The effect of conservative financial reporting and tax aggressiveness on the market valuation of unrecognized tax benefits

Abstract

Purpose

Prior studies document a positive relation between stock prices and tax-related contingent liability, unrecognized tax benefits (UTBs) and interpret the finding as evidence that investors reward tax aggressiveness. The purpose of this paper is to explore the nature of this puzzle finding by considering a link between UTBs and financial reporting strategy and propose that financial reporting conservatism may explain the positive association between UTBs and stock prices.

Design/methodology/approach

To estimate the incremental valuation weights on UTBs, the authors employ the Ohlson (1995) valuation model and regress stock prices on UTBs and its interactions with the proxies for financial reporting conservatism and tax aggressiveness. Further, the authors adopt a UTB estimation model to decompose its balance into the predicted and unpredicted components.

Findings

The authors find that the reporting conservatism has a positive effect on the market valuation of UTBs. The authors also find some evidence that tax aggressiveness increases the valuation weight of UTBs. When UTBs are decomposed into predicted and unpredicted components, the authors find that the effect of financial reporting conservatism is more pronounced for the market valuation of predicted UTBs. Collectively, the evidence suggests that conservative financial reporting is a major driver of the positive valuation of UTBs and that tax aggressiveness plays a less significant role in investors' valuation decisions.

Originality/value

While prior studies focus on how UTBs are associated with stock prices, this paper is the first attempt to explain why UTBs are positively valued by investors.

Citation

Jiménez-Angueira, C.E., Nwaeze, E. and Park, S.-J. (2021), "The effect of conservative financial reporting and tax aggressiveness on the market valuation of unrecognized tax benefits", Asian Review of Accounting, Vol. 29 No. 2, pp. 150-172. https://doi.org/10.1108/ARA-07-2020-0111

1. Introduction

In the United States, the Financial Accounting Standards Board's [FASB] Accounting Standard Codification (ASC) 740 governs the accounting for income taxes. In 2006, FASB pronounced Financial Interpretation No. 48, Accounting for Uncertainty in Income Taxes (FIN 48, currently codified as ASC 740–10), establishing new reporting requirements for unrecognized tax benefits (UTBs). The reporting requirements under US GAAP are to improve the reliability and comparability of the provisioning for uncertain tax positions across reporting entities and time. Given that International Financial Reporting Standards (IFRS) has provided no direct guidance on accounting for uncertainty in income taxes, empirical evidence on the value relevance of tax information in financial statements under FIN 48 would be useful to standard setters around the world.

Conceptually, UTBs are a tax-related contingent liability that exposes the reporting entity to potential future economic sacrifices. The extant literature establishes that the contingent liability is negatively associated with firm value (e.g. Barth and McNichols, 1994; Cohen et al., 2011). Nonetheless, several studies surprisingly find a positive association between UTBs and firms' stock prices (e.g. Koester, 2012; Koester et al., 2015; Brushwood et al., 2018). Varying explanations have been advanced for such findings. Koester (2012), for instance, suggests that a positive valuation of UTBs is consistent with UTBs representing past and present tax avoidance, a signal for future tax avoidance, and/or investors valuing managers' tax planning ability. Hanlon and Heitzman (2010) suggest that UTBs may represent either aggressive tax planning or conservative financial reporting of the benefits of a tax position. Considering that accounting for income taxes including UTBs are at the intersection of a firm's financial reporting practices and aggressive tax avoidance activities (Graham et al., 2012; Koh and Lee, 2015), we explore the incremental effects of both tax aggressiveness and financial reporting conservatism on the valuation implications of the UTBs. In particular, we explore the possibility that differences in a firm's tax aggressiveness and financial reporting conservatism practices moderate the valuation of UTBs.

In theory, the market valuation of the UTB liability may depend on investors' inferences about firms' future tax-related cash benefits from the uncertain (aggressive) tax positions. If investors view UTBs as a delayed realization of benefits from aggressive tax positions, then they would incorporate the expected future release from the reserve into stock prices. On the other hand, a tax aggressive firm with UTBs may face potential tax and disclosure costs. A firm taking aggressive tax positions may give away information about its tax planning strategies and other related proprietary information in their FIN 48 disclosures and, in that sense, suffer an economic loss that could result in the lower valuation of the UTBs. For example, Dyreng et al. (2019) find that firms working to obtain lower cash effective tax rates must address the tradeoff of financial reporting costs of disclosing these activities.

We examine the effect of financial reporting conservatism and tax aggressiveness on the valuation of UTBs by following the Ohlson (1995) framework used in Koester (2012) and Koester et al. (2015). We expect the valuation of UTBs is moderated by cross-sectional differences in tax aggressiveness and financial reporting conservatism. To test our conjecture, we regress firms' stock price per share on UTBs, net income, assets and liabilities excluding UTBs while including proxies for tax aggressiveness and financial reporting conservatism and their interaction with UTBs. We use the industry-adjusted cash effective tax rate to capture the magnitude of a firm's tax aggressiveness exceeding the industry benchmark of tax planning activities. As our proxy for financial reporting conservatism, we employ the ratio of the market value to its book value of equity as a measure of balance sheet conservatism (Beaver and Ryan, 2000) as well as the measure of unconditional conservatism developed by Easton and Pae (2004). In our analysis, we first confirm the positive association between UTBs and stock prices documented by Koester (2012) and Koester et al. (2015). Our test of the interaction of conservatism and UTB suggests that the positive valuation of UTBs is explained by firms' conservatism profile. Contrary to our expectation, we find some evidence suggesting that a firm's tax aggressiveness may also escalate the positive valuation of UTBs. The results are robust to several sensitivity checks.

As a supplementary analysis, we decompose UTBs into their predicted and unpredicted components based on a firm's degree and scope of business activities. This analysis is motivated by the reasoning that normal (i.e. predicted) levels of UTBs based on more conservative accounting would have less risk in that they reflect possible tax liabilities more exhaustively while deferring much of the tax positions that have future tax benefits. In addition, we posit that the abnormal (i.e. unpredicted) levels of UTBs may be regarded as a trigger for the regulatory scrutiny, in turn, resulting in real costs for the firm in the form of additional tax payments.

Our results indicate that the positive valuation of UTBs increasing in financial reporting conservatism is concentrated on the predicted component of UTBs. This empirical pattern suggests that investors appear to attach more positive weight to the economic attributes of the predicted component of UTBs than on the attributes of the less predictable component as a firm's financial reporting conservatism increases. By contrast, we find that the positive valuation of UTBs increasing in tax aggressiveness is concentrated on the unpredicted component of UTBs.

