Taxation, Dividends, and Share Repurchases: Taking Evidence Global

2013, Journal of Financial and Quantitative Analysis

visibility

…

description

29 pages

link

1 file

Abstract

We compile a comprehensive international dividend and capital gains tax data set to study tax-based explanations of corporate payout for a panel of 6,035 firms from 25 countries for the period 1990–2008. We find robust evidence that the tax penalty on dividends versus capital gains corresponds closely with firms’ propensity to pay dividends and repurchase shares, and with the amount of dividends and shares repurchased. Our coefficient estimates suggest a smaller tax effect than reported in recent single-country, single-event studies. Instead, our results correspond more closely with historic long-term estimates of the elasticity of dividends.

Figures (13)

Table 1 presents the number of firms and firm-year observations for each country in our sample. The sample consists of 6,035 firms in 25 countries for 1990-2008. We start our analysis in 1990 for two reasons. First, WorldScope does not provide comprehensive coverage of individual data items for non-U.S. firms prior to 1990. An earlier start may thus have biased our results for earlier subperiods toward evidence from North America. Second, 1990 is a historically logical year to begin. With their transformation into capitalist, democratic systems, many former communist countries began to address dividends and capital gains in tax legislation in 1990.

In Panel C of Table 3, we provide univariate tests of differences in payout between observations where the dividend tax penalty is negative, that is, divi- dends are taxed preferentially, and observations where the dividend tax penalty is nonnegative. The nonparametric tests provide a Ist indication that the fraction of dividend-paying firms and the average dividend yield are significantly higher in country-years with negative dividend tax penalties. In contrast, the percentage of share-repurchasing firms and the average repurchase yield are significantly lower in country-years with a preferential tax treatment of dividends. !”

Table 5 is based upon the regression coefficients reported in Table 4. It presents the effects of a 1-standard-deviation increase in the dividend tax penalty upon firm payout choices. Column 1 presents the dependent variable (see Table 3 for a description of the dependent variables). Column 2 replicates the coefficient estimates and significance levels. Column 3 features the effects of a 1-standard-deviation increase of the independent variable upon dividend policy (expressed in percentage points). Column 4 summarizes the sample averages of the dependent variables. Finally, column 5 presents the ratio of the effect in column 3 relative to the sample average of the respective payout variable.

Table 6 reports coefficient estimates of the dividend tax penalty variable for various model specifications. Column 1 presents the dependent variables, which are defined as in Table 3. Model 1 gives coefficients from OLS regressions without firm- and country-level controls. In model 2, we present estimates from OLS regressions with firm-level controls as defined in Table 4. In model 3, regression coefficients for 5'V are based on OLS with country-level controls only. Model 4 uses cash flow (CF) instead of cash holdings in the firm-level controls. Model 5 features OLS regression coefficients, where we use capital expenditures (CapEx) instead of Tobin’s q to measure growth opportunities. Firm- and industry-year fixed effects are used throughout. Standard errors (shown in parentheses) allow for heteroskedasticity and are clustered at the firm and the country level. ** and * denote significance at the 1% and 5% levels, respectively. 6Note that there are some concerns about interpreting firm-level controls in a multi-country set- ting, as they might jointly reflect firm and country-level characteristics (see, e.g., Demirguc-Kunt and Maksimovic (1998)). Thus, some cross-country differences are observed as differences in the country means of EBITDA. We rerun all our regressions using firm-level variables that are demeaned for each country-year cell. Results are very similar.

Table 7 reports fixed effects (FE) panel regression results for corporate payout behavior, estimated over the panel of 6,035 firms in 25 countries for 1990-2008. Dependent and independent variables are defined as in Table 3. Corporate payout is measured 1 year after firm- and country-level controls. OE as is the dividend tax penalty for substantial shareholdings CLOSELY is the percentage of shares held by substantial shareholders. FLOAT is defined as 1 — CLOSELY. Firm- anc industry-year fixed effects are included in all regressions. Standard errors (shown in parentheses) allow for heteroskedas- ticity and are clustered at the firm and the country level. ** and * denote significance at the 1% and 5% levels, respectively. In Table 8, we address another central concern about effective tax rates. The statutory dividend and capital gains tax rates that we use to compute the dividend tax penalty may be an inaccurate representation of the “effective” tax rates faced by shareholders. If, for example, state capacity to collect taxes in a country is low and tax evasion widespread, the divergence between the statutory and the effective tax rates on dividends and capital gains is large. The larger this divergence, the more inaccurate is the measurement of our tax variables. Consequently, we opt for robustness checks in which we weight the dividend tax penalty with a measure of state capacity to collect taxes. A popular measure of state capacity in the literature is the Polity index (Polity IV Project (2010)). Our results from this additional check (Panel A of Table 8) are very similar to our baseline results. A final concern about our tax preference measure is that capital gains are 4 aq 7 qe 7 Pt aie eae * er ,

