Agency Costs, Capital Structure Decisions and the Interaction with Payout Decisions: Empirical Evidence from Brazil
Antonio Zoratto Sanvicente
Full Professor, Ibmec São Paulo
São Paulo, SP, Brazil
January, 2013
Abstract
This paper analyzes the determinants of capital structure in a sample of 167 publiclyowned Brazilian firms with data for 2010 and 2011. The approach is empirical, emphasizing the interdependency between debt and dividend policies and the agency cost of equity. The point is made, and the results indicate the validity of the point that most of the Brazilian literature not only ignores those interdependencies but also uses the incorrect estimator for the determinants of capital structure. With the estimation of a threeequation system by threestage least squares, with lagged dependent variables, whose importance is also ignored in the Brazilian literature, the paper obtains significant results, and the results confirms unambiguously the existence of a convergenceofinterest effect in the agency relationship involving controlling and minority shareholders.
Keywords: capital structure, dividend policy, ownership concentration, agency relationship, endogeneity bias.

Introduction
Since Modigliani and Miller (MM) (1958; 1963) published their classical “irrelevance theorem” papers on capital structure decisions and their possible impact on firm value, a large portion of the financial economics and corporate finance literature has been concerned with (a) building on the MM propositions with the addition of capital and managerial labor market imperfections, and (b) testing the implications of the theories arising from such additions.
In short, it has been variously shown that market imperfections cause the capital structure decision to be relevant, i.e., that there is an optimal capital structure in the sense that its implementation maximizes the value of the firm. This is the contribution of the socalled static trade off theory of capital structure, whose main driver is the existence of market imperfections in the form of costs of financial distress, that arise from information and transactions costs which in turn generate an agency problem involving firm owners/managers and creditors. Similarly, a dynamic version of the theory, most commonly known as the pecking order theory of capital structure, deals with the choice between debt and equity when the firm needs new funds to finance longterm investments. This theory, in turn, is predicated on the acknowledgement of information asymmetry between managers and owners, on one hand, and outside equity and debt suppliers, on the other hand.
In several cases documented in the abundant literature on capital structure decisions, two other important corporate decisions are used as independent or control variables. One is the firm’s dividend policy or payout decision, also dealt with in classical fashion by Miller and Modigliani (1961) in another “irrelevance theorem” paper. The other is the ownership or property concentration decision, most notably discussed by Jensen and Meckling (1976), as one of the main aspects of the agency relationships involving owners, managers and creditors in a corporation.
As pointed out in the next section of this paper, the literature that discussed the payout and ownership decisions often included the capital structure decision as an important factor. This clearly has the potential of creating an endogeneity problem, best dealt with, in empirical tests, with the use of simultaneous equation models. This was pointed out very well by Kim, Rhim and Friesner (2007), referred to as KRF from this point on, in their paper on South Korea.
The objective of this paper is to add to the Brazilian evidence on the capital structure decision of publiclyowned firms with the use of the KRF empirical approach, and also add to the methodological menu described and discussed in Rocha and Amaral (2007), in their empirical paper on the determinants of Brazilian firms’ indebtedness. However, instead of focusing on a measure of ownership concentration as the third main variable in the analysis, the paper uses a proxy for the agency cost of equity, namely, the inverted asset turnover ratio. This variable was empirically demonstrated to have, as reported in Florackis and Ozkan (2009), a significant and direct relationship with a composite entrenchment index, constructed with the use of principal component analysis, and including several of the usual measures of governance quality, such as ownership concentration, variable executive compensation, the size and the presence of independent directors in a firm’s board of directors, among others.
The remainder of this paper is organized as follows. Section 2 presents a review of the relevant literature, covering the fundamental theoretical discussions that lead us to consider that the capital structure, payout and ownership decisions are in fact interdependent, and the specific Brazilian literature that has tested one or more theories of the capital structure decision. Hence, the focus of attention in this paper, in terms of its results, will be on the determinants of capital structure in Brazil. It concludes with the specification of the main hypotheses to be tested. Section 3 describes the sample of firms, variable definitions, data sources, and the methodology. Section 4 presents the results and section 5 concludes.

