Print Reference This Reddit This

The Real Drivers Of Dividend Distribution Finance Essay

With the use of the anti-directors rights index constructed by La Porta, Lopez-de-Silanes, Shleifer & Vishny (1997) this paper shows that countries operating under the common law legal system offer better protection to investors. Furthermore, companies in countries with a civil law background pay higher dividends to shareholders than those in a common law system. Using a sample of 2770 companies over 39 countries, this paper concludes that dividend payouts depend majorly on the legal system in operation.

I would like to thank Queen’s University Belfast for both the provision and use of the software systems Thomson One Banker and STATA. I would also like to extend my gratitude to Christopher Coyle and Professor John D. Turner for their academic guidance and personal assistance.

Introduction

The aim of this paper is to analyse the distribution of dividends across legal systems around the world. The general argument revolving this topic is that dividend payout widely depends on the level of protection against possible expropriation which shareholders are exposed to. Evidence constructed by Rafael La Porta, Florencio Lopez-de-Silanes, Andrei Shleifer and Robert Vishny, henceforth LLSV (1997) shows that common law countries offer better investor protection to shareholders than those countries with a civil law legal system. With this idea as a basis to my own research it can be predicted that firms located in countries operating under a civil law legal system pay higher dividends to shareholders.

To decipher what the real driver(s) behind dividend payout may be, a review must first of all be carried out of the differences in the amount of dividends paid out amongst companies in countries operating under both common and civil law legal systems. It is proved true that investor protection affects dividend payout and LLSV’s (1997) theory that common law countries provide greater protection is evidently backed up. As the majority of literature suggests the main driver behind dividend payout is either investor protection or the legal system, tests will also be performed showing that this is theoretically a true reflection of dividend policies. It seems realistic to assume that factors such as a company's size, growth or profit would affect the amount of money paid out to shareholders however; the only one of these factors found to be significant is company size.

To begin this paper I give an extensive overview of previous research in the literature review in section two to highlight the main arguments surrounding the topic of dividend policy. In Section three there is a detailed description of data used and the methods by which the hypotheses are tested. Results are set out and analysed in section four with the study being brought to a conclusion in section five. The final section is a complete list of references of papers used to help construct this paper.

Literature Review

There has been extensive research carried out in the past into dividend policies. Further examination into the reasons for the payment of dividends is still being currently carried out and this paper looks into what really drives dividends as an extension to what has previously been suggested. The ongoing argument is that dividend payments are used as a solution to the agency problem. This important issue in financial environments where investors have difficulties with insuring that a manager will work in their best interests is thought by many to be partially solved with dividend policies. Easterbrook (1984) explains how the payment of dividends increases the need for managers to enter the capital markets to raise funds where they are subjected to higher levels of monitoring and thus “aligning managers’ interests with those of shareholders.”

The research of Jensen (1986) is conclusive with the view of dividend policies acting in a way to ameliorate this principal-agent problem. In his work he explains that dividend policies reduce the power that managers maintain by reducing the amount of resources which they control and subsequently resulting in the need to return to the capital market. Jensen (1986) concentrates his study on companies with large amounts of free cash flow [1] . Because these firms have the ability to increase the amount of dividends they pay to investors this decreases the amount of funds that can be inefficiently used or wasted and in turn reduces the agency problem. The promise of an increase in payout may not be a long term solution however, and Jensen (1986) concludes that the issue of debt instead could have a more successful reduction in the expropriation of shareholders as it provides them with rights that dividend policies cannot. In the event of failure to pay the principal, Jensen (1986) explains that the firm can be taken to bankruptcy court creating a “bond” between managers and shareholders that “cannot be accomplished by a simple dividend.”

