Amsterdam Business School
Master Thesis
Does Government Injection Affect Dividend Policy - A Study on
Capital Purchase Programme
MSc Business Economics, Finance Track
Yang, Jingxian
June 2015
Statement of Originality
This document is written by Jingxian Yang who declares to take full
responsibility for the contents of this document.
I declare that the text and the work presented in this document is
original and that no sources other than those mentioned in the text and its
references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the
supervision of completion of the work, not for the contents.
Abstract
I examine the impact of the Capital Purchase Programme on the dividend payment of those recipients. No significant difference between the CPP participants and non-participants is found in 2008Q4, the quarter when CPP was announced. The financials were the determinants of dividend payment for banks that were out-performing before the crisis while the historical dividend was the key determinant for under-performing firms. In the six quarters after CPP was announced, those CPP recipients on average handed out less dividend that non-recipients did. I observed no abuse of the government injection by the means of paying cash dividend.
Table of Content
Thesis ... 1
1. Introduction ... 1
2. Background on Trouble Asset Relief Programme and Capital
Purchase Programme ... 3
3. Literature Review ... 5
4. Methodology ... 10
5. Data ... 14
6. Result ... 18
7. Conclusion ... 24
Reference ... 26
Appendix ... 28
1
1. Introduction
In October 2008, the United States initiated the Troubled Asset Relief Program (TARP) in hope of stabilizing and strengthening its financial sector. The Capital Purchase Program (CPP), one of the major components of TARP, was designed to help banks extend lending in tough economic times. Based on the principle of voluntary participation, from October 2008 to December 2009, 707 banks received in total $205 billion funds through CPP. Regulations are made for these recipients. When staying in the programme, the top executives face a cap on their compensation and firms need to ask the Treasury for permission if they want to raise dividend on common stock. The year after the fall of Lehman Brothers is regarded as a period when funding was difficult, and the accumulated loss of banks was greater than new capital raised (Acharya et al 2011). Questions came as CPP was revealed. In the eyes of the public, it is the financial institutions who played risky games that lead to the crisis, but it is the public who burden the consequences. When the crisis came, money from the taxpayers was required to save those financial institutions, but the public didn’t seem to believe that their help was cherished as news on the dividend payment of those CPP recipients came out. One professor from Harvard University argued on New York Times that continuously dividends payment on common stocks among the CPP participants transferred the wealth of taxpayer and debt holder to the equity holders. According to his estimation, the dividends that the top executives of the 9 initial recipients receive in the following year would exceed $25 million, almost one-fifth of the aggregate injection of $125 billion for these 9 financial institutions. Statistic also shows that in some banks, for example the first financial bank, the dividend handout in the quarter they participate the programme is larger than the amount of injection received.
The other side of the picture is that banks argue they have to maintain the same level of dividend to support stock price and make further capital-raising easier. It is important for them to tell investors that the bank is still in good condition when the tide of mistrust and doubt is high. This argument is supported by Bessler and Nohel (1996) that dividend change has a more severe effect in banking system. In the situation of a
2
declining market, dividend signalling plays a more important role than in normal times, as mentioned in Fuller and Goldstein (2011).
Despite the existing arguments above, it is interesting to take an academic approach to better understand the situation. In this article, I try to seek if CPP has an effect on the dividend policy of the receiving firms by using the difference-in-differences method. I first analyse if announcing CPP programme made a difference on the dividend payout of the potential participates. That is to say, if the potential participates changed their dividend policy before entering the programme as a response to Treasury’s regulation on increasing dividend per share. I look at both the dividend per share on common stocks and the total dividend payout. The former is regulated by the terms of CPP while the latter is what the media and public more concerned about. Secondly, I also analyse what happens to the dividend payout after the firms received the injection to see if the regulation on the dividend per share truly had its affect. Furthermore, the banks are divided into two groups by their profitability to see if firms with different profitability behave differently. They are divided into out-performer and under performer based on the net-income-to-asset ratio.
The data needed for the article are gathered from several sources. ProPublica provides a precise record on the timeline of the financial institutions entering CPP. The bank financials come from Compustat-bank, and Bureau van Dick provides with additional information on the ownership these banks.
My results give no support to the view that the CPP injection was abused by transferring to equity holders. In fact, the dividend policy of CPP participants didn’t differ significantly from the non-recipients within the quarter that CPP was announced. Though not significant, it seems the under-performing participants were more reluctant to reduce their dividend compared with under-performing non-participants and out-performing participants. More surprising results come when analysing the dividend payout during the period when the injection is in place. The dividend per share of the recipients turns out to be less than the non-recipients. One the one hand, it can be viewed as the compliance of the recipients to put the money where it belong; on the other hand, it suggests the efforts of the non-recipients to distinguish themselves during the tough
3 economic time.
The existing literatures on TARP and CPP focus on the allocation and distribution of funds, and the effect of executive compensation. This article hopes to add to the existing literature the effect of the regulation on dividend payment. This article is also linked to the literature on the dividend policy of firms. The dividend policy of financial institutions has been studied by many, covering periods of both good time and bad time. This article provides analyse on the dividend payout of the beneficiaries under the largest government rescue mission in the face of the worse financial crisis since the Great Depression.
The article is arranged as follows. Section 2 depicts a broad background of TARP and CPP. Section 3 and 4 explains the methodology of this research. The main results are explained in section 5 and section 6 is the conclusion.
2. Background on Trouble Asset Relief Programme and Capital
Purchase Programme
Enacted on 3rd October 2008 by the U.S government, the Trouble Asset Relief
Programme (TARP) expresses the effort to combat the worst financial crisis since the Great Depression after the collapse of Lehman Brothers and the rescue of AIG. The programme was authorized by Congress through the Emergency Economic Stabilization Act of 2008 (EESA) and signed into law by the president. Overseen by the Office of Financial Stability at the U.S. Department of the Treasury, the programme was authorized to invest up to $700 billion to strengthen the financial sectors. The programme was conducted in two stages with focus on five main areas: auto industry, bank investment, credit market, housing and investment in AIG.
The Capital Purchase Programme (CPP), announced on 14th October 2008, is under
the bank investment programme. According to U.S. Treasury, this programme aims to help bolster the capital position of viable institutions of all sizes and build confidence in these institutions and the financial system as a whole. Compared with other TARP programmes, it is also interpreted that CPP aims to help “healthy” bank extend lending
4
through providing capital. Initially, the CPP was committed $250 billion, more than a third of total TARP. In March 2009, this amount was reduced to $218 billion.
