Taxation of Dividend in the Banking Sector
Cordy van Werven, s3526291
First supervisor: Prof. Dr. L.J.R. Scholtens
Co-supervisor: dr. N. Selmane
Field keywords,
Banks, Dividend Pay-out Policy, Dividend Tax Policy, Cash Dividends, Share Repurchases, 2008 Financial Crisis, Law systems
Abstract,
This thesis examines the effect of dividend taxation policy of countries on the dividend pay-out policy of banks in OECD countries. It finds that the relevance of cash dividend pay-outs is significant. However, the substitution theory that is expected to go along with this finding does not hold. Banks seem to have a preference for share repurchases in the case of excessive cash. This study also takes into account the legal origin of the banks, indicating both findings mentioned better hold for code law countries than common law countries. Finally, this study takes into account the impact of the financial crisis on dividend policy of banks and finds that banks pay significant less dividends as a result of the financial crisis with as main factor regulation.
1
1. Introduction
"The harder we look at the dividends picture, the more it seems like a puzzle, with pieces that just do not fit together" (Black, 1976). This thesis starts with a famous quote of Fisher Black about dividend policy. This quote is an inspiration to narrow down the dividend topic, trying to solve a small piece of the overall puzzle. The piece of the puzzle tried to be solved in this thesis is the effect of dividend tax policies of OECD countries on the dividend policy of banks. Dividend tax policies have been a topic of discussion since their existence. It has been reviewed as an unfair double taxation of the investor (Chetty, and Saez, 2005). The profit of a company is taxed twice, at the company level as corporate income tax and when this profit is distributed to shareholders as personal income tax. The dividend payment of banks is in particular interesting as more banks pay-out dividends relative to firms in other industries. Dickens, Casey, and Newman (2002) found that in the year 2000 92 percent of the U.S. banks paid dividend, while 49 percent of the non-financial firms paid dividend. The scope of this thesis will be how banks cope with the dividend taxation policies of countries. In the decision to pay out dividend, banks can either pay its shareholders in cash dividend or, repurchase shares to increase the stock value of outstanding shares and result in capital gains for its shareholders. Both forms of dividend pay-out have different tax consequences for the shareholders, in general capital gains are taxed at a lower tax rate compared to cash dividend. This dividend decision will impact the net profit of investors.
2
effects tax reforms (Casey, and Dickens, 2000; Grullon, and Michaely 2002). This thesis is new in its kind as it will examine the direct effect of tax rates on the dividend pay-out policy of banks. External factors, such as the crisis and law system of countries, will be examined separately to see their influence as well. The overall question answered in this thesis will be:
“What effects do a country’s dividend tax policies have on bank shareholder pay-out policies?”
This topic is applicable to the International Financial Management subject. The international aspect is the comparison of different taxation policies of countries, based on their tax rates as well as their legal origin. The financial and management aspects are the dividend pay-out forms and the managerial decision to cope with different levels of tax rates. This thesis will be done with an unbalanced panel data collected over the period of 2001- 2017 from OECD countries. From this data multiple regressions will be constructed comparing different types of income taxes. The results of the regressions will be compared with existing literature to see whether this literature is applicable for banks on the matter of dividend pay-out policy in form.
3
2 Background
This paragraph will explain the exceptional role of banks in the economy and describe the reason that banks are distinctively different from other industries.
Banks are an important actor within the transmission mechanism. The transmission mechanism is the process through which monetary policy decisions affect the economy in general and the price levels in the economy. A characteristic of the transmission mechanism, as described by the European Central Bank (2018), is that time lags are long, variable, and uncertain. This makes it difficult to predict the precise effect of monetary policy actions, such as taxes, on the economy or price levels. An example of such a policy is the supply of money from central banks affecting the interest rates that banks have to pay on money they borrow. Actual effects are hard to measure due too many factors influencing money demand. The effects of policies from a central bank can often only be used for short-term effects (Taylor, 1995). For long-term results regulation from governments or international agreements (constructed by the Basel committee for example) are used. The short-term influencing decisions of central banks to steer the monetary market is driven by economic factors such as GDP and inflation rates (Taylor, 1995). The significance of the effects of monetary policy on the lending task of a bank is confirmed by Kashyap and Stein (2000), who also found that smaller banks are more impacted by these policies than bigger banks. Pruteanu‐
4
Banks play an important role in the economy, such as their role in the transmission mechanism. A survey by Fry, Julius, Mahadeva, Roger, and Sterne (2000) among 89 central banks underlines the statement, that banks play an important role in the economy, in which they found that ‘analysis of the banking sector’ was ranked high (seventh) on the most important elements of the monetary framework. Another part of the function of banks is to provide money to the public through loans and investments. The profit made from investments and return on loans is distributed among employees, managers, and shareholders, to the extent that regulation requirements provide space. The pay-out policy of these profits are the object of investigation in this thesis.
Banks have contradictory goals simultaneously. They have to apply to special regulation to ensure their stability and integrity. On the other hand, they are a profit seeking privatized organization that have as goal to maximize the return on investments for shareholders. The goal to maximise profit, as desired by the owners, is in contrast with the goal of stability and integrity as it implies taking risks and a focus on gains. A bank has to make a trade-off in the share of the profits either distributed to the shareholders and management, and the investments in stability. The stability of a bank is also indicated by its valuation by the market as this will either increase or lower the cost of capital due reputation. The dividend payed is one of the factors determining the valuation of banks. Dividend pay-outs can be used for strategic goals, such as signing stability to the market. This information is used for the valuation of the bank by outsiders.
5
6
3. Literature review
This chapter will discuss existing literature on dividend policy, tax policy, and connected theories, with the focus on the banking sector. In order to include relevant studies for this particular research with its sample data from OECD countries, most of the studies will be focused on one or more OECD countries. The hypotheses used to answer the overall question of this thesis will be reasoned for as well in this section.
The best applicable theory explaining why banks pay dividend is the signal theory. As mentioned in the introduction, within the banking sector dividend payments are twice as common relative to other industries (Dickens, Casey, and Newman, 2002). This suggests a higher importance for banks to signal to the capital market than other industries. Another reason to pay dividends is to get rid of excessive cash within a company. This excessive cash would otherwise be wasted by management for own interest or less profitable projects (Jensen, 1986). The products offered by a bank might also influence the dividend pay-out level of banks, Ameer (2008) finds differences in the pay-out of banks that sell a non-interest based banking product and banks that sell a mix of both interest and non-interest based banking products. A possible reason for this, as mentioned by Ameer (2008), is the risk factor with the offered products. This thesis will focus on more general applicable theories.
7
requirements lowers the incentive to exceed these requirements as it already gives a feeling of security for banks. Lowering dividends will give negative messages (signal) to the market about its financial conditions. Even if the dividends are not reduced, it is difficult for the management of a bank to raise provisions, because of the need to avoid sending negative signals about the quality of its loan portfolio. As a consequent of this, banks will choose a level of provisions that minimizes negative effects (Rajan, 1994).
According to taxes are an important factor for banks in their dividend policy. Taxes reduce the profit of firms and shareholders. These are paid at multiple levels, on a corporate level and on an individual level, and the effect of taxes on dividends therefore can be measured with multiple tax rates. To see whether these tax policies are indeed an important factor for dividend pay-out of banks, which is assumed as higher tax rates reduce profits and therefore reduces available resources for dividends, the first hypothesis will be:
- H01: Higher tax levels in OECD countries will lead to significant lower pay-outs of dividend in the banking sector.
8
changes in tax system to generate significant observable effects in managerial behaviour. Their paper investigates whether new tax credit systems for investors would increase the dividend pay-out of firms and so show a preference of investors towards cash dividends. Even after allowing for new equity issues, the Australian tax reform increased dividend pay-outs. Casey, and Dickens (2000) reflect upon the U.S. Tax Reform Act equalizing the tax rates for cash dividends as well as realized capital gains at 33 percent as of January 1st, 1988. Their paper is rare as it is one of the few papers concentrating on the dividend pay-out policies in the banking sector with as main independent variable taxes. Three possible reactions in the banking sector are examined, which are increased, decreased, or no changes in cash dividend pay-out policies. An Argument supporting an increase is the greater preference for cash dividend payments as they are, from then on, no longer taxed at a higher rate than capital gains (positive tax argument). Because of the repeal of investment tax credit due the act, this could decrease investments, as greater cash flows for projects are desired, “left-over” funds are bigger and can be paid to investors. On the other hand, these investment tax credit repeal could decrease dividend pay-outs. The losses from the investment tax credit could be compensated with less dividend pay-outs to maintain cash flow levels. Finally, they argue that due tax avoidance vehicles the tax irrelevance theory is a possible scenario as well (Casey, and Dickens, 2000).
