Dividend payers, firm characteristics, and the propensity to pay in the United Kingdom

0

Dividend payers, firm characteristics, and the

propensity to pay in the United Kingdom

Ya Yang (s1906445) Faculty of Economic and Business

University of Groningen

First supervisor: Prof. dr. J.H. von Eije

Second supervisor: Prof. dr. L.J.R. Scholtens

1

Abstract

The fraction of British firms paying dividends declines from 76.6% in 1998 to 40.5% in 2009. After doing research, I come to the conclusion that this situation is due in part to the changing characteristics of publicly traded firms and in part to firms’ lower propensity to pay dividends. Both factors are important in explaining the declining incidence of dividend-paying firms. Also, the lifecycle theory and pecking order theory are supported by my research. Besides, last year’s payment status also plays an important role in determining the next year’ payment status for a firm, and dividend stickiness presented by Lintner (1956) is still important.

Key words: Dividends, likelihood to pay, firm characteristics, the propensity to pay, last year payment status, lifecycle theory, signaling theory, pecking order theory, and dividend stickiness

JEL code: G3

2

Content

2.1.2 Pecking order theory ... 7

2.1.3 Life cycle theory ... 8

2.2 METHODOLOGY ... 8

3 SAMPLE SELECTION AND DATA DESCRIPTION ... 11

4 METHODOLOGY ... 15

4.1 LOGISTIC REGRESSION MODEL ... 15

4.2 QUANTIFICATION OF EFFECTS OF CHARACTERISTICS AND PROPENSITY TO PAY ON THE LIKELIHOOD TO PAY ... 16

4.3 STICKINESS ... 17

5 EMPIRICAL RESULTS ... 18

5.1 UNIVARIATE ANALYSIS FOR LIKELIHOOD TO PAY ... 18

5.2 MULTIVARIATE ANALYSIS FOR THE LIKELIHOOD TO PAY ... 20

5.3 DIVIDEND STICKINESS... 22

5.4 QUANTITATIVE EVIDENCE OF THE PROPENSITY TO PAY... 22

5.4.1 Propensity to pay with different logistic variables--- 22

5.4.2 Test of the hypothesis of propensity to pay--- 28

5.4.3 Propensity to pay for different dividend groups--- 28

5.4.4 Test of the hypothesis of propensity to pay--- 33

5.4.5 Implication for dividend stickiness --- 33

6 CONCLUSION ... 35

REFERENCE

3

1 Introduction

Using Thompsons’ DataStream, I study the incidence of dividend payers for non-financial non-utility firms during the 1998-2009 period in Britain. The proportion of firms paying dividends declines continuously during the whole period. In 1998, there are 81.5% of publicly traded firms who pay dividends, and then there are 59.3% of firms paying dividends in 2002. Until 2009, the proportion keeps decreasing and reaches to 45.4%.

The decline of the likelihood to pay dividends of British firms across time generates two questions. 1) is the decline of the likelihood to pay dividends due to the changing characteristics of firms, or 2) are firms with similar characteristics less likely to pay dividends?

I use univariate analysis and multivariate analysis to examine the characteristics of dividend payers. Univariate analysis is conducted by comparing the average characteristics of dividend paying groups and non dividend paying groups over time. Multivariate analysis is conducted by a logistic regression model. There are four variable to test for the impact of the characteristics, namely size, profitability and investment opportunities. The variable of total assets (TA) is used as a proxy for the characteristic of size. The variable of the ratio of earnings before interest and tax (EBIT)/TA is used as a proxy for the characteristic of profitability. The variable of market-to-book value (MTB) is calculated by market capitalization divided by common equity, which together with the variable of asset growth rate (DTA) is used as proxy for the characteristic of investment opportunities. Finally, the variable of last year’s payment status (LP) is added to test if dividend stickiness holds.

Univariate analysis shows that prior payers are always larger, more profitable and have less investment opportunities than prior nonpayers. Besides, according to the analysis across time, the result does not follow regular pattern and this finding contradicts Denis and Osobov (2008)’s results which show that firms tend to be smaller, less profitable and have more investment opportunities over time. For the multivariate analysis, Fama and French (2001) conclude that size, profitability and investment opportunities can influence the likelihood to pay dividends for a firm and I use the four variables mentioned above as proxies for these three characteristics to analyze the relationship between the three characteristics and the likelihood to pay.

4 After logistic regression, I come to the conclusion that the characteristics of size and profitability have a positive relationship with the likelihood to pay while the characteristic of investment opportunities has a negative relationship with the likelihood to pay. Although there are some abnormal results which are not consistent with the conclusion above, most of results are consistent. Also, the variable of last year’s payment status (LP) is added in the logistic regression to test the relationship between the likelihood to pay and last year’s payment status. The relationship is expected to be positive in this paper and this result implies that dividend stickiness holds. Finally, the group of all firms is divided into two groups: prior payers and prior nonpayers in order to analyze the development in the propensity to pay for the two groups.

Besides considering the characteristics of firms on the likelihood to pay, I further study if firms have a lower propensity to pay dividends, irrespective of their characteristics. I use the difference between the actual ratio and the expected ratio of payers for a year to measure the propensity to pay dividends. The changes between actual and expected ratio of payers across years can be represented as the effect of changes in the propensity to pay dividends. My results show a lower propensity to pay over time. This result is consistent with papers conducted by Fama and French (2001), Twu (2010) and Von Eije and Megginson (2008). Hypotheses with t-tests are also used to test whether the changes of propensity to pay are significant.