The CPP is a voluntary programme that open to all qualified financial institutions (QFIs) that is established and operating in the United States. Bank holding companies, financial holding companies, insured depository institutions and some savings and loan holding companies fall in this range. At that time, some investment banks changed into bank holding companies in order be eligible for the programme. The financial institutions willing to participate must submit their applications through appropriate Federal banking agency (FBA) and were required to consult with their FBA before application. The selection criteria to the programme was never made public. Within 30 days after the notification of preliminary approval from the Treasury, the applicant was required to submit the investment agreements and related documents. It is noticed that several banks, though approved by the Treasury, chose to withdraw from the programme during this period of time.
The investment was made through purchasing senior preferred stocks. As designed, the Treasury would purchase non-voting senior preferred stock that counts 1% to 3% (or $25 million, the lesser) of the risk weighted assets from those participants at the price of $1,000 per share, on condition that the participants sign a warrant that they would purchase those preferred stock back at a set price. The investment doesn’t come without a price. Out of the intention to make good use of the taxpayer’s money, the participants are required to pay a 5% annual dividend on the preferred stock during the first five years in the programme, and after five years, the dividend rate jumps to 9%. Apart from the annual dividend and repurchase, there are other requirements that the participants should fulfil. During the first three years in the programme, the participant is not allowed to increase the dividend per share on common stocks unless it is permitted by the Treasury. Nor can the participant repurchase shares (other than the senior preferred stocks) without the consent of the Treasury. Another widely discussed term is the cap on executive compensation. In the initial design, the tax deduction of top executives was limited to $500,000 and the participants were prohibited from making any golden parachute agreement with future executives as long as the Treasury
5
still holds the preferred stocks. In February 2009, it was further amended that the annual compensation of top executives to be limited to $500,000. An office of Special Master was established to review and approve the compensation of the five senior executive officers and 20 next most highly paid employees.
On 14th October 2008, the day that CPP was announced, 9 big financial institutions - Citigroup, Wells Fargo, JPMorgan, Bank of America, Goldman Sachs, Morgan Stanley, State Street, Bank of New York Mellon, and Merrill Lynch - which were considered too big to fail, received an aggregate amount of injection of $125 million, almost half of the CPP funds. Because of their special status, the injection came quick and they didn’t undergo the standard procedure as other CPP recipients later did. For the other potential participants, they had to go through a standard application procedure and wait for the approval from the Treasury. In 2009, some of the existing CPP recipients transferred to another programme: the Community Development Capital Initiative (CDCI) that is designed to help viable certified Community Development Financial Institutions and the communities they serve.
From October 2008 till the end of 2009, the Treasury invested approximately $205 billion in 707 U.S financial institutions in 48 states, with the amount of single investment ranging from $301,000 to $25 billion.
3. Literature review
The dividend policy of firms have been a puzzle in financial studies for decades. Miller and Modigliani (1961) argue that dividend policy is irrelevant assuming that market is perfect, i.e. no tax, no transaction cost. When the market is imperfect, as it always have been, academics argue that various factors affect the dividend policy. Bhattacharya (1979) finds that in the existence of high tax and information asymmetry, dividend can signal expected cash flow. John and Williams (1985) also find that firms distribute larger dividends when managers have private information that firm value is likely to increase. Through an empirical model, Rozeff (1982) concludes that an optimal dividend payout ratio minimize the agency cost and transaction cost of external financing. His study also suggests a relationship among profitability, growth and
6
dividend. Easterbrook (1984) intuitively explains that continuously dividend payment can reduce agency cost. With dividend payout, firms are frequently forced to raise new capital in the market where all the firm financials will be examined. This frequent examination reduce the cost for monitor and cost induced by managers’ risk-aversion at the same time. Jensen and Solberg (1992) find empirical support for the relationship between insider ownership and dividend payout. Barclay, Smith, and Watts (1995) combine several factor together to examine the determinants of dividend yield among different industries. Their study suggests that dividend yield is connected with investment opportunity, regulation, size and signalling factor.
Several studies provide specific factors that affect the dividend in financial industry. Dickens, Casey and Newman (2002) confirm the model of Barclay, Smith, and Watts (1995) in banking industry using the data on bank holding company during 1998 to 2000 from Morningstar. They further modify the model to fit the characteristics of banking industry. Of the seven variables in their modified model, five are significant. They find that dividend yields are negatively related with investment opportunities, future earning, and ownership while positively related to size and dividend history. When future earning is considered a variable to test signal effect, the study provides no support for signal effect. Theis and Dutta (2009) tested the model of Dickens et al. (2003) using another set of data containing 99 bank holding companies in 2006 from Yahoo!Finance. They also find dividend yields have significant relation with size, ownership and dividend history. In addition, capital ratio, which is taken as a proxy of regulatory pressure, also has a significant influence on dividend while investment opportunity and future earning don’t. The study also concludes a non-liner relationship between dividend yield and insider holding.
The different results of the same model using different data suggests that the factors affecting dividend payout are changing over time. Abreu and Gulamhussen (2013) test their model on the data of 462 publicly traded U.S bank holding companies before and during the crisis. While size, profitability and historical growth are significant in both period, expected growth and capital ratio are significant only in the crisis period. Their study suggests that during crisis period banks feel an increasing importance to signal
7
itself as better than average through the way of dividend payment.
The dividend payment of the firms during the recent financial crisis draws the attention of academics. Previous studies suggest that during the crisis period banks tend to hand out more dividend, and this behaviour increased the risk within the system. One literature that explains the seemingly “excess” dividend payment during the crisis years is the work by Acharya, Le and Shin (2013). They did a study on the dividend payment behaviour of ten largest U.S commercial banks and security firms during the period of 2007 to 2009 using a model with financial externalities. The dividend payment of one firm affects the franchise value of others as the others are the debt holders of this particular firm in most cases. They show that firms with a not too low franchise value tend to pay out excess dividends if this externality is strong. And firms with higher leverage ratio also tend to pay out higher dividend. Their sample period overlaps with the research period of this paper, but only focuses on a small groups of big firms and the externalities among the firms. When the sample is extended to include a large number of small-sized and mid-sized firm, the externality might be weakened and it is questionable whether the results and explanations will still hold. In another study by Acharya et al (2011), which also focuses on the dividend and bank capital during 2007 to 2009, they argue that the inadequate dividend payment during the crisis time endangers the whole banking system. They show that while banks raised capital in the forms of hybrid claims, the continuously high dividend payout deprived the bank of the much needed common equity and left behind risky assets. These behaviour increased the leverage among banks, increased systemic risk and benefited the shareholder at the cost of debt holder and taxpayers. Intuitively, this paper explains how the high dividend payout during the crisis period can increase systemic risk, thus provides a reason why we should take a closer look at the dividend payment of the financial firms.
