Rebound Effect of Allowance for Corporate Equity on Debt Bias

We analyse if the Brazilian Allowance for Corporate Equity (ACE-type system) reduces the debt tax bias. Specifically, we study if the continuous treatment effect of interest on equity negatively affects the level of financial leverage. We find that the tax policy implemented is similar to the deductible cash dividends and not to an ACE. The empirical implication is that the interest on equity treatment increases the debt tax bias, producing a rebound effect to what is expected for this policy on the risk-taking behaviour and corporate capital structure. This rebound effect is homogeneous in firms with different financial constraints status. There are evidences that shareholders influence the cash distribution policy, adjusting the later to their own tax preferences. Therefore, there may be a "ACE clientele effect" induced by heterogeneity in tax preferences among shareholders.


Introduction
The tax deductibility of interest expenses has been considered to be an historical accident (Schoenmaker & Goodhart, 2010), with no legal or economic reasons for its existence as policy (Hemmelgarn & Nicodeme, 2010;Warren Jr., 1974). The tax discrimination between debt and equity has been identified as a policy leading to harmful effects on the legal distinction between the two types of financing (Warren Jr., 1974), on how the firms choose their financing sources and on how much risk they bear (Hemmelgarn & Nicodeme, 2010). To De Mooij (2011), the deductibility of debt has created a debt tax bias favouring the excessive use of leverage, which has been associated with a higher vulnerability of firms in the presence of credit crunches. When there is rationing in the supply of credit or equity, a tax distortion may favour firms with better 3 The present study contributes to the existing literature on tax policy, on capital structure and on cash distribution policies. At the tax level it contributes to a better understanding on how different ACE systems affect the effectiveness on reducing the debt tax bias, subject to institutional factors specific of emerging economies. Mainly, it clarifies that the Brazilian tax system cannot be considered an ACE-type system experiment. To the literature on capital structure and cash distribution, it allows for a better understanding on how the tax factor affects both policies simultaneously. This research also contributes to better understand how shareholders may influence the cash distribution policy, adjusting the later to their own tax preferences. The Brazilian case is the only one known that allows us to investigate the effect of dosage of the ACE policy. This study is the first to explore this opportunity and adequately address confounding factors.
This paper is structured as follows: in section 2, we present the main characteristics of the ACE systems and the specificities of the Brazilian ACE-type system, and also develop our working hypotheses; in section 3, we outline the empirical strategy; in section 4, we report the results and proceed with the discussion; and, in section 5, we make the final remarks.

Brazilian ACE-Type System
ACE is a variant of Allowance for Corporate Capital (ACC) first proposed by (Boadway & Bruce, 1984). The main idea behind ACC is to abolish the deductibility of the interest expenses and replace it by a notional risk-free return on the book value of the firm's capital. The ACE-type system, first introduced by the Capital Taxes Committee of the Institute for Fiscal Studies (Devereux & Freeman, 1984;IFS Capital Taxes Group, 1991) differs from the ACC by still allowing the deductibility of the interest expenses, while adding the tax deductibility against corporate profits of a notional risk-free rate on equity. Both the ACC and the ACE systems assume the notional return as being risk free, since the tax advantage from its deductibility is taken for granted (Bond & Devereux, 1995).
The Comprehensive Business Income Tax (CBIT) has been considered as an alternative to ACE system. CBIT system eliminates the deductibility of the interest expenses, and thus eliminates the debt tax bias, it leads to negative collateral effects by increasing the firm's cost of capital and by exacerbating the distortion on marginal investments. De Mooij & Devereux (2010) consider the ACE system preferable to the CBIT system because of the following four favourable properties: first and foremost, it allows for the neutrality between debt and equity; second, it neutralizes the tax effect on marginal investment decisions; third, it avoids the distortion on investment induced by differences between economic and tax depreciation; and fourth, it lowers the firm's cost of capital.

