Expropriation of minority shareholders and payout policy

Expropriation of minority shareholders and payout policy

The British Accounting Review 44 (2012) 207–220 Contents lists available at SciVerse ScienceDirect The British Accounting Review journal homepage: w...

209KB Sizes 3 Downloads 123 Views

The British Accounting Review 44 (2012) 207–220

Contents lists available at SciVerse ScienceDirect

The British Accounting Review journal homepage: www.elsevier.com/locate/bar

Expropriation of minority shareholders and payout policy Amedeo De Cesari* Finance and Accounting Group, Aston Business School, Aston University, Aston Triangle, Birmingham B4 7ET, UK

a r t i c l e i n f o

a b s t r a c t

Article history: Received 13 January 2011 Received in revised form 29 March 2012 Accepted 5 April 2012

This paper studies the payout policy of Italian firms controlled by large majority shareholders (controlled firms). The paper reports that a firm’s share of dividends in total payout (dividends plus repurchases) is negatively related to the size of the cash flow stake of the firm’s controlling shareholder and positively associated with the wedge between the controlling shareholder’s control rights and cash flow rights. These findings are consistent with the substitute model of payout. One of the implications of this model is that controlled firms with weak corporate governance set-ups, in which controlling shareholders have strong incentives to expropriate minority shareholders, tend to prefer dividends over repurchases when disgorging cash. Ó 2012 Elsevier Ltd. All rights reserved.

JEL classification: G32 G35 Keywords: Dividend Repurchase Controlling shareholder Minority shareholder Family firm

1. Introduction The main objective of this study is to investigate whether the payout policy of Italian controlled firms is affected by the agency conflicts between controlling and minority shareholders. These conflicts arise because controlling shareholders can expropriate minority shareholders by using firm cash and resources to pursue private interests. Expropriation of minority shareholders is likely to be common in Italy. The existing literature in fact shows that Italian corporations are often characterized by concentrated ownership structures, significant private benefits accruing to controlling shareholders, and the usage of control-enhancing devices to magnify control rights relative to cash flow rights (e.g., Bianchi, Bianco, & Enriques, 2001; Faccio & Lang, 2002). Payouts can be effective control mechanisms to reduce minority shareholder expropriation because they draw down the corporate cash reserves available to controlling shareholders. A key goal of this study is to analyse whether Italian controlled firms actually use payouts to protect minority shareholders from expropriation. Both payout levels and the choice between dividends and stock repurchases are studied to test the validity of the two alternative agency payout models of La Porta, Lopez-de-Silanes, Shleifer, and Vishny (2000) (LLSV (2000)). In LLSV’s (2000) outcome model, corporate outsiders utilise their legal powers to force firms and insiders to disgorge cash. In contrast, in their substitute model, firms distribute cash to shareholders to mitigate agency conflicts between corporate insiders and outsiders and payouts are seen as one of several alternative governance mechanisms that can be used to reduce agency costs. While the outcome model is consistent with a negative relation between the magnitude of payouts and the depth of the agency conflicts, the substitute model implies that firms facing higher agency costs and characterised by weaker corporate governance set-ups should pay out more.

* Tel.: þ44 (0) 121 2043028. E-mail address: [email protected]. 0890-8389/$ – see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.bar.2012.09.002

208

A. De Cesari / The British Accounting Review 44 (2012) 207–220

John and Knyazeva (2006) extend the substitute model by suggesting that, besides the level of payout, the composition of total pay out also has an influence on agency costs. They argue that dividend payouts, unlike stock repurchases, carry implicit commitments to similar or larger payouts in future periods. They conclude that dividend payments, owing to the conveyed pre-commitment signal to future cash distributions, are more powerful mechanisms than repurchases to attenuate agency conflicts. The view that the governance of Italian corporations is plagued by agency conflicts is quite common and has received clear support from the academic literature. The most compelling evidence is that Italy scores low in international studies of shareholder protection (e.g., La Porta, Lopez-de-Silanes, Shleifer, & Vishny, 1998). Also, the premium attached to voting shares is particularly large in Italy (e.g., Nenova, 2003; Zingales, 1994) probably because large private benefits can be gained through voting rights and corporate control. Mancinelli and Ozkan (2006) find that an Italian firm’s largest shareholder has significantly more voting rights than other major shareholders. This finding is in line with the idea that minority shareholders are unlikely to have the power to curb controlling shareholders’ attempts to engage in self-serving behaviours. Apart from weak shareholder protection, another feature of Italian corporations is ownership concentration and the existence of a controlling shareholder (e.g., Bianchi et al., 2001). In the sample period of this study, 86% of Italian listed companies have a controlling shareholder. It can be hypothesised that Italian controlled firms affected by serious agency conflicts have two main features. First, their controlling shareholders hold a small fraction of cash flow rights. A controlling shareholder with a small cash flow stake receives a small part of the cash paid out by its firm to shareholders. They, therefore, have a strong incentive to divert firm resources. Second, the difference between the control rights and the cash flow rights held by their controlling shareholders is large. When the wedge between a controlling shareholder’s control and cash flow stakes is large, controlling shareholders have the power to expropriate minority shareholders owing to their significant control stake and their small cash flow stake does not curb their tendency to indulge in self-serving behaviours. A sample of observations for Italian non-financial controlled firms listed on the Milan stock exchange (sample period 1999–2004) is analysed to test the validity of the two competing LLSV’s (2000) agency models of payout. These analyses essentially aim to test the joint hypothesis that the protection of minority shareholders is weak in the Italian corporate governance system and that the payout policies of Italian controlled firms can be explained by one of the two agency explanations. Evidence is found that is consistent with the substitute model and in favour of the notion that a firm’s payout policy is designed to reduce agency conflicts and costs. The key reported finding is that, when choosing between dividends and repurchases, firms with lower cash flow stakes held by their controlling shareholders and higher wedges between controlling shareholders’ control and cash flow rights are more likely to prefer dividends. This result indicates that firms adjust their dividend–repurchase mix to lower agency costs. In contrast, there is no strong evidence that firms increase payouts to mitigate agency conflicts. Firms with deeper governance problems seem reluctant to raise payouts possibly because the costs of higher cash disbursements outweigh the benefits in terms of reduced agency costs. This paper offers several significant contributions to the existing literature. First, the paper sheds light on the association between payouts and both the share of cash flow rights held by a firm’s controlling shareholder and the wedge between the controlling shareholder’s control and cash flow rights. There is very limited prior evidence on this issue (e.g., Faccio, Lang, & Young, 2001; Gugler & Yurtoglu, 2003). Second, contrary to most previous literature (e.g., Faccio et al., 2001; Gugler & Yurtoglu, 2003; LLSV, 2000), the study uses both dividend and repurchase data and shows that misleading conclusions can be drawn on the validity of the agency explanations of payout policy if researchers rely only on dividend data. Also, it shows that more comprehensive tests of these explanations should entail analyses of a firm’s choice between dividends and repurchases. Third, there are no previously published studies documenting that firms with controlling shareholders with small cash flow stakes and large differences between their control and cash flow rights are characterized by larger fractions of dividend payout in total payout. Finally, the research complements and extends Mancinelli and Ozkan’s (2006) study on the impact of ownership structure on dividend policy in Italy.1 2. Literature review and research questions 2.1. Agency problems and payout policy LLSV (2000) suggest two competing agency views of payout policy. In their outcome model of payout they posit that corporate outsiders use their legal powers to force their firms to distribute cash in order to prevent insiders from using firm cash to pursue private interests. In contrast, in LLSV’s (2000) substitute model, payouts are seen as a corporate governance control mechanism that can substitute alternative mechanisms. Firms distribute cash to shareholders to mitigate agency conflicts between corporate insiders and outsiders, especially when the latter group of investors is not already protected by other governance mechanisms. Insiders may want to alleviate agency conflicts by paying out cash possibly because firms affected by serious governance problems need to establish a reputation of fair treatment of corporate outsiders in order to raise external capital on favourable terms.

1 This study extends their work by also considering stock repurchases besides cash dividends, by using alternative ownership structure variables that fully reflect the impact of the control-enhancing mechanisms used by controlling shareholders, and by analysing data from a longer sample period.

