The association between book-tax conformity and earnings management

For an in-depth discussion of this debate, see Hanlon and Heitzman (2010). For examples of proponents of book-tax conformity, see Desai (2003, 2005, 2006), Graetz (2005), Whitaker (2005), and Chan et al. (2010). For examples of opponents, see Shackelford (2006), Hanlon and Shevlin (2005), Hanlon et al. (2008), and McClelland and Mills (2007).

Like Healy and Wahlen (1999) and Leuz et al. (2003), we define earnings management as alterations in firms’ reported economic performance by insiders to either mislead stakeholders or influence contractual outcomes.

According to Whitaker (2005), “By substantially eliminating deferred tax expenses, uniform accounting would reduce such opportunities for potentially misleading earnings management.” (p. 708).

The exception relates to the effect of increased conformity on earnings smoothing. While proponents of conformity argue that book-tax differences help firms smooth book income, a subset of the accounting literature argues that managers’ incentives to smooth taxable income will carry over to smoother accounting earnings under high book-tax conformity (Alford et al. 1993; Lang et al. 2012). Supporting this view, Lang et al. (2012) find a positive correlation between their measure of smoothing and a dummy indicating high book-tax alignment. Issues regarding smoothing, however, should not affect earnings management with respect to the probability of small losses or the magnitude of signed accruals at the country-year level. We note that our conclusions are quite similar when we eliminate earnings smoothing from our study and focus on nonsmoothing-related earnings management.

Note that, if we find less accruals management, one might ask whether firms switched to more real earnings management. Real earnings management (cutting R&D, advertising, sales promotions, etc.) generally involves book-tax conforming behavior, and thus its relative cost declines as book-tax conformity on accruals increases. If book-tax conformity increases the costs of accruals earnings management—reducing earnings management—then we might expect to observe more real earnings management. However, our results suggest more, not less, accruals earnings management in book-tax conformed systems.

Initially our results may appear to be inconsistent with Atwood et al. (2010) because firms with smoother earnings are likely to have more rather than less persistent earnings. However, while we find a positive relation between book-tax conformity and smoothing, we also find a positive relation between book-tax conformity and other forms of earnings management that are likely to decrease earnings persistence. Thus we conclude that effects of persistence decreasing earnings management outweigh the effects of earnings smoothing in our sample.

A contemporaneous working paper by Watrin et al. (2012) also tests for the relation between book-tax conformity and earnings management in a European setting using book-tax conformity measures based on whether the firm’s country uses a one-book, two-book, or three-book system. They observe less earnings management (measured as discretionary accruals) in two- and three-book firms, consistent with a positive relation between book-tax conformity and earnings management. We acknowledge that this measure of book-tax conformity overcomes some difficulties of the Atwood et al. (2010) measure that we use. We note, however, that the distinction between a one-book or two-book system exists at the single-entity level and that multinational corporations would likely operate in various one-book as well as two- and three-book systems. At the consolidated level, the distinction between a one-book versus two-book firm is less clear and assumes that the consolidated entity does not have any levers to influence reported earnings that will not affect taxable income at the single-entity level. Since the capital market pressures that influence earnings management are felt at the consolidated rather than single-entity level, it is difficult to infer the effect of book-tax conformity on earnings management by examining single-entity financial statements. We view our analysis as complementing theirs by using a measure of book-tax conformity that can be applied to many countries and by including both upward and downward earnings management in our tests to consider the effects of book-tax conformity on managers’ reporting behavior more broadly.

All financial variables are drawn from Compustat’s Global Vantage files. We use Compustat Global Vantage instead of the more recent Compustat Global files because the recent version does not report foreign tax expense, which Atwood et al. (2010) require to compute book-tax conformity.

Like Atwood et al. (2010), we use current tax expense (if available) when either total tax expense or deferred tax expense is missing. Atwood et al. (2010) include ForPTBI, because foreign earnings of multinationals may be taxed at different rates than their domestic statutory rate, and DIV to control for potential cross-sectional differences in current tax expense arising from dividend distributions (e.g., such as the transition between the imputation system to the current system in Germany). Atwood et al. (2010) find that book-tax conformity rankings are unchanged after excluding either ForPTBI or DIV from the model.

Given that we use financial statement disclosures to estimate taxable income rather than directly examining taxable income from tax returns, some of the cross-country differences in book-tax conformity may be captured by the intercept term in model 1 rather than in the root mean-squared error. In untabulated analysis, we find a higher than 99 % correlation between the book-tax conformity measure when we omit the intercept term from model 1 and the book-tax conformity measure when the intercept is included. Thus we conclude that the intercept is not capturing a meaningful amount of cross-country variation in book-tax conformity.

This procedure converts the ranks into percentiles. We do this conversion because the number of included countries varies by year as in Atwood et al. (2010). This transformation gives the book-tax conformity variable a consistent scale across years.

Barth et al. (2008) use modified versions of these measures to test for changes in earnings management around an event. They note that “An alternative approach used in some prior research (Dechow 1994; Leuz et al. 2003) is to base comparisons on alternative metrics constructed using a time series of firm-specific data. Data limitations preclude this approach because it requires a time series of observations for each firm that is not overlapping in the pre- and postadoption periods” (p. 481). Since we are not testing changes in earnings management around an event and we have the firm-specific time series of data, we follow Leuz et al. (2003) in constructing our measures of earnings management.