The signal effect of dividend is one of the continuously tests hypothesis in related literature and the results vary greatly. As implied in Bhattacharya (1979), Miller and Rock (1985) and John and Williams (1985), higher dividend should signal better quality of the firm. Filbeck and Mulineaux (1993), Collins et al. (1994) and Boldin and Leggett (1995) find support in the banking sector that dividend is used as a signal for bank
8
performance. Among many studies that either support or reject this hypothesis, two particular papers are of great interest. Bessler and Nohel (1996) show that banks suffer from more severe announcement effect on dividend reductions compared to non-financial firms, suggesting that in banking industry investors react stronger to negative news than in non-banking industry and such reaction is even stronger for larger banks. Their conclusion is draw form a study of 81 dividend cuts or omissions announced by 56 banks in a sample formed by 132 U.S. commercial banks during 1974 to 1991. This research provides explanation on why banks are reluctant to reduce dividends, especially those large banks. Through a study on monthly stock return between 1970 and 2007, Fuller and Goldstein (2011) not only find evidence to support signal hypothesis, but conclude that an increase in dividend is of greater importance in declining market than in advancing market. This paper gives reason to increase dividend in time of crisis.
The existing literature on CPP and TARP mainly focus on two broad aspects: the distribution of TARP fund, the characteristic of banks participating and leaving the programme. These literatures help to better understand CPP and the behaviour of firms under CPP.
Shortly after the TARP academics started to analyse the programme, the early literatures assess the two aspects at the same time. Taliaferro (2009) uses a dataset from September 2006 to June 2009 on both recipients and non-recipients to study the allocation of CPP fund and how banks use such new equity. The study shows banks applied for CPP fund to help realize optimal capital structure. Those applying banks either have high commitments and opportunities for new lending, or have exposures to certain troubled loan classes. As a result only a modest thirteen percent of new equity was used for lending, while about sixty percent of the money was retained to improve regulatory ratio. Li (2010) also studies the allocation of injection money, but the object extends to the whole TARP. The result shows that the TARP investment resulted in a six present annualized growth rate in lending among those banks with a Tier 1 ratio below median level, but the quality of those loans remained ambiguous. Unlike
9
Taliaferro (2009) that traces the evidence in bank financials, in contrast, this study provides another approach to assess the allocation of TARP money - political and regulatory connection are positively related.
Later literatures dig deeper into bank financials, trying to get more clues on the publicly unavailable selection criteria to CPP. Bayazitova and Shivdasani (2012) assess the characteristics of banks that applied for CPP, approved for CPP and voluntarily opt out CPP. For the applying banks, the ratio of wholesale deposit funding and Tier 1 ratio are important factors to consider. The authority’s decision is more dependent on the size and derivative exposure. And the cap on executive compensation plays an important role in banks’ rejection towards CPP. Cornett, Li and Tehranian (2013) further divide the recipients into healthy and unhealthy groups based on pre-crisis performance to see how the size and pre-pre-crisis health of banks are related to receiving injection through a sample from 2007Q1 to 2011Q1. They find that for relatively unhealthy bank, a better financial performance before the CPP increases the probability of getting the investment as CPP favoured banks whose loans were performing well. And for those relatively large and healthy banks, liquidity is the determinant in assigning CPP.
The size of the participants matters in more than one way. Taliaferro (2009) concludes large and small banks apply for CPP for different purpose, Cornett, Li and Tehranian (2013) suggest large and small banks seem to face different entry criteria, and Black and Hazelwood (2012) show that large and small banks differ in risk taking after receiving injection. Using a difference-in-differences analyses, Black and Hazelwood (2012) finds that the average risk in large TARP recipients increased compared with non-recipients, but the average risk decreased in small TARP recipients. The divergence in size doesn’t seem too surprising. Given the fact that TARP aims to stabilize economy and extend lending at the same time, it seems reasonable that money was assigned to different firm with different purpose. The big and healthy banks use the injection to overcome liquidity problem, and then extend lending, which involves risk taking, while the small and unhealthy banks use the injection to strengthen their own financials, which reduce the risk.
10
The motivation of banks’ early exit from CPP is associated with the restriction on the executive compensation. Wilson and Wu (2012) studied the banks’ early exit from CPP. While this studied confirms that the cap on executive compensation do increase the probability of exit, it also shows that those early exited banks have better earning performance in 2008 and raise substantial equity in 2009. The financial strength of the banks is closely related to the exit decision. Cadman, Carter and Lynch(2012) show a decreasing likelihood of participation and increasing probability of early exit among banks given the strong impact of the restriction. Banks feel participation too costly as they fear a talent drain in the company caused by the compensation restriction. The study even finds a higher executive turnover in participating firms. The study also suggests that the executive compensation restriction might have helped a more efficient distribution of CPP investment as the financial performance of those turning down CPP is no worse than the participants.
4. Methodology
To assess whether and to what extend does the CPP programme affect the dividend payout of the banks, I adopt a difference-in-differences approach that contrast the change in the recipients and the change in the non-recipients of the dividend payment. The CPP was initiated in October 2008, the last quarter of 2008, and the last investment was made in the last quarter of 2009. Quarterly data of the recipients and non-recipients during 2008Q3 to 2010Q2 is gathered to build the estimation. During this period, every CPP participants received the injection for the Treasury. Through two different models, I’ll analyse the recipients’ reaction to the CPP when they knew about the project and when they officially received the injection from the government.
As mentioned above, 9 big financial institutions that entered the programme at the very beginning received a large proportion of the fund and didn’t go through the standard selecting procedure. I exclude these 9 firms from my research. Because the CPP is only applicable to the financial institutions that have a “significant operation” in the United States, only the American financial institutions are considered in the
11 model.
A. Banks’ reaction on CPP announcement
According to the terms of the TARP, the consent of the US Treasury is required if any participant is to increase the common dividends per share within the first three years in the programme. I suspect that participating firms would increase their dividend right before entering the programme to avoid lobbying among the Treasury if they were going to do it afterwards, or they would be more reluctant to reduce their dividend per share compared with non-recipients.