4
De Mooij & Devereux (2010) use a general equilibrium model to simulate the implementation of both the ACE and CBIT systems in the European Community. Given the need to adjust the statutory corporate tax in order to compensate for the reduction on tax revenue on the ACE system, there is an increase in the effective average tax rate (EATR). A higher EATR decreases the internal funds and intensifies the international profit shifting and discrete location, which affect corporate investment decisions. These collateral effects of the ACE system make the CBIT system more efficient and welfare improving, at the country level, and also in the presence of capital-market imperfections (Hubbard, 1998). Meanwhile, the ACE system becomes more efficient than the CBIT system when implemented in a jointly manner through all the European Community. In addition to the experiences that have been made regarding the ACE system, this is also the preferred system by some of the proponents of fiscal policies to reduce the debt bias (IFS Capital Taxes Group, 1991;FSB Financial Stability Board, 2015).
Simulation models evaluating individual fiscal reforms using variants of the ACE system in Switzerland (Keuschnigg & Dietz, 2007) and Germany (Radulescu & Stimmelmayr, 2007) suggest a reduction on the cost of capital and an increase in the investment level. However, the results from real experiments implementing the ACE-type systems (e.g. in Austria, Belgium, Croatia and Italy) show no clear-cut evidence on the economic impact of these reforms (De Mooij & Devereux, 2010). This lack of evidence is mostly due to two factors: first, the implementation of the ACE system on these countries was part of a package of multiple reforms, making it hard to identify the impact of a specific reform; and second, the reforms did not include all the characteristics of the ACE system, thus being only an ACE-type system. The exception to both these factors was the implementation of the ACE system in Belgium, in 2006.
Brazil is the only emerging economy with an ACE system. The country implemented in 1996 a tax reform, Law 9.249/95 (Brasil, 1995), which has been classified as an ACE-type system (Klemm, 2007;IMF, 2009;De Mooij, 2012;ICC Commission on Taxation, 2012;Panier, González-Pérez & Villanueva, 2013). However, and although sharing common goals, the Brazilian system diverges in fundamental properties from the other implemented systems and from the original proposal by IFS Capital Taxes Group (1991).
The Brazilian ACE-type system is characterized by three distinctive elements. First, it only allows for the tax deductibility of the notional interest, denominated as interest on equity (IOE), when this is paid out to the shareholders. The IOE is deducted when considering the corporate income tax (CIT). Brazilian CIT is 34% and includes an income tax rate of 25%, Law nº 9. 249/1995249/ (Brasil, 1995, and a social contribution tax rate of 9%, Law nº 7. 689-88 (Brasil, 1988). To determine the IOE we have to consider the amount of equity (E) minus the revaluation Electronic copy available at: https://ssrn.com/abstract=2743192 reserves (RR), both from the previous fiscal year, i.e. (Et-1 -RRt-1). The notional risk-free rate used is the Long-Term Interest Rate (LTIR) 2 in t-1. Second, the system includes eligibility criteria that limits the level of notional interest accepted as tax deductible. Eligible firms are those that cumulatively have in t-1 positive equity and also fulfil one of the following criteria: having positive earnings before interest and taxes (EBIT), or having positive retained earnings (RE). The IOE accepted for deductibility must not exceed 50% of the EBIT or 50% of the RE, the highest.
Third, there is a tax rate of 15% on the interest on equity, which is retained by the paying firm. Additionally, the personal income tax rate on the interest on equity received varies in accordance to the nature of the beneficial shareholder. When the shareholder is a physical entity or the government, the retained tax rate of 15% is definitive. Institutional investors and pension funds are exempt from income tax, Law 9.532/1997 art. 28 and 33 (Brasil, 1997), and thus compensation is allowed. When the shareholders are firms, these are taxed through the real income system and may be compensated for the retained tax rate of 15%, but are subject to an income tax rate of 34% on the financial income from the interest on equity received (firms with an annual income lower than R$ 240.000 have a lower tax rate of 24%). Additionally, when the shareholders are firms, these must pay for social purposes 1.65% and 7.6% on the interest on equity received, corresponding to the contribution to the Programa de Integração Social (Social Integration Program) and Contribuição para Financiamento da Seguridade Social (Contribution to Social Security Financing), respectively.
Overall, the tax savings 3 for physical entities and government is 19% (34%-15%), to institutional investors and pension funds is 34% (34%-0%), e to legal entities it may be -9.25% (34%-15%-34%-9.25%) or 0.75% (34%-15%-24%-9.25%), depending on the income level of the firm. Besides, in the case where the shareholders are firms, the disincentive increases as one uses corporate control structures of pyramidal form, due to the double taxation (or more, depending of the control-enhancing mechanism level adopted). So, depending on the legal form of the shareholder, one may find significant differences in the tax incentives, and thus on the propensity to be willing to receive the interest on equity. Boulton, Braga-Alves & Shastri (2012) and Colombo & Terra (2014) find evidence on Brazilian firms that the legal form of the controlling shareholders affect not only the propensity to pay, but also the amount of interest on equity paid.
The authors point out the positive effect on the IOE paid when institutional investors, with a minimum stake in the firm of 5% of shares with voting rights, are present. 6 2.2 Hypotheses development Implementing an ACE system is expected to reduce the debt tax bias, which in turn should lead to a lower debt-to-equity ratio. That would happen because the firm would be using more internal or external equity instead of debt. In fact, Panier et al. (2013) find this expected result when analysing Belgium firms after the implementation of the ACE system in that country.
Meanwhile, the characteristic of the Brazilian ACE system, to only allow for the deductibility of the interest on equity paid out to shareholders, may have three important implications. First, firms need to cash out to obtain the deductibility, thus making the latter dependent on existing internal funds and directly competing with growth opportunities (Fazzari, Hubbard & Petersen, 1988). In frictionless capital markets firms would have the necessary conditions to access the equity market and adjust their capital structure when needed (Modigliani & Miller, 1958). However, in the presence of markets imperfections, this adjustment may be slower or even impossible to follow (Almeida & Campello, 2010).
Second, there is an effect derived from accounting mechanics on the debt-to-equity ratio. This is due to the fact that even if the deductibility of interest on equity leads to a reduction on the debt tax bias, the cash out needed lowers the retained earnings and, consequently, lowers the equity level. Thus, the tax incentives to use equity instead of debt, in the presence of imperfect capital markets, may not be sufficient to compensate the reduction on internal equity.
Third, when internal funds become insufficient, firms may access the debt market to obtain the necessary funds to pay out interest on equity to shareholders. If this is the case, firms may benefit twice from the allowed deductibility, one from debt and another from interest on equity tax shield. Thus, we may observe a multiplying effect of the tax benefit. Additionally, the simulations performed by Almeida & Paes (2013) suggest that the IOE in Brazil, despite reducing the cost of new equity, is not sufficient to compensate the tax benefit of debt financing.
Considering the three stated implications from the specificities of the Brazilian ACE-type system, we expect the existence of interest on equity to intensify the debt tax bias, instead of reducing it. The first research hypothesis is thus the following:

