The manipulation of financial information characterized the 2008 global financial crisis. In the case of Lehman Brothers, which went bankrupt during this crisis, such manipulation was aimed at omitting liabilities from its financial statements. In Brazil, financial frauds involving financial institutions occurred in Banco Nacional, Noroeste, and Panamericano, causing financial losses to their investors (Dantas & Medeiros, 2015; Martins & Ventura Júnior, 2020). That becomes relevant as Brazil has the five largest banks in Latin America regarding assets (Standard and Poor's, 2022).

This context refers to problems of market inefficiency, especially those associated with information asymmetry (Coase, 1937, 2013; Akerlof, 1970). These problems are established in two moments: before the asset's acquisition (adverse selection) and after (moral hazard). At first, investors may incur adverse selection as they do not have a level of information equivalent to that of asset managers. In addition, all information about the asset is not reflected in its price. Therefore, it will be up to the investor to acquire the asset and assume the risk of eventual losses (Akerlof, 1970). Secondly, when acquiring the asset, the investor will be exposed to moral hazard arising from the misalignment of interests between their expectations and the decisions made by the asset managers (Arrow, 1963).

Quality financial information mitigates information asymmetry problems by reducing the cost of capital and the cost of monitoring asset managers (Leuz & Verrecchia, 2000; Verdi, 2006; Daske et al., 2008; Kim, Taylor & Verrecchia, 2021). However, accounting manipulations are more problematic in banks than in non-financial companies due to their high leverage and dependence on depositors and central banks for financing. Such manipulations occur, especially, in loan loss allowance accounts (Salem, Usman & Ezeani, 2021). In addition, factors such as corporate governance, auditing, regulations and enforcement can also affect the quality of companies' financial information, including banks (Moura et al., 2017; Abed et al., 2022).

Therefore, this study aimed to analyze the determinants of the quality of financial information of banks listed on Brazil Bolsa Balcão (B3). To that end, we began our research by measuring the quality of financial information from 20 banks listed on B3 through a set of 12 indicators from the study by Gabriel and Silveira (2011). Then, an econometric model was used to identify the factors determining financial information quality.

The results of this study indicate that a differentiated corporate governance index is not enough to improve the quality of the accounting statements by itself and, consequently, it cannot mitigate the problems of information asymmetry. This in turn means that it requires more effective monitoring through an audit carried out by the Big Four and an audit committee in the banks. In addition, it shows that the lower the indebtedness, the lower the probability of manipulations in the accounting statements of banks.

The contribution of this study is to include Brazilian banks in quality of accounting information research, given that previous studies were carried out only in non-financial companies in the Brazilian context (Moura, Mazzioni, & Ziliotto, 2016; Nunes et al., 2016; Silva, Mazzioni & Vargas, 2020). It also reveals that adopting an audit committee and the type of audit are essential factors in improving the quality of banks’ financial information. This can help Brazilian supervisory and regulatory bodies, especially the Central Bank of Brazil (BACEN), to identify the practices that may improve the financial information disclosed by banks to their investors.

The study proceeds in the following structure. Section 2 presents the literature review and the construction of the hypothesis. Section 3 demonstrates the methodology used in the study and section 4 the respective results. Lastly, section 5 presents the final considerations.

Adam Smith’s Classical Theory did not consider the behavioral problem in organizations, dealt with in the work of Coase (1937), which maintains that the firm is configured as a set of contracts between factors of production driven by self-interest. Later, Williamson (1993) deepened the research on the behavioral problem when dealing with market failure, which comes from bounded rationality, uncertainty and opportunism, increasing transaction costs. Finally, Fama (1980) explains that "behavioral" and "managerial" theories of the company were developed, which reject the classical model of entrepreneur or manager-entrepreneur treated in the classical approach since there was a separation between ownership and control of the company, as a way of seeking efficiency in the economic organization within the perspective of the "set of contracts".