Another repetition of this argument is that of Cheffins (2006) who finds that “paying dividends acts as a check on the skimming and squandering of corporate profits.” This again, is a result of the exposure to the “screening and scrutiny of financial intermediaries” in capital markets because managers must find funds for projects following the reduction in their retained earnings due to the payment of dividends as put by Cheffins (2006). By constraining and disciplining managers’ behaviour dividends can limit the amount of inefficient performance carried out and therefore reducing the principal-agent problem.

The other most prominent side to the argument which is evident in empirical research is that the amount dividends are dependent on investor protection. The elaborate body of work carried out over years by Rafael La Porta, Florencio Lopez-de-Silanes, Andrei Shleifer and Robert Vishny (LLSV) provides an excellent insight into the reasoning behind dividend policies. LLSV (1997) investigate the development of capital markets across countries with two types of legal systems; common law and civil law. The anti-director rights index composed by LLSV (1997) is a means of measuring the level of shareholder protection in a country, it ranges from zero to six and the better protection there is against the expropriation of shareholders the higher the index. In general they find that this index is higher in countries which operate under a common law system and consistently lower in civil law countries. This brings them to the conclusion that because common law countries better protect investors then their capital markets are also better developed than those in civil law countries. LLSV (1997) explain that “stronger anti-director rights are associated with larger and broader equity markets” which leads us to ask - what affect does the legal system consequently have (if any) on dividend payout? LLSV (2000) further develop this idea coming to a consistent conclusion stating that financial market development increases as investor protection increases. When investors can avoid risk because of better protection they are more willing to pay higher prices for securities which encourage the issue of such securities further developing the market for equity.

LLSV (Feb. 2000) suggest two models for explaining the payment of dividends: the “outcome model” and the “substitute model”. These two models propose conflicting relationships between investor protection and dividend payout. The “outcome model” argues that dividends are distributed as a result of the pressure placed on managers by monitoring minority shareholders and because better protective rights enables increased monitoring possibilities then higher protection should result in higher dividend payouts. On the other hand, “the substitute model” is based on the credibility managers want to attain in the capital market. LLSV (Feb. 2000) explain that agents who want to build a reliable reputation for the “decent treatment of minority shareholders” only payout dividends because they “are interested in issuing equity in the future.” Their results overall support the “outcome model”. They found that dividend payouts are generally higher in companies in countries operating under a common law country than those in civil law countries. Further to this, they found that the growth of a firm was also significant in their regressions finding that fast growing companies in common law countries “make higher dividend payouts than slow growth firms.” Where others have suggested that dividends are a way of ameliorating the agency problem for those investors who are poorly protected, LLSV (Feb. 2000) say that when an investor is offered better protection they “use their legal powers to extract dividends from firms.”

The most recent literature relevant to my own research is that of Campbell & Turner (2011) which by using historical data in Victorian Britain comes to a contrasting conclusion to that of LLSV (Feb. 2000). Campbell & Turner (2011) insist that dividend payout provides protection to shareholders due to the submission of managers “to the discipline of the capital markets” but as opposed to LLSV (Feb. 2000) suggests that this would lead to higher dividends in lower protection countries as they would act “as a substitute to legal protection.” Campbell & Turner’s (2011) evidence provides the same argument as Easterbrook (1984), Jensen (1986) and Cheffins (2006) advising that the need to enter the capital market keeps managers from “diverting cash flows” and ameliorating the potential advantage that insiders may have over minority shareholders. Their research claims that when investors are exposed to poor legal protection then the managers will pay out higher dividends to almost reassure them and in turn improving their own reputation within the market. Campbell & Turner (2011) regression analysis showed that companies with higher corporate values pay significantly higher dividends but only showed weak evidence that “firms were submitting themselves to market discipline by paying dividends.” Furthermore, they found investment opportunities (growth variable) and age of the companies to not affect dividend policy. While they show that it was necessary to attract investors with “high and stable dividend payments” to raise capital their results show that dividends were only a “partial substitute” for poor investor protection.