Literature Review and Testable Hypotheses
In the discussion that follows, as in the rest of this paper, controlling shareholders will be referred to as “insiders”. As pointed out in KRF (2007), the ownership decision has been basically cast in an agency cost framework, and here, as in Jensen and Meckling (1976), the agency relationship involves a principal (minority shareholder) and an agent (the controlling shareholder, or “insider”). This is what generates the agency cost of equity. Obviously, when a firm uses some external equity, and there are outside shareholders, the controlling shareholder will attempt to maximize his/her wealth at the expense of minority shareholders, with the creation of conflicts of interest and the socalled agency cost of equity, which increases with the proportion of external equity used by the firm. The literature (for example, Copeland et al. 2005) mentions the possibility of using independent auditors to reduce that agency cost. As is also claimed in the literature reviewed in what follows, certain important policies, such as debt financing and dividend payment policies can be used as external control mechanisms.
In the case of debt financing, the need to make the required interest and amortization payments helps to limit the “free cash flow” at the disposal of controlling shareholders; analogously, a high payout ratio limits the firm’s ability to finance profitable new investment opportunities without accessing the capital markets and having to provide information about its intentions.
Concerning the relationship between controlling and minority shareholders, one speaks of two possibilities or “effects”: (1) entrenchment by insiders, namely, the situation in which they hold a sufficiently large proportion of the firm’s shares to feel immune to the discipline imposed by the entrepreneurial market, since their control over the firm makes it increasingly difficult to replace them, in case the firm runs into difficulties or its economic performance begins to suffer; (2) convergence of interests, when insiders hold a sufficiently large stake in the firm to become interested in maximizing overall shareholder wealth, since this also contributes to maximizing their own wealth. Therefore, a higher proportion of ownership by insiders will induce them to act in such a way as to maximize shareholder wealth, up to a certain point, at which entrenchment sets in.
As pointed out by Florackis and Ozkan (2009), the insideroutsider conflict may be controlled by corporate governance mechanisms, thereby contributing to the reduction of agency costs. Both external and internal governance mechanisms can be used; the internal mechanisms include the composition of the board of directors (e.g., the presence of independent directors), managerial incentives (such as variable compensation), capital structure and dividend policies. Therefore, the firm’s observed capital structure may be the result of the existing level of agency cost and, at the same time, determine the observed agency cost of equity.
Florackis and Ozkan (2009) construct an entrenchment index for UK firms, using data on governance mechanisms such as board size and composition, ownership concentration ratios, magnitude of variable compensation, and the nature of block holdings, and find that their index is significantly and positively related to proxies of agency cost, especially the inverted asset turnover ratio. This result will enable us to use that agency cost proxy directly in the analysis that follows.
In that context, both debt and dividend policies may be used as monitoring tools regarding the actions by insiders. The issuance of debt securities creates greater opportunities for outside monitoring, in this case by creditors, as discussed by Jensen and Meckling (1976). In turn, higher payout, since it is usually followed by the issuance of new securities in order to finance investments, also creates greater opportunities for outside monitoring, as pointed out by Rozeff (1982). Hence, since two of the tools used for reducing agency costs are associated with better outside monitoring, they may be substitutes or complements for each other, and thus they are interdependent: the observed capital structure may be partly determined by dividend policy, and vice versa.
In this sense, then, this paper, which is an empirical analysis of debt policy choices by publiclyowned Brazilian firms, will take into account the interaction of both debt and dividend policies, as their interaction with observed agency cost, since the two policies may be used as tools for reducing agency costs of equity.
The previous discussion, in addition, indicates that debt and dividend policies may be substitutes for each other, if there is convergence of interests, and complementary, if there is entrenchment. This leads, for example, to two different testable hypotheses regarding the relationship between debt and dividend policies: the association between indebtedness and payout will be positive if there is entrenchment (because the proportion of ownership by insiders is high and the agency cost of equity is also high), or negative if there is convergence of interests (when the proportion of ownership by insiders is low, and so is the level of the agency cost of equity). Clearly, any hypothesis about the relationship between debt and dividend policies must be tested controlling for the level of agency cost.
Finally, even though the interactions described herein and in KRF (2007) had already been examined, for example, in Jensen et alii. (1992), they were already present before the development of the socalled “modern finance theory”, when the “residual theory” of dividend policy was postulated. That theory proposed that payout was decided upon as a “residual” in the sense that the firm would pay dividends only after it determined how much of its income in a given period would be left over after (1) financing new positive net present value investments so as to (2) maintain the optimal balance between equity (such as internallygenerated funds) and debt. If the firm decided to payout a proportion of income different from that determined by this procedure, it would force itself to make a suboptimal investment decision, a suboptimal capital structure decision, or both. Therefore, it is clear that debt and dividend policies already interacted in this early imperfect market theory of dividend policy. For a discussion, see Gitman (2004), for example.
The literature involving Brazilian firms is mostly concerned with explaining either the choice of debt policy, with payout and proxies for ownership policies as control variables, or the choice of payout policy, with debt and ownership policies as controls. In this paper, even though the main objective is to explain debt policy, the starting point is the level of the agency cost of equity. Given the possible interdependency of the three variables, it really does not matter much which variable is assigned the role of the “dependent” variable. It is simply more convenient to begin with an agency cost of equity story associated with ownership policy.
Rocha and Amaral (2007) discuss in an appendix to their paper, and with illustrations from the Brazilian debt policy choice literature, the various empirical methodology alternatives that could be used. They are (the papers which illustrate their discussion are cited in parentheses along with each alternative):

