- Email: [email protected]
The link between dividend policy and institutional ownership
Journal of Corporate Finance 8 (2002) 105 – 122 www.elsevier.com/locate/econbase
The link between dividend policy and institutional ownership Helen Short *,1, Hao Zhang 1, Kevin Keasey
1
The International Institute of Banking and Financial Services (IIBFS), Leeds University Business School, University of Leeds, Leeds, LS2 9JT, UK Accepted 22 June 2001
Abstract This paper examines the relatively neglected link between dividend policy and institutional ownership. It is also the first example of using well-established dividend payout models to examine the potential association between ownership structures and dividend policy. Moreover, the paper presents the first results for the UK, where the institutional framework and ownership structures are different from those of the US. Using a UK panel data set, the role of institutional ownership in association to dividend payout ratios is analysed within the context of the dividend models of Lintner [American Economic Review, 46 (1956) 97], Waud [Journal of the American Statistical Association, 1996] and Fama and Babiak [Journal of the American Statistical Association, 63 (1968) 1132]. The results consistently produce strong support for the hypothesis that a positive association exists between dividend payout policy and institutional ownership. Furthermore, the results for an earnings trend model suggest a positive earnings trend component to the association between institutional ownership and the dividend payout ratio. In addition, there is some evidence in support of the hypothesis that a negative association exists between dividend payout policy and managerial ownership. D 2002 Elsevier Science B.V. All rights reserved. JEL classification: G32 Keywords: Dividend policy; Institutional ownership
*
Corresponding author. Tel.: +44-113-233-4471; fax: +44-113-233-4459. E-mail address: [email protected] (H. Short). 1 Helen Short, Hao Zhang and Kevin Keasey are, respectively, Senior Lecturer in Finance and Accounting, Senior Lecturer in Finance and Accounting, Halifax plc Professor of Financial Services, at IIBFS, University of Leeds. 0929-1199/02/$ - see front matter D 2002 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 9 - 11 9 9 ( 0 1 ) 0 0 0 3 0 - X
106
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
1. Introduction The link between a firm’s financial policy and its ownership structure is recognised in the established literature (for example, Williamson, 1964; Leland and Pyle, 1977; Jensen, 1986). While the empirical evidence of the relation between dividend policy and management ownership has also been well documented within this literature (for example, Rozeff, 1982; Jensen et al., 1992; Eckbo and Verma, 1994; Moh’d et al., 1995), the potential relation between dividend policy and institutional ownership has been somewhat neglected and this is especially the case for non-US firms where the institutional framework and ownership structures are different from those of the US. In terms of the institutional ownership of firms, there are significant differences between the US and UK. These differences result from the differences in legal restrictions and tax incentives. In fact, institutional shareholdings in the US are only about two-thirds of those in the UK (Kester, 1992). The significantly lower concentration of US institutional shareholdings reflects both the legal restrictions preventing US institutions from building significant stakes in individual firms (see Roe, 1990, for a discussion) and the tax system being largely neutral in relation to dividend policy. In contrast, the UK tax system has (in the period under consideration) provided clear incentives for tax-exempt institutions to demand dividends. Using a UK panel data set, this paper examines the possible link between dividend policy and institutional ownership. In particular, the role of institutional ownership in relation to dividend payout ratios is analysed within the context of the dividend models of Lintner (1956), Waud (1966) and Fama and Babiak (1968). The evidence from four wellknown types of dividend models consistently shows positive and statistically significant associations between institutional ownership and dividend payout ratios and thus suggests a link between institutional ownership and dividend policy. The paper is organised as follows. Section 2 discusses the relation between dividend policy and institutional ownership in the context of theories that emphasise the importance of taxation, agency and signalling for dividend decisions. A discussion of the statistical models of dividend payout behaviour is provided in Section 3. Section 4 discusses the sample, variables and empirical method to be used in examining the association between dividends and institutional ownership. Section 5 provides the results and a discussion of the empirical analysis. Finally, Section 6 presents the conclusions.
2. Institutional shareholders and dividend payout policy The relatively high level of dividends paid by UK quoted firms has been the subject of concern for a number of years. In particular, the dividend payout rates of UK firms are significantly greater than those of other countries, such as Germany and Japan, and dividends in the UK are considered to be downwardly inflexible (Mayer and Alexander, 1990; Mayer, 1994). The relatively high levels of dividends paid by UK firms have been argued to reduce the amount of funds available for long-term investment, to the detriment of long-term economic performance (Bond and Meghir, 1994). In particular, attention has focused on the alleged role of institutional shareholders in forcing firms to maintain high
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
107
dividends, particularly in the face of falling profits in the UK recession of the late 1980s/ early 1990s (for a detailed discussion, see the Report of the Trade and Industry Committee on Competitiveness, 1994). While the high level of dividends paid by firms could be construed as reflecting the short-term attitudes held by institutions (see, for example, Hutton, 1995; Haskins, 1995), it could equally be construed as the result of institutional investors’ effort to reduce ‘‘free’’ cash flows available to the management (Jensen, 1986). The ownership of equity in the UK is dominated by institutional shareholders and this may affect the dividend payout ratio in a number of ways. First, a bias in the UK tax system which favours dividends over capital gains, particularly for tax-exempt institutions such as pension funds, provides incentives for institutions to demand relatively high levels of dividends. Second, dividends may be used as a mechanism to reduce agency problems existing between shareholders and managers. Third, given the presence of information asymmetries existing between shareholders and managers as a result of the capital market’s emphasis on arm’s length funding and liquidity, dividends may be used to signal future prospects of the firm. The following subsections detail the theoretical and empirical implications of taxation, agency costs and signalling considerations on the relation between dividends and institutional shareholders. 2.1. Taxation The US operates a classical company tax system whereby companies are taxed separately from their shareholders. Firms pay a flat rate of corporation tax on their profits and shareholders pay income tax on the dividend income they receive at their marginal rates of income tax. This means that dividends are essentially taxed twice, first in the form of corporation tax on firm profits, and second, in the form of income tax on dividend income. An important outcome of the classical system is that both basic and high rate income taxpayers will prefer profits to be retained in the firm rather than paid out in dividends, with tax-exempt shareholders being neutral with respect to dividends and retained profits. This leads to the so-called classical dividends puzzle prevalent in the US literature (Black, 1976); namely, why do firms pay dividends at all? In contrast, the UK operates a partial imputation company tax system whereby corporation tax is charged on firm profits but part of the corporation tax paid is taken into account when assessing shareholders’ liability to income tax. The workings of the UK system of corporate taxation are detailed in Appendix A. An important outcome of the UK system is that tax-exempt shareholders prefer dividends to retentions, basic rate taxpayers are neutral with respect to dividends and retentions, whilst higher rate taxpayers prefer retentions to dividends. The bias in the UK system for tax-exempt shareholders to prefer dividends to retentions may result in relatively higher dividend payments. Bond et al. (1995) argue, To the extent that tax-exempt shareholders such as pension funds are now the most influential investors in many UK companies, their tax preference for dividend income is likely to result in significantly higher dividend payout ratios than would be chosen by companies in the absence of this tax bias. (p. 17)
108
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
A related issue is the need of institutional shareholders for funds on an ongoing basis. Clearly, institutions invest in equities in order to provide returns to fund their activities, such as funding pensions, paying out in insurance policies, etc. Regardless of the tax bias in favour of dividends, institutions cannot simply rely on capital gains to fund their liabilities and, hence, require dividend payments. Furthermore, UK pension funds are, on average, mature in the sense that pensions are paid out of investment income rather than from new cash flow from contributions. Given the investment profile of pension funds with their reliance on equities, pension funds require a certain level of dividends to match their liabilities. Moreover, the actuarial value of pension funds is in part based on dividend income. Institutions’ requirements for certain levels of dividends to meet their own liabilities may force companies to pay out dividends at a level higher than they would otherwise prefer, particularly in times of recession and low corporate profitability. Hence, as a result of the tax bias in favour of dividends for tax-exempt institutions and the need of institutions to maintain income flows in the form of dividends received, it is hypothesised that institutional shareholdings will have a positive effect on the dividend payout ratio. 2.2. Agency theory The payment of dividends may act to help reduce agency costs which arise between managers and external shareholders. Rozeff (1982) and Easterbrook (1984) argue that the payment of dividends forces firms to go to the external capital markets for additional funding and, hence, undergo monitoring by the capital market. In this respect, Zeckhauser and Pound (1990) suggest that institutional shareholders may act as a substituting monitoring device, hence reducing the need for external monitoring by the capital markets. However, the arm’s length view of investment held by many institutional investors, coupled with the incentives to free ride with respect to monitoring activities, suggests that institutional shareholders are unlikely to provide direct monitoring themselves. Indeed, Zeckhauser and Pound suggest that institutions, rather than providing monitoring themselves, force firms to increase their dividends in order that they are subsequently forced to go to the external capital market for future funds. Jensen’s (1986) free cash flow theory suggests that managers are reluctant to pay out dividends, preferring instead to retain resources under their control. Eckbo and Verma (1994) argue that institutional shareholders will prefer free cash flow to be distributed in the form of dividends in order to reduce the agency costs of free cash flow. From this perspective, it may be argued that institutional shareholders may counter a tendency for managers to prefer the excessive retention of cash flow and, by virtue of their voting power, force managers to pay out dividends. The agency perspective, therefore, hypothesises a positive relation between institutional shareholders and the dividend payout ratio, as institutions demand dividends and force firms to the capital market for future funds in order to reduce the agency costs of free cash flow. 2.3. Signalling The signalling perspective suggests that dividends are used to signal managements’ private information regarding the future earnings of the firm (Bhattacharya, 1979, 1980;
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
109
Miller and Rock, 1985). Zeckhauser and Pound (1990) suggest that dividends and institutional shareholders may be viewed as alternative signalling devices. The presence of a large shareholder may mitigate the use of dividends as a signal of good performance, as the large shareholders themselves can act as a (more) credible signal. Under this hypothesis, Zeckhauser and Pound suggest that a negative relation is predicted to exist between dividends and large shareholders. However, it is not clear exactly how institutional shareholders would act as a signal of future prospects of the firm to the market. There are two possible scenarios. First, the presence of institutional shareholders may signal to the market that agency costs are reduced due to the monitoring activities of the institutional shareholders. However, as already discussed, due to free rider problems it is unlikely that institutional shareholders are willing to provide direct monitoring. Second, given the superior information that institutional shareholders are likely to have concerning the future prospects of the firm (because of their easier access to and control of the management, etc.), it is possible that the market may interpret the presence of an institutional shareholder as signalling good news regarding the firm’s future prospects. However, whilst this explanation holds some attractions, the available evidence is not strongly in favour of this scenario. Given the emphasis placed on dividends as a signal of future prospects by institutions themselves (see Trade and Industry Committee on Competitiveness, 1994), it would appear that institutions do not regard themselves as substitute signalling mechanisms. In addition, the institutions need to be wary of being privy to inside information because of the insider dealing laws. Furthermore, the growth in indexed funds means that the presence of an institutional shareholder does not necessarily imply that the particular institution believes that the firm has better than average prospects. Overall, the notion that dividends and institutional shareholders act as substituting signalling devices, while possible, is less than convincing. 2.4. Summary and empirical evidence In summary, the relation between dividends and institutional ownership would appear to depend on the relative weights of tax, agency and signalling considerations. From the tax perspective, there are clear incentives for (tax-exempt) institutions to demand high levels of dividends as a result of the bias in the UK tax system in favour of dividends for tax-exempt shareholders. Furthermore, institutions require certain levels of dividends to meet their own liabilities. From the agency perspective, institutions may demand high levels of dividends in order to force firms to go to the capital for external funding and, hence, be subject to monitoring by the external market. From the free cash flow perspective, institutions may counter management’s tendency to retain excess free cash flow. Hence, both the tax and agency related perspectives suggest that a positive association exists between dividends and institutional shareholdings. In contrast, the signalling arguments suggest that dividends and institutions may act as substitute signalling devices. However, as discussed, this is possible but less than convincing in the context of the UK. Therefore, on the basis of tax and agency considerations, we hypothesise the following: A positive association is expected to exist between dividend changes and institutional shareholdings.