H1:
The interest on equity positively affects the preference for debt financing instead of equity.
Firms may become financial constrained in the presence of capital markets imperfections (Hubbard, 1998;Almeida, Campello & Weisbach, 2004;Crisóstomo, López-Iturriaga & Vallelado González, 2014). Financially constrained firms may have difficulty in accessing the debt market, making internal and external funds not fully replaceable (Almeida & Campello, 2010). Since internal funds are relatively less costly (Myers & Majluf, 1984;Fazzari et al., 1988), 7 this may lead to a preference for cash holdings instead of cash distribution to shareholders (Vogt, 1994;Hubbard, 1998;Almeida et al., 2004;Arslan, Florackis & Ozkan, 2006). This behaviour is expected to be more severe in countries with weak investor protection, bank-oriented financial system and low level of development, similar to Brazil's (Khurana, Martin & Pereira, 2009).
Although the low use of tax benefits has been associated with conservative financial policies (Graham, 2000), recent evidence shows that when all tax shields and distress costs are accordingly considered, firms produce tax-efficient capital structures (Blouin, Core & Guay, 2010).
Additionally, simulations show that in Brazil the tax distortion is amplified when the access to less costly financing sources increases (Almeida & Paes, 2013), condition favourable to financially unconstrained firms.
Therefore, the effect of interest on equity may be different in firms subject to different levels of financial constraints, with different financing costs and accessibility to external funds.
The second research hypothesis is thus the following:

H2:
The effect of interest on equity on the preference for debt instead of equity is higher in financially unconstrained firms.

