## The Moderating Role of Internal Control in Tax Avoidance: Evidence from a COSO-Based Internal Control Index in China

To address the concern that tax risk management and tax internal controls were not well understood even by those in the tax function (PwC 2004, 2), Big 4 accounting firms recommend using the Internal Control—Integrated Framework by the Committee of Sponsoring Organizations of the Treadway Commission (COSO 1992, 2013) to reduce the uncertainty over the facts themselves or uncertainty as to how well systems operate to arrive at the tax results (PwC 2004; EY 2014).

We thus study the role of internal control in tax avoidance by evaluating the efficacy of the COSO framework in tax risk management. First, we use a comprehensive COSO-based index in China that covers a firm's internal control over not only financial reporting, but also operations and compliance. Second, we perform quantile regressions to account for the entire tax avoidance distribution. These two key features enable us to find a nonlinear relation between internal control and tax avoidance by showing that internal control quality enhances tax avoidance for under-sheltered firms but curbs tax avoidance for over-sheltered firms. Different from prior studies that document a one-sided role of internal control in risk reduction, we find a two-sided moderating role of internal control in tax avoidance.

Specifically, our quantile regression analyses show a significantly positive (negative) relation between internal control quality and tax avoidance in the lower (upper) tail of the tax avoidance distribution. This finding indicates that high-quality internal control can subtly balance the benefits and costs to achieve an optimal level of tax avoidance commensurate with a firm's risk tolerance level and risk capacity. Among the five COSO components of internal control, we find that Control Environment, Control Activities, and Information and Communication have a moderating effect on tax avoidance, whereas Monitoring reduces tax avoidance for over-sheltered firms only. Consequently, the moderating role of internal control in tax avoidance alleviates tax volatility, supporting the accounting firms' recommendation to use COSO-based internal control in tax risk management.

Our results are robust to different regression methods and alternative measures of tax avoidance and internal control quality. Further analyses show that our results are not driven by alternative explanations. Our findings have important policy implications. Public firms should design internal control for not only the reliability of financial reporting, but also the effectiveness of nonfinancial operations. Our evidence suggests that a comprehensive internal control system can moderate a firm's tax avoidance behavior and thus enhance firm value through risk management.

## Trust and Compliance Effects of Taxpayer Identity Theft: A Moderated Mediation Analysis

Tax identity theft is an ongoing problem for tax authorities, especially the Internal Revenue Service (IRS). Tax identity theft occurs when someone uses the personal identification information of another taxpayer to file a fictitious tax return and obtain an unauthorized refund. Taxpayer service delivery occurs primarily through online technologies, making the online infrastructure of a tax authority vulnerable to hackers, especially if the online infrastructure is antiquated, as is the case at the IRS.

The purpose of our research is to examine how taxpayers react to the tax authority after their identity has been stolen and the tax authority is aware of their situation. We conduct a 2 × 3 between-subjects experiment using adult American taxpayers and manipulate two organizational accountability factors over which the tax authority has some degree of control: responsibility and responsiveness (Koppell 2005). Responsibility is the extent to which an organization is perceived as blameworthy for a mishap. Responsiveness is what an organization did to make amends following a mishap. We also measure perceptions of trust in the tax authority and taxpayers' compliance intentions.

The accountability literature suggests that individuals are more likely to trust an organization if it is accountable to them (i.e., responds to them). There is also a stream of tax compliance literature showing a positive association between trust and compliance. Combining these streams of literature, we predict that trust mediates the relation between responsiveness and compliance. Moreover, the trust repair literature suggests that an organizational response to a trust violation is conditional upon individuals' perceptions of that organization's blameworthiness for a mishap. Thus, we predict that our mediation model is conditional upon perceptions of organizational responsibility.

We find strong empirical support for our moderated mediation model. Specifically, we find that the indirect effect of IRS responsiveness on taxpayers' compliance intentions through trust in the IRS is stronger when IRS responsibility is lower (i.e., not blameworthy) versus higher. Collectively, our results suggest that when a tax authority is perceived as blameworthy for identity theft, there is little it can do to increase taxpayers' trust and subsequent compliance.