I suspect the change in dividend payment happened in 2008Q4 for the following reasons. Though the selection criteria to the programme was closely guarded by both the Treasury and the banks, rumour has it that the programme preferred those financially better off banks or those too big to fail. In this case, banks with a Camel rating of 1 or 2 were almost guaranteed participation if they were to apply. Given the massive amount of communications happened between the applicants and the authority during this special period of time, and the bank’s understanding of its own financial situation, I presume that most of the banks eventually receiving the injection knew that they have a high probability of receiving it when the applied for the funding, or even before. As a result, most firms that entered the CPP should have responded the programme by changing their dividend policy in the last quarter of 2008. Though the programme was signed into law on 8th October, 2008, and quite a number of firms
signed contract with the Treasury at the end of 2008, they should have already adjusted their dividend policy right before signing the contract if they intended to do it. Before the fall of the Lehman Brothers on 15th September, 2008, there wasn’t much information on whether and how the government would intervene. So there didn’t seem to be a special reason for banks to change their dividend deliberately in the third quarter of 2008. As for those non-recipients, either they didn’t apply for the programme in the first place, or they were rejected. In both cases they didn’t have incentive to increase dividend as a rejected bank was uncertain about the injection at the very beginning. From this standpoint it is reasonable to think that those recipients changed dividend policy while non-recipients didn’t.
12
Based on these information above, the first hypothesis is that, in comparison with 2008 Q3, in the last quarter of 2008, those CPP recipients paid more cash dividend compared with non-recipients. and this is to be tested by the following model:
DPS𝑖𝑡 = 𝛽0+ 𝛽1𝑅𝑒𝑐𝑖𝑝𝑖𝑒𝑛𝑡𝑖 + 𝛽2𝑃𝑒𝑟𝑖𝑜𝑑𝑡+ 𝛽3𝑅𝑒𝑐𝑖𝑝𝑖𝑒𝑛𝑡𝑖∙ 𝑃𝑒𝑟𝑖𝑜𝑑𝑡+ 𝛽4𝑋𝑖𝑡+ 𝜀𝑖𝑡(1) 𝑅𝑒𝑐𝑖𝑝𝑖𝑒𝑛𝑡𝑖 is a dummy variable that equals 1 if the bank is a CPP recipient and equals 0 otherwise. And 𝑃𝑒𝑟𝑖𝑜𝑑𝑡 is a time dummy that equals 1 for 2008 Q4 and equals 0 for 2008 Q3. This regression also includes a set of control variables, 𝑋𝑖𝑡 , that might affect the dividend payment of banks.
The coefficient of interest, 𝛽3, captures the difference between the change in dividend of the recipients and the change in dividend of the non-recipients in 2008 Q4. A positive 𝛽3 is expected if the CPP recipients actually increased their DPS as a
response to enter the programme.
Based on the studies of Barclay, Smith and Watts(1995), Fama and Fench (2001), and Abreu and Gulamhussen (2012), the control variables include size, profitability, growth opportunity, signal factor, regulatory pressure, dividend history and independency. Size is measured by the natural logarithm of total assets. Net-income-to-total-assets ratio stands for profitability. The growth opportunity is captured by the market-to-book ratio. The signal factor is the percentage change in net income over the next quarter. And the regulatory pressure is captured by the tier 1 risk-adjusted capital ratio. The dividend in the previous quarter is used as historical dividend, and the independence indicator for Bureau van Dijk is used to measure the independency of the firm.
B. Banks’ reaction when injection is in effective
It is also possible that banks increased their dividend payment after they joined CPP. Though the consent from US Treasury is needed, there’s no source suggesting that this is a difficult task.The second model analyse if there was significant change in the dividend payment among the recipient after they entered the programme.
13
There is no aggregate unit where the treatment is assigned. Instead, individual banks received the injection at a particular point of time. The following difference-in-difference model in regression formation is used to estimate if there were changes after receiving the injection.
DPS𝑖𝑡 = 𝜆t+ 𝛽𝐸𝑓𝑓𝑒𝑐𝑡𝑖𝑣𝑒𝑖𝑡+ δ𝑍𝑖𝑡+ 𝑢𝑖𝑡 (2) 𝐸𝑓𝑓𝑒𝑐𝑡𝑖𝑣𝑒𝑖𝑡is a dummy variable that indicates whether the injection is in place for firm 𝑖 at time 𝑡. Given consideration of the time lag between the entry date (the date that the Treasury committed injection) and the actual date that firm received the injection, I take the quarter after the date of the entry as the time that the injection was in place. 𝜆t is the time effect, representing the effect that changes over time but remains constant among individuals. The covariance 𝑍𝑖𝑡 contains the factors that determine whether a bank can receive the injection.
The model can also be considered as the random effect model. Because the sample focuses on a specific industry in a specific country, the time-invariant characters of these firms shouldn’t have a huge difference. So the individual effect is treated as a random disturbance, not a parameter to be estimated. In order to identify the treatment effect, the treatment should only be determined by the covariance. As stated above, the selection criteria for the programme is not yet publicized, but the study of Bayazitova and Shivdasani (2012) on the characteristics that the recipients share shed light upon this issue. The size of the bank, the wholesale debt ratio and the expected loan loss have a significant effect on the Treasury’s decision, thus will be included in 𝑍𝑖𝑡. Another important factor that determined whether a bank received injection is whether the bank applied for the injection itself. A binary variable indicating whether the bank applied for the injection is also included in 𝑍𝑖𝑡. Unfortunately this information is not available. It’s not clear how many of the non-recipients were rejected by the programme, but of the 193 applications submitted by December 10, 2008, 87% were approved.1 To
simplify the model, I assume those non-recipients didn’t apply for the programme at all. The hypothesis of this model is that after receiving the CPP injection, the recipients
1 Taliaferro, R. (2009). How do banks use bailout money? Optimal capital structure, new equity, and the
14
handed out more dividend compared with non-recipients. If this is true, a positive 𝛽 can be observed.
The whole sample is divided in two subgroups based on profitability before the CPP, that is, the net-income-to-asset ratio in 2008Q. Previous studies suggest that firms with different size were given the fund for different reasons – extend lending, cover potential loss or strengthen regulatory ratio. It is possible that they differ from each other in terms of dividend payout.
Apart from using DPS as depend variable, I also use the cash dividend on common stocks as depend variable. While DPS is to some extend controlled by terms of CPP, general public is relatively more concerned about the total amount of dividend because it is directly comparable with the amount of injection the bank received. I’ll also take a look at this side.
5. Data
The research is basically built on a combined dataset from two separated sources – a second hand data on CPP recipient comes from ProPublica and the fundamentals of the banks come from COMPUSTAT. The independency indicator from Bureau van Dijk (BvD) is used as supplement.
