Introduction

One of the barriers still faced by the International Accounting Standards Board (IASB) in its efforts to converge global accounting standards is the existence of alternative presentation, recognition, measurement and disclosure practices. While flexibility in accounting choice allows companies to represent events and transactions more faithfully, it also affects the comparability of the information disclosed.

IAS 7 (^{IASB, 2010}) allows some items in the Statement of Cash Flows (SCF) to be classified under different subgroups. This subjectivity can result in adjustments to economic reality or simply the company preference. It also gives rise to the possibility of accounting choice with respect to these items. If the goal of IASB was to segregate the amounts paid or received by a company into operating, financing or investing categories, then this goal implies that segregation is important for user decisions. If not, input and output values would simply be listed, for example, in alphabetical order or by relevance, or even by whatever model a company deemed appropriate. Therefore, to be useful, it is assumed that cash flow values from operating activities must be comparable among companies.

According to ^{IAS 7 (2010}), Statement of Cash Flows help increase the comparability of business performance by decreasing the use of different accounting criteria for the same transactions and events. Furthermore, according to IAS 07, one of the main improvements provided by the SCF is increased comparability among different companies regarding the presentation of operating performance. ^{Macedo, Machado, Murcia, and Machado (2011}) states that the options allowed by ^{IAS 7 (2010}) hinder comparability among companies facilitate cash flows management and consequently affect the relevance of accounting information for decision-making.

The convergence of Brazilian accounting standards with International Financial Reporting Standards (IFRS) in 2010 was an important milestone in Brazilian accounting. IASB should examine the remaining choices in IFRS in order to make financial statements truly comparable. Accounting choice allows more faithful representations of accounting facts, but also allows greater discretion. One of the accounting choices still allowed in IFRS is the ability to classify items in the SCF.

Given these facts and a desire to contribute to this discussion, our objective is to identify the level of comparability in the accounting choices of Brazilian companies in their statements of cash flows and to identify the factors that might explain these choices. We selected a sample of 354 Brazilian public companies with shares traded on the BM&FBOVESPA from 2010 to 2015 and analyzed accounting choices related to the disclosure of the following SCF items: interest expense, interest income, income tax and social contribution on net income, dividends and interest on own capital (IOOC) received and dividends and IOOC paid. After identifying these choices, the Herfindahl index (H index) was calculated to determine the degree of comparability among the choices made by the companies in the sample. Statistical tests were used to determine whether company size, indebtedness, profitability, growth opportunities and negative operating cash flows affected the choices made by companies in the disclosure of the previously mentioned SCF items.

The importance of this study stems mainly from the desire to understand the decisions taken by managers regarding the classification of SCF items and the consequential impacts on comparability within our sample of Brazilian companies. These results may be useful for investors, managers, and in standardized analyses of the impacts of the choices allowed in the accounting regulations regarding the disclosure of SCF items. ^{Silva, Martins, and Lima (2014}) identified the factors that affect the accounting choices selected by 107 Brazilian companies listed in 2010 in one segment of the BM&FBovespa regarding the disclosure of interest expense, income, dividends and interest on own capital in the Statement of Cash Flows. The period (2010 a 2015) and sample (354 companies) were widened in this study to more robustly evaluate changes in accounting practices over six years.

After this introduction, the paper will present the theoretical reference for the study. The third section will then describe the methods used to carry out the study, the fourth section will present results and the fifth section will present final considerations.

Review of the literature

Methodology and data

The research sample consisted of Brazilian public companies listed on BM&FBOVESPA from 2010 to 2015, except those in the finance and insurance industries and companies with missing dates of the operating cash flows (Table 2). Financial and insurance companies were excluded because their classification of interest is distinct from that of companies in other industries. Thus, the valid sample size was 354 Brazilian public companies. The financial statements of these companies were obtained from the Brazilian Securities Commission (CVM - Comissão de Valores Mobiliários) website.