Data
The initial sample included all non-financial firms listed on São Paulo's Stock Exchange In order to reduce potential bias we applied the following restrictions: a) to exclude firms that have been involved in some sort of M&A, all firm-year observations with an assets annual growth or sales growth higher than 100% were eliminated (Acharya, Almeida & Campello, 2007); b) to minimize existing measurement errors on the market-to-book variable, all firm-year observations with a value higher than 10 were eliminated; c) all firm-year observations that did not fulfil the requirements for the tax deductibility of interest on equity were excluded. This 4 The Brazil, from 1994 to 2000, experienced a major inflation control program, Real Plan in 1994, and financial crises in emerging markets in Asia, Russia and a domestic currency crisis. Therefore, these confounding factors can distort the purpose of assessing ACE implanting in Brazil, which prevents a reliable ex-ante vs. ex-post analysis. 5 http://www.cvm.gov.br/ingl/indexing.asp. 6 http://www.bmfbovespa.com.br/en-us/markets/equities/companies/corporategovernance.aspx?Idioma=en-us.
8 includes firm-year observations with negative Equity, and that cumulatively have negative Earnings Before Interest and Taxes and negative Retained Earnings; d) all continuous variables were winsorized at the top and bottom 2.5% level to avoid the effect of potential outliers.

Empirical Strategy
Implementing the ACE-type system is the remedy used for the treatment of the debt tax bias. The first stage of the empirical strategy addresses the first hypothesis (H1) in which the payment of interest on equity is expected to intensify the debt tax bias, and where this effect increases with higher payment levels. In order to do so, we look at the causal effect of the treatment variable interest on equity on an outcome representing the debt tax bias. In this research the outcome variable is proxied by two measures of financial leverage: the Financial debt-toassets ratio and the financial debt-to-capital ratio. Although the financial debt-to-capital ratio is more suitable, because it only considers the two deductible financing sources (Welch, 2011), we keep the Financial debt-to-assets ratio to allow the comparability with previous studies (Klemm, 2007;Ness Junior & Zani, 2001;Zani, Leites, Macagnan, & Portal, 2014) 7 , where this variable was used for similar purposes.
The first test involves a model estimation (see Equation 1), using ordinary least squares (OLS), where the dependent variable is leverage and the main predictor is interest on equity (IOE).
As previously stated leverage is proxied by two different measures: the Financial debt-to-assets ratio and the financial debt-to-capital ratio. The variable IOE is defined as the ratio between interest on equity and total assets. Indicative of a reduction on the debt tax bias, we expect the coefficient of IOE to be negative and statistically significant. Following the existing literature on capital structure (Booth, Aivazian, & Demirguc-kunt, 2001;Frank & Goyal, 2009;Graham & Leary, 2011;Harris & Raviv, 1991) we include additional control variables in our model. The level of cash flow is measured by the ratio of cash flow-to-assets (Cash flow). CAPEX is the ratio of capital expenditures-to-assets while the level of cash holdings is measured as the ratio of cash holdings-to-assets (Cash holdings). We define firm size as the natural logarithm of assets (Size).
To proxy for investment opportunities we use the market-to-book ratio of assets (Market-tobook). To control for differences between industries, and following (Frank & Goyal, 2009), we include industry dummies according to the NAICS level 1 industrial classification system. Given the specific purpose of the present study, we include a control variable for the tax shield (Tax shield) in order to control for capital tax shields associated to the deductibility of interest expenses and interest on equity, but also to control for the non-capital tax shields, such as depreciation expenses and investment tax credit. Following Carvalhal & Leal, 2013 and Costa, Paz & Funchal 9 (2008), we use a dummy for ADR firms (Financial unconstraint) in order to control for the different access levels to the capital markets. All variables definitions are presented in Appendix 1.
Given the characteristics of the Brazilian ACE-type system presented in Section 2, we know that the interest on equity treatment is not random. In fact, firms must comply with a set of eligibility criteria, have available internal funds or access to the external funds market, and also take into consideration the tax preferences of their shareholders. These may induce a potential problem of selection bias/self-selection bias, which violates the unconfoundedness assumption (Rosenbaum & Rubin, 1983). Therefore, estimation through OLS is unappropriated because treated and non-treated firms may respond differently to specific observables and unobservable confounders, as well as to treatment intensity (t). Given the specific characteristics of the study, we apply continuous treatment effects estimated using a dose-response function (DRF) on endogenous and exogenous treatment, as proposed by Hirano & Imbens (2004) and implemented by Cerulli (2014). The DRF enable us to observe the effect of a continuous treatmentinterest on equitywhile the treatment reacts to observable (exogenous treatment) or to unobservable (endogenous treatment) confounders, and also enable us to evaluate the effect not only of the payment decision (w) of IOE, but also the effect of the its level of payment (t) on the debt tax bias. The implementation by Cerulli (2014) Table 1 reports the descriptive statistics of all the variables of interest. We can observe that capital financing (equity and financial debt) represents, on average, 38% of firms' total assets, and that the financial debt constitutes 50% of the capital financing. The payment of interest on equity represents, on average, 0.5% the firms' total assets, while the maximum deductible value is, on average, 4.23% of total assets. Considering cash flow and cash holdings, these represent on average 9.15% and 10.27% of total assets, respectively. Firms seem to be very conservative when considering the potential tax benefit they could get from the payment of interest on equity. Interest on equity only represents 15.26% of the cash distribution to shareholders (IOE + cash dividends).