## The Unintended Consequences of Internal Controls Reporting on Tax Decision Making

Sarbanes-Oxley Section 404 requires management and the external auditor to report on the adequacy of the company's internal controls over financial reporting (ICFR). Material weaknesses in internal controls reduce the overall reliability of financial reporting and, as such, are required to be disclosed when identified (PCAOB 2007, ¶ 90). However, in a study of companies that file restatements while disclosing an underlying control weakness, Rice and Weber (2012) find that only 32.4 percent of these firms reported a material weakness in internal controls during the period that was later restated. Rice and Weber (2012) suggest that managers' incentives to detect and disclose weaknesses play an important role in whether detected internal control weaknesses are reported. The current study explores this suggestion by examining the disclosure decisions of tax executives who possess private information that a material weakness exists.

I use a 2 × 2 fully crossed experiment, varying the type of internal control deficiency (significant deficiency or material weakness) and error direction (tax refund or amount due) to investigate whether the ICFR regulations impact tax executives' decisions to amend a prior year return and disclose an internal control deficiency. One hundred twenty-three tax executives assume the role of a tax manager who has private information about an error that existed on the prior year tax return, and because of an internal control deficiency was also reported on the prior year's financial statements. Tax executives must choose whether to amend the prior year tax return, with the knowledge that amending will trigger disclosure of the internal control deficiency to the external auditors and the reporting requirements under Section 404.

While the overall compliance rate is high, the classification of a deficiency under ICFR as a material weakness significantly reduces the likelihood that tax executives will disclose the material weakness and amend the prior year tax return. One in six tax executives in this study would not amend (disclose) if the internal control deficiency would be considered a material weakness in internal control. When amending the tax return would result in both a material weakness in internal control and additional tax due, nearly one in four executives chose not to disclose the error and amend the tax return.

These findings suggest the ICFR regulations may have unintended consequences: tax executives may be less likely to report and correct an error (and the associated internal control deficiency) because of the disclosure requirements of the ICFR rules. Companies need to be concerned about incorrect financial reporting related to taxes and the potential for a material misstatement on the financial statements to go undetected. In addition, while the focus of the current study is on tax executives' decision making, if the results are generalizable to other corporate executives, there may be concern that additional financial statement misstatements, unrelated to taxes, may occur because of the reporting requirements of ICFR.

## Analysts, Taxes, and the Information Environment

We examine the extent to which analysts incorporate tax-based earnings information in their earnings forecasts relative to other earnings information. The notion that taxes and tax disclosures are complex and more difficult to understand relative to other accounting information is prevalent throughout the tax literature and supported by evidence. However, recent research suggests that information advantages not only mitigate analysts' EPS forecast errors (Hutton, Lee, and Shu 2012), but also mitigate analysts' ETR forecast errors (Bratten, Gleason, Larocque, and Mills 2017). Therefore, understanding the extent to which analysts incorporate tax-based earnings information in their earnings forecasts relative to other earnings information should help to broaden our understanding of the “black box” decision processes through which analysts add value to capital markets as information intermediaries (Bradshaw 2009).

To investigate whether analysts process tax-based earnings information differently relative to other accounting information, we regress analysts' forecast errors on three articulating earnings components, operating cash flows, accruals, and earnings generated by changes in the ETR (i.e., the tax change component of earnings [TCC]) using a broad sample of U.S. corporations from 1994 to 2017. We show that analysts' misestimation of tax-based earnings information, measured as the TCC, is statistically different from the misestimation of other, nontax accounting information, measured as accruals or operating cash flows. Our coefficient estimates suggest that, on average, analysts misforecast approximately $0.03 ($0.005) of future earnings due to the TCC (accruals and operating cash flows). These results suggest that, on average, analysts' reaction to tax-based earnings information is distinct from their reaction to other (nontax) accounting information.

We then examine the extent to which information environment attributes (i.e., the quality of accounting information, the extent of external monitoring, and the quality of external monitoring) (1) moderate the association between analysts' forecast errors and tax-based earnings information, and (2) affect analysts' processing of tax-based earnings information relative to other accounting information. We partition our sample into observations with strong and weak information environments and find that analysts misestimate tax-based earnings information only when firms have weak information environments; when firms have strong information environments, analysts' forecasts fully incorporate the implications of tax-based earnings information. Further, the difference in analysts' misestimation of tax-based earnings information between our strong and weak information environment settings is economically and statistically significant. In weak information environment settings, we document tax-related forecast errors in the range of 3.8 to 6.9 cents per share; these forecast errors largely disappear in strong information environment settings. Finally, we only observe a significant difference in analysts' misestimation of tax-based earnings information relative to accruals and operating cash flows when information environments are weak; the difference in the relative analysts' misestimation is also statistically and economically significant. The tax-related forecast errors are approximately three to five cents per share larger than the accrual and operating cash flow forecast errors in weak information environment settings. These results suggest that improving attributes of the information environment mitigates analysts' misestimation of tax-based earnings information compared to other accounting information.