110
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
The empirical evidence concerning the link between institutional shareholders and dividend payments is extremely limited. Eckbo and Verma (1994), from a sample of Canadian quoted companies, report that the cash dividend yield increases significantly with the voting power of corporate/institutional shareholders and decreases significantly with the voting power of management. Moh’d et al. (1995) report a significant and positive association between dividends and institutional shareholdings. Alternatively, Zeckhauser and Pound (1990), investigating US companies, find no evidence to support the view that institutional shareholders have an impact on dividend policy.2
3. Models Four types of dividend models are used to test the hypothesised positive link between institutional ownership and dividend policy: the Full Adjustment Model, the Partial Adjustment Model (Lintner, 1956), the Waud Model (1966), and the Earnings Trend Model (Fama and Babiak, 1968). These models are modified by interactive dummy variables to account for the possible effects of institutional ownership. In addition, we also allow for the potential link between managerial ownership and dividend policy (Rozeff, 1982; Easterbrook, 1984), as there is a significant body of empirical literature to suggest a negative link between the two (Rozeff, 1982; Jensen et al., 1992; Eckbo and Verma, 1994; Agrawal and Jayaraman, 1994; Moh’d et al., 1995). 3.1. The full adjustment model (FAM) If changes in income are considered permanent and a firm has a desired payout ratio r, then the association between changes in earnings (E) and changes in dividends (D), for firm i at time t, will be given by: Dti Dðt1Þi ¼ a þ rðEti Eðt1Þi Þ þ lti If we assume that firms with significant institutional ownership and/or managerial ownership may have a different r, then the model becomes: Dti Dðt1Þi ¼ a þ rðEti Eðt1Þi Þ þ rI ðEti Eðt1Þi Þ Inst þ rM ðEti Eðt1Þi Þ MDum þ lti
ðModel 1; FAMÞ
Inst is a dummy variable denoting the presence of significant institutional ownership and MDum is a dummy variable denoting the presence of significant managerial ownership. The coefficients rI and rM denote the respective impacts of institutional ownership and managerial ownership in association to the dividend payout ratio of the firm. 2 However, the Zeckhauser and Pound study utilises an industry aggregate sample (as opposed to individual firm level data) and, moreover, use a threshold of 15% to denote the presence of an institutional shareholder, both of which may be inappropriate.
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
111
As discussed in the previous section, institutional ownership is hypothesised to be positively linked to the dividend payout ratio, whereas managerial ownership is expected to be negatively linked to dividend payout ratio. Hence, the coefficient rI is expected to be positive and statistically significant and the coefficient rM is expected to be negative and statistically significant. 3.2. The partial adjustment model (PAM) The partial adjustment model assumes that for any year, t, the target level of dividend, D for firm i at time t is related to profits, Eti, by a desired payout ratio, r: Dti ¼ rEti If firms with significant institutional ownership and/or managerial ownership have a different r, then the model becomes: Dti ¼ rEti þ rI Eti Inst þ rM Eti MDum In any given year, the firm adjusts only partially to the target dividend level. Thus: Dti Dðt1Þi ¼ a þ cðDti Dðt1Þi Þ where a=a constant representing the resistance of management to reduce dividends, and c equals the ‘speed of adjustment coefficient’ which represents the extent to which the management wishes to ‘play-safe’ by not adjusting to the new target immediately. Substitution yields the following reduced form: Dti Dðt1Þi ¼ a þ crEti þ crI Eti Inst þ crM Eti MDum cDðt1Þi þ lti ðModel 2; PAMÞ As before, the coefficient (crI) on the interactive variable EtiInst variable is expected to be positive while that on the EtiMDum variable is expected to be negative. 3.3. The Waud model (WM) The Waud model incorporates elements of both partial and full adjustment. It assumes that the target dividends, D , for firm i at time t, are proportional to the long-run expected earnings, E : Dti ¼ rEti and that the actual dividend change follows a partial adjustment mechanism: Dti Dðt1Þi ¼ a þ cðDti Dðt1Þi Þ þ lti The formation of expectations follows an adaptive expectation model: Eti Eðt1Þi ¼ dðEti Eðt1Þi Þ
112
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
Assuming a possible difference in payout ratio for firms with significant institutional and/ or managerial ownership, the reduced form becomes:3 Dti Dðt1Þi ¼ ad þ cdrEti þ cdrI Eti Inst þ cdrM Eti MDum þð1 d cÞDðt1Þi ð1 dÞð1 cÞDðt2Þi lti ðModel 3; WMÞ Again the coefficient of EtiInst variable is expected to be positive whilst that of EtiMDum variables is expected to be negative. 3.4. Earnings trend model (ETM) The earnings trend model is a modified partial adjustment model. It assumes a specific profit generating process, for firm i at time t, of the form: Eti ¼ ð1 þ jÞEðt1Þi where j is an earnings trend factor. Assuming a possible difference in the earnings trend factor for firms with significant institutional and/or managerial ownership, the profit generating process becomes:4 Eti ¼ Eðt1Þi þ jEðt1Þi þ jI Eðt1Þi Inst þ jM Eðt1Þi MDum Target dividends are given by: Dti ¼ rEti It is assumed that there is full adjustment of dividends to the expected change, that is, jEðt1Þ þ jI Eðt1Þ Inst þ jM Eðt1Þ MDum and partial adjustment to the reminder. The resulting reduced form is: Dti Dðt1Þi ¼ a þ c½rðEti jEðt1Þi jI Eðt1Þi Inst jM Eðt1Þi MDumÞ cDðt1Þi þ rjEðt1Þi þrjI Eðt1Þi Inst þ rjM Eðt1Þi MDum þ lti Arranging, the reduced form becomes: Dti Dðt1Þi ¼ a þ rcEti þ rjð1 cÞEðt1Þi þ rjI ð1 cÞEðt1Þi Inst þrjM ð1 cÞEðt1Þi MDum cDðt1Þi þ lit ðModel 4; ETMÞ 3
See Waud (1966) for detailed derivation. It turned out that, in ETM, introducing dummy variables to account for the possible difference in payout ratio (as opposed to earnings trend factor) for firms with significant institutional and/or managerial ownership produces no reduced form. 4
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
113
Once again the coefficient of E(t1)i*Inst is expected to be positive, whereas the cofficient of E(t1)i*MDum is expected to be negative.
4. Sample, variables and empirical method 4.1. Sample The empirical analysis of the association between dividends and equity ownership is conducted on a sample of 211 firms listed on the London Stock Exchange Official List for the period 1988 to 1992. For a firm to be included in the sample, several criteria had to be met. First, the firm had to be listed on the Official List for the whole of the period under consideration. In addition, the firm would also be required to be listed in the year 1987. This condition was imposed to ensure that dividend policy was not distorted by the effects of a recent Official Listing. Second, firms in the financial and oil and gas sectors were excluded due to the different income measuring rules governing such firms, as compared to those in the manufacturing and service sectors. Third, privatised firms (such as water, electricity, gas, telecommunications) were excluded as their operating conditions (in terms of regulation and monopoly markets) are usually atypical. Fourth, firms operating in the broadcasting sectors were excluded due to the regulatory nature of the market and the changes in regulation that occurred during the period (for example, tendering for operating licences and changes in ownership rules which are largely government controlled). Finally, firms with ownership structures that did not conform to the typical one vote for each ordinary share were also excluded. As this paper specifically examines the potential link between ownership structure and dividend policy, it is important that the ownership structure and the associated voting structure of the equity shares of a particular firm can be determined. As a result, firms whose share structure included non-ordinary voting shares (such as voting A shares, management shares, founders’ shares) were excluded.5 Other than being subject to the criteria outlined above, a sample of 225 firms was selected entirely at random from all firms quoted on the London Stock Exchange Official List. The sample was further reduced to 211 firms, as a result of missing data. 4.2. Variables In order to examine empirically the dividend models discussed in Section 3, the key variables of interest are measures of dividends (D), earnings (E), institutional shareholdings (Inst) and managerial shareholdings (MDum). This section outlines the variables to be used to examine the association between institutional shareholdings and the dividend payout ratio. The dividend and profit variables were derived from data collected from the 5 Non-ordinary voting shares often carry more votes per share than ordinary shares and would considerably complicate the determination of the ownership structure of individual firms. As the trend is towards share structures including only ordinary voting shares and, in addition, in many cases, firms with such atypical share structures often change the nature of the shares in the period under consideration, exclusion of such firms should not materially affect the results.