9
are paid by companies. This thesis will research this effect in the banking sector with the following hypothesis:
- H02: In OECD countries with lower dividend tax relative to other OECD countries, banks will have significant higher cash dividend pay-outs.
10
led to a tripling of stock repurchase programs in the year after. The transaction costs associated with raising capital in the signalling theory indicates that stock repurchases and cash dividends are perfect substitutes as these costs are not related to pay-out policy. Some studies contradict this by taking taxes into account as well and suggest therefore that stock repurchases are a less costly alternative to cash dividends, as these are often submissive to higher tax rates (Grullon, and Michaely, 2002).
Stock repurchase announcements are an alternative way to give positive signals to the market and increase the stock value. As stock repurchases are occasionally while cash dividends are expected to be steady every year, they have a lower and more short-term impact on the stock value. According to Stephens and Weisbach (1998), firms on average acquire 74 to 82 percent of the shares announced as repurchase targets within three years of the repurchase after the announcement. Furthermore they find that share repurchases are positively related to the cash flow level and negatively related to prior stock price performance, suggesting that firms increase their share repurchasing depending on its degree of perceived undervaluation.
According to Skinner (2008) three groups of dividend payers have evolved from the past three decades. These are firms that pay dividend and make regular repurchases, firms that only make regular repurchases, and firms that make occasional repurchases. The firms that only pay dividends are largely extinct these days. Share repurchases are now the dominant form of dividend pay-out method. The level of share repurchases is fundamentally determined by earnings, while the timing is subject to multiple other factors such as stock prices and cash holdings (Skinner, 2008).1
11
For the period of 2001-2005, Eije, and Megginson (2008) found that in Europe as well as in the U.S. dividend payments and stock repurchases are complements of each other. In the years prior to 2000, U.S. had a higher proportion of dividend in stock repurchases relative to Europe, but stock repurchases made a more rapidly grow in Europe in order to catch up with the U.S. pay-out policy. Grullon, and Michaely (2002) indicate in their study that in the two decades prior of their study share repurchases in the U.S. experienced an extraordinary growth. Share repurchase expenditures grew from 4,8 percent in 1980 towards 41,8 percent in 2000, and an average annual growth rate of 26,1 percent. While dividends grew at an average annual rate of 6,8 percent in the same time. This resulted in share repurchases being 113,1 percent relative to cash dividends in 2000, while being only 13,1 percent relative to cash dividends in 1980. In the two decades before 2000, firms did not cut or reduce cash dividend payment but instead lowered the growth rate of cash dividends and shifted this growth rate towards share repurchases (Grullon, and Michaely 2002). These findings indicate a preference of a pay-out policy increasingly focused on share repurchases.
12
The policy in many countries to tax dividends at a higher rate than capital gains promotes share repurchases as a manner of investment return. Therefore it is expected that higher tax rates will lead to more share repurchases in order to optimize returns and keep tax payments low as desired by investors (Dittmar, 2000). This will be tested with the following hypothesis:
- H03: Higher dividend tax rates in OECD countries will result in significantly more share repurchases by banks.
13
effectiveness of this approach in the banking regulation has earlier been confirmed by a study of Shrieves, and Dahl (1992). A counterargument is that regulatory restrictions increase bank’s risk-taking incentives by reducing their charter value. Therefore capital restrictions for banks do not have the intended effect and do not increase a bank’s stability and should not be considered a good indicator of regulation pressure (Gonzalez, 2005). Guntay, Jacewitz, and Pogach (2015) underline the importance of regulation system for banks on the matter of dividend payment .
14
Common law countries, such as the U.S., still seem to pay more cash dividends relative to code law countries, such as the mainland of Europe, but both law types have an equal level of share repurchases (Eije, and Megginson, 2008). Stock repurchases are not always a fitting substitute for cash dividends, this does not hold for special dividends. The core value of special dividends is often not the dividend itself, but the rights attached to them (Grullon, and Michaely, 2002). La Porta, et al. (2000) find an insignificant weak positive association between dividend pay-outs and growth rates in code law countries. This is consistent with the “substitutes” theory applying to these countries. This theory suggests that shareholders in code law countries prefer to keep cash in the firm (La Porta, et al., 2000). One explanation for the substitute of cash dividends by share repurchases is that it makes it easier for companies to get rid of excessive cash without making dividend highly volatile due excessive cash difference per year (Jensen, 1986). As dividends are used to signal the status of a company, a high volatility is not desired and can cause a penalty for the firm in terms of valuation. Therefore firms use stock repurchases instead, as the market does not expect such a distribution on a regular, e.g. annual, basis.
Chetty and Saez (2005) furthermore argue that tax on dividends can have the political desired effect of wealth redistribution. However it also reduces the net return of investors, potentially reducing savings and capital stock in the economy or counter cash flows in the economy as companies retain their profits. The type of law system can be viewed upon as an indication of difference in culture between countries. This difference could be found in how banks in these two type of law systems cope with tax rates. In order to find this difference the following hypotheses has been constructed:
15
The financial crisis of 2008 was largely blamed upon the banking sector. This is an example of how risk-shifting by banks make them a governmental concern. When banks fall depositors will lose their deposits (except for the ensured part by the government) or the banks should be bailed out by the government, shifting the risk towards the tax-payers (Kanas, 2013;Onali, 2014). As a response to the crisis of 2008 regulators took action by enforcing stricter regulations on banks, in order to prevent such a crisis from happening again. Before the crisis of 2008, there was a deregulation going on for banks in the U.S. Examples of this deregulation are the Riegle-Neal Act of 1994 and the Gramm-Leach Bliley Act in 1999. After the crisis risk measurements were taken into inspection to see whether they appropriately measure the risk factor for banks. The goal of a risk measurement for banks is to assign a single numerical value to the random loss of an assets portfolio. This value will give an estimation of what future outcome can be expected from an assets portfolio in terms of riskiness. The riskiness can be countered with an assurance of none or low level risk holding assets. This method of risk measurement used for banks is known as the Tier-1 capital ratio in which the equity and disclosed reserves are set off against risk-weighted assets. A good external risk measure should be robust with respect to model misspecification and small changes in data. This measurement in turn can be used by regulators to determine appropriate level of risks a bank should be able to take (Kou, Peng, and Heyde, 2013).
In 1974 the bank governors of the G-102, Luxembourg, and Spain established the Basel Committee on Banking Supervision (BCBS), after a period of disturbances in international currencies and the banking market. The goal of this committee was to enhance financial stability by improving the quality of banking supervision and serve as a forum for cooperation on banking
16
regulation matters among member states. This committee does not issue binding regulations itself, but instead gives standards and guidelines which are expected to be followed up by member states. The membership consists of 45 institutions with 28 jurisdictions, among others China, the E.U., and the U.S. It should be kept in mind that, even while common as well as code law accepted the Basel accords, there is a morale difference between the two types of law. Code law countries are more focused on fairness and equal wealth distribution compared to common law countries. This can be noticed in differences in law on a national level La Porta, Lopez-de-Silanes, and Shleifer (2008). The first Basel accord of 1988 was focused on capital adequacy stating that banks were required to have a minimum ratio of capital to risk-weighted assets by 1992. The Basel II accord was published in 2004. The new accord gave some more depth into the regulation and consisted of three different pillars, which were a minimum capital requirements, supervisory review of capital adequacy and internal assessment process, and effective use of disclosure to strengthen market discipline and encourage sound banking practices. As a reaction to the financial crisis of 2008, that particularly hit banks hard, the Basel committee implemented the final and currently active Basel III accord in 20103. The new accord mainly reinforced the existing Basel accords with higher requirements and lessons learned from the financial crisis. That it is focused on the financial crisis of 2008 is noticeable due the measurements to counter poor risk management, low capital buffers, and inappropriate incentive structure. In 2022 a Basel IV accord is planned to be implemented with as core measurement higher capital buffers. One of the counter measurements in Basel III, to counter inappropriate incentive structure, is a minimum of capital buffers (a common equity requirement of 7%). When this requirement is not met, banks face restrictions on pay-out of cash dividends, share buybacks, and management bonuses. (Bank for International
17
Settlements, 2018). Before the final Basel accord a study of Guntay, Jacewitz, and Pogach (2015) shows that banks already took capital adequacy levels into account on matters such as dividend policy. The capital adequacy level was already regulated before the Basel accords on a national level.