Altogether, our test results indicate that the positive valuation of UTBs is linked to firms' financial reporting conservatism, and that tax aggressiveness has, if any, a positive effect on investors' valuation of UTBs. These results suggest that financial reporting conservatism has a greater weight on the positive valuation of UTBs and complement Koester et al.'s (2015) by providing additional evidence that firms' financial reporting strategies have a strong effect on the valuation implications of UTBs. Given our results indicate tax aggressiveness has a less prominent effect on the valuation of UTBs, researchers need to be cautious about characterizing UTBs as a manifestation of firms' tax avoidance/aggressive practices only. Furthermore, our results suggest that the positive valuation of UTBs is more pronounced for the expected UTBs (derived as the predicted value of UTB based on several economic determinants) as a firm's financial reporting conservatism increases.

We contribute to the understanding of the market valuation of UTBs by analyzing the extent that the reported positive association between UTBs and a firm's stock price is affected by the level of a firm's conservative financial reporting and aggressive tax reporting. Our results suggest that cross-sectional differences in firms' conservative financial reporting practices and tax aggressiveness are associated with the positive valuation of UTBs. These findings are important because they add to the stream of the literature that treats UTBs in the context of a firm's broader financial reporting disclosures (Lisowsky et al., 2013; Koester et al., 2015; Nesbitt, 2016; Robinson et al., 2016). Besides, our study informs the FASB's debate about the usefulness of FIN 48 for investors, addressing a part of the scope of FIN 48's Post-Implementation Review (PIR; Blouin and Robinson, 2014).

In the next section, we discuss the background for the development of our hypothesis. The third section describes our research method. The fourth section presents the data selection and descriptive statistics. We present our empirical analyses in the fifth section, and our conclusions and limitations in the sixth section.

2. Background and hypotheses development

2.1 FIN 48 and the valuation of UTBs

The main objective of FIN 48 is to standardize the measurement and disclosure of tax reserves among public companies for the financial statements covering the periods after December 15, 2006 (FASB, 2006). The interpretation establishes a two-step process to measure and disclose uncertain tax benefits in the firms' financial reports. The first step, the recognition step, requires that for a tax position to be recognized in the financial statement, it must possess a more likely than not probability of being sustained upon taxing authority examination based only on its technical merits, regardless of detection risk and any other factors. A tax position that fails the recognition step must be fully accrued in the financial statements as an uncertain tax liability. For tax positions that pass the recognition test, the measurement step (the second step) requires firms to determine and recognize the amount of benefit that has a greater than fifty percent probability of realization using the cumulative probability distribution of potential outcomes for a given position. Any remaining uncertain tax benefit identified in the measurement process is recorded as an uncertain tax liability.

Arguably, the uncertain tax liability resulting from applying the two-step process jointly reflects a firm's tax aggressiveness and its financial reporting policy (Hanlon and Heitzman, 2010), which indicates that both aspects are important when evaluating the relevance of UTB disclosure under FIN 48. Therefore, the value that stockholders attach to UTBs may depend not only on inferences regarding a firm's tax aggressiveness but also on the inferences on firms' financial reporting conservatism since the reported UTB for a certain level of tax aggressiveness vary depending on the firms' financial reporting policies.

To date, the valuation of UTBs remains an open area of inquiry. In their PIR of FIN 48, Blouin and Robinson (2014) suggest that the positive association between UTBs and firm value can be explained either by the notion that UTBs are an indicator of value-enhancing tax avoidance activities or that UTBs reflect conservative financial reporting that overstates the reserve for uncertain tax positions and understates earnings. However, early research on FIN 48 seems to interpret the positive association of UTBs as a consequence of a firm's tax avoidance activities. Frischmann et al. (2008) show that the initial disclosure of UTBs in 2007 leads to positive abnormal stock returns when the reversal of the reported UTBs would affect effective tax rates. Koester (2012) was the first to document a positive association between UTBs and stock prices and argued such association could be explained as current UTBs being considered by investors as predictors of uncertain tax avoidance activities to reduce tax burdens.

Robinson and Schmidt (2013) find that UTBs are positively associated with abnormal stock returns. However, they also find that highly detailed disclosure on UTBs is perceived negatively by the market, suggesting that, at the margin, too much disclosure pertaining to uncertain tax benefits is perceived by investors as value destructing, consistent with the theory of disclosure and proprietary costs. More recently, Koester et al. (2015) document that firms' tax-related internal control material weakness disclosures moderate the positive association between UTBs and stock prices. Notably, both Robinson and Schmidt (2013) and Koester et al. (2015) suggest potential factors that modify the association between UTBs and stock prices. In particular, both studies raise the prospect that the reported positive association between UTBs and firm value is affected by cross-sectional differences in other aspects of firms' financial reporting practices. However, the separate effects of tax aggressiveness and financial reporting conservatism on the market valuation of firms' UTBs are still to be fully explored.

2.2 Accounting conservatism and the valuation of UTBs

Accounting conservatism involves the use of higher standards of verifiability for recognizing gain than for recognizing losses (Basu, 1997; Watts, 2003), thereby affecting the information content of accounting information (Smith et al., 2011). Application of the principle gives rise to more timely and exhaustive recognition of losses, with the consequence that negative (positive) earnings news is more (less) fully reflected in earnings. From a valuation perspective, accounting conservatism understates positive earnings news, which in equilibrium, will lead to a more positive market reaction per unit of reported earnings. This arises because price changes that more fully incorporate both negative and positive earnings news will be more positive relative to earnings that only captures negative news more fully (e.g. Beaver and Ryan, 1983). Under this view, the application of accounting conservatism portends a valuation premium on UTBs. For instance, for firms with greater accounting conservatism, UTBs would more exhaustively reflect the uncertain tax liability (e.g. the firm may apply higher thresholds in classifying tax benefits as realizable) [1]. In such contexts, the firm will overstate the unrecognized tax liability, raising the prospect that investors will adjust the UTBs more positively because of the perceived overstatement (understatement) of the potential tax liability (expected tax benefits) [2]. Hanlon and Heitzman (2010) also highlight the importance of context in the valuation of tax information and point out that UTBs are jointly determined by a firm's tax avoidance activities and financial reporting incentives. As a result, investors may condition their interpretation of UTBs on the firm's tax aggressiveness and/or degree of the firm's accounting conservatism.

From a contracting perspective, accounting conservatism imposes greater risk on managers with respect to earnings-based performance evaluation. In particular, favoring speedier and more complete recognition of negative earnings news implies that reported earnings will understate managers' stewardship and impose costs on the managers equal to the incentive value of the unrecognized good news. This creates an incentive for the managers to pad reported earnings to mitigate such costs (e.g. Fan and Zhang, 2012). However, Bertomeu et al. (2017) argue that firms select a level of conservatism and provide appropriate effort-incentives to achieve the desired level of accounting quality. In their setting, greater conservatism yields higher accounting quality. In the context of our study, their analysis suggests that conservatism in the measurement of UTBs could affect investors' assessment of the firm's potential future tax benefits and cash flow implications.