ProPublica is an independent and no-profit news room that focuses on investigative journalism. One of their projects, the Bailout Tracker is keeping track of every dollar and every recipient for the broader $ 700 billion TARP (including CPP) and Fannie Mae and Freddie Mac. Most of their data come directly from the Treasury Department where the documents relating to the programme and contract for each injection are disclosed. A few of their data were collected by themselves through other government agencies, press releases or regulatory filings from bailout recipients. From their database I get the data on the name, time of entry and the amount of injection received of each CPP recipient. The date of entry in the data is the date that the Treasury committed injection, also the date that contract was signed, instead of the actual date when injection was
15
received. Unfortunately, the data itself doesn’t provide any identifier of those participants other than names. In order to merge the data with COMPUSTAT data, this bailout list has to be matched with the COMPUSTAT by company name manually to get the ticker of these banks. The ProPublica database indicates whether a bank is private or public and provides the state of registration of each bank. These information are of help in the process of matching.
The original list contains 709 recipients. From this list, I excluded those 9 original participants. Then, banks that switched to another program, the Community Development Capital Initiative (CDCI), after receiving CPP were also deleted. Excluded both types of recipients, 676 banks left on the list.
The quarterly data on bank fundamentals comes from COMPUSTAT-Bank. Those non-US banks are excluded. Only the cash dividend on common stocks is used, because the banks receiving injecting have to pay 5% dividends on the preferred stocks they sold to the Treasury, thus including the cash dividends on preferred stocks would affect the accuracy of the outcome. Given the fact that CPP was launched in October 2008, and the last company entered the programme at the end of 2009, the time horizon is between July 2008 and June 2010.
One missing piece in the quarterly data is the book value. This can be found in the COMPUSTAT-bank-annual. This annual data and the quarterly data of closing price is used to calculate market-to-book ratio.
The ProPublica data and COMPUSTAT data are then merged to delete those non-publicly traded firms. The data on several large investment banks, like Morgan Stanley, Goldman Sachs isn’t available in COMPUSTAT-bank and thus excluded from the sample. This unbalanced panel data contains 274 CPP participants and 389 non-participants over 8 calendar quarters. Table I displays the distribution of the CPP recipients by time of entry. Of the 275 recipients, 159 entered the programme in the fourth quarter of 2008, account for 57.82% in total, 98 entered the programme in the first quarter of 2009, account for 35.64%. By the end of the first quarter 2009, 93.46% of the firms in the sample have already received government injection.
16
Table I: Distribution of Recipients by Time of Entry
The data on CPP recipients from ProPublica is merged with COMPUSTAT to delete the non-publicly traded banks. Table I displays the distribution of time of entry of the 274 publicly traded firms. The date of entry is the date that the Treasury committed injection, also the date that contract was signed, instead of the actual date when injection was received. Of the 274 recipients, 159 entered the programme in the fourth quarter of 2008, account for 57.82% in total, 98 entered the programme in the first quarter of 2009, account for 35.64%. By the end of the first quarter 2009, 93.46% of the firms in the sample have already received government injection.
Total 2008Q4 2009Q1 2009Q2 2009Q3 2009Q4
Number 275 159 98 12 4 2
Fraction 100.00% 57.82% 35.64% 4.36% 1.45% 0.73%
The independence indicator from BvD is added to the dataset too. The banks is labelled for A+ to D by increasing level of ownership concentration. These alphabetic label is coded into 1 to 10. 1 stands for A+ and 10 stands for 10. The larger the number, the more concentrated the ownership it has.
Using the unbalanced data I check the change in the dividend on common stock and DPS for both CPP recipients and non-recipients. Though there are 274 recipients in the data, only 265 of them have record at the last quarter of 2008. Table II presents the summary statistics of the change in dividend and DPS in both cases. When taking the initiative of CPP as experiment, 39.25% recipients increased their total dividend payout while only 27.06% of the non-recipients did the same thing. Almost the same percentage of both recipients and non-recipients decreased the total dividend. 16.60%of the recipients handed out the same amount of dividends as previous quarter while among the non-recipients the fraction is 28.61%. In terms of DPS, the recipients were also more likely to increase DPS. 32.83%of the recipients increased DPS while the figure for non-recipients is 16.75%. The recipients were more reluctant to cut DPS compared to non-recipients. Overall, while both recipients and non-recipients were cautious about cutting dividends, it seems that the recipients were more eager to increase the dividend payout, no matter the total amount or dividend per share. When considering the entry of CPP as experiment, it is hard to form a proper control group as the treatment happened in multiple times. The figure of those recipients are
17
summarized in the table. Of those 266 CPP recipients, 31.58% increased the total dividend payout and 23.31% increased the DPS.
Table II: Summary Statistics on the Change of Dividend and DPS
Using the unbalanced data I check the change in the dividend on common stock and DPS for both CPP recipients and non-recipients. Though there are 274 recipients in the data, only 265 of them have record at the first quarter of 2009, possibly due to the M&A that took place. When taking the initiative of CPP as experiment, Overall, while both recipients and non-recipients were cautious about cutting dividends, it seems that the recipients were more eager to increase the dividend payout, no matter the total amount or dividend per share. When considering the entry of CPP as experiment, it is hard to form a proper control group as the treatment happened in multiple times. The figure of those recipients are summarized in the table.
Dividend DPS
Recipient Non-recipient Recipient Non-recipient (a) the initiative of CPP as experiment
Increase Number 104 105 87 65 Fraction 39.25% 27.06% 32.83% 16.75% Decrease Number 52 80 8 36 Fraction 19.62% 20.62% 3.02% 9.28% Stable Number 44 111 119 196 Fraction 16.60% 28.61% 44.91% 50.52% NAP Number 65 92 51 91 Fraction 24.53% 23.71% 19.25% 23.45% Total 265 388 265 388 (b) enter CPP as experiment Increase Number 84 62 Fraction 31.58% 23.31% Decrease Number 65 26 Fraction 24.44% 9.77% Stable Number 64 141 Fraction 24.06% 53.01% NAP Number 53 37 Fraction 19.92% 13.91% Total 266 266
The sample group are divided into two groups based on the net-income-to-asset ratio of 2008Q3. Those whose profitability is under average is defined as under performer while those above average are defined as out-performer. Table III provides the summary statistic of the whole sample and two groups of 2008Q4. The number of out-performers
18
tripled compared with under-performers. A particularity interesting discovery is that the average profitability of the out-performers becomes negative, implying huge loss of some out-performers during this period of time.