114
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
financial database Datastream. Data on ownership was hand collected from the sample firms’ annual reports. 4.2.1. Dividends Dividends (D) are calculated as the total amount of ordinary dividends relating to the accounting year. 4.2.2. Earnings The earnings variable (E) is calculated as net profit derived from normal trading activities after depreciation and other operating provisions. A similar measure was utilised by Jensen et al. (1992). 4.2.3. Institutional ownership Institutional ownership is defined as the percentage of equity held by institutions owning 5% or more of equity at the beginning of the accounting year. For the period 1988 to 1990, firms were obliged to provide information in the annual report of external shareholdings amounting to 5% or more of equity shares in issue. The reporting requirements changed in June 1990, such that from 1990 to 1992, firms were obliged to provide information in the annual report of external shareholdings amounting to 3% or more of equity shares in issue. In order to avoid complications caused by this change in reporting requirements, a cut-off point of 5% institutional ownership was used for the whole time period. The analysis uses dummy variables based on the percentage of institutional ownership; Inst is coded 1 if an institution owning 5% or more of equity is present and 0 otherwise. 4.2.4. Management ownership Management ownership is defined as the total percentage of equity owned by directors and their immediate families at the beginning of the accounting year. This measure includes directors’ ownership via corporate vehicles, for example, where directors are majority shareholders in other firms that have direct ownership stakes in the sample firm under consideration.6 The analysis uses a dummy variable based on the percentage of management ownership. MDum is coded 1 if management ownership is greater or equal to 5% of equity and 0 otherwise. 4.3. Empirical method As noted in Section 4.1, the data consist of a sample of 211 firms covering the period from 1988 to 1992. The dividend models FAM, PAM and ETM focus on the changes in dividends and earnings between 2 years (for example, from t1 to t). For each period (for example, t1 to t), ownership measures that are already in place at the start of the period (for example, at t1) are used. Hence, for these three models, there are four periods of
6 Directors’ non-beneficial ownership of equity is included in the measure of management ownership, as directors have voting control over these shares.
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
115
observations (1988 – 1989, 1989– 1990, 1990– 1991 and 1991 – 1992). However, since dividend information going back two periods is required for the dividend model (WM), there are three periods of observations (i.e. 1988 – 1990, 1989 –1991 and 1990 –1992) in this instance. The fixed effect panel data estimation method is used to test the stated hypothesis regarding institutional ownership and the dividend payout ratio (see Baltagi, 1995). The resulting firm-specific intercepts account for firm-specific characteristics in the sample.
5. Empirical analysis and results Summary statistics relating to the dependent and explanatory variables are presented in Table 1. The mean dividend payment has increased in every year of the period under consideration. Mean earnings, while increasing over the period 1988– 1989, declined in each year thereafter; evidence of the economic recession that occurred in the UK over that period. The increase in dividends occurring while earnings declined is indicative of the reluctance of firms to decrease dividends in line with earnings reductions. The data pertaining to ownership demonstrates interesting changes in the nature of ownership of the period. The concentration of institutional ownership appears to have increased significantly, with 84% of the sample having an institution owning 5% or more of equity in 1992 compared with 57% of the sample having such an institutional owner in 1988. Whilst institutional ownership has increased, management ownership has remained relatively static over the period, with approximately 51% of the sample having directors owning 5% or more of the equity for the 1988 – 1991 period. However, there are
Table 1 Summary statistics Variable
Year 1988
Period 1989
1990
1991
1992
1988 – 1992
Dividend (D) (£’000’s) Mean 7054.12 Median 1556.00 SD 14,513.29
8637.20 1963.00 17,701.53
9610.50 2349.00 20,022.38
10,154.50 1999.00 22,542.83
11,005.22 2220.00 25,285.72
9491.18 2122.00 20,545.92
Earnings (E) (£’000’s) Mean 32,809.40 Median 7923.00 SD 69,863.96
37,516.24 9755.00 79,785.45
37,381.35 9461.00 81,671.03
34,012.71 7142.00 84,128.86
33,063.33 6484.00 86,097.08
36,002.72 8740.00 81,198.51
Institutional ownership (Inst=1) Number 121 % 57.3
129 61.1
153 72.5
161 76.3
177 83.9
n/a n/a
Management ownership (MDum=1) Number 109 110 % 51.7 52.1
108 51.2
109 51.7
98 46.4
n/a n/a
116
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
indications that this level of ownership is on the decline, as there is a drop in such ownership to 46% of the sample in 1992. Clearly, the difference between institutional ownership and management ownership (in terms of 5% or more equity ownership) widens significantly over the sample period. The results of the full adjustment model (FAM) are shown in Table 2. The coefficients for (EtEt1) and (EtEt1)*Inst are positive and statistically significant. The dividend payout ratio is, on average, significantly higher for firms with more than 5% institutional ownership. The results are consistent with the stated hypothesis of a positive association between dividend policy and institutional ownership. It appears that the presence of institutional ownership does have a positive impact in increasing dividend payout ratios. In addition, the significant and negative coefficient on the variable (EtiE(t1)i)*MDum supports the hypothesis of a negative association between managerial ownership and the dividend payout ratio. The results of the partial adjustment model (PAM) are presented in Table 3. The coefficient (c) for Dt1 is negative and statistically significant whereas the coefficients (cr and crI) for Et and Et*Inst are positive and statistically significant. The results, again, suggest that firms with 5% or more institutional ownership have a significantly higher dividend payout ratio. The evidence is consistent with the stated hypothesis. However, it appears that the association between managerial ownership and dividend payout ratios, while negative as expected, is not significantly so. The results of the Waud Model (WM), shown in Table 4, are also consistent with the stated hypothesis. The parameters c and d, jointly determined by the statistically significant coefficients of Dt1 and Dt2, are both positive. The coefficients (cdr and cdrI) for Et and Et*Inst are both positive and statistically significant, while the coefficient (cdrM) for Et*MDum is negative and statistically significant. The implication is that the presence of institutional ownership, on average, significantly increases a firm’s dividend payout ratio while the presence of managerial institutional ownership, on average, significantly reduces a firm’s dividend payout ratio.
Table 2 Generalised least squares (GLS) regression estimates using full adjustment model (FAM) based on 211 UK firms from 1988 to 1992 Dti Dðt1Þi ¼ a þ rðEti Eðt1Þi Þ þ rI ðEti Eðt1Þi Þ Inst þ rM ðEti Eðt1Þi Þ MDum þ lti Variable
Coefficient
t-statistic
EtEt1 (EtEt1)*Inst (EtEt1)*MDum
0.0626 0.0040 0.0477
15.498* 7.367* 11.805*
Fixed effects significant? Adjusted R square
Yes* 0.993
The variables are defined as follows: DtDt1 is the change in Dividend. EtEt1 is the change in Earnings. Inst equals 1 if Institutional Ownership is greater than 5%; 0 otherwise. MDum equals 1 if Managerial Ownership is greater than 5%; 0 otherwise. t-Statistics are heteroskedasticity consistent statistics. * Indicates statistical significance at the 1% level.
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
117
Table 3 Generalised least squares (GLS) regression estimates using partial adjustment model (PAM) based on 211 UK firms from 1988 to 1992 Dti Dðt1Þi ¼ a þ crEti þ crI Eti Inst þ crM Eti MDum cDðt1Þi þ lti Variable
Coefficient
t-statistic
Et Et*Inst Et*MDum Dt1
0.0633 0.0103 0.0031 0.3779
22.045* 7.976* 0.964 27.917*
Fixed effects significant? Adjusted R square
Yes* 0.847
The variables are defined as follows: DtDt1 is the change in Dividend. Et denotes earnings for year t. Inst equals 1 if Institutional Ownership is greater than 5%; 0 otherwise. MDum equals 1 if Managerial Ownership is greater than 5%; 0 otherwise. Dt1 denotes dividend for t1. t-Statistics are heteroskedasticity consistent statistics. * Indicates statistical significance at the 1% level.
Table 5 provides the results of the modified earnings trend model (ETM). The coefficient (c) for Dt1 is negative and statistically significant. The coefficient (rc) for Et is positive and statistically significant. The coefficients, rj(1c) and rjM(1c), for Et1 and Et1*MDum are not statistically different from zero. The positive and statistically significant coefficient, rjI(1c), for Et1*Inst, on the other hand, suggests that jI is positive and statistically significant. It appears that the earnings generating process for firms without significant institutional ownership on average follows a random walk without trend. However, the current evidence indicates that the earnings generating process for
Table 4 Generalised least squares (GLS) regression estimates using Waud Model (WM) based on 211 UK firms from 1988 to 1992 Dti Dðt1Þi ¼ ad þ cdrEti þ cdrI Eti Inst þ cdrM Eti MDum þ ð1 d cÞDðt1Þi ð1 dÞð1 cÞDðt2Þi lti Variable
Coefficient
t-statistic
Et Et*Inst Et*MDum Dt1 Dt2
0.0788 0.0115 0.0202 0.5940 0.13959
22.172* 9.009* 5.445* 38.843* 8.787*
Fixed effects significant? Adjusted R square
Yes* 0.960
The variables are defined as follows: DtDt1 is the change in Dividend. Et denotes Earnings at t. Inst equals 1 if Institutional Ownership is greater than 5%; 0 otherwise. MDum equals 1 if Managerial Ownership is greater than 5%; 0 otherwise. Dt1 is the Dividend at t1. Dt2 is the Dividend at t2. t-Statistics are heteroskedasticity consistent statistics. * Indicates statistical significance at the 1% level.
118
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
Table 5 Generalised least squares (GLS) regression estimates using earnings trend model (ETM) based on 211 UK firms from 1988 to 1992 Dti Dðt1Þi ¼ a þ rcEti þ rjð1 cÞEðt1Þi þ rjI ð1 cÞEðt1Þi Inst þ rjM ð1 cÞEðt1Þi MDum cDðt1Þi þ lit Variable
Coefficient
t-statistic
Et Et1 Et1*Inst Et1*Mdum Dt1
0.0659 0.0063 0.00912 0.0273 0.3599
22.506* 1.599 6.188* 0.796 22.094*
Fixed effects significant? Adjusted R square
Yes* 0.843
The variables are defined as follows: DtDt1 is the change in Dividend. Et denotes Earnings at t. Et1 denotes Earnings at t1. Inst equals 1 if Institutional Ownership is greater than 5%; 0 otherwise. MDum equals 1 if Managerial Ownership is greater than 5%; 0 otherwise. Dt1 is the Dividend at t1. t-Statistics are heteroskedasticity consistent statistics. * Indicates statistical significance at the 1% level.
firms with significant institutional ownership contains a positive and statistically significant trend. The results from the ETM imply that the impact of institutional ownership may go beyond increasing the dividend payout ratio; it may produce a positive and statistically positive earnings trend component. In summary, the results from the four dividend models are robust. They consistently produce strong support for the hypothesis that a positive association exists between dividend payout policy and institutional ownership. Furthermore, the results for the ETM suggest that dividend payouts have a positive and significant earnings trend component given higher institutional ownership. In addition, there is also some evidence in support of the hypothesis that a negative association exists between dividend payout policy and managerial ownership.