Dividends can be used as valuation of a bank as it is a source for information about the liquidity, stability, and profitability of a bank (Kauko, 2011; Oliveira, Schiozer, and Barros, 2014). A practical example of this information is the liquidity of a bank during a crisis. Depositors desire banks to be liquid during a crisis to cope with possible bank runs as a result of this crisis. In order to signal liquidity to shareholders and keep depositors calm, a bank might increase its dividend pay-outs as a sign of liquidity. Forti, and Schiozer (2015) confirm that one of the main targets of dividend signalling are depositors. Signalling is a tool used by banks to counter the effects of a crisis. But the crisis also resulted in stricter regulation and dividend cuts as form of punishment making this tool ineffective for banks. However, according to Basse, Reddemann, Riegler, and Von der Schulenburg (2014), who write about the worries of the banking sector about dividend cuts after the 2008 financial crisis, banks should not fear dividend cuts or dividend omissions. Their study indicates that the economic phenomena does not hold for European banks contradicting the earlier stated importance of signalling in the banking sector.
18
including less profit distribution to management and shareholders. This should be represented in the dividend pay-out policy of bank as well, therefore a fourth hypothesis is constructed:
19
4 Model
This chapter will describe the model that will be used for the regression analyses, on what base it is constructed, and its assumptions.
In order to test the hypotheses the following three simple linear models will be used:
𝑇𝑂𝑇𝐴𝐿𝐷𝑖𝑡 = 𝑎0,𝑖𝑡+ 𝑎1𝑇𝑗𝑘𝑡+ 𝑎2𝐿𝑗+ 𝑎3𝑃𝑖𝑡+ 𝑎4𝐶𝑖𝑡+ 𝑢𝑖𝑡 (1)
𝐶𝐴𝑆𝐻𝐷𝑖𝑡 = 𝑎0,𝑖𝑡 + 𝑎1𝑇𝑗𝑘𝑡+ 𝑎2𝐿𝑗+ 𝑎3𝑃𝑖𝑡+ 𝑎4𝐶𝑖𝑡+ 𝑢𝑖𝑡 (2)
𝑅𝐸𝑃𝑈𝑅𝐶𝐻𝐴𝑆𝐸𝐷𝑖𝑡 = 𝑎0,𝑖𝑡+ 𝑎1𝑇𝑗𝑘𝑡 + 𝑎2𝐿𝑖𝑡+ 𝑎3𝑃𝑖𝑡+ 𝑎4𝐶𝑖𝑡+ 𝑢𝑖𝑡 (3)
TOTALD= The total dividend paid, in cash dividend as well as share repurchases, in ratio to assets.
CASHD= The dividend paid in in the form of cash dividend in ratio to assets.
REPURCHASED= The dividend paid in the form of share repurchases in ratio to assets.
𝑎0 = Constant, 𝑎1,2,3… = Variable coefficients, 𝑖 = bank, 𝑗 = 𝑐𝑜𝑢𝑛ry, 𝑡 =year, 𝑘 = type of tax
T= Taxes, measured at three different levels (𝑘) at a company level (corporate income tax), at an
20
L= Law form as dummy variable, either common law (1) or code law (0) (+, expected to have a positive effect on every type of dividend) (La Porta, Lopez-de-Silanes, and Shleifer, 2008).
P= Profit in the form of total income before tax as a ratio to assets (+, expected to have a positive impact on every type of dividend) (Lintner, 1956).
C= Control variables which will be:
- Log of the total assets (+, expected to have a positive impact on all forms of dividend) (Casey, and Dickens, 2000).
- Cash on accounts ratio to assets (-, expected to have a negative impact on all forms of dividend).
- Growth over the past 5 years calculated as an average compound over the past 5 years (-, expected to have a negative impact on all forms of dividend) (Casey, and Dickens, 2000). - Market- to- book value (+-, expected to have a positive influence on cash and total dividend
payments, but a negative influence on share repurchases. The reason for the last statement is that share repurchases are used when shares are undervalued according to theory) (Dittmar, 2000).
- Return on assets (+, expected to have a positive effect on all forms of dividend) (Kanas, 2013).
- Tier- 1 capital ratio (+, expected to have a positive effect on all forms of dividend) (Forti, and Schiozer, 2015)
21
22
banking sector are taken into account in this thesis. The choice to use different variables than Lintner’s (1956) original model is commonly done by articles cited in this paper. Linter’s model gives a good indication on how dividend payments are influenced, in particular by profits. The model used in this paper is adjusted according to many papers citing Lintner (1956) with adjustments based on articles about tax influence on dividends and the banking sector to fit the objective of research, namely the effect of tax policies on dividend pay-out policies in the banking sector in OECD countries.
The study of the direct effect of tax on dividend payment and form of dividend payment is rare, in particular the focus on the banking sector. Rehman, and Takumi (2012) is one of the few studies found using dividend pay-outs as dependent variable and taxes as independent variable to measure a direct effect and comparing different tax regimes, but with a more complicated model. Law is taken into account as an important variable to measure regulation, which is an important aspect in the highly regulated banking sector. Studies that take law into account as independent variable on dividend tax are for example Barclay, Smith, and Watts (1995), Casey, and Dickens (2000),La Porta, Lopez‐de‐Silanes, Shleifer, and Vishny (2000), Dickens, Casey, and Newman (2002), and
Kanas (2013). The effect of profit on dividend is taken into account by almost every study measuring effect on dividends, e.g. Lintner (1956). The log of assets is used in order to take the size of banks into account as done by, among others, Grullon, and Michaely (2002), and Pattenden and Twite (2008). The excessive cash theory is controlled for with the total available cash on accounts as done by Grullon, and Michaely (2002). Past growth is taken into account as done by Von Eije, and Megginson (2008), and La Porta, Lopez‐de‐Silanes, Shleifer, and Vishny (2000)
23
24
5 Data
In this chapter the data characteristics will be described. It will exemplify changes made on the data and implications in the research as a result of the sample data.
The data used in this paper is retrieved from Datastream and the official online database of the OECD countries (OECD database), gathered in the month October 2018. Table 1 gives an overview of the variables used including a description based on the description provided by the source, the expected sign according to theory, the unit it might have been transformed to, whether it was winsorized, measurement of frequency, and the source.4 All continues variables used, except for the dependent variables and the log of assets have been winsorized at 1% to transform outliers at bottom and top of the data. This means that all outliers, the top and bottom 1%, are transformed to the value of the observation at marginal 1% and 99%. The reason that the independent tax variables (CIT, PIT, and NIT) and law dummy variable are not winsorized is that these cannot be considered as continues variables, but rather discrete variables as they are a given and not a random picked value (Brooks, 2014). The log of assets was also not winsorized as there were no significant outliers after applying a log transformation and so winsorizing did not make sense. The process of removing outliers will alter the data set in a way that it is less representative for the actual sample used. This is why this process should be kept to a minimum and winsorizing is done at only 1%. As all continues variables had outliers, they all had to be winsorized. In order to make the data comparable, the dependent variables are transformed into a ratio to total assets and so are the control variables PROFIT and CASH. The dependent variables are stated as percentage. The
25
dummy variable accounts an additional value to the regression when these have a common law system.