A. De Cesari / The British Accounting Review 44 (2012) 207–220

209

The outcome model implies that firms with fewer agency conflicts and stronger protection of corporate outsiders should pay out more. The substitute model predicts that when corporate insiders can more easily expropriate outsiders by using firm resources to benefit themselves, firms disgorge more cash through payouts. LLSV (2000) test the validity of their models on a large sample of firms from 33 countries and find clear support for the outcome model of payout. Some further support for the outcome model is offered by Gugler and Yurtoglu (2003), Adjaoud and Ben-Amar (2010), and Chae, Kim, and Lee (2009). In contrast, some empirical evidence is consistent with the substitute model. For instance, it is often reported that firms with larger cash flows and fewer growth opportunities (e.g., Dittmar, 2000; Fama & French, 2001), lower insider ownership (Jensen, Solberg, & Zorn, 1992; Noronha, Shome, & Morgan, 1996; Rozeff, 1982), and more entrenched managers (Hu & Kumar, 2004) choose larger payouts. Further, John and Knyazeva (2006) use two corporate governance indices and find that firms with weak governance mechanisms tend to pay out more. John and Knyazeva (2006) extend the substitute model of payout by arguing that dividend payments are more effective than repurchases to address agency conflicts because current dividends are associated with an implicit pre-commitment to future dividends. This argument derives from the fact that dividend cuts are greeted negatively by the market (e.g., Aharony & Swary, 1980). Also, firms are reluctant to increase dividends to levels that cannot be sustained in the future (Brav, Graham, Harvey, & Michaely, 2005). A firm that provides dividends today is, therefore, implicitly pre-committing itself to paying similar or larger dividends tomorrow. In contrast, repurchases are a more flexible way to distribute cash than dividends (e.g., Brav et al., 2005; Jagannathan, Stephens, & Weisbach, 2000). The completion rates of announced repurchase programs are in many cases low (Stephens & Weisbach, 1998), and there is no evidence that failure to complete repurchase programs triggers negative reactions from the market. On the whole, the market sees repurchases as far weaker pre-commitment signals of future cash distributions than dividend payments. John and Knyazeva (2006) find empirical support for their argument and show that weakly governed firms are more likely to prefer dividends over repurchases when paying out cash. 2.2. Agency conflicts in Italian controlled firms: cash flow and control rights of controlling shareholders Most Italian firms are characterised by concentrated ownership structures and are controlled by a majority shareholder. Two key variables are used in the finance literature to evaluate a controlling shareholder’s incentives to expropriate minority shareholders: The size of their cash flow stake and the wedge between their control rights and cash flow rights. The share of cash flow rights or ownership rights held by a firm’s controlling shareholder determines the fraction of the cash distributed by the firm to shareholders that accrues to the controlling shareholder. For instance, if a shareholder owns 30% of the cash flow rights of a firm that pays out 100 Euros through dividend payments, the shareholder receives 30 Euros from the firm. If the firm distributes 100 Euros through repurchases, the shareholder may get less or more than 30 Euros because shareholders have the option not to tender their shares in repurchase programs.2 An inverse relation is expected between the size of a controlling shareholder’s cash flow rights (ownership stake) and their propensity to expropriate minority shareholders. Controlling shareholders with large (small) ownership stakes have weak (strong) incentives to avoid payouts and, instead, use their firms’ cash to pursue private benefits. This is the case because they receive a large (small) part of the cash that is distributed to shareholders. Hence, the depth of the agency problems and the size of the agency costs are negatively related to the size of the ownership stake of the controlling shareholder.3 Another variable that is often suggested as measure of potential minority shareholders’ expropriation is the wedge between a controlling shareholder’s control rights and cash flow rights (e.g., Claessens, Djankov, Fan, & Lang, 2002; Faccio et al., 2001). A shareholder’s control rights in a firm are defined as the fraction of the firm’s voting shares that the shareholder owns. Controlling shareholders with more control rights benefit from a stronger degree of control. Non-voting shares and pyramids are control-enhancing devices that are used in Italy by controlling shareholders to gain control rights in excess of cash flow rights (e.g., Bianchi et al., 2001; Faccio & Lang, 2002). Non-voting shares, which carry cash flow rights, may be sold to minority shareholders in order to allow controlling shareholders to boost their degree of control without requiring further investments. Similarly, pyramidal structures allow investors to control companies at the bottom of the pyramids without large direct equity investments. Overall, agency costs are expected to be a positive function of the wedge between controlling shareholders’ control rights and cash flow rights.4 2.3. Research questions This study investigates whether agency conflicts have a significant impact on payout policy and if the relation between payout policy and agency costs is consistent with the outcome model or the substitute model of LLSV (2000).

2 As suggested by an anonymous referee, this caveat is particularly meaningful for controlling shareholders. These shareholders are unlikely to tender shares when this can endanger their ability to control their company. They may also avoid taking part in repurchase programs to strengthen their control. 3 Several empirical studies find a positive relation between a firm’s value and performance and the cash flow rights of its controlling shareholder (e.g., Claessens et al., 2002; Joh, 2003; La Porta et al., 2002; Laeven & Levine, 2008; Mitton, 2002). 4 This expectation is consistent with several studies documenting that firms with larger differences between control and cash flow rights have worse performances and lower market valuations (e.g., Claessens et al., 2002; Joh, 2003; Laeven & Levine, 2008; Mitton, 2002).

210

A. De Cesari / The British Accounting Review 44 (2012) 207–220

Italy is a country characterized by firms with concentrated ownership structures, significant private benefits accruing to controlling shareholders, and the usage of control-enhancing devices to magnify control rights relative to cash flow rights; also, family-controlled firms are prevalent in Italy (e.g., Bianchi et al., 2001; Faccio & Lang, 2002). As discussed in Section 2.2, Italian controlled firms with serious agency problems and weak corporate governance share the following two features. First, their controlling shareholders hold small shares of their firms’ cash flow rights. Second, the wedges between the control rights and the cash flow rights held by the controlling shareholders are large. If the outcome model was supported, Italian firms with serious agency problems would be expected to distribute less cash than counterparts with more limited problems. In contrast, based on the substitute model, one would expect Italian firms afflicted by deeper agency conflicts to pay out more. As for the trade-off between dividends and stock repurchases, no clear predictions can be drawn in light of the outcome model of payout. However, based on the implications of the substitute model and John and Knyazeva’s (2006) arguments, the following prediction can be suggested: When choosing between dividends and stock repurchases, firms with significant agency problems tend to prefer dividends because they carry an implicit pre-commitment to future dividend payments. 3. Data collection and variable construction Information from the website of the Italian stock exchange (www.borsaitaliana.it) and Thomson Datastream is used to identify a list of 276 Italian non-financial firms listed on the Italian exchange in the sample period January 1999–December 2004. For each firm, data on dividends, repurchases, and other variables are collected for the sample period. After excluding firm-years for widely held firms and observations with missing data, the base dataset employed in the empirical tests comprises 176 firms and 630 firm-year observations. The data sources and the variables used in the empirical analyses are described below. 3.1. Dividend and repurchase payout Data are taken from Worldscope to compute DIV, a measure of dividend payout. DIV is the total cash dividends paid by the firm in the fiscal year multiplied by 100 and divided by the firm’s market capitalization at the end of the previous fiscal year. There is also the dummy variable DDIV that takes the value one if DIV is non-zero; else, DDIV is set to zero. Annual reports provide information on the number of ordinary shares repurchased on the open market.5 Since firms are not required to disclose data on the prices at which they repurchase stock it is not possible to obtain exact estimates of the cash spent to repurchase stock. Hence, for a certain firm-year, the value of the distributed cash through repurchases is assumed to equal the product between the number of the firm’s shares bought back in the year and the average split-adjusted price (from Thomson Datastream) of the firm’s stock over the year. In order to find the repurchase payout (REP) for the firmyear, the value of the distributed cash is scaled by the firm’s market capitalization (at the end of the previous year) and multiplied by 100. If REP is non-zero, the dummy variable DREP is equal to one; else, DREP is equal to zero. The total payout measure PAYOUT is the sum of DIV and REP and the dummy DPAYOUT takes the value one when PAYOUT is non-zero. A measure of the fraction of total payout that is paid out through dividends is also used. This measure (%DIV), which can be computed only for positive values of total payout, is the ratio between DIV and PAYOUT, times 100. In this study, the payout measures DIV, DDIV, REP, DREP, PAYOUT, DPAYOUT, and %DIV are the dependent variables. Panel A of Table 1 contains some descriptive statistics for these variables. Dividend payouts are larger and more common than repurchase payouts. The average value of DIV is 2.17, whereas that of REP is only 0.35. The average of DDIV is 0.72. This value implies that in 72% of the firm-years in the sample there is a positive dividend payout. In contrast, firms distribute cash through repurchases only in 30% of the years. As for the total payout measures, the average value of PAYOUT is 2.52 and the dummy DPAYOUT is equal to one in 76% of the observations. Finally, the mean value of %DIV is 86. This finding shows that 86% of the distributed cash is paid out using cash dividends. The number of observations for the variable %DIV is only 478 because the variable is not defined when PAYOUT is equal to zero. 3.2. Main ownership variables The cash flow and the control rights held by controlling shareholders in firms are measured by following a methodology similar to that used by Claessens et al. (2002) and Faccio and Lang (2002). For a particular firm-year, ownership variables are built using the latest available information on ownership before the start of the fiscal year. Ownership data are gathered from Consob’s (“National Commission for Companies and Stock Market”) website,6 firms’ annual reports, the database Factiva, and the World Wide Web. These data are used to identify the links of the control chains used by investors to control stakes in firms, separate holdings of shares that carry voting or control rights from those that do not, and determine whether a listed

5 In the period 1999–2004, repurchases on the open market were, by far, the most common type of buyback in Italy. Using the database Factiva, only three off-market repurchase transactions can be found in the period. 6 www.consob.it; Consob is the “public authority responsible for regulating the Italian securities markets”. The website publishes semi-annual reports on large shareholdings (greater or equal to 2%) held in firms listed in Italy.