Then dividend stickiness is also considered in this paper. Lintner (1956) first proposes dividend stickiness which shows that last year’s payment status can have influence on the next year’s payment. Twu (2010) has also shown that last year’s payment status does matter when analyzing the next year’s payment. Based on the two papers, firstly, I divide the full sample of firms into prior payers and prior nonpayers. The firms which paid dividend last year are more likely to pay dividend the next year while the firms which did not pay dividend last year are more likely to not pay dividend the next year. This result is consistent with dividend stickiness. Then, I extend the logistic regression by adding last year’s payment status as another variable. The positive coefficient between dividend payment status and last year’s payment status shows a positive relationship, which coincides with Lintner (1956) and Twu (2010). Therefore, dividend stickiness also holds in the recent period of time. Also I follow Twu (2010)’s

5 approach to conduct logistic regression by dividing firms into prior payers and prior nonpayers. The results also support dividend stickiness.

The research contributes to current research related to dividends in four aspects: firstly, I choose the United Kingdom while most researches are for the United States or other countries. Secondly, I study the most current period which is from 1998 to 2009 which may show new light on the impact of firm characteristics and on the propensity to pay. Thirdly, I add last year payment status (LP) as variable to test dividend stickiness and this shows that stickiness is an important factor related to dividend payment. Fourthly, I find that the decline of propensity to pay does also exist in the United Kingdom. Those recent results can be used in the practical life in particular for the United Kingdom when firms want to make a decision of dividend payment or when shareholders want to analysis their firms’ dividend payment.

The structure of this paper is as follows: section 2 provides the literature review to conduct this research; section 3 describes the sample selection and data description; section 4 gives the methodology to conduct the research; section 5 shows the empirical results of my research and section 6 concludes the paper.

6

2 Literature review

2.1 Theory base

Miller and Modigliani (1961) initially conduct research with the assumption of perfect capital markets. However, the markets are imperfective and dividend payment is influenced by many factors under market imperfection. Signaling theory, pecking order theory and lifecycle theory are considered here as they are the main theories related to dividend payment.

2.1.1 Signaling theory

Lintner (1956) states that most managers regard their investors as having interest in earnings, and specially special interest in getting earnings in dividends. He interviewed managers and those managers think that their responsibilities and standards of fairness require them to distribute part of any substantial increase in earnings to investors as dividends. The study of dividend policy by Lintner (1956) shows that firm dividends are only changed when there is a permanent change in earnings. According to this view, earnings - also regarded as profitability in this paper - is positive related to the payment of dividends. And for outside investors, it can be used as a signal for firm earnings. Under asymmetric information, a firm’s dividend policy can conveys information about future earnings. An increase (or a decrease) of current dividends indicates the rise (fall) of future earnings. This view is consistent with the signaling theory. Bhattacharya (1979) and Miller and Rock (1985) both assume that under information asymmetric, dividend can function as a signal of firms expected cash flow. Also, John and Williams (1985) state that dividend announcement

effectsbelong to the basic decision model for a firm, and that dividend can be used as

a signal. Garrett and Priestley (2000) find evidence that dividends convey information about previously unexpected permanent earnings. And their analysis indicates that mainly positive changes to permanent earnings affect the current dividend. The results suggest that for an unexpected increase in permanent earnings, dividends will increase by around 30% of the increase in permanent earnings. The result demonstrates signaling theory. However, Bhattacharya (1979) also states that there is signaling cost and the cost is mostly caused by the fact that cash dividends are taxed at a higher rate

7 than capital gains. Asymmetric information may however also generate agency cost. Easterbrook (1984) states that dividend can reduce the agency cost and so firms may pay dividends despite the tax disadvantage. He states that dividends may keep firms involved in such capital markets where monitoring of managers is available at low cost, and therefore dividend may be used to adjust the level of risk taken by managers and investors. Such an explanation also explains why firms pay out dividends despite the tax disadvantage in the capital market and at the same time go to the capital market for additional funds. In this paper, I only do research about whether signaling theory holds related dividend payment and do not consider why firms pay dividend instead of repurchasing stock. I use earnings before interest and tax divided by total assets (EBIT/TA) as the proxy for profitability (earnings) to test the relationship between the likelihood to pay and profitability by using the method of logistic regression. If a positive relationship is found, signaling theory is assumed to hold. Besides, according to signaling theory, less (more) investment opportunities should result in a lower (higher) likelihood to pay dividends. Therefore, the likelihood to pay dividends should positively correlate with investment opportunities. Then, I use market-to-book value (MTB) and asset growth rate (DTA) as proxies for investment opportunities to test the relationship between the likelihood to pay dividends and investment opportunities. If positive relationships are found, signaling theory is assumed to hold.

2.1.2 Pecking order theory

Pecking order theory suggests that firms prefer to use internal funds to pay dividends and to implement growth opportunities. And if external finance is needed, firms prefer to raise debt instead of external equity. Based on this theory, Myers and Majluf (1984) consider that if a firm must issue common stock to raise cash to undertake a valuable investment opportunity, the firms would prefer to forego good investments rather than issue risky securities to finance them. That is, if prospective investments are larger than operating cash flows, and if the firm has used up its ability to issue low-risk debt, and then the firm may forego good investments rather than issue risky securities. The pecking order theory thus comes to the conclusion that there is a negative relationship between the likelihood to pay dividend and investment opportunities because a high dividend payout ratio will lead to a low level of retained earnings and a greater need

8 to raise money from external sources to finance investment opportunities. And this will result in forgoing the investment opportunities. Adedeji (1998) also concludes that there is a negative interaction between dividend payout ratio and investment in the UK over the period 1993-1996 and which is consistent with pecking order theory. However, the results of the signaling theory of dividends are opposite to the pecking order theory. In this paper, I use market-to-book value (MTB) and asset growth rate (DTA) as proxies for investment opportunities to test the relationship between likelihood to pay dividends and investment opportunities. If negative relationships are found, the pecking order theory is assumed to hold.