The entirety of empirical research carried out on what affects dividend policies suggest that the main drivers are investor protection and the legal system which can be argued are both entwined and dependent of each other. This shows that I must expect to come to a similar solution regarding the significance of investor protection and the legal system but can only find out what affect this will be following the regression tests. Because the majority of literature explains how dividends are a way of ameliorating the principal-agent problem, I anticipate that because civil law countries have lower investor protection (LLSV (1997)) they will have higher dividend payouts as Campbell & Turner (2011) suggest should be the apparent relationship with investor protection, legal system and dividend policies.

Data & Methodology

3.1 Data Collection

The foundation of this entire dataset relies on the LLSV (1997) anti-director rights index which shows the level of protection against the expropriation of investors in a country. The index ranges from 0-6 and illustrates how well a country's shareholders are protected against the inappropriate behaviour of managers carried out for personal gain. If there is no protection at all the index would equal zero and for this specific test it is suggested that a value of greater than four shows a well protected country. As this anti-director rights index is only applied to a specific selection consisting of 49 countries, the Thomson One Banker database was used to extract the top 150 companies based on year end market capitalisation in 2010 in each of the 49 countries. A report was then drawn up for each country with the necessary variables for the regressions shown below.

Dividend per share = β + β1 legal system + β2 risk + β3 size + β4 profit + β5 growth

Dividend per share = β + β1 investor protection + β2 risk + β3 size + β4 profit + β5 growth

3.2 Variables

The dependent variable being the dividend per share, I gathered information on the variables expected to affect this. The legal system variable will be a dummy variable which will equal one if the country operates under a civil law legal system and equal to zero if it practises common law. The investor protection variable will also take the form of a dummy variable equalling one if there is good protection (anti-director rights index of greater than four) and zero otherwise. From the Thomson One Banker system, Beta is used for the risk variable, yearend market capitalisation for size and net profit margin and total assets one year growth for profit and growth variables respectively.

A log of dividend per share is used to reduce the dispersion of the data and to make a more accurate testing model. With regard results it must be noted that the coefficients are interpreted differently after forming a log-linear model such as this [2] . There are several companies in the sample which pay zero dividends but as a log of zero is impossible to calculate I leave these as zero in the dataset. This is because otherwise these observations would be removed from the regression in STATA and it is vital that this does not occur and that they are included in this cross-country analysis.

3.3 Problems

Several of the countries in the sample which were developing countries had no information at all available (possibly due to poor resources). These were Ecuador, Kenya, Nigeria and Uruguay and are therefore extracted from the data set. Others had incomplete listings of variables with several showing “n/a”. Any incomplete company listings were removed from the country's data which consequently left more countries with no remaining data. These countries were Chile, Egypt, Israel, Jordan, Venezuela and Zimbabwe and so will also not appear in the sample. The resulting sample consists of a total of 2770 companies (reduced from 5759) across 39 countries to carry the research out on.

3.4 Methodology

As this paper is seeking to find an explanation for why dividend payout differs across countries, several regressions are carried out with and without the dummy variables. Separate regressions are used to test investor protection and the legal system to find out which has a greater affect on dividend payout. Regressions are first carried out without the dummy variable and then again with the addition of the dummy to determine whether including it has a significant effect on dividend per share.

Regression testing common and civil law:

Ln(Dividend per share) = β + β1 risk + β2 size + β3 profit + β4 growth

Ln(Dividend per share) = β + β1 legal system + β2 risk + β3 size + β4 profit + β5 growth

Regression testing good and poor investor protection:

Ln(Dividend per share) = β + β1 risk + β2 size + β3 profit + β4 growth

Ln(Dividend per share) = β + β1 investor protection + β2 risk + β3 size + β4 profit + β5 growth

I expect that results will show that the legal system has a more significant effect on dividend per share than any other variable and that companies in civil law countries have higher dividend payouts.