A. De Cesari / The British Accounting Review 44 (2012) 207–220

211

Table 1 Descriptive statistics for all of the dependent and independent variables used in the study. Variable

Observations

Panel A: Payout variables DIV 630 DDIV 630 REP 630 DREP 630 PAYOUT 630 DPAYOUT 630 %DIV 478 Panel B: Ownership variables CRIGHTS 630 ORIGHTS 630 WEDGE 630 FAMILY 630 Panel C: Other variables OI 630 NOI 630 MB 630 TA 630 LEV 630 VAR 630 OPT 630

Mean

Median

Standard deviation

Minimum

Maximum

First quartile

Third quartile

2.171 0.717 0.347 0.297 2.519 0.759 85.915

1.655 1 0 0 1.883 1 100

3.236 – 1.131 – 3.487 – 27.393

0 – 0 – 0 – 0

47.141 – 13.058 – 47.141 – 100

0 – 0 – 0.018 – 85.825

3.072 – 0.091 – 3.413 – 100

50.082 44.195 5.887 0.805

51.267 49.875 0 1

15.414 19.834 10.277 –

20.321 0.347 3.695 –

91.143 91.143 45.548 –

37 29.776 0 –

59.683 57.57 9.288 –

0.015 0.006 1.086 13.166 0.334 0.0007 0.23

0.027 0.014 0.841 12.917 0.301 0.0006 0

0.134 0.124 0.894 1.807 0.236 0.0006 –

2.608 0.618 0.052 9.755 0 0.0001 –

0.192 2.652 11.21 18.362 0.967 0.0109 –

0.014 0.04 0.629 11.823 0.14 0.0004 –

0.068 0.009 1.241 14.341 0.508 0.0008 –

This table reports the descriptive statistics for a set of variables for samples of firm-years (fiscal years) belonging to the period 1999–2004. Only observations for controlled firms are included in the samples. A firm is defined as controlled if at least one of the firm’s shareholders holds 20% or more of the firm’s control rights. The data necessary to build the variables is obtained from Thomson Datastream, Worldscope, firms’ annual reports, Consob’s website, the database Factiva, and the World Wide Web. For a particular firm-year, DIV (REP) is the ratio between the cash distributed by the firm in the year through dividend payments (repurchases) multiplied by 100 and the firm’s market capitalization at the end of the previous year. The binary variable DDIV (DREP) is equal to one if DIV (REP) is non-zero, otherwise it is set to zero. PAYOUT is the sum between DIV and REP, and the dummy DPAYOUT equals one when PAYOUT is nonzero. %DIV is computed dividing the product between DIV and 100 by PAYOUT. For each firm-year, the ownership variables are constructed based on the latest available information before the start of the fiscal year. CRIGHTS and ORIGHTS are, respectively, the fraction of the firm’s control rights and the fraction of the firm’s cash flow rights held by the shareholder with the largest control stake (i.e. the controlling shareholder). WEDGE is the difference between CRIGHTS and ORIGHTS. The binary variable FAMILY is set to one if the firm is controlled by a family. For a particular firm-year, the variables OI, NOI, MB, TA, and LEV are computed using information as of the end of the previous fiscal year. TA is the natural logarithm of book value of total assets in thousands. OI and NOI are operating and non-operating income respectively scaled by total assets. MB is the sum of market capitalization and total debt divided by total assets. LEV is the ratio between total debt and the sum of market capitalization and total debt. VAR is the sample variance of the daily log-returns in the previous fiscal year. If, in the current fiscal year, there are ongoing stock option and/or stock grant plans for managers and directors, OPT is set to one; otherwise, it is equal to zero. For each of the variables above, the table reports the number of observations, the mean value, the median value, the standard deviation, the minimum value, the maximum value, the value of the first quartile, and the value of the third quartile.

firm has a controlling shareholder. As in La Porta, Lopez-de-Silanes, and Shleifer (1999) and Faccio and Lang (2002), a firm is defined as being controlled if there is at least one investor (individual or firm) that, through one or several control chains, holds a fraction of the firm’s control rights that is equal to or larger than 20%. Observations for widely held firms are excluded from the sample. Following previous literature (e.g., Claessens et al., 2002; Faccio & Lang, 2002), the “weakest link” principle is applied to estimate the degree of control of an investor over a stake in a firm. According to this principle, if an investor has some voting rights in a firm through a pyramidal control chain, the investor’s control rights are equal to the weakest link in the control chain.7 If an investor uses several control chains to hold a control stake in a firm, the investor’s control rights in the firm amount to the sum of the weakest links of all the control chains. The variable CRIGHTS is the percentage of the control rights held by the investor with the largest control stake, i.e. the controlling shareholder. Given the focus of this study on the Italian economy and stock market, firm-year observations belonging to controlled firms whose controlling shareholder is a foreign firm are excluded. The variable ORIGHTS is the percentage of cash flow rights held by a firm’s controlling shareholder. An investor with cash flow rights in a firm has the right to receive a fraction of the firm’s cash distributions to shareholders. Hence, holdings of both voting and non-voting shares are considered when measuring cash flow rights. The cash flow stake of a controlling shareholder is estimated by multiplying the cash flow rights of all the links of the chain the shareholder uses to control the stake.8 If the shareholder uses several control chains, the cash flow rights from all of them are added together to calculate the shareholder’s overall stake. Finally, the variable WEDGE is created by subtracting ORIGHTS from CRIGHTS. This variable is a measure of the discrepancy between the control rights and the cash flow rights held by a controlling shareholder.

7 The weakest link is the smallest control stake along a chain. For example, assume that Firm A holds 30% of the control rights of Firm B that, in turn, has a 40% control stake in Firm C. The control rights of Firm A in Firm C amount to 30% because the weakest link in the control chain is 30%. 8 For example, a controlling shareholder owns 15% of the cash flow rights of Firm A if the shareholder has a 50% cash flow stake in Firm B that, in turn, owns a 30% cash flow stake in Firm A (50%  30% ¼ 15%).

212

A. De Cesari / The British Accounting Review 44 (2012) 207–220

Some descriptive statistics for the ownership variables of this section can be found in Panel B of Table 1. As expected, the mean of CRIGHTS (50%) is larger than that of ORIGHTS (44%). Also, the average value of WEDGE is 6%.9