Simple linear regression estimated by OLS (SILVA and VALLE, 2005).

Multiple linear regression with the historical means of each variable (GOMES and LEAL, 1999; PEROBELLI and FAMÁ, 2002).

The twostage Fama and MacBeth (1973) methodology (BRITO and LIMA, 2005).

Static panel estimation, with fixed or random effects (TERRA, 2002; SOARES and KLOECKNER, 2006).

Dynamic panel estimation with the Arellano and Bond (1991) or Blundell and Bond (1998) methodologies (BARROS et alii. 2006).

Threestate least squares estimation for a system of equations (SILVEIRA et alii. 2008).
In turn, the Rocha and Amaral (2007) methodology uses twostage least squares for estimating a single equation with various specifications.
It is apparent from their descriptions that all of the abovelisted methodologies may correct, to a greater or smaller extent, for autocorrelation or heteroscedasticity, but, with the exception of Silveira et alii. (2008), none uses an estimator that considers the full interdependency possibilities involving all policy variables. Hence, they all may suffer from simultaneity and/or measurement error bias (for example, in the case of the Brito and Lima (2005) paper). This includes the Rocha and Amaral (2007) analysis.
In particular, the Silveira et alii. (2008) paper uses the threestage least squares estimator to examine a twoequation specification in which the endogenous variables are debt (proxied by socalled “financial” liabilities, namely, notes payable to financial institutions, which excludes accounts payable to suppliers, wages and taxes payable), and a “quality of governance” variable, whose value is determined with the use of a questionnaire developed by Silveira (2004). Hence, this paper does not even consider a proxy for agency cost that is relevant to the entrenchment versus convergenceofinterests discussion, and a payout policy proxy is not even used as a control variable in the two equations.
The results of Brazilian research on the determinants of debt policy, in addition to problems with the choice of specification and estimation method, reveal the following:

The papers by Gomes and Leal (2001), Perobelli and Famá (2002), Terra (2002), Basso et alii. (2004), Brito et alii. (2005), Martin et alii. (2005), Moraes and Rhoden (2005), Famá and da Silva (2005), and Barros et alii. (2006) did not find any significant determination of capital structure by either payout or ownership policy. As discussed above, when such policy variables were included, there may have been specification problems and the use of an inappropriate estimator. This, once more, makes the importance of modeling their interdependency very clear.

A paper by Procianoy and Schnorrenberger (2004), made an attempt at evaluating the possible association between ownership and capital structures. It tested the hypothesis that higher ownership concentration was associated with more limited use of debt in the firm’s capital structure, and obtained evidence in that direction. It used as a proxy for the ownership decision the proportion of voting shares held by up to the top five shareholders, creating a variable for the proportion held by the top shareholder, a second variable for the proportion held by the top two shareholders, and so on; however, in effect only the results for the proportion held by the top three investors are reported. In contrast, this paper considers a proxy for the agency cost of equity directly. Their results, given the discussion of testable hypotheses above, are consistent with the convergenceofinterests explanation. It is just the opposite of what has been found in this paper, as indicated by our results. However, their paper used a singleequation specification with the debt level as the dependent variable, various cumulative levels of ownership as independent variables and several control variables, none of which measures payout directly. The equation was estimated by OLS. Hence, the discrepant results may be attributed to substantial specification error.