Data collection involved the identification of the classification for the following five items in the Statement of Cash Flows of the sample companies: interest income (II), interest expense (IE), income tax and social contribution on net income, (ITC) dividends and interest on own capital received (DIR) and dividends and interest on own capital paid (DIP). Table 1 shows the choices analyzed in data collection. Each of these items could be classified in operating (O), investment (I) or financing (F) cash flows. In addition, three other classifications were included in the research: not mentioned (NM), mixed (M) and not applicable (NA). The option ‘not mentioned’ was recorded when an item was recognized in the balance sheet, income statement or notes, but was unrecognizable in the SCF. The option ‘not applicable’ was noted when one of the itens was not mentioned in the SCF, balance sheet, income statement or notes. The mixed option was used to group companies that chose more than one rating for the same accounting practice, which is not allowed by the nomenclature used or the notes disclosed by the company.

After collecting the classifications of the items in the SCF, the Herfindahl index (H index) was used to calculate comparability. The H index is an accumulated index that is calculated from the relative frequency of comparison between alternatives and the same alternatives (^{Tas, 1988}). The H index was chosen because of its applicability to the five possible SCF accounting choices considered by the present study. Thus, the data collected from the financial reports was compared among all sample companies to determine the degree of comparability. Equation 1 shows the Van Der Tas (Tas, 1988) formula for the H index:

Here H represents the Herfindahl index, n is the number of alternative accounting methods and pi is the relative frequency of accounting method i. H varies from 0 (no harmony, with an infinite number of alternative methods having the same frequency) to 1 (all companies using the same method) (^{Tas, 1988}).

The H index uses all the elements introduced in this equation to generate a value that represents the level of comparability of the practices evaluated. ^{Lemes, Costa, and Martins (2014}) state that researchers must use judgment to interpret this comparability value as high, moderate or low. ^{Taplin (2011}) highlights the difficulty of evaluating harmony due to the lack of benchmarks, but notes that this criticism extends to all indices and statistical summaries. The author cites the traditional benchmark for the p-value (0.05). This value is accepted by convention but not universally accepted as appropriate in all areas. Table 3 summarizes some of the parameters used to classify comparability.

As in the ^{Taplin (2011}) study, we adopted the ^{Ali, Ahmed, and Henry (2006}) classifications to determine the degree of comparability of the SCF accounting choices.

In addition, statistical analyses were performed to identify the factors that affect the classification choices of the items analyzed in this study. First, statistical analysis consisted of the mean differences test for two independent samples. Then, logistic regression was used to test the hypotheses of the study. Equation 2 shows the model used in this study that was first specified by ^{Silva, Martins, and Lima (2014}) and based on research by ^{Lee (2012}), ^{Gordon et al. (2017}) and ^{Baik et al. (2016}).

The CLAS variable represents the classification of items in SCF (1 when the company classified the items described in Table 1 according to the first classification suggested by the IASB; 0 otherwise). The first classification suggested by the IASB was also presented in Table 1. SIZE represents the size of a company measured by the logarithm of total assets in reais (BR$). IND is indebtedness measured by the ratio of total liabilities to total capital. ROE symbolizes return on equity as measured by net income divided by shareholder equity and is a proxy for the company’s profitability. NOPC is negative operating cash flows, dummy 1 if the company had negative operating cash flows and 0 otherwise. BP refers to book-to-price or growth opportunities, which is the calculated market price of the share divided by the book value per share.

It is expected that larger firms, companies with negative operating cash flows, greater indebtedness, greater profitability and more growth opportunities will tend to choose classifications that increase operating cash flows. Thus, the expectation is that companies with these characteristics prefer to classify interest and dividends paid from financing activities rather than operating activities and also prefer to classify interest and dividends received under operating activities and not investing activities. It is also expected that these companies would prefer to classify the payment of income tax and social contributions on net income from financing activities or investment rather than operating activities.