3.5 Diagnostic tests

A test for multicollinearity is carried out to determine whether the explanatory variables in the chosen OLS estimation method of regression are correlated. This assumption of the model which states that the variables have no relationship will be violated if the results show that any of the following are correlated with dividend per share: the legal system, investor protection, company size, profit or growth.

To test the normality of the model as a measure of the goodness-of-fit a Bera-jarque test is applied. This type of test based on the skewness and kurtosis of the data sample, will ascertain whether the model is that of a normal distribution. Finally, an “Eyeball test” is carried out on the residuals calculated from them Bera-Jarque method to establish any outliers in the data and re-run the regression to analyse their significance (if any).

Results & Analysis

4.1 Data summary

Table 4.1 Summary of averages

Countries

Observations

Dividend per Share

Risk

Size

Profit

Growth

Legal System

Investor Protection

Civil Law

1484

2.17

5.75

7063.89

-1927.27

36.02

1

0.223

Common Law

1286

0.33

1.02

13246.21

837.80

14.69

0

0.876

All

2770

1.32

3.56

9934.09

-643.56

26.12

0.536

0.526

Table 4.1 shows the spread of each of the variables and the differences across the separate legal systems. There are 198 more observations in civil law countries than in common law countries due to the problems experienced in collecting data. This is the reason why the average of the legal system dummy variable has a value of 0.536 over all countries rather than 0.5 as would be expected. The removal of incomplete data has left an uneven sample and although this could have created bias, I am confident that it is not too serious an issue as the sample is large and the difference is small in relation.

The average dividend per share is a lot greater in civil law countries at 2.17 compared to 0.33 in common law countries. This contradicts the findings of LLSV (Feb. 2000) where they found common law countries to pay higher dividends because of the idea of the “outcome model”. If the “outcome model” was at play then there should be higher dividends in common law countries as they have higher investor protection. I can however, wholly agree with LLSV (1997) and their statement that there is a much higher level of investor protection in common law countries (0.876) than in civil law (0.223). This varied investor protection would also explain the evident difference in risk across the legal systems as civil law countries have a high average risk of 5.75 whereas the common law average is a lot lower at 1.02.

The larger size variable for common law countries could be in agreement with LLSV’s (1997) explanation that “stronger anti-director rights are associated with larger and broader equity markets.” It can be suggested that these “broader equity markets” would also accommodate generally larger companies.

Table 4.2 Summary statistics

Observations

Mean

Standard Deviation

Min

Max

Skewness

Kurtosis

Ln(Div per Share)

2770

-1.00946

1.611639

-4.60517

6.002034

-0.30172

3.163827

Risk

2770

3.556318

36.43207

-386.43

675.68

7.90912

117.847

Size

2770

9934.093

24891.63

2.24

364064.5

5.816761

48.63174

Profit

2770

-643.558

59170.49

-2963108

947829.9

-43.8415

2293.856

Growth

2770

26.11645

578.4998

-77.63

30109.65

50.90756

2642.685

Legal System

2770

0.53574

0.498811

0

1

-0.14333

1.020543

Investor Protection

2770

0.526354

0.499395

0

1

-0.10556

1.011143

Table 4.2 shows that both legal system and investor protection are slightly negatively skewed and have kurtosis of just above one. This low value of kurtosis means that the deviations around the mean of this sample are not extreme.

4.2 Regression results

Table 4.3 Legal system results

Ln(Dividend per share) = β + β1 risk + β2 size + β3 profit + β4 growth

Ln(Dividend per share) = β + β1 legal system + β2 risk + β3 size + β4 profit + β5 growth

(1)

(2)

VARIABLES

lndivpershare

VARIABLES

lndivpershare

Risk

0.000580

Risk

-0.000646

(0.700)

(-0.872)

Size

1.10e-05***

Size

1.46e-05***

(9.075)

(13.37)

Profit

-6.80e-07

Profit

-4.00e-07

(-1.333)

(-0.879)

Growth

3.73e-05

Growth

1.51e-05

(0.715)

(0.324)