Data and Methodology
Given the literature review, three simultaneouslydetermined equations are proposed below, including the expected signs for the coefficients.
Capital structure decision equation:
(1)
Where,
DEBT_{t} = total debt ratio = total debt/total assets at the end of year t;
PAYOUT_{t} = dividends paid/net income during year t;
INVTURN_{t} = 1/total asset turnover = total asset/total sales revenue during year t;^{1}
CFLOW_{t} = operating cash flow = earnings before interest, taxes, depreciation and amortization/total assets in year t;
CURR_{t} = current liquidity ratio = current assets/current liabilities at the end of year t;
MARGIN_{t} = operating margin = earnings before interest and taxes/total assets in year t;
SIZE_{t} = natural logarithm of total sales revenue (million Reais) during year t;
INTANG_{t} = intangible assets/total assets at the end of year t;
OPNPV_{t} = proxy for positive net present value investment opportunities = enterprise value/total assets at the end of year t.
Expected signs for the coefficients of equation (1):^{2}
c_{12}: positive (if there is entrenchment); negative (if there is convergence of interests)
c_{13}: positive (if there is entrenchment); negative (if there is convergence of interests)
c_{14}: negative
c_{15}: negative
c_{16}: negative
The CFLOW, CURR and MARGIN variables are proxies for the availability of internallygenerated funds, which are a preferred source of financing, according to the pecking order theory of Myers and Majluf (1984). Hence, the expectation of negative signs for the corresponding coefficients.
Payout decision equation:
(2)
The expected signs for the coefficients of equation (2) are as follows:^{3}
c_{22}: positive (if there is entrenchment); negative (if there is convergence of interests)
c_{23}: positive (if there is entrenchment); negative (if there is convergence of interests)
c_{24}: positive
c_{25}: positive
c_{26}: positive
c_{27}: negative
Agency cost equation:
(3)
In equation (3), the additional variable STDEBT corresponds to the ratio between current liabilities and total assets.
The expected signs for the coefficients of equation (3) are:
c_{32}: positive (if there is entrenchment); negative (if there is convergence of interests)
c_{33}: positive (if there is entrenchment); negative (if there is convergence of interests)
c_{34}: negative^{4}
In all three equations, a lagged value for the dependent variable was included, since it is expected that the levels of such variables are not susceptible to significant changes from one year to the next, as observed by Florackis and Ozkan (2009). If this is true, the coefficient of the lagged dependent variable will be positive and the variable itself will be significant.
The values for all variables correspond to t = 2011 and t1 = 2010, and were obtained from the Economática ® database for Brazilian publiclyowned firms, as well as the Reference Forms submitted to the Comissão de Valores Mobiliários (CVM), available at the CVM website and/or the sample firms’ own websites.
In the 3SLS estimation of the system of equations above, the instruments used included (a) the lagged variables of DEBT, PAYOUT and INVTURN; (b) the exogenous variables in equations (1)(3), i.e., CFLOW, CURR, MARGIN, SIZE, OPNPV, INTANG and STDEBT.
The sample comprises 167 Brazilian publiclyowned firms. The sample does not include financial institutions, since the nature of their financial statements differs very much, particularly in terms of capital structure, from those of industrial/commercial/service firms. It also excludes firms whose payout ratio in 2011 was negative. The sample’s corresponding descriptive statistics and correlation coefficients are displayed in Tables 1 and 2.
Table 1. Descriptive statistics for the 167 firms included in the sample, 2011.
Variable

Mean

Median

Maximum

Minimum

Standard
Deviation

DEBT (%)

43.9150

42.3957

88.0202

0.2717

17.0058

PAYOUT (%)*

59.8768

34.5689

698.6020

0.0000

90.0022

INVTURN

2.3831

1.7288

18.7483

0.2824

2.5778

CFLOW (%)

12.2786

10.5694

55.4151

7.5285

8.3464

CURR

1.8789

1.6933

11.9534

0.3389

1.3229

MARGIN (%)

19.5691

12.8678

213.4355

12.1987

25.2643

SIZE

7.4024

7.3373

12.4057

2.4114

1.5994

* 32 of the sample firms did not pay any dividends in 2011.
Table 2 displays the Pearson correlation coefficients for all variables included in equations (1)(3), that is, for the three dependent variables (DEBT, PAYOUT and INVTURN) and four control variables (CFLOW, CURR, MARGIN and SIZE) associated with common hypothesis about determinants of firm indebtedness.
Table 2. Pearson correlation coefficients for all variables used in equations (1)(3).