Thus, the expected signals for the variables of Equation 2 are presented in Table 4. When the variable CLASS is represented by the choices related to Interest Income (II), the expected signal for the five independent variables is positive. In other words, larger, more indebted, more profitable companies with higher book-to-market and negative operations cash flows choose to classify item II in the operations cash flows, increasing this flows. The positive sign is also expected for the five independent variables when the dependent variable is Dividends and Interest on Own Capital Received (DIR).

When the dependent variables are Interest Expense (IE) and Income Tax and Social Contribution on Net Income (ITC), the expected signal for the five independent variables is negative. Considering that the classification of these items into the cash flows of operations would reduce this flows, companies are expected to classify these items into another flows (investment, financing, or otherwise).

The dependent variable Dividends and Interest on Own Capital Paid (DIP) has a value of 1 when the company classifies this item in the financing cash flows (primary classification) and 0 otherwise. Thus, the positive sign of the five independent variables indicates that companies are classifying the DIP item in financing activities, which increases the cash flows of operations.

Data analysis

Table 5 shows the choices made by the publicly traded sample companies regarding the classification related to SCF items. Most of the sample companies (282 companies or 80%) did ‘not mention’ (NM) in the notes or in the SCF the classification of interest income (II) in 2010. This behavior was repeated in every year of the study. Few Brazilian public companies (around 3%) have classified interest income in the cash flows of operations as suggested by the IASB as the primary classification.

The classification of Income Tax and Social Contribution on Net Income (ITC) was also reported under operating cash flows (36% in 2010 e 55% in 2015), as recommended in IAS 7 (2010) (Table 5). Next to it, despite a high number of companies that did not mention (NM) the classification of this item, this percentage has been reducing over the period (47% in 2010 e 28% in 2015),

Most companies chose to classify dividends and interest on own capital received (DIR) under investment cash flows (29% in 2015), contradicting IAS 7 (2010), which recommended operating cash flow for this classification (Table 5). Most statements classified dividends and interest on own capital paid (DIP) under financing cash flows (64% in 2015), similar to the findings of ^{Silva, Martins, and Lima (2014}) and as suggested by the IASB as the primary classification.

Of the total number of companies disclosing the classification of items in the SCF, most classified the items as income, interest expense, income tax and social contribution on net income in operating activities as suggested by the IASB as the primary classification. In addition, most companies classified the item dividends and interest on own capital paid as a financing activity, as also recommended by the IASB. Thus, the H1 hypothesis that Brazilian public companies classify the accounting choices in SCF in the primary classifications suggested by the IASB can be accepted for items II, IE, ITC and DIP.

On the other hand, most of the companies in the sample did not follow the IASB’s recommendation to classify dividends and interest on own capital received in the cash flows of operations, classifying this item in the investment cash flows. This did not allow the hypothesis H1.

It is important to note the significant number of companies that did not disclose (not mentioned) the classification of the items analyzed in this study of SCF. This fact may suggest poor quality of the information disclosed.

From the data in Table 5, it was possible to calculate the H index (Equation 1). These H indexes were then used to determine the degree of comparability of the choices of SCF (Table 6). For this calculation, the companies that did not mention the items analyzed in this research of the balance sheet, income statement, notes or the SCF (not applicable - NA - Table 5) were not considered.

The comparability index for interest income (II) in 2010 was 0.89 (Table 6). This level of comparability for this choice is high; however, the ratio is due to non-disclosure (not mentioned- NM) of interest income in SCF, although this item was disclosed in the balance sheet, in the DRE or in the explanatory notes. In the other periods, the comparability index was similar, ranging from 0.89 to 0.91 due to the high volume of companies that did not mention (NM) the classification of interest income in SCF. Although the general index of the comparability (GCI) of interest income (0.90) is high, this index represents a poor quality of the disclosure of the accounting choices since most companies (1,766, approximately 95%) did not mention (NM) the classification of interest income in SCF during the period from 2010 to 2015.