Constant

-1.122***

Legal System

1.458***

(-34.32)

(26.74)

Constant

-1.933***

Observations

2,770

(-45.96)

R-squared

0.030

Observations

2,770

R-squared

0.229

t-statistics in parentheses

*** p<0.01, ** p<0.05, * p<0.1

The regression was ran first of all without the dummy variable for the legal system and then again with the dummy to see whether it’s inclusion was relevant. As before, the legal system dummy variable equalled one if the company was in a civil law country and zero if it was in a common law country. The original regression results on the left hand side of table 4.3 show the only variable to be significant to dividend per share (lndivpershare) is the size of the company. Even though the size variable is significant at the 1% level, the R2 is very low at 0.030 which shows that this model is a very poor fit and that we should not rely too much on the results it shows. The (2) lndivpershare regression shows the effect that the legal system has on dividend per share and as the R2 has increased to 0.229 these results are more credible. The positive coefficient of 1.458 shows that there is also a positive relationship between civil law and dividend per share; there are higher dividends within countries operating under the civil law legal system. This is true to say because this dummy variable is significant at the 1% level. The size variable is again significant at the 1% level but because the coefficient is very small, an increase in size will only have a small increase in dividend payout. Risk and profit both have a negative effect on dividend per share but again because the coefficient is very small and because they are not shown to be significant I cannot draw much of a relationship from these results.

Table 4.4 Investor protection results

Ln(Dividend per share) = β + β1 risk + β2 size + β3 profit + β4 growth

Ln(Dividend per share) = β + β1 investor protection + β2 risk + β3 size + β4 profit + β5 growth

(1)

(2)

VARIABLES

lndivpershare

VARIABLES

lndivpershare

Risk

0.000580

Risk

-0.0338

(0.700)

(-0.427)

Size

1.10e-05***

Size

1.38e-05***

(9.075)

(11.79)

Profit

-6.80e-07

Profit

-4.85e-07

(-1.333)

(-0.996)

Growth

3.73e-05

Growth

1.95e-05

(0.715)

(0.392)

Constant

-1.122***

Investor Protection

-0.960***

(-34.32)

(-16.40)

Constant

-0.641***

Observations

2,770

(-14.95)

R-squared

0.030

Observations

2,770

R-squared

0.116

t-statistics in parentheses

*** p<0.01, ** p<0.05, * p<0.1

The regression on the left hand side of table 4.4 is the same as in table 4.3 and is only included again to make a clear comparison between the regression without the dummy variable and the regression with the dummy.

This regression is carried out to see whether the Campbell & Turner’s (2011) idea that “dividends are a substitute of legal protection” is a plausible idea. The size variable shows the same results as in table 4.3. As before, the investor protection dummy variable equals one if there is good protection and zero otherwise. This variable shows to be significant at the 1% level with a coefficient of -0.960; the greater the investor protection the lower the dividend per share. This could mean that Campbell & Turner (2011) could be right in saying that greater dividends are paid out because there is lower investor protection. Furthermore, it is a widespread argument (see 2. Literature Review) that these dividends are paid out to unprotected investors to align managers’ interests with theirs but this is something I cannot prove.

The following series of graphs (graphs 4.1, 4.2 and 4.3) illustrate this relationship between investor protection and the average dividend per share in all countries and also in both common law countries and civil law countries. A graphical representation cannot be achieved for the relationship between the legal system and dividend per share because the latter is simply a dummy variable whereas we can plot investor protection as the anti-director rights index by LLSV (1997) on a scale from 0 to 6.

Graph 4.1

The graph above shows that investor protection has a negative relationship with average dividend per share. The line of best fit added to the graph shows this very clearly as it has a negative slope.

