Introduction

The capital structure has been gaining prominence over the years, and it is one of the most important and complex contents in corporate finance (^{Espinola, 2013}). The concept of capital structure is related to the forms of financing of companies. There are two forms of financing, equity and debt. But there is not consensus in financial theory about the best form of financing (^{Santos et al., 2009}).

^{Durand (1952)} was one of the leading researchers in a problem in a quantitative form. However, ^{Modigliani and Miller (1958)} created the first theory of capital structure. They argued that the capital structure was not essential to measure the value of companies. Recently, authors refuted this theory. They proposed a new theory. It showed an increased value of the companies (^{Fama; Barros; Silveira, 2001}).

There are several external elements that affect the form of financing of companies. Credit offers and interest rates are some examples of these external elements. However, Intrinsic organizational factors also have effects. Internal elements that affect the form of financing are size of company, operating margin, and leverage. Therefore, capital structure theories analyze the main causes of the companies’ indebtedness. The main theories are Static Trade-off Theory, Agency Theory, Pecking Order Theory, and Market Timing Theory (^{Espinola, 2013}).

Considering previous empirical studies about capital structure theory, this research attempts to answer the question: are the determinants of the capital structure in Brazilian companies different in times of crisis?

The main motivation for this paper is to assess whether there is a distinct behavior among the determinants of the capital structure in the above periods, in order to compare the results with the research hypotheses, which were built grounded on the four main theories of capital structure (Static Trade-off theory, Agency theory, Pecking Order Theory, and Market Timing Theory).

Finally, this paper aims to identify the determinants of the capital structure of Brazilian publicly-held companies and, specifically, through the results of the regressions, determine whether there were deviating behaviors between the periods. Therefore, the general hypothesis tested shows that companies may present different relationships in some determinants of the capital structure. It occurs in both periods, normality and crisis periods.

Theoretical background

The capital structure research is grounded on a central issue that concerns the company’s value in relation to the way it is financed. Numerous variations, innovations, and theoretical perspectives on indebtedness of organizations have emerged in opposition to the pioneer and classical theories of ^{Modigliani and Miller (1958)}. There are four main theories (Static Trade-off Theory, Agency Theory, Pecking Order Theory, and Market Timing Theory), which try to consistently establish an optimal point of debt for companies to take full advantage of the benefits offered by debts (^{Brito, 2007}).

*Static trade-off theory*

The theory presents the idea that there is an optimal capital structure. There is a combination of equity and debt that is able to increase the value of the company. Debt utilization generates two opposite implications: tax benefit (positive effect) and bankruptcy costs (negative effect) (^{Santos et al., 2009}). It is necessary for companies to seek a capital structure that will maximize the benefits and minimize debt costs (^{Bastos; Nakamura, 2009}).

According to ^{Myers (1984)}, companies should gradually replace equity with debt and vice versa to an equilibrium point at which it is possible to reduce the costs and increase the value of the company.

*Asymmetric information and Pecking Order Theory*

Pecking Order theory is the approach that determines the optimal capital structure. It is based on minimizing inefficiency in the companies’ investment decisions (^{Myers; Majluf, 1984}; ^{Myers, 1984}).

The idea presented by ^{Myers and Majluf (1984)} is based on information asymmetry between managers and investors, i.e., managers have more and better information on investment options and financing, to the point of prioritizing the opportunity that will bring lower risk to the company.

Thus, ^{Myers (1984)} developed the Pecking Order Theory that determines a hierarchy in the sources of funding: firstly, using the retained earnings as sources of financing; secondly, acquiring new debt; and, ultimately, issuing shares to obtain resources. This structure is maintained even if the shares are overvalued.

*Market timing theory*

In a globalized environment it is clear that a range of factors can influence the value and the capital structure of an organization, and according to ^{Frank and Goyal (2009}) managers follow two lines regarding capital structure decisions: suspend the financing decision for as long as possible if the conditions are not favorable or take advantage of unusual opportunities even if there are no needs.