Empirical results and discussion
Looking at the profile of the firm's shareholders, institutional investors are present in 13% of the firms, while 55% have pyramidal control structures. Ownership by firms and family is present in about 33% and 27% of firms, respectively. About 74% of all firm-year observations include the payment of dividends, but only about 22% include the payment of IOE.    Zani et al. (2014). However, as discussed in Section 2, OLS estimation becomes inconsistent when the unconfoundedness assumption is violated. To re-establish the condition of unconfounded treatment assignment, and assuming an exogenous treatment, we estimate an OLS dose response function (DRF-OLS).   Table 3 reports the results of the tests obtained through the DRF-OLS estimation for both proxies of financial leverage as the outcome variables. The results show that when the condition of unconfoundedness on specific observables is reintroduced, there is a positive and statistically significant effect of the treatment IOE on the Financial debt-to-capital measure. Conversely, the effect of the IOE is not statistically significant when the proxy Financial debt-to-assets is considered. As previously stated, as a proxy for the debt tax bias this ratio shows measurement deficiencies (Welch, 2011), and is only included in this study for comparison purposes with previous literature. Thus, we cannot reject our first hypothesis (H1) that the treatment by IOE tends to increase the debt tax bias instead of reducing it.
The test for our second hypothesis (H2) includes the interaction between the IOE treatment and the financial constrained status, proxied by the Financial unconstraint dummy variable. The results of this interaction, presented in Table 3 for the DRF-OLS estimation, show an inexistent effect of the treatment IOE on the debt tax bias when the financial constraint status of the firm is accounted for. The coefficient of the variable Financial debt-to-assets is negative, while positive for the Financial debt-to-capital variable. In both instances the coefficient is not statistically significant. Table 3 also reports the estimation results for the control variables. These results are in line with the existing literature not only in the expected sign, but also regarding the statistical significance. We can observe that the three polynomial factors from the DRF show statistical significance. Note: *,** and *** indicate statistical significance at the 10%, 5% and 1% level, respectively. Robust t-statistics adjusted for firm-level clustering are presented in parentheses.  (Equation 2). The selection and treatment level models are the same for all the outcome models (Financial debt-to-assets ratio and Financial debt-to-capital ratio).
The results from the selection model show that the variables Cash flow, Financial unconstraint dummy, Funds dummy and Dividend dummy (Size, Firm dummy and Pyramidal control dummy) are positively (negatively) related to higher propensity to pay IOE, with the corresponding coefficients presenting statistical significance. These results suggest that firms with higher cash flows and financially unconstrained tend to have higher payment of interest on equity. The ownership structure shows the expected effect on the propensity to pay IOE.
Investment funds have more tax benefits with the IOE, and thus positively influence its payment.
Since pyramidal ownership structures and firms shareholders have less tax benefits, or even additional costs with IOE, they tend to negatively influence its payment The results from the treatment level model indicate that the level of payment of IOE is increasing (decreasing) with higher cash flow, growth opportunities, tax shield, maximum deductible amount of IOE and payment ratio (  show that for both the Financial debt-to-assets and Financial debt-to-capital ratios, the estimated coefficients were positive but not statistically significant. Therefore, we cannot accept Hypothesis 2 (H2) that the rebound effect of the IOE on the debt tax bias is more severe on financially unconstrained firms. The results reported may be influenced by the costs of adjusting the capital structure. If this is indeed the case, firms subject to less adjustment costs should respond more quickly and contemporaneously to treatment (Almeida & Campello, 2010;Strebulaev, 2007).
Empiricaly, we should observe a negative coefficient in the tests performed for the heterogeneous effect resulting from firms with different financial constrained status (reported in Tables 3 and 4).
However, these coefficients are positive, while not statistically significant. Therefore, we may discard the hypothesis that the results were driven by the costs of adjustment of the financial structure. Note: *,** and *** indicate statistical significance at the 10%, 5% and 1% level, respectively. Robust t-statistics adjusted for firm-level clustering are presented in parentheses.
Additionally, the results found may be being driven by accounting mechanics and managerial inertia (Welch, 2004), and not necessarily by tax factors. To rule out this hypothesis, we perform an additional test using dividend payment as treatment, instead of interest on equity 9 .
This robustness test is adequate since the payment of dividends is an alternative way of cash distribution to shareholders that follows the same payment mechanisms as interest on equity. If the results are in fact being driven by this accounting mechanics associated with managerial inertia, than it would be reasonable to expect similar results between both treatments, both in direction as in magnitude. The coefficients for the variable of interest -Dividend treatment -from the estimation of the robustness tests are reported in  Note: *,** and *** indicate statistical significance at the 10%, 5% and 1% level, respectively. Robust t-statistics adjusted for firm-level clustering are presented in parentheses.