DEBT

PAYOUT

INVTURN

CFLOW

CURR

MARGIN

SIZE

DEBT

1,0000







PAYOUT

0.0518

1,0000






INVTURN

0.1150

0.0665

1,0000





CFLOW

0.0571

0.0278

0.2001*

1,0000




CURR

0.5071*

0.2926*

0.1452

0.1535*

1,0000



MARGIN

0.1355

0.0123

0.7291*

0.3232*

0.0625

1,0000


SIZE

0.1681*

0.0215

0.2485*

0.2424*

0.2235*

0.0627

1,0000

* Statistically significant at the 5% level. Sample size = 167 observations.
The application of a 5%significance ttest to the Pearson correlation coefficients shows that DEBT and INVTURN are positively associated, as would be predicted by the entrenchment hypothesis for the agency relationship, but the correlation is not significant. There does not appear to be a significant association between the third variable (PAYOUT) and the DEBT and INVTURN variables. These results, however, does not yet take into account the full possibilities of interaction, as well as the influence of control variables, some of which appear to be associated with the dependent variables, but, again, before taking into account the interdependency of the three dependent variables. This is accounted for in the simultaneousequation setup based on the equation system (1)(3). Those results follow.

Results
Since an important part of this paper’s objective is to ascertain the importance of treating capital structure, payout and the proxy for agency cost as interdependent variables, the results presented in Tables 35 provide the reader with a comparison of results for when the three variables are not treated as interdependent, even though, for example, one usually tests for the determination of the firm’s capital structure by dividend and ownership policies, in addition to other factors. Thus, the results provided in the second and third columns of Tables 35 are those obtained estimated equations (1)(3) as separate empirical models, by ordinarily least squares (OLS). The fourth and fifth columns provide the results obtained when equations (1)(3) are estimated by threestage least squares (3SLS), with the instruments listed above.
In addition, a choice was made to use the 3SLS estimator, as opposed to estimation by 2SLS, as in Rocha and Amaral (2007), because the 2SLS estimator will be inefficient, especially when the equations contain different independent variables, as in this case, and the error terms in the system are heteroscedastic.
Table 3. Results for equation (1). DEBT is the dependent variable. Sample size = 167 observations.
Variable

Estimated coefficient (OLS)

Prob(tstatistic)

Estimated coefficient (3SLS)

Prob(tstatistic)

Intercept

0.0169

0.5653

0.0419

0.2439

DEBT_{t1}

0.8428

0.0000

0.8016

0.0000

INVTURN

0.0025

0.3628

0.0200

0.0038

PAYOUT

0.0054

0.3650

0.0498

0.0121

CFLOW

0.3090

0.0074

0.7012

0.0001

CURR

0.0237

0.0000

0.0352

0.0000

MARGIN

0.0303

0.3478

0.1995

0.0038

SIZE

0.0011

0.6937

0.0016

0.6291

INTANG

0.0338

0.3633

0.0381

0.1807

OPNPV

0.0179

0.0868

0.0263

0.0015

Adj. R^{2}

0.8529

0.7955

The results for the main specification considered in this paper indicate the following:

The signs of all coefficients are unchanged from OLS to 3SLS, an indication that simultaneity bias may not be a serious problem, although magnitudes change substantially in most cases.

The standard errors of most coefficients, particularly those associated with the INVTURN and PAYOUT variables, are lower with 3SLS than with OLS. In fact, these variables, while not significant under OLS, are significant with 3SLS.as an indication that 3SLS is effectively a more efficient estimator, as expected.

The negative sign of the coefficient association with the PAYOUT variable indicates the net existence of convergence of interests, since the negative sign corresponds to substitution between debt and dividend policies).

The positive and significant result for the proxy for agency cost of equity (INVTURN) indicates that, the higher the cost of agency, the greater the tendency for using debt as a cost monitoring mechanism.

The significant result obtained for the CFLOW variable is consistent with the static tradeoff hypothesis of capital structure. This is in substantial divergence with the results in Rocha and Amaral (2007), where a negative coefficient was obtained, though not always significant. However, it is also consistent with the prediction from the pecking order theory, as discussed above.