Graph 4.2

Graphs 4.2 and 4.3 show the distribution of dividends across the different legal systems. It can be seen that the data isn’t as concentrated as previous literature would suggest. LLSV (1997) find that common law countries have better investor protection than civil law countries but the graphs show that it is not as simple as that. The scatter graphs are more dispersed across the x-axis showing that investor protection varies even across separate legal systems. There are countries in civil law legal system operation with an anti-director rights index equal to four and countries under common law system with an index value of two. It is interesting to note that the lines of best fit in graphs 4.2 and 4.3 show that dividends decrease as investor protection increases in civil law countries and vice versa for common law.

Graph 4.3

4.3 Diagnostic test results

Table 4.5 Legal System multicollinearity

Ln(Div per Share)

Risk

Size

Profit

Growth

Legal System

Ln(Div per Share)

1

Risk

0.0083

1

Size

0.1695

-0.0282

1

Profit

-0.0242

0.0008

0.0045

1

Growth

0.0122

-0.0011

-0.0066

0.001

1

Legal System

0.4227

0.0648

-0.1239

-0.0233

0.0184

1

Both tables 4.5 and 4.6 show that this regression model violates the OLS assumption that the variables are orthogonal; or have no relationship.

The results in table 4.5 show again that size and legal system are the most significant as they have the highest correlation with dividend per share. The legal system dummy variable has the most correlation and it’s positive value shows again that civil law is positively related with the amount of dividends paid out to shareholders.

Table 4.6 Investor Protection multicollinearity

Ln(Div per Share)

Risk

Size

Profit

Growth

Investor Protection

Ln(Div per Share)

1

Risk

0.0083

1

Size

0.1695

-0.0282

1

Profit

-0.0242

0.0008

0.0045

1

Growth

0.0122

-0.0011

-0.0066

0.001

1

Investor Protection

-0.2662

-0.0738

0.1471

0.0247

-0.0223

1

Table 4.6 reiterates that which table 4.5 shows. As before with the regression analysis, investor protection and size are the most significant variables. Investor protection has a negative relationship with dividend per share backing up the statement that the higher the investor protection, the lower the dividends paid out.

Table 4.7 With outliers

Test Statistic

P Value

Legal System

35.09

0.0008051

Investor Protection

1750

0

Table 4.7 shows the result of the Bera-Jarque normality test. As the p values are lower than 0.5 the normality of the models is rejected. This rejection could suggest that there are outliers in the sample which could bias the data.

To determine what these outliers in the separate regressions for the legal system and investor protection are, the residuals are predicted and the outliers are picked from these. A graphical representation of this test on the legal system is shown in graph 4.4. This shows that the outliers were almost as high as a value of 4 where the mean of the normal is around -1. The corresponding test for investor protection is shown in graph 4.5.

Graph 4.4 Distribution of legal system outliers

Graph 4.5 Distribution of investor protection outliers

Within the legal system regression there are three apparent outliers and only one in the investor protection regression. Each of these outliers is re-entered into the appropriate regression as a dummy variable, equalling one for that specific company and zero for all other companies. The regression is repeated and the results of the legal system regression with the three outliers as separate dummy variables are shown in table 4.8.

Table 4.8 Regressing legal system outliers as dummy variables

Ln(Dividend per share) = β + β1 legal system + β2 risk + β3 size + β4 profit + β5 growth + β6 outlier1 dummy+ β7 outlier2 dummy+ β8 outlier3 dummy

(1)

VARIABLES

lndivpersharey10

Risk

-0.000635

(-0.859)

Size

1.59e-05***

(13.67)

Profit

-4.00e-07

(-0.881)

Growth

1.54e-05

(0.330)

Legal System

1.467***

(26.90)

id2768dum

-2.262

(-1.578)

id2769dum

-2.924**

(-2.038)

id2770dum

-3.279**

(-2.227)

Constant

-1.948***

(-46.11)

Observations

2,770

R-squared

0.232

t-statistics in parentheses

*** p<0.01, ** p<0.05, * p<0.1

The legal system is yet again the most significant and because the coefficient and the R2 have slightly increased it is clear that these outliers have an effect however small it is, on the regression.