3.3. Other variables The control dummy FAMILY is equal to one for observations related to family-controlled firms; else, FAMILY is set to zero. A firm that is controlled by an Italian individual investor or by an Italian unlisted firm is defined as family-controlled. Control variables are created for firm profitability, growth opportunities, firm size, and leverage. OI and NOI are profitability measures. OI is equal to operating income scaled by total assets and NOI is non-operating income scaled by total assets. The control variable for growth opportunities is the market-to-book ratio (MB) that is computed by scaling the sum of market capitalization and total debt by total assets. The firm size measure used in this study is TA, which is the natural logarithm of total assets in thousands. Leverage (LEV) is total debt over the sum of market capitalization and total debt. For each firm-year observation, the variables OI, NOI, MB, TA, and LEV are calculated using Worldscope end-of-year data from the previous fiscal year. Also, a measure of firm risk (VAR) is included. This is the sample variance of all the daily log-returns (based on Thomson Datastream data) in the previous period. Further, data from annual reports are used to build the dummy variable OPT. This variable is equal to one in years in which there are ongoing stock option and/or stock grant plans for managers and directors. OPT is the only independent variable for which contemporaneous values are used. Finally, Table A1, in Appendix, provides values for the variable TAX over the period 1999–2004. This variable is the after-tax value (after both corporate and personal taxes) of before-tax earnings if the earnings are paid out to shareholders through cash dividends over the after-tax value of the same earnings if they are used to repurchase stock (or if they are retained). TAX varies in relation to whether the recipient of the cash distribution is an individual investor, a corporation or a government body. For a particular firm-year, the value of TAX that is used is the one which is relevant for the shareholder that has a direct stake in the firm and is part of the control chain used by the majority shareholder to control it. On the right hand side of every regression model, a constant term and sets of year and industry dummies are included. Coefficients on these variables are not reported in the paper. Based on the previous literature, payouts are expected to be positively related to OI, NOI, and TA and negatively associated with MB, LEV, and VAR (Dittmar, 2000; Fama & French, 2001; John & Knyazeva, 2006; Skinner, 2008). Also, the propensity to pay dividends rather than to repurchase stock should be negatively related to NOI, VAR, and OPT whereas increases in TAX should boost such propensity (e.g., Chay & Suh, 2009; Fenn & Liang, 2001; Jagannathan et al., 2000; Kahle, 2002). Previous research does not provide any clear guidance as to the effects of family control on payout policy. Descriptive statistics for the variables FAMILY, OI, NOI, MB, TA, LEV, VAR, and OPT are presented in Panels B and C of Table 1. 4. Results 4.1. Univariate differences The two panels of Table 2 include a series of univariate t-tests to investigate whether the means of the payout variables used in this study are related to the cash flow rights held by a firm’s controlling shareholder (ORIGHTS) and the difference between the control rights and the cash flow rights of a firm’s controlling shareholder (WEDGE). For robustness purposes, Table 2 also includes non-parametric Mann–Whitney–Wilcoxon tests (MWW tests) on the distributions of the payout variables. Panel A presents tests of whether the means and the distributions of the payout variables for observations belonging to firms with above-median values of ORIGHTS are different from the means and the distributions of the same variables for firmyears with below-median values of ORIGHTS. The cash flow stake of a controlling shareholder does not have a statistically significant influence either on the level of total payout (PAYOUT) or on the probability of a positive payout (DPAYOUT). The tests provide insignificant results also for the level of dividend payout (DIV), the probability of a positive dividend payout (DDIV), and the probability of repurchases (DREP). Based on the t-tests, it can be concluded that the level of repurchase payout (REP) is significantly larger and the value of %DIV is significantly smaller for observations with above-median cash flow rights. However, only the second of these findings is supported by the MWW tests at a 10% level of statistical significance. Panel B includes some tests of the impact of the variable WEDGE on the firm’s payout policy. Findings for the variable PAYOUT are mixed. According to the results of the t-test, there is a statistically insignificant relation between WEDGE and PAYOUT. However, the MWW test indicates that the distribution of PAYOUT is significantly different between firm-years with an above-median WEDGE and observations with a below-median WEDGE. DPAYOUT is significantly larger, at a 1% level, in

9 The minimum value of WEDGE is negative (4%). For a few observations in the sample, the controlling shareholder has fewer control rights than cash flow rights. A shared feature of these observations is that a stake in the controlled firm is held through a wholly-owned subsidiary of the firm itself. Following the Italian law, the voting rights of the subsidiary cannot be exercised in shareholder meetings of the parent firm. At the same time, the subsidiary keeps its cash flow rights attached to the stake in the parent firm. In this scenario, the controlling shareholder has, through the subsidiary, a cash flow stake in the controlled firm that does not carry any voting rights.

A. De Cesari / The British Accounting Review 44 (2012) 207–220

213

Table 2 Univariate tests on the payout variables. Variable

Observations

Panel A: Cash flow rights PAYOUT 630 DPAYOUT 630 DIV 630 DDIV 630 REP 630 DREP 630 %DIV 478 Panel B: Wedge PAYOUT 630 DPAYOUT 630 DIV 630 DDIV 630 REP 630 DREP 630 %DIV 478

Mean for observations with cash flow rights above median

Mean for observations with cash flow rights below median

Median for observations with cash flow rights above median

Median for observations with cash flow rights below median

p-value of t-test

p-value of MWWtest

2.482 0.772 2.022 0.711 0.46 0.317 82.025

2.555 0.745 2.318 0.723 0.237 0.277 89.804

1.679 1 1.485 1 0 0 100

2.125 1 1.835 1 0 0 100

0.7955 0.4266 0.252 0.7441 0.013 0.2657 0.0018

0.4945 0.4262 0.2006 0.7439 0.1125 0.2654 0.0885

2.733 0.8240 2.511 0.807 0.222 0.283 91.61

2.393 0.72 1.972 0.665 0.421 0.305 82.091

2.309 1 2.112 1 0 0 100

1.519 1 1.372 1 0 0 100

0.2374 0.0033 0.0436 0.0001 0.0331 0.5688 0.0002

0.0009 0.0034 0.0001 0.0001 0.2815 0.5684 0.0071

In the table, the mean values and the distributions of the payout variables PAYOUT, DPAYOUT, DIV, DDIV, REP, DREP, and %DIV are compared across different sub-samples through standard t-tests on means and non-parametric Mann–Whitney–Wilcoxon tests on distributions. Information from Worldscope and firms’ annual reports for the period 1999–2004 is used to build the payout variables. For a particular firm-year, DIV (REP) is the ratio between the cash distributed by the firm in the year through dividend payments (repurchases) multiplied by 100 and the firm’s market capitalization at the end of the previous year. The binary variable DDIV (DREP) is equal to one if DIV (REP) is non-zero, otherwise it is set to zero. PAYOUT is the sum between DIV and REP, and the dummy DPAYOUT equals one when PAYOUT is non-zero. %DIV is computed dividing the product between DIV and 100 by PAYOUT. In Panel A, the mean values (distributions) of the payout variables for observations related to firms with controlling shareholders holding above-median fractions of cash flow rights are compared to the mean values (distributions) of the same variables for observations of firms with controlling shareholders possessing belowmedian cash flow stakes. In Panel B, the mean values and the distributions of the payout variables are compared across two sub-samples: One with above-median values of the variable wedge, which is the difference between the controlling shareholder’s control and cash flow rights, and one with belowmedian values of this variable. In each of the two panels, the last two columns report the two-sided p-values of the t-tests and of the Mann–Whitney– Wilcoxon (MWW) tests.

firm-years with above-median values of WEDGE. These firm-years are also characterized by statistically significant (at least at a 5% level) larger values of the variables DIV and DDIV. It is unclear whether WEDGE is significantly related to REP given that the t-test and the MWW test provide opposite answers. Owing to the lack of statistically significant results, nothing can be concluded with respect to DREP. Finally, the value of %DIV is significantly larger, at a 1% level, for observations with an abovemedian WEDGE. On the whole, the results of the univariate tests support some of the implications of the substitute model of LLSV (2000). To be more specific, these results indicate that firms with weak governance set-ups tend to replace repurchases with dividend payments. 4.2. Regression analyses 4.2.1. Probabilities of positive dividend and repurchase payouts This section presents an investigation of whether the probability of a positive payout is dependent on the size of the control stake of a firm’s controlling shareholder and on the difference between their control rights and their cash flow rights. More specifically, non-overlapping sets of observations are created based on payout outcomes. One set comprises firmyears with zero payouts (firm-years with DDIV ¼ 0 and DREP ¼ 0). A second set is made up of observations with positive dividend payout and zero repurchase payout (DDIV ¼ 1 and DREP ¼ 0). A third set comprises observations with positive dividend and repurchase payouts (DDIV ¼ 1 and DREP ¼ 1). The remaining observations with zero dividend payout and positive repurchase payout (DDIV ¼ 0 and DREP ¼ 1) are included in the final set. A Multinomial Logit regression is estimated with four possible outcomes, one for each of the four sets above. The base outcome is chosen to be that in which both types of payouts are zero. This choice allows a direct study of the influence of ownership variables on firms’ propensity to make payouts. Results for the Multinomial Logit model are reported in Table 3.10 Coefficients on the variables ORIGHTS and WEDGE are statistically significant when the alternative outcome is positive dividends and no repurchases (DDIV ¼ 1 and DREP ¼ 0). In light of these significant results it can be concluded that the odds of paying cash dividends without making any repurchases rather than not making any payouts at all is negatively (positively) related to ORIGHTS (WEDGE). This finding is consistent with the notion that firms with weak governance set-ups are more likely to pay dividends to attenuate agency conflicts.

10 As explained in the Appendix, values of the variable TAX are computed for four alternative scenarios. In Table 3, only the specifications that comprise the variable TAX for the first scenario are reported. Findings for the other three scenarios are qualitatively similar. Likewise, Table 4 only shows regressions with TAX for the first scenario because results are not qualitatively different when alternative scenarios are considered.