With the emergence of corporations, ownership and control assume different roles. The principal is no longer the holder of decision-making, which has become the agent's role (Berle & Means, 1988), configuring what Jensen and Meckling (1976) recognize as an agency problem. However, the agent will not always act in the principal's best interest, which makes the emergence of a conflict of interest possible. Thus, to minimize the divergences of interest, the principal may establish appropriate incentives for the agent, reflected in monitoring costs to limit the agent's opportunistic attitudes. Known as agency costs, in addition to the expenses with monitoring by the principal, there are expenses with contractual guarantees by the agent and the residual loss (Jensen & Meckling, 1976).

Monitoring costs are necessary, given that the agent has more information than the principal and other stakeholders, resulting in an asymmetry between the parties involved in this contractual relationship. Adverse selection and moral hazard correspond to two problems arising from information asymmetry. Adverse selection is an ex-ante problem because, at the negotiation time, the buyer does not have the same level of information as the seller. In addition, all information about the asset is not reflected in its price. Therefore, the buyer will pay an average price for the asset and bear the risk of losses. On the other hand, moral hazard is an ex-post problem and will come up if the agent takes advantage of the information asymmetry and acts opportunistically to obtain personal benefits at the expense of other stakeholders (Arrow, 1963; Akerlof, 1970).

The problems created by information asymmetry prevent the efficient allocation of resources in the capital market. To reduce the information asymmetry between managers and investors, the company must be efficient in disclosing information. Therefore, adopting corporate governance mechanisms is important since the literature suggests that such mechanisms can improve information transparency and mitigate information asymmetry problems (Healy & Palepu, 2001; Verrechia, 2001; Schiehll & Martins, 2016; Pardo López & Cortés, 2020).

A practical corporate governance framework is critical to increasing the credibility of financial reporting, as the authors found a negative relationship between audit committee effectiveness and the likelihood of fraudulent financial reporting (Razali & Arshad, 2014). Corporate governance mechanisms can help limit and explain some of the choices related to the identification, measurement and disclosure of economic-financial events, influencing the quality of financial information produced by the firm and disclosed to investors and other stakeholders (Gabriel & Silveira, 2011). That is, the higher the level of corporate governance, the higher the quality of financial information (Almeida, Lima & Lima, 2009; Seguí-Mas, Bollas-Araya & Polo-Garrido, 2018; Romero, Ruiz & Fernandez‐Feijoo, 2019).

The quality of financial information is responsible for the design of the primary business data in a transparent or even profound way, as it subsidizes managers in updating the organizational environment, providing the accuracy of actions relevant to this context, as well as its succession (Silva, Takamatsu, & Avelino, 2017). From this perspective, it is noteworthy that the quality of financial information has a direct relationship with the management, as they provide informational support to all areas of the organization, covering the stages of the management process, as pointed out by Queiroz and Almeida (2017).

Following this idea, it is possible to infer that the quality of financial information is essential for the business process, as it provides organizations with a foundation for planning, monitoring, communicating and controlling their actions through the handling and caring for financial information. All of these aspects represent the result of corporate financial and external reporting systems that measure and publicly disclose audited quantitative data relating to the financial position and performance of publicly traded companies, which may be reflected in share prices (Bushman & Smith, 2001; Farinha, Mateus, & Soares, 2018).

To assess the impact of adherence to financial information quality practices adapted to Brazilian companies, Gabriel and Silveira (2011) developed the Financial Information Quality Index (IQIC), validated internally by Cronbach's alpha and externally by eight researchers. The authors found that the shareholding concentration positively influences the IQIC and that the firm's corporate governance constitutes one of the main explanatory variables of the IQIC. Furthermore, the study shows the objective measurement of the quality of financial information and presents the result estimated by the Systemic Generalized Method of Moments (GMM). Lastly, it concludes that it is easier to overcome obstacles than to create a greater supply of resources for financing companies.

