How do CEO incentives affect corporate tax planning and financial reporting of income taxes?

Among 3309 firms with positive pre-tax income, operating cash flows, tax expense, and taxes paid. All values winsorized at 1 %.

Most CEOs do not have a tax background and are not often directly involved in corporate tax planning (Dyreng et al. 2010). However, CEOs provide input for incentivizing and evaluating other executives and managers, such as the CFO and tax director. CEOs also likely align the incentives of subordinates with their own and reward performance that increases their own incentive compensation. Therefore CEO incentives can influence taxes even without direct CEO involvement.

Besides performing a battery of tests to support this finding, we attempt to reconcile our results with those of Gaertner (2014). Section 4.2.2 provides details of this process and our findings.

We assume that all firms engage in some tax planning. For example, many firms claim accelerated tax depreciation. However, not all firms engage in cost segregation studies whereby tax depreciation is further accelerated by advantageously classifying depreciable property into shorter recovery periods. We would consider such a project to be an incremental tax planning strategy of sufficient cost to warrant additional consideration before implementation.

A cost segregation study is an example of a strategy that would not reduce reported tax expense because it is temporary. In contrast, claiming a research and development tax credit is an example of a strategy that would reduce reported tax expense.

Many components of the tax expense accrual require significant judgment. For example, two firms could engage in the same tax planning strategy and accrue different reserves for unrecognized tax benefits due to subjective differences in assessing the inherent risk of the strategy (e.g., De Simone et al. 2014). Our predictions remain unchanged if we allow firms to engage in a combination of tax planning and tax-advantageous financial reporting in each action presented in Exhibit 1.

Action 5 produces an equivalent amount of pre-tax income. However, we believe CEOs compensated on pre-tax earnings will prefer action 1 because making opportunistic financial reporting choices related to tax planning could come at a cost with no corresponding benefit (i.e., no incremental effect on his/her bonus).

One tax advisor revealed to us that a CEO chose not to implement a legal tax planning strategy that would have generated $20 million in tax savings because his bonus was determined exclusively based on pre-tax income. In this situation, the CEO’s performance metric would have reflected the cost of the strategy (i.e., a reduction to pre-tax income reflecting consulting fees paid to implement the strategy) but not the tax benefit.

If the CEO does not achieve his or her bonus, it is difficult to generate clear predictions about how accounting metrics affected taxes in nonbonus years. For example, if the CEO anticipates missing a target, he or she could be incentivized to take a pre-tax big bath or engage in other forms of downward pre-tax earnings management, which would introduce a denominator effect into our ETR measures. Eliminating firms with nonbonus years allows us to retain 75 % of otherwise includable sample firms. We therefore believe our results are broadly generalizable. Additionally, in an untabulated analysis, we relax this restriction for a sample of S&P 500 firms, and the results remain unchanged.

Companies often modify disclosed performance targets to reflect non-GAAP adjustments. For example, a firm can list EPS as the performance target, but a detailed examination of the metric’s calculation reveals that the actual target adjusts GAAP EPS to exclude income tax expense. We evaluate how each metric is defined in the proxy to ensure our classification reflects the actual characteristics of the performance target.

Because our sample period encompasses three years after the financial crisis of 2007–2008, short-term bonuses paid during our sample may not represent bonuses in other periods. For all firms with nonmissing data on CEO incentive compensation in Execucomp, bonuses were 40 % of incentive compensation during our sample (2009–2011) and 37 % during 2012–2013. Additionally, 88 % of Execucomp firms awarded CEO bonuses during our sample and 91 % awarded bonuses during 2012–2013. Therefore, during the two years since the end of our sample period, we see no substantial changes in CEO bonuses that would cause us to believe our results have limited generalizability.

The receipt of performance-based restricted stock is also often calculated using multi-year averages of the same metrics used for annual bonuses, which increases the salience of these metrics to the CEO. For example, in 2012 Priceline’s compensation committee established adjusted EBITDA as the performance metric to judge performance over both the annual (1-year) and long-term (3-year) periods. In this case, a 3-year average of annual EBITDA determines the long-term incentive, and this compensation is paid in equity rather than cash.

Although some studies suggest that firms with more tax avoidance opportunities are more likely to use after-tax metrics, Huang et al. (2015) present evidence that firms with higher GAAP ETRs are more likely to choose EPS, an after-tax metric, as a performance measure in CEO bonus contracts. Therefore it is not obvious how tax planning opportunities are associated with firms’ choice of performance metrics.

Greene (2003) argues that lagged values can also address measurement error.

Because we use robust regressions, we do not winsorize our variables. This design choice significantly affects the mean and standard deviation of variables presented. We therefore focus on median comparisons.

We use robust regressions to deal with influential observations. However, all inferences remain unchanged if we (1) retain all observations and winsorize all variables at 1 and 99 %, (2) eliminate observations where ETR or CETR is outside of [0, 1] before winsorizing at 1 and 99 %, or (3) eliminate observations where ETR or CETR is outside of [0, 1] before estimating robust regression.

Gaertner (2014) conducts his analysis using a sample of S&P 500 firms in 2010. He does not separately identify or control for the presence of after-tax cash flow metrics, and his sample includes firms frequently excluded from tax avoidance research including financial firms, utilities and those reporting losses. We exclude these firms, consistent with prior literature, because the CEO’s incentives to engage in tax avoidance are either limited or unclear. For example, managers of REITs have very little incentive to commit resources to tax avoidance (Manzon and Plesko 2001), and in 2010, managers of banks subject to TARP restrictions were not eligible to receive cash bonuses regardless of firm performance. In an untabulated analysis, we attempt to reconcile our findings to Gaertner’s (2014) and conclude that the difference in CETR results likely stems from differences in sample composition. When we re-estimate Model (4) from Table 4, Panel B, using a sample of 498 observations from S&P 500 firms from 2009 to 2011, including financial firms and utilities, we find a negative coefficient estimate of −0.03 on ATAX (two-tailed p value = 0.03). However, when we eliminate 71 observations from financial firms and utilities, we find an insignificant coefficient estimate on ATAX (−0.015, two-tailed p valued = 0.28).

We also re-estimate our main analysis after dividing the sample into foreign and domestic firms, because Newman (1989) suggests that firms with foreign operations are more likely to use after-tax incentives. Our results (untabulated) hold within both domestic and multinational subsamples, suggesting that foreign operations are not driving our results.

Results are robust to estimating UTBs using the methodology of Lisowsky et al. (2013) and Nesbitt (2014), who measure some control variables over multi-year horizons.