The positive and significant result for MARGIN, however, contradict the predictions from the pecking order theory, which would indicate that, the more profitable a firm, the higher the use of retained earnings and the lower its reliance on the use of debt. Again, this is a result in favor of the static tradeoff theory of capital structure.
Table 4. Results for equation (2). PAYOUT is the dependent variable. Sample size = 167 observations.
Variable

Estimated coefficient (OLS)

Prob(tstatistic)

Estimated coefficient (3SLS)

Prob(tstatistic)

Intercept

0.0380

0.9081

0.1596

0.5016

PAYOUT_{t1}

0.5272

0.0114

0.3705

0.0036

DEBT

0.8518

0.0869

1.2191

0.0108

INVTURN

0.1467

0.0931

0.2196

0.0002

CFLOW

3.8106

0.0195

5.9237

0.0001

CURR

0.2711

0.0910

0.3119

0.0000

MARGIN

1.5384

0.0604

2.1778

0.0000

OPNPV

0.1520

0.0058

0.1790

0.0358

Adj. R^{2}

0.1874

0.1632

The results in Table 4 indicate the following:

As in the case of the results for equation (1), no change in sign is observed, and the 3SLS results involve increased significance, as expected from a more efficient estimator.

The sign of the DEBT variable confirms the substitution role between debt and dividend policies, already observed in the results for equation (1). The positive and significant result for the proxy for agency cost indicates that dividend policy (a higher payout ratio) seems to be used as a voluntary control mechanism.

One of the control variable coefficients (CFLOW) has the expected positive sign. However, the coefficient for MARGIN is significant, but negative, apparently indicating the importance of dividend policy as a control mechanism, or the firm’s reaction to the dearth of profitable investment opportunities. However, the negative but significant result for OPNPV seems to contradict the latter conjecture: better growth opportunities lead to reduced payout, so that they can be financed with retained earnings.

When compared to the results presented in Table 2 (Pearson correlation coefficients), the significance of the PAYOUT policy variable (at any reasonable level) and the INVTURN proxy in equation (3) (and DEBT and INVTURN in equation (4)) is in stark contrast with the lack of partial association among the three variables, indicating that it is crucial that one takes their interdependency into account.
Table 5. Results for equation (3). INVTURN is the dependent variable. Sample size = 167 observations.
Variable

Estimated coefficient (OLS)

Prob(tstatistic)

Estimated coefficient (3SLS)

Prob(tstatistic)

Intercept

1.8238

0.0022

2.3238

0.0000

INVTURN_{t1}

0.7710

0.0000

0.7836

0.0000

DEBT

1.2622

0.0365

1.2293

0.0374

PAYOUT

0.2207

0.2869

0.8623

0.0000

STDEBT

2.4528

0.0024

3.0829

0.0000

Adj. R^{2}

0.8878

0.8374

The results presented in Table 5 indicate the following:

The use of 3SLS instead of OLS does not lead to changes in the signs of coefficients, but produces higher significance, particularly in the case of PAYOUT.

As shown by Florackis and Ozkan (2009), firms in Brazil also use shortterm debt as mechanism to control for agency cost of equity, in addition to total debt and payout. The negative coefficients for DEBT and PAYOUT in equation (3) confirm that higher agency cost tends to result from lower levels of DEBT and PAYOUT as control mechanisms.

The negative signs for both DEBT and PAYOUT are, once more, indications of convergence of interest in the sample firms. This result, as well as those in Tables 3 and 4, provide a stronger confirmation of convergence of interests than that obtained by KRF(2007), where the corresponding coefficients were in some cases significantly positive (for example, for their OWN, or ownership concentration variable in the DEBT equation), whereas others were negative (for example, for the same OWN variable in the PAYOUT equation).

Conclusion
This paper presents evidence indicating that the choice of capital structure is positively associated with the level of the agency cost of equity (as proxied by inverted asset turnover): this is consistent with the existence of convergence of interests, and with the intention of protecting minority investors. Overall, the results obtained in this paper contradict most of the evidence accumulated in several papers already presented, discussed and published involving the determinants of capital structure. The differences are initially caused by the application of incomplete theory, namely, failing to consider the interaction among agency cost of equity, debt and payout policy choices, described in section 2 of this paper. Following the failure to recognize that interaction, all previously available evidence is obtained with the use of obviously inappropriate specifications and the resulting biased and/or inefficient estimators. In some cases, even the direction of the association between two of the three policy variables is the opposite of this paper has revealed to be.
This paper replicates the KRF (2007) analysis for South Korea with data for Brazilian firms. Their results for the entrenchment versus convergenceofinterests dichotomy were not unambiguous, and they finished by claiming that these two explanations for agency costs are not mutually exclusive. The evidence in our paper, however, is much clearer in this respect, for the dominance of convergence of interests. In addition, this paper adjusted for the possibility that firms are not able to change the relevant policies or the level of agency cost very rapidly. The fact that all lagged terms are significant and positive is a testament to the truth of this conjecture.
In future research, we would be very happy if we could enlarge the sample size, when access to a longer period becomes possible, and a better dynamic approach can be used.