In general, the value of UTBs can be positive when a firm demonstrates a higher level of financial reporting conservatism (Penman and Zhang, 2002; LaFond and Watts, 2008; Kim et al., 2013) to the extent that investors unravel the excess reserves that will not result in tax-related cash outflows. Since the effect of accounting conservatism on enhancing the credibility of a firm's financial reports can complement the role of UTBs in mitigating uncertainties regarding future tax-related cash outlays, the valuation effect of UTBs would be more positive for more conservative firms. The preceding discussion is summarized with the following hypothesis:

H1.

Ceteris paribus, UTBs are valued more positively for firms with higher financial reporting conservatism.

To the extent that UTBs reflect the magnitude of a firm's tax aggressiveness, investors may perceive the disclosed UTB amount as a trigger for a future tax audit or another regulatory scrutiny. In such a context, larger UTBs may act as a signal of higher future tax payments and receive a negative valuation. Lisowsky et al. (2013), for instance, find that the reported UTBs are positively associated with a firm's use of tax shelters, which are often referred to as the most aggressive tax position available. Robinson and Schmidt (2013) find that the incremental effect of disclosure quality on the positive pricing of UTBs is negative, especially for a tax avoider firm because of the disclosure of proprietary information. In that sense, if market participants associate UTBs with an increased likelihood of tax audits and probable future cash outlays, then UTBs may receive a negative valuation. As such, our second hypothesis is summarized as follows:

H2.

Ceteris paribus, UTBs are valued less positively for firms with higher tax aggressiveness.

3. Method

3.1 Measures of financial reporting conservatism

Financial reporting conservatism has been manifested in the accounting literature as two distinct constructs: unconditional (or news independent) and conditional (or news dependent) conservatism (Beaver and Ryan, 2005). Unconditional conservatism represents the accounting process related to the recognition of assets and liabilities (or understatement of the book value of net assets) such as the immediate expensing of the cost of internally developed intangibles or the maintenance of historical costs of the assets in-use. By contrast, conditional conservatism focuses on more timely recognition of losses than gains summarized into the earnings (or differential timeliness of earnings). While both unconditional and conditional conservatism measures are widely used in the accounting literature, Givoly et al. (2007) report that conditional conservatism measures can be negatively associated with the measures of unconditional conservatism and caution that improper use of conservatism measures may lead to incorrect inferences.

In our empirical tests, we choose the ratio of market to book value of equity (CON-MB) as the primary measure of financial reporting conservatism based on two reasons. First, from a theoretical point of view, our focus on the market value of UTB as a component of net book value requires us to use a measure of unconditional conservatism, reflecting a firm's accumulated balance sheet conservatism. Second, Ashton and Wang (2013) show empirically that financial reporting conservatism is more clearly reflected in the linear information dynamics of the Ohlson model when the ratio of market to book value is used compared to the conditional conservatism measure such as Khan and Watt's (2009) C-SCORE.

As an alternative measure of conservatism, we use an unconditional conservatism measure (CON-EP) developed by Easton and Pae (2004), following Lisowsky et al. (2013) and Koester et al. (2015). Easton and Pae (2004) argue that the ratio of the market value of net operating assets to the book value of net operating assets can be a better a priori indicator of a firm's financial reporting conservatism than the ratio between the market value of common equity to its book value [3]. Our use of unconditional conservatism measures responds to the warnings of several studies, indicating econometric problems in estimating conditional conservatism metrics (Ball et al., 2013; Patatoukas and Thomas, 2011, 2016).

3.2 The measure of tax aggressiveness

Hanlon and Heitzman (2010) define tax aggressiveness as a more extreme version of tax avoidance that reduces explicit taxes and argue that various proxies of tax aggressiveness may or may not be appropriate for a given research question. Given that our research question revolves around the financial reporting of aggressive tax positions, our empirical proxies for tax aggressiveness should reflect the interrelation between financial reporting and tax reporting. Considering that the nonconformity between financial accounting and tax rules allows firms to simultaneously manage book income upward and taxable income downward (Plesko, 2007), we employ a nonconforming tax avoidance measure exploiting book-tax differences to identify a firm's tax aggressiveness.

While a large body of prior tax avoidance studies uses one or multiple forms of the effective tax rate (ETR), reflecting the outcome of a firm's aggressive tax planning (e.g. Rego, 2003; Dyreng et al., 2008; Rego and Wilson, 2012; Wang 2019), ETR measures span a continuum of tax planning strategies including benign tax-saving activities such as municipal bond investments, not necessarily representing tax aggressiveness. Therefore, we adopt the industry-adjusted cash effective tax rate (CashETR-IND) as our primary measure of tax aggressiveness [4]. To compute this measure, we first divide the sum of total cash taxes paid over the three years (from year t-2 to year t) by the sum of pretax income over the same period. Then, we subtract a firm's three-year average cash ETR from the mean of three-year average cash ETRs of other firms in the same industry. To the extent that the industry norm of tax planning activities is reflected in the industry mean of cash ETR, our measure may capture more aggressive tax avoidance activities. Especially, a firm with a cash ETR lower than the industry average will show a positive value of CashETR-IND, suggesting that our measure is increasing in tax aggressiveness.

Notably, one limitation of using the industry-adjusted cash ETR as a measure of tax aggressiveness is the fact that cash ETRs cannot be calculated for loss firms, creating potential sample selection bias. Indeed, Henry and Sansing (2018) highlight that loss firms are often excluded from prior studies using ETR measures and propose an alternative measure of tax avoidance that can be calculated for all firm-years regardless of reporting negative tax expenses or pre-tax book losses. We adopt Henry and Sansing's measure of tax avoidance as an alternative measure of tax aggressiveness for our sensitivity check. Considering that annual tax expense can be mismatched with the actual tax burden for the year and the temporary book-tax difference would take time to be reversed, the annual measure of tax avoidance may overstate or understate the actual magnitude of a firm's tax aggressiveness. To mitigate this measurement issue, we use the three-year average of Henry and Sansing's (2018) tax avoidance measure for our empirical tests.