Table 1. Variable description
Variable
name Variable description TOTDIV Exp. sign CASHDIV Exp. sign REPDV Exp. sign Unit Winsorized at 1% frequency Source Total
Dividend Paid (TOTDIV)
Cash dividend and share repurchases accumulated as a ratio to total assets
Ratio to total
assets No yearly Datastream (WC04551+ WC04751) Cash Dividend Paid (CASHDIV) Cash dividend as a ratio to total assets Ratio to total
assets No yearly Datastream (WC04551) Dividend in Share Repurchases (REPDIV) Share repurchases as a ratio to total assets Ratio to total
assets No yearly Datastream (WC04751) Corporate Income Tax (CIT) Percentage of corporate income tax on distributed profit
- - - Percentage No yearly OECD stats
Personal Income Tax(PIT) Percentage of personal income tax on dividend
- - + Percentage No yearly OECD stats
Net Personal Income Tax (NIT) Percentage of net personal income tax on dividend
- - + Percentage No yearly OECD stats
Law system
(LAW) Dummy variable for either code (0) or common (1) law
+ + + Dummy
variable No OECD info
Profit before Tax (PROFIT) Profit before taxes as a ratio to total assets + + + Ratio to total
assets Yes yearly Datastream (WC01250) Bank Size in
Assets (ASSETS0
The logarithm of
the total assets + + + Log of total assets no yearly Datastream (WC02999) Growth over
the past 5 years (GROWTH)
Growth over the past 5 years in net income in compound annual rate
- - - Percentage Yes yearly Datastream
(WC08638) Cash on Accounts (CASH) Free cash on accounts at other banks - - - Ratio to total
assets Yes yearly Datastream (WC02004)
Market-to-book value (MTBV)
Market-to-book
value + + - Ratio to book value Yes yearly Datastream (MTBV)
Return on
Assets (ROA) Return on assets + + + Return ratio to assets Yes yearly Datastream (WC08326) Risk
Measurement (TIER1)
Tier 1 capital ratio to total risk-weighted assets
+ + + Ratio to total
risk weighted assets
Yes yearly Datastream
26
The correlation between variables and the expected effect of variables on the dependent variables is checked through a Pearson correlation matrix, Table 2. The benchmark of a high correlation is set upon 0,4 or higher as commonly done by researchers in this area Values of 0,4 or higher will get additional attention to explain this correlation and check its impact on results. As TOTDIV is the accumulation of CASHDIV and REPDIV, we see a high significant correlation which makes sense that will not have any negative consequences in this paper’s results. There also is a high significant correlation between NIT and PIT, this is logical as NIT is PIT including the benefits of cash credits and will also have no negative consequences for the results. More surprising is the correlation between CIT and LAW, this suggests a link between law system and the height of corporate income tax. An explanation for this observation is the difference in tax culture.5 This link on itself will not directly influence the results of this research, but should be taken into account in the interpretation of the results. More problematic are the signs of CIT, PIT, and NIT which are all expected to have a negative effect on the depend variables TOTDIV and CASHDIC but according to the matrix only a direct negative correlation is found for CIT and CASHDIV. This does not have to mean that the regression coefficients have the same signs. PROFIT shows a positive significant correlation with TOTDIV and CASHDIV, this is explained in the theory as main influential factor on dividend and as theory already predicted this is not the case for REPDIV. ASSETS has a negative direct effect on the dependent variables TOTDIV and CASHDIV which also was not expected. MTBV has a high correlation with CASHDIV, this can be explained in the theory section as CASHDIV is used to give a positive signal to the market of the performance of a company and so relative improve MTBV. The actual objective performance of a bank is measured with PROFIT, this explains the high significant correlation with the
27
Table 2. Correlation matrix
T O T D I V 1 - - - - - - - - - - - - - - - - - - - - - - - - - - CA SHD I V 0,666*** 1 -- -- -- -- -- -- -- -- -- -- -- --RE PD IV 0, 793*** 0, 107*** 1 -- -- -- -- -- -- -- -- -- -- --CI T 0, 080*** -0, 101** 0, 198*** 1 -- -- -- -- -- -- -- -- -- --PI T 0, 067*** 0, 092*** 0, 013 - 0,327*** 1 -- -- -- -- -- -- -- -- --NIT 0,041** 0,003 0,068*** - 0,254*** 0,777*** 1 -- -- -- -- -- -- -- --LA W 0, 225*** 0, 189*** 0, 187 *** 0, 445*** 0, 251*** 0, 217*** 1 -- -- -- -- -- -- --PROF IT 0, 455*** 0, 568*** 0, 166*** 0, 011 0, 102*** 0, 036 0, 288 1 -- -- -- -- -- --AS SE TS -0, 009 - 0,118*** 0,086*** 0,054** 0,163*** 0,122*** 0,017 -0, 286 1 -- -- -- -- --G ROW T H -0, 013 -0, 015 -0, 005 0 ,046** 0,061*** 0,049** 0,052** 243***0, - 0,084*** 1 -- -- -- --CA SH - 0,092*** -0, 036 - 0,116*** - 0,274*** - 0,113*** - 0,094*** - 0,139*** -0, 009 0, 001 -0, 040* 1 -- -- --M TBV 0, 361*** 0, 480*** 0, 116*** -0, 003 0, 218*** 0, 100*** 0, 278*** 0, 620*** - 0,141*** 0,197*** - 0,131*** 1 -- --ROA 0,374*** 0,535*** 0,078*** - 0,239*** 0,227*** 0,083*** 0,091*** 0,715*** - 0,198*** 0,253*** -0, 027 0, 531*** 1 --TI ER1 0, 070*** 0, 114*** -0, 015 - 0,257*** 0,009 0,138*** - 0,172*** 0,127*** - 0,222*** 0,028 0,213*** -0, 031 0, 0 31 1 Var iab le s TOT DI V C ASHDI V RE P D IV C IT P IT NI T L AW P R OFI T ASS E T S GR OW T H CASH MT B V R OA T IE R 1
28
performance indicator as perceived by the market through MTBV. ROA is a second objective performance measurement. A steady CASHDIV requires a corresponding ROA to be able to pay-out the dividends in cash year by year. As it is a performance measurement just like PROFIT, it correlates highly with PROFIT. Finally TIER1 seems to be negatively correlated with REPDIV, which is not as expected and will have a further review in the results.
Table 3 shows the descriptive statistics of the variables used in this paper on an individual level. The data used is unbalanced panel data over 17 years including 262 individual banks. Examining the amount of observations it can be noticed that TIER1 has significant lower observations than other variables. The reason is unavailability in the data source of DataStream. As TIER1 is compulsory for banks to report due legislation, the unavailability is surprising. When examining the raw data it shows that this is particular missing for Asian countries such as Japan and Korea. A possible explanation could be the “reporting culture”6 of these countries in which the “reporting culture” is not as strong as in the rest of the world. Appendix 1 shows that these two countries represent 20,78% of the total observations. Besides this the raw data shows that most of the missing values for any variable are of the first years of the testing period (mainly 2001-2005). Apparently, in this period the sources used in this paper putted less effort in keeping track of the used data or this data was unavailable for this period. Most of the standard deviations are high relative to the average. This means that the measured data observations are far away from the mean. This shows that the data has a lot of variation but is not by definition a bad thing, and as the data is corrected for outliers through winsorizing the variation is partly corrected for. A negative value for the minimum of NIT is reported as Norway had a few years in which tax credits exceeded the PIT,
29
resulting in a surplus for domestic investors, to stimulate domestic investments from the population in Norway, this only held stand from 2001 till 2005 and affects 12 observations. The skewness and kurtosis indicate that the data is barely normal in most cases. Only the variables CIT, PIT, NIT, ASSETS, could be argued for to be normal according to this table. The data in this paper is normalized to minimize the impact of the non-normality issue.