The Market Timing Theory explains that a company with greater financial leverage obtains funds when its market value is low. Companies that show lower leverage capture external resources when the market value is high (^{Baker; Würgler, 2002}). ^{Hovakimiam (2006)} stated that in a warm market the issuance of shares that are held exerts no influences on capital structure. In contrast, ^{Alti (2006)} shows that in a warm market companies enjoy the opportunity to issue more shares, overvaluing them and reducing debt. However, after overvaluation period companies increase their debts.

However, ^{Alti (2006)} also shows that the theory is an important determinant of financing activity, but only for the short term.

*Agency theory*

Debt utilization is key to the emergence of conflicts of interest between shareholders and managers (^{Ross; Westerfield; Jaffe, 2002}). Thus, because of such conflicts, ^{Mendes et al. (2009)} show that the agency theory can be understood as the costs that come from the purpose of resolution of these conflicts, since managers and shareholders intervene in defense of their own interests.

^{Grinblatt and Titmman (2005)} elucidate that high company debt may be more favorable for shareholders as it would limit managers to adopt effective measures in personal terms and to expand the company more rapidly than advisable. On the other hand, managers prefer a smaller debt than the optimal point because it reduces the bankruptcy risk and also does not restrict management.

According to this theory, the optimal debt is defined by minimizing the total agency cost (^{Mendes et al., 2009}).

*The economic crisis variable*

Effects of the globalization of economic crises have many critical consequences for public companies (^{Berger; Turtle 2011}). According to ^{Espinola (2013)} the cause of crisis is related to financial bubbles. They create macroeconomic effects in chains that affect several countries. Their influence is negative on companies. This causes insolvency, affects interest rate, growth, and other microeconomic factors.

The study indicates financial innovations as one of the determinants of crisis. This is related to the nature of innovation in the competitive financial services industry (^{Thakor, 2012}). More generally, ^{Mitchell (1941)} pointed out that the scope of business crises is due to the fact they are an intrinsic part of the corporate market.

Therefore, global financial crises make it difficult for companies to maintain their capital structures. It can make companies make the wrong decisions (^{Espinola, 2013}).

Methodology

This methodology was based on papers about public companies’ capital structure. The objects of study were companies listed in the Brazilian stock exchange from 2007 to 2015. The sample consisted of all public companies in the aforementioned period. However, companies which had no financial statements available for the entire period of analysis were eliminated from the sample, therefore remaining 114 companies. The panel consists of 1,026 observations: 228 in crisis periods (2008 and 2015) and 798 in normality periods. The data were taken from the Bloomberg database.

Considering the research focus on evaluating the determinants of the capital structure in financial crisis periods, Brazilian companies were considered because of the crises that the country faced in 2008 (financial crisis) and 2015 (political and economic crisis).

*Operational definition of variables*

Variables analyzed in the present study were divided between dependent and independent variables.

*Dependent variables*

Regarding the dependent variables, the research was based on previous studies such as ^{Shyam Sunder and Myers (1999)}, ^{Frank and Goyal (2003)}, ^{Leary and Roberts (2010)}, and ^{Espinola (2013)}. Debt indicators were treated at book value.

Dependent Variable |
Formula |
---|---|

TD: Total Debt | =(Short-Term Debt+ Long-Term Debt) / Total Asset |

STD: Short-Term Debt | = Short-Term Debt / Total Asset |

LTD: Long-Term Debt | = Long-Term Debt / Total Asset |

Total Debt: it indicates the percentage of short-term and long-term debt in relation to the company’s assets; the higher this percentage, the higher the company’s leverage level. When the TD index is high, it shows that there is great dependence on debt.

Short-Term Debt: it indicates the percentage of short-term debt in relation to the company’s assets.

Long-Term Debt: it indicates the percentage of long-term debt in relation to the company’s assets.

The three variables of indebtedness were chosen to evaluate whether there are specific drivers, according to the type of indebtedness. The financial literature has more research evaluating the long-term debt. Short-term indebtedness is little discussed in the literature.

*Independent variables*

The definition of independent variables was performed through an adaptation of empirical studies (Booth et al., 2001; ^{Brito; Corrar; Batistella., 2007}, ^{Nakamura et al., 2007}; ^{Bastos; Nakamura, 2009}; ^{Frank; Goyal, 2003}; ^{Espinola, 2013}).