Moura, Mazzioni, & Ziliotto (2016) analyzed the determinants of the quality of financial information in publicly traded companies listed on the BMF&Bovespa, from 2008 to 2014, excluding those that carried out financial activities and did not have the information required for all the variables used. The study was descriptive and had a quantitative approach. The dependent variable was earnings management and the independent variables were the level of market competitiveness, proportions of intangible assets, ownership concentration and corporate governance. The results showed that, among the factors investigated in the period, only two confirmed the competitiveness and intangibility factors.

The study of Nunes et al. (2016) analyzed the quality of financial information in 9 mixed capital companies whose controller is the State and compared it to that produced by 9 private capital companies. A structured, quantitative and descriptive methodological approach is deployed, using document analysis to develop the research. The data collected were analyzed descriptively using the Mann-Whitney non-parametric statistical test. The results point to a non-significant difference in the quality of financial information between the two types of companies when examining the following factors: shareholder control; auditing; republication of accounting statements; preparation of financial information in an international standard; publication outside the legal deadline; revaluation of property, plant and equipment; disclosure of the Statement of Cash Flow; greater adoption of practices of disclosure of the Statement of Added Value than companies with private control; executive compensation and economic profit measure.

Moura et al. (2017) found that the fact that when a company is audited by the big four, has an audit committee, institutional investors' shareholding, and trades shares on the American stock exchange, it produces and discloses information of better quality. Martins and Ventura Júnior (2020) analyzed the influence of the corporate governance structure in mitigating the possibility of fraudulent financial reporting of Brazilian companies. They found that the corporate governance structure of the companies mitigate earnings manipulation. It was also noticed that the governance practices related to the board of directors were more efficient against the probability of bankruptcy. In contrast, the practices related to the audit were associated with the reduction of earnings manipulation.

Silva, Mazzioni & Vargas (2020) aimed to identify the factors that improve the quality of financial information of public utility companies listed on B3. They found that ownership concentration, company size and level of corporate governance contribute to improving the quality of financial information. Considering the empirical literature exposed and that banks have more significant incentives for earnings management due to their high leverage (Kanagaretnam, Lim & Lobo, 2010; Mangala & Singla, 2021; Salem, Usman & Ezeani, 2021), the following was established as our study’s hypothesis:

H1: Corporate governance mechanisms influence the quality of bank financial information.

Table 3 presents the descriptive statistics of the analysis, whose sample of selected variables is composed of 20 banks listed on B3 for four years, obtaining 80 observations in total. The dependent variable (IQIC) has a mean of 7.22 with a standard deviation of 0.86. This result confirms the similarity between the analyzed companies. The independent variable of shareholding concentration (CA) presented an average percentage of shares held by the largest shareholder corresponding to 53.51%. In its analysis, there was a standard deviation of 26.84%, the minimum rate of shares that the largest shareholder held is 7.07%, and the largest was very close to 100% with 99.98%.

The company size, operated as a proxy for the natural logarithm of the company's total assets, presented a maximum of 12.18 and a minimum of 7.88, with an average of 10.35 (corresponds to approximately USD 55,5 billion). The companies analyzed have an average of 74.45% of their total assets committed to short and long-term debt obligations, with a minimum of 0.71 and a maximum of 95.3. The return on equity (ROE) presented a minimum of -28.93, a maximum of 29.50, and an average of 10.53 with a standard deviation of 10.10.

Table 3. Table 3. Descriptive statistics of the variables used in the study (2014 to 2017) Source: elaborated by the authors. Note:

IQIC - Financial Information Quality Index; GC - Corporate Governance; TIPOAUD - Type of Audit; COMAUD - Audit Committee; CA - Shareholding Concentration; SIZE - Size; END - Indebtedness; ROE - Return on equity.

It was identified that only 2 banks weren’t audited by one of Brazil's four largest auditing companies between 2014 and 2017. One of them was ITAUSA (2015 to 2017) and another was ALFA HOLDING (2014). More than half of the companies, precisely 43, have an audit committee. Audit (COMAUD) with an average of 0.54 of the banking companies presenting 0.50 of their standard deviation and the 80 observations, only 32 companies were traded on the New Market and listed in levels 1 and 2 of corporate governance (GC). This corresponds to an average of 0.40 of banking companies with a standard deviation of 0.49. Regarding corporate governance, the result of this variable is justified by the fact that the analyzed companies present accounting statements different from the other publicly traded companies listed in B3, bearing in mind that these are financial companies.