2.1.3 Life cycle theory

Lifecycle theory related to dividend means that young firms are less likely to pay dividends because they have limited resources and face relatively abundant investment opportunities, whereas mature firms are more likely to pay dividends because they have a higher profitability and fewer attractive investment opportunities. The trade-off between profit retention and profit distribution evolves over time as profits accumulate and investment opportunities decline. Therefore, paying dividends becomes gradually desirable as firms mature over time. DeAngelo, DeAngelo, and

Stulz(2006) state that dividends tend to be paid by mature firms, and this supports the

lifecycle theory. DeAngelo, DeAngelo (2006) and Bulan, Subramanian and Tanlu (2007) also support the theory. Moreover, Fama and French (2001) find that firms with high profitability and low growth rates tend to pay dividends, while firms with low profitability and high growth rates tend to retain profits, and which is also in line with lifecycle theory because larger firms are in general older firms. In this paper, I use total assets (TA) as proxy for firm size to test the relationship between the likelihood to pay dividends and firm size and in order to see if the lifecycle theory holds. If positive relationship is found, lifecycle theory is assumed to hold.

2.2 Methodology

Fama and French (2001) conclude that the decline in the incidence of dividend payers is in part due to the changing characteristics of the publicly traded firms that have never paid dividends --- small size, low earnings, and large investments relative to

earnings, and in another part due to a lower propensity to pay. They use logistic

9 regressions and a portfolio approach to document that characteristics and propensity to pay make large contributions to the decline in the percent of payers. Von Eije and Megginson (2008) and Twu (2010) also quantify the firm characteristics and propensity to pay by using logistic regressions and come to the conclusion that the decline in the incidence of dividend payers is in part due to the changing characteristics and in another part due to the lower propensity to pay. In this paper, by following their approach, I use logistic regression to quantify the firm characteristics and then study the development in the propensity to pay.

Lintner (1956) first proposes the concept of dividend stickiness. He states that managers are reluctant to make dividend changes that might have to be reversed. This implies that firms that currently pay dividends prefer to continuously pay dividends. By contrast, firms that do not currently pay dividends prefer not to pay them. Based on Lintner’s paper, Twu (2010) examines the importance of prior payment status in determining the likelihood to pay dividends. She conducts an examination of the assumption that prior dividend payers and nonpayers are influenced similarly by the determinants of the likelihood to pay dividends. She categorizes firms into those that paid dividends previously and those that did not. Her results of logistic regressions show that strong dividend stickiness exists and the determinants to pay differ significantly for the two groups of firms. Prior payers prefer continuing dividend payment and prior nonpayers are inclined not to pay. Then, the change of characteristics and the propensity to pay dividends is also examined. It is shown that the sample firms tilt increasingly to the firms with the characteristics typical of nonpayers which is the same as Fama and French (2001)’s results. She states that taking the prior payment status into account can eliminate the problem of overestimating the portion of expected payers. The overestimation of the expected portion of payers is reduced significantly for prior payers and disappears for prior nonpayers. In this paper, I take the prior payment status into account by using two approaches. The first approach is that I add the variable of last payment status (LP) for logistic regression. The second approach is that I divide the whole sample into two groups-prior payers and prior nonpayers-to conduct logistic regression separately. Then I also measure the development of propensity to pay for the two groups. Fama and French (2001) study the incidence of dividend payers during the 1926-99 period, with special interest in the period after 1972, when the data cover NYSE, AMEX, and

10 NASDAQ firms and come to the conclusion of lower propensity to pay. In this paper, I study if this result is also the same in a more recent period and for a different country of the United Kingdom.

11

3 Sample selection and data description

The firm-level data are obtained from the Thomson DataStream. UK-Market firms are chosen from this database. Therefore, all the firms chosen are publicly traded ones. Standard Industrial Classification (SIC) codes are used to erase non-financial and non-utility firms from the UK-Market firms from the analysis. I exclude banking, insurance, and other financial companies (SIC codes 04-06) because their financial papering standards are different from those of the rest of the sample. I also exclude utilities (SIC codes 02) because their payment policies and the access to external financing are regulated. This sifting process is inspired by Fama and French (2001) as well as Von Eije and Megginson (2008). The data period is chosen from 1998 to 2009. I start from 1998 because the coverage of data for most firms is quite sparse before 1998.

Figure 1 shows, from 1998 to 2006, the total number of firms used for the analysis is around 1200. In this period, the number of payers decreases while the number of nonpayers increases.

Figure 1 Number of firms the number of payers and the number nonpayers in United Kingdom

12 Figure 2 shows that the percentage of firms paying dividends decreases continuously during the whole period. In 1998, there are 81.5% of publicly traded firms which pay dividends, and then there are 59.3% of firms paying dividends in 2002. Until 2009, the percentage keeps decreasing and reaches to 45.4%. Otherwise, the percentage of nonpayers increases during this period from 18.5% to 55.6%. According to Fama and French (2001) and Denis and Osobov (2008), the decrease of percentage of dividend payers results from changing characteristics of firms (there are more small size, less profitability and more investment opportunity firms) and the lower propensity to pay.