Independent Variable |
Formula |
---|---|

Operating margin | = Operating Incomes / Total Revenue |

Size | = In Total Revenue |

Tangibility | = Permanent Asset / Total Asset |

ROE: Return On Equity | = Net Profit / Equity |

Growth | = Total Revenue (t) / Total Revenue (t-1) |

Tobin's Q ratio | = (Total Market Value + Total Debt) / Total Asset |

Altman Z-score | =1.2X1 + 1.4X2 + 3.3X3 + 0.6X4 + 1.0X5 |

Note: EBIT corresponds to Earnings Before Interest and Taxes. X1 corresponds to working capital / total assets. X2 corresponds to retained earnings / total assets. X3 corresponds to earnings before interest and taxes / total assets. X4 corresponds to market value of equity / book value of total liabilities. X5 corresponds to sales / total assets

Operating margin: it indicates the percentage between the operating incomes and revenues. According to the Pecking Order Theory, presented by ^{Brealey and Myers (2006)}, companies that show profits tend to use equity instead of debt. Therefore, their relation is inversely related to the level of indebtedness.

Size: expressed by the log of the revenues. By theory, larger firms tend to have low default risk, lowering the cost of borrowing. It has a direct effect on indebtedness.

Tangibility: it indicates the percentage of fixed assets in relation to total assets. Companies with more physical structures have greater tangibility.

ROE: it reflects the percentage of return on equity. Theories show that there is a discordance of analysis. According to ^{Jensen and Meckling (1976)}, because of Agency problems, managers seek to retain the company’s excess profits to make investments. Researchers show that companies committed to debt tend to have agency problems. Thus, according to the trade-off model, the most profitable companies should be the most leveraged. However, according to the Pecking Order Theory, described by ^{Myers (1984)}, excess of profit generates high indexes of reinvestment, which eliminates the necessity of searching for third-party capital. Therefore, the ratio is inverse to the level of leverage.

Growth: it indicates the percentage of growth of revenue over time. According to the Pecking Order Theory: the higher the revenue growth, the lower the need for debt.

Tobin’s Q ratio: this ratio was developed by ^{Tobin (1969)} and shows the relation between the market value, given by the expectation of future earnings, and the company’s book value. According to theories, companies with high indebtedness tend to add more market value.

Altman Z-score: The Z-score was developed by ^{Altman (1968)}. It is an indicator of credit risk. It is used to forecast bankruptcies of companies. It is calculated from multiple income and corporate balance values to measure the financial health of a company.

*Hypotheses for normality periods*

H

_{1}: The relationship between operating margin and debt is negative;H

_{2}: The relationship between size and debt is positive;H

_{3}: The relationship between tangibility and debt is negative;H

_{4}: The relationship between ROE and debt is negative;H

_{5}: The relationship between Altman Z-score and debt is negative;H

_{6}: The relationship between expected growth (Tobin’s Q ratio) and debt is positive;H

_{7}: The relationship between growth and debt is negative;

*Data processing and statistical tests*

In the study it was used a regression analysis with panel data, unbalanced. To this end, it was tested the hypothesis of two regression models, Fixed and Random, through the Hausman specification test, proposed by Hausman (1978). It was used Eviews 7 software for test and regression statistics. It has been widely used by many researchers in the area (^{Nakamura et al., 2007}, ^{Brito, Corrar; Batistella, 2007}; ^{Bastos; Nakamura, 2009}; ^{Frank; Goyal, 2003}; ^{Almeida et al., 2004}; ^{Espinola, 2013}).

Regression model:

where:

*Validation of the regression model*

The validation of the regression model was made by testing for the absence of multicollinearity and independence of residuals. Multicollinearity is the phenomenon of correlation between variables. Their evidence occurs when the R² of the regression is high and the variables are not significant. To this end, we analyzed the correlation matrix between the variables and the significance of the parameters. Residual dependence occurs when the regression’s errors are autocorrelated. To this end, the serial autocorrelation test proposed by ^{Durbin and Watson (1971)} was used.