References
Arellano, M.; S. Bond. Some Tests of Specification for Panel Data: Monte Carlo Evidence and An Application to Employment Equations. Review of Economic Studies, v. 38, n. 194, Apr. 1991, p. 277297.
Barros, L. A. B. C.; A. M. Silveira; H. P. Silveira. Excesso de Confiança, Otimismo Gerencial e os Determinantes da Estrutura de Capital. Paper presented at the 6th Annual Meeting of the Brazilian Finance Association, Vitória, 2006.
Basso, L. F. C.; E. A. Mendes; E. K. Kayo. Teste da Teoria da Janela de Oportunidades para o Mercado Acionário Brasileiro. Paper presented at the 28th ANPAD Annual Meeting, Curitiba, 2004.
Blundell, R.; S. R. Bond. Initial Conditions and Moment Restrictions in Dynamic Panel Data Models. Journal of Econometrics, v. 87, n. 1, Nov. 1998, p. 115143.
Brito, G. A. S.; L. J. Corrar; F. D. Batistella. Fatores Determinantes da Estrutura de Capital das Maiores Empresas que Atuam no Brasil. Paper presented at the 5^{th} Annual Meeting of the Brazilian Finance Association, São Paulo, 2005.
Brito, R. D.; M. R. Lima. A Escolha da Estrutura de Capital sob Fraca Garantia Legal: o Caso do Brasil. Revista Brasileira de Economia, v. 59, n. 2, abr./jun. 2005, p. 177208.
Copeland, T. E.; J. F. Weston; K. Shastri. Financial Theory and Corporate Policy, 4^{th} Ed. Pearson Addison Wesley, Boston, 2005. 1000p.
Famá, R.; E. S. da Silva. Desempenho Acionário e a Estrutura de Capital as Companhias Abertas Brasileiras NãoFinanceiras. Paper presented at the 5th Annual Meeting of the Brazilian Finance Association, São Paulo, 2005.
Florackis, C.; A. Ozkan. The Impact of Managerial Entrenchment on Agency Costs: An Empirical Investigation Using UK Panel Data. European Financial Management, v. 15, n. 3, p. 497528, 2009.
Gitman, L. J. Princípios de Administração Financeira, 10th ed. São Paulo, Pearson, AddisonWesley, 2004. 745p.
Gomes, G. L.; R. P. C. Leal. Determinantes da Estrutura de Capital das Empresas Brasileiras com Ações Negociadas em Bolsa de Valores. In: Leal, R. P. C.; N. Costa, Jr.; E. F. Lemgruber. Finanças Corporativas, São Paulo, Atlas, 2001.
Jensen, G.; D. Solberg; T. Zorn. Simultaneous Determination of Insider Ownership, Debt, and Dividend Policies. Journal of Financial and Quantitative Analysis, v. 27, June 1992, p. 247263.
Jensen, M.; W. Meckling. Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure. Journal of Financial Economics, v. 4, Oct. 1976, p. 305360.
Kim, Y. H.; J. C. Rhim; D. L. Friesner. Interrelationships among Capital Structure, Dividends, and Ownership: Evidence from South Korea. Multinational Business Review, v.
15, n. 3, Fall 2007, p. 2542.
Martin, D. M. L.; D. Forte ; A. C. F. Costa; W. T. Nakamura; A. F. Carvalho Filho; A. C. Amaral. Determinantes da Estrutura de Capital no Mercado Brasileiro: Análise de Regressão com Painel de Dados no Período 19992003. Paper presented at the 5th Annual Meeting of the Brazilian Finance Association, São Paulo, 2005.
Miller, M.; F. Modigliani. Dividend Policy, Growth, and the Valuation of Shares. Journal of Business, v. 34, Oct. 1961, p. 411433.
Myers, S. C.; N. Majluf. Corporate Financing and Investment Decisions When Firms Have Information that Investors Do Not Have. Journal of Financial Economics, v. 13, 1984, p. 187221.
Modigliani, F.; M. Miller. The Cost of Capital, Corporation Finance and the Theory of Investment. American Economic Review, v. 48, June 1958, p. 261297.
Modigliani, F.; M. Miller. Corporate Income Taxes and the Cost of Capital: A Correction, American Economic Review, v. 53, June 1963, p. 433443.
Moraes, E. G.; M. I. S. Rhoden. Determinantes da Estrutura de Capital das Empresas Listadas na Bovespa. Paper presented at the 29th ANPAD Annual Meeting, Brasília, 2005.