Because the arguments in favor of increasing book-tax conformity focus on accrual (as opposed to real) earnings management, our tests focus on measures of accrual rather than real earnings management.

Firms can also use discretionary accruals to improve the usefulness of accounting information (Watts and Zimmerman 1986). However, Leuz et al. (2003) argue this may be a result of effective outside investor protection and therefore may not extend to countries with inferior investor protection. Consistent with their argument, they find a negative relation between investor protection and measures of accounting discretion.

This approach to identifying earnings management has recently come under attack in Durtschi and Eason (2005, 2009), who argue that the results of Burgstahler and Dichev (1997) are driven largely by scaling and sample selection issues. Burgstahler and Chuk (2012), however, continue to find earnings discontinuities using unscaled earnings with differing bin sizes as a function of firm size. They also find that the majority of firms in Durtschi and Easton’s (2005, 2009) sample are very small firms with stock prices of less than $5, with very little analyst following and generally much weaker incentives to manage earnings than sample firms in Burgstahler and Dichev (1997). Burgstahler and Chuk (2012) observe earnings discontinuities in firms that are likely to have sufficient capital market pressures to manage earnings, consistent with the earnings management interpretation of discontinuities around earnings benchmarks. Nevertheless, in untabulated tests, we omit EM4 from our analysis and find similar results.

In untabulated tests, we substitute our aggregate measure of earnings management with a single factor obtained from factor analysis and find similar results. We also use aggregates of EM1 and EM2 (which capture smoothing) and of EM3 and EM4 (which capture overall discretion in financial reporting) separately and again find similar results for both sets of earnings management measures.

Compustat Global Vantage begins its coverage in 1991. However, because our aggregate earnings management measure requires 5 years of lagged data our sample period starts in 1996.

Due to increases in Compustat’s data coverage over time, we can estimate book-tax conformity for three countries (Ireland, Pakistan, and Portugal) in addition to those included in Atwood et al. (2010). However, our results are very similar when we delete these three countries from our sample (untabulated). Also, China and Indonesia are included in Atwood et al. (2010) but are excluded from our sample due to missing data from La Porta et al. (1998) (China) or the CIFAR index (Indonesia). We discuss inclusion of China and Indonesia below.

While the conventional wisdom holds that Germany has high book-tax conformity, given the close tie between tax and book numbers in Germany, prior literature shows that such book-tax conformity requirements do not apply to the “group accounts” in the consolidated financial statements (Leuz and Wustemann 2004). Atwood et al. (2010) show that overall conformity in Germany can be quite low, similar to that of Australia, where conventional wisdom indicates a low level of book-tax conformity.

La Porta et al. (1998) define privately owned firms to be those in which the state is not a known shareholder. Under this definition, these firms may have publicly traded stock.

Since our earnings management and book-tax conformity variables are ranked by year, adding year fixed effects to the model should have little impact on our test results. Nevertheless, we include year fixed effects to capture yearly variation in our control variables that might vary from year to year.

In studying the relation between earnings persistence and book-tax conformity, Atwood et al. (2010) also control for earnings variability and find that it has no significant effect on the relation between earnings persistence and book-tax conformity. We omit earnings variability from our set of control variables because it likely over-controls for the effect we are trying to capture (e.g., earnings variability is the numerator in calculating EM1). However, in untabulated results we find that the coefficient on book-tax conformity is still positive and significant (coefficient = 0.241, p value < 0.01) even after controlling for earnings variability.

Whether moving from local GAAP to IFRS increases or decreases book-tax conformity also depends on how conformed local GAAP was to the tax rules before and after the change. We can only capture the differences between local GAAP and IFRS in this test. If a country used reported GAAP income as the basis for calculating taxes both before and after IFRS adoption, it would not matter how different local GAAP was from IFRS in the pre-IFRS period; book-tax conformity would be high both before and after IFRS adoption. While somewhat surprising, we interpret the insignificant coefficient on β₄ below as evidence that the positive relation between book-tax conformity and earnings management above is unlikely to be driven by IFRS adoption.

We thank Hollis Skaife for providing us with country-level data on required inventory and depreciation conformity.

Results from these alternative tests are available upon request.

We demean the book-tax conformity and statutory tax rate variables to enhance the interpretability of the main effects as well as the interaction between these variables (Burks et al. 2013).

\(ITCV = \left( {\sqrt {(1 - r_{x \cdot z}^{2} )(1 - r_{y \cdot z)}^{2} } } \right)\left( {\frac{{t^{2} + t\sqrt d }}{{ - \left( {n - q - 1} \right)}} + \left[ {\frac{ - d - t\sqrt d }{{ - \left( {n - q - 1} \right)}}} \right]r_{y \cdot x|z} ,} \right)\) where \(r_{x \cdot z}^{2}\) is the R² from a regression of x on all other control variables (0.2503); \(r_{y \cdot z }^{2}\) is the R² from a regression of y on all other control variables (0.487); t is the critical value from a T distribution (we use a t value of 1.96, where alpha = 5 % two-tailed); n is the number of observations (362); q is the number of independent variables included in the model (22 including the year dummy variables); \(r_{y \cdot x|z}\) is the partial correlation between x and y holding constant all control variables (0.433); and \(d = t^{2} + \left( {n - q - 1} \right).\) For more details on the calculation of ITCV see Frank (2000).