Tax-Related Restatements and Tax Avoidance Behavior

Abstract

Our research empirically examines the relationship between tax avoidance and the likelihood of incurring a tax-related restatement, as well as the effects tax restatements have on future tax avoidance behavior. We predict and find that the association between tax avoidance and the likelihood of a tax-related restatement is nonlinear. Specifically, both relatively high levels and relatively low levels of tax avoidance compared to peer firms increase the likelihood of incurring a tax-related restatement. We consider whether the increased likelihood for high avoiders is attributable to obfuscation necessary to escape detection of tax avoidance or weak corporate governance. For high avoiders, we find evidence that both obfuscation and weak corporate governance may contribute to the likelihood of a tax-related restatement. As low avoiders should not need to obfuscate, we focus on corporate governance and find an increased likelihood when governance is weak. In response to a tax-related restatement announcement, we document a decline in tax avoidance for firms that have relatively high tax avoidance prior to the restatement announcement. We attribute this to the increased level of Internal Revenue Service (IRS) monitoring and strengthening of corporate governance that we observe post-restatement announcement. In contrast, we do not find evidence that low avoidance firms alter their tax avoidance after a tax-related restatement announcement, consistent with our finding that corporate governance does not improve post-restatement for low avoiders.

Keywords

tax avoidance tax-related restatements obfuscation corporate governance

Introduction

Based on a review of the extant literature, Wilde and Wilson (2018) propose a framework of determinants of corporate tax planning that includes three types of costs that influence corporate tax planning decisions. The first of the three costs, agency costs, are defined as “the misalignment of incentives between the principal (shareholders) and the agent (executives)” (Wilde & Wilson, 2018, p. 64) and occur when shareholders’ and managers’ risk preferences for tax planning are not aligned. In the Wilde and Wilson (2018) framework, agency costs can be mitigated with strong corporate governance. Implementation costs are the second type of costs and are defined as “frictions and constraints associated with implementing tax planning strategies” (Wilde & Wilson, 2018, p. 64). Implementation costs are those related to executing and maintaining tax planning decisions and are influenced by firm attributes, business models, strategies, and financial or other constraints. Finally, the third type of costs, outcome costs, are defined as “the expected outcomes costs associated with engaging in a particular tax strategy” (Wilde & Wilson, 2018, p. 64) and may include interest and penalties from reversals of tax positions and potential reputational effects. Wilde and Wilson (2018) promote the importance of developing a greater understanding of real consequences, both direct and indirect, of corporate tax planning and suggest that understanding these consequences may help create better understanding of the determinants and outcomes of tax planning decisions. In addition, they call for more research into how the different types of costs work together as determinants of tax planning decisions.

Our research answers this call and empirically tests tax-related restatements as a potential outcome cost of corporate tax planning, or lack thereof, by examining whether the level of tax avoidance influences the likelihood of incurring a tax-related restatement. In addition, we examine whether incurring a tax-related restatement becomes a determinant of future tax avoidance behavior. Finally, we consider agency costs and outcome costs as explanations for the relationship between the level of avoidance and a tax-related restatement, both before and after the actual incurrence of the restatement. Consistent with Seetharaman et al. (2011), we define a tax-related restatement as one in which an unintentional or intentional misapplication of a generally accepted accounting principle in the tax area is listed as a cause for the observed restatement. We define tax avoidance as tax planning decisions which minimize current taxes paid and may include activities ranging from fraudulent to benign as well as general tax planning that creates temporary and permanent tax savings.

Prior research indicates an association between financial reporting quality and restatements (Dechow et al., 2011; Ettredge et al., 2010). Desai et al. (2007) propose a theory which suggests that firms engaging in tax avoidance would need to engage in behavior designed to obfuscate the avoidance to escape detection. This obfuscation behavior may reduce outcome costs related to detection but has the potential to increase other types of outcome costs by negatively affecting the financial reporting quality of income tax-related accounts and increasing the likelihood of a required tax-related restatement. In contrast, firms engaging in lower levels of tax avoidance would not need to engage in obfuscation behavior, which could decrease the likelihood of a tax-related restatement.

Prior research (Armstrong et al., 2015; Chen et al., 2020; Chyz & Gaertner, 2018; Hsu et al., 2018) provides evidence that strong corporate governance mitigates extreme levels of tax avoidance at both the high and low ends of the distribution. These results could be indicative of corporate governance better aligning tax avoidance with shareholder preferences. A firm may be either over or under avoiding taxes due to weak corporate governance failing to lead to proper incentive alignment. These same weak governance mechanisms could also result in poor quality financial reporting (Carcello et al., 2011; Cohen et al., 2004) of the tax accounts and an increased likelihood of restatement.

We identify a sample of firms reporting a tax-related restatement for reporting periods 2004–2013 using Audit Analytics. We hand collect information necessary for our analyses from 8-Ks and 10-Ks and supplement with Compustat and other common sources of financial data. Our results reveal that the association between tax avoidance, measured as the industry-size adjusted cash effective tax rate (CETR), and the likelihood of incurring a tax-related restatement is not linear. Specifically, using logistic regression with a propensity score–matched sample, we find that firms with tax avoidance at the high end of the distribution relative to their peers are more likely to incur a tax-related restatement, while firms with tax avoidance at the low end of the distribution relative to their peers are also more likely to incur a tax-related restatement. We confirm this result using an analysis with a squared term and find the likelihood of a tax-related restatement is lowest at the 61st percentile of the tax avoidance distribution.

In additional analysis, we explore the channels by which the level of tax avoidance influences the likelihood of a tax-related restatement.¹ We find that high tax avoidance only increases the likelihood of a tax-related restatement when obfuscation, proxied by either tax accrual quality or analyst forecast accuracy, is high. Thus, by engaging in behavior designed to reduce one outcome cost, the discovery of tax avoidance by the tax authority, firms may be increasing their exposure to a different outcome cost, a tax-related restatement. We do not find that corporate governance, proxied by institutional ownership, influences the relationship between tax avoidance and the likelihood of a tax-related restatement for high avoiders. We do find that the presence of a CEO who is also the chairman of the board, which is indicative of weak corporate governance (Brickley et al., 1997; Rechner & Dalton, 1991), is associated with a higher likelihood of a tax-related restatement for high avoiders. Low avoiders are more likely to experience a tax-related restatement only when corporate governance, proxied by both measures, is weak. These results highlight the complexity of the corporate governance system (i.e., Baber et al., 2012; Misangyi & Acharya, 2014) where there is not a one-size-fits-all solution to mitigating agency costs.

Next, we examine the tax avoidance behavior of firms after they announce a tax-related restatement. Studies show that firms take actions to improve their corporate governance in response to a restatement (Desai et al., 2006; Farber, 2005; Halperin & Lai, 2015; Mande & Son, 2013; Srinivasan, 2005). Extant research finds that a change in corporate governance will move tax avoidance in one direction or another (Bauer, 2016; Frank et al., 2018). Improved corporate governance following a tax-related restatement announcement could lead to heightened focus on the tax planning function, resulting in a change in tax avoidance to a level more consistent with shareholder preferences.

In addition to improved corporate governance, a tax-related restatement could increase management’s perception of firm risk from increased attention by external monitors. Prior research establishes that tax avoidance decreases when there is a real or perceived increase in external monitoring activities (Hoopes et al., 2012; Kubick et al., 2016). A perceived increase in monitoring after the announcement of a tax-related restatement would serve to further reduce tax avoidance behavior of pre-restatement announcement high tax avoiding firms and constrain, but potentially not eliminate, the expected increase in tax avoidance behavior of pre-restatement announcement low avoiding firms.

Using a difference-in-difference research design with tax avoidance as the dependent variable, we document that high tax avoidance firms reduce their level of tax avoidance following a tax-related restatement announcement. In a supplemental analysis, we find high avoiders also experience an improvement in corporate governance (i.e., decreased likelihood of a Dual CEO) and an increase in IRS monitoring (i.e., downloads of financial reports by the IRS following Bozanic et al., 2017²) following a tax-related restatement announcement, providing evidence that both improved governance and increased external monitoring are channels by which a tax-related restatement constrains future tax avoidance behavior for high avoiders. In contrast, we do not observe a change in behavior for low tax avoidance firms following a tax-related restatement announcement, consistent with our supplemental analysis showing no significant change in governance following a tax-related restatement for low avoiders.

We believe that our results of a change in tax avoidance following a tax-related restatement announcement are not due to mean reversion for three reasons. First, we find that the matched control firms that do not experience a tax-related restatement do not change their tax avoidance behavior over the same time period. Second, we find that firms incurring a non-tax-related restatement do not alter their tax avoidance behavior following the restatement announcement. Finally, in the spirit of Kim et al. (2019), we argue that the evidence of cross-sectional variation in tax avoidance trends following the restatement announcement is counter to mean reversion.

Our study addresses Wilde and Wilson’s (2018) call for further research into better understanding of the direct and indirect outcome costs of tax avoidance. Consideration of the role of agency costs and two potential outcome costs (detection costs and incurrence of a tax-related restatement) in the tax planning decision contributes to prior literature on the relationship between tax avoidance and financial reporting quality and transparency (Balakrishnan et al., 2019; Desai & Dharmapala, 2006; Li et al., 2018; Wilson, 2009) by examining the channels through which tax avoidance influences a firm’s information environment. We also extend prior literature on changes in tax avoidance behavior following a negative financial reporting event (Bauer, 2016; Kubick et al., 2016) by showing the importance of considering both the firm’s prior tax avoidance behavior and characteristics of the event itself. This finding is also relevant to studies examining tax avoidance behavior changes following other types of events.

Our study also has practical implications. Managers should be aware that a tax-related restatement is a potential outcome cost of tax avoidance outside of stakeholders’ expected range. Furthermore, shareholders should know that some, but not all, governance mechanisms mitigate the likelihood of a tax-related restatement for high avoiders. Finally, while a tax-related restatement leads a high avoiding firm to move tax avoidance toward the expected level, low avoiders appear to remain below the expected level, even in the absence of increased IRS monitoring.

Background and Hypotheses

Tax-Related Restatements as an Indirect Consequence of Tax Avoidance

Shareholders prefer firms to engage in tax planning to the extent that it increases the present value of future cash flows (Scholes et al., 2015). Empirical studies show that tax avoidance increases shareholder value, at least for well-governed firms (Desai & Dharmapala, 2009; Wilson, 2009). Shareholders weigh the benefits of tax savings against the costs of tax avoidance, including risk (Drake et al., 2019) and potential negative impacts to firm reputation (Austin & Wilson, 2017; Dyreng et al., 2016; Graham et al., 2014) and develop an expectation of the level of tax avoidance that will maximize shareholder value. Wilde and Wilson (2018) argue that, absent agency concerns, firms will pursue tax strategies that achieve a net marginal benefit. We use the term “expected” to refer to the level of tax avoidance that achieves that net marginal benefit. Thus, we are using the term “expected” consistently with how Kim et al. (2019) use the term “optimal.”