3.3 Testing the effect of conservatism and tax aggressiveness on the valuation of UTBs

Consistent with Koester (2012) and Koester et al. (2015), we use a modified version of the Ohlson (1995) valuation model to examine whether the market valuation of UTBs is conditional on tax aggressiveness and/or accounting conservatism. Barth and Clinch (2009) comment that the differences of firm size either across firms or over time may bias the statistical inferences from the Ohlson model. Therefore, we follow the recommendation of Barth and Clinch (2009) and use the number of common shares as the deflator of the Ohlson model. We use ordinary least squares (OLS) to estimate the following equation:

(1)PRICEi,t=β0+β1UTBi,t+β2CONi,t+β3UTBi,t×CONi,t+β4TAXi,t+β5UTBi,t×TAXi,t+β6ASSETSi,t+β7DTAi,t+β8LIABi,t+β9DTLi,t+β10EARNi,t+β11LOSSi,t+β12EARNi,t×LOSSi,t+εi,t

where PRICE is the common stock price one day after the Form 10-K filing date. UTB is defined as the year-end balance of uncertain tax benefits scaled by the number of common shares outstanding one day after the 10-K filing date. When UTB information is missing, we replaced the missing values with zeros [5]. CON is estimated as the ratio of market to book value of equity (CON-MB) and following Easton and Pae (2004, CON-EP). TAX is the measure of tax aggressiveness computed as the industry-adjusted cash ETR over a three-year period. Both CON and TAX are transformed into decile ranks by year to reduce the undue influence of extreme values and measurement errors [6]. The model includes a decomposition of the book value of equity where assets and liabilities are adjusted for deferred tax assets and liabilities as well as the balance of UTB, following the method used in Koester et al. (2015) and prior research (e.g. Barth and McNichols, 1994; Ayers, 1998). EARN is defined as income before extraordinary items scaled by the number of common shares outstanding one day after the Form 10-K filing date. We also include the indicator LOSS and its interaction with EARN to allow the coefficient on EARN to vary between profit and loss values; LOSS is defined as one when EARN is negative and zero otherwise. We include year and industry (two-digit SIC code) fixed effects in the estimation of equation (1) to control for time and industry-specific variation on the valuation model. We use robust standard errors clustered at the firm level in all of our estimations to control for serial correlation and heteroscedasticity. Detailed definitions of the variables are in Appendix.

The model allows the association between UTB and firm value to vary across the level of a firm's financial reporting conservatism and tax aggressiveness. A downside of this approach is that interpreting interactions of UTB with financial reporting conservatism and tax aggressiveness as continuous variables becomes more challenging (Jaccard and Turrisi, 2003) [7]. To alleviate this concern, we adopt several approaches. First, we convert CON and TAX into dummy variables (CON_DUM and TAX_DUM) for ease of interpreting the coefficients of the interactions between UTB and CON (UTB×CON_DUM) and TAX (UTB×TAX_DUM). However, several prior studies warn that dichotomizing continuous variables into dummy variables may induce the loss of power and serious biases depending on the cutoff points used for dichotomizing variables (Irwin and McClelland 2003; Royston et al., 2006). Therefore, as our second approach, we use CON and TAX as continuous variables after splitting our sample into two subsamples having the values of conservatism and tax aggressiveness higher than or lower than their median values, respectively. Subsequently, we estimate equation (3) using the full sample along with CON and TAX as continuous variables [8]. Based on equation (1), we predict that the coefficient of the interaction between UTB and conservatism will be positive and significant. Hypothesis 2 posits a reduction in the value that investors assign to UTBs for tax aggressive firms. Therefore, we predict a negative coefficient on the interaction between UTB and tax aggressiveness.

4. Data and descriptive statistics

4.1 Sample selection

Our sample consists of firms in the intersection of COMPUSTAT and CRSP for fiscal-years 2007–2018. Our initial sample contains 110,766 firm-year observations from which we remove: firms without stock trade information in CRSP, firms without the necessary information to compute the accounting conservatism and tax aggressiveness measures and firms without the necessary information to estimate equation (1). Our test sample consists of 16,167 firm-year observations.

4.2 Descriptive statistics

Table 1 reports the descriptive statistics of our main sample. The mean (median) of our dependent variable, UTB, is 0.231 (0.039), which is lower than the statistics reported in Koester et al. (2015) [9]. The mean (median) value of MTB is 3.210 (1.957). Although the values for the other variables are varied slightly from those reported in prior research (e.g. Koester, 2012; Koester et al., 2015), they can be arguably attributable to the difference in the sample period, i.e. prior studies' inclusion of a greater proportion of firm-years from the post-recession period.

The univariate correlations reported in Table 2 show positive and significant (at the five percent significance level) correlations between PRICE and the components of the book value of equity (ASSETS, DTA, LIAB, DTL and UTB), consistent with findings in prior studies (e.g. Koester, 2012; Koester et al., 2015).

5. Results

5.1 The effect of conservatism and tax aggressiveness on the valuation of UTBs

Table 3 presents the results of estimating equation (1) for the tests of H1 and H2 on the valuation implications of UTBs conditioning on firms' financial reporting conservatism and tax aggressiveness. As a baseline, we replicate the positive association between UTBs and stock price in column (1) as reported in earlier studies (e.g. Koester, 2012; Koester et al., 2015; Brushwood et al., 2018). The coefficient on UTB is positive and significant at one percent level (β₁ = 6.567, t = 4.10), corroborating the results documented in several prior studies.

To test our hypotheses on the valuation of UTB conditional to a firm's financial reporting conservatism and tax aggressiveness, we first create dichotomized variables representing conservatism (CON_DUM) and tax aggressiveness (TAX_DUM) by using the top decile values of conservatism and tax aggressiveness measures. To provide the results for the benchmark specifications, considering the effect of conservatism only, we use CON-MB as the measure of conservatism in column 2 and CON-EP in column 3. Results in columns 2 and 3 show positive coefficients on the interaction between UTB and conservatism (UTB×CON_DUM) for both conservatism proxies supporting H1, which suggests that a firm's financial reporting conservatism influences the positive valuation of UTBs. The joint test of the overall coefficient of UTB and UTB×CON_DUM (β₁+β₃ as reported at the bottom of Panel A) also shows that the aggregate valuation weight of UTB for more conservative firms is positive and significant at one percent level [10].

By contrast, we find that the coefficient of the interaction between UTB and tax aggressiveness (UTB×TAX_DUM) in column 4 is insignificant at the conventional significance level. Although the joint test of the overall coefficient of UTB and UTB×TAX_DUM (β₁+β₅ as reported at the bottom of panel A) is positive and significant at a one percent significance level, corroborating the positive valuation of UTB for more tax aggressive firms, we fail to find evidence suggesting an incrementally negative UTB valuation implication of tax aggressiveness. Therefore, our H2 is not supported.

The results for the model specification considering both conservatism and tax aggressiveness are reported in columns 5 and 6. Although we find that the coefficients of UTB×TAX_DUM become marginally significant (β₅ = 4.949, t = 1.86) in column 4, when the ratio of market to book value of equity is used as a proxy for balance sheet conservatism, all other results are consistent with our findings from the benchmark specifications, lending support for H1 but inconsistent with H2.