The comparability level of interest expense (IE) was lower in each of the years analyzed (0.33, 0.34, 0.34, 0.34, 0.33 e 0.35, respectively). Most of the companies classified interest expense under operating cash flows. Nevertheless, this classification is disperse. The index of comparability was also low for income tax and social contribution (ITC) from 2010 until 2015 (0.50, 0.49, 0.51, 0.51, 0.53 and 0.55, respectively). Most of the companies classified this item under operating activities, but many companies did not mention (NM) the classification of these items in the SCF.

The comparability level for dividends and interest on own capital received (DIR) was low throughout the entire period analyzed (between 0.36 and 0.43). Most companies chose to rate this item in investing cash flows. However, one possible explanation for these low comparability values could be that companies chose heterogeneous classifications for DIR.

The level of comparability for dividends and interest on own capital paid (DIP) was moderate throughout the study period (between 0.82 and 0.84). It can be inferred that this comparability index was maintained because the companies chose to classify DIP under financing activities, in accordance with IAS 7.

Table 6 shows that, in general, given the time horizon of the study, the comparability level was high for interest income (II = 90%), moderate for dividends and interest on own capital paid (DIP = 83%) and low for interest expense (IE = 33%), income tax and social contribution (ITC = 50%) and dividends and interest on own capital received (DIR = 41%).

Figure 1 also shows that comparability was lowest in 2010 (roughly 58%). However, in subsequent years, comparability gradually increased to moderate levels (59% to 62%), which allows the hypothesis H2 (Figure 1) not to be rejected. Despite moderate indexes, this upward trend may indicate an improvement in comparability over time, which may in turn suggest that these Brazilian companies had garnered more experience with IFRS.

Figure 1 General Comparability Index in SCF by year.  

Table 7 shows the descriptive statistics for the variables that comprise the model specified in Equation 2. Table 8 shows the difference of means test for the variables that represent the group of companies that classified interest income (II), interest expense (IE), income tax and social contribution on net income (ITC) and dividends and interest on own capital received (DIR) under operating cash flows and the group that classified these items under other items or that did not mention the item. Add to this, Table 8 also shows the difference of means test for the variables that represent the group of companies that classified dividends and interest on own capital paid (DIP) under financing cash flows and the group that classified these items under other items or that did not mention the item.

Table 7 shows that the mean SIZE is higher for the group that chose to classify interest income (II), interest expense (IE), ITC and DIR in the operating activities (dummy = 1). The same occurred with DIP, the average SIZE is higher for those companies that chose to classify this item in financing activities (dummy = 1). The mean test in Table 8 confirms these results, which that the company size may affect the choice of classification of these items in the Statement of Cash Flows.

Besides that, the tests in Table 8 show that companies size (SIZE) and book-to-price (BP) can determine the option used to classify interest income in SCF (95% confidence level) and indebtedness (IND) and profitability (ROE) can determine this classification with a 90% confidence level.

Size (SIZE), profitability (ROE), book-to-price (BP) and negative cash flows from operating activities (NOPC) can determine the classification choice for interest expense in SCF (95% confidence level).

In the same way, size (SIZE), profitability (ROE), book-to-price (BP) and negative cash flows from operating activities (NOPC) may affect the classification choice for taxes on income tax and social contribution (95% confidence level), but indebtedness (IND) can only determine this classification with a 90% confidence level.

Size (SIZE) and profitability (ROE) may affect the classification choice for dividends and IOOC received (95% confidence level), but indebtedness (IND) can only determine this classification with a 90% confidence level. Whereas size (SIZE), profitability (ROE), growth opportunities (BP, book-to-price) and negative cash flows from operating activities (NOPC) can affect the classification of dividends and interest on own capital paid.

Table 9 - Panels A to E - shows the logistic regressions. The absence of significant multicollinearity in the data set is verified through VIF and tolerance statistical tests. It was possible to verify that it is not multicollinearity in the models, since the VIF test indicated that the highest colinearity in the models is equal to 1.37. The significant LR chi2 test results (5) at the 5% level in Panels A to E suggest that the variables should not be dropped. In other words, it seems appropriate.

