Firm- and Country-Level Attributes as Determinants of Earnings Management: An Analysis for Latin American Firms

ABSTRACT

This article analyzes firm- and country-level determinants of the earnings management for a sample of Latin American companies from 1997 to 2015 by using panel data to deal with the endogeneity and heterogeneity problems. Results show that dividend pay-outs impact positively on earnings management. The ownership structure, however, is a double-edged sword as a controlling mechanism that may constrain earnings manipulation but may also exacerbate it. Concerning country-level variables, we found that the development of the financial system behaved opposite of expectation. Consequently, before inefficient financial markets in Latin America, managers had more room for manipulation of financial statements. The legal and regulatory system, however, proved itself to be efficient in reducing the opportunistic behavior of managers.

KEY WORDS:

• corporate governance
• discretionary accruals
• earnings management
• emerging markets
• legal system

JEL CLASSIFICATION:

• G30
• G32
• G34

Acknowledgments

We wish to thank Ali M. Kutan (editor), three anonymous referees, and seminar participants in the 15th Finance, Risk, and Accounting Perspectives (FRAP) Conference, celebrated in 2015 at University of Applied Sciences Upper, Austria, for their valuable comments and suggestions.

Supplementary Information

Supplemental data for this article can be accessed on the publisher’s website.

Notes

^1. There are two kinds of earnings management techniques: The accounting methods and the operating methods (Tirole 2006). This work is only focused on the accounting methods.

^2. In fact, Bebchuk and Hamdani (2009) argue that all the effort invested in creating a composite index of corporate governance is worthless unless ownership structure is not incorporated.

^3. This regulation came after a series of financial scandals involving accounting irregularities and share price manipulation in several leading companies of that time, such as the energy company Enron, Worldcom, and Tyco, among many others, which filed restated financial results with the Securities Exchange Commission (SEC) during 2001 and 2002.

^4. Updated information can be downloaded from the permanent URL http://go.worldbank.org/X23UD9QUX0.

^5. Updated information can be downloaded from www.govindicators.org.

^6. An exogenous variable is that whose values are given and are not affected by the variable to be explained, which is said to be endogenous. As a result, there is an endogeneity problem when some of the explanatory variables are not strictly exogenous.

^7. Several standard diagnostic tests for panel data are used. The Hansen/Sargan test assesses the model specification validity (Hansen, Heaton, and Yaron 1996). This test examines the lack of correlation between the instruments and the error term. The AR(2) statistic measures the second-serial correlation. The Wald test of joint significance is also used to assess the significance of all the independent variables in the sample. We conduct the variance inflation factor (VIF) as a formal test to ensure that multicollinearity does not bias our results in the models’ estimation. Finally, the Lind-Mehlum contrast is used to study the statistical significance of the non-monotonic relationships suggested in this study (e.g., the cases of leverage and the insider ownership). As observed in all the regression outputs, the results are robust according to these diagnostic tests.

^8. To do so, we must compute the first derivative of this regression relative to the $LevB$ variable, and then make it equal to zero. After that we must solve for $LevB$ that represents the point at which the discretional accruals are maximized. Specifically speaking, this solution takes the form: $\frac{\mathrm{\partial }\mathrm{D}\mathrm{A}\mathrm{C}\mathrm{C}1}{\mathrm{\partial }\mathrm{L}\mathrm{e}\mathrm{v}\mathrm{B}}=0.0122-2\times 0.0502\times \mathrm{L}\mathrm{e}\mathrm{v}\mathrm{B}=0$ from the first regression output in Table 4. Consequently, when $\mathrm{L}\mathrm{e}\mathrm{v}\mathrm{B}=12.15\mathrm{\%}$, the earnings management are maximized.

^9. As described at the lowermost of the table, we used the appropriate Lind and Mehlum (2010) test to accept the hypothesis that the there is a statistically significant inverse U-shaped relationship between the firm’s leverage and the earnings manipulation.

^10. Similar regressions were computed by using $Div2$ calculated as the absolute value of the cash dividends divided by the previous year after-tax income. The results were comparable to those shown in Table 4, 5, 7, 8, and 9 although with some loss of significance in some estimations. For space-saving reasons, these results are not included in this work, but they are available upon request to the corresponding author.

^11. To do so, we must compute the first derivative of this regression relative to the $InsOwn$ variable, and then make it equal to zero. After that, we must solve for $InsOwn$ that represents the point at which the discretional accruals are minimized. Calculations are like those already described in Footnote 8.

^12. Originally, the models used the $Z1$ variable. However, for most of the regressions, this variable was not statistically significant (not reported here for space-saving reasons). Consequently, we decided to use $Z2$ variable that had much better explicative power than $Z1$. As it was mentioned in the Supplementary Material available online, $Z1$ is based on the formulation for developed countries (Altman 1968), while $Z2$ is based on the computation for emerging economies (Altman 2005). Therefore, we believe that $Z2$ is more suitable for the goals of this article than $Z1$.

^13. Computed as the arithmetic mean of the coefficient for $IFRS$ variable among the 12 regressions in Table 5.