It is not possible to specifically identify expected tax avoidance for a given firm; however, it is likely that when compared with peer firms, on average, firms with tax avoidance at the high (low) end of the distribution are above (below) that expected level. Hanlon and Slemrod (2009) find that the market reacts negatively to announcements that firms engaged in a tax shelter; however, they find the reaction is less negative for firms with higher CETRs, suggesting that shareholders believe management should achieve a specific level of tax avoidance. In addition, Inger et al. (2018) find that when tax avoidance is high, investors value tax avoidance positively when tax footnotes are relatively complex and negatively when tax footnotes are relatively straightforward. This result indicates that investors’ evaluations of management’s tax avoidance activities vary based on their perception of whether management is taking actions to reduce detection risk.

We consider two possible situations, obfuscation of tax avoidance and weak corporate governance, which could lead to a relationship between tax avoidance and the likelihood of incurring a tax-related restatement. Desai et al. (2007) argue that tax avoidance requires management to engage in activities designed to obfuscate the avoidance to escape detection by the IRS.³ Prior literature provides evidence that tax avoidance can negatively affect the quality and transparency of the firm’s information environment (Balakrishnan et al., 2019; Desai & Dharmapala, 2006; Li et al., 2018; Wilson, 2009), supporting the Desai et al. (2007) argument that obfuscation occurs. Tax avoidance activities might affect financial reporting quality in general; however, based on evidence from prior literature (Hanlon et al., 2014; Inger et al., 2018; Kubick et al., 2016), we expect the most common effect to occur in the tax-related accounts. Thus, the need for obfuscation could lead to an increased likelihood of a tax-related restatement for high avoiders. Firms engaging in low levels of tax avoidance behavior would not need to obfuscate, which could lead to higher quality reporting of their tax accounts and a lower risk of tax-related restatement.

Each firm’s corporate governance should serve as an internal monitor that moves the firm’s tax avoidance behavior toward the expected level by aligning management incentives with shareholder interests (Bauer, 2016; Desai & Dharmapala, 2006; Desai et al., 2007). Prior studies have found mixed results in terms of the relationship between corporate governance and tax avoidance (Jimenez-Angueira, 2018; Khurana & Moser, 2013; Minnick & Noga, 2010). Armstrong et al. (2015) argue that the mixed results could be the consequence of focusing on mean or median tax avoidance and instead examine the relationship between corporate governance, operationalized as both board independence and financial sophistication, and tax avoidance at the high and low ends of the distribution. Their study finds that the relationship is positive for low tax avoiders and negative for high tax avoiders, consistent with under(over)-investment in tax avoidance in the absence of monitoring. Subsequent research has found a similar relationship between other measures of corporate governance and tax avoidance (Chen et al., 2020; Chyz & Gaertner, 2018; Hsu et al., 2018). These studies are consistent with the idea that weak corporate governance mechanisms could lead to tax avoidance either above or below the expected level. A weak governance environment may also lead to improper applications of US Generally Accepted Accounting Principles (GAAP).

Finally, prior research has also directly examined the association between financial reporting quality and restatements in general (Choudhary et al., 2016; Dechow et al., 2011). Dechow et al. (2011) find a negative relationship between general restatements and financial reporting quality, while Choudhary et al. (2016) document an association between general restatements and tax reporting quality. Regardless of whether high tax avoidance is creating a need for obfuscation or whether tax avoidance above the expected level is an outcome of weak corporate governance, we expect it to also be associated with decreased financial reporting quality. Thus, we anticipate that higher than expected tax avoidance will be associated with a higher likelihood of tax-related restatement. This expectation leads us to H1a for high tax avoiders. For low tax avoiders, we state H1b in the null form due to the competing arguments for the effect of lower than expected tax avoidance on the likelihood of a tax-related restatement.

Hypothesis 1a (H1a): Relatively high levels of tax avoidance will increase a firm’s likelihood of incurring a tax-related restatement.

Hypothesis 1b (H1b): Relatively low levels of tax avoidance will not impact a firm’s likelihood of incurring a tax-related restatement.

Tax Avoidance Following the Announcement of a Tax-Related Restatement

Anecdotal evidence suggests that a tax-related restatement is a serious event for corporations with potentially important consequences,⁴ and prior research demonstrates firm behavioral changes after the detection of financial misreporting (Desai et al., 2006; Halperin & Lai, 2015; Mande & Son, 2013; Srinivasan, 2005). Most relevant to our research, Farber (2005) provides evidence that firms act to improve their corporate governance after being identified by the Securities and Exchange Commission (SEC) as fraudulently manipulating their financial statements.

A related line of research provides evidence that a firm’s tax avoidance behavior changes in reaction to a change in the firm’s corporate governance. Bauer (2016) finds that an increase in corporate governance (as evidenced by remediation of a tax-related internal control weakness) increases tax avoidance behavior. Frank et al. (2018) use the post-Sarbanes Oxley Act (SOX) period as a proxy for increased corporate governance and find that tax avoidance behavior decreases when corporate governance is improved. Regardless of the varying results, these studies uniformly provide evidence that a firm’s tax avoidance reacts to a change in corporate governance. A change in corporate governance post restatement should reduce agency costs and better align tax planning with shareholders’ preferences, such that firms who were under avoiding prior to the restatement announcement will increase their tax avoidance behavior while the firms who were over avoiding will decrease their tax avoidance behavior.

Another important form of influence on a firm’s tax avoidance relates to external monitors, who should serve to restrict the firm’s tax avoidance behavior by increasing monitoring activities (Hanlon et al., 2014; Hoopes et al., 2012; Kubick et al., 2016) and therefore the risk and cost of detection (Allingham & Sandmo, 1972). Fox and Wilson (2020) find that IRS downloads of firm financial statements increase in the periods immediately following a general (rather than tax specific) restatement announcement, providing evidence that external monitoring activities increase post-restatement. Hoopes et al. (2012) find that public firms undertake less tax avoidance when the expected likelihood of an IRS audit increases. An increase in detection risk post-restatement due to an increase in IRS monitoring should serve to further decrease avoidance behavior for high avoiding firms and constrain, but not eliminate, the increase in tax avoidance behavior for low avoiding firms.

Shareholders are willing to bear some level of risk in exchange for an increase in shareholder value. Kim et al. (2019) observe that the optimal level of tax avoidance may vary across time. The level of tax avoidance expected by shareholders may be lower after a restatement due to the increase in detection risk; however, the expected level would likely still be higher than the pre-restatement avoidance of most low avoiding firms. Therefore, low avoiding firms should still be likely to increase their tax avoidance behavior post-restatement. Thus, we present our second hypothesis:

Hypothesis 2a (H2a): Firms engaging in relatively high tax avoidance prior to the restatement announcement will decrease their tax avoidance behavior subsequent to the restatement announcement.

Hypothesis 2b (H2b): Firms engaging in relatively low tax avoidance prior to the restatement announcement will increase their tax avoidance behavior subsequent to the restatement announcement.

Research Design and Results

Sample Selection Procedures and Sample Profile

We create a database of information using Audit Analytics as the initial source to identify firms with tax-related restatements announced through 2014 for restated years 2004–2013.⁵ We focus on annual restatements both because quarterly reports contain reduced tax disclosures compared to annual reports and because the level of tax disclosures reported in quarterly reports varies among firms. The Non-Reliance Restatement section of Audit Analytics provides a reason code for each restatement; reason code 18 specifically relates to restatements due to “Tax Expense/benefit/deferral/other (FAS 109) issues.”⁶ We use this information to identify tax-related restatements for our sample. Specifically, we require there to be a tax-specific issue and not just a change to a tax account resulting from an error or misapplication of US GAAP in another account. We examine restatement filings (8-Ks, 10-Ks, and 10-KAs) for each firm to confirm that the information provided in the 8-K disclosing the intent to restate is followed by an actual restatement in a subsequent 10-K or 10-KA which includes a tax issue. We also determine the specific tax issue that generated the restatement using a list of categories of tax-related restatements provided by Deloitte (2008) and augmented with additional categories identified during our collection process, summarized in Supplemental Appendix A. After removing firms without data necessary to calculate regression variables, our sample consists of 397 restatement observations from 228 unique firms.

Tax Avoidance and Likelihood of Tax-Related Restatement

To investigate the association between tax-related restatements and tax avoidance, we estimate a logistic regression model of a binary variable (Tax-related Restatement = 1 or 0) on tax avoidance and control variables⁷ predicted to be associated with tax-related restatements:

\begin{matrix} Logit (Tax__Restate__Year = 1) & = F (α + β_{1} High__Tax__Avoi d_{it} \\ + β_{2} Low__Tax__Avoi d_{it} + β_{3} M {W_{i}}_{t} + β_{4} Tax__M W_{it} + β_{5} TA Q_{it} \\ + β_{6} Accrual s_{it} + β_{7} Discretionary__Accrual s_{it} + β_{8} Big 4_{it} \\ + β_{9} Specialis t_{it} + β_{10} Tenur e_{it} + β_{11} Leverag e_{it} + β_{12} Eq__In c_{it} \\ + β_{13} Los s_{it} + β_{14} Tax__Fee s_{it} + β_{15} NO L_{it} + β_{16} Merge r_{it} \\ + β_{17} Goodwil l_{it} + β_{18} For__Comple x_{it} + β_{19} For__In c_{it} \\ + β_{20} Inst__Ow n_{it} + β_{21} Dual__CE O_{it} + β_{22} FinExp__A C_{it} \\ + β_{23} Ind__Boar d_{it} + β_{24} Forecast__Accurac y_{it} \\ + β_{25} IRS__Attentio n_{it} + Industrydummies \\ + Yeardummies + ε) \end{matrix}

Model 1

The dependent variable, Tax_Restate_Year, is set to 1 for firm-years that are restated in a subsequent 10-K, 10-KA, or 8-K for a tax-related reason. Potential control firms consist of all U.S. Compustat firms with necessary data to calculate the model variables and excludes firms not subject to 404. Control firms have not experienced any type of restatement during the sample period. Restatement firms are matched to control firms using propensity score matching to ensure that the non-restating control firms have a similar likelihood of incurring a tax-related restatement based on firm characteristics that are associated with restatements. This technique is common in restatement research (Amel-Zadeh & Zhang, 2015; Guo et al., 2016; among others) to address the endogeneity concern that firm-specific attributes influence the variable of interest and are also correlated with a non-randomly assigned treatment. A first-stage logit model is estimated to determine propensity scores, with the dependent variable coded 1 if an observation has a tax-related restatement (0 otherwise). Excluding the variable of interest (i.e., tax avoidance), the independent variables are all firm-specific control variables in Model 1, following Francis et al. (2013). We then match treatment and control firms based on industry and year and impose a caliper (i.e., maximum allowed difference in propensity score between the treatment and matched firm) of .01.⁸ This matching procedure reduces our sample to 221 tax-related restatement firm year observations.

We measure tax avoidance as the industry-size adjusted CETR⁹ following Balakrishnan et al. (2019) whose measure captures cross-sectional variation in a firm’s tax planning and benchmarks a firm’s level of tax avoidance relative to similar sized firms in their industry. Following Balakrishnan et al. (2019), we subtract a firm’s CETR from the median CETR of the portfolio of firms comprised of firms in the same quintile of assets, Fama French 48 industry, and year (Tax_Avoid). Positive numbers of Tax_Avoid indicate tax avoidance is above the size-industry median. We use the entire sample of Compustat firms with necessary data to calculate Tax_Avoid.