Figure 2 Percentage of firms in the United Kingdom that pay dividends and that do not pay dividends

Both figures thus show a decline in payers or paying percentages in the United Kingdom and my question is whether this is caused by changes in firm characteristics or by changes in the propensity to pay or by both.

In this paper, dividend payment status is regarded as the dependent variable. Cash Dividend Payment-Total (Field 04551) of UK-market firms is chosen from Thomson DataStream. Then, I set dividend payment status equal to 1 when Cash Dividend Payment-Total is bigger than zero, and equal to zero when Cash Dividend Payment-Total is zero. Therefore, I get a binary variable for dividend payment status. The binary dependent variable of dividend payment status is used in logistic regressions.

13 Five variables are chosen as independent variables. They are total assets (Field 02999), the ratio of EBIT (Field 18191)/total assets, market-to-book value, assets growth rate and last year’s payment status. The variable of total assets represents the characteristic of size which influences dividend payment status. The ratio of EBIT /total assets represents the characteristic of profitability. The variable of market-to-book value is calculated by market capitalization (Field 03501) divided by common equity (Field 08001), and the variable of assets growth is the change in the percentage of total assets. Both market-to-book value and assets growth rate represent the characteristic of investment opportunities. The variable of last year’s payment status (LP) is added to test if dividend stickiness really holds.

Table 1, Panel A, Panel B, and Panel C show the summary statistics for all firms, prior payers and prior nonpayers with the four variables as proxies for the three characteristics of size, profitability and investment opportunities. It shows mean, median, maximum, minimum and standard deviation of the variables of TA, EBIT/TA, MTB and DTA. Panel A shows the summary statistics for all firms. Panel B shows the summary statistics for prior payers and Panel C shows the summary statistics for prior nonpayers. By comparing the means of prior payers and prior nonpayers, prior payers have larger TA, EBIT/TA and smaller MTB and DTA which implies that prior payers are larger, more profitable and have less investment opportunities. The result is consistent with the lifecycle theory for size, the signaling theory for earnings, and the pecking order theory for investment opportunities but not consistent with signaling theory for investment opportunities.

15

4 Methodology

4.1 Logistic regression model

Following Fama and French (2001) and Twu (2010), I estimate the likelihood to pay using the logistic regression model. The logistic regression model is used to test if there is relationship between dividend payment status and the characteristics of the firms such as size, profitability, investment opportunity and last year’s payment status.

Model 1: Pit = α0 + α1 * (TA) it-1 + α2 * (EBIT/TA) it-1 + α3 * (MBF) it-1 + α4 *

(DTA) it-1

Model 2: Pit = β0 + β1 * (TA) it -1 + β2 * (EBIT/TA) it-1 + β3 * (MBF) it-1 + β4 *(DTA)

it-1 + β5 * (LP) it-1

In Model 1:

Pit is dummy dependent variable and represents dividend payment statusfor firm i and

in year t. Pit equals to 1 when cash dividend payment is bigger than zero, and equals to

zero when cash dividend payment is zero. (TA) it-1 is an independent variable and

represents total assets for firm i in year t, and is used as the proxy of size. TA has been

logged and winsorised at 0.005 and 0.995 and lagged one year. (EBIT/TA) it-1 is an

independent variable and is calculated as earning before interest and tax divided by total assets for firm i in year t and is used as the proxy of profitability. EBIT/TA has

been winsorised at 0.005 and 0.995 and lagged one year. (MBF) it-1 is an independent

variable and represents market-to-book value for firm i in year t, and is used as the proxy of investment opportunities. MBF has been logged and winsorised at 0.005 and

0.995 and lagged one year. (DTA) it-1 it is an independent variable and represents

assets growth ratio for firm i in year t, and is also used as the proxy of investment opportunities. DTA has been winsorised at 0.005 and 0.995 and lagged one year. In Model 2:

(LP) it-1 is added as independent variable and represents last year’s payment status for

firm i. (LP) it-1 equals to 1 when cash dividend payment is bigger than zero, and

16

equals to zero when cash dividend payment is zero. If β5 is positive, it means last year

payment status has a positive effect on likelihood to pay and if β5 is negative, it means

last year payment status has a negative effect on likelihood to pay. I expect the positive effect because that is consistent with dividend stickiness.

Twu (2010) also considers that prior payment status has an important effect on the likelihood to pay. She categorizes firms into those that paid dividends previously and those that did not. And then she uses logistic regressions to analysis the relationship between likelihood to pay and above four variables (TA, EBIT/TA, MBF and DTA) separately for the two groups. I follow the same approach of Twu (2010) to conduct the logistic regression by considering prior payment status. Model 3 and model 4 are logistic regression models respectively for prior payers and prior nonpayers.

Model 3: Pit (Prior payers) = γ0 + γ1 * (TA) it-1 + γ2 * (EBIT/TA) it-1 + γ3* (MBF) it-1

+ γ4 * (DTA) it-1

Model 4: Pit (Prior nonpayers)= δ0 + δ1* (TA) it-1 + δ2 * (EBIT/TA) it-1 + δ3*

(MBF) it-1 +δ4 *(DTA) it-1

In Model 3 and Model 4:

Pit (Prior payers) is the dependent dummy variable of payment status for the firms that

paid dividends last year. Pit (Prior payers) equals to 1 when cash dividend payment is

bigger than zero, and equals to zero when cash dividend payment is zero. Pit (Prior

nonpayers) is dependent dummy variable of payment status for the firms that did not

pay dividends last year. Pit (Prior nonpayers) equals to 1 when cash dividend payment

is bigger than zero, and equals to zero when cash dividend payment is zero. The four independent variables are the same as described in Model 1.