Table 3 Multinomial Logit models: The controlling shareholder’s cash flow rights and the wedge between the controlling shareholder’s control and cash flow rights. Base outcome: DDIV ¼ 0 and DREP ¼ 0 Alternative outcomes:

Panel A: Cash flow rights Independent variables: ORIGHTS FAMILY OI NOI MB TA LEV TAX VAR OPT Observations Log-likelihood value Panel B: Wedge Independent variables: WEDGE FAMILY OI NOI MB TA LEV TAX VAR OPT Observations Log-likelihood value

DDIV ¼ 1 and DREP ¼ 0

DDIV ¼ 1 and DREP ¼ 1

DDIV ¼ 0 and DREP ¼ 1

0.022** (2.03) 0.453 (0.87) 48.889*** (6.44) 48.169*** (6.32) 0.324 (1.58) 0.0003 (0) 0.612 (0.7) 0.091 (0.07) 2069.21*** (4.88) 1.089** (2.43)

0.012 (1) 1.572*** (2.63) 52.575*** (6.91) 48.105*** (6) 0.236 (0.94) 0.167 (1.14) 0.456 (0.41) 1.617 (1.04) 3264.444*** (5.12) 1.82*** (3.96)

0.025 (1.16) 14.718*** (15.43) 0.609 (0.19) 0.516 (0.27) 0.061 (0.19) 0.33 (1.43) 1.323 (1.01) 2.443 (1.42) 522.03 (0.83) 0.732 (1.2)

630 472.633

630 472.633

630 472.633

0.051** (2.25) 0.498 (1.03) 48.338*** (6.35) 47.723*** (6.2) 0.38* (1.79) 0.002 (0.02) 0.628 (0.72) 0.193 (0.15) 2187.585*** (4.65) 1.115** (2.39)

0.029 (1.12) 1.516*** (2.64) 52.1*** (6.82) 47.898*** (5.89) 0.254 (1) 0.17 (1.23) 0.459 (0.41) 1.654 (1.1) 3368.537*** (5.11) 1.829*** (3.88)

0.05 (-0.83) 14.4*** (11.8) 0.763 (0.29) 1.073 (0.53) 0.0007 (0) 0.379 (1.41) 1.723 (1.33) 2.357 (1.55) 535.092 (0.92) 0.686 (1.13)

630 472.584

630 472.584

630 472.584

*Significant at 10%; **significant at 5%; ***significant at 1%. This table shows the estimates for two Multinomial Logit models. The models use four alternative outcomes based on the variables DDIV and DREP and are estimated using samples of firm-years (fiscal years) belonging to firms with controlling shareholders. The sample period is 1999–2004. The data necessary to build the variables used in the Multinomial Logit models is obtained from Thomson Datastream, Worldscope, firms’ annual reports, Consob’s website, the database Factiva, and the World Wide Web. For a particular firm-year, the binary variable DDIV (DREP) is equal to one if the firm pays dividends (repurchases stock) in the fiscal year, otherwise it is set to zero. Observations with DDIV ¼ 0 and DREP ¼ 0 (i.e. with zero dividends and no stock repurchase activity) constitute the base outcome of the models. There are three alternative outcomes: The first is DDIV ¼ 1 and DREP ¼ 0, the second is DDIV ¼ 1 and DREP ¼ 1, and the third is DDIV ¼ 0 and DREP ¼ 1. ORIGHTS is the fraction of the firm’s cash flow rights held by the controlling shareholder (i.e. the shareholder with the largest control stake above 20%). WEDGE is the difference between the fraction of the firm’s control rights possessed by the controlling shareholder and ORIGHTS. The binary variable FAMILY is set to one if the firm is controlled by a family. These ownership variables are constructed based on the latest available information before the start of the fiscal year. For a particular firm-year, the variables OI, NOI, MB, TA, and LEV are computed using information as of the end of the previous fiscal year. TA is the natural logarithm of the book value of total assets. OI and NOI are operating and non-operating income respectively scaled by total assets. MB is the sum of market capitalization and total debt divided by total assets. LEV is the ratio between total debt and the sum of market capitalization and total debt. TAX is the ratio between the after-tax (after both corporate and personal taxes) value of beforetax earnings if they are distributed by the firm through dividend payments and the after-tax value of the same earnings if they are used to repurchase stock (or if they are retained). For a particular firm-year, the value of TAX used is that relevant to the shareholder who has a direct stake in the firm and is part of the control chain used by the majority shareholder to control it. Four different scenarios are considered to compute TAX. In this table, only results for the first scenario are reported. The Appendix provides detailed information on the construction of TAX for the four alternative scenarios. VAR is the sample variance of the daily log-returns in the previous fiscal year. If, in the current fiscal year, there are ongoing stock option and/or stock grant plans for managers and directors, OPT is set to one; otherwise, it is equal to zero. Both year and industry dummies are included among the independent variables; however, results for these dummies are not reported. For each independent variable, the table shows the coefficient estimate and the z-statistic robust to heteroscedasticity and within-firm correlation (in parentheses). The table also reports the number of observations and the value of the log-likelihood function.

A. De Cesari / The British Accounting Review 44 (2012) 207–220

215

Unfortunately, doubts are cast on this conclusion if the other two alternative outcomes are considered. For these outcomes, the coefficients on ORIGHTS and WEDGE are not statistically different from zero, indicating that these ownership variables do not affect payout probabilities. In un-tabulated analyses, Logit models with binary regressands are used to study the relationships between the independent variables and either the probability of a payout (DPAYOUT), the probability of a dividend payout (DDIV) or the probability of a repurchase payout (DREP). On the whole, the results of the Logit models are similar to the multinomial findings. It is worth nothing that in Table 3 the signs of the coefficients on both ownership variables are different between the first (DDIV ¼ 1 and DREP ¼ 0) and the third (DDIV ¼ 0 and DREP ¼ 1) alternative outcome. This finding seems to suggest that among firms with positive payout the choice between dividends and repurchases is likely to be determined by ORIGHTS and WEDGE. In particular, an increase (a decrease) in ORIGHTS (WEDGE) seems to lead firms to prefer repurchases over dividends. In order to formally test this assertion, in un-tabulated analyses a new Multinomial Logit model is estimated with the same four outcomes as those in Table 3 but with the following base outcome: Zero dividend payments and positive repurchase payout (DDIV ¼ 0 and DREP ¼ 1). The slopes of this new model are related to the coefficients reported in Table 3. In particular, the coefficient on ORIGHTS is 0.047 (0.022  0.025) for the alternative outcome with positive dividend payments and no repurchases (DDIV ¼ 1 and DREP ¼ 0).11 The coefficient on this variable is 0.037 (0.012  0.025) if the other alternative outcome with positive payouts (DDIV ¼ 1 and DREP ¼ 1) is chosen. With respect to the same alternative outcomes, the coefficients on WEDGE are 0.101 (0.051  (0.05)) and 0.079 (0.029  (0.05)). Wald tests are run to investigate the statistical significance of the four coefficients above. With the exception of the second coefficient on WEDGE that is marginally insignificant (the p-value is equal to 0.168), the other three slopes are significantly different from zero at a significance level of either 5% or 10%. The estimates from the new multinomial model imply that when ORIGHTS increases by one unit the ratio between the probability of paying dividends without making repurchases (DDIV ¼ 1 and DREP ¼ 0) and the probability of just repurchasing shares (DDIV ¼ 0 and DREP ¼ 1) declines by 4.6% (e0.047  1). If WEDGE goes up by one, the same ratio increases by 10.6% (e0.101  1). These results support the notion that in firms in which controlling shareholders are more likely to expropriate minority shareholders (i.e. firms with small values of ORIGHTS and large values of WEDGE) dividends are preferred over repurchases when paying out cash. 4.2.2. Payout yields and the share of dividends in total payout Table 4 reports the estimates for the coefficients, the t-statistics (robust to heteroscedasticity and within-firm correlation), and the marginal effects of the final set of regressions. The dependent variables are PAYOUT, DIV, REP, and %DIV, and the main independent variables are ORIGHTS and WEDGE. Since the dependent variables are censored either at zero or at zero and 100, a Tobit estimation method is used. Findings for the Tobit regressions are reported in Panels A and B of Table 4. From these regressions, it can be concluded that both ORIGHTS and WEDGE are not statistically significant predictors of PAYOUT. What this finding suggests is that the magnitude of a controlled firm’s payout is not dependent on either the size of the cash flow stake of the firm’s controlling shareholder or the wedge between the controlling shareholder’s control and cash flow rights. Coefficients on ORIGHTS and WEDGE are insignificant in DIV specifications too, whereas the two regressors appear to be significantly related to REP. In particular, the magnitude of repurchase payout is positively associated with ORIGHTS and negatively related to WEDGE. A crucial finding in Table 4 is that the fraction of dividends in total payout (%DIV) is a negative function of the controlling shareholder’s cash flow rights (ORIGHTS) and a positive function of the difference between the control and the cash flow rights of the controlling shareholder (WEDGE). Considering the reported marginal effects, it can be stated that when ORIGHTS (WEDGE) increases by one, %DIV changes by 0.198 (0.423). 4.2.3. Summary and discussion of the main multivariate findings Like the univariate findings, the multivariate results are also consistent with LLSV’s (2000) substitute model. These results provide overall support for the notion that weakly governed firms use payout policy to mitigate corporate governance conflicts. Two key results can be highlighted. First, the odds of paying dividends rather than repurchasing own stock are negatively related to the cash flow stake of the firm’s controlling shareholder and positively associated with the wedge between the control and the cash flow rights of the controlling shareholder. Second, the weight of a firm’s dividend payout in total payout (dividends plus repurchases) decreases when the size of the cash flow rights of the firm’s controlling shareholder increases and when the difference between the controlling shareholder’s control rights and cash flow rights decreases. Agency costs are reduced mainly by increasing the fraction of dividend payments in total payout rather than by paying out more. The results of Table 3 show that firms with more serious agency problems are more likely to pay cash dividends without