To test Hypothesis 1, we create indicator variables set to 1 when Tax_Avoid is in the top quartile (High_Tax_Avoid) and bottom quartile (Low_Tax_Avoid)¹⁰ relative to industry peers in the year that was restated. Under Hypothesis 1a we expect a positive coefficient on High_Tax_Avoid, suggesting that firms engaging in high levels of tax avoidance relative to their peers are more likely to incur a tax-related restatement. Under null Hypothesis 1b, we do not predict a coefficient on Low_Tax_Avoid. We include numerous variables identified by prior research to be associated either with the likelihood of incurring a restatement or with tax avoidance. For purposes of brevity, we do not discuss the control variables; however, specific variable definitions and references to prior literature are presented in Supplemental Appendix B.

Results for Determinants of Tax-Related Restatements—Hypotheses 1a and 1b

Univariate results

Table 1 presents descriptive statistics for the model of the determinants of tax-related restatements. Column 1 presents means, standard deviations, and medians for tax-related restatement observations, and Column 2 presents means for the matched sample. Column 3 presents a covariate balance check for the matched sample. With few exceptions (MW, Leverage, NOL, and Inst_Own),¹¹ our matched pairs are balanced as evidenced by the statistically insignificant differences in the means (and untabulated medians) of our dependent and independent variables between tax-related restatement and matched control observations.

Table 1.

Tax-Related Restatement and Matched Sample Descriptive Statistics.

Variable	Column 1				Column 2	Column 3
	n	Tax-related restatements			Control (one-to-one PSM match)	Difference
	n	M	SD	Median	M	Difference	t-value	p-value
Tax_Avoid	221	−0.061	0.216	−0.062	−0.050	−0.012	0.640	.523
MW	221	0.127	0.333	0.000	0.059	0.068	2.470	.014
Tax_MW	221	0.036	0.187	0.000	0.027	0.009	0.540	.588
TAQ	221	−0.007	0.018	−0.003	−0.005	−0.002	1.600	.129
Accruals	221	−0.004	0.193	−0.009	0.014	−0.019	0.820	.411
Discretionary_Accruals	221	0.021	0.081	0.007	0.008	0.093	1.500	.135
Big4	221	0.851	0.357	1.000	0.815	0.036	1.020	.309
Specialist	221	0.181	0.386	0.000	0.127	0.054	1.580	.114
Tenure	221	11.362	8.930	8.000	11.045	0.317	0.370	.708
Leverage	221	0.197	0.213	0.159	0.158	0.039	2.020	.044
Eq_Inc	221	0.158	0.366	0.000	0.136	0.023	0.670	.503
Loss	221	0.276	0.448	0.000	0.290	−0.014	0.320	.752
Tax_Fees	221	0.115	0.128	0.081	0.099	0.016	1.330	.186
NOL	221	0.507	0.501	1.000	0.615	−0.109	2.310	.021
Merger	221	0.529	0.500	1.000	0.525	0.004	0.100	.924
Goodwill	221	0.151	0.154	0.102	0.179	−0.028	1.450	.181
For_Complex	221	0.389	0.489	0.000	0.344	0.045	0.990	.325
For_Inc	221	0.016	0.037	0.002	0.018	−0.002	0.620	.536
Inst_Own	221	0.259	0.358	0.000	0.345	−0.087	2.430	.016
Dual_CEO	221	0.520	0.501	1.000	0.466	0.054	1.140	.255
FinExp_AC	221	0.454	0.263	0.333	0.491	−0.037	1.430	.154
Ind_Board	221	0.817	0.154	0.818	0.830	−0.013	0.960	.336
Forecast_Accuracy	221	−0.031	0.238	−0.001	−0.027	−0.004	0.140	.890
IRS_Attention	221	5.891	14.2049	2.000	7.701	13.433	1.100	.274

Note.Table 1 presents descriptive statistics. Column 1 presents descriptive statistics for tax-related restatement observations included in multivariate analyses. Column 2 presents descriptive statistics for a matched control sample based on one-to-one propensity score matching (radius = .01). Specifically, a first-stage logit model is estimated, the dependent variable is coded 1 if an observation has a tax-related restatement (0 otherwise), and the independent variables are all of the firm-specific control variables, excluding the variable of interest, Tax_Avoid. Match firms have not experienced a restatement in any year of sample period 2004–2013. Column 3 presents a covariate balance check of a comparison of means between tax-related restatement observations and the matched control sample. Variables are defined in Supplemental Appendix B.

Multivariate results

Next, we discuss the multivariate results presented for the model of determinants of tax-related restatements in Table 2. Column 1 presents the results of the first-stage logit model used to estimate propensity scores with 228 restatement firms having sufficient data for the analysis. For brevity, we only discuss results of control variables in Column 1. Firms with material weaknesses are more likely to incur a tax-related restatement (p < .0001). However, firms with tax-specific material weakness are not more likely to incur a tax-related restatement (p = .5873).¹² Consistent with Choudhary et al. (2016), we find TAQ decreases the likelihood of a tax-related restatement (p = .0462). While Discretionary_Accruals are marginally associated with an increased likelihood of restatement (p = .1301), Accruals are insignificant. We attribute the lack of statistical significance on these variables to TAQ being a better measure of the firm’s tax reporting environment. Interestingly, Big4, Specialist auditors, and Tax_Fees¹³ increase the likelihood of a tax-related restatement (p < .0694), possibly due to the complex clients served by these categories of auditors. Consistent with prior literature, we find that Leverage is positively associated (p = .001) with the likelihood of incurring a tax-related restatement. In contrast, NOL is negatively associated with the likelihood of incurring a tax-related restatement (p < .0937), possibly due to the reduced need to avoid tax, and subsequently obfuscate, when NOLs are available to offset taxable income. Consistent with prior literature, strong governance, proxied by Inst_Own and lack of a dual CEO, reduces the likelihood of incurring a tax-related restatement (p < .0583).

Table 2.

Tax-Related Restatement Logistic Regression Model.

Parameter	Column 1: First-stage logit model			Column 2: Baseline industry-size adjusted CETR			Column 3: High/low avoidance based on quartile industry-size adjusted CETR			Column 4: High/low avoidance based on decile industry-size adjusted CETR			Colum 5: Square tax avoidance term
Parameter	Estimate	Wald chi-square	Pr > chi-square	Estimate	Wald chi-square	Pr > chi-square	Estimate	Wald chi-square	Pr > chi-square	Estimate	Wald chi-square	Pr > chi-square	Estimate	Wald chi-square	Pr > chi-square
Intercept	−3.6470	30.1218	<0.0001	−0.3579	0.1186	0.7305	−0.6200	0.3459	0.5564	−0.6398	0.3175	0.5731	4.0935	3.3091	0.0689
High_Tax_Avoid							0.7790	9.1220	0.0025	1.0093	8.2273	0.0041
Low_Tax_Avoid							0.6572	4.7201	0.0298	0.9172	5.7195	0.0168
Tax_Avoid				−0.2558	0.1979	0.6564							−0.0902	4.9044	0.0268
Tax_Avoid × Tax_Avoid													0.0005	4.7905	0.0286
MW	1.0444	17.3351	<0.0001	1.3874	7.6797	0.0056	1.3716	7.2008	0.0073	−0.0565	0.0179	0.8937	1.3051	6.4913	0.0108
Tax_MW	−0.2276	0.2945	0.5873	−1.1543	2.3305	0.1269	−1.1065	2.0939	0.1479	0.8316	1.2522	0.2631	−1.0691	1.9596	0.1616
TAQ	−4.7249	3.9750	0.0462	−11.5966	1.7698	0.1834	−10.1691	1.3320	0.2485	−54.3096	6.7601	0.0093	−11.5608	1.7983	0.1799
Accruals	0.1025	0.1475	0.7010	0.0339	0.0052	0.9423	0.0540	0.0132	0.9087	−0.1964	0.1400	0.7083	0.0140	0.0009	0.9765
Discretionary_Accruals	1.2623	2.2916	0.1301	2.2111	2.5950	0.1072	2.4723	3.1409	0.0763	−0.5553	0.1431	0.7052	2.2824	2.7033	0.1001
Big4	0.4377	4.2905	0.0383	0.3135	0.9184	0.3379	0.2453	0.5412	0.4620	−0.3168	0.7717	0.3797	0.2707	0.668	0.4137
Specialist	0.3378	3.2976	0.0694	0.4885	2.4091	0.1206	0.5095	2.5706	0.1089	0.6325	3.6764	0.0552	0.4458	1.9821	0.1592
Tenure	0.0021	0.0723	0.7880	0.0074	0.3262	0.5679	0.0053	0.1647	0.6848	−0.0101	0.7108	0.3992	0.0060	0.2148	0.643
Leverage	0.9661	10.7547	0.0010	1.4987	5.8509	0.0156	1.5694	6.2415	0.0125	0.2359	0.1500	0.6986	1.4320	5.2477	0.022
Eq_Inc	0.0597	0.0916	0.7622	0.1722	0.3057	0.5803	0.2122	0.4490	0.5028	−0.1064	0.1212	0.7277	0.2175	0.4832	0.487
Loss	−0.0538	0.0881	0.7666	−0.3868	1.8102	0.1785	−0.5447	3.1568	0.0756	−0.2273	0.5034	0.4780	−0.5902	3.7569	0.0526
Tax_Fees	1.2637	4.9779	0.0257	1.1901	1.7830	0.1818	1.3581	2.2951	0.1298	1.5368	2.4879	0.1147	1.2761	2.0347	0.1537
NOL	−0.2380	2.8096	0.0937	−0.5701	6.3953	0.0114	−0.4975	4.7088	0.0300	−0.1409	0.3810	0.5371	−0.5482	5.7961	0.0161
Merger	0.0256	0.0283	0.8663	0.2168	0.7808	0.3769	0.2077	0.7007	0.4025	0.1589	0.4677	0.4940	0.2169	0.7769	0.3781
Goodwill	−0.5589	1.4059	0.2357	−1.4643	3.7459	0.0529	−1.3582	3.1967	0.0738	0.0262	0.0011	0.9736	−1.3753	3.2746	0.0704
For_Complex	0.0751	0.2495	0.6174	0.1595	0.4527	0.5010	0.1706	0.5061	0.4768	−0.2741	1.3128	0.2519	0.1625	0.4637	0.4959
For_Inc	0.7091	0.1669	0.6829	−1.1707	0.1664	0.6834	−0.8807	0.0903	0.7638	−0.9834	0.1599	0.6892	−1.3176	0.2049	0.6508
Inst_Own	−0.5170	6.0549	0.0139	−0.6222	4.1383	0.0419	−0.6322	4.1639	0.0413	−0.1809	0.3384	0.5607	−0.5240	2.8706	0.0902
Dual_CEO	0.2628	3.5855	0.0583	0.2162	0.9203	0.3374	0.1539	0.4540	0.5005	0.6306	7.2305	0.0072	0.2157	0.9099	0.3401
FinExp_AC	−0.1293	0.2453	0.6204	−0.5726	2.0149	0.1558	−0.6417	2.4641	0.1165	−0.3834	0.8808	0.3480	−0.6163	2.3048	0.129
Ind_Board	−0.4618	0.9652	0.3259	−0.3378	0.1978	0.6565	−0.3972	0.2675	0.6050	0.0029	0.0000	0.9969	−0.4381	0.33	0.5656
Forecast_Accuracy	0.0010	0.0479	0.8268	−0.1256	0.1085	0.7419	−0.1042	0.0768	0.7817	−0.0580	0.1777	0.6734	−0.1650	0.1889	0.6638
IRS_Attention	−0.0004	0.0108	0.9173	−0.0057	0.6449	0.4219	−0.0068	0.8989	0.3431	0.0009	0.0139	0.9062	−0.0054	0.5893	0.4427
Industry	Yes			Yes			Yes			Yes			Yes
Year	Yes			Yes			Yes			Yes			Yes
N	13,620	(228 tax restatements)		442	(221 tax restatements)		442	(221 tax restatements)		442	(221 tax restatements)		442	(221 tax restatements)
Model chi-square/p-value	142.8507	<0.0001		47.8	0.0647		60.5263	0.0739		50.4626	0.0337		55.5693	0.0576
ROC curve	0.806			0.7152			0.7092			0.7060			0.7065
Max rescaled Rsq	.0669			.1423			.1706			.1476			.1575