Table 3. Descriptive statistics
TOTDIV CASHDIV REPDIV CIT PIT NIT LAW PROFIT ASSETS GROWTH CASH MTBV ROA TIER1
N 3566 3566 3566 3566 3566 3554 3566 3564 3566 3974 3489 3380 3247 2696 Average 0,427 0,268 0,146 30,801 27,565 23,791 0,349 0,976 17,736 12,683 4,438 1,366 1,048 11,918 Median 0,244 0,172 0,001 30,000 26,000 22,000 0,000 0,861 17,543 10,275 2,946 1,195 0,960 11,500 Std. 0,548 0,320 0,348 7,753 12,530 10,462 0,477 0,962 1,713 26,806 4,418 0,796 0,984 3,814 Minimum 0,000 0,000 0,000 9,000 0,000 -0,010 0,000 -2,239 12,063 -53,048 0,036 0,160 -1,983 6,045 Maximum 3,176 1,780 2,142 46,100 55,640 51,0000 1,000 4,293 22,052 124,980 23,722 4,137 5,161 27,008 Skewness 2,503 2,234 3,589 -0,268 0,168 0,180 0,634 0,289 0,242 1,140 2,059 1,021 0,826 1,167 Kurtosis 10,907 9,227 17,554 1,837 2,110 2,892 1,402 5,468 2,996 6,691 7,765 3,987 6,525 5,133
This table shows the descriptive statistics of all variables included in this paper. The columns show the values per variable and the rows show the statistics in number of observations (N), average, median, standard deviation (Std.), minimum, maximum, skewness, and kurtosis.
30
disturb the results for this research with taxation as important independent variable. Appendix 1 shows the U.S. and Japan are overrepresented in this study, with accumulated representing almost 40% of all the observations taken into account. Code law is represented by 67,92%, and common law by almost the half of code law observations, 32,08%. A comparison of dividend and taxation policy calculated from the raw data can be found in table 4. The table shows that code law countries have lower average taxes compared to common law countries. This is in line with differences on a national level between the two types of legal origins as described by La Porta, Lopez-de-Silanes, and Shleifer (2008). The total dividend payments and cash dividend payments of common law countries is higher compared to code law countries. This is in line with the theory of Eije, and Megginson, 2008. Nevertheless, common law countries also seem to pay more dividend in the form of share repurchases, while Grullon, and Michaely (2002) argue that share repurchases should be on an equal level nowadays.
Table 4. Averages of the tax and dividend variables per law system.
Variables Code law Common law
Average CIT 28,84% 34,48% Average PIT 25,33% 31,50% Average NIT 21,96% 27,34% Average TOTDIV 0,49% 0,60% Average CASHDIV 0,39% 0,34% Average REPDIV 0,10% 0,26%
31
6 Methodology
In this chapter the methodology to calculate the results in this thesis will be explained, including the assumptions.
This thesis uses unbalanced panel data, which is a dataset existing of time-series and cross-sectional elements and missing some data values. An OLS estimation will be used to calculate regressions. The OLS method is justified as the model estimated is a linear model. Unfortunately, the data is significant for heteroscedasticity. In order to reduce this problem, this did not remove the heteroscedasticity, the data has been winsorized at 1% and normalized. Because of the heteroscedasticity the standard errors, and so the significance levels, of the coefficients might be wrong. An additional advantage of normalized data is that the coefficients are better comparable relative to each other in terms of p-value to measure the impact. The data also has positive autocorrelation in it as the Durbin-Watson tests indicate a value below 1,42. The values of the Durbin-Watson tests can be found at the bottom row of tables 5,6, 7, 9, 10, and 11. The consequence of failing these two tests are that, even while the OLS coefficients are still unbiased, the errors could be wrong and so a wrong inferences could be made on whether a variable is important in the regression. In order terms, the significance level cannot be taken fully serious anymore. However, the data is distributed more or less normal due to the law of large numbers. The average of the sample mean will converge to the population mean and so converges to a normal distribution. This is confirmed by plots of the data7 (Brooks, 2014). Time fixed effects are used to control for aggregate trends that do not have an impact on the relationship, for example the financial crisis of 2008 and its ravage in the following years.
32
33
7 Results
In the tables 5, 6, and 7 the OLS regression with fixed time effects and normalized data can be found. The first models show the effect of the control variables on a dependent variable to see whether these fit the model. For model 1(a) all control variables are highly significant and are a good predictor for the dependent variable, as expected according to the theory. For model 1(b) only the variable CASH is not significant. For model 1(c) GROWTH and ROA are less significant and ROA shows an unexpected sign. Model 1(d) confirms the findings of model 1(c). ROA seems not to be a good estimation for share repurchases. This can be reasoned for as the decision to repurchase share is the result of multiple years while ROA only measures one year. Therefor CASH would be a better predictor for share repurchases as this is the result of savings over multiple years. All models of table 5, 6, and 7 show that profit is the main factor explaining dividend showing that Lintner’s (1956) theory is still relevant these days and also applicable for the banking sector. The variable cash is of higher and more significant impact for the models on share repurchases compared to the models on cash dividends. This is contradictory to the “outcome” theory of La Porta, Lopez‐de‐Silanes, Shleifer, and Vishny (2000) that does not make a distinction
34
confirms the theory of Grullon, and Michaely (2002) for banks. Tier 1 capital seems to be a good predictor of any form of dividend payment, which is the result of regulation that sets this capital adequacy level at a minimum and that it is a good measurement for risk. Banks with higher tier 1 capital ratios experience less risk and therefor are able to pay out more dividends according to regulation (Basel accord) and according to policies described by Guntay, Jacewitz, and Pogach (2015). On the matter of taxes table 6 gives the expected results and shows that cash dividends are
Table 5. The relationship between different types of taxes and the total dividend payed by banks in OECD countries.
Variable Model 1 (a) Model 2 (a) Model 3 (a) Model 4 (a) Model 5 (a)
INTERCEPT 0,369*** (0,014) 0,316*** (0,077) 0,209*** (0,051) 0,453*** (0,027) 0,425*** (0,030) CIT 0,004* (0,002) 0,006*** (0,002) PIT -0,003** (0,002) -0,003*** (0,001) NIT 0,002 (0,002) -0,003** (0,001) LAW 0,117*** (0,023) 0,109*** (0,028) 0,084*** (0,025) 0,138*** (0,023) 0,123*** (0,023) PROFIT 0,261*** (0,024) 0,237*** (0,024) 0,244*** (0,024) 0,242*** (0,024) 0,253*** (0,024) ASSETS 0,083*** (0,011) 0,087*** (0,011) 0,082*** (0,011) 0,090*** (0,011) 0,086*** (0,011) GROWTH -0,086*** (0,012) -0,088*** (0,012) -0,089*** (0,012) -0,086*** (0,012) -0,086*** (0,012) CASH -0,032*** (0,011) -0,031** (0,012) -0,024** (0,012) -0,038*** (0,012) -0,036*** (0,012) MTBV 0,023 (0,016) 0,032** (0,016) 0,029* (0,016) 0,028* (0,016) 0,023 (0,016) ROA 0,089*** (0,020) 0,124*** (0,022) 0,115*** (0,022) 0,113*** (0,021) 0,097*** (0,021) TIER1 0,079*** (0,012) 0,082*** (0,013) 0,084*** (0,012) 0,083*** (0,012) 0,085*** (0,013) Time fixed effects
Yes Yes Yes Yes Yes
Normalized data
Yes Yes Yes Yes Yes
𝑅2 0,290 0,295 0,294 0,294 0,292
Adjusted 𝑅2 0,282 0,286 0,285 0,285 0,283
F- value 34,901*** 31,680*** 34,095*** 34,037*** 33,741***
N 2074 2070 2074 2070 2074
Durbin-Watson 0,918 0,918 0,923 0,932 0,916
35
influenced by taxes. CIT, PIT, and NIT all have a negative influence on the dividends paid by banks and so the theory of Casey, and Dickens (2000) that tax have an impact on dividend policy has been confirmed, supporting the positive tax theory. However, dividend tax seems not to be a good predictor for share repurchases. As the signs and significance levels for the variables of taxes in table 7 differ from expectations. This can also be partly found back in the results of table 5. Therefore the substitution theory of Grullon, and Michaely (2002) cannot be confirmed based on taxes by this thesis.
Because the overall dividends paid are a combination of cash dividends and share repurchases, we could not safely state that taxes negatively impact the overall dividend payment of banks. But only cash dividend are directly impacted by dividend tax policies of countries. Most studies take the combined dividend and as shown by these results, the effect differs per form of dividend pay-out. The results on the total dividend paid do confirm the statement that dividend taxes influence dividend pay-out height in total significantly as described by studies done through the impact of tax reforms (e.g. Casey, and Dickens, 2000; Grullon, and Michaely 2002).
