“Tax aggressiveness and CEO overconfidence in the stock market: Evidence from Brazil”

This study examines the association between tax aggressiveness and overconfidence in 277 Brazilian stock market listed companies from 2010 to 2017, with the supposition (based on optimal capital ownership structure theory) that the greater a manager’s over- confidence, the more aggressive the company’s tax decisions. Overconfidence is measured in an innovative way in which normalizing excess acquisitions and excess invest- ments using the company’s market value and then combining these two variables with indebtedness to capture, more directly, the possible effects of overconfidence on the cor- poration’s operations. Tax aggressiveness is computed using a tax burden on earnings and value-added. The variables included in the model were obtained from data contained in the selected companies’ financial statements. Data analysis was performed by multiple linear regression. The parameters used combined and fixed effects methods to identify an association between tax aggressiveness and overconfidence. Data related to corporate governance, CEO’s characteristics, and capital concentration were used as control vari- ables. The study’s main finding does not show any significant relationship between fiscal aggressiveness and overconfidence; however, they did show a significant association with tax aggressiveness, the company’s size, the return on shares, and the education level of the CEO. An interesting finding of the robustness tests is the stationarity of tax aggressive- ness, which could partially explain the non-significance of the main finding.


INTRODUCTION
Tax aggressiveness is influenced by issues related to the separation of ownership and control (Slemrod, 2004). In general, shareholders expect managers who act on their behalf to focus on maximizing profits, including taking actions that result in reduced tax liabilities. However, the separation between ownership and control can generate corporate tax decisions that reflect an administrator's private interests (Hanlon & Heitzman, 2010).
Separation of ownership and control, a context in which owners delegate the power of decisions about their assets to a third party (Berle & Means, 1991), causes conflicts of interest. Because an agent does not necessarily make decisions compatible with the principal's interests, formalized compensation contracts to align the interests of the two parties are required, which generates agency costs. Agency costs are incurred to address issues related to ownership structure, management incentives, corporate governance, and the characteristics of the organization's chief executive (Wilde & Wilson, 2018).
Overconfidence is considered one of the characteristics that lead company managers to develop greater tax aggressiveness (Hsieh et al., Jensen and Meckling's (1976) theory of optimal capital ownership structure is adopted as the basis for this study's development. This theory predicts that corporate governance mechanisms will align the interests of managers and investors. In the context of this study, tax aggressiveness would result from managerial behaviors that conflict with investors' interests since reducing the tax burden on the company's operations could destroy the firm's value in the long run. Therefore, this study hypothesizes is that managerial overconfidence will result in greater tax aggressiveness in the sample companies.

Determinants of tax aggressiveness
A conceptual structure distinguishes categories of tax aggression determinants: a) implementation costs; b) costs of results; and c) agency costs. Agency costs are those associated with minimizing the conflicts of interest between executives and shareholders generated by the separation of capital and property, such as the firm's ownership structure, management incentives, corporate governance, and executive characteristics (Wilde & Wilson, 2018).
The costs associated with tax aggressiveness may differ depending on a firm's ownership structure; for example, the costs are less than the benefits for family businesses in a concentrated property environment. Thus, it is expected that there will be a higher level of tax aggressiveness in family businesses. For public companies, the benefits seem greater than the associated costs, so public companies are expected to be tax aggressive. Finally, multinational companies generally pay low taxes in their host countries (Annuar et al., 2014).
Also, companies that compensate their managers with incentives based on performance after taxes tend to be more tax aggressive (Hanlon & Heitzman, 2010). In contrast, companies with higher internal controls levels are expected to be less prone to tax aggressiveness (Desai & Dharmapala, 2006

Overconfidence
Overconfidence is defined as an optimistic estimate of one's own abilities and the results related to one's situation, thus characterizing the "better than average" effect (Langer, 1975). In general, individuals overestimate their ability to control results and underestimate the probability of failure (Alicke et al., 1995;Langer, 1975;March & Shapira, 1987;Weinstein, 1980 According to the literature, this study tests the hypothesis: H1 Overconfidence is positively associated with tax aggressiveness.