Results and discussion

The correlation matrix was performed (Table 3 and Table 4) between the independent variables in order to determine multicollinearity (if it is established, there is a problem of almost exact linear relationship between these variables, reducing the quality of the regression). However, there is a correlation between ROE and Operating Margin variables. The variables have the same input (net profit) in the calculation. However, as the correlation was lower than 0.5, as recommended by ^{Gujarati and Porter (2008}) the effect of multicollinearity cannot be disregarded.

Operating Margin | Size | Tangibility | ROE | Growth | Tobin's Q | Altman Z | |
---|---|---|---|---|---|---|---|

Operating Margin |
1.000 | ||||||

Size |
0.186 | 1.000 | |||||

Tangibility |
-0.063 | 0.160 | 1.000 | ||||

ROE |
0.314 | 0.148 | -0.061 | 1.000 | |||

Growth |
-0.047 | -0.053 | -0.038 | 0.005 | 1.000 | ||

Tobin's Q |
0.012 | 0.036 | 0.027 | 0.043 | 0.000 | 1.000 | |

Altman Z |
0.057 | 0.100 | -0.041 | 0.164 | -0.018 | 0.037 | 1.000 |

Operating Margin | Size | Tangibility | ROE | Growth | Tobin's Q | Altman Z | |
---|---|---|---|---|---|---|---|

Operating Margin |
1.000 | ||||||

Size |
0.192 | 1.000 | |||||

Tangibility |
0.103 | 0.152 | 1.000 | ||||

ROE |
0.256 | 0.134 | 0.112 | 1.000 | |||

Growth |
-0.060 | -0.208 | -0.109 | -0.007 | 1.000 | ||

Tobin's Q |
0.135 | 0.173 | 0.068 | 0.125 | -0.083 | 1.000 | |

Altman Z |
0.236 | 0.088 | -0.120 | 0.341 | 0.093 | 0.262 | 1.000 |

The low values of the correlations between the dependent variables suggest that there is no multicollinearity effect on the data. Thus, it is not necessary to test multicollinearity using the ^{Farrar and Glauber (1967)} test.

It was also performed the Durbin-Watson test (Table 5), which is based on the assumption that the errors in the regression model are generated from the previous observation, i.e., it correlates the errors of the current regression model with the previous one. Results close to 2 indicate independence of the residuals, validating the proposed regression model (^{Brito, 2007}).

Normality Period | Crisis Period | ||
---|---|---|---|

Durbin-Watson |
Short-Term Debt | 2.07 | 1.83 |

Long-Term Debt | 1.97 | 1.90 | |

Total Debt | 2.04 | 2.06 |

Durbin-Watson statistic for all models indicated the acceptance of null hypotheses for 1% levels. It indicates that regression residuals are independent. There is no serial autocorrelation of residuals. DW statistic numbers are between 1.746 and 2.397 (4-DL).

*Descriptive statistics*

Table 6 shows the mean and standard deviation of the debt variables (dependent variables), which were calculated at book value, and Table 7 shows independent variables: operating margin, size, tangibility, ROE, growth, Tobin’s Q, and Altman Z.

Normality Period | Crisis Period | |||
---|---|---|---|---|

Dependent Variable | Mean | Standard Deviation | Mean | Standard Deviation |

Short-Term Debt | 0.091 | 0.077 | 0.103 | 0.096 |

Long-Term Debt | 0.203 | 0.131 | 0.226 | 0.169 |

Total Debt | 0.294 | 0.157 | 0.329 | 0.193 |

Normality Period | Crisis Period | |||
---|---|---|---|---|

Independent variable | Mean | Standard Deviation | Mean | Standard Deviation |

Operating margin | 0.112 | 0.412 | 0.111 | 0.202 |

Size | 21.516 | 1.600 | 21.651 | 1.644 |

Tangibility | 0.412 | 0.334 | 0.470 | 0.342 |

ROE | 0.105 | 0.253 | 0.042 | 0.390 |

Growth | 0.269 | 1.376 | 0.627 | 5.064 |

Tobin's Q | 1.750 | 7.816 | 1.106 | 0.550 |

Altman Z | 2.002 | 1.980 | 2.059 | 2.006 |

Debt indicators show changes during the analyzed periods. It can be observed that in times of crisis the means are higher. It shows that debt was higher in this period. The standard deviation does not influence the analysis, since the sample was diverse and composed of companies from various industries and sizes.