Note.Table 2 presents the results of a logistic regression modeling the likelihood of incurring a tax-related restatement. p-values are two-tailed. Tax-related restatement firms are matched to a control sample based on one-to-one propensity score matching (caliper = .01). Specifically, a first-stage logit model is estimated, the dependent variable is coded 1 if an observation has a tax-related restatement (0 otherwise), and the independent variables are all of the firm-specific control variables in Table 1, but excluding the variable of interest, Tax_Avoid. Match firms have not experienced a restatement in any year of the sample period 2004–2013. Column 1 presents the results of the first-stage logit model using the population of firms with data necessary to calculate variables. Column 2 presents a baseline model with the propensity score–matched sample. Column 3 defines High_Tax_Avoid (Low_Tax_Avoid) as observations in the top (bottom) quartile of the industry-size adjusted cash effective tax rate. Column 4 defines High_Tax_Avoid (Low_Tax_Avoid) as observations in the top (bottom) decile of the industry-size adjusted cash effective tax rate. Column 5 includes a square-term of Tax_Avoid. We add 1 to Tax_Avoid to account for the industry-adjustment resulting in negative values for some observations and then multiply by 100 to account for decimal variables in a quadratic model. Variables are defined in Supplemental Appendix B. CETR = cash effective tax rate.

Using the propensity score–matched sample, we next run a baseline test of the association between tax avoidance and the likelihood of a tax-related restatement. Results are presented in Table 2, Column 2. The insignificant coefficient on Tax_Avoid (p = .6564) suggests that tax avoidance does not have a linear association with the likelihood of a tax-related restatement. Column 3 presents results using a quartile definition of high and low tax avoidance to test Hypothesis 1. Consistent with Hypothesis 1a, the positive and significant coefficient of .779 on High_Tax_Avoid (p = .0025) suggests that firms with high tax avoidance relative to their peers are more likely to incur a tax-related restatement. The coefficient of .6572 on Low_Tax_Avoid is also positive and significant (p = .0298), suggesting that firms with low tax avoidance relative to their peers are also more likely to incur a tax-related restatement. These results are economically meaningful. The odds ratio indicates that firms with high tax avoidance were 2.179 times more likely to have a tax-related restatement than firms that had neither high nor low tax avoidance. In addition, firms with low tax avoidance were 1.929 times more likely to have a tax-related restatement than firms that have neither low nor high tax avoidance. Column 4 presents results using a decile definition of high and low tax avoidance. Again, the coefficients on High_Tax_Avoid and Low_Tax_Avoid are positive and significant (p = .0041 and .0168, respectively), confirming the results in Column 3.

We include a square term of Tax_Avoid in Column 5¹⁴ and find further evidence of a nonlinear association. Specifically, the coefficient of −.0902 on Tax_Avoid is negative and significant (p = .0268), whereas the coefficient of .0005 on the square-term Tax_Avoid × Tax_Avoid is positive and significant (p = .0286). This suggests that as tax avoidance increases from a low level, the likelihood of a tax-related restatement declines, but that tax avoidance eventually reaches a level that increases the likelihood of a tax-related restatement. We calculate the critical point at which the likelihood of tax-related restatement is lowest at Tax_Avoid of .0022, which is approximately the 61st percentile on the distribution of tax avoidance for observations in the matched sample. We rerun the analysis replacing our dependent variable with non-tax-related restatements and find that our measures of tax avoidance are not significant across all model specifications. This suggests that our results of a nonlinear association between tax avoidance and the likelihood of a tax-related restatement are limited to tax-specific restatements.

Next, in Table 3, we explore the two explanations from our hypothesis development that may explain the relationship between tax avoidance and the increase in likelihood of a tax-related restatement, obfuscation, and weak corporate governance. In Panel A, we include proxies for obfuscation, specifically tax accrual quality and analyst forecast accuracy.¹⁵ Obfuscation is low (Low_Obfuscation) when either TAQ (Column 1) is above the sample median or the absolute value of Forecast_Accuracy (Column 2) is below the sample median.¹⁶ We interact Low_Obfuscation with High_Tax_Avoid and Low_Tax_Avoid based on tax avoidance quartiles and deciles (consistent results are untabulated for brevity).¹⁷ We find that high tax avoidance only increases the likelihood of a tax-related restatement when obfuscation is high. Specifically, the coefficient on High_Tax_Avoid, which reflects the influence of high tax avoidance on the likelihood of a restatement for firms with high obfuscation, is positive and significant across both measures of obfuscation (p < .0030). Panel B presents joint significance tests showing that the sum of the coefficients on High_Tax_Avoid and High_Tax_Avoid × Low_Obfuscation is insignificant across both measures (p > .2428), consistent with high avoiding firms with low levels of obfuscation not having an increased likelihood of a tax-related restatement.¹⁸ These results suggest that engaging in obfuscation to reduce detection costs increases the outcome cost of a tax-related restatement.

Table 3.

Tax-Related Restatement Logistic Regression Model With Governance and Obfuscation Analysis.

Panel A: Obfuscation analysis with tax accrual quality (Column 1) and analyst forecast accuracy (Column 2)
Parameter	Column 1: Tax accrual quality				Column 2: Analyst forecast accuracy
Parameter	Estimate	Wald chi-square	Pr > chi-square	Change in conditional probability	Estimate	Wald chi-square	Pr > chi-square	Change in conditional probability
Intercept	−0.3332	0.1024	0.7450		−0.7443	0.0676	0.7970
High_Tax_Avoid	0.8167	8.3521	0.0030		1.2333	11.8336	0.0000
Low_Tax_Avoid	0.6338	3.4596	0.0590		0.7175	4.2025	0.0370
High_Tax_Avoid × Low_Obfuscation	−0.0866	0.0169	0.8990	−0.0188	−0.9124	3.3856	0.0660	−0.1994
Low_Tax_Avoid × Low_Obfuscation	−0.0704	0.0100	0.9190	−0.0153	−0.0871	0.0361	0.8490	−0.0190
Low_Obfuscation (High_TAQ or High_Forecast_Accuracy)	0.3845	0.7744	0.3770		0.2845	0.9409	0.3310
TAQ					−11.0342	1.5625	0.2080
Forecast_Accuracy	−0.0829	0.0484	0.8270
Other Controls	Yes				Yes
Industry	Yes				Yes
Year	Yes				Yes
N	442	(221 tax restatements)			442	(221 tax restatements)
Model Chi-sq/p-value	62.16	0.0982			59.32	0.1072
ROC curve	0.7069				0.7097
Max rescaled Rsq	.1686				.1736
Panel B: Joint significance tests
	Tax accrual quality				Analyst forecast accuracy
Sum of coefficients	Sum	Wald chi-square	Pr > chi-square		Sum	Wald chi-square	Pr > chi-square
High_Tax_Avoid + High_Tax_Avoid × Low_Obfuscation	0.7301	1.36	0.2428		0.3209	0.81	0.3669
Low_Tax_Avoid + Low_Tax_Avoid × Low_Obfuscation	0.5634	0.81	0.3675		0.6305	2.30	0.1293
Panel C: Governance analysis with institutional ownership (Column 1) and No_Dual-CEO (Column 2)
Parameter	Column 1: Tax accrual quality				Column 2: Analyst forecast accuracy
	Estimate	Wald chi-square	Pr > chi-square	Change in conditional probability	Estimate	Wald chi-square	Pr > chi-square	Change in conditional probability
Intercept	−0.2626	0.0625	0.8030		−0.7125	0.3364	0.5630
High_Tax_Avoid	0.6685	2.8224	0.0920		1.5766	17.1396	<0.0001
Low_Tax_Avoid	0.9077	5.0625	0.0250		1.2039	6.7081	0.0100
High_Tax_Avoid × Strong_Govern	0.1494	0.0841	0.7720	0.0327	−1.5631	8.8804	0.0030	−0.3331
Low_Tax_Avoid × Strong_Govern	−0.5536	1.0201	0.3120	−0.1211	−0.8402	1.9321	0.1640	−0.1629
Strong_Govern (High_Inst_Own or No_Dual_CEO)	−0.1039	0.1024	0.7500		0.4648	1.8496	0.1740
Inst_Own					−0.7213	4.7961	0.0290
Dual_CEO	0.1433	0.3844	0.5330
Other control variables	Yes				Yes
Industry	Yes				Yes
Year	Yes				Yes
N	442	(221 tax restatements)			442	(221 tax restatements)
Model chi-square/p-value	58.61	0.1636			65.77	0.0233
ROC curve	0.7011				0.7240
Max rescaled Rsq	.1651				.1943
Panel D: Joint significance tests
Sum of coefficients	Institutional ownership				No dual CEO
	Sum	Wald chi-square	Pr > chi-square		Sum	Wald chi-square	Pr > chi-square
High_Tax_Avoid + High_Tax_Avoid × Strong_Govern	0.8179	6.02	0.0141		0.0135	0.00	0.9480
Low_Tax_Avoid + Low_Tax_Avoid × Strong_Govern	0.3541	0.72	0.3977		0.3637	0.82	0.4069

Note.Table 3 presents the results of a logistic regression modeling the likelihood of incurring a tax-related restatement. p-values are two-tailed. Tax-related restatement firms are matched to a control sample based on one-to-one propensity score matching (caliper = .01) from Table 2. Panel A includes an analysis of the influence of obfuscation on the association between tax avoidance and the likelihood of incurring a tax-related restatement. Column 1 uses tax accrual quality as a proxy for obfuscation, with Low_Obfuscation defined above the sample median of TAQ. Column 2 uses analyst forecast accuracy as a proxy for obfuscation, with Low_Obfuscation defined as the absolute value of Forecast_Accuracy below the sample median. Panel B presents joint significance tests of High (Low) Tax Avoidance and Low Obfuscation. Panel C includes an analysis of the influence of corporate governance on the association between tax avoidance and the likelihood of incurring a tax-related restatement. Column 1 uses institutional ownership as a proxy for corporate governance, with Strong_Govern defined above the sample median for the percentage of the firm owned by institutional investors (High_Inst_Own). Column 2 uses the presence of a CEO who is not also the chairman of the board (No_Dual_CEO) as a proxy for Strong_Govern. Panel D presents joint significance tests of High (Low) Tax Avoidance and Strong Governance. Since the p-values of interaction terms in nonlinear regression models cannot be interpreted without adjustment, Panels A and C present the change in conditional probabilities for the interaction terms. Joint significance tests are not affected by the change in conditional probabilities (Mao, 2014). Variables are defined in Supplemental Appendix B.

