since a company can only leave the program once it has entered before. Table 5 provides a summary as well as the number of observations made in each period.
The critical question for the determination of the time-variables is when the restrictions and amendments of the TARP apply. Similar to other sections, section 111(b)(2)(A) states that elements of compensation contracts that induce excessive risk have to be excluded “[…] during the period that the Secretary holds an equity or debt position in the financial institution;” Therefore the initial payments from the secretary to the bank was used as entry date. A bank that has fully paid back governmental funds has left the program. Therefore the last refund payment marks the exit of a company out of the program. The technical determination of the time-variables was possible without any major constraint, because the website provides all payments between banks and the Secretary. However, there have to be made some underlying assumptions which potentially affect the outcome of this study. The general problem yields from the fact that one observation was made for one year. In contrast to regulatory changes like the introduction of SOX or IFRS, there is no distinct date on which all companies have to
28 comply with the new regulation. Each bank individually entered and left the program at a certain date that is most likely not the date of their year-end closing. Therefore an observation covering one year has to be classified as either before, during or after the participation although the company entered or left the program during that year.
Imagine bank A and B. Bank A enters the program on 20-Dec-2008 and bank B on 03-Jan-2009. The actual difference between the dates is only 3 weeks, however, the 2008 observation will be treated as participation for A but not for B. The first question arising from that is, if the different treatment of A and B can be justified. It could be justified if the 3 weeks difference potentially have an impact on the dependent variable. So, are compensation structure and risk profile of A in 2008 affected by the entry in late 2008? Or more abstract: When does the TARP potentially affect a company’s risk profile and compensation structure? The risk profile is measured by stock volatility, a market measure which includes the expectations of investors. Announcing to join the TARP in December 2008 has no influence on the volatility for the previous 11 month. Therefore classifying 2008 as TARP year would not be accurate. The signaling effect of TARP will affect stock prices in 2009 the earliest.
Since the effects of TARP on compensation structure are first examined by this study, it is hard to predict when the TARP will affect compensation. As outlined in the literature review, the TARP is expected to have multiple effects on compensation structure. First of all, immediately after entering the program, compensation committees and risk managers have to review compensation contracts and eliminate elements that induce excessive and unnecessary risk. Entering in 2008 instead of 2009 would also affect the compensation contracts of 2008. Therefore, declaring 2008 as a TARP year would make sense. On the other hand it is questionable if in practice compensation contracts will be adjusted for past events.
Secondly, compensation structure might change due to different incentives. As the amended Internal Revenue Code (IRC) prohibits the deduction of equity based compensation in excess of $500,000 the incentives to pay executives with options or stocks are reduced. The underlying question stays the same: When do these incentives affect compensation structure? The amendments of IRC 162(m) are effective if the financial institution is treated as ‘applicable employer’, the SEOs are treated as ‘covered employees’ and if “any taxable year that includes any portion of that period is
29 treated as an ‘applicable taxable year’ […]” (Interim Final Rule, p. 23) The quote states that the amendment introduced by the TARP are applicable for any taxable year in which the Secretary holds equity or debt of the bank. Following that rule strictly would imply that bank B would be allowed to deduct $1 million from their tax base in 2008 whereas bank A could only deduct $500,000 although A joined the program only 3 weeks earlier. The Interim Final Rule furthermore states that the deductibility should be allocated on a pro-rata basis for the portion of the taxable year covered by the TARP. However, the analysis of the IRC does not suggest any pro-rata allocation.
As a conclusion from the discussion above: Since the effects of the TARP on compensation structure are unknown, 2 sets of time-variables are implemented in the study. The first “conservative” set assumes that the effects of TARP will always affect compensation structure and risk-profile, no matter how short the covered period is. That means, joining the TARP in late 2008 will result in a declaration of 2008 as participating year; ‘ENTER’ is set to 1 in 2008. If a company leaves the TARP in early 2010, is assumed that the effects will still prevail in 2010. Therefore 2011 is the first non-participating year; ‘EXIT’ is set to 1 in 2011. The more “progressive” set of time-variables assumes that the effects of TARP will only be measureable if at least half of the observed year is covered by the TARP. Therefore 2008 will never be classified as TARP year because banks started to enter the TARP in October 2008. ‘ENTER’ is set to 1 in 2009. If a bank leaves the TARP in the first half of a year, the specific year will be declared as non-TARP year. If the company leaves the program in the second half of the year, the according year is classified as TARP year. The following year will be the first non-participating year; ‘EXIT’ will turn 1 in the following year.
As can be seen in table 5, the more progressive classification of observations results in an overall decrease in observations made during the participation in the TARP. This is due to the reclassification of 2008 observations (62) as well as reclassifications related to the date of exit (25). The more progressive set of time-variables will be used in a robustness test to verify the findings. Furthermore the time variables were used reduce bias. To examine the effects of entering the TARP, the pre-TARP period has to be compared to the pre-TARP period. This can be achieved by ignoring all observations that were made in the post-TARP period. For H1 all observations with
30 ‘EXIT = 1’ were ignored. Accordingly for H2 all observations with ‘ENTER = 0’ were not taken into account.
Table 5 - Time-variables
The following model was employed: 𝐶𝑂𝑁𝑆𝐸𝑄𝑈𝐸𝑁𝐶𝐸
= 𝛽0+ 𝛽1𝑃𝑎𝑟𝑡𝑖𝑐𝑖𝑝𝑎𝑛𝑡 + 𝛽2𝑇𝑖𝑚𝑒𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒 + 𝛽3𝑇𝑖𝑚𝑒𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒 ∗ 𝑃𝑎𝑟𝑡𝑖𝑐𝑖𝑝𝑎𝑛𝑡 + ∑𝛽𝑘𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠 + ∑𝛾𝑘𝑌𝑒𝑎𝑟
Relative equity compensation and the natural logarithm of absolute compensation in equity were used as dependent variable to test the compensation structure. The risk profile was measured by the natural logarithm of volatility in daily stock. ‘Participant’ is set 1 for companies participating in the program and 0 for non-participants. ‘TimeVariable’ can be either ‘ENTER’ or ‘EXIT’, depending on the event observed. The time-variable will also be adjusted for non-participating banks. So if a company enters the TARP in 2008, ‘ENTER’ will also be 1 for the paired non-participating bank. The coefficient of the interaction variable 𝛽3 represents the difference-in-differences and turns 1 if the according time-variable and PARTICIPANT are 1 at the same time and 0 otherwise. The coefficient indicates the difference in the dependent variable between participants and non-participants before and after the event. Therefore 𝛽3 is the coefficient which is relevant to test the hypothesis. Firm size as measured by the natural logarithm of assets, the reflected natural logarithm of return on assets, the natural logarithm of funding mix, and income diversification were added as control variables for both dependent variables. The natural logarithm of the total compensation was added to the model to estimate compensation structure more accurately. The natural logarithm of the equity ratio and capital tier 1 ratio were added to improve the estimation of risk profile. Year dummies also increase the predictability of the model.
Conservative Progressive ENTER EXIT An observation with the following indicators … 150 212 0 0 … was made before a company entered the TARP 246 159 1 0 … was made during the participation in the TARP 114 139 1 1 … was made after the company left the TARP
- - 0 1 Only theoretically possible
Description Number of observations (N) Indicator