In panel B, we repeat the same test using the dichotomized variables, representing conservatism (CON) and tax aggressiveness (TAX) based on their sample median values of conservatism and tax aggressiveness. To the extent that the valuation implication of conservatism and tax aggressiveness is not limited to extreme cases but more pervasive across our sample, we expect that the positive incremental effect of conservatism on the pricing of UTBs would also be captured when the indicator variables based on the sample medians are used. As expected, we find that the coefficients on the interaction between UTB and conservatism (UTB×CON_DUM) are positive and significant in panel B, but insignificant for the coefficients of UTB×TAX_DUM, corroborating our findings in panel A. Notably, however, the joint test of overall coefficients of UTB and UTB×TAX_DUM (β₁ + β₅ as reported at the bottom of Panel B columns 5 and 6) lost their statistical significance alluding to the loss of power from using dichotomized variables based on the median values as a cutoff (Irwin and McClelland 2003; Royston et al., 2006).

To mitigate the concern from using dichotomized variables, we repeat the estimation of equation (1) using the continuous variables of financial conservatism and tax aggressiveness. For this, we first adopt a step-wise approach and test for the UTB valuation effects of conservatism and tax aggressiveness separately by using the subsamples partitioned by tax aggressiveness and conservatism, respectively. The results from using the partitioned sample by the level of tax aggressiveness are summarized in Table 4. Corroborating our findings in Table 3, we find in columns 2 and 3 that the coefficients of UTB×CON are positive and significant, especially of the firms with high tax aggressiveness (in panel A column 3, β₃ = 1.886, t = 3.02; in panel B column 3, β₃ = 3.001, t = 4.37). Notably, we find that the positive valuation effect of financial conservatism is more pronounced at firms with higher tax aggressiveness.

In columns 4 and 5, we use alternative subsamples partitioned by the level of conservatism and test for the UTB valuation effect of tax aggressiveness. Consistent with our findings in Table 3, the incremental effect of tax aggressiveness on UTB valuation is generally silent.

When the full sample is used to estimate the Ohlson model of UTB valuation in Table 5, we find that the positive effect of financial conservatism on UTB valuation remains persistent. However, the coefficient of UTB×TAX becomes positive and significant in panel A column 3 (β₅ = 0.748, t = 2.02), when a firm's financial conservatism is measured based on the market to book ratio of common equity. Collectively, our empirical findings suggest that the positive valuation of UTBs documented in prior research is potentially explained by firms' financial reporting conservatism and, to a lesser extent, by tax aggressiveness policies.

5.2 Sensitivity check

To check whether our findings are robust to the inclusion of loss firms in the sample, we replace the industry-adjusted cash ETR measure (TAX-IND) with the tax avoidance measure developed by Henry and Sansing (2018) as an alternative measure of tax aggressiveness (TAX-HS). Ultimately, nonconforming tax avoidance measures intend to gauge the deviation of a firm's actual tax burden (either the reported tax expense or cash tax paid) from the maximum statutory tax burden. Although a firm's aggressive tax planning can expand the wedge by reducing the actual tax burden, an extremely low ETR by itself may not necessarily indicate a firm's tax aggressiveness. As an illustration, Schwab et al. (2020) report that about 70% of extremely low ETRs (i.e. ETRs of 5% or lower) include the release of valuation allowance, reducing the reported tax burden, but not related to a firm's effort to avoid taxes. Schwab et al. (2020) also find that Henry and Sansing's (2018) measure is more effective in capturing the consequences of tax avoidance activities than GAAP and Cash ETR measures. This finding supports the validity of our sensitivity check using Henry and Sansing's tax avoidance measure.

As reported in Table 6, we confirm that the coefficients of UTB×CON remain positive and significant, whereas the coefficients of UTB×TAX are largely insignificant even after expanding our test sample to include loss firms.

5.3 Supplementary analysis

In valuing UTBs, it is likely that investors have judgments about what can be considered normal and abnormal levels of a firm's UTBs, given the firm's environment and the nature, scale and complexity of the firm's business. With respect to the normal level of UTB, investors' expectations are likely to depend on firms' economic environments, which will include tax jurisdictions and economic profiles such as the size and complexity of the business. However, the accounting policies and financial reporting practices that underlie the measurement of UTBs provide important context for establishing the level of UTB that is normal for a firm. In this sense, the expected or normal level of UTBs conveys the size of tax positions that are in-line not only with the nature, scope and level of a firm's operations but also with the accounting and financial reporting policies of the firm. Further, it is possible that conservative firms experience relatively higher positive valuation for the predicted UTBs because their UTBs have a lower chance of being realized as future tax payments or negative cash flows. If so, we expect that the positive valuation of the predicted component of UTB will be strengthened as the level of financial reporting conservatism increases. On the contrary, the unpredicted UTBs should carry a lower valuation impact since the unpredicted component may result from estimation errors and/or white noise, not relevant for firm valuation (Rego and Wilson, 2012).

To explore this conjecture on the differential valuation effect of the predicted and the unpredicted UTBs, we decompose the UTB balance between the predicted (i.e. the amount of UTB that is associated with the firm underlying economic characteristics) and the unpredicted components of UTB by estimating the following equation [11]:

(2)UTBi,t=λ0+λ1PT_ROAi,t+λ2SIZEi,t+λ3FOR_SALEi,t+λ4R&Di,t+λ5LEVi,t+λ6DISC_ACCRi,t+λ7SG&Ai,t+λ8MTBi,t+λ9SALES_GRi,t+λ10PPEi,t+λ11NOLi,t+υi,t

We base equation (2) in models used by Cazier et al. (2009), Rego and Wilson (2012) and Nestbitt (2016) and estimate it by year. In our estimation, we depart from Nesbitt (2016) that explicitly controls for the effects of tax avoidance. In particular, we exclude a tax aggressiveness measure in our specification of the normal determinants of UTB and separately test for the impact of tax avoidance on the valuation of the UTB components.

In Table 7 column 1, we estimate the Ohlson model of UTB valuation in equation (1) after replacing UTB with PREDUTB (the predicted component of UTB) and UNPREDUTB (the unpredicted component of UTB) as a baseline. Consistent with our expectation, we find the coefficient of PREDUTB being positive significant (coefficient = 37.40, t = 12.95), whereas the coefficient of UNPREDUTB is statistically insignificant. When we interact with predicted and unpredicted components of UTBs with our measures of financial conservatism and tax aggressiveness, we find that the positive valuation effect of conservatism is driven by the valuation weight on the predicted UTB. In comparison, the positive valuation effect of tax aggressiveness appears to be driven by the valuation weight on the unpredicted UTB.

6. Conclusion

This paper examines the effect of financial reporting conservatism and tax aggressiveness on investors' valuation of UTBs and contributes to our understanding of why UTBs carry a positive weight in market valuations tests (Koester, 2012). Given a firm's UTBs can be influenced by financial reporting policy and may not solely represent its tax aggressiveness, we argue that it is necessary to maintain a balanced view of FIN 48 disclosure as a financial reporting disclosure as well as a consequence of a firm's tax aggressiveness.