Table III: Summary Statistics on the Sample Group
This table provides the summary statistic of the whole sample and two groups of 2008Q4. The sample group are divided into two groups based on the net-income-to-asset ratio of 2008Q3. Those whose profitability is under average is defined as under performer while those above average are defined as out-performer. The number of out-performers tripled compared with under-performers. A particularity interesting discovery is that the average profitability of the out-performers becomes negative, implying huge loss of some out-performers during this period of time.
All banks Under-performers Out-performers
N Mean Median N Mean Median N Mean Median Dividend 4662 5.8471 0.2825 1028 1.0634 0 3540 7.3737 0.472 DPS 4787 0.1007 0.05 1039 0.0474 0 3689 0.1170 0.07 Size 4967 7.2481 6.9728 1113 7.0348 6.7917 3781 7.3277 7.0339 Profitability 4961 0.0019 0.0008 1110 0.0097 -0.0016 3779 -0.0003 0.0010 Investment opportunity 4740 0.8781 0.8312 1021 0.6988 0.6468 3670 0.9290 0.8878 Signalling 4338 -1.8793 0.0049 951 -1.6699 0.0894 3297 -2.0101 -0.0072 Independence 3060 2.1278 1 593 2.9730 1 2451 1.8914 1
6. Results
A. Banks’ reaction on CPP announcement
Table IV reports the estimates of the effects that the CPP announcement has on the dividend and DPS change among banks. Column (1)-(4) displays the regression results on the whole sample. Column (5) and (6) are the results for under-performers and column (7) and (8) are the results for the out-performers with whole set of control variables.
There is no significant result on the coefficient of interest, the coefficient of
Period*Recipient, which tells the difference in the change of the dividend from 2008Q3
to 2008Q4 between recipients and non-recipients. This implies that the dividend policy of the recipients is not affected by CPP announcement. Though not significant, I’ll still take a look at what these figures tell. On the aggregate level, while the non-recipients seems increased their dividends and DPS, the dividends amount the recipients increased
19
is less than those of non-recipients and the DPS even decreased. Estimations on the CPP out-performers show the same trend, but the situation seems the other way round among the under-performs. Among the group of under-performers, though not significant, the increase in dividend among recipients is greater than non-recipients and the decline in the DPS among recipients is less that of non-recipients.
The coefficients of the control variables are reported in Table VI，signal plays a significant part in the determination of DPS among all three sample groups. This is in consistence with the conclusion of Abreu and Gulamhussen (2013) that signalling bank’s quality is of great importance in the time of crisis. However the effect is not as strong among under-performers as it is among out-performers and the aggregate group. Another special characteristic among under-performers is that size no longer matters significantly.
These tendencies are contradictory to the expectations. It is reasonable that banks, regardless of recipients or non-recipients, did not increase DPS and dividend in such difficult time. And especially it is a great relief that we don’t find evidence showing that CPP recipients increased their dividend and DPS more than they should, thus no obvious transfer of wealth from debt holders and taxpayers to shareholders. But among those under-performers, I observed a different trend –less cut in DPS and even more increase in dividends in total amount compared with non-recipients. It is precise these banks, banks that suffered great losses, that public would expect to make good use of the injection to save the business. It turns out their dividend policy falls out of public expectation and general tendency of the whole industry. In the determinants of their dividend policy, past dividend is the most important factor, profitability and signal plays a less significant role than in other two groups and size matters nothing at all. This can be explained by Bessler and Nohel (1996), as banks suffers from more severe reaction on dividend cut, those under-performing banks cannot afford to lose any investors induced by dividend reduction. So rather than size, signal, historical dividend determines their dividend payout.
Overall, the empirical study doesn't reject the hypothesis that the announcement of CPP programme has no direct effects on the dividend on common stocks of potential
20
participants. In other words, the dividend payout of the CPP recipients, though different from the under-performers and out-performers, are not affected by the expectation that they would receive injection from the government, both in terms of dividend per share and total dividend payout.
B. Banks’ reaction when injection is in effect
Table V reports the estimations of the effects of receiving CPP fund on the dividend and DPS of the full sample, under-performers and out-performers. This estimation result reports the differences between recipients and non-recipients during the period in which the injection is actually effective. The coefficient of interest – the coefficients of
Effective are negative and significant at 0.001 level in all three sample groups. These
negative coefficients imply that when the injection is effective, recipients hand out less DPS and total dividend compared with non-recipients controlling for other factors. For the full sample, as displayed in column (1) and (3), on average, the recipients hand out 0.0385 dollar less dividend per share compared with non-recipients and 6.943 million dollar less in total dividend when ignoring the factors affecting receiving CPP. When taking the factors that affect receiving CPP, i.e. size, loan loss provision, wholesale deposit into consideration, the differences in DPS and dividend become even larger. In column (2) we see those recipients hand out 0.0471 dollar less dividend per share than non-recipients and in column (4), the total cash dividend on common stock is 13.96 million dollar less.
In this estimation results, the differences in dividend payout between under-performers and out-under-performers still exist, but the divergence in different directions is gone. Both in under-performers and out-performers DPS and total dividend of recipients are less than those of non-recipients. In under-performers group, the differences in DPS and dividend of recipients is less than that of the non-recipients in absolute value. The under-performing recipients paid 0.023 dollar less dividend per share than under-performing non-recipients, but in contrast the out-performing recipients pay out 0.0521 dollar less dividend than out-performing non-recipients. In terms of total amount of dividend, the difference in under-performer group is 0.801
21
million while the difference in out-performer group is 17.19 million.
These absolute value doesn’t explain too much since two groups differs in dividend payment fundamentally during the whole sample period. As Table III presents, on average, the DPS of under-performer is 0.0474 while of out-performers is 0.1170. Compared with this average, the difference in under-performers group is 48.52% (0.023/0.0474) and in out-performers group is 44.53% (0.0521/0.1170).
These statistic results allow us to reject the hypothesis that banks increased their dividend payout using CPP money when the injection is in their bank account. Furthermore, they show that those recipients actually handed out less dividend and DPS compared with non-recipients.