4.2 Quantification of effects of characteristics and propensity to pay on the likelihood to pay

The effects of changing characteristics and propensity to pay on the percentage of dividend payers can be quantified. I quantify those effects by following the approach

17

of Fama and French (2001) as well as Twu (2010). Firstly, I estimate the probabilities1

that firms with given characteristics (independent variable in logistic regression) pay dividends during 1998-2000. Secondly, I apply the probabilities from the base period of 1998-2000 to the samples of firm characteristics observed in subsequent years and estimate the expected percentage of dividend payers for the years after 2000. Since the probabilities associated with characteristics are fixed at their base period values, variation in the expected percent of payers after 2000 is due to the changing characteristics of sample firms. The changes of expected percentage across years can be represented as the effect of changing characteristics on the percentage of firms paying dividends. Then I use the difference between the actual percentages of payers and the expected percent for a year to measure the propensity to pay dividends. The changes of the difference between actual and expected percent of payers across years can be represented as the effect of changes of propensity to pay dividends.

I use the following hypothesis to test if the propensity to pay changes across years.

H0: the difference between the actual and expected percentage of payers for all years

is similar to the difference for 2001.

If H0 is rejected, the difference between actual and expected percentage is different

over time. If the difference is significantly larger, the propensity to pay increases over time. If the difference is significantly smaller, the propensity to pay decreases over time.

4.3 Stickiness

I divide observed firms into four groups: firms that paid dividends last year and continue to pay the next year, firms that paid dividends last year but stop paying the next year, firms that did not pay dividends last year but start paying the next year and firms that did not pay dividends last year and do not pay the next year either. Then I assume that firms that paid dividends last year are more likely to continue to pay the next year and firms that did not pay dividends in year t-1 are more likely to still not

1_{For the base period 1998-2000, the coefficients between the independent variables as well as the intercept and} the likelihood to pay can be calculated by logistic regression in Eviews. And the coefficients can be used in calculating as the probability of expected percentage of dividend payers.

18 pay dividends the next year. This assumption means that dividend stickiness holds. Besides, I also add last year status as an independent variable for the logistic regression analysis. The coefficient between last year status and likelihood to pay for this year is expected to be positive, and it shows that the last year status has positive effect on the likelihood to pay. If the firm pays dividends last year, it is more likely for the firm to pay the current year, and if the firm does not pay dividends last year, it is more likely for the firm to not pay the current year. This is also in line with dividend stickiness proposed by Lintner (1956).

5 Empirical results

5.1 Univariate analysis for likelihood to pay

In table 2, I use means to show the summary statistics for each variable. Panel A, Panel B and Panel C show the means of firm characteristics of all firms, dividend payers and nonpayers separately. Characteristics are size, profitability and investment opportunities. TA is total assets of the firm and it is used as a proxy for characteristics of size. EBIT/TA is calculated as EBIT divided by total assets, and it is used as a proxy for characteristics of profitability. MBF is the market-to-book value of the firm and is calculated as market capitalization divided by common equity. DTA is assets growth rate of the firm. MBF and DTA are together used as a proxy for characteristics of investment opportunities. Fama and French (2001) and Denis and Osobov (2008) as well as Twu (2010) conclude that dividend payers tend to be larger, more profitability and less valuable investment opportunities firms. In Panel A, the mean of TA of payers is $1070201 thousand which is more than 10 times larger than $95884 thousand of nonpayers. The mean of EBIT/TA of payers is 0.070 which is more than the profitability of -1.075 of nonpayers. The mean of MTB of payers is 3.954 which is smaller than 17.341 of nonpayers and the mean of DTA of payers is 0.324 which is smaller than 3.552 of nonpayers. Therefore, Dividend payers have larger average size and higher average profitability, but less average investment opportunities than dividend nonpayers. My results show the same tendency as Fama and French (2001) and Denis and Osobov (2008). Panel B shows the evolution of firm characteristics over time for each characteristic. I show the averages of annual means in three sub-periods: 1997-2000, 2001-2004, and 2005-2009. From the first period to the

20 second period, firms tend to be larger, less profitable and to have more investment opportunities. From the second to the third period, the firms tend to be larger, more profitable and to have less investment opportunities. The result is inconsistent with Denis and Osobov (2008)’s results which shows that firms tend to be smaller, less profitable and having more investment opportunities over time.

5.2 Multivariate analysis for the likelihood to pay

I also quantify the marginal effects of size, profitability, investment opportunities on the likelihood to pay by using multivariate analysis. Annual logistic regressions are used similar to those in Fama and French (2001) as well as Denis and Osobov (2008). The dependent variable in my regressions equals one if a firm pays dividends in year t and zero otherwise. I use the total assets (TA) as proxy for firm size and earnings before interest divided by total assets (EBIT/TA) as a proxy for its profitability, and market-to-book value (MTB) together with asset growth rate (DTA) as proxies for

investment opportunities.2

In table 3, I estimate intercept and slope coefficients from annual regressions and then paper the averages of the annual coefficients for five periods-1998-2009, 1998-2000, 2001-2003, 2004-2006, 2007-2009-following the approach of Fama and MacBeth (1973). Three groups of all firms, prior payers and prior nonpayers are included in this table. Consistent with the findings of Fama and French (2001) as well as Denis and Osobov (2008), the likelihood to pay is positively related to firm size and profitability for all different kinds of payers across years. However, the effect of investment opportunities on likelihood to pay is not always negative. For the group of all firms, the coefficients on both investment opportunity proxies (MTB and DTA) are mostly significantly negative. There is only one exception of 0.032 which is the average coefficient of MTB for the subperiod 2007-2009, but it is not significant. For the group of prior payers, the effect of MTB on the likelihood to pay is positive and the effect of DTA on likelihood to pay is negative except for the sub-period 2007-2009. For the group of prior nonpayers, the effect of MTB on likelihood to pay is negative

2

Size and MBF have been logged and winsorised and lagged one year. DTA and EBIT/TA have been winsorised and lagged one year.