Our study, along with Weber (2009) and Bratten et al. (2017), indicates that while analysts still on average do not fully incorporate tax-based information into their forecasts, there is mounting evidence to suggest that the notion that analysts do not understand taxes (Graham, Raedy, and Shackelford 2012) or that tax disclosures are inscrutable (McGill and Outslay 2004) may be outdated. For example, we show that analysts (1) view tax-based earnings and earnings quality information as persistent earnings forecasting components, and (2) can differentiate tax-based earnings information from tax-based earnings quality information. Both results are inconsistent with the maintained hypothesis in prior fundamental analysis studies that analysts treat all non-statutory ETR changes as transitory (Lev and Thiagarajan 1993; Abarbanell and Bushee 1997, 1998). Further, our results suggest that given appropriate information and resources, analysts fully incorporate the implications of tax-based earnings information for future earnings and exhibit no difference in their ability to incorporate the implications of tax-based earnings information relative to other accounting information. These results, along with the evidence in Bratten et al. (2017) that analysts' quarterly ETR forecasts are relatively more accurate than management as complexity increases, provide a more complete understanding of the source of analysts' tax-related forecast errors.

## Capital Gains Taxes and the Market Response to Earnings Announcements

In this paper, we investigate how stock price and trading volume are affected by the different capital gains tax rates applied to short-term and long-term capital gains when firms disclose public information. We extend the theoretical framework developed in Shackelford and Verrecchia (2002) linking differential capital gains taxes to price and volume, but allow for negative news in addition to positive news and also incorporate exogenous non-taxable, uninformed traders into the model. We are motivated by the likelihood that the short-term and long-term capital gains tax rate difference effects on price and volume will differ between positive and negative news announcements, and will be influenced by the presence of non-tax, uninformed trading.

To empirically investigate our model's predictions, we use stock price and taxable institutional shareholder trading data associated with firms' public quarterly earnings announcements. We follow the approach taken by Blouin, Smith Raedy, and Shackelford (2003) in their empirical investigation of the Shackelford and Verrecchia (2002) model, and extend it to investigate the additional predictions generated by our model. Specifically, our model predicts that the different capital gain tax rates applied to short-term and long-term capital gains magnify equity price responses surrounding earnings announcements in model regions where taxable investors sell. Using institutional shareholding trading data as a proxy for taxable investors' trades, we find that for capital gains, the equity price response to earnings changes is generally greater in those regions when this condition holds consistent with our model's prediction. In contrast, the price response empirical results for capital losses are not consistent with our model's predictions.

Regarding trading volume, our model predicts that short-term and long-term capital gains tax rate differences inhibit trading volume surrounding earnings announcements only in regions where there are capital gains and where different types of informed, taxable investors trade in opposite directions. Our empirical results are mixed relating to this prediction. When we employ a proxy for capital gains that does not rely on institution-specific shareholdings information, the results weakly support the model's prediction when capital gains are present.

In additional analysis, we also make predictions regarding the effect of the long-term and short-term capital gains tax rate differences on the association between the firm's two-day cumulative abnormal return and unexpected earnings, or the earnings response coefficient (ERC). Specifically, our model predicts that the estimated ERC will be greater for more positive earnings news and a larger capital gains tax rate difference. Our results are consistent with this prediction. In summary, our research contributes to the tax and capital market literatures by expanding the theoretical model in Shackelford and Verrecchia (2002) to show that the price and volume effects of capital gain tax rate differences are potentially more nuanced than Shackelford and Verrecchia (2002) suggest.

## Why Pay Our Fair Share? How Perceived Influence over Laws Affects Tax Evasion

We identify a negative effect of political activities such as lobbying: it leads other firms to not pay their taxes. We find that when firms perceive their competitors as being influential for governmental policymaking, they are less likely to pay their own taxes. This finding is important because governments worldwide lose significant tax revenue to tax evasion, with revenue losses estimated in the hundreds of billions of dollars annually. Further, as taxpayers' trust in the government sinks to an all-time low, it is important to understand the negative consequences of perceived unfairness. Our results suggest that a rise in some firms' perceived influence over governments leads to increased tax evasion by other firms, unfairly penalizing firms that fully comply with laws (pay their taxes) and placing an increased strain on already constricted government budgets.