36
Table 6. The relationship between different types of taxes and the cash dividend payed by banks in OECD countries.
Variables Model 1 (b) Model 2 (b) Model 3 (b) Model 4 (b) Model 5 (b)
INTERCEPT 0,249 (0,007) 0,540*** (0,038) 0,320*** (0,026) 0,301*** (0,014) 0,321*** (0,058) CIT -0,006*** (0,001) -0,002*** (0,001) PIT -0,002** (0,001) -0,002*** (0,000) NIT -0,003*** (0,001) -0,003*** (0,001) LAW 0,032*** (0,011) 0,094*** (0,014) 0,046*** (0,013) 0,044*** (0,012) 0,048*** (0,012) PROFIT 0,121*** (0,012) 0,119*** (0,012) 0,128*** (0,012) 0,109*** (0,012) 0,110*** (0,012) ASSETS 0,016*** (0,006) 0,025*** (0,006) 0,017*** (0,006) 0,020*** (0,006) 0,021*** (0,006) GROWTH -0,067*** (0,006) -0,063*** (0,006) -0,065*** (0,006) -0,066*** (0,006) -0,067*** (0,006) CASH -0,007 (0,006) -0,024*** (0,006) -0,011* (0,006) -0,011* (0,006) -0,012** (0,006) MTBV 0,060*** (0,008) 0,056*** (0,008) 0,057*** (0,008) 0,063*** (0,008) 0,060*** (0,008) ROA 0,111*** (0,010) 0,106*** (0,011) 0,099*** (0,011) 0,126*** (0,011) 0,121*** (0,010) TIER1 0,032*** (0,006) 0,036*** (0,006) 0,030*** (0,006) 0,034*** (0,006) 0,039*** (0,006) Time fixed effects
Yes Yes Yes Yes Yes
Normalized data
Yes Yes Yes Yes Yes
𝑅2 0,430 0,448 0,432 0,435 0,438
Adjusted 𝑅2 0,423 0,441 0,425 0,428 0,431
F- value 64,415*** 61,419*** 62,388*** 62,924*** 63,822***
N 2074 2070 2074 2070 2074
Durbin-Watson 0,509 0,502 0,467 0,500 0,491
37
Table 7. The relationship between different types of taxes and the dividend payed in share repurchases by banks in OECD countries.
Variable Model 1 (c) Model 2 (c) Model 3 (c) Model 4 (c) Model 5 (c)
INTERCEPT 0,106*** (0,010) -0,165*** (0,058) -0,084** (0,038) 0,147*** (0,021) 0,080*** (0,023) CIT 0,007*** (0,001) 0,007*** (0,001) PIT -0,003*** (0,001) -0,002** (0,001) NIT 0,007*** (0,001) 0,001 (0,001) LAW 0,100*** (0,017) 0,046** (0,021) 0,061*** (0,019) 0,110*** (0,018) 0,094*** (0,018) PROFIT 0,119*** (0,018) 0,098*** (0,018) 0,098*** (0,018) 0,110*** (0,018) 0,123*** (0,018) ASSETS 0,060*** (0,008) 0,058*** (0,009) 0,059*** (0,008) 0,064*** (0,009) 0,058*** (0,008) GROWTH -0,016* (0,009) -0,021** (0,009) -0,019** (0,009) -0,016* (0,009) -0,016* (0,009) CASH -0,027*** (0,009) -0,013 (0,009) -0,018** (0,009) -0,030*** (0,009) -0,025*** (0,009) MTBV -0,032*** (0,012) -0,020 (0,012) -0,025** (0,012) -0,030** (0,012) -0,033*** (0,012) ROA -0,027* (0,016) 0,011 (0,017) 0,004 (0,017) -0,015 (0,016) -0,031* (0,016) TIER1 0,039*** (0,009) 0,035*** (0,000) 0,044*** (0,009) 0,041*** (0,009) 0,036*** (0,010) Time fixed effects
Yes Yes Yes Yes Yes
Normalized data
Yes Yes Yes Yes Yes
𝑅2 0,106 0,127 0,117 0,108 0,106
Adjusted 𝑅2 0,095 0,115 0,106 0,097 0,095
F- value 10,084*** 10,979*** 10,867*** 9,858*** 9,750***
N 2074 2070 2074 2070 2074
Durbin-Watson 1,133 1,164 1,152 1,136 1,134
38
Table 8. A TOBIT regression of the relationship between different types of taxes and the dividend payed in share repurchases by banks in OECD countries.
Variable Model 1 (d) Model 2 (d) Model 3 (d) Model 4 (d) Model 5 (d)
INTERCEPT -0,134*** (0,018) -0,857*** (0,098) -0,790*** (0,067) 0,025 (0,032) -0,075** (0,036) CIT 0,023*** (0,002) 0,023*** (0,002) PIT -0,007*** (0,002) -0,007*** (0,001) NIT 0,011*** (0,002) -0,003* (0,001) LAW 0,183*** (0,028) 0,057* (0,034) 0,063** (0,030) 0,229*** (0,030) 0,197*** (0,029) PROFIT 0,186*** (0,030) 0,098*** (0,030) 0,106*** (0,030) 0,147*** (0,031) 0,177*** (0,030) ASSETS 0,101*** (0,014) 0,108*** (0,014) 0,106*** (0,013) 0,117*** (0,014) 0,105*** (0,014) GROWTH -0,029** (0,014) -0,042*** (0,014) -0,040*** (0,014) -0,025* (0,014) -0,028** (0,014) CASH -0,063*** (0,014) -0,028* (0,014) -0,031** (0,014) -0,071*** (0,014) -0,066*** (0,014) MTBV -0,014 (0,018) 0,015 (0,018) 0,008 (0,018) 0,002 (0,018) -0,011 (0,018) ROA -0,125*** (0,028) 0,001 (0,029) -0,013 (0,028) -0,082*** (0,028) -0,117*** (0,028) TIER1 0,033** (0,014) 0,056*** (0,015) 0,070*** (0,014) 0,048*** (0,014) 0,041*** (0,015) Time fixed effects
Yes Yes Yes Yes Yes
Normalized data
Yes Yes Yes Yes Yes
N 2074 2070 2074 2070 2074
Censored N 910 906 910 906 910
This table shows a TOBIT regression on the dependent variable dividend payed in share repurchases of banks in OECD countries. The rows show the value per variable included and model characteristics and the columns different models. The p-value is given. *, **, *** Show the significant level at 10%, 5%, 1% respectively. The standard errors are given within parentheses.
39
common law countries and insignificant for code law countries. Code law countries do not apply to the theory that taxes are an important predictor on dividends as stated by Casey, and Dickens (2000). Casey, and Dickens (2000) however studied an U.S. sample. Only CIT does not give the expected sign for common law countries, which could indicate a mistake in the model as this does not make sense. The difference in TIER 1 indicates a higher focus on regulations in code law countries compared to common law countries. The theories that dividend tax influences cash dividend as seen before holds for code law countries, but not for common law countries. Again this could be the result of a mistake in the model as CIT again does not show the expected sign. The substitution theory of Grullon, and Michaely (2002) holds for code law countries as well, but again not for common law countries. This is strange as Grullon, and Michaely (2002) focus in their study on a U.S. sample. However the model on share repurchases has many flaws as many of the control variables are not determined significant. The insignificance found between a positive growth and dividend payments in code law countries (La Porta, et al., 2000) is
40
Table 9. The relationship between different types of taxes and the total dividend payed by banks in OECD countries split in type of law system.