22 for the whole period and the first two subperiods, but positive for the next two subperiods. The effect of DTA on the likelihood to pay is mostly negative only except for the sub-period 1998-2000.

5.3 Dividend stickiness

Lintner (1956) firstly bring up the idea of dividend stickiness. He states that managers are reluctant to make dividend changes that might reversed. This implies that firms that currently pay dividends prefer continuous paying dividends. By contrast, firms that do not currently pay dividends prefer not paying them. Twu (2010) also develops this idea.

I also analyse dividend stickiness. In table 4, the group of firms that paid dividends in last year and that continues to pay in next year is bigger than the group of firms that pay dividends in last year but stop paying in next year. (e.g. in 1997, 96.3%>3.7%; in 2009, 84.9%>15.1%) The group of firms that did not pay dividends in last year still do not pay in the next year is bigger than the group of firms that did not pay dividends in last year and start paying in next year. (e.g. in 1997, 79.762%>20.283%; in 2009, 95.629%>4.371%) Therefore, stickiness holds. The firms who paid dividend last year are more likely to pay dividend the next year while the firms who did not pay dividend last year are more likely not to pay dividend the next year.

5.4 Quantitative evidence of the propensity to pay

5.4.1 Propensity to pay with different logistic variables

The effects of changing characteristics and propensity to pay on the percent of dividend payers can be quantified. Firstly, I estimate the probabilities that firms with given characteristics (size, profitability and investment opportunities) pay dividends during 1998-2000. Secondly, I apply the probabilities from the base period of 1998-2000 to the samples of firm characteristics observed in subsequent years and estimate the expected percentage of dividend payers for the years after 2000. Since the probabilities associated with the characteristics are fixed at their base period values, variation in the expected percentage of payers after 2000 is due to the

24

changing characteristics of sample firms.3 _{The changes of expected percentage across}

years can be represented as the effect of changing characteristics on the percentage of firms paying dividends. I then use the difference between the actual ratio of payers for a year and the expected ratio to measure the change in the propensity to pay dividends.

Table 5 shows the average coefficients from year-to-year logistic regressions for the base period 1998-2000. Two sets of results are shown. In panel A, the regressions use size (TA), profitability (EBIT/TA), and investment opportunities (MBF and DTA) to estimate the probability that a firm pays dividends. In panel B, the variable of last year’s payment status (LP) is added. The likelihood to pay is significantly positively related to firm size and profitability and significantly negatively related to investment opportunities for the logistic regressions in Panel A which is consistent with the results of Fama and French (2001). However, in Panel B, the likelihood to pay is significantly positively related to firm size and profitability, but not significantly negatively related to investment opportunities. Besides, in Panel B, the likelihood to pay is significantly positively correlated with the variable of last year payment status, and which implies that the last year payment status do has a positive influence on the likelihood to pay and that dividends are also sticky in United Kingdom.

Table 6, in Panel A, shows the actual ratio of dividend payers. The actual ratio of dividend payers declines from 64.4% in 2001 to 45.3% in 2009. In panel B, I consider

first the regression from table 5 for base period 1998-2000 that only useSIZE, EBIT/TA,

MTB, and DTA to explain the probabilities that a firm pays dividends. Then I apply the base period 1998-2000 average regression function to estimate the expected ratios of payers for each of the following years. The change in the expected ratios after 2000 is

due to the changing characteristics of sample firms, and the change of differences

between actual and expected ratios is due to the changing propensity to pay. In 2001, the expected percent of payers is 76.8%. The expected percent declines after 2001 until 66.5% in 2003. Then it begins to fluctuate and reaches to 72.2% in 2009. The 4.6% percentage point decline in the expected percent of payers between 2001 and 2009 is an estimate of the effect of changing characteristics on the ratio of firms

3 _{In Table 4, for base period 1998-2000, the coefficients between the four variables as well as the intercept and the} likelihood to pay can be calculated by logistic regression in Eviews. And the coefficient can be represented as the probability of characteristics being used for calculating expected percent of dividend payers.

27 paying dividends. Then, by comparing the actual ratio of payers and the expected ratio of payers for each year, the effect of the change in the propensity to pay dividends can be inferred. In 2001, the actual percent is only 12.5% below the expected ratio. By 2009, the actual percent is 26.9% below the expected ratio. The 14.4% decline of difference between actual and expected percentage to pay during the period 2001-2009 shows a decline of propensity to pay in the United Kingdom.