METHODS
The data were analyzed using panel data analysis, respecting all premises and specifications appropriate for the examined corporations. Also, tests were carried out to verify the consistency of the results obtained. In all cases, the coefficients obtained were controlled using the traditional determinants found in finance studies.
The sample of companies is composed of 277 Brazilian companies listed on the B3 Stock Exchange, excluding financial institutions. The variables included in the model were obtained from data contained in the selected companies' financial statements. The dependent and independent variables are composed using financial data collected from the Economatica® database. The control variables, except market to book, size, and return, are composed of non-financial data collected from the ComDinheiro database.
The effective tax rate (ETR) is calculated by dividing tax expense by earnings before taxes (Armstrong et al., 2012). The cash effective tax rate (CASHETR) measures the effective cash taxes paid during the year instead of using the tax expense measured on an accrual basis (Dyreng et al., 2008).
Metrics of tax aggressiveness that only measure the impact of taxes on income can limit the results in a hypothesis where indirect taxation exists, as is the case in Brazil. Thus, the CTAadj measure includes direct taxes, calculated on earnings, and indirect taxes, calculated on aggregate values. The measure uses the division between total taxes and contributions, disclosed in the Value-Added Statement (DVA), and the company's gross revenue, also disclosed in the DVA (Ignacio, 2018).
The overconfidence variable proposed in this study is based on the measure created by Schrand and Zechman (2012). In this study, excess acquisitions and excess investments are normalized using the company's market value; these two variables were then combined with indebtedness.
The control variables selected for the tax aggressiveness model are those related to agency conflicts that arise from the separation of ownership and control: corporate governance, ownership structure, the concentration of capital, and executive characteristics.
The variables Family, Government, and ForeignCapital represent companies controlled by members of a family, the government, and international investors, respectively. MainShareholder indicates the control is concentrated in the main shareholder. FEM indicates that the company's CEO is female. Executive education is indicated by the variables Graduation, MBA, MSc, and Ph.D., referring to undergraduate degree, master's degree in business administration, master's degree, and doctorate. CEO_Duality indicates whether the CEO is a company founder, and CEO_Turnover represents the length of time the CEO has been in office.

Descriptive statistics
The descriptive statistics of the variables and their correlations, found in this section, are followed by the model estimation results.  Table 3 shows the analysis of the ownership concentration variables suggesting similar proportions of family and foreign companies in the sample of 13.38% and 12.20%, respectively, with public companies representing about 5.50% of the sample. Besides, less than 50% of the companies have the main shareholder. These results seem to indicate a sample composed of professional companies whose capital is dispersed.
As for personal characteristics, just over 3% of the observations used in the data sample have female CEOs. This confirms the result of another study by Dalcero, Fabrício, and Ferreira (2020) that investigates the influence of gender on accrual quality, which indicated approximately 3% of CEOs in a  An analysis of the correlations shown in Table 4 indicates no strong linear association between the variables used in the tax aggressiveness model and overconfidence and between the variables used in the robustness tests. Such results suggest the possibility of no significance between the The correlations between the variables CASHETR, ETR, and CTAadj and OverConfidence are low, indicating no association between tax aggressiveness and overconfidence.
The corporate governance variables Market to book, Size, and Return even show low correlation with the tax aggressiveness and overconfidence variables; this seems to make sense regarding the trend of expectations in the literature. The ownership concentration variables confirm the expectation that family and public companies are more tax aggressive. CASHETR and ETR confirm the expectation that foreign companies pay less tax in host countries.
The ETR and CTAadj measures support the expectation that female executives are less tax aggressive. The levels of MBA and MSc training confirm the expectation that a higher level of knowledge results in lower taxation for the variables CASHETR and CTAadj. The variable CEO_Duality tends to reverse for ETR and CTAadj, meaning that there could be greater tax aggressiveness if the executive is the company's founder. Finally, the CEO_ Turnover variable has a reverse trend in the ETR tax measure, meaning that higher executive turnover decreases the tax burden, representing greater aggressiveness in the payment of taxes.

Tax aggressiveness and overconfidence
The analysis in Table 5 shows that overconfidence does not explain tax aggressiveness for any of the measures used -CASHETR, ETR, and CTAadjsince the OverConfidence variable is not significant in any of the models presented.

