Evidence of differing market responses to beating analysts’ targets through tax expense decreases

Throughout the paper, including our sample selection and partitioning, we use “beat” to refer to “meet or beat.”

See also Krull (2004), Frank and Rego (2006), Schrand and Wong (2003), Blouin and Tuna (2007) for other papers that discuss opportunistic use of the tax expense.

Our conclusions are unchanged if we include firms that beat the forecast by more than five cents.

We estimate firm-specific prior persistence of tax changes for each year using the tax changes from the prior eight years. We require a minimum of five observations for each firm-year estimate. These constraints significantly reduce the sample used in our tests. The conclusions regarding the effect of tax decreases that permit firms to beat the forecast are unchanged when we use the full sample and exclude persistence measures.

Indirectly related, researchers use book-tax differences as a proxy for discretionary accruals to study pretax earnings management. Phillips et al. (2003) find that deferred tax expense is incrementally useful beyond accruals in detecting earnings management. Lev and Nissim (2004) and Hanlon (2005) find that firms with large book-tax differences have lower growth or less persistent earnings.

In 2002, Karen Pincus (Auditing professor, University of Arkansas) related the following anecdote from her participation in training sessions for auditors serving as the “second partner reviewer.” When asked what area of the financial statement these partners had the least comfort, they responded “the tax account” because there was no easy smell test for whether it was correct. Her observation is consistent with recent evidence that informal income tax expense procedures generated material control weaknesses post-Sarbanes-Oxley.

In untabulated tests, we observe that nearly all of our firms report nonmissing foreign tax expense (Compustat #64) or foreign pretax income (Compustat #273), suggesting that foreign tax rates affect their ETRs. Our results are unaffected by repatriations of foreign earnings following the American Jobs Creation Act of 2004, because only a handful of firms repatriated in 2004 (our last sample year). Nearly all firms waited until 200. Fewer than ten repatriations occurred in 2004 in Albring et al.’s (2007) sample of approximately 300 repatriations.

Starting in 2007, the FASB requires that corporations record the best estimate of the impact of a tax position only if that position is more likely than not of being sustained on [IRS] audit based solely on the technical merits of the position, thus reducing the flexibility that management judgment previously permitted. (Financial Interpretation No. 48, Accounting for Uncertainty in Income Taxes: an Interpretation of FASB Statement No. 109. Uncertain tax benefits became more of an issue following the perception that certain corporate tax shelters were increasing in use during our sample period (U.S. Treasury 1999). McGill and Outslay (2004) note that the ideal tax shelter would be a transaction that reduces taxable income without decreasing book income, generating an effective tax rate benefit. Evidence of firms releasing cushion prior to and as a result of adopting FIN 48 provides additional evidence that many large firms maintained excess reserves (Blouin et al. 2007).

Managers will use tax expense to beat targets only if their incentives are aligned with after-tax income. Phillips (2003) finds that firms whose CEOs’ bonuses are based on after-tax earnings report lower effective tax rates than do firms that base bonuses on pretax earnings. In addition, compensation plans in our time period are increasingly stock-based, and thus implicitly after-tax. Core et al. (2003) document that the average annual salary and bonus is only 30% of total CEO pay and the change in value of CEO equity holdings is more than eight times total CEO pay. This is consistent with the notion that net earnings play an important role in CEO motivation. Thus, although managers may prefer to manage pretax earnings first, they are still generally motivated to use tax expense if it represents their last chance to manage earnings (Dhaliwal et al. 2004).

Some details concerning tax characteristics, like deferred tax and ETR reconciliation components, are likewise only available in the tax footnote in the annual report.

We confirm that our results are qualitatively similar using the last forecast issued by an individual I/B/E/S analyst at least four trading-days prior to the annual earnings announcement. Our results are also robust to requiring that the consensus include at least five analysts and to eliminating the top and bottom deciles for the standard deviation of analyst forecasts. We use actual pretax earnings (Compustat data item 170) to calculate earnings prior to tax changes because the earnings reported by I/B/E/S differ from net income (Abarbanell and Lehavy 2002). Our results are robust to omitting firms with special items, extraordinary items or discontinued operations amounts, so we conclude that write-offs, often concentrated in the fourth quarter (Hayn and Watts 2002), do not affect our conclusions.

The integral method prescribed by APB No. 28 applies to other income statement accounts including cost of goods sold and selling, general, and administrative expenses. We focus on tax expense because the complexity and discretion in this account permit “last chance” earnings management.