Panel A - Table 9, where the dependent variable is a dummy with a value 1 when the company chose to classify Interest Income in the operating activities and 0 if other classifications were chosen. The results of Table 9 - Panel A suggests that companies with higher size (SIZE), lower indebtedness (IND), higher book-to-price (BP) and negative operating cash flows chose to classify interest income under operating cash flows, which allows us initially to accept the hypotheses H 3, H6 and H7. That is, public companies with negative operating cash flows (H7), larger public companies (H3) and more growth opportunities (H6) are more likely to make choices that increase operating cash flows, evidencing interest income in operating activities. Hypothesis H4 was not confirmed because companies with greater indebtedness seem not to classify interest income under operating cash flows. Similarly, the hypothesis H5 was not confirmed since the coefficient of the profitability (ROE) was not significant at the 5% level. The ‘Correctly Classified’ of 96.68% shows the overall efficiency of the model, that is, 96.68% of the observations were classified correctly (Table 9 - Panel A).

The “odds ratio” presented in Table 9 should be interpretd as follows: the coefficient 1.87, in Table 9 - Panel A, means that the probability of a larger (SIZE) company choosing to classify the item “Interest Income” in operating cash flows is 1.87 times more likely than the larger company to classify this item in another flow or not mention the rating. Similarly, the odds ratio 0.10, being less than 1, in Tabel 9 - Panel A, indicates that companies with higher indebtedness (END) is 0.10 times less likely to classify “Interest Income” in operating cash flows as compared to a company less indebted.

Table 9 - Panel B considers classification choices in SCF related to interest expense as a dependent variable. Here, the company size (SIZE), indebtedness (END) and book-to-price (BP) affected the option for the classification of interest expense in SCF (at the 95% confidence level). There are indications that larger Brazilian public companies, companies with lower indebtedness and companies with higher growth opportunities opted to classify interest expense as an operating activity. Considering that when the company classifies interest expense into the operating cash flows, the balance of that flows decreases, the hypotheses H3 and H6 was not confirmed because larger companies and companies with more growth opportunities classify interest expense into operating cash flows, reducing this flow. H4 can be confirmed because companies with higher indebtedness did not classify interest expense into operating cash flows, which increases this flow. The ROE and NOCP coefficient were not significant at the 5% level.

Regarding the dependent variable choices related to the classification of income tax and social contribution (ITC), Table 9 - Panel C, the results suggest that larger companies (SIZE), companies with lower indebtedness (IND), higher profitability (ROE) and when operating cash flows are not negative (NOCP) to register this item under operating cash flows. Considering that when the company classifies ITC into the operating cash flow these flows decrease, the hypotheses H4 and H7 were confirmed because companies with higher indebtedness (IND) and companies with negative operating cash flows (NOCP) do not seem to register ITC in operating cash flows. But H3, H5 and H6 were not confirmed. Larger companies (SIZE) and companies with higher profitability (ROE) allocated ITC in operating cash flows, which reduced these flows. The book-to-price (BP) coefficients were not significant at the 5% level.

Table 9 - Panel D - considers classification choices in SCF related to Dividends and IOOC Received (DIR) as a dependent variable. The SIZE and END coefficients are significant at the 5% level. These results suggest that larger companies (SIZE) and companies with lower indebtedness (IND) to register DIR under operating cash flows. Therefore, the hypothesis H3 was confirmed because larger companies make choices that increase operating cash flows.

Table 9 - Panel E - considers classification choices in SCF related to Dividends and IOOC Paid (DIP) as a dependent variable. The SIZE, END and ROE coefficients are significant at the 5% level. These results suggest that larger companies (SIZE), companies with lower indebtedness (IND) and higher profitability companies (ROE) to register DIR under financing cash flows. Therefore, the hypotheses H3 and H5 were confirmed because larger companies and profitable companies make choices that increase operating cash flows, allocating this item in the financing cash flows.