In Panel C, we proxy for strong corporate governance (Strong_Govern) with Inst_Own above the sample median in Column 1 and the presence of a CEO who is not also the chairman of the board (No_Dual_CEO) in Column 2. We interact Strong_Govern with High_Tax_Avoid and Low_Tax_Avoid, measured based on tax avoidance quartiles and deciles (consistent results are untabulated for brevity). We find that low tax avoidance increases the likelihood of a tax-related restatement only when governance is weak. Specifically, in Columns 1 and 2 of Panel C the coefficients on Low_Tax_Avoid, which reflects the influence of low tax avoidance on the likelihood of a tax-related restatement in firms with weak governance, are positive and significant across both measures of governance (p < .0220). In contrast, the joint significance test in Panel D shows that firms with strong governance and low tax avoidance are not more likely to incur a tax-related restatement, as evidenced by the insignificant sum of coefficients on Low_Tax_Avoid and Low_Tax_Avoid × Strong_Govern across both measures of governance (p > .3977). It appears that for low-avoiding firms, strength of corporate governance is a channel through which tax avoidance affects the likelihood of incurring a tax-related restatement.

For high tax avoiders, we find that our first proxy for governance, the level of institutional ownership,¹⁹ does not reduce the likelihood of a tax-related restatement, as the coefficient on High_Tax_Avoid in Column 1 of Panel C and the joint significance test of the sum of the coefficients on High_Tax_Avoid and High_Tax_Avoid × Strong_Govern in Panel D are both positive and significant (p < .0920).²⁰ In contrast, we find that that our second proxy for strong governance, the presence of a CEO who is not also the chairman of the board (No_Dual_CEO), does influence the likelihood of a high tax avoider incurring a tax-related restatement. In particular, in Column 2 of Panel C, the coefficient on High_Tax_Avoid, which reflects the impact of high tax avoidance on a tax-related restatement when there is a dual CEO, is positive and significant (p < .0001). However, the sum of the coefficients on High_Tax_Avoid and High_Tax_Avoid × Strong_Govern in Panel D is insignificant (p = .9480). These results suggest that the presence of a dual CEO increases the risk of a tax-related restatement for high avoiders. The differing results using institutional ownership and the presence of a dual CEO as proxies for governance for high avoiders is in line with prior research highlighting that corporate governance is a complex system (i.e., Baber et al., 2012; Misangyi & Acharya, 2014) where a variety of procedures and characteristics are necessary to address multiple issues for which there is not a one-size-fits-all solution. It appears that for high avoiders, both obfuscation and weak corporate governance can be channels through which tax avoidance affects the likelihood of incurring a tax-related restatement.

Tax Avoidance Behavior after a Tax-Related Restatement Announcement

Next, we investigate whether tax avoidance behavior changes after a firm announces a tax-related restatement by utilizing the following ordinary least squares (OLS) regression which employs a difference-in-difference (“DID”) design:

\begin{matrix} Tax__Avoi d_{it} & = α + β_{1} Tax__Restate__Announc e_{it} + β_{2} Post__Announc e_{it} \\ + β_{3} Tax__Restate__Announc e_{it} * Post__Announc e_{it} + β_{4} Tax__M W_{it} \\ + β_{5} Accrual s_{it} + β_{6} NO L_{it} + β_{7} Eq__In c_{it} + β_{8} For__In c_{it} \\ + β_{9} R D_{it} + β_{10} Leverag e_{it} + β_{11} PP E_{it} + β_{12} RO A_{it} + β_{13} CAS H_{it} \\ + β_{14} De p_{it} + β_{15} Inst__Ow n_{it} + β_{16} Industr y_{it} + β_{17} Yea r_{it} + ε \end{matrix}

Model 2

In our DID design, we are interested in a firm’s behavior after the discovery and subsequent announcement of a restatement and thus define Tax_Restate_Announce as an indicator variable set to 1 for the year the firm announces a tax-related restatement. We match restatement to control firms based on tax avoidance prior to the restatement to compare outcomes of firms that have similar tax profiles before the restatement announcement (“DID match”). Tax-related restatement firms (Tax_Restate_Announce = 1) are matched to control firms (Tax_Restate_Announce = 0) based on year of restatement announcement, Fama French 48 industry code, total assets within 50%, and closest industry size-adjusted CETR averaged over the 3-year period ending the year prior to the restatement announcement. Tax_Avoid remains the industry-size adjusted CETR utilized in Model 1.²¹

Post_Announce is an indicator variable equal to one for years after the restatement announcement.²² We examine the time period that includes 2 years prior to the restatement announcement, the announcement year, and 2 years after the restatement announcement, leaving 136 restatement firms with the necessary data to calculate model variables over the time period examined. The choice of the period of examination was selected to minimize the loss of restatement observations.²³ The interaction of Tax_Restate_Announce and Post_Announce (“DID estimator”) captures the change in tax avoidance observed after a tax-related restatement announcement compared to a matched control firm.

To examine our second set of hypotheses, we split the sample into groups based on their level of tax avoidance prior to the restatement announcement. Specifically, firms above (below) the sample median or in the top (bottom) quartile of industry-size adjusted CETR in the year prior to the restatement announcement are classified into our high (low) tax avoidance samples.²⁴ Under Hypothesis 2a, we expect a negative coefficient on the DID estimator in the high tax avoidance sample, suggesting that firms that exhibited high tax avoidance before the restatement announcement decrease their tax avoidance after the restatement announcement. Under Hypothesis 2b, we expect a positive coefficient on the DID estimator in the low tax avoidance sample, suggesting that firms that exhibited low tax avoidance before the restatement announcement increase their tax avoidance after the restatement announcement. We control for a series of factors found by prior research to be associated with tax avoidance behavior. For brevity, variable definitions and references to prior literature are presented in Supplemental Appendix B.

Results for Behavior Following Tax-Related Restatements—Hypotheses 2a and 2b

Univariate results

Table 4, Panel A presents covariate balance in the means of the dependent and independent variables for tax-related restatement and control observations in the year of the restatement announcement. Panel B presents descriptive statistics before the restatement announcement by including the 2 years leading up to the announcement. With few exceptions (Tax_MW, R&D, and Cash), variable means are not statistically different between treatment and control firms in the year of the restatement announcement or the years leading up to the announcement, suggesting our multivariate results are not attributable to mean reversion. In addition, we note an indistinguishable difference between restatement and control firms’ growth in the dependent variable during the 2 years leading up to the restatement announcement (p = .6679), demonstrating the required parallel trends for a difference-in-difference design (Roberts & Whited, 2012).

Table 4.

Tax-Related Restatement Announcement and Matched Sample Covariate Balance Descriptive Statistics.

Panel A: Announcement year tax-related restatement and matched sample descriptive statistics
Tax-related restatement announcement year
	Tax-related restatement			Control			Difference
Variable	n	M	SD	n	M	SD	Difference	t-value	p-value
Tax_Avoid	136	−0.0501	0.2183	136	−0.0545	0.1745	0.0044	0.1800	.8545
Tax_MW	136	0.1103	0.3144	136	0.0143	0.1191	0.0960	3.3700	.0009
Accruals	136	−0.0204	0.1678	136	−0.0059	0.0165	−0.0145	0.6600	.5067
NOL	136	0.5515	0.4992	136	0.4786	0.5013	0.0729	1.2100	.2272
Eq_Inc	136	2.4268	13.1048	136	3.0154	20.4188	−0.5886	0.3000	.7633
For_Inc	136	0.0128	0.0516	136	0.0336	0.0028	−0.0208	0.1700	.8644
R&D	136	0.0296	0.0509	136	0.0456	0.0072	−0.0160	1.8900	.0592
Leverage	136	0.2280	0.2392	136	0.2299	0.0194	−0.0019	0.8700	.3838
PPE	136	0.2762	0.2399	136	0.2645	0.2499	0.0117	0.4000	.6920
ROA	136	−0.0031	0.1620	136	0.2200	0.1874	−0.2231	1.1900	.2350
CASH	136	0.1920	0.1879	136	0.1980	0.2067	−0.0060	0.2500	.7992
Dep	136	0.0459	0.0274	136	0.0455	0.0317	0.0004	0.1100	.9160
Inst_Own	136	0.2078	0.3286	136	0.2143	0.3710	−0.0065	0.1500	.8770
Panel B: Two years prior to tax-related restatement announcement and matched sample descriptive statistics
Prior to tax-related restatement announcement year
	Tax-related restatement			Control			Difference
Variable	n	M	SD	n	M	SD	Difference	t-value	p-value
Tax_Avoid	272	−0.0639	0.2075	272	−0.0589	0.1730	−0.0050	0.3100	.7594
Tax_MW	272	0.1136	0.3179	272	0.0144	0.1193	0.0992	4.8600	<.0001
Accruals	272	0.0063	1.1605	272	0.0195	0.3633	−0.0132	0.1800	.8565
NOL	272	0.5128	0.5008	272	0.4748	0.5003	0.0380	0.8900	.3733
Eq_Inc	272	1.9361	15.4439	272	3.8568	25.9872	−1.9207	1.0500	.2912
For_Inc	272	0.0113	0.0439	272	0.0126	0.0329	−0.0013	0.4000	.6888
R&D	272	0.0316	0.0565	272	0.0557	0.1554	−0.0241	2.4100	.0164
Leverage	272	0.2258	0.2526	272	0.1817	0.2121	0.0441	2.2200	.0266
PPE	272	0.2854	0.2553	272	0.2663	0.2454	0.0191	0.9000	.3708
ROA	272	0.0111	0.2327	272	−0.0071	0.3944	0.0182	0.6600	.5122
CASH	272	0.1953	0.2151	272	0.2359	0.2627	−0.0406	1.9800	.0481
Dep	272	0.0483	0.0304	272	0.0461	0.0344	0.0022	0.8000	.4244
Inst_Own	272	0.2033	0.3527	272	0.1840	0.3142	0.0193	0.6800	.4975

Note.Table 4 presents descriptive statistics for the dependent and independent variables for Model 2 defined in Supplemental Appendix B. The one-to-one matched control sample matches a control firm to each tax-related restatement firm based on industry, year, assets within 50% and closest industry-size adjusted CETR the year prior to the restatement announcement and necessary data over required time span (i.e., 2 years prior to, during, and 2 years after the restatement announcement). Panel A presents the covariate balance in the means of the dependent and independent variables for the year of the tax-related restatement announcement. Panel B presents means for the 2 years leading up to the restatement announcement.