In the descriptive statistics of the independent variables, it may be observed that all have changed in the compared periods. The operating margin shows a substantial change. It shows negative percentage in times of global crisis and in Brazilian economic recession. Considering the other indicators, several show lower mean in the crisis period; thus it may be considered that these indicators were affected by government and situational macro environment.

*Estimated coefficients*

Table 8 shows the results of the Hausman specification test statistics.

Normality Period | Crisis Period | |||||
---|---|---|---|---|---|---|

Dependent Variable | Chi-Sq. Statistic |
Prob. | Effect. Indicated |
Chi-Sq. Statistic |
Prob. | Effect. Indicated |

Short-Term Debt | 4.832 | 0.681 | Random | 10.584 | 0.158 | Random |

Long-Term Debt | 16.197 | 0.023 | Fixed | 13.947 | 0.052 | Random |

Total Debt | 13.981 | 0.052 | Random | 18.103 | 0.012 | Fixed |

The results of Long-Term Debt in normality period and Total Debt in crisis period reject the null hypothesis. It shows that there are correlations between the regressors and error term. It indicates to estimate of models in the fixed form. The other models, however, accept the null hypothesis, indicating for an estimate of random models.

Tables 9, 10, and 11 consider the estimated coefficients of the proposed regression model of dependent variables (short-term debt, long-term debt, and total debt). In the Short-Term Debt model, the variables that have significance effects are those with p-value lower than 10%. Analyzing the normality period, operating margin, size, ROE, growth, and Altman Z were the variables that influenced the dependent variable. In the long-term debt it was noticed that operating margin, tangibility, ROE, growth, and Tobin’s Q were not significant. Finally, analyzing the total debt, operating margin, growth, and Tobin’s Q did not influence the regression model.