11 In the original Multinomial Logit model of Table 3, 0.022 is the change in the logarithm of the ratio between the probability of the first alternative outcome (DDIV ¼ 1 and DREP ¼ 0) and the probability of the base outcome (DDIV ¼ 0 and DREP ¼ 0) caused by a unit increase in ORIGHTS. On the other hand, the logarithm of the ratio Pr(DDIV ¼ 0 and DREP ¼ 1)/Pr(DDIV ¼ 0 and DREP ¼ 0) increases by 0.025 if ORIGHTS changes by one. Through simple algebraic manipulations it can be shown that a unit change in ORIGHTS causes a variation in the logarithm of Pr(DDIV ¼ 1 and DREP ¼ 0)/Pr(DDIV ¼ 0 and DREP ¼ 1) that is equal to 0.022  0.025.

216

A. De Cesari / The British Accounting Review 44 (2012) 207–220

Table 4 Tobit models: The controlling shareholder’s cash flow rights and the wedge between the controlling shareholder’s control and cash flow rights. Dependent variables:

Panel A: Cash flow rights Independent variables: ORIGHTS

FAMILY

OI

NOI

MB

TA

LEV

TAX

VAR

OPT

Observations Log-likelihood value Panel B: Wedge Independent variables: WEDGE

FAMILY

OI

NOI

MB

TA

LEV

TAX

VAR

OPT

PAYOUT

DIV

REP

%DIV

0.003 (0.27) [0.002] 1.65** (2.41) [1.159] 24.82*** (5.39) [16.267] 22.264*** (4.57) [14.592] 0.478 (1.59) [0.313] 0.139 (0.77) [0.091] 1.425 (1.03) [0.934] 1.842 (1.37) [1.208] 3035.548*** (4.22) [1989.459] 1.015* (1.96) [0.692]

0.007 (0.74) [0.004] 2.029*** (2.93) [1.338] 29.122*** (5.27) [17.327] 26.994*** (4.66) [16.061] 0.424 (1.4) [0.252] 0.219 (1.25) [0.13] 0.959 (0.74) [0.571] 2.357* (1.78) [1.403] 3049.076*** (3.75) [1814.172] 0.568 (1.11) [0.347]

0.018** (2.3) [0.004] 2.023*** (3.6) [0.307] 6.591* (1.88) [1.413] 0.885 (0.31) [0.19] 0.233 (1.24) [0.05] 0.046 (0.4) [0.01] 0.434 (0.45) [0.093] 1.812* (1.67) [0.388] 1398.518** (2.53) [299.838] 1.158*** (2.64) [0.3]

0.593*** (2.7) [0.198] 54.689*** (3.9) [13.515] 379.094*** (2.66) [126.666] 407.172*** (2.63) [136.048] 0.185 (0.04) [0.062] 1.965 (0.67) [0.657] 10.604 (0.46) [3.543] 54.287** (2.13) [18.139] 11,683.22 (0.82) [3903.69] 21.164** (2.44) [7.7]

630 1385.412

630 1298.68

630 580.318

478 1040.214

0.0006 (0.03) [0.0004] 1.629** (2.34) [1.143] 24.904*** (5.44) [16.318] 22.339*** (4.59) [14.638] 0.478 (1.58) [0.313] 0.134 (0.75) [0.088] 1.395 (1) [0.914] 1.787 (1.24) [1.171] 3045.034*** (4.27) [1995.232] 0.998* (1.91) [0.68]

0.014 (0.86) [0.009] 2.021*** (2.84) [1.331] 28.97*** (5.29) [17.224] 26.858*** (4.66) [15.968] 0.406 (1.35) [0.241] 0.221 (1.28) [0.131] 0.967 (0.75) [0.575] 2.225 (1.62) [1.323] 3065.321*** (3.8) [1822.486] 0.581 (1.12) [0.355]

0.034** (2.19) [0.006] 2.175*** (3.94) [0.281] 7.043* (1.89) [1.334] 2.06 (0.63) [0.39] 0.28 (1.38) [0.053] 0.051 (0.42) [0.01] 0.359 (0.35) [0.068] 1.646 (1.38) [0.312] 1399.43** (2.58) [265.086] 1.098** (2.55) [0.251]

1.277*** (3.18) [0.423] 55.337*** (4.11) [13.486] 381.131*** (2.64) [126.26] 403.643*** (2.68) [133.717] 1.125 (0.28) [0.373] 2.337 (0.8) [0.774] 9.867 (0.43) [3.269] 42.244* (1.7) [13.994] 11,161.17 (0.79) [3697.427] 20.143** (2.27) [7.243]

A. De Cesari / The British Accounting Review 44 (2012) 207–220

217

Table 4 (continued ) Dependent variables:

Observations Log-likelihood value

PAYOUT

DIV

REP

%DIV

630 1385.456

630 1298.609

630 580.318

478 1039.477

*Significant at 10%; **significant at 5%; ***significant at 1%. This table reports the estimates for eight Tobit regressions (four in Panel A and four in Panel B) of the dependent variables PAYOUT, DIV, REP, and %DIV on several independent variables. The regressions are run on samples of firm-years (fiscal years) belonging to firms with controlling shareholders. The sample period is 1999–2004. The data necessary to build the variables is obtained from Thomson Datastream, Worldscope, firms’ annual reports, Consob’s website, the database Factiva, and the World Wide Web. For a particular firm-year, DIV (REP) is the ratio between the cash distributed by the firm in the year through dividend payments (repurchases) multiplied by 100 and the firm’s market capitalization at the end of the previous year. PAYOUT is the sum between DIV and REP, and %DIV is computed dividing the product between DIV and 100 by PAYOUT. ORIGHTS is the fraction of the firm’s cash flow rights held by the controlling shareholder (i.e. the shareholder with the largest control stake above 20%). WEDGE is the difference between the fraction of the firm’s control rights possessed by the controlling shareholder and ORIGHTS. The binary variable FAMILY is set to one if the firm is controlled by a family. These ownership variables are constructed based on the latest available information before the start of the fiscal year. For a particular firm-year, the variables OI, NOI, MB, TA, and LEV are computed using information as of the end of the previous fiscal year. TA is the natural logarithm of the book value of total assets. OI and NOI are operating and non-operating income respectively scaled by total assets. MB is the sum of market capitalization and total debt divided by total assets. LEV is the ratio between total debt and the sum of market capitalization and total debt. TAX is the ratio between the after-tax (after both corporate and personal taxes) value of before-tax earnings if they are distributed by the firm through dividend payments and the after-tax value of the same earnings if they are used to repurchase stock (or if they are retained). For a particular firm-year, the value of TAX used is that relevant to the shareholder who has a direct stake in the firm and is part of the control chain used by the majority shareholder to control it. Four different scenarios are considered to compute TAX. In this table, only results for the first scenario are reported. The Appendix provides detailed information on the construction of TAX for the four alternative scenarios. VAR is the sample variance of the daily log-returns in the previous fiscal year. If, in the current fiscal year, there are ongoing stock option and/or stock grant plans for managers and directors, OPT is set to one; otherwise, it is equal to zero. Both year and industry dummies are included among the independent variables; however, results for these dummies are not reported. For each independent variable, the table shows the coefficient estimate, the t-statistic robust to heteroscedasticity and within-firm correlation (in parentheses), and the marginal effect (in brackets). When the independent variable is a dummy, the marginal effect is for a discrete change in the variable from zero to one. The table also reports the number of observations and the value of the log-likelihood function for every regression.