6. Conclusions This paper examines the relatively neglected association between dividend policy and institutional ownership. In fact, it presents the first results for the UK, where the institutional framework and ownership structures are different from those of the US. Furthermore, it is the first example of using well-established dividend payout models to examine the impact of ownership structures. Using a UK panel data set, the link between institutional ownership and dividend payout ratios is analysed within the context of the dividend models of Lintner (1956), Waud (1966) and Fama and Babiak (1968). The results from the four dividend models consistently produce strong support for the hypothesis that a positive association exists between dividend payout policy and institutional ownership. Furthermore, the results
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
119
from the ETM suggest that the influence of institutional ownership may go beyond increasing the dividend payout ratio; it may, in fact, produce a positive and significant earnings trend component. In addition, there is also some evidence in support of the hypothesis that a negative association exists between dividend payout policy and managerial ownership. Further research might usefully extend the present use of dividend payout models to examine the impact of institutional ownership (and other ownership structures) to the US and other non-UK markets. Such work will help to further establish the robustness of the current results of a strong positive association between institutional ownership and dividend payout ratios.
Appendix A The US operates a classical company tax system whereby companies are taxed separately from their shareholders. Companies pay a flat rate of corporation tax on their profits and shareholders pay income tax on the dividend income they receive at their marginal rates of tax. Essentially, dividends are taxed twice, first in the form of corporation tax on company profits and second, in the form of income tax on dividend income. For example, if a company makes a pre-tax profit of $100 and the corporation tax rate is c, the after-tax profit of the firm is $100 (1c). Assuming that the firm pays out all its after-tax profits as dividends to its sole shareholder, who faces a marginal personal income tax rate of p, the shareholder will receive $100(1c)(1p). In contrast, the UK has, since 1973, operated a partial imputation system whereby corporation tax is charged on company profits but part of the corporation tax paid is taken into account when assessing shareholders’ liability to income tax. The company pays a shareholder a cash dividend net of the basic rate of income tax (which is equal to the rate of imputation7). At the time a dividend is paid, the company pays Advance Corporation Tax (ACT) to the Inland Revenue. The ACT paid is equal to the rate of imputation times the gross dividend and represents an advance payment against its total corporation tax for the year. The shareholder receiving the net cash dividend also receives a tax credit equivalent to the basic rate of income tax on that dividend, which is used to offset his/her income tax liability. Therefore, under the partial imputation system, if a company makes a pre-tax profit of £100, the after-tax profit is equal to £100 (1c), where c equals the rate of corporation tax. Assuming that the company pays out all its after-tax profits as dividends to its sole shareholder, the shareholder receives a cash dividend of £100(1c) and a tax credit equal to £100[(1c)i]/[1i], giving a gross dividend (net cash dividend plus tax credit) of £100[(1c)]/[(1i)] (where i is the rate of imputation). The gross dividend is subject to
7 Prior to March 1993, the rate of imputation was always equal to the basic rate of income tax. However, from March 1993, the rate of imputation has been 20% (the same as the lower rate of income tax) compared to a basic rate of income tax of 25%. However, both lower rate and basic rate income taxpayers are deemed to be subject to a tax on dividend income of 20% (the rate of imputation).
120
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
Table 6 Dividend received after tax on a distribution of £100 pre-tax profit Corporate tax system
Classical system Preference Partial imputation system Preference
Formula
£100(1c)(1p) [£100(1c)(1p)]/ (1i)
Tax status of shareholder Exempt p=0
Basic or lower rate p = 0.25
Higher rate p = 0.4
£67.00 N £89.33
£50.25 R £67.00
£40.20 R £53.60
D
N
R
Assuming c=0.33 (the rate of corporation tax) and i=0.25 (the rate of imputation). Preferences: D=dividends; R=retentions; N=neutral.
income tax at the shareholder’s marginal income tax rate of p. Hence, after income tax, the shareholder receives £100[(1c)(1p)]/[(1i)]. For example, consider the following: a company earns a pre-corporation tax profit of £100,8 the corporation tax rate is 33% and the company pays out all of its profits as a cash dividend of £67. Assuming the imputation rate is 25%, the shareholder will receive a net cash dividend of £67 and a tax credit of £22.33, and the company will pay ACT of £22.33. Hence, the shareholder is deemed to have received a gross dividend of £89.33 on which £22.33 income tax has already been paid by the company in the form of ACT. The remaining tax liability or tax refund due to the shareholder depends upon the marginal income tax rate faced by the shareholder. For a basic rate taxpayer (facing a marginal income tax rate of 25%), the tax credit received is deemed to fully discharge his/her liability to income tax and no further income tax is due. For a higher rate taxpayer (subject to a marginal tax rate of 40%), an additional tax income charge arises equal to 15% of the gross dividend received (that is, an additional tax charge of £13.40). Finally, tax-exempt shareholders (such as pension funds and the pension business of insurance companies) receive a cash refund of the tax credit from the Inland Revenue.9 Table 6 illustrates the difference between the classical and partial imputation systems by calculating the after tax dividend which would be received by the different groups of taxpayers on a distribution of £100 pre-tax profits. In order to appreciate the incentives inherent in the two systems towards the payment of dividends or the retention of profits, the value of after-tax dividends is compared with the taxation of retained profits (ignoring
8
The following example describes the tax rates operating over the period of the empirical study, 1988 to 1992. In 1992, an additional income tax band was introduced, the lower rate band of 20%, and the rate of imputation was reduced to 20%. However, these changes do not affect the conclusions drawn in this section. 9 The July 1997 Budget has abolished tax credits to pension schemes and UK companies. These changes in the operation of the imputation tax system will provide opportunities for future researchers to examine whether the present tax bias in favour of dividends for institutional shareholders does have a significant effect on UK dividend policy. The potential consequences of such a change are complex: it may remove the tax bias in favour of dividends for institutional shareholders, but by reducing the after-tax value of the dividend, may lead to such institutions demanding higher levels of dividends to cover their cash short-fall. Clearly, the changes in the dividend tax system will allow researchers to better isolate the agency and signalling effects on dividend policy and to provide a richer test of the determination of dividend policy.
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
121
the effects of capital gains tax). Assuming a corporation tax rate of 33%, pre-tax profits of £100 are equivalent to £67 if retained in the company. As Table 6 illustrates, it is only taxexempt shareholders in the UK who will prefer dividends to retentions. Although the preferences for dividends or retentions are potentially complicated by the imposition of capital gains tax (CGT), the effective rate of CGT, in both the US and UK, is much lower than the nominal rate, due to the provision of annual exemptions from CGT, the ability of taxpayers to tax plan with regard to the realisation of capital gains and the fact that CGT can be deferred indefinitely.10 Therefore, following the lead of prior authors, the current analysis assumes that the effective rate of CGT is zero (see Bond et al., 1995).
References Agrawal, A., Jayaraman, N., 1994. The dividend policies of all-equity firms: a direct test of the free cash flow theory. Managerial and Decision Economics 15, 139 – 148. Baltagi, B., 1995. Econometric Analysis of Panel Data Wiley, New York. Bhattacharya, S., 1979. Imperfect information, dividend policy and the ‘‘bird in the hand’’ fallacy. Bell Journal of Economics 10 (1), 259 – 270. Bhattacharya, S., 1980. Nondissipative signalling structure and dividend policy. Quarterly Journal of Economics 95, 1 – 24. Black, F., 1976. The dividend puzzle. Journal of Portfolio Management 2, 72 – 77. Bond, S., Meghir, C., 1994. Financial constraints and company investment. Fiscal Studies 15 (2), 1 – 18. Bond, S., Chennells, L., Devereux, M., 1995. Company dividends and taxes in the UK. Fiscal Studies 16 (3), 1 – 18. Easterbrook, F.H., 1984. Two agency-cost explanations of dividends. American Economic Review 74 (4), 650 – 659. Eckbo, B.E., Verma, S., 1994. Managerial shareownership, voting power, and cash dividend policy. Journal of Corporate Finance 1, 33 – 62. Fama, E.F., Babiak, W., 1968. Dividend analysis: an empirical analysis. Journal of the American Statistical Association 63, 1132 – 1161. Haskins, C., 1995. A company should be run primarily in the interests of its shareholders. Enterprise and Governance: The Proceedings of a Conference Held at the Institute of Directors on 24 October, IoD, London. Hutton, W., 1995. The State We’re In Jonathan Cape, London. Jensen, M.C., 1986. Agency costs of free cash flow, corporate finance and takeovers. AEA Papers and Proceedings 76 (2), 323 – 329. Jensen, G.R., Solbreg, D.P., Zorn, T.S., 1992. Simultaneous determination of insider ownership, debt, and dividend policies. Journal of Financial and Quantitative Analysis 27 (2), 247 – 263. Kester, W.C., 1992. Industrial groups as systems of contractual governance. Oxford Review of Economic Policy 8, 24 – 44. Leland, H., Pyle, D., 1977. Information asymmetries, financial structure and financial intermediation. Journal of Finance, 371 – 387, May. Lintner, J., 1956. Distribution of incomes of corporations among dividends, retained earnings and taxes. American Economic Review 46, 97 – 113. Mayer, C., 1994. Stock markets, financial institutions, and corporate performance. In: Dimsdale, N., Prevezer, M. (Eds.), Capital Markets and Corporate Governance. Clarendon Press, Oxford, pp. 179 – 194. Mayer, C., Alexander, I., 1990. Banks and markets: corporate financing in Germany and the UK. Journal of Japanese and International Economics 3, 450 – 475.
10 In the UK, CGT raised only £0.9 billion from individuals in the tax year 1992 – 1993 (HMSO, Economic Trends, October 1994).
122
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
Miller, M.H., Rock, K., 1985. Dividend policy and asymmetric information. Journal of Finance 40, 1031 – 1051. Moh’d, M.A., Perry, L.G., Rimbey, J.N., 1995. An investigation of the dynamic association between agency theory and dividend policy. Financial Review 30 (2), 367 – 385. Roe, M.J., 1990. Political and legal restraints on ownership and control of public companies. Journal of Financial Economics 27, 7 – 42. Rozeff, M.S., 1982. Growth, beta and agency costs as determinants of dividend payout ratios. Journal of Financial Research 5, 249 – 259. Trade and Industry Committee, 1994. Competitiveness of UK Manufacturing Industry. Second Report, HC 41.I, HMSO, London. Waud, R., 1966. Small sample bias due to misspecification in the ‘partial adjustment’ and ‘adapted expectations’ models. Journal of the American Statistical Association, pp. 134 – 145, December. Williamson, O.E., 1964. The Economics of Discretionary Behaviour: Managerial Objectives in a Theory of the Firm. Prentice-Hall, New York. Zeckhauser, R.J., Pound, J., 1990. Are large shareholders effective monitors? an investigation of share ownership and corporate performance. In: Hubbard, R.G. (Ed.), Asymmetric Information, Corporate Finance and Investment. University of Chicago Press, Chicago, pp. 149 – 180.