In our analysis, we confirm the positive valuation of UTBs reported in the prior literature (Koester, 2012; Koester et al., 2015) and find evidence that such an effect may be fully explained by financial reporting conservatism and tax aggressiveness. We find consistent evidence indicating that conservative financial reporting may be a powerful driver in the valuation of UTBs. Our results suggest that once conservatism is considered the balance of UTB has little relevance on its own. Furthermore, our results suggest that tax aggressiveness has little incremental effect on the valuation of UTBs, which should be of interest to researchers using UTB as a proxy for tax aggressiveness.

Our results indicate that there are differential valuation implications for the predicted and unpredicted components of UTBs. Our evidence suggests that the predicted component of UTBs (i.e. the portions of UTB that are explained by the firm's economic characteristics) carries value relevant information, and it is likely the contributor to the positive valuation of UTBs. However, our evidence suggests that the unpredicted component of UTB is not relevant for valuation purposes and that it may just represent noise in the financial reporting process. The evidence also suggests that the positive valuation effect of the expected component of UTBs increases with the firm level of financial reporting conservatism.

In general, our findings indicate that financial reporting practices have a significant influence on the valuation of UTBs. The results align with the explanation that investors use the information in excess UTB reserves to adjust their expectations regarding future earnings and cash flow streams. Our results complement Koester et al. (2015) by showing that cross-sectional differences in firms' financial reporting environment help explaining the valuation implications of UTBs. On the other hand, our results indicate that tax aggressiveness plays a less significant role in the valuation of UTBs. This is important because even though UTBs can be indicative of a firm's tax shelter activities (Lisowsky et al., 2013), tax aggressiveness does not seem to be the primary factor affecting investors' valuation of UTBs. We believe our findings expand our understanding of the valuation implications of the FIN 48 disclosures and make an important contribution to the literature by highlighting the effect of financial reporting conservatism on the market valuation of UTBs.

Table 1

Descriptive statistics

Variable	N	Mean	Std.Dev	Q1	Median	Q3
PRICE	16,167	37.508	37.931	13.290	26.260	49.060
UTB	16,167	0.231	0.443	0.000	0.039	0.251
MTB	16,167	3.210	14.339	1.301	1.957	3.148
CON-MB	16,167	4.803	2.542	3.000	5.000	7.000
Easton_Pae	16,167	1.226	1.692	0.493	1.016	1.753
CON-EP	16,167	4.487	2.596	3.000	5.000	7.000
CashETR-IND	16,167	−0.047	0.306	−0.109	−0.016	0.085
TAX-IND	16,167	5.500	2.872	3.000	5.000	8.000
TaxAvoid-HS	19,800	0.003	0.018	−0.006	−0.001	0.004
TAX-HS	19,800	4.971	2.508	3.000	5.000	7.000
ASSETS	16,167	66.920	91.359	17.896	36.355	75.448
DTA	16,167	1.937	2.550	0.503	1.122	2.302
LIAB	16,167	46.900	77.692	7.638	19.479	45.804
DTL	16,167	2.868	4.693	0.261	1.044	3.120
EARN	16,167	1.889	2.734	0.505	1.413	2.753
LOSS	16,167	0.115	0.319	0.000	0.000	0.000

Note(s): Variables are defined in Appendix. The sample period includes 2007–2018. All continuous variables are winsorized at the 1st and 99th percentiles

Table 2

Pearson correlation

	(1) PRICE	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)
(2) UTB	0.377*	1.000
(3) CON-MB	0.355*	0.150*	1.000
(4) CON-EP	0.313*	0.131*	0.769*	1.000
(5) TAX-IND	0.036*	0.059*	0.034*	0.052*	1.000
(6) TAX-HS	0.123*	0.088*	0.160*	0.203*	0.414*	1.000
(7) ASSETS	0.303*	0.187*	−0.274*	−0.477*	−0.056*	−0.046*	1.000
(8) DTA	0.363*	0.394*	−0.017*	−0.120*	0.094*	0.008	0.445*	1.000
(9) LIAB	0.204*	0.116*	−0.277*	−0.519*	−0.072*	−0.061*	0.972*	0.380*	1.000
(10) DTL	0.394*	0.285*	−0.026*	−0.092*	0.113*	−0.001	0.380*	0.618*	0.257*	1.000
(11) EARN	0.667*	0.340*	0.214*	0.153*	0.043*	0.098*	0.405*	0.385*	0.303*	0.392*	1.000
(12) LOSS	−0.201*	−0.013	−0.156*	−0.104*	−0.085*	−0.079*	−0.094*	−0.022*	−0.079*	−0.061*	−0.464*

Note(s): Variables are defined in Appendix. ^*represents significant correlations at the 5 percent level or better. The sample period includes 2007–2018. The sample size is 16,167 firm-years. All continuous variables are winsorized at the 1st and 99th percentiles to reduce the undue influence of extreme values in regression analyses