Table 10 is a summary of the rejected and non-rejected hypotheses based on the logistic regressions in Table 9.

Thus, the results suggest that larger, more indebted, more profitable Brazilian public companies with larger book-to-market and negative cash flows from operating activities make choices that increase operating cash flows. Size, book-to-market, and negative operating cash flows affect the choice of allocating interest income to the operating cash flows. Meanwhile, indebtedness affects the choice of allocating interest expense in operating activities.

Indebtedness and negative operating cash flows mean that the company chooses to register ITC in its operating cash flows. Size influences the decision to allocate the DIR and the size and profitability influence the decision to allocate the DIP to the operating cash flows.

These results corroborate the study by ^{Gordon et al. (2017}) with regard to profitability and size of companies and the study of ^{Baik et al (2016}) in relation to the growth-opportunities (book-to-price). However, the results of this research contradict those of ^{Gordon et al. (2017}) as to profitability. These results may help complement the ‘accounting choice theory’, since it shows that in Brazil there are also significant evidences of the use of accounting choices to manage cash flows.

Conclusions

Our objective was to identify the comparability level of accounting choices selected by Brazilian public companies for their Statement of Cash Flows and the factors that might explain these choices. We evaluated accounting choices related to the disclosure of the following items in SCF: interest expense (IE), interest income (II), income tax and social contribution on net income (ITC), dividends and interest on own capital received (DIR) and dividends and interest on own capital paid (DIP). The Herfindahl index (H index) was used to calculate comparability and logistic regression was used to identify variables that might affect the classification of accounting choices in SCF for the 354 sample companies.

Our results suggest a high level of comparability (0.90 from 2010 to 2015) for the classification of interest income in the SCF. The level of comparability was moderate (0.83) for dividends and interest on own capital paid during the time horizon of the study. Conversely, the comparability index was low for the classification of interest expense, income tax and social contribution, and dividends and IOOC received (0.33, 0.50 and 0.41, respectively).

It should be noted that the high indexes of the comparability of the items interest income and interest expense are derived from the high number of companies that registered these ítems in the balance sheet, income statement or notes, but was unrecognizable in the SCF. Thus, these companies are comparable because they made the choice not to show the classification of these two items in the SCF, but this does not represent the quality of the accounting information.

The comparability index for the classification of accounting choices was low in 2010 (0.58), but it has gradually increased over time to a moderate level (0.62 in 2015), which might suggest improvements in the quality of information reported by Brazilian public companies. This trend also suggests a migration to similar choices, with lesser dispersion among the classifications in the SCF.

The results of our quantitative analysis suggest that the size of the Brazilian public company, indebtedness, profitability, book-to-market, and negative operating cash flows affect the choice of item classification in SCF. Larger, more indebted, more profitable Brazilian public companies with larger book-to-market and negative cash flows from operating activities make choices that increase operating cash flows. Size, book-to-market, and negative operating cash flows affect the choice of allocating interest income to the operating cash flows. Meanwhile, indebtedness affects the choice of allocating interest expense in operating activities. Indebtedness and negative operating cash flows mean that the company chooses to register ITC in its operating cash flows. Size influences the decision to allocate the DIR and the size and profitability influence the decision to allocate the DIP to the operating cash flows.

These results may be useful for accounting choice theory since they suggest that Brazilian listed companies use accounting choices to manage cash flows.

The fact that most of the companies in the sample did not mention the option used to classify, for example, interest income in the SCF is one limitation for the research. Nevertheless, this limitation does not invalidate the results of the study given that the results of the logistic regressions indicate variables that might affect the classification options chosen for the items in the SCF. Furthermore, ^{Taplin (2003}) points out that bias in the estimation of populations based on the index calculated for the sample are negligible and that the index based on the sample is, on average, close to the population index.