Robustness tests
To verify the consistency of the non-significance observed in the main model, the other specifications examined internal variables that could influence the tax aggressiveness of the sample corporations, including the personal characteristics of managers, capital concentration, and CEO duality and turnover. Table 6 presents the results of the models, including the variables corresponding to CEO characteristics.  For the CASHETR and ETR measurements, none of the response variables show significant results. The constant and Ph.D. and Size variables are significant for CTAadj. These results indicate that the higher the CEO's level of education, the greater the company's tax aggressiveness, and the larger the company, the less tax aggressive it is. Table 7 presents the results of the models with corporate governance variables that demonstrate significance in response to the measures of tax aggressiveness. Due to the perfect multicollinearity with the property concentration indexes, the results of this table also apply to the analysis of the possible effects of concentrated capital.

Corporate governance and ownership concentration
For the CASHETR and ETR measures, the Return variable is significant, indicating that greater tax aggressiveness is associated with a greater return on shares. In comparison, for the CTAadj measure, the constant and the Size variable are significant, indicating that larger companies are less tax aggressive.
The effect of property concentration was especially likely for the CASHETR model, but it was not significant; further, there are no other significant variables in that model. For the ETR model, the variable Return is significant, indicating greater tax aggressiveness generates a greater return on shares. For the CTAadj measure, the constant and the Size variable are significant, with the Size variable indicating larger companies are less tax aggressive. The Return variable is significant, indicating more tax-aggressive companies have a higher return on shares. Table 8 presents the results of the tax aggressiveness models considering CEO duality and turnover in the sample companies.

Duality and turnover
Duality indicates independence between the chairman and the CEO, an appropriate corporate governance mechanism. For this reason, the expectation is that greater duality would result in less aggressiveness. Also, turnover represents the length of the CEO's term in the corporation: the longer the term, the less tax aggressiveness there would be.
No significant variables were found for the CASHETR and ETR models. For the CTAadj tax measure, the constant and the Size variable are Note: * 5% < p-value < 10%; *** p-value < 1%.

Stationarity
There is a possibility that overconfidence and tax aggressiveness may not occur in the same period, or that tax aggressiveness is a characteristic that does not depend on other variables. One way to measure the exogeneity of tax aggressiveness is by regressing the variable itself with its coefficients from the previous period. The results in all models analyzed suggest that current tax aggressiveness is significantly influenced by tax aggressiveness in previous periods, varying only in the direction of the association (Table 9).
For the CASHETR and CTAadj measures, the time-adjusted variable is significant; however, the interpretation of these measures differs. For CASHETR, the time-lagged variable indicates an inverse relationship with the current variable, which means that the company tends to decrease aggression after a year with intense tax aggressiveness. For CTAadj, the time-adjusted variable suggests a direct relationship, indicating increasing tax aggressiveness.

DISCUSSION
The  Finally, the results of the CASHETR and ETR models, indicating stationarity with a negative association, may indicate that Brazilian corporations have a tax target, possibly analogous to that suggested by capital structure trade-off theory. If such evidence is confirmed, a new research field in taxation may be opened related to the optimal taxation structure.

CONCLUSION
The present study examined the association between tax aggressiveness and overconfidence in a sample of 277 Brazilian corporations for over eight years. The assumption was that there would be a positive association between tax aggressiveness and overconfidence, considering the theory of the optimal structure of capital ownership since tax aggressiveness can result from the misalignment of the interests of managers and investors, with the former assuming higher tax risk increasing the variable portion of their compensation at the expense of creating long-term value for investors. However, this hypothesis cannot be confirmed given the lack of significance observed in the various models and specifications used for the data analysis, including those in most of the robustness.
The theoretical implications could indicate that sample corporations have corporate governance mechanisms that appropriately align the interests of managers and investors. Since the results show no association between tax aggressiveness and excess risk-taking by managers, tax aggressiveness would create value for investors.
This possibility becomes more robust considering the stationarity of the tax aggressiveness variables. This suggests corporations have an optimal taxation structure that, in theory, would maintain taxation at levels investors accept and that allow the corporation to continue to create value. Therefore, this supposedly optimal tax structure can become a useful field for theoretical development in future research.
Concomitantly, this evidence was obtained from an analysis of the data's stationarity, which also enables directly applying time-series econometric techniques to future studies. Using time-series econometrics would have the same impact as adopting the overconfidence variable proposed in this study by combining the variables that have been individually and less accurately examined in the literature.
There are obvious limitations in the present study. Although the tax measures used are almost standard in the literature on tax aggressiveness, there is a possibility, albeit mitigated, that the measures fail to adequately capture the concept. Again, this is a relevant methodological issue that should be addressed in future studies.