The link between dividend policy and institutional ownership Helen Short *,1, Hao Zhang 1, Kevin Keasey
1
The International Institute of Banking and Financial Services (IIBFS), Leeds University Business School, University of Leeds, Leeds, LS2 9JT, UK Accepted 22 June 2001
Abstract This paper examines the relatively neglected link between dividend policy and institutional ownership. It is also the first example of using well-established dividend payout models to examine the potential association between ownership structures and dividend policy. Moreover, the paper presents the first results for the UK, where the institutional framework and ownership structures are different from those of the US. Using a UK panel data set, the role of institutional ownership in association to dividend payout ratios is analysed within the context of the dividend models of Lintner [American Economic Review, 46 (1956) 97], Waud [Journal of the American Statistical Association, 1996] and Fama and Babiak [Journal of the American Statistical Association, 63 (1968) 1132]. The results consistently produce strong support for the hypothesis that a positive association exists between dividend payout policy and institutional ownership. Furthermore, the results for an earnings trend model suggest a positive earnings trend component to the association between institutional ownership and the dividend payout ratio. In addition, there is some evidence in support of the hypothesis that a negative association exists between dividend payout policy and managerial ownership. D 2002 Elsevier Science B.V. All rights reserved. JEL classification: G32 Keywords: Dividend policy; Institutional ownership
*
Corresponding author. Tel.: +44-113-233-4471; fax: +44-113-233-4459. E-mail address: [email protected] (H. Short). 1 Helen Short, Hao Zhang and Kevin Keasey are, respectively, Senior Lecturer in Finance and Accounting, Senior Lecturer in Finance and Accounting, Halifax plc Professor of Financial Services, at IIBFS, University of Leeds. 0929-1199/02/$ - see front matter D 2002 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 9 - 11 9 9 ( 0 1 ) 0 0 0 3 0 - X
106
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
1. Introduction The link between a firm’s financial policy and its ownership structure is recognised in the established literature (for example, Williamson, 1964; Leland and Pyle, 1977; Jensen, 1986). While the empirical evidence of the relation between dividend policy and management ownership has also been well documented within this literature (for example, Rozeff, 1982; Jensen et al., 1992; Eckbo and Verma, 1994; Moh’d et al., 1995), the potential relation between dividend policy and institutional ownership has been somewhat neglected and this is especially the case for non-US firms where the institutional framework and ownership structures are different from those of the US. In terms of the institutional ownership of firms, there are significant differences between the US and UK. These differences result from the differences in legal restrictions and tax incentives. In fact, institutional shareholdings in the US are only about two-thirds of those in the UK (Kester, 1992). The significantly lower concentration of US institutional shareholdings reflects both the legal restrictions preventing US institutions from building significant stakes in individual firms (see Roe, 1990, for a discussion) and the tax system being largely neutral in relation to dividend policy. In contrast, the UK tax system has (in the period under consideration) provided clear incentives for tax-exempt institutions to demand dividends. Using a UK panel data set, this paper examines the possible link between dividend policy and institutional ownership. In particular, the role of institutional ownership in relation to dividend payout ratios is analysed within the context of the dividend models of Lintner (1956), Waud (1966) and Fama and Babiak (1968). The evidence from four wellknown types of dividend models consistently shows positive and statistically significant associations between institutional ownership and dividend payout ratios and thus suggests a link between institutional ownership and dividend policy. The paper is organised as follows. Section 2 discusses the relation between dividend policy and institutional ownership in the context of theories that emphasise the importance of taxation, agency and signalling for dividend decisions. A discussion of the statistical models of dividend payout behaviour is provided in Section 3. Section 4 discusses the sample, variables and empirical method to be used in examining the association between dividends and institutional ownership. Section 5 provides the results and a discussion of the empirical analysis. Finally, Section 6 presents the conclusions.
2. Institutional shareholders and dividend payout policy The relatively high level of dividends paid by UK quoted firms has been the subject of concern for a number of years. In particular, the dividend payout rates of UK firms are significantly greater than those of other countries, such as Germany and Japan, and dividends in the UK are considered to be downwardly inflexible (Mayer and Alexander, 1990; Mayer, 1994). The relatively high levels of dividends paid by UK firms have been argued to reduce the amount of funds available for long-term investment, to the detriment of long-term economic performance (Bond and Meghir, 1994). In particular, attention has focused on the alleged role of institutional shareholders in forcing firms to maintain high
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
107
dividends, particularly in the face of falling profits in the UK recession of the late 1980s/ early 1990s (for a detailed discussion, see the Report of the Trade and Industry Committee on Competitiveness, 1994). While the high level of dividends paid by firms could be construed as reflecting the short-term attitudes held by institutions (see, for example, Hutton, 1995; Haskins, 1995), it could equally be construed as the result of institutional investors’ effort to reduce ‘‘free’’ cash flows available to the management (Jensen, 1986). The ownership of equity in the UK is dominated by institutional shareholders and this may affect the dividend payout ratio in a number of ways. First, a bias in the UK tax system which favours dividends over capital gains, particularly for tax-exempt institutions such as pension funds, provides incentives for institutions to demand relatively high levels of dividends. Second, dividends may be used as a mechanism to reduce agency problems existing between shareholders and managers. Third, given the presence of information asymmetries existing between shareholders and managers as a result of the capital market’s emphasis on arm’s length funding and liquidity, dividends may be used to signal future prospects of the firm. The following subsections detail the theoretical and empirical implications of taxation, agency costs and signalling considerations on the relation between dividends and institutional shareholders. 2.1. Taxation The US operates a classical company tax system whereby companies are taxed separately from their shareholders. Firms pay a flat rate of corporation tax on their profits and shareholders pay income tax on the dividend income they receive at their marginal rates of income tax. This means that dividends are essentially taxed twice, first in the form of corporation tax on firm profits, and second, in the form of income tax on dividend income. An important outcome of the classical system is that both basic and high rate income taxpayers will prefer profits to be retained in the firm rather than paid out in dividends, with tax-exempt shareholders being neutral with respect to dividends and retained profits. This leads to the so-called classical dividends puzzle prevalent in the US literature (Black, 1976); namely, why do firms pay dividends at all? In contrast, the UK operates a partial imputation company tax system whereby corporation tax is charged on firm profits but part of the corporation tax paid is taken into account when assessing shareholders’ liability to income tax. The workings of the UK system of corporate taxation are detailed in Appendix A. An important outcome of the UK system is that tax-exempt shareholders prefer dividends to retentions, basic rate taxpayers are neutral with respect to dividends and retentions, whilst higher rate taxpayers prefer retentions to dividends. The bias in the UK system for tax-exempt shareholders to prefer dividends to retentions may result in relatively higher dividend payments. Bond et al. (1995) argue, To the extent that tax-exempt shareholders such as pension funds are now the most influential investors in many UK companies, their tax preference for dividend income is likely to result in significantly higher dividend payout ratios than would be chosen by companies in the absence of this tax bias. (p. 17)
108
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
A related issue is the need of institutional shareholders for funds on an ongoing basis. Clearly, institutions invest in equities in order to provide returns to fund their activities, such as funding pensions, paying out in insurance policies, etc. Regardless of the tax bias in favour of dividends, institutions cannot simply rely on capital gains to fund their liabilities and, hence, require dividend payments. Furthermore, UK pension funds are, on average, mature in the sense that pensions are paid out of investment income rather than from new cash flow from contributions. Given the investment profile of pension funds with their reliance on equities, pension funds require a certain level of dividends to match their liabilities. Moreover, the actuarial value of pension funds is in part based on dividend income. Institutions’ requirements for certain levels of dividends to meet their own liabilities may force companies to pay out dividends at a level higher than they would otherwise prefer, particularly in times of recession and low corporate profitability. Hence, as a result of the tax bias in favour of dividends for tax-exempt institutions and the need of institutions to maintain income flows in the form of dividends received, it is hypothesised that institutional shareholdings will have a positive effect on the dividend payout ratio. 2.2. Agency theory The payment of dividends may act to help reduce agency costs which arise between managers and external shareholders. Rozeff (1982) and Easterbrook (1984) argue that the payment of dividends forces firms to go to the external capital markets for additional funding and, hence, undergo monitoring by the capital market. In this respect, Zeckhauser and Pound (1990) suggest that institutional shareholders may act as a substituting monitoring device, hence reducing the need for external monitoring by the capital markets. However, the arm’s length view of investment held by many institutional investors, coupled with the incentives to free ride with respect to monitoring activities, suggests that institutional shareholders are unlikely to provide direct monitoring themselves. Indeed, Zeckhauser and Pound suggest that institutions, rather than providing monitoring themselves, force firms to increase their dividends in order that they are subsequently forced to go to the external capital market for future funds. Jensen’s (1986) free cash flow theory suggests that managers are reluctant to pay out dividends, preferring instead to retain resources under their control. Eckbo and Verma (1994) argue that institutional shareholders will prefer free cash flow to be distributed in the form of dividends in order to reduce the agency costs of free cash flow. From this perspective, it may be argued that institutional shareholders may counter a tendency for managers to prefer the excessive retention of cash flow and, by virtue of their voting power, force managers to pay out dividends. The agency perspective, therefore, hypothesises a positive relation between institutional shareholders and the dividend payout ratio, as institutions demand dividends and force firms to the capital market for future funds in order to reduce the agency costs of free cash flow. 2.3. Signalling The signalling perspective suggests that dividends are used to signal managements’ private information regarding the future earnings of the firm (Bhattacharya, 1979, 1980;
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
109
Miller and Rock, 1985). Zeckhauser and Pound (1990) suggest that dividends and institutional shareholders may be viewed as alternative signalling devices. The presence of a large shareholder may mitigate the use of dividends as a signal of good performance, as the large shareholders themselves can act as a (more) credible signal. Under this hypothesis, Zeckhauser and Pound suggest that a negative relation is predicted to exist between dividends and large shareholders. However, it is not clear exactly how institutional shareholders would act as a signal of future prospects of the firm to the market. There are two possible scenarios. First, the presence of institutional shareholders may signal to the market that agency costs are reduced due to the monitoring activities of the institutional shareholders. However, as already discussed, due to free rider problems it is unlikely that institutional shareholders are willing to provide direct monitoring. Second, given the superior information that institutional shareholders are likely to have concerning the future prospects of the firm (because of their easier access to and control of the management, etc.), it is possible that the market may interpret the presence of an institutional shareholder as signalling good news regarding the firm’s future prospects. However, whilst this explanation holds some attractions, the available evidence is not strongly in favour of this scenario. Given the emphasis placed on dividends as a signal of future prospects by institutions themselves (see Trade and Industry Committee on Competitiveness, 1994), it would appear that institutions do not regard themselves as substitute signalling mechanisms. In addition, the institutions need to be wary of being privy to inside information because of the insider dealing laws. Furthermore, the growth in indexed funds means that the presence of an institutional shareholder does not necessarily imply that the particular institution believes that the firm has better than average prospects. Overall, the notion that dividends and institutional shareholders act as substituting signalling devices, while possible, is less than convincing. 2.4. Summary and empirical evidence In summary, the relation between dividends and institutional ownership would appear to depend on the relative weights of tax, agency and signalling considerations. From the tax perspective, there are clear incentives for (tax-exempt) institutions to demand high levels of dividends as a result of the bias in the UK tax system in favour of dividends for tax-exempt shareholders. Furthermore, institutions require certain levels of dividends to meet their own liabilities. From the agency perspective, institutions may demand high levels of dividends in order to force firms to go to the capital for external funding and, hence, be subject to monitoring by the external market. From the free cash flow perspective, institutions may counter management’s tendency to retain excess free cash flow. Hence, both the tax and agency related perspectives suggest that a positive association exists between dividends and institutional shareholdings. In contrast, the signalling arguments suggest that dividends and institutions may act as substitute signalling devices. However, as discussed, this is possible but less than convincing in the context of the UK. Therefore, on the basis of tax and agency considerations, we hypothesise the following: A positive association is expected to exist between dividend changes and institutional shareholdings.
110
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
The empirical evidence concerning the link between institutional shareholders and dividend payments is extremely limited. Eckbo and Verma (1994), from a sample of Canadian quoted companies, report that the cash dividend yield increases significantly with the voting power of corporate/institutional shareholders and decreases significantly with the voting power of management. Moh’d et al. (1995) report a significant and positive association between dividends and institutional shareholdings. Alternatively, Zeckhauser and Pound (1990), investigating US companies, find no evidence to support the view that institutional shareholders have an impact on dividend policy.2
3. Models Four types of dividend models are used to test the hypothesised positive link between institutional ownership and dividend policy: the Full Adjustment Model, the Partial Adjustment Model (Lintner, 1956), the Waud Model (1966), and the Earnings Trend Model (Fama and Babiak, 1968). These models are modified by interactive dummy variables to account for the possible effects of institutional ownership. In addition, we also allow for the potential link between managerial ownership and dividend policy (Rozeff, 1982; Easterbrook, 1984), as there is a significant body of empirical literature to suggest a negative link between the two (Rozeff, 1982; Jensen et al., 1992; Eckbo and Verma, 1994; Agrawal and Jayaraman, 1994; Moh’d et al., 1995). 3.1. The full adjustment model (FAM) If changes in income are considered permanent and a firm has a desired payout ratio r, then the association between changes in earnings (E) and changes in dividends (D), for firm i at time t, will be given by: Dti Dðt1Þi ¼ a þ rðEti Eðt1Þi Þ þ lti If we assume that firms with significant institutional ownership and/or managerial ownership may have a different r, then the model becomes: Dti Dðt1Þi ¼ a þ rðEti Eðt1Þi Þ þ rI ðEti Eðt1Þi Þ Inst þ rM ðEti Eðt1Þi Þ MDum þ lti
ðModel 1; FAMÞ
Inst is a dummy variable denoting the presence of significant institutional ownership and MDum is a dummy variable denoting the presence of significant managerial ownership. The coefficients rI and rM denote the respective impacts of institutional ownership and managerial ownership in association to the dividend payout ratio of the firm. 2 However, the Zeckhauser and Pound study utilises an industry aggregate sample (as opposed to individual firm level data) and, moreover, use a threshold of 15% to denote the presence of an institutional shareholder, both of which may be inappropriate.
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
111
As discussed in the previous section, institutional ownership is hypothesised to be positively linked to the dividend payout ratio, whereas managerial ownership is expected to be negatively linked to dividend payout ratio. Hence, the coefficient rI is expected to be positive and statistically significant and the coefficient rM is expected to be negative and statistically significant. 3.2. The partial adjustment model (PAM) The partial adjustment model assumes that for any year, t, the target level of dividend, D for firm i at time t is related to profits, Eti, by a desired payout ratio, r: Dti ¼ rEti If firms with significant institutional ownership and/or managerial ownership have a different r, then the model becomes: Dti ¼ rEti þ rI Eti Inst þ rM Eti MDum In any given year, the firm adjusts only partially to the target dividend level. Thus: Dti Dðt1Þi ¼ a þ cðDti Dðt1Þi Þ where a=a constant representing the resistance of management to reduce dividends, and c equals the ‘speed of adjustment coefficient’ which represents the extent to which the management wishes to ‘play-safe’ by not adjusting to the new target immediately. Substitution yields the following reduced form: Dti Dðt1Þi ¼ a þ crEti þ crI Eti Inst þ crM Eti MDum cDðt1Þi þ lti ðModel 2; PAMÞ As before, the coefficient (crI) on the interactive variable EtiInst variable is expected to be positive while that on the EtiMDum variable is expected to be negative. 3.3. The Waud model (WM) The Waud model incorporates elements of both partial and full adjustment. It assumes that the target dividends, D , for firm i at time t, are proportional to the long-run expected earnings, E : Dti ¼ rEti and that the actual dividend change follows a partial adjustment mechanism: Dti Dðt1Þi ¼ a þ cðDti Dðt1Þi Þ þ lti The formation of expectations follows an adaptive expectation model: Eti Eðt1Þi ¼ dðEti Eðt1Þi Þ
112
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
Assuming a possible difference in payout ratio for firms with significant institutional and/ or managerial ownership, the reduced form becomes:3 Dti Dðt1Þi ¼ ad þ cdrEti þ cdrI Eti Inst þ cdrM Eti MDum þð1 d cÞDðt1Þi ð1 dÞð1 cÞDðt2Þi lti ðModel 3; WMÞ Again the coefficient of EtiInst variable is expected to be positive whilst that of EtiMDum variables is expected to be negative. 3.4. Earnings trend model (ETM) The earnings trend model is a modified partial adjustment model. It assumes a specific profit generating process, for firm i at time t, of the form: Eti ¼ ð1 þ jÞEðt1Þi where j is an earnings trend factor. Assuming a possible difference in the earnings trend factor for firms with significant institutional and/or managerial ownership, the profit generating process becomes:4 Eti ¼ Eðt1Þi þ jEðt1Þi þ jI Eðt1Þi Inst þ jM Eðt1Þi MDum Target dividends are given by: Dti ¼ rEti It is assumed that there is full adjustment of dividends to the expected change, that is, jEðt1Þ þ jI Eðt1Þ Inst þ jM Eðt1Þ MDum and partial adjustment to the reminder. The resulting reduced form is: Dti Dðt1Þi ¼ a þ c½rðEti jEðt1Þi jI Eðt1Þi Inst jM Eðt1Þi MDumÞ cDðt1Þi þ rjEðt1Þi þrjI Eðt1Þi Inst þ rjM Eðt1Þi MDum þ lti Arranging, the reduced form becomes: Dti Dðt1Þi ¼ a þ rcEti þ rjð1 cÞEðt1Þi þ rjI ð1 cÞEðt1Þi Inst þrjM ð1 cÞEðt1Þi MDum cDðt1Þi þ lit ðModel 4; ETMÞ 3
See Waud (1966) for detailed derivation. It turned out that, in ETM, introducing dummy variables to account for the possible difference in payout ratio (as opposed to earnings trend factor) for firms with significant institutional and/or managerial ownership produces no reduced form. 4
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
113
Once again the coefficient of E(t1)i*Inst is expected to be positive, whereas the cofficient of E(t1)i*MDum is expected to be negative.
4. Sample, variables and empirical method 4.1. Sample The empirical analysis of the association between dividends and equity ownership is conducted on a sample of 211 firms listed on the London Stock Exchange Official List for the period 1988 to 1992. For a firm to be included in the sample, several criteria had to be met. First, the firm had to be listed on the Official List for the whole of the period under consideration. In addition, the firm would also be required to be listed in the year 1987. This condition was imposed to ensure that dividend policy was not distorted by the effects of a recent Official Listing. Second, firms in the financial and oil and gas sectors were excluded due to the different income measuring rules governing such firms, as compared to those in the manufacturing and service sectors. Third, privatised firms (such as water, electricity, gas, telecommunications) were excluded as their operating conditions (in terms of regulation and monopoly markets) are usually atypical. Fourth, firms operating in the broadcasting sectors were excluded due to the regulatory nature of the market and the changes in regulation that occurred during the period (for example, tendering for operating licences and changes in ownership rules which are largely government controlled). Finally, firms with ownership structures that did not conform to the typical one vote for each ordinary share were also excluded. As this paper specifically examines the potential link between ownership structure and dividend policy, it is important that the ownership structure and the associated voting structure of the equity shares of a particular firm can be determined. As a result, firms whose share structure included non-ordinary voting shares (such as voting A shares, management shares, founders’ shares) were excluded.5 Other than being subject to the criteria outlined above, a sample of 225 firms was selected entirely at random from all firms quoted on the London Stock Exchange Official List. The sample was further reduced to 211 firms, as a result of missing data. 4.2. Variables In order to examine empirically the dividend models discussed in Section 3, the key variables of interest are measures of dividends (D), earnings (E), institutional shareholdings (Inst) and managerial shareholdings (MDum). This section outlines the variables to be used to examine the association between institutional shareholdings and the dividend payout ratio. The dividend and profit variables were derived from data collected from the 5 Non-ordinary voting shares often carry more votes per share than ordinary shares and would considerably complicate the determination of the ownership structure of individual firms. As the trend is towards share structures including only ordinary voting shares and, in addition, in many cases, firms with such atypical share structures often change the nature of the shares in the period under consideration, exclusion of such firms should not materially affect the results.
114
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
financial database Datastream. Data on ownership was hand collected from the sample firms’ annual reports. 4.2.1. Dividends Dividends (D) are calculated as the total amount of ordinary dividends relating to the accounting year. 4.2.2. Earnings The earnings variable (E) is calculated as net profit derived from normal trading activities after depreciation and other operating provisions. A similar measure was utilised by Jensen et al. (1992). 4.2.3. Institutional ownership Institutional ownership is defined as the percentage of equity held by institutions owning 5% or more of equity at the beginning of the accounting year. For the period 1988 to 1990, firms were obliged to provide information in the annual report of external shareholdings amounting to 5% or more of equity shares in issue. The reporting requirements changed in June 1990, such that from 1990 to 1992, firms were obliged to provide information in the annual report of external shareholdings amounting to 3% or more of equity shares in issue. In order to avoid complications caused by this change in reporting requirements, a cut-off point of 5% institutional ownership was used for the whole time period. The analysis uses dummy variables based on the percentage of institutional ownership; Inst is coded 1 if an institution owning 5% or more of equity is present and 0 otherwise. 4.2.4. Management ownership Management ownership is defined as the total percentage of equity owned by directors and their immediate families at the beginning of the accounting year. This measure includes directors’ ownership via corporate vehicles, for example, where directors are majority shareholders in other firms that have direct ownership stakes in the sample firm under consideration.6 The analysis uses a dummy variable based on the percentage of management ownership. MDum is coded 1 if management ownership is greater or equal to 5% of equity and 0 otherwise. 4.3. Empirical method As noted in Section 4.1, the data consist of a sample of 211 firms covering the period from 1988 to 1992. The dividend models FAM, PAM and ETM focus on the changes in dividends and earnings between 2 years (for example, from t1 to t). For each period (for example, t1 to t), ownership measures that are already in place at the start of the period (for example, at t1) are used. Hence, for these three models, there are four periods of
6 Directors’ non-beneficial ownership of equity is included in the measure of management ownership, as directors have voting control over these shares.
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
115
observations (1988 – 1989, 1989– 1990, 1990– 1991 and 1991 – 1992). However, since dividend information going back two periods is required for the dividend model (WM), there are three periods of observations (i.e. 1988 – 1990, 1989 –1991 and 1990 –1992) in this instance. The fixed effect panel data estimation method is used to test the stated hypothesis regarding institutional ownership and the dividend payout ratio (see Baltagi, 1995). The resulting firm-specific intercepts account for firm-specific characteristics in the sample.
5. Empirical analysis and results Summary statistics relating to the dependent and explanatory variables are presented in Table 1. The mean dividend payment has increased in every year of the period under consideration. Mean earnings, while increasing over the period 1988– 1989, declined in each year thereafter; evidence of the economic recession that occurred in the UK over that period. The increase in dividends occurring while earnings declined is indicative of the reluctance of firms to decrease dividends in line with earnings reductions. The data pertaining to ownership demonstrates interesting changes in the nature of ownership of the period. The concentration of institutional ownership appears to have increased significantly, with 84% of the sample having an institution owning 5% or more of equity in 1992 compared with 57% of the sample having such an institutional owner in 1988. Whilst institutional ownership has increased, management ownership has remained relatively static over the period, with approximately 51% of the sample having directors owning 5% or more of the equity for the 1988 – 1991 period. However, there are
Table 1 Summary statistics Variable
Year 1988
Period 1989
1990
1991
1992
1988 – 1992
Dividend (D) (£’000’s) Mean 7054.12 Median 1556.00 SD 14,513.29
8637.20 1963.00 17,701.53
9610.50 2349.00 20,022.38
10,154.50 1999.00 22,542.83
11,005.22 2220.00 25,285.72
9491.18 2122.00 20,545.92
Earnings (E) (£’000’s) Mean 32,809.40 Median 7923.00 SD 69,863.96
37,516.24 9755.00 79,785.45
37,381.35 9461.00 81,671.03
34,012.71 7142.00 84,128.86
33,063.33 6484.00 86,097.08
36,002.72 8740.00 81,198.51
Institutional ownership (Inst=1) Number 121 % 57.3
129 61.1
153 72.5
161 76.3
177 83.9
n/a n/a
Management ownership (MDum=1) Number 109 110 % 51.7 52.1
108 51.2
109 51.7
98 46.4
n/a n/a
116
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
indications that this level of ownership is on the decline, as there is a drop in such ownership to 46% of the sample in 1992. Clearly, the difference between institutional ownership and management ownership (in terms of 5% or more equity ownership) widens significantly over the sample period. The results of the full adjustment model (FAM) are shown in Table 2. The coefficients for (EtEt1) and (EtEt1)*Inst are positive and statistically significant. The dividend payout ratio is, on average, significantly higher for firms with more than 5% institutional ownership. The results are consistent with the stated hypothesis of a positive association between dividend policy and institutional ownership. It appears that the presence of institutional ownership does have a positive impact in increasing dividend payout ratios. In addition, the significant and negative coefficient on the variable (EtiE(t1)i)*MDum supports the hypothesis of a negative association between managerial ownership and the dividend payout ratio. The results of the partial adjustment model (PAM) are presented in Table 3. The coefficient (c) for Dt1 is negative and statistically significant whereas the coefficients (cr and crI) for Et and Et*Inst are positive and statistically significant. The results, again, suggest that firms with 5% or more institutional ownership have a significantly higher dividend payout ratio. The evidence is consistent with the stated hypothesis. However, it appears that the association between managerial ownership and dividend payout ratios, while negative as expected, is not significantly so. The results of the Waud Model (WM), shown in Table 4, are also consistent with the stated hypothesis. The parameters c and d, jointly determined by the statistically significant coefficients of Dt1 and Dt2, are both positive. The coefficients (cdr and cdrI) for Et and Et*Inst are both positive and statistically significant, while the coefficient (cdrM) for Et*MDum is negative and statistically significant. The implication is that the presence of institutional ownership, on average, significantly increases a firm’s dividend payout ratio while the presence of managerial institutional ownership, on average, significantly reduces a firm’s dividend payout ratio.
Table 2 Generalised least squares (GLS) regression estimates using full adjustment model (FAM) based on 211 UK firms from 1988 to 1992 Dti Dðt1Þi ¼ a þ rðEti Eðt1Þi Þ þ rI ðEti Eðt1Þi Þ Inst þ rM ðEti Eðt1Þi Þ MDum þ lti Variable
Coefficient
t-statistic
EtEt1 (EtEt1)*Inst (EtEt1)*MDum
0.0626 0.0040 0.0477
15.498* 7.367* 11.805*
Fixed effects significant? Adjusted R square
Yes* 0.993
The variables are defined as follows: DtDt1 is the change in Dividend. EtEt1 is the change in Earnings. Inst equals 1 if Institutional Ownership is greater than 5%; 0 otherwise. MDum equals 1 if Managerial Ownership is greater than 5%; 0 otherwise. t-Statistics are heteroskedasticity consistent statistics. * Indicates statistical significance at the 1% level.
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
117
Table 3 Generalised least squares (GLS) regression estimates using partial adjustment model (PAM) based on 211 UK firms from 1988 to 1992 Dti Dðt1Þi ¼ a þ crEti þ crI Eti Inst þ crM Eti MDum cDðt1Þi þ lti Variable
Coefficient
t-statistic
Et Et*Inst Et*MDum Dt1
0.0633 0.0103 0.0031 0.3779
22.045* 7.976* 0.964 27.917*
Fixed effects significant? Adjusted R square
Yes* 0.847
The variables are defined as follows: DtDt1 is the change in Dividend. Et denotes earnings for year t. Inst equals 1 if Institutional Ownership is greater than 5%; 0 otherwise. MDum equals 1 if Managerial Ownership is greater than 5%; 0 otherwise. Dt1 denotes dividend for t1. t-Statistics are heteroskedasticity consistent statistics. * Indicates statistical significance at the 1% level.
Table 5 provides the results of the modified earnings trend model (ETM). The coefficient (c) for Dt1 is negative and statistically significant. The coefficient (rc) for Et is positive and statistically significant. The coefficients, rj(1c) and rjM(1c), for Et1 and Et1*MDum are not statistically different from zero. The positive and statistically significant coefficient, rjI(1c), for Et1*Inst, on the other hand, suggests that jI is positive and statistically significant. It appears that the earnings generating process for firms without significant institutional ownership on average follows a random walk without trend. However, the current evidence indicates that the earnings generating process for
Table 4 Generalised least squares (GLS) regression estimates using Waud Model (WM) based on 211 UK firms from 1988 to 1992 Dti Dðt1Þi ¼ ad þ cdrEti þ cdrI Eti Inst þ cdrM Eti MDum þ ð1 d cÞDðt1Þi ð1 dÞð1 cÞDðt2Þi lti Variable
Coefficient
t-statistic
Et Et*Inst Et*MDum Dt1 Dt2
0.0788 0.0115 0.0202 0.5940 0.13959
22.172* 9.009* 5.445* 38.843* 8.787*
Fixed effects significant? Adjusted R square
Yes* 0.960
The variables are defined as follows: DtDt1 is the change in Dividend. Et denotes Earnings at t. Inst equals 1 if Institutional Ownership is greater than 5%; 0 otherwise. MDum equals 1 if Managerial Ownership is greater than 5%; 0 otherwise. Dt1 is the Dividend at t1. Dt2 is the Dividend at t2. t-Statistics are heteroskedasticity consistent statistics. * Indicates statistical significance at the 1% level.
118
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
Table 5 Generalised least squares (GLS) regression estimates using earnings trend model (ETM) based on 211 UK firms from 1988 to 1992 Dti Dðt1Þi ¼ a þ rcEti þ rjð1 cÞEðt1Þi þ rjI ð1 cÞEðt1Þi Inst þ rjM ð1 cÞEðt1Þi MDum cDðt1Þi þ lit Variable
Coefficient
t-statistic
Et Et1 Et1*Inst Et1*Mdum Dt1
0.0659 0.0063 0.00912 0.0273 0.3599
22.506* 1.599 6.188* 0.796 22.094*
Fixed effects significant? Adjusted R square
Yes* 0.843
The variables are defined as follows: DtDt1 is the change in Dividend. Et denotes Earnings at t. Et1 denotes Earnings at t1. Inst equals 1 if Institutional Ownership is greater than 5%; 0 otherwise. MDum equals 1 if Managerial Ownership is greater than 5%; 0 otherwise. Dt1 is the Dividend at t1. t-Statistics are heteroskedasticity consistent statistics. * Indicates statistical significance at the 1% level.
firms with significant institutional ownership contains a positive and statistically significant trend. The results from the ETM imply that the impact of institutional ownership may go beyond increasing the dividend payout ratio; it may produce a positive and statistically positive earnings trend component. In summary, the results from the four dividend models are robust. They consistently produce strong support for the hypothesis that a positive association exists between dividend payout policy and institutional ownership. Furthermore, the results for the ETM suggest that dividend payouts have a positive and significant earnings trend component given higher institutional ownership. In addition, there is also some evidence in support of the hypothesis that a negative association exists between dividend payout policy and managerial ownership.
6. Conclusions This paper examines the relatively neglected association between dividend policy and institutional ownership. In fact, it presents the first results for the UK, where the institutional framework and ownership structures are different from those of the US. Furthermore, it is the first example of using well-established dividend payout models to examine the impact of ownership structures. Using a UK panel data set, the link between institutional ownership and dividend payout ratios is analysed within the context of the dividend models of Lintner (1956), Waud (1966) and Fama and Babiak (1968). The results from the four dividend models consistently produce strong support for the hypothesis that a positive association exists between dividend payout policy and institutional ownership. Furthermore, the results
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
119
from the ETM suggest that the influence of institutional ownership may go beyond increasing the dividend payout ratio; it may, in fact, produce a positive and significant earnings trend component. In addition, there is also some evidence in support of the hypothesis that a negative association exists between dividend payout policy and managerial ownership. Further research might usefully extend the present use of dividend payout models to examine the impact of institutional ownership (and other ownership structures) to the US and other non-UK markets. Such work will help to further establish the robustness of the current results of a strong positive association between institutional ownership and dividend payout ratios.
Appendix A The US operates a classical company tax system whereby companies are taxed separately from their shareholders. Companies pay a flat rate of corporation tax on their profits and shareholders pay income tax on the dividend income they receive at their marginal rates of tax. Essentially, dividends are taxed twice, first in the form of corporation tax on company profits and second, in the form of income tax on dividend income. For example, if a company makes a pre-tax profit of $100 and the corporation tax rate is c, the after-tax profit of the firm is $100 (1c). Assuming that the firm pays out all its after-tax profits as dividends to its sole shareholder, who faces a marginal personal income tax rate of p, the shareholder will receive $100(1c)(1p). In contrast, the UK has, since 1973, operated a partial imputation system whereby corporation tax is charged on company profits but part of the corporation tax paid is taken into account when assessing shareholders’ liability to income tax. The company pays a shareholder a cash dividend net of the basic rate of income tax (which is equal to the rate of imputation7). At the time a dividend is paid, the company pays Advance Corporation Tax (ACT) to the Inland Revenue. The ACT paid is equal to the rate of imputation times the gross dividend and represents an advance payment against its total corporation tax for the year. The shareholder receiving the net cash dividend also receives a tax credit equivalent to the basic rate of income tax on that dividend, which is used to offset his/her income tax liability. Therefore, under the partial imputation system, if a company makes a pre-tax profit of £100, the after-tax profit is equal to £100 (1c), where c equals the rate of corporation tax. Assuming that the company pays out all its after-tax profits as dividends to its sole shareholder, the shareholder receives a cash dividend of £100(1c) and a tax credit equal to £100[(1c)i]/[1i], giving a gross dividend (net cash dividend plus tax credit) of £100[(1c)]/[(1i)] (where i is the rate of imputation). The gross dividend is subject to
7 Prior to March 1993, the rate of imputation was always equal to the basic rate of income tax. However, from March 1993, the rate of imputation has been 20% (the same as the lower rate of income tax) compared to a basic rate of income tax of 25%. However, both lower rate and basic rate income taxpayers are deemed to be subject to a tax on dividend income of 20% (the rate of imputation).
120
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
Table 6 Dividend received after tax on a distribution of £100 pre-tax profit Corporate tax system
Classical system Preference Partial imputation system Preference
Formula
£100(1c)(1p) [£100(1c)(1p)]/ (1i)
Tax status of shareholder Exempt p=0
Basic or lower rate p = 0.25
Higher rate p = 0.4
£67.00 N £89.33
£50.25 R £67.00
£40.20 R £53.60
D
N
R
Assuming c=0.33 (the rate of corporation tax) and i=0.25 (the rate of imputation). Preferences: D=dividends; R=retentions; N=neutral.
income tax at the shareholder’s marginal income tax rate of p. Hence, after income tax, the shareholder receives £100[(1c)(1p)]/[(1i)]. For example, consider the following: a company earns a pre-corporation tax profit of £100,8 the corporation tax rate is 33% and the company pays out all of its profits as a cash dividend of £67. Assuming the imputation rate is 25%, the shareholder will receive a net cash dividend of £67 and a tax credit of £22.33, and the company will pay ACT of £22.33. Hence, the shareholder is deemed to have received a gross dividend of £89.33 on which £22.33 income tax has already been paid by the company in the form of ACT. The remaining tax liability or tax refund due to the shareholder depends upon the marginal income tax rate faced by the shareholder. For a basic rate taxpayer (facing a marginal income tax rate of 25%), the tax credit received is deemed to fully discharge his/her liability to income tax and no further income tax is due. For a higher rate taxpayer (subject to a marginal tax rate of 40%), an additional tax income charge arises equal to 15% of the gross dividend received (that is, an additional tax charge of £13.40). Finally, tax-exempt shareholders (such as pension funds and the pension business of insurance companies) receive a cash refund of the tax credit from the Inland Revenue.9 Table 6 illustrates the difference between the classical and partial imputation systems by calculating the after tax dividend which would be received by the different groups of taxpayers on a distribution of £100 pre-tax profits. In order to appreciate the incentives inherent in the two systems towards the payment of dividends or the retention of profits, the value of after-tax dividends is compared with the taxation of retained profits (ignoring
8
The following example describes the tax rates operating over the period of the empirical study, 1988 to 1992. In 1992, an additional income tax band was introduced, the lower rate band of 20%, and the rate of imputation was reduced to 20%. However, these changes do not affect the conclusions drawn in this section. 9 The July 1997 Budget has abolished tax credits to pension schemes and UK companies. These changes in the operation of the imputation tax system will provide opportunities for future researchers to examine whether the present tax bias in favour of dividends for institutional shareholders does have a significant effect on UK dividend policy. The potential consequences of such a change are complex: it may remove the tax bias in favour of dividends for institutional shareholders, but by reducing the after-tax value of the dividend, may lead to such institutions demanding higher levels of dividends to cover their cash short-fall. Clearly, the changes in the dividend tax system will allow researchers to better isolate the agency and signalling effects on dividend policy and to provide a richer test of the determination of dividend policy.
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
121
the effects of capital gains tax). Assuming a corporation tax rate of 33%, pre-tax profits of £100 are equivalent to £67 if retained in the company. As Table 6 illustrates, it is only taxexempt shareholders in the UK who will prefer dividends to retentions. Although the preferences for dividends or retentions are potentially complicated by the imposition of capital gains tax (CGT), the effective rate of CGT, in both the US and UK, is much lower than the nominal rate, due to the provision of annual exemptions from CGT, the ability of taxpayers to tax plan with regard to the realisation of capital gains and the fact that CGT can be deferred indefinitely.10 Therefore, following the lead of prior authors, the current analysis assumes that the effective rate of CGT is zero (see Bond et al., 1995).
References Agrawal, A., Jayaraman, N., 1994. The dividend policies of all-equity firms: a direct test of the free cash flow theory. Managerial and Decision Economics 15, 139 – 148. Baltagi, B., 1995. Econometric Analysis of Panel Data Wiley, New York. Bhattacharya, S., 1979. Imperfect information, dividend policy and the ‘‘bird in the hand’’ fallacy. Bell Journal of Economics 10 (1), 259 – 270. Bhattacharya, S., 1980. Nondissipative signalling structure and dividend policy. Quarterly Journal of Economics 95, 1 – 24. Black, F., 1976. The dividend puzzle. Journal of Portfolio Management 2, 72 – 77. Bond, S., Meghir, C., 1994. Financial constraints and company investment. Fiscal Studies 15 (2), 1 – 18. Bond, S., Chennells, L., Devereux, M., 1995. Company dividends and taxes in the UK. Fiscal Studies 16 (3), 1 – 18. Easterbrook, F.H., 1984. Two agency-cost explanations of dividends. American Economic Review 74 (4), 650 – 659. Eckbo, B.E., Verma, S., 1994. Managerial shareownership, voting power, and cash dividend policy. Journal of Corporate Finance 1, 33 – 62. Fama, E.F., Babiak, W., 1968. Dividend analysis: an empirical analysis. Journal of the American Statistical Association 63, 1132 – 1161. Haskins, C., 1995. A company should be run primarily in the interests of its shareholders. Enterprise and Governance: The Proceedings of a Conference Held at the Institute of Directors on 24 October, IoD, London. Hutton, W., 1995. The State We’re In Jonathan Cape, London. Jensen, M.C., 1986. Agency costs of free cash flow, corporate finance and takeovers. AEA Papers and Proceedings 76 (2), 323 – 329. Jensen, G.R., Solbreg, D.P., Zorn, T.S., 1992. Simultaneous determination of insider ownership, debt, and dividend policies. Journal of Financial and Quantitative Analysis 27 (2), 247 – 263. Kester, W.C., 1992. Industrial groups as systems of contractual governance. Oxford Review of Economic Policy 8, 24 – 44. Leland, H., Pyle, D., 1977. Information asymmetries, financial structure and financial intermediation. Journal of Finance, 371 – 387, May. Lintner, J., 1956. Distribution of incomes of corporations among dividends, retained earnings and taxes. American Economic Review 46, 97 – 113. Mayer, C., 1994. Stock markets, financial institutions, and corporate performance. In: Dimsdale, N., Prevezer, M. (Eds.), Capital Markets and Corporate Governance. Clarendon Press, Oxford, pp. 179 – 194. Mayer, C., Alexander, I., 1990. Banks and markets: corporate financing in Germany and the UK. Journal of Japanese and International Economics 3, 450 – 475.
10 In the UK, CGT raised only £0.9 billion from individuals in the tax year 1992 – 1993 (HMSO, Economic Trends, October 1994).
122
H. Short et al. / Journal of Corporate Finance 8 (2002) 105–122
Miller, M.H., Rock, K., 1985. Dividend policy and asymmetric information. Journal of Finance 40, 1031 – 1051. Moh’d, M.A., Perry, L.G., Rimbey, J.N., 1995. An investigation of the dynamic association between agency theory and dividend policy. Financial Review 30 (2), 367 – 385. Roe, M.J., 1990. Political and legal restraints on ownership and control of public companies. Journal of Financial Economics 27, 7 – 42. Rozeff, M.S., 1982. Growth, beta and agency costs as determinants of dividend payout ratios. Journal of Financial Research 5, 249 – 259. Trade and Industry Committee, 1994. Competitiveness of UK Manufacturing Industry. Second Report, HC 41.I, HMSO, London. Waud, R., 1966. Small sample bias due to misspecification in the ‘partial adjustment’ and ‘adapted expectations’ models. Journal of the American Statistical Association, pp. 134 – 145, December. Williamson, O.E., 1964. The Economics of Discretionary Behaviour: Managerial Objectives in a Theory of the Firm. Prentice-Hall, New York. Zeckhauser, R.J., Pound, J., 1990. Are large shareholders effective monitors? an investigation of share ownership and corporate performance. In: Hubbard, R.G. (Ed.), Asymmetric Information, Corporate Finance and Investment. University of Chicago Press, Chicago, pp. 149 – 180.