Variable Model 1 (e) Model 2
(e)
Model 3 (e) Model 4 (e) Model 5 (e) Model 6 (e) Model 7 (e) Model 8 (e) INTERCEPT -2,112*** (0,296) -2,107*** (-0,301) -2,114*** (0,300) -2,112*** (0,296) -0,792*** (0,300) -1,557*** (0,397) -0,739** (0,298) -0,517 (0,319) CIT -0,000 (0,005) 0,016*** (0,005) PIT -0,001 (0,002) -0,011*** (0,003) NIT -0,001 (0,003) -0,016** (0,006) PROFIT 0,370*** (0,055) 0,371*** (0,056) 0,371*** (0,055) 0,368*** (0,056) 0,684*** (0,069) 0,605*** (0,074) 0,579*** (0,075) 0,641*** (0,071) ASSETS 0,113*** (0,016) 0,113*** (0,017) 0,114*** (0,017) 0,114*** (0,017) 0,050*** (0,017) 0,063*** (0,017) 0,066*** (0,017) 0,057*** (0,017) GROWTH -0,128*** (0,026) -0,127*** (0,026) -0,128*** (0,026) -0,128*** (0,026) -0,267*** (0,037) -0,261*** (0,037) -0,261*** (0,037) -0,264*** (0,037) CASH -0,015 (0,028) -0,015 (0,028) -0,017 (0,028) -0,018 (0,029) -0,186*** (0,033) -0,148*** (0,035) -0,194*** (0,033) -0,176*** (0,033) MTBV 0,186*** (0,036) 0,186*** (0,036) 0,186*** (0,036) 0,186*** (0,036) -0,207*** (0,046) -0,188*** (0,047) -0,161*** (0,048) -0,195*** (0,047) ROA 0,144*** (0,043) 0,142*** (0,047) 0,148*** (0,045) 0,147*** (0,044) 0,232*** (0,077) 0,266*** (0,077) 0,277*** (0,077) 0,238*** (0,077) TIER1 0,158*** (0,025) 0,158*** (0,026) 0,159*** (0,026) 0,161*** (0,026) 0,062 (0,049) 0,047 (0,049) 0,041 (0,049) 0,060 (0,049) Time fixed effects
Yes Yes Yes Yes Yes Yes Yes Yes
Normalized data
Yes Yes Yes Yes Yes Yes Yes Yes
𝑅2 0,294 0,294 0,293 0,294 0,297 0,304 0,307 0,302 Adjusted 𝑅2 0,280 0,279 0,278 0,279 0,278 0,284 0,287 0,282 F- value 21,190*** 20,290*** 20,169*** 20,299*** 15,675*** 15,508*** 15,699 15,358*** N 1197 1197 1193 1197 877 877 877 877 Durbin-Watson 0,631 0,631 0,632 0,632 1,316 1,327 1,333 1,325
41
Table 10. The relationship between different types of taxes and the cash dividend payed by banks in OECD countries split in type of law system.
Variable Model 1 (f) Model 2
(f)
Model 3 (f) Model 4 (f) Model 5 (f) Model 6 (f) Model 7 (f) Model 8 (f) INTERCEPT -0,033 (0,286) 0,036 (0,290) 0,072 (0,285) -0,038 (0,284) -0,813*** (0,215) -0,450 (0,286) -0,835*** (0,215) -0,670*** (0,230) CIT -0,006 (0,004) -0,008* (0,004) PIT -0,013*** (0,002) 0,005* (0,002) NIT -0,010*** (0,002) -0,008* (0,005) PROFIT 0,381*** (0,053) 0,388*** (0,054) 0,366*** (0,053) 0,356*** (0,053) 0,384*** (0,050) 0,421*** (0,053) 0,427*** (0,054) 0,362*** (0,051) ASSETS 0,000 (0,016) 0,006 (0,016) 0,012 (0,016) 0,012 (0,016) 0,050*** (0,012) 0,043*** (0,012) 0,043*** (0,012) 0,054*** (0,012) GROWTH -0,196*** (0,025) -0,192*** (0,025) -0,198*** (0,025) -0,199*** (0,025) -0,251*** (0,027) -0,253*** (0,027) -0,253*** (0,027) -0,249*** (0,027) CASH 0,000 (0,027) -0,003 (0,027) -0,031 (0,027) -0,027 (0,027) -0,108*** (0,024) -0,126*** (0,026) -0,105*** (0,024) -0,103*** (0,024) MTBV 0,325*** (0,035) 0,318*** (0,035) 0,325*** (0,034) 0,324*** (0,034) -0,056* (0,033) -0,065* (0,034) -0,075** (0,035) -0,050 (0,024) ROA 0,246*** (0,041) 0,222*** (0,045) 0,328*** (0,043) 0,280*** (0,042) 0,546*** (0,055) 0,530*** (0,056) 0,527*** (0,056) 0,549*** (0,055) TIER1 0,082*** (0,024) 0,077*** (0,025) 0,100*** (0,024) 0,111*** (0,025) 0,084** (0,035) 0,091*** (0,035) 0,093*** (0,035) 0,083** (0,035) Time fixed effects
Yes Yes Yes Yes Yes Yes Yes Yes
Normalized data
Yes Yes Yes Yes Yes Yes Yes Yes
𝑅2 0,441 0,442 0,458 0,449 0,431 0,434 0,434 0,433 Adjusted 𝑅2 0,430 0,431 0,447 0,437 0,416 0,418 0,418 0,417 F- value 40,282*** 38,706*** 41,099*** 39,722*** 28,100*** 27,168*** 27,176*** 27,127*** N 1197 1197 1193 1197 877 877 877 877 Durbin-Watson 0,575 0,571 0,606 0,586 0,438 0,446 0,449 0,434
42
Table 11. The relationship between different types of taxes and dividend payed in share repurchases by banks in OECD countries split in type of law system.
Variable Model 1 (g) Model 2
(g)
Model 3 (g)
Model 4 (g) Model 5 (g) Model 6 (g) Model 7 (g) Model 8 (g) INTERCEPT -2,825*** (0,315) -2,769*** (0,320) -2,915*** (0,317) -2,820*** (0,314) -0,518 (0,408) -2,018*** (0,538) -0,419 (0,403) -0,240 (0,435) CIT -0,005 (0,005) 0,031*** (0,007) PIT 0,010*** (0,002) -0,021*** (0,004) NIT 0,009*** (0,003) -0,016* (0,009) PROFIT 0,124** (0,059) 0,130** (0,059) 0,138** (0,059) 0,149** (0,059) 0,695*** (0,094) 0,541*** (0,100) 0,497*** (0,102) 0,652*** (0,097) ASSETS 0,151*** (0,018) 0,156*** (0,018) 0,143*** (0,018) 0,140*** (0,018) 0,035 (0,023) 0,061*** (0,023) 0,066*** (0,023) 0,043* (0,023) GROWTH -0,004 (0,028) -0,001 (0,028) -0,003 (0,028) -0,001 (0,028) -0,181*** (0,051) -0,169*** (0,050) -0,169*** (0,050) -0,178*** (0,051) CASH -0,034 (0,029) -0,036 (0,029) -0,011 (0,030) -0,007 (0,030) -0,189*** (0,045) -0,116** (0,048) -0,205*** (0,045) -0,180*** (0,045) MTBV 0,012 (0,038) 0,007 (0,039) 0,011 (0,038) 0,012 (0,038) -0,270*** (0,063) -0,232*** (0,063) -0,182*** (0,065) -0,257*** (0,064) ROA -0,012 (0,045) -0,032 (0,049) -0,076 (0,048) -0,045 (0,046) -0,109 (0,105) -0,044 (0,105) -0,025 (0,105) -0,103 (0,104) TIER1 0,148*** (0,027) 0,143*** (0,027) 0,134*** (0,027) 0,120*** (0,028) 0,017 (0,066) -0,012 (0,066) -0,023 (0,066) 0,015 (0,066) Time fixed effects
Yes Yes Yes Yes Yes Yes Yes Yes
Normalized data
Yes Yes Yes Yes Yes Yes Yes Yes
𝑅2 0,099 0,100 0,113 0,108 0,139 0,157 0,162 0,142 Adjusted 𝑅2 0,082 0,082 0,095 0,090 0,116 0,133 0,138 0,118 F- value 5,619*** 5,427*** 6,200*** 5,938*** 5,992*** 6,597*** 6,841*** 5,894*** N 1197 1197 1193 1197 877 877 877 877 Durbin-Watson 0,791 0,790 0,807 0,803 1,394 1,426 1,440 1,403
This table shows an OLS regression on the dependent variable total dividend payment, with the first four models representing code law countries and the second four models representing common law countries. The rows show the value per variable included and model characteristics and the columns different models. The p-value is given. *, **, *** Show the significant level at 10%, 5%, 1% respectively. The standard errors are given within parentheses.