Table 3

Valuation of UTBs and the effect of conservatism and tax aggressiveness

	Parameters	Pred	(1)	CON-MB	CON-EB	(4)	CON-MB	CON-EB
				(2)	(3)		(5)	(6)
			PRICE	PRICE	PRICE	PRICE	PRICE	PRICE
Panel A: CON_DUM and TAX_DUM based on highest decile
UTB	β1	+	6.567***	5.092***	5.863***	6.092***	4.555**	5.439***
			(1.602)	(1.671)	(1.635)	(1.691)	(1.776)	(1.738)
CON_DUM	β2	?		16.56***	20.90***		16.55***	20.84***
				(1.989)	(1.846)		(1.987)	(1.848)
UTB × CON_DUM	β3	+		7.607*	18.12***		7.764**	18.29***
				(3.899)	(5.008)		(3.892)	(4.999)
TAX_DUM	β4	?				−4.214***	−4.464***	−4.210***
						(0.870)	(0.845)	(0.808)
UTB × TAX_DUM	β5	−				4.535	4.949*	3.803
						(2.798)	(2.661)	(2.560)
ASSETS	β6	+	0.0403	0.126**	0.103*	0.0387	0.125**	0.101*
			(0.0522)	(0.0573)	(0.0538)	(0.0520)	(0.0570)	(0.0536)
DTA	β7	+	−0.116	−0.189	−0.102	−0.0819	−0.153	−0.0670
			(0.314)	(0.317)	(0.314)	(0.314)	(0.317)	(0.314)
LIAB	β8	−	−0.0503	−0.138**	−0.103*	−0.0488	−0.137**	−0.102*
			(0.0561)	(0.0618)	(0.0583)	(0.0559)	(0.0616)	(0.0582)
DTL	β9	?	0.709***	0.696***	0.831***	0.719***	0.707***	0.840***
			(0.226)	(0.230)	(0.225)	(0.225)	(0.229)	(0.224)
EARN	β10	+	9.729***	9.051***	8.697***	9.686***	9.004***	8.648***
			(0.437)	(0.419)	(0.402)	(0.437)	(0.419)	(0.403)
LOSS	β11	−	−0.671	−0.269	−0.201	−0.644	−0.241	−0.181
			(0.871)	(0.841)	(0.820)	(0.871)	(0.841)	(0.819)
EARN × LOSS	β12	−	−10.78***	−10.02***	−9.597***	−10.71***	−9.945***	−9.520***
			(0.521)	(0.499)	(0.478)	(0.522)	(0.499)	(0.479)
INTERCEPT	β0	?	8.246***	6.571***	5.995***	8.618***	6.966***	6.376***
			(0.891)	(0.912)	(0.909)	(0.908)	(0.929)	(0.926)
Industry-fixed effects			Included	Included	Included	Included	Included	Included
Year-fixed effects			Included	Included	Included	Included	Included	Included
UTB for higher conservatism	β1 + β3			12.699***	23.983***		12.318***	23.732***
				F = 11.02	F = 24.03		F = 10.38	F = 23.57
UTB for higher tax aggressiveness	β1 + β5					10.627***	9.503***	9.242***
						F = 16.47	F = 15.24	F = 16.01
Total effect of UTB	β1 + β3 + β5						17.267***	27.535***
							F = 15.65	F = 26.62
N			16,167	16,167	16,167	16,167	16,167	16,167
adj. R-sq			0.600	0.620	0.637	0.601	0.621	0.638
Panel B: CON_DUM and TAX_DUM based on higher/lower than median
UTB	β1	+	6.567***	0.888	1.425	4.999**	−0.313	0.547
			(1.602)	(1.740)	(1.915)	(1.979)	(2.173)	(2.402)
CON_DUM	β2	?		13.69***	14.53***		13.60***	14.36***
				(0.827)	(0.834)		(0.826)	(0.832)
UTB × CON_DUM	β3	+		7.537***	7.532***		7.600***	7.688***
				(2.668)	(2.359)		(2.677)	(2.346)
TAX_DUM	β4	?				−2.863***	−2.401***	−2.075***
						(0.684)	(0.647)	(0.645)
UTB × TAX_DUM	β5	−				2.706	2.039	1.417
						(2.407)	(2.243)	(2.282)
ASSETS	β6	+	0.0403	0.136**	0.0812	0.0390	0.134**	0.0797
			(0.0522)	(0.0587)	(0.0535)	(0.0520)	(0.0585)	(0.0534)
DTA	β7	+	−0.116	0.0608	0.178	−0.0804	0.0909	0.203
			(0.314)	(0.309)	(0.301)	(0.313)	(0.308)	(0.301)
LIAB	β8	−	−0.0503	−0.140**	−0.0757	−0.0484	−0.137**	−0.0738
			(0.0561)	(0.0633)	(0.0579)	(0.0559)	(0.0631)	(0.0579)
DTL	β9	?	0.709***	0.762***	0.910***	0.727***	0.778***	0.923***
			(0.226)	(0.219)	(0.211)	(0.225)	(0.219)	(0.211)
EARN	β10	+	9.729***	8.573***	8.697***	9.674***	8.531***	8.661***
			(0.437)	(0.445)	(0.437)	(0.436)	(0.444)	(0.436)
LOSS	β11	−	−0.671	0.909	1.842**	−0.905	0.695	1.631*
			(0.871)	(0.838)	(0.845)	(0.867)	(0.833)	(0.839)
EARN × LOSS	β12	−	−10.78***	−9.834***	−9.910***	−10.70***	−9.770***	−9.853***
			(0.521)	(0.529)	(0.516)	(0.520)	(0.528)	(0.515)
INTERCEPT	β0	?	8.246***	0.199	−0.618	9.477***	1.281	0.362
			(0.891)	(0.984)	(1.045)	(0.900)	(0.997)	(1.044)
Industry-fixed effects			Included	Included	Included	Included	Included	Included
Year-fixed effects			Included	Included	Included	Included	Included	Included
UTB for higher conservatism	β1 + β3			8.425***	8.957***		7.287***	8.235***
				F = 15.02	F = 22.34		F = 8.70	F = 14.28
UTB for higher tax aggressiveness	β1 + β5					7.705***	1.726	1.963
						F = 15.15	F = 0.75	F = 0.88
Total effect of UTB	β1 + β3 + β5						9.327***	9.652***
							F = 15.11	F = 19.47
N			16,167	16,167	16,167	16,167	16,167	16,167
adj. R-sq			0.600	0.632	0.632	0.601	0.632	0.633

Note(s): Variables are defined in Appendix. ^*, ^** and ^*** indicates significance at the 10%, 5% and 1% levels using two-tailed tests, based on clustered-robust standard errors (reported in parentheses) at the firm-level to control for heteroscedasticity and serial correlation. The sample period includes 2007–2018. In panel A, CON_DUM is an indicator variable equals to one when a conservatism measure (CON-MB or CON-EP) has the value at the highest decile; and zero otherwise. TAX_DUM is an indicator variable equals to one when the industry-adjusted cash effective rate as a tax aggressiveness measure (TAX-IND) has the value at the highest decile and zero otherwise. In panel B, CON_DUM and TAX_DUM equal to one when a conservatism measure (CON-MB or CON-EP) or tax aggressiveness measure (TAX-IND) has the value greater than the median and zero otherwise. All continuous variables are winsorized at the 1st and 99th percentiles

Table 4

Partitioned sample analysis using continuous conservatism and tax aggressiveness measures

Variables	(1)	(2)	(3)	(4)	(5)
	Full sample	Partitioned by tax aggressiveness		Partitioned by conservatism
	Baseline	Low	High	Low	High
Panel A. CON-MB as balance sheet conservatism measure
UTB	6.567***	−2.481	−3.582	2.825	−1.343
	(1.602)	(3.882)	(3.076)	(3.367)	(3.937)
CON-MB		3.551***	3.712***
		(0.255)	(0.253)
UTB × CON-MB		0.887	1.886***
		(0.734)	(0.625)
TAX				−0.394***	−0.325
				(0.0948)	(0.225)
UTB × TAX-IND				0.168	1.012*
				(0.468)	(0.574)
Control variables	YES	YES	YES	YES	YES
Industry-fixed effects	YES	YES	YES	YES	YES
Year-fixed effects	YES	YES	YES	YES	YES
N	16,167	8,084	8,083	8,084	8,083
Adj. R-squared	0.598	0.681	0.634	0.676	0.595
Panel B. CON-EP as unconditional conservatism measure
UTB	6.567***	−3.293	−8.147**	4.853	−0.00875
	(1.602)	(4.602)	(3.488)	(4.246)	(3.755)
CON-EP		4.755***	5.123***
		(0.318)	(0.322)
UTB × CON-EP		1.500*	3.001***
		(0.838)	(0.688)
TAX				−0.399***	−0.138
				(0.108)	(0.209)
UTB × TAX-IND				−0.00691	0.772
				(0.544)	(0.510)
Control variables	YES	YES	YES	YES	YES
Industry-fixed effects	YES	YES	YES	YES	YES
Year-fixed effects	YES	YES	YES	YES	YES
N	16,167	8,084	8,083	8,083	8,084
Adj. R-squared	0.598	0.692	0.652	0.661	0.614