making any repurchases rather than following a zero-payout policy. Further evidence consistent with this result cannot be reported when repurchase payout probabilities (in Table 3) and/or payout levels (in Table 4) are considered. Firms seem unwilling to boost overall payout probabilities and levels to mitigate agency costs. A possible explanation for this finding is that, on average, the benefits of increased payouts in terms of decreased agency conflicts are smaller than their costs. For instance, increases in payouts can be costly because higher cash distributions may jeopardize firms’ ability to invest in positive net present value projects. Furthermore, the equity reductions caused by payouts can result in sub-optimal levels of leverage. 5. Robustness checks A battery of tests is run to investigate the robustness of the multivariate findings presented in Section 4.2. Findings are not significantly qualitatively different if a) the regressions are estimated by including both ORIGHTS and WEDGE as regressors; b) the dependent variables PAYOUT, DIV, and REP are constructed by using net income rather than market value as deflator; c) it is assumed that companies repurchase stock at the lowest price during the year to estimate the annual amount of cash used to make repurchases; d) the following two alternative measures of the wedge between a controlling shareholder’s control and cash flow rights are adopted: Firstly, the ratio between cash flow rights and control rights; secondly, a binary variable that is equal to one when WEDGE is positive; e) it is required that to control a firm an investor needs at least either 35% or 50% of the firm’s voting rights; f) firms from the utilities industry are excluded from my sample; g) all of the continuous variables are winsorised at 2.5% and 97.5%. The Multinomial Logit estimates of Table 3 are consistent as long as the independence from irrelevant alternatives (IIA) assumption holds. Hausman–McFadden tests are run to assess the validity of the IIA assumption. The tests confirm the validity of this assumption in the dataset. 6. Conclusions The focus of this study is on the payout policy of firms controlled by shareholders with large control stakes. In controlled firms, the main agency conflict is that between controlling shareholders and minority shareholders. The conflict arises because controlling shareholders have an incentive to engage in self-dealing activities. This study investigates the relation between payout policy and the strength of the agency conflicts between controlling and minority shareholders using a sample of Italian firms. The main findings of the research are consistent with the substitute model of payout (John & Knyazeva, 2006; LLSV, 2000) and support the notion that a firm’s payout policy is designed to reduce its agency problems. Even though these problems do not to lead Italian companies to pay out more, firms in which minority shareholders are more likely to be expropriated by controlling shareholders tend to prefer dividends over repurchases more

218

A. De Cesari / The British Accounting Review 44 (2012) 207–220

and have larger shares of dividends in total payout than counterparts with stronger governance set-ups. In the study, firms with weak governance and severe agency problems are defined as those having controlling shareholders that own small fractions of their cash flow rights and posses control rights that significantly exceed cash flow rights. While most previous empirical research has attempted to test the validity of agency models of payout by studying only dividend payouts (e.g., Faccio et al., 2001; Gugler & Yurtoglu, 2003; LLSV, 2000), the findings of this paper highlight the importance of the dividend–repurchase trade-off. They also imply that researchers should jointly consider all the different types of cash distributions when studying the relationship between agency costs and payout policy. Furthermore, this study has important policy implications in that it documents the existence of attempts by controlled firms to reduce agency conflicts through payouts. Even though this evidence cannot completely allay concerns regarding the potential expropriation of minority shareholders, the results of this study are certainly of great importance to rule-setting bodies in Italy and beyond.

Acknowledgements I thank two anonymous referees, Henk von Eije, Ronald Giammarino, Nathan Joseph, Usha Mittoo, James Steeley, and seminar participants at Aston Business School, the 2010 World Finance Conference in Viana do Castelo, the 2010 European Financial Management Association annual meeting in Aarhus, and the 2010 Financial Management Association annual meeting in New York for helpful comments. However, the author is solely responsible for any remaining errors in the paper. Appendix12 The taxation of the following three types of shareholders is considered: Individual investors with large shareholdings, corporations, and government bodies (e.g., city councils) that are tax-exempt. Further, as baseline scenario, it is assumed that a firm holds V100 of before-tax earnings in the generic year t  1. The firm pays corporate taxes on these earnings in year t  1 and then can either use all the after-tax earnings to pay cash dividends to a shareholder in year t or to purchase own stock from the shareholder in the same year. For each type of shareholder, TAX is the ratio between the after-tax cash the shareholder receives if earnings are distributed through dividend payments and the after-tax cash the shareholder obtains if the earnings are disbursed through repurchases.13 Between 1999 and 2003, dividend taxes were based on an imputation system. A key characteristic of this system was the tax credit (scred,t  1), which was equal to the ratio between the corporate tax rate of the company that distributed dividends (scorp,t  1) and one minus the same rate (1  scorp,t  1). Assuming that in the generic year t  1 the corporate tax rate was 37%, then the after-tax earnings in the same year and the dividend payments in year t would have been V63, and the tax credit would have been V37 (63*scred,t  1). If the recipient of the dividends was an individual with a marginal personal income tax rate (sdiv,t) of 45% in year t, the recipient would have paid V8 in personal taxes. This sum is the product between the personal tax rate and the dividend payment plus the tax credit (45%*(63 þ 37) ¼ 45), minus the tax credit (45  37 ¼ 8). Overall, after deducting both corporate and personal taxes, the individual would have received V55, which was exactly the initial beforetax earnings times one minus their marginal personal tax rate (100*(1  sdiv,t)). In other words, the tax credit system ensured that individuals paid taxes on the before-tax earnings produced by their firms at rates that were exactly their personal tax rates. The same principle applied to shareholders that were corporations with the only difference that these shareholders paid corporate taxes and not personal taxes on the received cash dividends. In contrast, dividends received by tax-exempt government bodies did not carry any tax credits. In the period 1999–2003, an individual paid capital gain taxes at a rate of 27% (scap,t). If the firm distributed all the after-tax earnings through repurchases and the corporate tax rate was 37% in year t  1, the individual would have received V45.99 (100*(1  scorp,t  1)*(1  scap,t)) net of any taxes.14 On the other hand, a shareholder that was a corporation was required to pay corporate taxes on its capital gains. Hence, this shareholder would have obtained 100*(1  scorp,t  1)*(1  scorp,t). Finally, government bodies were tax-exempt and were not liable for capital gain taxes on sales of shares. Starting from January 2004, the imputation system for dividend taxes was not in place anymore and cash dividends did not carry any tax credits. Recipients who were individual investors paid taxes at their marginal personal income tax rate on 40% of the received dividends. Assuming a corporate tax rate (scorp,t  1) of 37% and a personal tax rate (sdiv,t) of 45%, the individual investor would have obtained V51.66 (100*(1  scorp,t  1)*(1  0.4*sdiv,t)) net of taxes. When the recipient was a corporation, the formula would have been 100*(1  scorp,t  1)*(1  0.05*scorp,t) because corporations were liable to pay taxes at the corporate tax rate only on 5% of the received dividends. No additional tax obligations arose for government bodies upon receipt of cash dividends.

12 The information on dividend and capital gain taxes that is used in this Appendix is primarily collected from the Italian publication Memento Pratico Fiscale for the years 1998–2004. 13 Repurchases are not the only alternative to cash dividends. Lower dividends in year t  1 lead to higher capital gains on share sales executed by shareholders in year t even if the buyer of the shares is not a repurchasing firm. In any case, the analyses are not dependent on the exact identity of the buyer of the shares sold by shareholders. 14 This formula is based on the assumption that the shareholder pays capital gain taxes on 100% of the cash received. This assumption is likely to be unrealistic as explained in a subsequent section of this Appendix.

A. De Cesari / The British Accounting Review 44 (2012) 207–220

219

In 2004, individual investors were obliged to pay capital gain taxes at their marginal personal income tax rate only on 40% of the capital gain. Hence, after both corporate and personal taxes, an individual investor would have received 100*(1  scorp,t  1)*(1  0.4*scap,t). Notice that in 2004 scap,t and sdiv,t were exactly the same. For shareholders that were corporations there were no substantial changes in 2004. Out of before-tax earnings of V100 they would have received 100*(1  scorp,t  1)*(1  scorp,t). As usual, government bodies were tax-exempt and did not pay any taxes on capital gains. Table A1 includes four panels that report the values of the variable TAX for four different scenarios. Individual investors with large shareholdings in listed firms are likely to be wealthy individuals. Hence, the top marginal personal income tax rate (sdiv,t) is considered in all four scenarios. In Panel A, the scenario is based on two assumptions: scorp,t is the maximum corporate tax rate and capital gain taxes are paid on 100% of the cash a shareholder receives when selling shares. This latter assumption, which is clearly unrealistic, is amended in Panel B where, following LLSV (2000), it is assumed that the effective tax rate on capital gains is 25% of the nominal tax rate. Panels C and D reflect the circumstance that in the period 1999–2003 a Dual Income Tax (DIT) system was in place that potentially allowed corporations to pay taxes on all or part of their taxable earnings at a rate that was lower than that assumed in Panels A and B. Owing to the DIT, a firm’s average corporate tax rate could range from the top to the lower rate. Since a firm’s eligibility to pay taxes at the lower rate was dependent on complex firm-specific conditions, it is not possible to determine the effective average tax rate for each corporation in the sample. Hence, in Panels C and D the assumption is that the corporate tax rate is the minimum DIT rate. The difference between these two panels is that in Panel C the effective capital gain rate is assumed equal to the nominal rate whereas in Panel D the effective rate is 25% of the nominal rate.