Table 4 identifies the frequency for better analysis of the variables measured by the dummy variable.

Table 4. Table 4. Frequency of variables analyzed by the dummy variable (2014 to 2017) Source: elaborated by the authors. Note:

GC - Corporate Governance; TIPOAUD- Type of Audit; COMAUD - Audit Committee

As shown in Table 4, the independent variables CG, TIPOAUD, and COMAUD were measured by the dummy variable, where one corresponds to a positive response and 0 to a negative one.

The variable that stood out the most was TIPOAUD, with 76 positive responses, characterizing that 95% of audits were performed by the big four firms. According to the dummy variable, the second variable that scored the most positively was COMAUD, totaling that a little more than half of the companies have an audit committee with 43 positive responses. Then the GC variable was the least scored, showing that of the 80 observations, only 32 scored positively.

Table 5. Table 5. Matrix of correlations of the research variables Source: elaborated by the authors Note:

IQIC - Financial Information Quality Index; GC - Corporate Governance; TIPOAUD - Type of Audit; COMAUD - Audit Committee; CA - Shareholding Concentration; SIZE - Size; END - Indebtedness; ROE - Return on equity.

Table 5 analyzes the correlation matrix to show if there are problems with autocorrelation between the variables. This result allows us to conclude that the financial information quality index (IQIC) showed a more substantial degree of correlation with the variables of corporate governance (GC) at 14.90%, indebtedness (END) at 13.60%, and return on equity (ROE) with 17.90%. As for the independent variables TIPOAUD and COMAUD, there was a low correlation of 5.90% and 6.10%. It is also possible to observe that the size variable (TAM), despite having almost no correlation with IQIC, presents a strong correlation with CG, COMAUD, CA, and ROE with 51.70%, 44.41%, 40.40%, 31,20 %, respectively. However, for this analysis, no problems with autocorrelation were identified.

After these descriptive analyses, regressions were applied with the Chow (F), Breusch-Pagan, and Hausman LM tests. Finally, to determine which model best fits this research, POLS, Fixed Effects, or Random effects.

Table 6. Table 6. Result of Indicated Panel Data Specification Models (2014-2017) Source: elaborated by the authors.

Analyzing Table 6 shows that the Chow test (F) resulted in an acceptance of the POLS model. Furthermore, the test of Breisch-Pagan LM evaluated the adequacy of the random model to the POLS model, rejecting the H0. Lastly, the Hausman test does not reject H0, which allows us to infer that the most appropriate model for this test is the random effect model.

Table 7 below shows the model used to verify the relationship between the determinants of the quality of financial information and the quality index of financial information – IQIC. The relationship between the variables was performed using panel data regression, using the Random Effects model with robust standard errors. This panel data model was chosen due to the Wald test, which indicated the presence of heteroscedasticity in the model used in this research. Weighted models were estimated and observed according to their variances to correct this problem.

Table 7. Table 7. Result of estimating the determinants of the quality of financial information through panel data for Random Effects (2014-2017) Source: elaborated by the authors. Note:

IQIC - Financial Information Quality Index; GC - Corporate Governance; TIPOAUD - Type of Audit; COMAUD - Audit Committee; CA - Shareholding Concentration; SIZE - Size; END - Indebtedness; ROE - Return on equity. * Significant normal level 0.05 or 5%

It is possible to observe in Table 7 that the R² is 18.62% and the Adjusted R² of 13.23%, which can be considered satisfactory. That is, the model has an explanatory capacity of 13.23% on the dependent variable as recorded in other previous studies, such as that of Lopes et al. (2017), which presents regression with R² and Adjusted R² of 16.70% and 12.90% and Moura et al., (2016) with Adjusted R² of 17 %.