Both models give no support for the argument that the money of tax payers is transferred to equity holders of the CPP recipients, at lease within these bank holding companies. Though this question is not explicitly examined before, pieces of evidence can still be found that lead to the expectation of this result. First is the term that those participants hoping to have a rise in the dividend per share must have a consent from the U.S Treasury. Banks concerned about this term may not participate the programme in the first place. Participants may have given up the idea of rise in DPS because of this, and the Treasury may have declined the request of DPS increase from those “unqualified” banks. Bayazitova and Shivdasani (2012) already show that banks’ participation decision is most significantly related to their Tire 1 ratio. The negative coefficient in their study suggests the lower the Tier 1 ratio, the more willingness for the bank to participate CPP, which implies that the banks applying for CPP money are in need of strengthening their capital ratio in order to reduce their risks as well as meet the regulatory standards. Treasury’s decision is positively related to the derivative exposures and provisional loss, trying to save the banks dancing on the edge of bankruptcy. Taliaferro (2009) also concludes that the recipients are either going to lend new loans, or in need of strengthening capital ratio. In either case, they have a clear purpose for using the injection.
22
Table IV：the Estimates of the Effects of the CPP Announcement
Table IV reports the estimates of the effects of the CPP announcement has on the dividend and DPS change among banks. Column (1)-(4) displays the regression results on the whole sample. Column (5) and (6) are the results for under-performers and column (7) and (8) are the results for the out-performers with whole set of control variables. There is no significant result on the coefficient of interest, the coefficient of Period*Recipient, which tells the difference in the change of the dividend from 2008Q3 to 2008Q4 between recipients and non-recipients. Overall, the empirical study doesn't reject the hypothesis that the announcement of CPP programme has no direct effects on the dividend on common stocks of potential participants.
All Banks Under-Performer Out-performer DPS DPS Dividend Dividend DPS Dividend DPS Dividend
(1) (2) (3) (4) (5) (6) (7) (8) Constant 0.119*** -0.177*** 2.322*** 15.13** 0.000236 -4.614 -0.195*** 16.96** (8.72) (-4.94) (4.83) (3.10) (0.00) (-1.61) (-4.90) (2.95) Period -0.0116 0.00994 -0.0648 0.631 -0.00942 0.885 0.0153 0.484 (-0.75) (0.84) (-0.10) (1.61) (-0.57) (1.93) (1.10) (1.02) Recipient 0.0370 0.0280 20.05* 0.262 -0.0131 -0.411 0.0241 0.491 (1.54) (0.81) (2.41) (0.33) (-0.81) (-0.85) (0.56) (0.48) Period*Recipient -0.0231 -0.0354 2.919 -0.581 0.00752 1.055 -0.0409 -1.046 (-0.87) (-1.15) (0.24) (-0.37) (0.29) (1.35) (-1.09) (-0.52) Controls No Yes No Yes Yes Yes Yes Yes
N 1274 709 1229 660 145 134 563 526
t statistics in parentheses
23
t statistics in parentheses
* p < 0.05, ** p < 0.01, *** p < 0.001
Table V: Estimations of the Effects of Receiving CPP Fund
Table V reports the estimations of the effects of receiving CPP fund on the dividend and DPS of the full sample, under-performers and out-performers. This estimation result reports the differences between recipients and non-recipients during the period in which the injection is actually effective. The coefficient of interest – the coefficients of Effective are negative and significant at 0.001 level in all three sample groups. These negative coefficients imply that when the injection is effective, recipients hand out less DPS and total dividend compared with non-recipients controlling for other factors.
All Banks Under-Performer Out-Performer DPS DPS dividend dividend DPS dividend DPS dividend
(1) (2) (3) (4) (8) (9) (10) (11) Constant 0.0830*** -0.0821* 5.744* -31.46*** -0.0225 -2.549* -0.0816 -41.34*** (7.80) (-2.23) (2.48) (-7.52) (-0.68) (-2.11) (-1.72) (-7.59) Effective -0.0385*** -0.0471*** -6.943*** -13.96*** -0.0230*** -0.801*** -0.0521*** -17.19*** (-4.10) (-4.42) (-3.30) (-5.65) (-4.92) (-4.49) (-3.83) (-5.37) Random Effect
YES YES YES YES YES YES YES YES
Controls NO YES NO YES YES YES YES YES
24
7. Conclusion
I use a difference-in-differences approach to analyse the effect of CPP injection has on the dividends payout of those recipients compared with non-recipients. Because of the terms and conditions on increasing dividend per share after participating in the programme, which gives the expectation that intended participants would increase the dividend right before signing the agreement to participate, two models are used to analyse the effect of two situations: first, the reaction of the intended banks when they know such programme is available, and second, the response of the actual participants after they receive the injection. While the regulation from the treasury emphasises on dividend per share, mass media and the public are more concerned about the total amount of cash dividend payout, hence, both the change in dividend per share and total cash dividend payout are analysed.
The empirical results give no supports to the suspected view that the injection benefit the shareholders a great deal at the cost of taxpayers and credit holders, as shareholders receive extra dividend payment coming out of CPP injection. No significant change in DPS and total dividend is observed in the quarter when CPP was established, but a divergence in the determinants of dividends between under-performers and out-performers is observed. In general, size, growth opportunity, profitability and signal are the main determinants of DPS of banks, despite the differences. Among the under-performers, past dividend is the most important factor in determine the DPS, while size has no effect on it. The signal factor is also weaker compared with out-performers. These imply the effort of the under-performers’ to maintain a comparable amount of dividend with previous period, possibly out of the fear of losing investors. In the out-performer group, size and profitability remain the most important determinants of dividend per share. It is reasonable to say that the within these out-performers, the dividend policy are affected by financial performance.
An interesting result is found when looking at the dividend payout after the injection was received. On average, the dividend payout of the recipients is significantly less than those non-recipients. It is reasonable to say that the regulation on DPS of CPP recipients
25
served its purpose as participants, whether voluntarily or not, comply with it, and those firms intended to increase DPS opted out the programme themselves. Other studies have suggested that banks applying for CPP were in one way or another, in need of capital, which give them no reason to distribute the CPP fund to their shareholders directly.
Together I don't observe the abuse of government injection by the means of handing out “excess” dividend during the first year of the CPP among the bank holding companies. Though extreme case do exist, like the First Financial Bank, it is not the representative of the CPP recipients in general. It is observed that the dividend payout of those recipients are well regulated, in terms of both self-regulation and government regulation.
This study only reveals a small fraction of the dividend payment of the CPP recipients on a short-term basis. The sample can be extended to the whole period of CPP to study the long-term effect of the injection. It is also interesting to see if the early exit of the recipients can be attributable to the regulation on dividend policy. Furthermore, as some of the top executives from the TARP receiving firms argued, maintaining high dividend payment is a method that makes future fund raising easier. Further research can be conducted to test this argument.