In Panel C, with the variable of last year payment status (LP) added in the logistic regression, the expected percent of dividend payers declines from 68.1% in 2001 to 47.4% in 2007 and increase slightly until 50.4% in 2009. The 17.7% percentage point decline in the expected percent of payers between 2001 and 2009 is an estimate of the effect of changing characteristics on the percent of firms paying dividends. Then, for the difference between actual and expected ratios of payers, in 2001, the actual ratio is 3.7% below the expected ratio, and then the difference is going to decrease until in 2003, the actual ratio is 0.6% lower than expected ratio. From 2004 to 2007, the actual ratio is higher than expected ratio and the differences are 0.008, 0.008, 0.001 and 0.015 respectively. Then, in 2008 and 2009, actual ratio returns to be lower than the expected ratio again. The fluctuation in the difference reflects a fluctuation in the propensity to pay. Besides, by comparing the difference between actual and expected percents in Panel B and in Panel C, we can see that the expected percents predicted in Panel C are much more closed to the actual percents compared to the expected percents predicted in Panel B through 2001-2009. Therefore, we can come to the conclusion that the last year payment status can positively influence the likelihood to pay. And it can imply that dividend stickiness holds. Moreover, it shows that the propensity to pay has not declined if we take stickiness into account. This counter- intuitive effect is primarily caused by the fact that the number of dividend payers is declining over time if we consider the payers of last years. Then, when the decline in the last year payers is taken into account in the forward analysis, the expected ratio of dividend payers will also become smaller. And the propensity to pay will become less low. This result means that the idea of Fama and French (2001) is not likely to be captured very well if we take previous payment status into account.

28 5.4.2 Test of the hypothesis of propensity to pay

H0: the difference between the actual and expected ratio of payers for all years are

similar to the difference for 2001.

H1: the difference between the actual and expected ratio of payers for all years are not

similar to the difference for 2001.

Table 7 gives the significance probabilities for t-tests for the propensities to pay with or without last year payment status from table 6. If these probabilities are smaller than 0.05, they are significant and it means that the propensity to pay is significant different from that for 2001. In panel A, the difference is significantly smaller, the propensity to pay decreases over time. In panel B, the propensity to pay differs over time and does not decrease over time.

5.4.3 Propensity to pay for different dividend groups

Following Twu (2010)’s approach, the sample firms are divided into two different groups based on their last year’s payment status: those firms that paid dividend in the last year (prior payers) and those that did not pay dividend in the last year (prior nonpayers). Table 8 summarizes logistic regressions estimated separately for firms classified as prior payers and prior nonpayers. For Panel A and Panel B, the difference of the base period (1998-2000) average coefficients between prior payers and prior nonpayers shows that the decision to pay dividends will be influenced by last year’s

payment status. For the independent variables of SIZE, EBIT/TA, and MBF, the

probability that a dividend payer continues to pay is higher than the probability that a nonpayer with the same characteristics starts paying. (0.355>0.143, 8.465>7.027, 0.165>-0.197) However, for the independent variable of DTA, the probability that a dividend payer continues to pay is lower than the probability that a nonpayer with the same characteristics starts paying. (-0.191<0.016)

Table 9 shows up how changing characteristics and propensity to pay affect the likelihood to pay across the groups of prior payers and prior nonpayers. I use average logistic coefficients of base period of 1998-2000 for each dividend group

32 from table 8 to estimate expected ratios of payers for each group in subsequent years. In Panel A, for prior payers, the ratio of year t-1 dividend payers is expected to continue paying in year t increases from 97.9% in 2001 to 98.5% in 2009. Therefore, the characteristics of former payers do not change much through time, and the changes of characteristics have a small positive effect on the likelihood to pay. The result is different from Denis and Osobov (2008) who point that there are growing number of firms with typical characteristics (low size, low profitability and high investment opportunity) of those that have never paid dividend. However, the actual ratio of continuing payers falls from 91.8% in 2001 to 86.1% in 2009. The difference between actual and expected ratio is negative and it shows that there is an overestimation of the ratio of expected payers. For the difference between actual and expected ratio, in 2001, the actual percent is only 5.9% below the expected percent. By 2009, the actual percent is 12.4% below the expected percent. The 6.5% decline of difference between actual and expected percentage to pay during the period 2001-2009 shows the decline of the propensity to pay. This decline in the propensity to pay results from the decrease of the actual ratio of dividend payers.

In Panel B, for prior nonpayers, changing characteristics and lower propensity to pay also have effects on the dividend decisions of prior nonpayers. When the average coefficients of the base period of 1998-2000 regressions for dividend prior nonpayers from table 8 are applied to the sample of prior nonpayers for later years, the percent of year t-1 dividend payers who are expected to start paying in year t ranges from 0.7% to 3.1% due to changes in characteristics. The difference between actual and expected ratios is positive and it shows that there is an underestimation of the ratio of expected payers. The differences range from 3.1% to 7.5% with an average of 5.5% in the forecasting period. The fluctuation of the difference implies a fluctuation of the propensity to pay for nonpayers. And it does not show a lower propensity to pay but only a fluctuation for nonpayers. Besides, the average ratio of prior payers who continue to pay is 93.1% while the average ratio of prior nonpayers who start to pay is 5.5%. The result shows dividend stickiness.

33 5.4.4 Test of the hypothesis of propensity to pay

H0: the difference between the actual and expected ratio of payers for all years are

similar to the difference for 2001.

H1: the difference between the actual and expected ratio of payers for all years are not

similar to the difference for 2001.

Table 10 gives the significance probabilities for t-tests for the propensities to pay for prior payers and prior nonpayers from table 9. If these probabilities are smaller than 0.05, they are significant and it means that the propensity to pay is significantly different from that for 2001. In panel A and panel B of table 10, the propensity to pay differs over time and the probabilities are not always significant, and so the propensity to pay does not show to a decrease over time.