Multivariate results

Table 5 presents the multivariate results of the examination of tax avoidance behavior following a tax-related restatement announcement. Column 1 shows that there is not a general trend in tax avoidance behavior for tax-restatement or control firms in the period after a restatement announcement (insignificant coefficients on Post_Announce and DID estimator Tax _Restate_Announce × Post_Announce). We next examine tax avoidance behavior after a tax-related restatement announcement conditional on tax avoidance prior to the restatement. Columns 2 and 3 present the high tax avoidance sample (defined as above the median and in the top quartile respectively²⁵) which includes both tax-restatement and corresponding matched control observations. The negative coefficients on the DID estimators in Columns 2 and 3 are significant (p < .0894), suggesting that tax-restatement firms exhibiting relatively high tax avoidance prior to the restatement announcement decrease tax avoidance in the period following the restatement announcement, consistent with Hypothesis 2a. The economic effect of the change in tax avoidance for high avoiders following a tax-related restatement announcement is meaningful. The coefficient estimate on the DID estimator in Column 3 suggests that on average high avoiders experience an increase in their size-adjusted CETR of 6.2 percentage points following a tax-related restatement. Given that the mean value of pretax income for our sample is $141.796 million, a difference of 6.2 percentage points is associated with a $8.791 million increase in cash taxes paid.²⁶

Table 5.

Difference-in-Difference Design Modeling Impact of Tax-Related Restatement on Tax Avoidance.

Parameter	Column 1: Full sample			Column 2: High avoid observations median industry-size adjusted CETR			Column 3: High avoidance observations based on quartile industry-size adjusted CETR			Column 4: Low avoid observations median industry-size adjusted CETR			Column 5: Low avoid observations based on quartile industry-size adjusted CETR
Parameter	Estimate	t-value	Pr >\|t\|	Estimate	t-value	Pr >\|t\|	Estimate	t-value	Pr >\|t\|	Estimate	t-value	Pr >\|t\|	Estimate	t-value	Pr >\|t\|
Intercept	0.0088	0.0900	0.9313	0.0062	0.0400	0.9671	0.0334	0.1700	0.8679	−0.1261	−2.4600	0.0146	−0.1492	−1.9400	0.0541
Tax_Restate_Announce	−0.0007	−0.0500	0.9640	0.0179	0.8000	0.4223	0.0323	1.0700	0.2876	−0.0009	−0.0400	0.9651	−0.0175	−0.5700	0.5700
Post_Announce	0.0036	0.2500	0.8052	0.0159	0.8200	0.4155	0.0219	1.0000	0.3201	0.0048	0.2100	0.8324	−0.0182	−0.5400	0.5918
Tax_Restate_Announce × Post_Announce	−0.0268	−1.4000	0.1620	−0.0591	−2.0800	0.0389	−0.0620	−1.7100	0.0894	−0.0030	−0.1000	0.9199	0.0075	0.1700	0.8670
Tax_MW	0.0015	0.0500	0.9568	0.0022	0.0600	0.9509	−0.0260	−0.6400	0.5231	−0.0057	−0.1500	0.8795	−0.0415	−0.8300	0.4072
Accruals	−0.0041	−0.6600	0.5104	−0.0088	−0.1800	0.8535	−0.1138	−1.6600	0.0993	0.0099	1.3500	0.1770	0.0032	0.4200	0.6736
NOL	−0.0672	−4.5200	<.0001	−0.0369	−1.9100	0.0577	−0.0167	−0.5700	0.5682	−0.0406	−2.5200	0.0126	−0.0113	−0.4300	0.6668
Eq_Inc	−0.0001	−0.2500	0.7990	−0.0002	−0.8500	0.3946	−0.0005	−1.5200	0.1313	0.0004	1.4700	0.1431	0.0008	0.5000	0.6183
For_ Inc	0.1645	0.9200	0.3609	−0.0985	−0.4200	0.6783	−0.3544	−1.3600	0.1748	0.2669	1.2700	0.2068	0.2396	0.7900	0.4322
R&D	0.0480	0.6300	0.5314	−0.1506	−1.0700	0.2843	−0.1583	−0.5300	0.5963	0.1197	1.6900	0.0926	0.0882	0.8100	0.4184
Leverage	−0.0716	−2.1300	0.0337	−0.1179	−2.7400	0.0066	−0.1203	−2.0900	0.0379	−0.0139	−0.3800	0.7007	−0.0580	−1.3200	0.1883
PPE	−0.0391	−1.0700	0.2848	−0.0785	−1.4500	0.1481	−0.0645	−0.9100	0.3662	0.0399	1.1400	0.2560	0.0787	1.5000	0.1351
ROA	0.0825	2.2700	0.0238	0.1229	2.3200	0.0213	0.5128	4.3600	<.0001	0.0342	1.2100	0.2259	0.0710	1.9000	0.0600
CASH	−0.0693	−2.2000	0.0285	−0.0310	−0.6500	0.5173	−0.0237	−0.3100	0.7586	−0.0402	−1.1900	0.2342	−0.0366	−0.6400	0.5241
Dep	−0.2431	−1.0900	0.2770	0.1903	0.5800	0.5618	−0.0586	−0.1200	0.9059	−0.3562	−1.5000	0.1362	−0.5044	−1.7000	0.0911
Inst_Own	0.0384	1.9200	0.0555	0.0102	0.4100	0.6809	−0.0225	−0.6700	0.5043	0.0545	2.2800	0.0235	0.0575	1.6600	0.0987
Industry fixed effects?	Yes			Yes			Yes			Yes			Yes
Year fixed effects?	Yes			Yes			Yes			Yes			Yes
Adj. R ²	7.89%			5.57%			9.05%			6.46%			4.96%
N	1,360			700			310			660			350
Restatement observations	136			70			31			66			35

Note.Table 5 presents results of Model 2 examining tax avoidance behavior in the 2 years prior to, during, and 2 years after a tax-related restatement announcement. The difference-in-difference regression uses matched pairs. The one-to-one matched sample matches a control firm to each tax-related restatement firm based on industry, year, assets within 50%, and closest industry-size adjusted CETR the year prior to the restatement announcement and necessary data over required time span (i.e., 2 years prior to, during, and 2 years after the restatement announcement). The announcement year is included in the pre-restatement announcement period. Column 1 presents results of the difference-in-difference regression model for the full sample. Column 2 (3) presents results of the model for high tax avoidance observations where high tax avoidance is defined above the median (top quartile) of the industry-size adjusted CETR based on the population of observations with necessary data. Column 4 (5) presents results of the model for low tax avoidance observations where low tax avoidance is below the median (bottom quartile) of the industry-size adjusted CETR based on the population of observations with necessary data. Variables are defined in Supplemental Appendix B. CETR = cash effective tax rate.

Columns 4 and 5 present the low tax avoidance sample (defined as below the median and in the bottom quartile, respectively) which includes both tax-restatement and corresponding matched control observations. The coefficients on the DID estimators in Columns 4 and 5 are insignificant, thus we do not find evidence consistent with Hypothesis 2b which predicts that improved governance following a restatement event will lead firms with relatively low tax avoidance toward a higher level of tax avoidance.

In an untabulated analysis, we find no evidence of a change in tax avoidance behavior for firms experiencing a non-tax-related restatement announcement. Specifically, in a matched sample of non-tax restatement and control firms, the DID estimator is insignificant across the high and low tax avoidance samples.

In Table 6, we present the results of an analysis exploring the channels by which we argue a tax-related restatement affects future tax avoidance. First, in Panel A, we examine whether IRS scrutiny increases after a tax-related restatement. Fox and Wilson (2020) find an increase in IRS monitoring after restatements. We conduct a similar analysis to confirm that their results hold in our sample with only tax-related restatements and to explore whether the increase in attention is dependent on tax avoidance behavior prior to the restatement. We replace the dependent variable from Model 2 with IRS_Attention and maintain the same set of control variables and sample partitions.²⁷ While we do not find evidence of an increase in IRS attention after a tax-related restatement in Column 1 with our full sample, we do find evidence of an increase in IRS monitoring for firms that were avoiding higher levels of taxes prior to the restatement. Specifically, in Column 2 the coefficient of 3.0673 on the DID estimator is positive and significant (p = .0818). In contrast, the coefficient is not significant for our low avoider samples in Columns 4 and 5. These results provide evidence consistent with our argument that the IRS acts as an external monitor after the announcement of a tax-related restatement and constrains tax avoidance for firms that were avoiding higher levels of tax prior to the restatement announcement.

Table 6.

Difference-in-Difference Design Modeling Impact of Tax-Related Restatement on Channels Impacting Tax Avoidance.