Normality Period | Crisis Period | |||||
---|---|---|---|---|---|---|

Coeff. | t-stat | p-value | Coeff. | t-stat | p-value | |

Const | 0.239 | 6.366 | 0.000 | 0.342 | 3.722 | 0.000 |

Operating margin | -0.016 | -2.297 | 0.022 | -0.034 | -0.960 | 0.339 |

Size | -0.006 | -3.169 | 0.002 | -0.012 | -2.695 | 0.008 |

Tangibility | -0.012 | -1.410 | 0.159 | -0.015 | -0.746 | 0.457 |

ROE | -0.058 | -5.006 | 0.000 | 0.035 | 1.856 | 0.065 |

Growth | 0.003 | 1.660 | 0.097 | -0.002 | -1.379 | 0.170 |

Tobin's Q | 0.000 | -0.509 | 0.611 | 0.044 | 3.396 | 0.001 |

Altman Z | -0.007 | -5.326 | 0.000 | -0.012 | -3.272 | 0.001 |

R-squared | 0.140 | 0.146 | ||||

Adjusted R-squared | 0.131 | 0.112 | ||||

F-statistic | 16.228 | 4.236 | ||||

Prob (F-statistic) | 0.000 | 0.000 | ||||

Effects Specification | Random | Random |

Normality Period | Crisis Period | |||||
---|---|---|---|---|---|---|

Coeff. | t-stat | p-value | Coeff. | t-stat | p-value | |

Const | -0.202 | -2.888 | 0.004 | -0.297 | -1.998 | 0.047 |

Operating margin | -0.013 | -1.015 | 0.310 | -0.046 | -0.816 | 0.416 |

Size | 0.021 | 6.329 | 0.000 | 0.019 | 2.797 | 0.006 |

Tangibility | -0.016 | -1.065 | 0.287 | 0.034 | 1.073 | 0.285 |

ROE | -0.007 | -0.293 | 0.769 | 0.013 | 0.424 | 0.672 |

Growth | 0.002 | 0.560 | 0.575 | 0.000 | -0.197 | 0.844 |

Tobin's Q | 0.000 | -0.179 | 0.858 | 0.128 | 6.186 | 0.000 |

Altman Z | -0.019 | -7.815 | 0.000 | -0.026 | -4.274 | 0.000 |

R-squared | 0.291 | 0.254 | ||||

Adjusted R-squared | 0.146 | 0.224 | ||||

F-statistic | 2.002 | 8.429 | ||||

Prob (F-statistic) | 0.000 | 0.000 | ||||

Effects Specification | Fixed | Random |

Normality Period | Crisis Period | |||||
---|---|---|---|---|---|---|

Coeff. | t-stat | p-value | Coeff. | t-stat | p-value | |

Const | -0.022 | -0.289 | 0.773 | -0.189 | -0.733 | 0.466 |

Operating margin | -0.021 | -1.509 | 0.132 | 0.088 | 0.997 | 0.322 |

Size | 0.018 | 5.135 | 0.000 | 0.018 | 1.519 | 0.134 |

Tangibility | -0.035 | -2.132 | 0.033 | 0.028 | 0.604 | 0.548 |

ROE | -0.091 | -3.938 | 0.000 | -0.313 | -3.051 | 0.003 |

Growth | 0.004 | 0.959 | 0.338 | -0.029 | -1.179 | 0.243 |

Tobin's Q | 0.000 | -0.500 | 0.617 | 0.199 | 5.553 | 0.000 |

Altman Z | -0.025 | -8.930 | 0.000 | -0.038 | -4.385 | 0.000 |

R-squared | 0.156 | 0.857 | ||||

Adjusted R-squared | 0.147 | 0.604 | ||||

F-statistic | 18.402 | 3.396 | ||||

Prob (F-statistic) | 0.000 | 0.000 | ||||

Effects Specification | Random | Fixed |

The behavior of indicators in crisis period was similar to the normality period. Both indicators were statistically significant. Independent variables with p-values greater than 10% are not statistically significant. In both periods the variables that most influenced the independent variables and had more differentiation between the periods were operating margin, growth, and Tobin’s Q.

When the variable becomes statistically insignificant in a period, this variable is discarded, as it is not possible to consider it in comparison.

In the long-term debt (for the normality period), size is determinant of capital structure; However, in crisis period, Tobins’s Q also becomes a determinant of long-term debt, confirming the findings of ^{Brito, Corrar, and Batistella (2007)}.

*Hypotheses analysis*

Through the analysis of the regression coefficients, it was possible to determine the relationship between the dependent variables and each of the explanatory factors (independent variables). Negative sign (-) shows a negative relationship, positive sign (+) shows a positive relationship and SI (statistically insignificant) means that there was no significance. Table 12 shows three columns of data: the first shows the expected relationship according to the hypotheses mentioned in the Methodology, the second shows the observed relationship in the normality period, and, finally, the third shows the observed relationship in the crisis period.

Variable | Expected Relationship | Observed Relationship in the normality periods (2007, 2009-2014) | Observed Relationship in the crisis periods (2008 and 2015) | ||||
---|---|---|---|---|---|---|---|

Debt | Short-term Debt | Long-Term Debt | Total Debt | Short-term Debt | Long-Term Debt | Total Debt | |

Operating margin | - | - | SI | SI | SI | SI | SI |

Size | + | - | + | + | - | + | SI |

Tangibility | - | SI | SI | - | SI | SI | SI |

ROE | - | - | SI | - | + | SI | - |

Growth | - | + | SI | SI | SI | SI | SI |

Tobin's Q | + | SI | SI | SI | + | + | + |

Altman Z | - | - | - | - | - | - | - |

The result shows that the determinants of the capital structures are different between the normality period and the crisis period. Some variables did not reject the hypotheses mentioned in the Methodology. Following the premises of the Pecking Order Theory, the Operating margin variable, which measures the operating profit margin, should be negative.^{Brealey and Myers (2006)} point out that, according to the Pecking Order Theory, companies with higher operating margin tend to finance themselves (reinvestment) rather than through third-party capital.Table 12 shows that in Brazil this assumption is true, since the coefficient is negative. The greater the company’s operating margin, the lower the debt. However, in the normality period, the regression model showed a negative relationship: thus the hypothesis was not rejected. In fact, the model showed that there is a proportional relationship between the operating margin and debt indicators, but it was observed only in normality periods for a short-term debt. In other models statistics were not significant.