Conclusions
Our results suggest that the Allowance for Corporate Equity (ACE) system implemented in Brazil leads to an increase on the debt tax bias. This is contrary to what is expected from this type of tax policy. It also goes against the existing literature where Brazil has been referenced as an example of an emerging country using the ACE system as a way to mitigate the debt tax bias. The results found are not heterogeneous regarding the firm's ability to access the external capital market. The tests are robust to different measures of financial leverage, and to the possibility that they are being driven by costs of adjusting the capital structure or accounting mechanisms.
The empirical results suggest that the Brazilian ACE system is flawed, mainly because of two factors. The first is the fact of being required to distribution cash to shareholders for the tax deductibility. In this case, the ACE incentives do not compensate for the reduction in equity due to cash distribution. The second is due to the existence of tax benefits heterogeneous among shareholders firms, which favors the controlling shareholders which adjust the cash distribution according to their tax preferences. The empirical results suggest that firms with participation of pension funds and shareholders users of pyramidal structure as control enhancing mechanisms are more and less likely to use ACE system, respectively. Therefore, there may be a "ACE clientele effect" induced by heterogeneity in tax preferences among shareholders.
Policy makers have recognized the effect of the interest tax deductibility on the asymmetric treatment of debt and equity as one the factors that promote the distortion in incentives to leverage (FSB Financial Stability Board, 2015) being intensified in developing countries due to high inflation (Abramovsky, Klemm, & Phillips, 2014). Indeed, non-financial companies from emerging countries have experienced large increases in their leverage above levels considered normal. This increase in leverage has been recognized as an issue that has implications on economic growth and financial system stability. In response to this situation, policy makers have suggested the adoption of allowance for corporate equity as a way address the debt-equity tax bias in emerging markets (FSB Financial Stability Board, 2015). The implementation of ACE system in emerging economies is more challenging, but it can bring great advances in these economies (Abramovsky et al., 2014). Differences between advanced and emerging economies could lead to different trade-offs between the designs of tax systems (Miller, 2014), so even though Brazil is the only emerging economy with ACE system, it is expected that their experience be a lesson for ACE systems implementations expected in the near future.
The empirical results about the Brazilian experience suggest that the new ACE system designs must address concerns about the effects of eligibility criteria. In Brazil, distribution cash requirement causes negative net benefits to the tax treatment of debt bias. The necessity of distribution of cash makes the ACE systems to compete with the preference of firms to retain internal funds. In addition, the capacity to controlling shareholders in setting the cash distribution policy according to their tax preferences can have two effects: make the effectiveness of the ACE system dependent on the ownership structure and substantially change the ownership patterns of firms by tax-induced clientele effect. Therefore, the underlying mechanism of the Brazilian ACE system is flawed and it needs to be redesigned to fulfill its role. In this reformulation Brazil can follow the current ACE systems of Belgium and Italy that have shown positive results in the treatment of debt tax bias (Panier et al., 2013;Panteghini, Parisi, & Pighetti, 2012) and do not require the distribution of funds as a criterion for obtaining tax deductibility of equity.
Our empirical results also contribute to the literature on corporate governance, on capital structure and cash distribution policies when considering the way firms react to tax incentives in the presence of tax preferences by controlling and influent shareholders, by adjusting their decisions on the level of debt, equity and cash distribution to shareholders.