Panel A: Influence of tax-related restatement on IRS attention
Parameter	Column 1: Full sample			Column 2: High avoid observations median industry-size adjusted CETR			Column 3: High avoidance observations based on quartile industry-size adjusted CETR			Column 4: Low avoid observations median industry-size adjusted CETR			Column 5: Low avoid observations based on quartile industry-size adjusted CETR
Parameter	Estimate	t-value	Pr >\|t\|	Estimate	t-value	Pr >\|t\|	Estimate	t-value	Pr >\|t\|	Estimate	t-value	Pr >\|t\|	Estimate	t-value	Pr >\|t\|
Intercept	7.9836	2.7800	0.0059	10.0831	2.1100	0.0360	8.2953	1.4000	0.1634	4.1152	1.0000	0.3182	9.1188	1.7600	0.0812
Tax_Restate_Announce	0.4056	0.4400	0.6630	−1.6577	−1.3700	0.1734	−2.0597	−1.3700	0.1739	3.3366	2.5700	0.0111	1.3214	0.8300	0.4063
Post_Announce	0.5714	0.7600	0.4484	−0.1054	−0.1100	0.9153	−1.0005	−0.7700	0.4433	1.2611	0.9600	0.3369	0.1555	0.0700	0.9404
Tax_Restate_Announce × Post_Announce	1.9294	1.3900	0.1654	3.0673	1.7500	0.0818	2.6746	1.1900	0.2376	0.3348	0.1500	0.8846	2.4222	0.7200	0.4735
Tax_MW	0.4605	0.3400	0.7365	2.9352	1.3100	0.1924	1.1194	0.8100	0.4198	−2.4938	−1.5100	0.1323	−0.3030	−0.1200	0.9060
Accruals	−0.0463	−0.2400	0.8128	−0.8472	−0.3100	0.7603	0.8060	0.2800	0.7786	−0.0748	−0.2700	0.7873	−0.1135	−0.2700	0.7867
NOL	−0.9930	−1.0700	0.2847	0.2876	0.2300	0.8202	−0.1639	−0.1200	0.9009	−1.7353	−1.0700	0.2865	−1.7091	−0.6000	0.5524
Eq_Inc	0.0347	1.9700	0.0506	0.0143	0.9600	0.3369	−0.0217	−1.4100	0.1599	0.0714	2.4000	0.0173	0.0094	0.1200	0.9033
For_ Inc	22.5695	1.9300	0.0553	19.5544	1.2100	0.2281	39.3892	2.0900	0.0384	34.1599	2.0400	0.0431	49.0398	2.1200	0.0363
R&D	1.0292	0.2600	0.7932	5.3193	0.6700	0.5015	16.3724	1.2300	0.2193	1.0848	0.2600	0.7930	−4.9860	−0.5600	0.5781
Leverage	−1.0098	−0.6100	0.5405	0.3293	0.1200	0.9049	4.5785	1.5200	0.1316	−1.3704	−0.7000	0.4853	−3.5218	−1.2600	0.2085
PPE	0.3603	0.1500	0.8775	5.9021	1.4600	0.1462	2.3418	0.4900	0.6222	−0.5541	−0.2100	0.8313	−0.9805	−0.2000	0.8387
ROA	0.8988	0.9000	0.3678	2.5176	1.5500	0.1226	−1.7135	−0.5000	0.6168	−0.0665	−0.0600	0.9545	0.5186	0.1700	0.8648
CASH	−3.4030	−2.0200	0.0443	−3.8514	−1.8300	0.0690	−5.3395	−1.9700	0.0510	−1.9601	−0.7000	0.4837	−4.6019	−1.1400	0.2570
Dep	−4.4335	−0.2500	0.8049	−17.6314	−0.6400	0.5252	−3.1893	−0.1300	0.8971	−4.0146	−0.1600	0.8696	13.3171	0.3200	0.7460
Inst_Own	1.4799	1.0800	0.2807	1.1899	0.7900	0.4314	0.7888	0.3300	0.7411	1.3100	0.5800	0.5627	1.5708	0.6500	0.5187
Industry fixed effects?	Yes			Yes			Yes			Yes			Yes
Year fixed effects?	Yes			Yes			Yes			Yes			Yes
Adj. R ²	12.39%			12.71%			19.28%			10.65%			17.95%
N	1,060			570			310			470			350
Restatement observations	106			57			31			47			35
Panel B: Influence of tax-related restatement on dual CEO/chairman of the board
Parameter	Column 1: Full sample			Column 2: High avoid observations median industry-size adjusted CETR			Column 3: High avoidance observations based on quartile industry-size adjusted CETR			Column 4: Low avoid observations median industry-size adjusted CETR			Column 5: Low avoid observations based on quartile industry-size adjusted CETR
Parameter	Estimate	t-value	Pr >\|t\|	Estimate	t-value	Pr >\|t\|	Estimate	t-value	Pr >\|t\|	Estimate	t-value	Pr >\|t\|	Estimate	t-value	Pr >\|t\|
Intercept	0.3178	1.8400	0.0665	0.0306	0.1200	0.9053	−0.0055	−0.0200	0.9841	0.5726	4.1700	<0.0001	0.6464	4.6900	<0.0001
Tax_Restate_Announce	0.0854	1.5400	0.1257	0.0701	1.0300	0.3062	0.1357	1.7200	0.0876	0.0947	1.3000	0.1966	0.1769	2.1300	0.0343
Post_Announce	0.0136	0.4400	0.6621	−0.0300	−0.7000	0.4825	−0.0128	−0.2400	0.8072	0.0445	0.8100	0.4173	0.0628	0.8700	0.3869
Tax_Restate_Announce × Post_Announce	−0.0134	−0.3500	0.7267	−0.0303	−0.5300	0.5944	−0.1307	−1.7400	0.0838	0.0346	0.5100	0.6086	0.0645	0.7500	0.4570
Tax_MW	−0.0525	−0.8200	0.4107	−0.0192	−0.2100	0.8361	0.0983	0.8600	0.3882	−0.0878	−0.8400	0.3994	−0.1717	−1.1000	0.2713
Accruals	0.0199	1.4300	0.1540	−0.0409	−0.6400	0.5220	0.0270	0.4500	0.6552	0.0259	1.7100	0.0879	−0.0446	−0.3500	0.7264
NOL	−0.0682	−1.5300	0.1262	−0.0427	−0.7800	0.4361	−0.0778	−1.2100	0.2260	−0.0841	−1.5700	0.1177	−0.1055	−1.5200	0.1304
Eq_Inc	0.0002	0.2500	0.7996	−0.0001	−0.1700	0.8684	0.0001	0.1500	0.8848	0.0013	1.2500	0.2138	−0.0008	−0.2900	0.7729
For_ Inc	−0.4743	−1.2000	0.2328	−0.2704	−0.5100	0.6090	0.1620	0.2700	0.7886	−0.8747	−1.6900	0.0915	−0.9193	−1.2900	0.1998
R&D	0.0064	0.0200	0.9811	0.2674	0.7200	0.4731	−0.1046	−0.1300	0.8997	−0.2206	−0.7700	0.4393	0.2781	0.6400	0.5243
Leverage	0.0267	0.2500	0.8053	−0.0102	−0.0700	0.9459	−0.0231	−0.1200	0.9020	0.0809	0.6600	0.5090	0.1262	0.8300	0.4094
PPE	0.1190	1.0100	0.3119	0.1546	0.8900	0.3728	0.0172	0.0800	0.9399	0.1278	1.0000	0.3184	0.0766	0.5400	0.5883
ROA	0.0343	0.5100	0.6134	0.0214	0.2700	0.7850	−0.0684	−0.3300	0.7446	0.0208	0.2100	0.8324	0.1471	1.1700	0.2452
CASH	−0.2537	−2.7900	0.0057	−0.2586	−2.0300	0.0434	−0.3822	−2.3800	0.0182	−0.1691	−1.4400	0.1511	−0.3297	−2.1700	0.0315
Dep	−0.0774	−0.1000	0.9210	0.5037	0.4400	0.6625	1.1351	0.7500	0.4534	−0.5091	−0.6300	0.5265	−0.5242	−0.4900	0.6240
Inst_Own	0.1098	1.5800	0.1141	0.1256	1.4800	0.1409	0.1485	1.5100	0.1324	0.0815	0.8900	0.3741	0.0627	0.6100	0.5449
Industry fixed effects?	Yes			Yes			Yes			Yes			Yes
Year fixed effects?	Yes			Yes			Yes			Yes			Yes
Adj. R ²	15.58%			14.96%			16.34%			17.42%			23.67%
N	1,360			700			310			660			350
Restatement observations	136			70			31			66			35

Note.Table 6 presents the results of an examination of monitoring behavior in the 2 years prior to, during, and 2 years after a tax-related restatement announcement using the matched sample from Table 5. Panel A examines the impact of a tax-related restatement announcement on IRS monitoring proxied by IRS_Attention. Panel B examines the impact of a tax-related restatement announcement on corporate governance proxied by Dual_CEO. Column 1 presents results of the difference-in-difference regression model for the full sample. Column 2 (3) presents results of the model for high tax avoidance observations where high tax avoidance is defined above the median (top quartile) of the industry-size adjusted CETR based on the population of observations with necessary data. Column 4 (5) presents results of the model for low tax avoidance observations where low tax avoidance is below the median (bottom quartile) of the industry-size adjusted CETR based on the population of observations with necessary data. Variables are defined in Supplemental Appendix B. CETR = cash effective tax rate.

Next, in Panel B, we examine whether governance proxied by Dual_CEO changes after a tax-related restatement.²⁸ We replace the dependent variable (Tax_Avoid) from Model 2 with Dual_CEO and maintain the same set of control variables and sample partitions for consistency with our tests of H2. In Column 1 with our full sample, the coefficient on the DID estimator is not significant (p = .0114). However, in Column 3 the coefficient of −.1307 on the DID estimator is significant (p = .0838), suggesting that, on average, firms avoiding the highest level of tax prior to a tax-related restatement separate the roles of management and board leadership after the event. This result is consistent with our argument that improved governance following the event moves the firm toward the expected level of tax avoidance. In contrast, the coefficients on the DID estimators in Columns 4 and 5 with only our low tax avoiders are insignificant. Thus, we conclude that increased governance in our high avoiding restatement firms may explain their decrease in tax avoidance post-restatement, whereas there is not a strong enough increase in governance in our low avoiding restatement firms to have a meaningful effect on tax avoidance.

Conclusion

This study examines the association between tax avoidance and a tax-related restatement. We find that both relatively high and relatively low levels of tax avoidance compared to peer firms are associated with an increased likelihood of incurring a tax-related restatement. We then examine two possible explanations for the relationship between tax avoidance and the likelihood of a tax-related restatement, obfuscation to minimize detection risk and weak corporate governance. For high avoiders, we find evidence supporting the argument that obfuscation is related to the increase in likelihood of incurring a tax-related restatement. We also find some evidence that weak corporate governance is related to the increase in likelihood of incurring a tax-related restatement for high avoiding firms. Specifically, we find that high tax avoidance is only associated with an increased likelihood of a tax-related restatement when the CEO is also the chairman of the board. For low avoiding firms, we find evidence that weak corporate governance proxied by institutional ownership and presence of a CEO who is also chairman of the board is associated with tax avoidance below the expected range and poor financial reporting quality.

We also examine whether the outcome cost of incurring a tax-related restatement influences future tax avoidance behavior. Our results show that firm behavioral response to a tax-related restatement announcement is dependent on the relative level of tax avoidance prior to the restatement. High avoiders decrease tax avoidance after announcing a tax-related restatement, while low avoiders do not change their tax avoidance behavior. We attribute these results to our finding of an increase in both IRS monitoring and strength of corporate governance in high avoiding firms after the restatement announcement, whereas we do not observe such changes for low avoiders.

We recognize that our results are potentially sensitive to research design choices, variable estimation techniques, and sample size. Like Caskey (2019), we recognize that the temporary difference component of CETR has inherently mean reverting characteristics. However, our sample exhibits cross-sectional variation, providing comfort that our results are not attributable to mean reversion (Kim et al., 2019). In addition, we recognize that the small sample size of tax-related restatements due to the necessary data requirements and the resulting loss of observations may affect the generalizability of our results. Finally, while our models include an extensive set of determinants and control variables, we acknowledge that we may not be able to control for managers’ ability or inability to react to all external forces, internal complexities, and unusual events.

Our results provide practical insight for firm managers and shareholders. Managers of high avoidance firms may increase the likelihood of the outcome cost of a tax-related restatement by engaging in obfuscation to escape the detection of tax avoidance. Shareholders of high avoidance firms should be aware that this outcome cost may be mitigated by some, but not all, corporate governance mechanisms. Once a high avoiding firm has incurred the outcome cost of a tax-related restatement, it appears that increases in external monitoring and improvements in corporate governance are effectively moving tax avoidance toward the shareholders’ expected level.

It is important for shareholders of low avoiding firms to understand that the likelihood of incurring a tax-related restatement is greater when corporate governance is weak. In addition, we do not observe changes in either IRS monitoring or corporate governance after the announcement of a tax-related restatement for low avoiding firms. These results indicate that there is an opportunity for the management of low avoiding firms to increase tax avoidance to a level more in line with the shareholders’ expectations.

Supplemental Material

sj-docx-2-jaf-10.1177_0148558X221115482 – Supplemental material for Tax-Related Restatements and Tax Avoidance Behavior

Supplemental material, sj-docx-2-jaf-10.1177_0148558X221115482 for Tax-Related Restatements and Tax Avoidance Behavior by Mollie T. Adams, Kerry K. Inger, Michele D. Meckfessel and John J. Maher in Journal of Accounting, Auditing & Finance

Supplemental Material

sj-xlsx-1-jaf-10.1177_0148558X221115482 – Supplemental material for Tax-Related Restatements and Tax Avoidance Behavior

Supplemental material, sj-xlsx-1-jaf-10.1177_0148558X221115482 for Tax-Related Restatements and Tax Avoidance Behavior by Mollie T. Adams, Kerry K. Inger, Michele D. Meckfessel and John J. Maher in Journal of Accounting, Auditing & Finance

Footnotes

Acknowledgements

The authors would like to thank the anonymous reviewers and the editorial staff of the Journal of Accounting, Auditing and Finance for their helpful comments.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Michele D. Meckfessel

Supplemental Material

Supplemental material for this article is available online.