Considering the results shown in Table 7, it was found that the coefficients of the independent variables, type of audit (TIPOAUD) and audit committee (COMITAUD) are significant for the IQIC, indicating that when the company is audited by the big four, it has a higher quality index of financial information, as in the research Moura et al. (2017). The same happened with the audit committee. The survey results show that companies with excellent audit committee performance have a higher quality index of financial information. This result confirms the result of the study by Moura, Ziliotto & Mazzioni (2016), which indicates the influence of the audit committee on the quality of financial information.

The coefficients were significant but negative for the variables of indebtedness (END) and return on equity (ROE), meaning that the less indebted the bank is, the less likely it is to carry out financial manipulation in its financial reports, corroborating the findings on various works on the subject (Watts & Zimmerman, 1986; Gabriel, 2011; Martins & Ventura Júnior, 2020). As for the ROE result, it indicates that high performance is related to excellent earnings management.

Regarding shareholding concentration (CA), the result is like that of Moura, Ziliotto & Mazzioni (2016), who did not indicate the influence of shareholding concentration on the quality of financial information. It is also like that of Moura et al. (2017), stating that the company with shareholding concentration does not have the best quality of financial information. Lastly, the size variable (TAM) result was like that of Lopes et al. (2017), which shows that size is unrelated to the quality of financial information.

As for corporate governance (GC), the result differs from those found in Gabriel (2011), who found that governance could be seen as a dimension of composition and management, meaning that some recommended GC practices positively impact the quality of financial information. However, the results are in line with other studies (Moura, Ziliotto & Mazzioni, 2016; Moura et al., 2017) which show that the presence of corporate governance does not influence the quality of financial information, as well as other studies (Boycko, Shleifer, & Vishny, 1996; Core, Holthausen, & Larcker, 1999) which point out that weak corporate governance does not help to mitigate the problem of information asymmetry caused by the manipulation of financial reports.

The study aimed to analyze the factors determining the financial information quality in banks listed on B3. For this, the quality of financial information was analyzed through the financial information quality index (IQIC). The determining factors observed were corporate governance (GC), audit committee (COMAUD), type of audit (TIPOAUD), shareholding concentration (CA), size (TAM), indebtedness (END), and return on PL (ROE). To achieve the proposed objective, the quality of financial information of 20 banks listed on B3 was measured through a set of twelve indicators from the study by Gabriel and Silveira (2011). Then, an econometric model was used to identify the factors determining financial information quality. As for the quality of financial information, the results were like those of Nunes et al. (2016) and Gabriel (2011), who investigated, through a questionnaire, the quality of financial information in publicly traded companies listed in B3.

The result shows that among the factors of corporate governance, audit committee, type of audit, shareholding concentration, size, indebtedness and ROE, which were pointed out in the literature as influencing the quality of financial information, four sustained the theoretical assumption in the investigated sample. Two factors of these, COMITAUD and TIPOAUD, with positive and significant coefficients, indicate that the greater the performance of the audit committee and audited by one of the four largest auditing companies, the higher the quality of the financial information. The others are END and ROE, which showed a negative relationship with the quality of financial information, demonstrating that the lower the indebtedness and the company's performance, the higher the quality of financial information.

The contribution of this research is to show that the adoption of the audit committee and type of audit are determining factors to improve the quality of financial information of Brazilian banks. In addition, the study shows it is possible to analyze the control variables indebtedness and performance, noting that improving the quality of financial information requires the variables of indebtedness and return on equity to be reduced.

As limitations, we highlight that other factors could be considered to explain the quality of financial information. For example, industry characteristics, since it could be a variable that captures different business contexts. Also, and no less critical, analyze other populations, as the institutional system of each country varies and can influence the quality of financial information. The sample size can also be considered an aspect that could be improved, giving greater robustness to the results. Thus, for future research, it is suggested to include other explanatory factors that have not been contrasted in here or the consulted literature of this study. Furthermore, the research should be developed with a broader sample, including developing countries with capital markets similar to the Brazilian.