26
Reference
Abreu, J. F., & Gulamhussen, M. A. (2013). Dividend payout: Evidence from US bank holding companies in the context of the financial crisis.Journal of Corporate
Finance,22, 54-65.
Acharya, V. V., Le, H., & Shin, H. S. (2013). Bank capital and dividend externalities.
NBER Working Paper
Acharya, V. V., Gujral, I., Kulkarni, N., & Shin, H. S. (2011). Dividends and bank capital in the financial crisis of 2007-2009.NBER Working paper
Basse, T., Reddemann, S., Riegler, J. J., & von der Schulenburg, J. M. G. (2014). Bank dividend policy and the global financial crisis: Empirical evidence from Europe. European Journal of Political Economy, 34, S25-S31.
Bayazitova, D., & Shivdasani, A. (2012). Assessing TAPR.Review of Financial
Studies,25(2), 377-407.
Bessler, W., & Nohel, T. (1996). The stock-market reaction to dividend cuts and omissions by commercial banks.Journal of Banking & Finance,20(9), 1485-1508. Bhattacharya, S. (1979). Imperfect information, dividend policy, and “the bird in the hand” fallacy. Bell journal of economics, 10(1), 259-270.
Black, L. K., & Hazelwood, L. N. (2013). The effect of TARP on bank risk-taking. Journal of Financial Stability, 9(4), 790-803.
Cadman, B., Carter, M. E., & Lynch, L. J. (2012). Executive Compensation Restrictions: Do They Restrict Firms Willingness to Participate in TARP?’.Journal of Business
Finance & Accounting,39(7‐8), 997-1027.
Cornett, M. M., Li, L., & Tehranian, H. (2013). The performance of banks around the receipt and repayment of TARP funds: Over-achievers versus under-achievers.Journal
of Banking & Finance,37(3), 730-746.
Dickens, R. N., Casey, K. M., & Newman, J. A. (2002). Bank dividend policy: explanatory factors. Quarterly journal of Business and Economics, 3-12. Dickens, R.N., K.M. Casey and J.A. Newman. (2003). Bank dividend policy: Explanatory factors. Quarterly Journal of Business and Economics 41, 3-12
27
Easterbrook, F. H. (1984). Two agency-cost explanations of dividends. The American
Economic Review, 650-659.
Fuller, K. P., & Goldstein, M. A. (2011). Do dividends matter more in declining markets?. Journal of Corporate Finance, 17(3), 457-473.
Jensen, G. R., Solberg, D. P., & Zorn, T. S. (1992). Simultaneous determination of insider ownership, debt, and dividend policies. Journal of Financial and Quantitative
analysis, 27(02), 247-263.
John, K., & Williams, J. (1985). Dividends, dilution, and taxes: A signalling equilibrium. Journal of finance, 1053-1070.
Li, L. (2010). TARP funds distribution and bank loan growth. Working paper Rozeff, M. S. (1982). Growth, beta and agency costs as determinants of dividend payout ratios. Journal of financial Research, 5(3), 249-259.
Taliaferro, R. (2009). How do banks use bailout money? Optimal capital structure, new equity, and the TARP.Optimal Capital Structure, New Equity, and the TARP, working
paper
Theis, J., & Dutta, A. S. (2009). Explanatory factors of bank dividend policy: revisited. Managerial Finance, 35(6), 501-508.
Veronesi, P., & Zingales, L. (2010). Paulson's gift.Journal of Financial
Economics,97(3), 339-368.
Wen, Y., & Jia, J. (2010). Institutional ownership, managerial ownership and dividend policy in bank holding companies, International Review of Accounting, Banking and
Finance, 2(1), 8-21
Wilson, L., & Wu, Y. W. (2012). Escaping TARP.Journal of Financial Stability,8(1),
28
Appendix
Table VI：the Estimates of the Effects of the CPP Announcement with Control Variables
Table VI reports the estimates of the effects of the CPP announcement has on the dividend and DPS change among banks with control variables. Column (1)-(4) displays the regression results on the whole sample. Column (5) and (6) are the results for under-performers and column (7) and (8) are the results for the out-performers with whole set of control variables. There is no significant result on the coefficient of interest, the coefficient of Period*Recipient, which tells the difference in the change of the dividend from 2008Q3 to 2008Q4 between recipients and non-recipients.
All Banks Under Performer Out-performer DPS DPS Dividend Dividend DPS Dividend DPS Dividend
(1) (2) (3) (4) (5) (6) (7) (8) Constant 0.119*** -0.177*** 2.322*** 15.13** 0.000236 -4.614 -0.195*** 16.96** (8.72) (-4.94) (4.83) (3.10) (0.00) (-1.61) (-4.90) (2.95) Period -0.0116 0.00994 -0.0648 0.631 -0.00942 0.885 0.0153 0.484 (-0.75) (0.84) (-0.10) (1.61) (-0.57) (1.93) (1.10) (1.02) Recipient 0.0370 0.0280 20.05* 0.262 -0.0131 -0.411 0.0241 0.491 (1.54) (0.81) (2.41) (0.33) (-0.81) (-0.85) (0.56) (0.48) Period*Recipient -0.0231 -0.0354 2.919 -0.581 0.00752 1.055 -0.0409 -1.046 (-0.87) (-1.15) (0.24) (-0.37) (0.29) (1.35) (-1.09) (-0.52) Size 0.0243*** -2.535** 0.00148 0.373 0.0290*** -2.875** (3.95) (-2.96) (0.26) (1.17) (4.20) (-2.80)
29 Profitability 1.993*** -8.586 0.739** 14.00 2.648*** -2.037 (5.61) (-0.31) (2.86) (1.65) (4.19) (-0.02) Opportunity 0.0705** 2.941* 0.0330* 2.094 0.0620* 3.526* (3.23) (2.23) (2.20) (1.96) (2.53) (2.19) Signal 0.0000690** 0.00467 0.00238* 0.124 0.0000707** 0.00472 (3.07) (1.13) (2.08) (1.21) (3.03) (1.17) Regulation 0.00329 -0.0815 -0.00205 -0.00108 0.00365 -0.0890 (1.95) (-0.95) (-0.66) (-0.02) (1.78) (-0.82) Independence 0.000841 0.0618 0.000971 0.0680 0.000126 0.0738 (0.66) (0.81) (0.54) (1.52) (0.07) (0.66) Past dividend 1.040*** 0.719*** 1.042*** (60.10) (4.86) (60.19) Past DPS 0.0879 0.766*** 0.0578 (1.13) (6.43) (0.92) N 1274 709 1229 660 145 134 563 526