Notes

Author Biographies

Mollie T. Adams, PhD is an Assistant Professor of Accounting at Missouri State University. Her research and teaching specialties are financial reporting and taxation.

Kerry Inger is an Associate Professor of Accounting in the Harbert College of Business at Auburn University. Her research interests center around taxation and sustainability issues.

Michele D. Meckfessel, PhD, CPA is an Associate Professor of Accounting at University of Missouri – St. Louis. Her research interests center around taxation and financial reporting.

John J. Maher holds the endowed Tom Wells/Kathy Dargo Professorship in the Department of Accounting and Information Systems at Virginia Tech. He specializes in teaching and research on capital markets and accounting information.

References

Adams

R. B.

Santos

J. A.

(2006). Identifying the effect of managerial control on firm performance. Journal of Accounting & Economics, 41(1/2), 55–85.

Norton

E. C.

(2003). Interaction terms in logit and probit models. Economics Letters, 80(1), 123–129.

Allingham

Sandmo

(1972). Income tax evasion: A theoretical analysis. Journal of Public Economics, 1(3/4), 323–338.

Amel-Zadeh

Zhang

(2015). The economic consequences of financial restatements: Evidence from the market for corporate control. The Accounting Review, 90(1), 1–29.

Armstrong

Blouin

Jagolinzer

Larcker

(2015). Corporate governance, incentives, and tax avoidance. Journal of Accounting and Economics, 60(1), 1–17.

Austin

Wilson

(2017). An examination of reputational costs and tax avoidance: Evidence from firms with valuable consumer brands. The Journal of the American Taxation Association, 39(1), 67–93.

Baber

W. R.

Liang

Zhu

(2012). Associations between internal and external corporate governance characteristics: Implications for investigating financial accounting restatements. Accounting Horizons, 26(2), 219–237.

Balakrishnan

Blouin

Guay

(2019). Tax aggressiveness and corporate transparency. The Accounting Review, 94(1), 45–69.

Bauer

(2016). Tax avoidance and the implications of weak internal controls. Contemporary Accounting Research, 33(2), 449–486.

10.

Blaylock

(2016). Is tax avoidance associated with economically significant rent extraction among US firms? Contemporary Accounting Research, 33(3), 1013–1043.

11.

Bozanic

Hoopes

J. L.

Thornock

J. R.

Williams

B. M.

(2017). IRS attention. Journal of Accounting Research, 55(1), 79–114.

12.

Brickley

J. A.

Coles

J. L.

Jarrell

(1997). Leadership structure: Separating the CEO and Chairman of the Board. Journal of Corporate Finance, 3(3), 189–220.

13.

Carcello

Hermanson

(2011). Corporate governance research in accounting and auditing: Insights, practice implications and future research directions. Auditing: A Journal of Practice & Theory, 30(3), 1–31.

14.

Caskey

(2019). Discussion of “How quickly do firms adjust to optimal levels of tax avoidance?” Contemporary Accounting Research, 36(3), 1855–1860.

15.

Chen

Yang

Zhang

Zhou

(2020). The moderating role of internal control in tax avoidance: Evidence from a COSO-based internal control index in China. Journal of the American Taxation Association, 42(1), 23–55.

16.

Choudhary

Koester

Shevlin

(2016). Measuring income tax accrual quality. Review of Accounting Studies, 21(1), 89–139.

17.

Chyz

Gaertner

(2018). Can paying “too much” or “too little” tax contribute to forced CEO turnover? The Accounting Review, 93(1), 102–130.

18.

Cohen

Krishnamoorthy

Wright

(2004). The corporate governance mosaic and financial reporting quality. Journal of Accounting Literature, 23(1), 87–152.

19.

Dechow

Larson

Sloan

(2011). Predicting material accounting misstatements. Contemporary Accounting Research, 28(1), 17–82.

20.

DeFond

Erkens

D. H.

Zhang

(2017). Do client characteristics really drive the Big N audit quality effect? New evidence from propensity score matching. Management Science, 63(11), 3628–3649.

21.

Deloitte

L. L. P.

(2008). Material weaknesses and restatements: Why is tax still in the hot seat?Deloitte Development LLC.

22.

Desai

Dharmapala

(2006). Corporate tax avoidance and high powered incentives. Journal of Financial Economics, 79(1), 145–179.

23.

Desai

Dharmapala

(2009). Corporate tax avoidance and firm value. The Review of Economics and Statistics, 91(3), 537–546.

24.

Desai

Dyck

Zingales

(2007). Theft and taxes. Journal of Financial Economics, 84(3), 591–623.

25.

Desai

Hogan

Wilkins

(2006). The reputational penalty for aggressive accounting: Earnings restatements and management turnover. The Accounting Review, 81(1), 83–112.

26.

Drake

Lusch

Stekelberg

(2019). Does tax risk affect investor valuation of tax avoidance? Journal of Accounting, Auditing & Finance, 34(1), 151–176.

27.

Dyreng

Hanlon

Maydew

(2008). Long-run corporate tax avoidance. The Accounting Review, 83(1), 61–82.

28.

Dyreng

Hoopes

Wilde

(2016). Public pressure and corporate tax behavior. Journal of Accounting Research, 54(1), 147–185.

29.

Ettredge

Scholz

Smith

Sun

(2010). How do restatements begin? Evidence of earnings management preceding restated financial reports. Journal of Business Finance & Accounting, 37(3–4), 332–355.

30.

Farber

D. B.

(2005). Restoring trust after fraud: Does corporate governance matter? The Accounting Review, 80(2), 539–561.

31.

Fox

Wilson

(2020). Double trouble: An analysis of IRS attention and financial reporting. SSRN. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3317839

32.

Francis

Michas

(2013). Office size of big 4 auditors and client restatements. Contemporary Accounting Research, 30(4), 1626–1661.

33.

Frank

Lynch

Rego

Zhao

(2018). Are corporate risk taking practices indicative of aggressive reporting practices? Journal of the American Taxation Association, 40(1), 31–55.

34.

Graham

Hanlon

Shevlin

Shroff

(2014). Incentives for tax planning and avoidance: Evidence from the field. The Accounting Review, 89(1), 991–1023.

35.

Guo

Huang

Zhang

Zhou

(2016). The effect of employee treatment policies on internal control weaknesses and financial restatements. The Accounting Review, 91(4), 1167–1194.

36.

Halperin

Lai

(2015). The relation between auditor-provided tax service fees and audit fees after the Sarbanes-Oxley Act: From the perspective of cross-selling of services. Journal of Accounting, Auditing & Finance, 30(3), 341–372.

37.

Hanlon

Hoopes

Shroff

(2014). The effect of tax authority monitoring and enforcement on financial reporting quality. Journal of the American Taxation Association, 36(2), 137–170.

38.

Hanlon

Slemrod

(2009). What does tax aggressiveness signal? Evidence from stock price reactions to news about tax shelter involvement. Journal of Public Economics, 93(1–2), 126–141.

39.

Hoopes

Mescall

Pittman

(2012). Do IRS audits deter corporate tax avoidance? The Accounting Review, 87(5), 1603–1639.

40.

Hsu

Moore

Neubaum

(2018). Tax avoidance, financial experts on the audit committee, and business strategy. Journal of Business Finance & Accounting, 45(9–10), 1293–1321.

41.

Inger

Meckfessel

Zhou

Fan

(2018). An examination of the impact of tax avoidance on the readability of tax footnotes. Journal of the American Taxation Association, 40(1), 1–29.

42.

Jimenez-Angueira

(2018). The effect of the interplay of corporate governance and external monitoring regimes and firms’ tax avoidance. Advances in Accounting, 41, 7–24.

43.

Khurana

Moser

(2013). Institutional shareholders’ investment horizons and tax avoidance. Journal of the American Taxation Association, 35(1), 111–134.

44.

Kim

McGuire

Savoy

Wilson

(2019). How quickly do firms adjust to optimal levels of tax avoidance? Contemporary Accounting Research, 36(3), 1824–1860.

45.

Kubick

Lynch

Mayberry

Omer

(2016). The effects of regulatory scrutiny on tax avoidance: An examination of SEC comment letters. The Accounting Review, 91(6), 1751–1780.

46.

Lang

Lins

Maffett

(2012). Transparency, liquidity, and valuation: International evidence on when transparency matters most. Journal of Accounting Research, 50, 729–774.

47.

Omer

Sun

(2018, May 21–23). How does tax avoidance affect transparency? [Conference session]. ABFER 6TH Annual Conference, Research Collection School of Accountancy, Singapore.

48.

Mande

Son

(2013). Do financial restatements lead to auditor changes? Auditing: A Journal of Practice and Theory, 32(2), 119–145.

49.

Mao

(2014). Testing for joint significance in nonstationary binary choice model. Economic Letters, 122(2), 311–313.

50.

Minnick

Noga

(2010). Do corporate governance characteristics influence tax management? Journal of Corporate Finance, 16(5), 703–718.

51.

Misangyi

Acharya

(2014). Substitutes or complements? A configurational examination of corporate governance mechanisms. Academy of Management Journal, 57(6), 1681–1705.

52.

Rechner

Dalton

(1991). CEO duality and organizational performance: A longitudinal analysis. Strategic Management Journal, 12(2), 155–160.

53.

Roberts

Whited

(2012, October 5). Endogeneity in empirical corporate finance (Simon School Working Paper No. FR 11-29). https://ssrn.com/abstract=1748604 or https://doi.org/10.2139/ssrn.1748604

54.

Scholes

Wolfson

Erickson

Hanlon

Maydew

Shevlin

(2015). Taxes and business strategy (5th ed.). Prentice Hall.

55.

Seetharaman

Wang

Sun

(2011). Tax-related financial statements restatements and auditor-provided tax services. Journal of Auditing, Accounting and Finance, 26(4), 677–698.

56.

Seidman

Stomberg

(2017). Equity compensation and tax avoidance: Disentangling managerial incentives from tax benefits and reexamining the effect of shareholder rights. Journal of the American Taxation Association, 39(2), 21–41.

57.

Srinivasan

(2005). Consequences of financial reporting failure for outside directors: Evidence from accounting restatements and audit committee members. Journal of Accounting Research, 43(2), 291–334.

58.

Wang

T. S.

Lin

Y. M.

Chao

C. F.

(2013). Board independence, executive compensation and restatement. Applied Financial Economics, 23(11), 963–975.

59.

Wilde

Wilson

(2018). Perspectives on corporate tax planning: Observations from the past decade. Journal of the American Taxation Association, 40(2), 63–81.

60.

Wilson

(2009). An examination of corporate tax shelter participants. The Accounting Review, 84(3), 969–999.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.02 MB

0.08 MB