Table 8 reports fixed effects (FE) panel regression results for corporate payout behavior, estimated over the panel of 6,035 firms in 25 countries for 1990-2008, but it includes the effective dividend tax penalty (oe). In Panel A, we weight our dividend tax penalty with the state capacity to collect taxes. Panel B uses the effective capital gains tax rate, which is 14 of the statutory tax rate, to compute the dividend tax penalty. Dependent variables and baseline regression controls are as in Table 4. Corporate payout is measured 1 year after firm- and country-level controls. Firm- and industry-year fixed effects are included in all regressions. Standard errors (shown in parentheses) allow for heteroskedasticity and are clustered at the firm and the country level. ** and * denote significance at the 1% and 5% levels, respectively. rate may be significantly lower. To control for this effect, we follow prior analyses of effective capital gains taxation and assume that the effective capital gains tax rate is equivalent to / of the statutory tax rate (see, e.g., La Porta et al. (2000), Poterba (1987)). Panel B of Table 8 reports baseline regression results with the capital gains tax rate adjustment taken into consideration. The economic magni- tude of the tax effect increases when we employ effective capital gains tax rates.

Panel B. Controlling for Endogeneity of Dividend Tax Penalty Panel A. Determinants of Dividend Tax Penalty Table 9 reports results for corporate payout behavior from a 2-stage regression, estimated over the _panel of 6,035 firms in 25 countries for 1990-2008. Panel A reports results of the cross-sectional OLS regression using 5°") as the dependent variable. The residuals of this regression (RESIDUALS_OF_5°'V) are used in the 2nd-stage regression (Panel B). Dependent and independent variables are defined as in Table 4. COMMON is a dummy variable equal to 1 for common law countries, and 0 otherwise. Corporate payout is measured 1 year after firm- and country-level controls. Firm- and industry-year fixed effects (FE) are included in all regressions. Standard errors (shown in parentheses) allow for heteroskedasticity and are clustered at the firm and the country level. ** and * denote significance at the 1% and 5% levels, respectively.

We also consider another feature of the tax system more specifically: the corporate tax rate. Corporate taxes may be connected to payout taxes for many reasons, including government budget trade-offs and political preferences (i.e., pro-business). Corporate taxes might also affect the importance of internal re- sources for firms. To ensure that our results are not biased in either direction, we include the corporate tax rate as an additional control. The results are reported in Table 10. Coefficient estimates for our dividend tax penalty remain significant and very close in magnitude to the baseline regressions.

Table 10 reports fixed effects (FE) panel regression results for corporate payout behavior, estimated over the panel of 6,03% firms in 25 countries for 1990-2008, but it includes the additional control variable CORPORATE-TAX-RATE. Dependen variables and baseline regression controls are as in Table 4. Corporate payout is measured 1 year after firm- and country level controls. Firm- and industry-year fixed effects are included in all regressions. Standard errors (shown in parentheses allow for heteroskedasticity and are clustered at the firm and the country level. ** and * denote significance at the 1% anc 5% levels, respectively.

Table 11 reports fixed effects (FE) panel regression results for corporate payout behavior, estimated over the panel of 6,035 firms in 25 countries for 1990-2008, but it includes the additional control variable insider shareholdings (INSIDER), and its interaction with 6° V Corporate payout is measured 1 year after firm- and country-level controls. Firm- and industry-year fixed effects are included in all regressions. Standard errors (shown in parentheses) allow for heteroskedasticity and are clustered at the firm and the country level. ** and * denote significance at the 1% and 5% levels, respectively.

Table 12 reports fixed effects (FE) panel regression results for corporate payout behavior, estimated over the panel ot 6,035 firms in 25 countries for 1990-2008. DIVIDEND_INIT_100 (DIVIDEND_INIT_15) is an indicator variable equal to 10C if a nonpayer initiates dividend payments or if a dividend-paying firm increases the dividend volume by more than 100% (15%), and 0 otherwise. REPURCHASE_INIT_100 (REPURCHASE_INIT_15) is an indicator variable equal to 100 if a firm initiates share repurchases or if a firm that repurchased shares in the prior period increases the share repurchase volume by more than 100% (15%), and 0 otherwise. DIV-TO-PAYOUT-_RATIO is the fraction of total payout that is distributed as dividends. We set DIV-TO-PAYOUT-_RATIO to 0 if the firm does not distribute cash to shareholders. Independent variables are defined as in Table 4. Firm- and industry-year fixed effects are included in all regressions. Standard errors (shown in parentheses) allow for heteroskedasticity and are clustered at the firm and the country level. ** and * denote significance at the 1% and 5% levels, respectively.