Variables Definition
Financial Debt-to-Assets Ratio of financial debt to total assets.
Financial Debt-to-Capital Ratio of financial debt to capital. Capital is defined as equity plus financial debt.

IOE
Ratio of interest on equity to total assets.

IOE normalized
Ratio of interest on equity to total assets. (IOE) normalized in order to only vary between 0 (inexistence of treatment) and 100 (maximum level of observed treatment). CAPEX Ratio of capital expenditures to total assets.
Cash flow Ratio of net income plus depreciation to total assets.
Cash holding Ratio of cash holdings to total assets.

Control rights
Percentage of shares with voting rights owned by the controlling shareholder. Deductible IOE Ratio of Deductible IOE to total assets. Deductible IOE is the Product between the difference of the previous year equity and revaluation reserves and the long-term interest rate, limited to the highest of 50% of the net income or retained earnings of the previous year. Dividend Ratio of cash dividend to total assets.

Dividend normalized
Ratio of cash dividend to total assets. (Dividend) normalized in order to only vary between 0 (inexistence of treatment) and 100 (maximum level of observed treatment).

Dividends dummy
Dummy variable that takes the value of one if the firm pays dividends and zero otherwise.
Financial uncontraint dummy Dummy variable that takes a value of one if the firm is not financial constrained and zero otherwise. Firms with ADR are considered not financial contrained. Firm dummy Dummy variable that takes the value of one if the controlling shareholder is a firm and zero otherwise.
Funds dummy Dummy variable that takes the value of one if an investment fund owns more than 5% of shares with voting rights and zero otherwise.
Governance quality dummy Dummy variable that takes the value of one if the firm is listed on the three high-governance listing of BM&Fbovespa, and zero otherwise.
Individual dummy Dummy variable that takes the value of one if the controlling shareholder is an individual and zero otherwise.

IOE dummy
Dummy variable that takes the value of one if the firm pays interest on equity and zero otherwise.
Market-to-book Ratio of market value of assets to total assets. The market value of assets is defined as total assets minus equity plus the market value of equity. Payout ratio Ratio of cash dividend to net income.
Pyramid Control Dummy variable that takes the value of one if the ultimate controlling shareholder uses a pyramid structure and zero otherwise.
Size Natural logarithm of total assets.
Tax shield Ratio of the difference of EBIT and the ratio of income tax payments to corporate tax rate to sales.