Table A1. Values of the dividend tax-preference variable TAX. Year (t)

scorp,t

sdiv,t

scap,t

Individual investors: TAX

Corporations: TAX

Government bodies: TAX

Panel A: Scenario based on top corporate tax rates and capital gain taxes paid on 100% of cash distributions 1998 37% – – – – 1999 37% 45.5% 27% 1.19 1.59 2000 37% 45.5% 27% 1.19 1.59 2001 36% 45% 27% 1.20 1.59 2002 36% 45% 27% 1.18 1.56 2003 34% 45% 27% 1.18 1.56 2004 33% 45% 45% 1 1.47

– 1 1 1 1 1 1

Panel B: Scenario based on top corporate tax rates and capital gain taxes paid on 25% of cash distributions 1998 37% – – – – 1999 37% 45.5% 27% 0.93 1.10 2000 37% 45.5% 27% 0.93 1.10 2001 36% 45% 27% 0.94 1.12 2002 36% 45% 27% 0.92 1.10 2003 34% 45% 27% 0.92 1.13 2004 33% 45% 45% 0.86 1.07

– 1 1 1 1 1 1

Panel C: Scenario based on minimum corporate tax rates and capital gain taxes paid on 100% of cash distributions 1998 27% – – – – 1999 27% 45.5% 27% 1.02 1.37 2000 27% 45.5% 27% 1.02 1.37 2001 19% 45% 27% 1.03 1.37 2002 19% 45% 27% 0.93 1.23 2003 19% 45% 27% 0.93 1.23 2004 33% 45% 45% 1 1.47

– 1 1 1 1 1 1

Panel D: Scenario based on minimum corporate tax rates and capital gain taxes paid on 25% of cash distributions 1998 27% – – – – 1999 27% 45.5% 27% 0.80 1.07 2000 27% 45.5% 27% 0.80 1.07 2001 19% 45% 27% 0.81 1.16 2002 19% 45% 27% 0.73 1.05 2003 19% 45% 27% 0.73 1.05 2004 33% 45% 45% 0.86 1.07

– 1 1 1 1 1 1

This table includes values of the variable TAX. The variable reflects the tax preference of dividends when compared to cash distributed through repurchases (or retained earnings) for three classes of shareholders (individual investors, corporations, and government bodies) and four different scenarios. scorp,t, sdiv,t, and scap,t are, for fiscal year t, the corporate tax rate, the tax rate on cash dividends paid by individual investors, and the tax rate on capital gains paid by individual investors respectively. Individuals are assumed to hold comparatively large shareholdings, which carry at least 2% of the voting rights, and to be subject to the top marginal personal tax rate. Government bodies are tax-exempt. The variable TAX is the ratio between the after-tax (after both corporate and personal taxes) value of before-tax earnings if they are distributed by the firm through dividend payments and the after-tax value of the same earnings if they are used to repurchase stock (or if they are retained). In the period 1999–2003, TAX is (1  sdiv,t)/((1  scorp,t  1)*(1  scap,t)) for individual investors, (1  scorp,t)/((1  scorp,t  1)*(1  scorp,t)) for corporations, and (1  scorp,t  1)/(1  scorp,t  1) for government bodies. In 2004, TAX is ((1  scorp,t  1) *(1  0.4sdiv,t))/((1  scorp,t  1)*(1  0.4scap,t)) for individual investors, ((1  scorp,t  1)*(1  0.05scorp,t  1))/((1  scorp,t  1)*(1  scorp,t)) for corporations, and (1  scorp,t  1)/(1  scorp,t  1) for government bodies. In Panels A and B, the top corporate tax rate is used whereas in Panels C and D for the period 1999–2003 it is assumed that corporate taxes are paid at the lower Dual Income Tax rate. In Panels A and C, it is assumed that the effective tax rate on capital gains is equal to the nominal tax rate. In contrast, in Panels B and D the effective rate is 25% of the nominal rate.

220

A. De Cesari / The British Accounting Review 44 (2012) 207–220

References Adjaoud, F., & Ben-Amar, W. (2010). Corporate governance and dividend policy: shareholders’ protection or expropriation. Journal of Business Finance and Accounting, 37, 648–667. Aharony, J., & Swary, I. (1980). Quarterly dividend and earnings announcements and stockholders’ returns: an empirical analysis. Journal of Finance, 35, 1–12. Bianchi, M., Bianco, M., & Enriques, L. (2001). Pyramidal groups and the separation between ownership and control in Italy. Available at SSRN. http://ssrn.com/ abstract¼293882. Brav, A., Graham, J. R., Harvey, C. R., & Michaely, R. (2005). Payout policy in the 21st century. Journal of Financial Economics, 77, 483–527. Chae, J., Kim, S., & Lee, E. J. (2009). How corporate governance affects payout policy under agency problems and external financing constraints. Journal of Banking and Finance, 33, 2093–2101. Chay, J. B., & Suh, J. (2009). Payout policy and cash-flow uncertainty. Journal of Financial Economics, 93, 88–107. Claessens, S., Djankov, S., Fan, J. P. H., & Lang, L. H. P. (2002). Disentangling the incentive and entrenchment effects of large shareholdings. Journal of Finance, 57, 2741–2771. Dittmar, A. K. (2000). Why do firms repurchase stock? Journal of Business, 73, 331–355. Faccio, M., & Lang, L. H. P. (2002). The ultimate ownership of Western European corporations. Journal of Financial Economics, 65, 365–395. Faccio, M., Lang, L. H. P., & Young, L. (2001). Dividends and expropriation. American Economic Review, 91, 54–78. Fama, E. F., & French, K. R. (2001). Disappearing dividends: changing firm characteristics or lower propensity to pay? Journal of Financial Economics, 60, 3–43. Fenn, G. W., & Liang, N. (2001). Corporate payout policy and managerial stock incentives. Journal of Financial Economics, 60, 45–72. Gugler, K., & Yurtoglu, B. B. (2003). Corporate governance and dividend pay-out policy in Germany. European Economic Review, 47, 731–758. Hu, A., & Kumar, P. (2004). Managerial entrenchment and payout policy. Journal of Financial and Quantitative Analysis, 39, 759–790. Jagannathan, M., Stephens, C. P., & Weisbach, M. S. (2000). Financial flexibility and the choice between dividends and stock repurchases. Journal of Financial Economics, 57, 355–384. Jensen, G. R., Solberg, D. P., & Zorn, T. S. (1992). Simultaneous determination of insider ownership, debt, and dividend policies. Journal of Financial and Quantitative Analysis, 27, 247–263. Joh, S. W. (2003). Corporate governance and firm profitability: evidence from Korea before the economic crisis. Journal of Financial Economics, 68, 287–322. John, K., & Knyazeva, A. (2006). Payout policy, agency conflicts, and corporate governance. Available at SSRN. http://ssrn.com/abstract¼841064. Kahle, K. M. (2002). When a buyback isn’t a buyback: open market repurchases and employee options. Journal of Financial Economics, 63, 235–261. La Porta, R., Lopez-de-Silanes, F., & Shleifer, A. (1999). Corporate ownership around the world. Journal of Finance, 54, 471–517. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., & Vishny, R. W. (1998). Law and finance. Journal of Political Economy, 106, 1113–1155. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., & Vishny, R. W. (2000). Agency problems and dividend policy around the world. Journal of Finance, 55, 1–33. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., & Vishny, R. W. (2002). Investor protection and corporate valuation. Journal of Finance, 57, 1147–1170. Laeven, L., & Levine, R. (2008). Complex ownership structures and corporate valuations. Review of Financial Studies, 21, 579–604. Mancinelli, L., & Ozkan, A. (2006). Ownership structure and dividend policy: evidence from Italian firms. European Journal of Finance, 12, 265–282. Memento Pratico Fiscale. Ipsoa-Francis Lefebvre, years 1998–2004. Mitton, T. (2002). A cross-firm analysis of the impact of corporate governance on the East Asian financial crisis. Journal of Financial Economics, 64, 215–241. Nenova, T. (2003). The value of corporate voting rights of control: a cross-country analysis. Journal of Financial Economics, 68, 325–351. Noronha, G. M., Shome, D. K., & Morgan, G. E. (1996). The monitoring rationale for dividends and the interaction of capital structure and dividend decisions. Journal of Banking and Finance, 20, 439–454. Rozeff, M. (1982). Growth, beta, and agency costs as determinants of dividend payout ratios. Journal of Financial Research, 3, 249–259. Skinner, D. J. (2008). The evolving relation between earnings, dividends, and stock repurchases. Journal of Financial Economics, 87, 582–609. Stephens, C. P., & Weisbach, M. S. (1998). Actual share reacquisitions in open-market repurchase programs. Journal of Finance, 53, 313–333. Zingales, L. (1994). The value of the voting right: a study of the Milan Stock Exchange experience. Review of Financial Studies, 7, 125–148.