43
significance level below 0,000. This means we reject the null-hypothesis and conclude that there is a significant difference between the two sub-samples. The next step will be to do a Wilcoxon rank-sum test8 to see the actual difference before and after the start of 2009. The test value is 31,308 at a significance level below 0,000.
Table 12. Financial crisis effect on dividend pay-out
Variable Count Median Overall median Mean rank Mean score
AFTER 2112 -0,429 574 1335,626 -0,370
BEFORE 1454 0,314 1209 2434,057 0,555
All 3566 0,053 1783 1783,500 0,007
This table shows the results of the Wilcoxon rank-sum test for the dependent variable total dividend pay-outs. The rows show the values after, before, and without breaking point taken into account. The columns show the values for count, median, overall median, mean rank, and mean score. The breaking point is the year 2009, due the financial crisis in the end of 2008, within the period of 2001-2017.
As can be found in table 12 the median of the variable AFTER is far lower than the variable BEFORE and with significant results in this test we can conclude that the total dividend payment of banks after the financial crisis is indeed lower than before the financial crisis. Table 13, 14 , and 15 show details of the dividend policy of banks in OECD countries before and after the crisis. It can be noticed that dividend tax became a more important predictor after the crisis than before the crisis. The substitution theory does not hold for banks after the crisis but does hold before the crisis. There could be a change in dividend form pay-out policy due to the crisis. The importance of dividend tax as factor on dividend pay-outs holds before and after the crisis confirming Casey, and Dickens (2000) again. It also can be noticed that common law countries seem more focused on share repurchases relative to code law countries after the crisis, while before the crisis their
44
focus lied more upon cash dividend payments relative code law countries. Appendix 2 states the period fixed effects and shows a relative high negative impact on dividends in the year the Basel III was implemented. Furthermore, it has been stated that regulation is of high impact for code law countries. Therefor regulation is the main factor found in this study that changed dividend policy in height after the crisis.
Table 13. The relationship between different types of taxes and the total dividend payed by banks in OECD countries before and after the 2008 financial crisis.
Variable Model 1 (h) Model 2
(h)
Model 3 (h)
Model 4 (h) Model 5 (h) Model 6 (h) Model 7 (h) Model 8 (h) INTERCEPT 0,171** (0,073) 0.012 (0,196) 0,035 (0,098) 0,077 (0,105) -0,207*** (0,033) -0,529*** (0,105) 0,175*** (0,066) 0,068 (0,071) CIT 0,005 (0,006) 0,012*** (0,004) PIT 0,006** (0,003) -0,015*** (0,002) NIT 0,005 (0,004) -0,013*** (0,003) LAW 0,021 (0,075) -0,001 (0,080) -0,017 (0,078) 0,010 (0,076) 0,284*** (0,050) 0,208*** (0,055) 0,383*** (0,051) 0,361*** (0,052) PROFIT 0,809*** (0,078) 0,789*** (0,082) 0,863*** (0,082) 0,828*** (0,080) 0,299*** (0,052) 0,268*** (0,052) 0,242*** (0,052) 0,272*** (0,052) ASSETS 0,242*** (0,035) 0,241*** (0,035) 0,234*** (0,035) 0,239*** (0,039) 0,096*** (0,025) 0,094*** (0,025) 0,140*** (0,025) 0,119*** (0,025) GROWTH -0,199*** (0,039) -0,200*** (0,039) -0,200*** (0,039) -0,197*** (0,039) -0,140*** (0,025) -0,146*** (0,021) -0,143*** (0,025) -0,138*** (0,025) CASH -0,094 (0,086) -0,099 (0,086) -0,055 (0,087) -0,069 (0,088) -0,049** (0,021) -0,031 (0,021) -0,066*** (0,021) -0,062*** (0,021) MTBV -0,036 (0,049) -0,031 (0,049) -0,045 (0,049) -0,038 (0,049) 0,094*** (0,036) 0,108*** (0,036) 0,124*** (0,035) 0,098*** (0,035) ROA 0,058 (0,060) 0,084 (0,066) 0,012 (0,064) 0,037 (0,062) 0,229*** (0,048) 0,281*** (0,050) 0,333*** (0,049) 0,256*** (0,048) TIER1 0,182*** (0,044) 0,186*** (0,044) 0,164*** (0,045) 0,168*** (0,045) 0,123*** (0,026) 0,130*** (0,026) 0,122*** (0,025) 0,141*** (0,026) Time fixed effects
yes Yes yes yes yes yes yes yes
Normalized data
yes Yes yes yes yes yes yes yes
𝑅2 0,318 0,319 0,322 0,319 0,263 0,269 0,286 0,273
Adjusted 𝑅2 0,304 0,304 0,307 0,305 0,254 0,259 0,277 0,264
F- value 23,373*** 21,953*** 22,273*** 22,023*** 28,702*** 27,822*** 30,326*** 28,519***
N 768 768 768 768 1306 1306 1302 1306
45
Table 14. The relationship between different types of taxes and the cash dividend payed by banks in OECD countries before and after the 2008 financial crisis.
Variable Model 1 (i) Model 2 (i) Model 3 (i) Model 4 (i) Model 5 (i) Model 6 (i) Model 7 (i) Model 8 (i) INTERCEPT 0,001 (0,054) 0,406*** (0,145) 0,057 (0,073) 0,143* (0,078) -0,054* (0,031) 0,066 (0,101) 0,218*** (0,063) 0,253*** (0,067) CIT -0,013*** (0,004) -0,004 (0,004) PIT -0,003 (0,002) -0,011*** (0,002) NIT -0,007** (0,003) -0,015*** (0,003) LAW 0,128** (0,056) 0,184*** (0,059) 0,143** (0,058) 0,145** (0,056) 0,071 (0,047) 0,0991* (0,052) 0,139*** (0,049) 0,156*** (0,049) PROFIT 0,470*** (0,058) 0,521*** (0,060) 0,448*** (0,061) 0,441*** (0,059) 0,319*** (0,049) 0,331*** (0,050) 0,280*** (0,049) 0,288*** (0,049) ASSETS 0,110*** (0,026) 0,112*** (0,026) 0,113*** (0,026) 0,115*** (0,026) 0,018 (0,023) 0,018 (0,023) 0,046* (0,024) 0,042* (0,024) GROWTH -0,250*** (0,029) -0,247*** (0,029) -0,250*** (0,029) -0,253*** (0,029) -0,178*** (0,024) -0,176*** (0,024) -0,178*** (0,024) -0,177*** (0,024) CASH -0,105* (0,064) -0,093 (0,063) -0,121* (0,065) -0,143** (0,065) -0,008 (0,020) -0,015 (0,021) -0,021 (0,020) -0,023 (0,020) MTBV 0,078** (0,036) 0,066* (0,036) 0,082** (0,036) 0,082** (0,036) 0,260*** (0,034) 0,255*** (0,034) 0,280*** (0,034) 0,265*** (0,034) ROA 0,347*** (0,044) 0,283*** (0,049) 0,367*** (0,047) 0,379*** (0,046) 0,334*** (0,045) 0,315*** (0,048) 0,408*** (0,047) 0,364*** (0,045) TIER1 0,131*** (0,033) 0,120*** (0,033) 0,138*** (0,033) 0,152*** (0,034) 0,081*** (0,025) 0,078*** (0,025) 0,080*** (0,024) 0,102*** (0,025) Time fixed effects
yes Yes yes yes yes yes yes yes
Normalized data
yes Yes yes yes yes yes yes yes
𝑅2 0,454 0,461 0,455 0,459 0,414 0,415 0,425 0,426
Adjusted 𝑅2 0,443 0,449 0,444 0,447 0,407 0,407 0,417 0,418
F- value 41,717*** 40,097*** 39,208*** 39,786*** 56,874*** 53,645*** 55,746*** 56,159
N 768 768 768 768 1306 1306 1302 1306