Perobelli, F. F. C.; R. Famá. Determinantes da Estrutura de Capital : Aplicação a Empresas de Capital Aberto Brasileiras. RAUSP, v. 37, n. 3, jul./set. 2002, p. 3346.
Silveira, A. M.; F. F. C. Perobelli; L. A. B. C. Barros. Governança Corporativa e os Determinantes da Estrutura de Capital: Evidências Empíricas no Brasil. RAC, v. 12, n. 3, jul./set. 2008, p. 763788.
Procianoy, J. L.; A. Schnorrenberger. A Influência da Estrutura de Controle nas Decisões de Estrutura de Capital das Companhias Brasileiras. Revista Brasileira de Economia, v. 58, n. 1, jan./mar. 2004, p. 122146.
Rocha, F. D.; H. F. Amaral. Análise dos Determinantes do Endividamento das Empresas Brasileiras à Luz de Abordagens Teóricas Tradicionais e Recentes. Paper presented at the 7th Annual Meeting of the Brazilian Finance Association, São Paulo, 2007.
Rozeff, M. Growth, Beta and Agency Costs as Determinants of Dividend Payout Ratios. Journal of Financial Research, v. 5, Fall 1982, p. 249259.
Saito, R.; A. A. M. Silveira. Governança Corporativa: Custos de Agência e Estrutura de Propriedade. RAE, v. 48, n. 1, 2008, p. 120.
Silva, A. F.; M. R. Valle. Análise da Estrutura de Endividamento: um Estudo Comparativo entre Empresas Brasileiras e Americanas. Paper presented at the 29th ANPAD Annual Meeting, Brasília, 2005.
Silveira, A. M. Governança corporativa e estrutura de propriedade: determinantes e relação com o desempenho das empresas no Brasil. Doctoral dissertation. Universidade de São Paulo, São Paulo, SP, Brasil, 2004.
Soares, R. O.; G. O. Kloeckner. Relações entre Estrutura de Propriedade e Estrutura de Capital no Brasil. Paper presented at the 6th Annual Meeting of the Brazilian Finance Association, Vitória, 2006.
Terra, P. R. S. An Empirical Investigation on the Determinants of Capital Structure in Latin America. Paper presented at the 26th ANPAD Annual Meeting, Salvador, 2002.
1 This is the proposed proxy for agency cost. The asset turnover ratio is a commonly used measure of the firm’s operating efficiency. According to Florackis and Ozkan (2009, p. 499): “A low asset turnover ratio indicates poor investment decisions, insufficient effort, and consumption of perquisites, and hence suggests that agency costs arising from the conflicts between managers and shareholders may not be insignificant”. The authors mention that another commonly used proxy for agency cost is the ratio of selling, general and administrative expenses to total sales revenue.
2 If the firm is not capable of generating funds for investment purposes in its own operations, and does not possess sufficiently large current resources, it is expected that the firm will be forced to resort to new debt as a source of funds. This would be predicted by the pecking order proposed by Myers and Majluf ( 1984). Therefore, the signs for the coefficients of CFLOW, CURR and MARGIN would be expected to be negative in equation (1).
3 The expected positive signs for the control variable coefficients correspond to the expectation that, the greater the capacity to generate cash flows from operations (CFLOW), the higher the firm’s liquidity (CURR), and the higher the firm’s profitability (MARGIN), the greater will the firm’s capacity to distribute current income to shareholders (PAYOUT). Since the retention of earnings is one important alternative for new investment financing, when the firm is able to generate funds in its operations, the higher its payout ratio can become. However, this will lead to the expectation that, with profitable opportunities available, the firm will retain profits to finance them. This is in accordance with the pecking order theory of capital structure determination.
4