The Size variable also adhered well to the assumptions of theories. The great size of a company makes it easier to obtain financing, making the company more leveraged. These results are also in agreement with studies of ^{Ozkan (2001)} and ^{Gaud et al. (2005)}. However, in the models with Short-Term Debt variable this hypothesis is rejected. The same occurred with the Tangibility variable. Less tangible companies tend to have higher level of leverage. However, the tangibility parameter can assume positive or negative values, as presented by ^{Titman & Wessels (1988)}.

In relation to Tobin’s Q, which demonstrates the expected investment opportunity, it is observed that the positive values of Tobin’s Q corroborate the Pecking Order Theory, which states that the investment opportunity ratio should be positive in relation to the leverage level.

Although the Growth variable was not significant in crisis periods, the positive value in the normality period does not corroborate the Pecking Order Theory. The higher the revenue growth, the lower will be the dependence on debts. Therefore, this leads to a reduction in the leverage level.

The ROE variable also confirms the Pecking Order Theory, which states that more profitable companies are less indebted. This theory argues that companies should not seek external financing, without first exhausting their own resources. Therefore, higher operating margin tends to decrease leverage.

Similar to the ROE variable, the Altman Z variable is also in agreement with the theory. High scores reduce the level of credit risk and leverage. Therefore, if companies have more debts, they will have worse scores of credit risk.

It may be observed that the relationship of some dependent variables (short-term debt, long-term debt, and total debt) with the independent variables in the analyzed periods was different. Thus, the negative macroeconomic influences affect the determinants of the capital structures. ^{Espinola (2013)} used a sample of US companies and confirmed that in times of financial crisis determinants also changed, and operating margin, size, and tangibility had more significance than the others.

The initial hypotheses of variables (operating margin and ROE) were not rejected. Regarding the operating margin variable, ^{Myers (1984)}, ^{Donaldson (1961)}, and ^{Brealey and Myers (2006)} suggest that firms prefer raising capital, firstly from retained earnings, secondly from debt, and thirdly from issuing new equity. This relationship occurs because of the transaction costs involved in issuing equity, according to ^{Myers and Majluf (1984)}. However, in periods of economic expansion in Brazil, companies tend to use both equity and short-term debts. During this period, the Brazilian Development Bank (BNDES) provides credit at low costs.

Conclusions

This paper analyzed the determinants of the capital structures of Brazilian companies. In addition, the study compared different economic periods (normality and crisis) in order to determine whether there was variation in indebtedness related to the independent variables. The results of the tests indicated a change in the ratio of some independent variables.

One of the contributions of this research is demonstrating that through statistical tests it is possible to observe changes in the capital structure determinants in different economic periods. In addition, it is emphasized that the factors chosen for the research show only some variables that influence corporate debt. The variables that explain the debt in the crisis period were size, ROE, Tobin’s Q, and Altman Z. According to ^{Espinola (2013)}, size was also a very significant variable in determining the capital structure.

Many of the hypotheses mentioned in the Methodology were not rejected by determining correlation with capital structure theories. The hypotheses were more substantiated in the Pecking Order Theory, Static Trade-off Theory, and Agency Theory.

The aims of research were achieved: through the proposed regression model, we found capital structure determinants of Brazilian publicly-held companies. The general hypothesis built in the introduction was not rejected because there was a change in the relationship of the capital structure determinants when comparing the normality period (2007, 2009-2014) with the crisis period (2008 and 2015).

Finally, this research field is large and diverse. Therefore, there is opportunity to develop new capital structure theories or delve into the existing theories, adapting them to increase the value of companies based on the choice of optimal debt point. In addition, this research is restricted to Brazilian companies. We also suggest testing the hypotheses presented here in companies of Latin American countries. Thus, it will be possible to determine whether the capital structure theories are also divergent from the American reality.