Comprix et al. (2006) find that firms also decrease ETRs in the second and third quarters to meet quarterly earnings targets, although tax rate changes vary less in those quarters. Our results are robust to including only those firms that had no tax rate change between the second and third quarters, eliminating any concern that earnings management during the third quarter could explain our results. Bauman and Shaw (2005) find that analysts appear to incorporate the third-quarter estimate of the annual ETR in forecasting annual earnings. Thus, our fourth quarter result should not be induced by analysts ignoring a third-quarter change. It is possible that analysts receive additional guidance about tax rates during the fourth quarter. Tasker (1998) documents that 20% of analyst questions in conference calls are focused on earnings guidance and include questions such as “What tax rate should we be using for fiscal ’98?” Such guidance would reduce the earnings surprise on announcement. This effect would work against detecting a difference in the market’s response to beating the forecast by decreasing tax expense.

I/B/E/S earnings forecasts are reported on a basic or diluted basis based on the predominant EPS used by analysts. We obtain the primary/diluted indicator from the I/B/E/S detail file and use the corresponding denominator to compute Earnings_ETRq3. I/B/E/S earnings forecasts also frequently differ from EPS numbers reported in Compustat (Abarbanell and Lehavy 2002). Analysts forecast earnings without discontinued operations, extraordinary charges and other non-operating items. Our results are robust to omitting firms with special items, extraordinary items or discontinued operations amounts as reported by Compustat.

Our regression results are robust to including controls for unanticipated tax planning. We use tax return data available through 2002 in untabulated tests to control for the net amount of estimated tax overpayment as a proxy for unanticipated or late-year tax planning. We continue to find a significant market discount for tax decreases necessary to beat the forecast. This result suggests either that late-year ETR changes are unrelated to tax planning or that the stock market reacts unfavorably to ETR changes that appear opportunistic even in the presence of higher tax refunds. Our results are also robust to excluding the top and bottom deciles either of the third-quarter ETR or of the change in the ETR from the third to the fourth quarter, to control for extreme levels or changes.

The change in the fourth quarter effective tax rate could be mechanical due to unexpected changes in pretax income adding more or fewer dollars of tax at the statutory rate to the overall effective tax rate. Our results are robust to including a control for the revision in analysts’ consensus annual earnings forecasts from just prior to the third-quarter earnings announcement to the last consensus before the earnings announcement. We also use the estimate of ‘induced tax change’ developed in Dhaliwal et al. (2004) and refer readers there for a full discussion. Our results are robust to redefining our measures of tax changes to exclude the induced effect. Finally, the accumulated fourth quarter ETR does not always strictly equal the annual ETR. The ETRs are not statistically different, however and our results are robust to adding a constraint to require the ETRs to be equal.

We also include net operating assets (NOA) as a control for balance sheet “slack” (Barton and Simko 2002). NOA is not significant in our regressions and the inferences regarding our variables of interest are unchanged.

The median values of AFE w/ ETRq3 and TCC for Beat w/ Tax firms are −0.074 and 0.088, respectively. Median values for Beat w/o Tax firms are 0.032 and −0.009, respectively.

In untabulated tests, we include firms that missed the forecast in the regression. Firms missed the forecast by an average of two cents per share and had an average increase in income of four cents per share due to a decrease in tax expense during the fourth quarter. The intercept for firms that missed is significantly negative (−0.007, p < 0.000). Thus the 0.005 (p = 0.035) market reward for beating the forecast by decreasing tax expense is significantly higher than the penalty for failing to beat the forecast. The difference between firms that missed the forecast and firms that decreased tax expense to beat the forecast is insignificant for forecast errors smaller than three cents when the forecast is based on the consensus. This may be due to the presence of stale forecasts. When we use the last forecast, we continue to observe a market reward for beating the forecast by decreasing tax expense that is significantly higher than the penalty for failing to beat the forecast.

In untabulated univariate comparisons of firms with tax expense decreases in the fourth quarter versus firms with tax expense increases, we confirm that there are no significant differences in AFE, abnormal accruals, prior persistence, book-to market ratio, momentum, announcement window returns or returns over the following year. Firms that decrease tax expense are larger on average than firms with tax expense increases, however our regressions include a control for size. By definition, firms that decrease their tax expense have lower per share forecast error without a tax change (AFE w/ ETRq3) and higher tax change components (TCC) relative to firms that increase tax expense. Firms that decrease tax expense also have higher ETRs in the third quarter and lower ETRs in the fourth quarter relative to increase firms.