5.4.5 Implication for dividend stickiness

Some results in previous test can imply dividend stickiness. In table 6, last year status is also added as an independent variable for logistic regression analysis. The coefficient between last year status and likelihood to pay for this year is positive, and this shows that the last year status has positive effect on the likelihood to pay. If the firm paid dividends last year, it is more likely for the firm to pay this year, and if the firm did not pay dividends last year, it is more likely for the firm to not pay this year. It is also consistent with dividend stickiness proposed by Lintner (1956). In table 9, the average ratio of prior payers who continue to pay is 93.1% while the average ratio of prior nonpayers who start to pay is 5.5%. This result also implies dividend stickiness. Moreover, when I consider the changes in the propensity to pay as shown in table 10, this does not give a systematic change in the propensity to pay. Again this suggests no wish to abandon the previous year’s payment status. Whether such a stickiness of the propensity to pay also captures the ideas of Fama and French (2001) can be doubted.

35

6 Conclusion

My analysis shows firms that are larger, more profitable and with less investment opportunities are more likely to pay dividends and this result is consistent with the results for the US and other countries related to this issue, such as the results of Fama and French (2010), Denis and Osobov (2008). However, my results for a more recent period do not show that firms tend to become smaller, less profitable and are getting more investment opportunities across time. The result of the logistic regression for size shows that it is positively related to the likelihood to pay and the result is consistent with lifecycle theory. The result of the logistic regression for profitability shows that profitability is positively related to the likelihood to pay and the result is consistent with view of Lintner (1956) and signaling theory. The result of the logistic regression for investment opportunities shows that investment opportunities are negatively related to the likelihood to pay and the result is consistent with the pecking order theory but goes against signaling theory. However, there are some exceptions, for prior payers, the relationship between MTB as proxy for investment opportunities and the likelihood to pay is positive, and for prior nonpayers, the relationship between DTA as proxy for investment opportunities and likelihood to pay is also positive. These results are in line with signaling theory. The opposite results for investment opportunities for the full sample or the two subgroups need further analysis.

The fraction of British firms paying dividends declines from 76.6% in 1998 to 40.5% in 2009. After analysis by using logistic regressions, I come to the conclusion that this is due in part to the changing characteristics of publicly traded firms and in part to firms’ lower propensity to pay dividends. Both factors are essential for explaining the declining incidence of dividend-paying firms. Besides, dividend stickiness presented by Lintner (1956) and in line with Twu (2010) has also been proved in this paper. Moreover, it shows that the propensity to pay has not declined if we take stickiness into account though it may be precise, the approach of Twu (2010) does not capture the basis concept of the declining propensity to pay of Fama and French (2001) because she takes each year the declined propensity of last year into account. I prefer the interpretation of Fama and French (2001) and conclude a declining propensity to pay for the United Kingdom.

36 Lower propensity to pay implicates that the perceived benefits of dividends have declined through time. We can guess that the decline may have resulted from lower transaction costs for selling stocks for consumption or that there are larger holdings of stock options by managers who prefer capital gains to dividends. The reasons remain to be analyzed.

37

Reference

Adedeji, A., 1998. Does the pecking order hypothesis explain the dividend payout ratios of firms in the UK, Journal of Business Finance & Accounting, 25(9) & (10), November/December 1998, 0306-686X

Bhattacharya, S., 1979. Imperfect information, dividend policy, and ‘The Bird-in-the-Hand’ fallacy, Bell Journal of Economics 10, 259–270.

Bulan, L., Subramanian N., and Tanlu L., 2007. On the timing of dividend initiations, Financial

Management 36, 31–65.

DeAngelo, H. and DeAngelo L., 2006. The irrelevance of the MM dividend irrelevance theorem,

Journal of Financial Economics 79 (2006) 293–315

DeAngelo, H., DeAngelo L., and Stulz, R. M., 2006. Dividend policy and the earned/contributed capital mix: A test of the life cycle theory, Journal of Financial Economics 81, 227–254.

Denis, D.J. and Osobov I., 2008. Why do firms pay dividend? International evidence on the determinants of dividend policy, Journal of Financial Economics 89, 62–82.

Fama, E. and MacBeth, J., 1973. Risk, return, and equilibrium: empirical tests. Journal of Political

Economy 81, 607-636.

Fama, E.F. and K.R. French, 2001. Disappearing dividends: Changing firm characteristics or lower propensity to pay? Journal of Financial Economics 60, 3–44.

Garrett, I. and Priestley, R., 2000. Dividend behavior and dividend signaling, Journal of financial

and quantitative analysis, Vol 3. 5, NO. 2

John, K. and Williams, J., 1985. Dividends, dilutions, and taxes: A signaling equilibrium, Journal

of Finance 40, 1053–1070.

Lintner, J., 1956. Distribution of incomes of corporations among dividends, retained earnings, and taxes, American Economic Review 46, 97–113.

Twu, M. 2010. Prior payment status and the likelihood to pay dividends: international evidence,

The Financial Review 45 (2010) 785–802

Miller, M. H. and Modjliani F. 1961. Dividend policy, growth, and the valuation of shares,

Journal of Business, Vol. 34, No. 4 (Oct., 1961), pp. 411-433

Miller, M.H. and K. Rock, 1985. Dividend policy under asymmetric information, Journal of

Finance 40, 1031–1051.

Myers, S. and Majluf, N., 1984. Corporate financing and investment decisions when firms have information the investors do not have. Journal of Financial Economics 13, 187-221.

Von Eije, H. and W. Megginson, 2008. Dividends and share repurchases in the European union,

Journal of Financial Economics 89, 347–374