Table 4.9 shows the investor protection regression results with the one outlier dummy variable. Again, the R2 has increased and the investor protection variable has decreased which would be expected as it has a negative coefficient and therefore negative relationship with dividend per share. Inserting the outlier as a dummy variable here has just as with the legal system regression, had a small but positive effect on the results making them slightly more accurate.

Table 4.9 Regressing investor protection outliers as dummy variables

Ln(Dividend per share) = β + β1 investor protection + β2 risk + β3 size + β4 profit + β5 growth + β6 outlier1 dummy

(1)

VARIABLES

lndivpershare

Risk

-0.000331

(-0.417)

Size

1.45e-05***

(11.89)

Profit

-4.86e-07

(-0.997)

Growth

1.97e-05

(0.395)

Investor Protection

-0.963***

(-16.45)

id2770dum

-3.100**

(-1.968)

Constant

-0.645***

(-15.04)

Observations

2,770

R-squared

0.117

t-statistics in parentheses

*** p<0.01, ** p<0.05, * p<0.1

The Bera-Jarque normality test is again applied following the presentation of the outliers and the results are shown in table 4.10.

Table 4.10 Without outliers

Test Statistic

P Value

Legal System

76.42

0.0000001076

Investor Protection

1157

6e-252

The P values in table 4.10 are again less than 0.5 so normality is still rejected. Unfortunately, the p values have decreased which shows that the extraction of the outliers has had a negative effect on the normality of the model.

Robustness of results

There are several limitations to this research which may affect the accuracy of the results. When collecting the data, several companies and also countries had to be removed due to incomplete information and this resulted in an uneven sample. There are 1286 common law companies and 1484 civil law companies which may have skewed the results. This unfortunately is an issue that could not be resolved if using the anti-director rights index constructed by LLSV (1997). As this index is only applied to 49 specific countries, when the data is not available for a country then it must be excluded. There are however other investor protection variables that could have been used which may not have encountered the same problem. Examples of these possible variables are risk of expropriation or the anti-self dealing index by La Porta et al. (2006).

A different selection of variables for either the dependent or independent factors could maybe improve the regression. Dividend payout ratio as used by Campbell & Turner (2011) could have been selected for a change in independent variable. There is always the possibility that there are other dependent variables which could have a significant effect on dividend policy such as performance. This could be measured by Tobin’s Q and included in the model.

The R2 for each of the regressions is not particularly strong. A value of one would show a perfect fit of the model and although the values achieved show a fair amount of explanation of the variance of the model, a higher value would suggest a higher relevancy of the results.

It is regrettable in hindsight that the Bera-Jarque normality testing of the regression did not prove to have a more positive effect on the model. Improvements were very slight and at times ambiguous, but it is possible that if the changes listed above were applied then this could show different results.

Summary & Conclusion

This research shows that dividend payout majorly depends on the legal system in operation in a country. Dividends received by shareholders are greater for those in civil law countries than those in common law countries. Results show that the legal system is more significant than the investor protection or the anti-directors rights index of the country. This is inconsistent with LLSV (Feb. 2000) but one could suggest that it agrees with Easterbrook (1984) and Jensen (1986) that dividends are a method of ameliorating the principal-agent problem. As it has been shown that the civil law legal system offers poorer protection to shareholders than common law and also that those in civil law countries benefit from higher dividends per share, it is very possible that these are to compensate for weak protection by reducing the agency problem although impossible to prove with quantitative data.

Of the other variables used in the regression (size, profit and growth) the only one to have a significant effect on dividend per share is the size of a company. Both profit and growth proved to be insignificant bringing me to conclude that the legal system is the single most important factor affecting dividend distribution.

Need help with your literature review?

Our qualified researchers are here to help. Click on the button below to find out more:

Literature Review Service

Related Content

In addition to the example literature review above we also have a range of free study materials to help you with your own dissertation: