Share Repurchases, Dividends and Executive Options: the Effect of Dividend Protection

Transcription

1 European Financial Management, Vol. 12, No. 1, 2006, 7 28 Share Repurchases, Dividends and Executive Options: the Effect of Dividend Protection Eva Liljeblom and Daniel Pasternack Swedish School of Economics and Business Administration, PO Box 479, Helsinki, Finland eva.liljeblom@hanken.fi; daniel.pasternack@hanken.fi Abstract We study the determinants of share repurchases and dividends in Finland. We find that higher foreign ownership serves as a determinant of share repurchases and suggest that this is explained by the different tax treatment of foreign and domestic investors. Further, we also find support for the signalling and agency cost hypotheses for cash distributions. The fact that 41% of the option programmes in our sample are dividend protected allows us to test more directly the substitution/ managerial wealth hypothesis for the choice of distribution method. When options are dividend protected, the relationship between dividend distributions and the scope of the options programme turns to a significantly positive one instead of the negative one documented in US data. Keywords: share repurchases; executive options; foreign owners. JEL classification: G12, G32, G35 1. Introduction The focus in studying share repurchases has recently turned from the announcement impact 1 to determinants for the choice of the cash distribution method (dividends versus share repurchases). Prior studies have documented support for several motives, such as the existence of executive stock option programmes. We would like to thank the editor, John Doukas, Steven Huddart, Sami Torstila, an anonymous referee as well as seminar participants of the Joint Finance Seminar in Helsinki and the Southern Finance Association 2002 Annual Meeting, for valuable comments. Financial support from the Academy of Finland is gratefully acknowledged. 1 Prior research typically has documented a positive price reaction, see e.g. Dann (1981) and Vermaelen (1981), less so to Dutch auction and open market repurchases (Comment and Jarrell, 1991). In Finland, a positive price reaction has been detected by Karhunen (2002). Journal compilation # 2006 Blackwell Publishing Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA.

2 8 Eva Liljeblom and Daniel Pasternack However, a common shortcoming in previous studies is the accuracy of measurement regarding both share repurchases and stock options. This has complicated direct testing of the tax and stock option motives for the choice of cash distribution method. Contrary to prior studies we use a comprehensive and unique Finnish data set, which allows us to measure share repurchases as well as the stock option motives for them in a more direct way compared to previous papers. We can therefore contribute by being able to, for the first time, explicitly test for whether the cash distribution policy is dependent on whether executive stock options are dividend protected or not. We further study the tax motive in the choice of the distribution method, and different determinants for the share repurchases actually performed. Jagannathan et al. (2000) list as non-mutually exclusive factors for the choice between dividends and share repurchases the following: (1) asymmetric information, (2) taxes, and (3) stock options. 2 First, asymmetric information can lead to signalling not only concerning the general level of the company performance, but also more specifically about the relative permanence of the cash flows (see, e.g., Guay and Harford, 2000; Jagannathan et al., 2000; Balachandran et al., 1996). Second, Talmor and Titman (1990) and Bagwell and Shoven (1988) argue that tax effects influence firms to use more repurchases. 3 However, the tax advantages of share repurchases in the USA were largely reduced with the 1986 tax reform. The remaining tax advantage for share repurchases is at the personal level and the magnitude of the advantage depends on investor characteristics. Empirical findings regarding the relation between these characteristics and the choice between dividends and repurchases remain mixed (see, e.g., Bartov et al. (1998) and Jagannathan et al. (2000)). Third, employee/executive stock options have been suggested as a reason for firms preferring share repurchases by Jolls (1998), Weisbenner (2000), Fenn and Liang (2001) and Kahle (2002). Stock options could influence payout decisions for two reasons. Dividend payments reduce the stock price on the ex-dividend date, and thus the option value unless it is dividend protected, whereas share repurchases might even improve the stock price. Kahle (2002) calls this the substitution/managerial wealth hypothesis and finds support for it, since unexercisable executive options are significantly positively related to the decision to repurchase. 4 Alternatively, share repurchases can be used to fund executive options by counteracting the dilution of the stock price otherwise caused by option exercises. Kahle (2002) calls this the option funding hypothesis. The economic rationale for this is unclear. A possible reason is that option exercises dilute all earnings multiples, and this dilution can be eliminated 2 Additional suggested reasons in the literature are, for example, adjusting capital structure or the bondholder wealth expropriation hypothesis (Vermaelen, 1981), or reducing agency costs i.e. the free cash flow problem (Nohel and Tarhan, (1998) also supported empirically in Stenius (2000) on Finnish data but not in Howe et al. (1992) for the USA). Also takeover defence has been suggested by Denis (1990) and Bagwell (1992). 3 Other cost aspects concerning the two distribution methods i.e. transaction cost differences are highlighted in Barclay and Smith (1988). 4 Jolls (1998) evidence points even more clearly towards the substitution hypothesis. She found evidence indicating that it is executive stock options, rather than employee stock options, that influence stock repurchases. A negative relationship between dividends and managerial stock options has also been found in Lambert et al. (1989) and Fenn and Liang (2001).

3 Share Repurchases, Dividends and Executive Options 9 by share repurchases. Although such cosmetic changes in reported earnings should be irrelevant, there is some empirical evidence supporting this view. 5 The contributions of this paper are several. First, contrary to US studies, we can directly measure share repurchases. As described in Jagannathan et al. (2000), in the US share repurchases are often measured using data from the Securities Data Company database, which however only reports intended, not actual, repurchases. Another source, the Compustat database, reports a measure for usage, but this variable is an aggregation of many other types of transactions and overstates (sometimes substantially) actual repurchases. Jagannathan et al. (2000) used changes in the decreases of shares outstanding as reported by CRSP but furthermore adjusted by stock splits, dividend reinvestment plans, and new stock issues. However, the effects of exercised stock options not accounted for blur even this variable. Contrary to the US studies, we have direct data on the maximum initial amount of shares that the company can repurchase 6 as well as the actual share repurchases made by the Finnish companies on a daily basis. 7 Therefore we can precisely measure the actual gross repurchases and their timing in order to investigate whether share repurchases are used to boost the stock price when options exercises are approaching. Second, our data allows us to measure the characteristics of the stock options programmes of each company, including whether the options are dividend protected or not. This information has not been used previously. According to Murphy (1999), only 1% of CEO options in the USA have dividend protection, and in Weisbenner s (1998) sample, only two companies out of 799 offered dividend protection. 8 In Finland, the use of dividend protection is increasing. As many as 41% of the options in our data set are dividend protected. This allows us to perform a more direct test of the substitution hypothesis, since companies with dividend protected options should have no reason to avoid dividend distributions. 9 5 See, for example, Weisbenner (2000) and Kahle (2002) for anecdotal and empirical evidence related to dilution avoidance as a rationale for the option-funding hypothesis. In Finland, the profit from executive option exercises is taxed as salary income at the personal level, and firms must pay social cost expenses (approximately 5%) on the sum of such distributed income. Some firms (e.g. Sonera Corp.) have motivated repurchasing shares by saying that the shares provide a hedge against such social cost expenses. Holding own stocks for the same amount as these expected expenses (repurchasing stocks for an amount equalling that given by the social cost percentage on outstanding options) provides a hedge against these expenses. However, the position involves a down-side risk and could also be interpreted as a (under)valuation signal. Other explanations for option funding behaviour may be that the management would not want to dilute its ownership stake, or that the firm would have an optimal capital structure which the option exercises would move them away from. 6 This is the number approved by the shareholders meeting. 7 In Finland, the company has to report these numbers on an ex post basis. 8 Fenn and Liang (2001) discuss the reasons for such low utilisation of dividend protection in the USA The most compelling explanation seems to be the fact that the accounting standards in the USA have defined dividend protected options as variable-plan options, the costs of which must be recognised on the income statement contrary to fixed-plan options, which do not result in a compensation expense as long as the option has an exercise price greater than or equal to the current market price. 9 Our test also differs from, for example, the one by Kahle (2002), who only compared share repurchasing versus dividend increasing firms (two non-overlapping samples) in that we investigate the whole stock market in Finland, i.e. our sample includes both dividend increasing as well as decreasing firms, active or not active at the same time with share repurchases. By having the whole empirical distribution of dividend changes, and the variable not used before i.e. information on dividend protection, we hope to capture a more detailed picture of the determinants of the corporate distribution policy, and the role of dividend protection in that.

4 10 Eva Liljeblom and Daniel Pasternack Third, in Finland domestic taxed investors have a tax-induced preference for dividends compared to capital gains, whereas foreign investors are subject at least to the withholding tax. Taxes thus influence the payout policy preferences of these two categories in a different way. Furthermore, these two categories have varying degrees of dominance in Finnish firms. We therefore expect that the degree of foreign ownership could in part explain the Finnish companies payout policy decisions (the choice between dividends and share repurchases). We also investigate this. The rest of the paper is organised as follows. In Section 2, we describe the Finnish market in terms of share repurchases and executive options, as well as the data used. The methods and results are reported in Section 3. Section 4 concludes the paper. 2. The Data and the Institutional Setting Share repurchases were allowed in Finland in Since then, a large number of companies on the Helsinki Stock Exchange s main list have announced share repurchase programmes, altogether 138 programmes by the end of September These programmes would have distributed approx. E10.7 billion (10,699,171,401) if fully utilised. Finnish companies have introduced altogether 257 executive option programmes during 1987 to October Share repurchases are governed by many rules both in the 1997 Companies Act, and the rules of the Helsinki Stock Exchange. The maximum number of shares that can be repurchased is limited to 5% of the total share capital. The shares can be bought only using free equity, i.e. proceedings that could also alternatively be paid out as dividends. The acquisition authorisation has to be given by the shareholders meeting. The authorisation is effective for one year. The shares can be repurchased through openmarket repurchases in amounts not seriously affecting the normal trading volumes of the stock, and the repurchases made during a day have to be reported immediately to the Helsinki stock exchange (before the next market opens). The repurchases can only start one week after the public announcement of the board s repurchase decision. In this study we use firms listed on the Helsinki Stock Exchange s main list during the time period from April 1996 to September The decision to decide to repurchase stocks is investigated by means of yearly tobit or probit regressions where the test years are the share repurchase/dividend decisions during the accounting years The sample is described in Table 1. In line with Jagannathan et al. (2000) and Fenn and Liang (2001), for example, we analyse the decision to repurchase stocks by modelling the repurchase and dividend decisions (estimating their determinants) in a yearly pooled framework. While Jagannathan et al. (2000) estimated multinomial logit models (explaining jointly the dividend and repurchasing decisions or combinations thereof), Fenn and Liang (2001) estimate separate tobit regressions for either the dividend or repurchase decisions, with actual share repurchase payout as the dependent variable in the share repurchase model. In the Finnish case, the law puts severe restrictions on the upper limit of the repurchase authorisation (a maximum of 5% of the share capital). Most companies therefore 10 Since companies usually decide on share repurchases or dividends in springtime (from 1997 to 2001 in our data set), the dependent variables are for these years. The accounting data based variables, on the other hand, are measured at the previous accounting year-end, while the pricerun up variable is measured using stock price data for a 6-month period prior to that.

5 Share Repurchases, Dividends and Executive Options 11 Table 1 Firms listed on the Helsinki Stock Exchange, , and the final sample. This table describes the construction of our final data set. Out of the total number of observations (firms which were listed a specific year), some observations were lost due to missing data on share repurchases, on foreign ownership, on previous year financial statement data, or on a price history of 6 months. Panel A describes how the sample is reduced because of these missing data items. For example in 1997, out of a total number of 78 firms, 39 had at least one (usually more than one) missing data item, leaving us with ¼ 39 firms in the final sample. Panel B reports the number of share repurchase and stock option programmes in the final sample Total Panel A. Sample selection Total no. of firms listed Missing data concerning: - Share Repurchase Stock Options Foreign Ownership Market-to-Book Income Statement Dividends Return Data Total no. of firms lost Firms left in final sample Panel B. The decomposition of the final sample Final sample includes: no. of firms repurchasing stock no. of firms with options both repurchases and options actually ask for an authorisation for the maximum amount, but then may later use less of it. Since we want to examine the actual usage separately, we first focus on whether an authorisation is sought for or not. 11 We therefore run a probit model for share repurchase authorisations, whereas for the dividend decision, we estimate a tobit model with the dividend payout (zero or a variable measuring relative dividends, defined as dividends/ assets) as the dependent variable. Later, we also estimate monthly repurchase timing regressions, where the dependent variable is the actual monthly repurchase divided by total repurchases. In Finland, dividends are effectively tax-free for domestic taxed investors because of the avoir-fiscal (imputation) system. 12 This leads to a preference for dividends compared to capital gains for domestic taxed investors. However, foreign investors are (depending of the country of origin) mostly subject to a source tax, most commonly on a level of 15%, and do not get the tax credit created by the imputation tax 11 So far, in Finland, when an authorisation is proposed at the shareholder s meeting, it has always been approved. 12 The investor can deduct the taxes paid by the company, and since the corporate tax rate is the same as the tax rate on personal capital income, the net dividend actually paid out by the company is tax-free. For further details, see Liljeblom et al. (2001).

6 12 Eva Liljeblom and Daniel Pasternack system. Tax arbitrage has also been detected on the Finnish market in Liljeblom et al. (2001), and low dividend yield also turns out to be significant when explaining capital flows to Finnish companies (Liljeblom and Lo flund, 2000). More evidence on tax related investor clienteles in the Finnish market is provided in a recent paper by Felixson and Liljeblom (2003), who study actual trading behaviour around ex-dividend days using daily transaction data for all investors on the Finnish market between 1996 and They find significantly higher selling behaviour just before the ex-dividend day in a group of foreign investors, the counterpart being dividend capturing domestic taxed companies and households. In line with these results, one would expect that share repurchases would be more favoured in companies with higher foreign ownership. We therefore include a variable measuring the degree of foreign ownership (the percentage of share capital owned by foreigners) at the start of the year. 13 If the argument for share repurchases is option funding, a relationship between the scope of the option programme and the share repurchase decision could be expected (if a company does not have an options programme, this variable takes the value of zero). On the other hand, since this variable includes executive options, a positive relationship between the scope of the options programme and share repurchases (and a negative between the scope and dividends) might be due to the substitution hypothesis. 14 We therefore included a variable for not-dividend protected option programmes, which might capture the effects of the substitution hypothesis. This variable is a multiplicative dummy, measured as the scope of the options programme multiplied by a dummy that takes the value of 1 if the options are not dividend protected. 15 Furthermore, in the 13 There may also, of course, be other than tax related reasons for foreign investors to prefer share repurchases for example. The foreign investors on the Finnish market seem largely to be institutional investors, as discussed in, for example, Grinblatt and Keloharju (2000). Bartov et al. (1998) found a positive relationship between institutional investors and share repurchases on the US market. This despite the fact that tax-exempt institutional investors in the US might be argued to be more attracted to dividend paying firms in a signalling equilibrium as in Allen et al. (2000). Since there thus may be other reasons for institutional investors to prefer share repurchases, reasons strong enough to counteract those for dividends in the USA, such forces may be at work in our case as well. One such may be transaction costs; dividends create a need for portfolio rebalancing while share repurchases do not, since they can be avoided. Unfortunately we have no information about the actual composition of the group of foreign investors (countrywise or in other aspects), and cannot separate between tax and other potential explanations for a preference for share repurchases. Since there is evidence from foreign investor trading behaviour on the Finnish market, which is highly likely to be tax related (Felixson and Liljeblom, 2003), we treat the foreign ownership variable mainly as a tax variable. 14 The Finnish option programmes always include as their major element executive options. Programmes are often also directed to both executives and employees/key personnel. However, the employee portion of the programme is usually small. The fact that executive options are always the main element makes it unfortunately impossible for us to test for the different predictions of the substitution hypothesis for executive versus employee options in the way Kahle (2002) did. 15 The incentives for the managers to reduce dividends are likely to be a function of their wealth loss, i.e. the reduced value of their option position when the stock goes ex-dividend. This wealth loss in turn is linked to the size of the option programme. We therefore use a multplicative dummy as there is a large variation in the scopes of the programmes in our sample, and we expect the reluctance for dividends to be especially great in companies where the non-protected option programmes are big.

7 Share Repurchases, Dividends and Executive Options 13 timing regressions, we use the degree of out-of-the-moneyness 16 of the option programme at the beginning of the year, and the time-to-maturity of the programme. If the purpose of the share repurchase activity is to boost the stock price, we expect a higher likelihood for share repurchases in companies with out-of-the-money options, and option programmes soon to expire. Finally, we include in the probit and tobit models several control variables to capture other factors affecting the share repurchase and dividend decisions. In line with Guay and Harford (2000) and Jagannathan et al. (2000), we test whether dividends are more often paid out of permanent earnings and share repurchases out of temporary earnings. We therefore include measures for both operating income (operating profit/sales) and nonoperating income (extraordinary income or charges/sales) in the share repurchase and dividend models, respectively. In Finland, however, there may be less clear reasons to expect that dividends would be, above all, related to operating income. This is because the avoir-fiscal tax rule favours, for domestic owners, the distribution of all taxable profits in the form of dividends since double taxation can thus be totally eliminated. 17 In order to test for the agency cost motive for cash distributions we include, in line with Opler and Titman (1993), 18 variables measuring high free cash flow, and a symmetrically constructed variable for low free cash flow. High free cash flow is a dummy that takes the value of 1 for firms with a market-to-book ratio less than the sample median and net operating cash flow greater than the sample median. This variable measures firms that should have the most severe free cash flow problems due to low growth and much cash. Firms with a market-to-book ratio greater than the sample median, and a free cash flow less than the sample median, represent a counter group. We expect that both share repurchases and dividend payments would be positively related to high free cash flow, and negatively to low free cash flow. The signalling motive for share repurchases and dividends is captured by two variables. First, we include the market-to-book ratio as a measure for potential undervaluation. 19 Second, we include the prior 6-month return for the stock. In line with the signalling hypothesis, we expect a negative relationship between cash distributions and both of these variables. Descriptive statistics for our data are reported in Appendix A. Specifically, we report separately for the full sample and for four subsamples, divided either on the basis of whether the companies repurchase stocks or not, and whether they have options or not. Appendix A shows that the group of repurchasing firms differs significantly from that of firms that do not repurchase in terms of lower dividends to total payout, higher operating 16 The out-of-the-moneyness is measured as: min[(s X)/X, 0], where S is the share price at time t and X is the lowest option strike in effect at time t (firms may have several stock option programmes in effect at the same time, furthermore there might be several nodes in the programmes with different strike prices). 17 When the company pays taxes on extraordinary income, an imputation tax credit is created, but this tax credit can only be used to offset taxes on dividends during a time period of 10 years. Thus high extraordinary income also gives an incentive for higher dividends now or in the future. 18 Analyses of the effects of a similar variable were also performed in Fenn and Liang (2001). 19 In constructing the free cash flow variables, market-to-book was used as an indicator of growth opportunities. Thus the variable all in all in our model acts as an indicator of both potential overvaluation and/or growth opportunities.

8 14 Eva Liljeblom and Daniel Pasternack profits, a higher foreign ownership, and a larger/smaller proportion of high cash flow/low cash flow firms. The group of firms with stock options differ significantly from firms without options in terms of more share repurchases, lower operating profits, a higher foreign ownership and market-to-book value, as well as a higher proportion of low free cash flow firms. Table 2 reports the correlations between the different explanatory variables used in our analyses. The correlation coefficients between the different explanatory variables are in general rather low except for the correlation between the different option variables. The option variables will therefore not be used simultaneously in the regressions with the exception of the scope of the options programme and the no-dividend-protection variable, which do not seem to be creating problems due to multicollinearity The Results First, we test whether there are any significant differences between the monthly repurchasing activities of option and non-option firms. The results of these tests are reported in Tables 3 and 4. Table 3 shows that the repurchasing activity in the option firms is significantly higher at the 1% level as compared to non-option firms. Table 4 furthermore divides the option firms in two groups depending on the in-the-moneyness of the outstanding options. In line with our hypothesis, the repurchasing activity is higher among the out-of-the money option firms, and using the first measure for the monthly repurchasing activity (monthly repurchases/total authorisation) in Panel A, the difference is significant at the 5% level. In Panel B of Table 4, when monthly repurchasing activity is measured by monthly repurchases/actual total repurchases, the difference between the two groups of firms is not significant. When investigating for the reason for the difference in results between Panels A and B, it turns out that, as one could expect, when the actual usage of the total repurchase authorisation is high (a high denominator for our measure in Panel B), the repurchases are also more spread out over time. 21 This is evidenced by a high positive correlation of 0.49 between total repurchases/total authorisation, and the number of months during which repurchases have been made. This means in turn that although the total amount of actual repurchases is higher in the group of firms with out-of-the-money options, the difference to the first group is not as great in relative terms anymore, since the repurchases are spread over a larger number of months. The results in Panel B of Table 4 hence indicate a somewhat (but not significantly) more concentrated relative repurchasing activity for firms with out-of-the-money options Later, instead, we will split the sample and do some tests for various subgroups. 21 Financial inspection requires that repurchases should be made without affecting the normal price behaviour of the stock. If larger amounts are bought, this means that the purchases have to be spread over a larger number of days. 22 Panel A can be considered to give a more proper picture of the actual size (in absolute terms) of the repurchasing activities in the two groups compared to the relative results in Panel B. Since companies mostly ask for a repurchase authorisation of 5% of share capital (which is the highest possible allowed by law), the denominator in Panel A will take values related to the size of the firm s equity, and only the nominator will be related to actual repurchasing activity. Then the variable monthly repurchases/total authorisation in Panel A will properly measure whether the actual monthly repurchases have been large or small in relation to a measure for the size of the firm.

9 Share Repurchases, Dividends and Executive Options 15 Table 2 Correlation coefficients between different variables used in the analyses. This table reports correlation coefficients between explanatory variables used in the analyses of share repurchasing/dividend decisions. Repurchase authorisation is a dummy variable that takes the value of 1 if the company has applied for an authorisation for share repurchases during the year. Dividends/assets measures cash dividends to the book value of the company s assets at the previous year-end. Operating income is measured as operating profit/sales, and non-operating income as extraordinary income and charges/sales. Other variables are the stock return during the prior 6 months, the percentage foreign ownership of the company, the market-to-book ratio, option (1/0) which is a dummy variable that takes the value of 1 if the company has an options programme, the scope of the options programme measured as the maximum number of shares that can be obtained by option exercise divided by the total number of shares outstanding, and a dummy taking the value of one if the options are not dividend protected, respectively. High free cash flow is a dummy that takes the value of 1 for firms with a market-to-book ratio less than the sample median and net operating cash flow greater than the sample median, and low free cash flow is a dummy taking the value of one for firms with a market-to-book ratio greater than the sample median, and a free cash flow less than the sample median. Finally, a multiplicative dummy measured as the scope multiplied by the variable for no dividend protection is included. All accounting data based variables are measured at the last year-end as is the prior 6 month return and the foreign ownership. Variables Repurchase authorisation Dividend/assets Dividend/total payout Operating income Non-operating income Prior 6 month return Foreign ownership Market-to-book Option (1/0) Scope of option program Scope*no dividend protection High free cash-flow Low free cash-flow

10 16 Eva Liljeblom and Daniel Pasternack Table 3 Average monthly repurchases for stock option firm-months and for no stock option firm-months. The average monthly repurchases are measured as the number of shares bought during the month in percentage of the total authorisation, either during months when an option programme has existed (Panel B) or when one has not existed (Panel A). The t-test compares the averages in the two groups assuming unequal variances. Panel A No stock option firm-months (n ¼ 121) Panel B Stock option firm-months (n ¼ 507) Mean Median Standard deviation Minimum Maximum t-value for equal means no stock option vs. stock option firms 3.91 probability value Table 4 Monthly timing of repurchases for in-the-money and out-of-the-money firm-months. The average monthly repurchases are measured as the number of shares bought during the month in percentage of either the total authorisation (Panel A) or the actual final number of shares bought during the programme (Panel B). This has been measured separately for two categories, companies with either options that are in-the-money, or out-of-the-money. The t-test tests for the differences between the groups assuming unequal variances. Panel A Monthly repurchase/ Total authorisation Panel B Monthly repurchase/ Total repurchase Stock option firm-months (in the money) (n ¼ 214) Mean Median Standard deviation Minimum Maximum Stock option firm-months (out of the money) (n ¼ 224) Mean Median Standard deviation Minimum Maximum Equal means t-value for in-the-money vs. out-of-the-money probability value

11 Share Repurchases, Dividends and Executive Options The repurchase and dividend decisions The repurchase and dividend decision are investigated by estimating a probit model for repurchase authorisations and a tobit model for the annual dividend decision 23 using a yearly panel data of all firms for the period, i.e. using all 296 yearly firm-specific observations. The results are reported in Table 5. The results in Table 5, Panel A, shows that of the control variables, contrary to the expectations, share repurchases seem to be more related to permanent income instead of temporary income, the latter variable obtaining a coefficient close to zero. The results in Table 5 also support signalling (in terms of a significantly negative marketto-book variable) and, again, agency cost avoidance (in terms of expected and significant signs for the low cash flow variable) as motives for share repurchases, all these being significant at least at the 10% level. As discussed earlier, the Finnish tax system gives, from the viewpoint of domestic investors, an incentive to distribute enough dividends so that all imputation tax credits can be used, in which case dividends could be expected to be positively related to both forms of income because of the tax motive. This is also the case in Table 5, Panel B, both variables are significantly related to dividends at the 5% level. One of the two specific variables focused on in this paper, the foreign ownership variable, which we mainly interpret as a variable measuring tax differences (a tax disadvantage for foreign investors in the form of a source tax for dividends), obtains an expected sign in the share repurchase regression and is highly significant at the 1% level. Foreign ownership is also of economic significance since a 1% increase in foreign ownership (from the mean value) indicates a 0.39% increase in the share repurchase decision. Since foreign owners on the Finnish market are largely institutional owners, this result is also in line with Bartov et al. (1998) indicating that institutional investors prefer share repurchases. However, the results for foreign ownership are weakened by the fact that foreign ownership is insignificant (and enters with a wrong sign) in the dividend model. Since Table 2 shows that foreign ownership is positively correlated with the option variables, i.e. companies with a higher foreign ownership are more likely to have executive option programmes, 24 there is a chance that there is a multicollinearity problem when all these are simultaneously included in the models. We investigated for this and found that once the option variables were removed from the models in Table 5, all other coefficients remained practically unchanged except for the foreign ownership variable in the dividend model. It has now the expected negative sign, but is still insignificant. The overall evidence for foreign ownership as an explanatory variable (through the tax hypothesis) for the choice between share repurchases and dividends therefore remains somewhat weak. The option programme dividend protection variable (the scope of the option programme multiplied by a dummy taking the value of 1 if the options are not dividend protected) is, as expected, significantly negatively related to dividends (significant at the 5% level), while the scope of the programme as such takes a positive and significant sign. This indicates a significant difference between the dividend 23 In Finland dividends are usually paid annually instead of quarterly. 24 The correlations between foreign ownership and the option variables in Table 2 range between 0.20 and 0.30, and the correlation (not reported in Table 2) between foreign ownership and the dummy variable for no dividend protection is 0.32.

12 18 Eva Liljeblom and Daniel Pasternack Table 5 Share repurchase authorisation probit and dividend decision tobit regressions. This table reports results from estimating probit/tobit models for the share repurchase authorisations/dividend decisions using annual panel data for , 296 observations. The dependent variable in the share repurchase probit regressions is a dummy taking the value of one if a share repurchase authorisation is sought for and zero otherwise. The dependent variable for the dividend decision tobit model is the dividend/assets ratio. Operating income is measured as operating profit/sales, and non-operating profit as extraordinary income or charges/sales. Other variables are the stock return during the prior 6 months, the percentage foreign ownership of the company, the market-to-book ratio, the scope of the options programme measured as the maximum number of shares that can be obtained by option exercise divided by the total number of shares outstanding, the scope multiplied by a dummy taking the value of one if the options are not dividend protected, and variables measuring high and low free cash flow, respectively. High free cash flow is a dummy that takes the value of 1 for firms with a market-to-book ratio less than the sample median and net operating cash flow greater than the sample median, and low free cash flow is a dummy taking the value of one for firms with a market-to-book ratio greater than the sample median, and a free cash flow less than the sample median. All accounting data based variables are measured at the last year-end as is the prior 6 month return and the foreign ownership. Statistical significance at the 1%, 5% and 10% levels is denoted by ***, ** or *, respectively. Panel A Panel B Repurchase authorisation (probit) Dividend/assets (tobit) Cragg-Uhler R Ù % % of right predictions 71.3% Corr.(observed,expected) Ù 2 9.1% Observations 296 Observations 296 Observations at one 97 Non-limit observations 272 Observations at zero 199 Limit observations 24 Elasticity at means t-value (Prob.) Elasticity of index t-value (Prob.) Operating income (0.0196)** (0.0293)** Non-operating income (0.8206) (0.8688) Prior 6 month return (0.2552) (0.2978) Foreign ownership (0.0000)*** (0.8767) Market-to-book (0.0763)* (0.2063) Scope of option program (0.1285) (0.0154)** Scope*no dividend protection (0.8323) (0.0314)** High free cash-flow (0.3067) (0.1004) Low free cash-flow (0.0200)** (0.2473) Constant (0.0000)*** (0.0000)***

13 Share Repurchases, Dividends and Executive Options 19 distribution behaviour between companies with and without dividend protection for their option programmes. These results thus refine the results on the reluctance for dividends, obtained, for example, by Weisbenner (2000), Lambert et al. (1989) and Fenn and Liang (2000) on data for the US market, by relating that behaviour to the non-existence of dividend protection. However, the results are weakened by the fact that for repurchases, none of the option variables are significant (although they have the expected positive signs). In order to test further for the importance of being able to dividend protect options, we split the sample of firms with options into two parts, those with dividend protected options and those without, and re-estimated the dividend and share repurchase equations in order to investigate how the dividend protection itself affects a company s distribution policy. The results are reported in Table 6 in Panels A and B for the dividend decision, and in Panels C and D for the share repurchase decision. Table 6 shows that once the dividend protection is accounted for, a company with options does not seem to have any tendency to avoid dividends. Dividend distributions are instead significantly positively related (at the 1% level) to the scope of the options programme if options are protected for dividends (perhaps due to managerial motives to distribute good news whenever possible because of the option rewards), whereas the coefficient for the scope of the options programme is negative but insignificant for companies with options not protected for dividends. For share repurchases, few variables are significant in the subsamples. The option variables still fail to explain repurchases. However, foreign ownership is significantly positive at the 1% level for share repurchases in companies with dividend protection while not even quite at the 10% level in the other subgroup. There is a substantial difference in the level of foreign ownership between the subsamples; therefore the split between dividend and not dividend protected options weakens the explanatory power of foreign ownership in general. Also surprisingly the prior 6 month return, which earlier obtained a positive sign, i.e. a sign contrary to the expected one, is now even significant at the 5% level for dividend protected companies. Finally, as additional specification tests, we combined the total distributions (dividends and share repurchases) into one variable called total payout, and ran regressions examining the decomposition of the total payout (using dividends/total payout as the dependent variable). When forming the total payout variable, we used the total share repurchase authorisation (and not the ex post usage) as a measure for the share repurchase component. We expect this to somewhat blur our results, since it is customary for companies to ask authorisation for the maximum amount of 5% of equity, while often using a much smaller proportion. 25 This specification thus leads to the fact that in share repurchasing companies, the share repurchase component can easily (because of the generally large size of the authorisation request) dominate over the dividend component. Also, for companies not repurchasing at all, this variable always takes the value of 1, making it impossible to test for the effect of variables mainly affecting the size of the dividend distribution. Results from these regressions are reported in Panels A to C in Table This is the reason why we earlier preferred to conduct probit, not tobit regressions for share repurchases.

14 20 Eva Liljeblom and Daniel Pasternack Table 6 Dividend decision tobit regressions (Panels A and B) and share repurchase authorisation probit estimations (C and D) for firms with dividend protected or not dividend protected stock options. This table reports results from estimating tobit models for the dividend decision (Panels A and B) and probit models for the share repurchase authorisation decision (Panels C and D) in two samples of option companies, companies with options either dividend protected or not, using annual panel data for , 217 observations. The dependent variable is the dividend/assets ratio. Operating income is measured as operating profit/sales, and non-operating profit as extraordinary income or charges/sales. Other variables are the stock return during the prior 6 months, the percentage foreign ownership of the company, the market-to-book ratio, the scope of the options programme measured as the maximum number of shares that can be obtained by option exercise divided by the total number of shares outstanding, and variables measuring high and low free cash flow, respectively. High free cash flow is a dummy that takes the value of 1 for firms with a market-to-book ratio less than the sample median and net operating cash flow greater than the sample median, and low free cash flow is a dummy taking the value of one for firms with a market-to-book ratio greater than the sample median, and a free cash flow less than the sample median. All accounting data based variables are measured at the last year-end as is the prior 6 month return and the foreign ownership. Statistical significance at the 1%, 5% and 10% levels is denoted by ***, ** or *, respectively. Panel A Panel B Options dividend protected Options not dividend protected Corr.(observed, expected) Ù % Corr.(observed, expected) Ù 2 9.4% Observations 88 Observations 129 Observations at one 82 Observations at one 120 Observations at zero 6 Observations at zero 9 Elasticity at means t-value (Prob.) Elasticity at means t-value (Prob.) Operating income (0.0043)*** (0.0103)** Non-operating income (0.0027)*** (0.2180) Prior 6 month return (0.3133) (0.9273) Foreign ownership (0.3024) (0.3036) Market-to-Book (0.8966) (0.1276) Scope of option program (0.0006)*** (0.6016) High free cash-flow (0.0286)** (0.7159) Low free cash-flow (0.4691) (0.7576) Constant (0.0087)*** (0.0002)***

15 Share Repurchases, Dividends and Executive Options 21 Panel C Panel D Options dividend protected Options not dividend protected Cragg-Uhler R Ù % Cragg-Uhler R Ù 2 8.2% % of Right Predictions 71.6% % of Right Predictions 64.3% Observations 88 Observations 129 Observations at one 31 Observations at one 46 Observations at zero 57 Observations at zero 83 Elasticity at means t-value (Prob.) Elasticity at means t-value (Prob.) Operating income (0.4442) (0.6679) Non-operating income (0.4612) (0.8099) Prior 6 month return (0.0480)** (0.8739) Foreign ownership (0.0014)*** (0.1473) Market-to-book (0.1193) (0.2834) Scope of option program (0.9452) (0.1686) High free cash-flow (0.2652) (0.8783) Low free cash-flow (0.1590) (0.1441) Constant (0.0491)** (0.1283)

16 22 Eva Liljeblom and Daniel Pasternack Table 7 Regression examining distribution of the total payout, using dividends/total payout as the dependent variable. This table reports results from estimating regressions for the distribution of total payout, using dividends/total payout as the dependent variable. Panel A reports results for the complete sample. Panels B and C report results for companies with options that are either dividend protected or not. Operating income is measured as operating profit/sales, and non-operating profit as extraordinary income or charges/sales. Other variables are the stock return during the prior 6 months, the percentage foreign ownership of the company, the market-to-book ratio, the scope of the options programme measured as the maximum number of shares that can be obtained by option exercise divided by the total number of shares outstanding, and variables measuring high and low free cash flow, respectively. High free cash flow is a dummy that takes the value of 1 for firms with a market-to-book ratio less than the sample median and net operating cash flow greater than the sample median, and low free cash flow is a dummy taking the value of one for firms with a market-to-book ratio greater than the sample median, and a free cash flow less than the sample median. All accounting data based variables are measured at the last year-end as is the prior 6 month return and the foreign ownership. Statistical significance at the 1%, 5% and 10% levels is denoted by ***, ** or *, respectively. Panel A Panel B Panel C Whole sample Options dividend protected Options not dividend protected Adjusted R Ù % Adjusted R Ù % Adjusted R Ù % Observations 296 Observations 88 Observations 129 Coefficient t-value (Prob.) Coefficient t-value (Prob.) Coefficient t-value (Prob.) Operating income (0.7524) (0.2234) (0.0020)*** Non-operating income (0.0048)*** (0.5734) (0.1662) Prior 6 month return (0.9331) (0.0082)*** (0.2344) Foreign ownership (0.0002)*** (0.0000)*** (0.0514)* Market-to-book (0.8330) (0.3019) (0.5292) Scope of option program (0.9760) (0.0877)* (0.0061)*** Scope*No dividend protection (0.1746) High free cash-flow (0.2261) (0.1821) (0.3513) Low free cash-flow (0.2127) (0.2919) (0.4930) Constant (0.0000)*** (0.0000)*** (0.0000)***