Note(s): Variables are defined in Appendix. ^*, ^** and ^*** indicates significance at the 10%, 5% and 1% levels using two-tailed tests, based on clustered-robust standard errors (reported in parentheses) at the firm-level to control for heteroscedasticity and serial correlation. The sample period includes 2007–2018. In panels A and B, the sample is partitioned by the median value of tax aggressiveness (TAX-IND) and conservatism measures. CON-MB is used in panel A as a conservatism measure. CON-EP is used in panel B as a conservatism measure. All continuous variables are winsorized at the 1st and 99th percentiles

Table 5

Valuation of UTBs and the effect of conservatism and tax aggressiveness using continuous measures

Variables	(1)	(2)	(3)
	PRICE	PRICE	PRICE
Panel A. CON-MB as balance sheet conservatism measure
UTB	−4.224	2.200	−8.857**
	(2.813)	(2.782)	(3.646)
CON-MB	3.677***		3.663***
	(0.194)		(0.194)
UTB×CON-MB	1.663***		1.687***
	(0.553)		(0.560)
TAX-IND		−0.498***	−0.434***
		(0.120)	(0.109)
UTB×TAX-IND		0.727*	0.748**
		(0.419)	(0.369)
Control variables	YES	YES	YES
Industry-fixed effects	YES	YES	YES
Year-fixed effects	YES	YES	YES
N	16,167	16,167	16,167
Adj. R-squared	0.649	0.599	0.650
Panel B. CON-EP as unconditional conservatism measure
UTB	−7.285**	2.200	−9.992***
	(3.268)	(2.782)	(3.867)
CON-EP	4.956***		4.924***
	(0.242)		(0.242)
UTB×CON-EP	2.626***		2.657***
	(0.635)		(0.637)
TAX-IND		−0.498***	−0.341***
		(0.120)	(0.104)
UTB×TAX-IND		0.727*	0.427
		(0.419)	(0.363)
Control variables	YES	YES	YES
Industry-fixed effects	YES	YES	YES
Year-fixed effects	YES	YES	YES
N	16,167	16,167	16,167
Adj. R-squared	0.663	0.599	0.664

Note(s): Variables are defined in Appendix. ^*, ^** and ^*** indicates significance at the 10%, 5% and 1% levels using two-tailed tests, based on clustered-robust standard errors (reported in parentheses) at the firm-level to control for heteroscedasticity and serial correlation. The sample period includes 2007–2018. In panels A and B, sample is partitioned by the media value of tax aggressiveness (TAX-IND) and conservatism measures. CON-MB is used in panel A as a conservatism measure. CON-EP is used in panel B as a conservatism measure. All continuous variables are winsorized at the 1st and 99th percentiles

Table 6

Valuation of UTBs and the effect of conservatism and tax aggressiveness excluding loss firms

	(1)	CON-MB	CON-EP	(4)	CON-MB	CON-EP
		(2)	(3)		(5)	(6)
	PRICE	PRICE	PRICE	PRICE	PRICE	PRICE
UTB	7.641***	−6.797***	−8.649***	1.299	−9.164***	−9.911***
	(1.489)	(2.510)	(2.909)	(2.613)	(3.127)	(3.251)
CON		3.022***	4.064***		3.022***	4.048***
		(0.154)	(0.193)		(0.154)	(0.193)
UTB × CON		2.405***	3.222***		2.231***	3.074***
		(0.503)	(0.602)		(0.512)	(0.629)
TAX				0.593***	0.458***	0.414***
				(0.0986)	(0.0942)	(0.0931)
UTB × TAX-HS				1.098**	0.571	0.341
				(0.435)	(0.418)	(0.456)
ASSETS	0.0779*	0.214***	0.130***	0.0783*	0.212***	0.129***
	(0.0439)	(0.0570)	(0.0479)	(0.0433)	(0.0566)	(0.0476)
DTA	0.119	0.298	0.434	0.114	0.293	0.430
	(0.286)	(0.283)	(0.269)	(0.284)	(0.283)	(0.269)
LIAB	−0.0916*	−0.222***	−0.103*	−0.0917*	−0.220***	−0.103*
	(0.0483)	(0.0627)	(0.0528)	(0.0479)	(0.0624)	(0.0525)
DTL	0.645***	0.683***	0.959***	0.625***	0.665***	0.939***
	(0.203)	(0.199)	(0.191)	(0.203)	(0.199)	(0.191)
EARN	9.301***	7.691***	7.467***	9.190***	7.654***	7.446***
	(0.423)	(0.425)	(0.408)	(0.426)	(0.427)	(0.408)
LOSS	−4.612***	−3.592***	−3.482***	−3.277***	−2.609***	−2.634***
	(0.796)	(0.782)	(0.778)	(0.754)	(0.733)	(0.728)
EARN × LOSS	−10.59***	−9.371***	−9.100***	−10.46***	−9.303***	−9.051***
	(0.500)	(0.498)	(0.475)	(0.502)	(0.498)	(0.476)
INTERCEPT	8.340***	−5.923***	−10.69***	5.441***	−8.190***	−12.68***
	(0.815)	(1.096)	(1.259)	(0.813)	(1.049)	(1.214)
Industry-fixed effects	YES	YES	YES	YES	YES	YES
Year-fixed effects	YES	YES	YES	YES	YES	YES
N	19,800	19,800	19,800	19,800	19,800	19,800
Adj. R-squared	0.589	0.638	0.649	0.592	0.639	0.650

Note(s): Variables are defined in Appendix. ^*, ^** and ^*** indicates significance at the 10%, 5% and 1% levels. Statistical inferences are made using clustered-robust standard errors (reported in parentheses) at the firm-level to control for heteroscedasticity and serial correlation. All continuous variables are winsorized at the 1st and 99th percentiles

Table 7

Valuation of components of UTBs and the effect of conservatism and tax aggressiveness

Variables	(1)	CON-MB	CON-EP	CON-MB	CON-EP
		TAX-IND	TAX-IND	TAX-HS	TAX-HS
		(2)	(3)	(4)	(5)
PREDUTB	37.40***	−8.716	−10.75*	−16.75***	−11.85*
	(2.888)	(5.931)	(6.050)	(6.083)	(6.152)
UNPREDUTB	2.338	−7.963**	−7.854*	−7.480**	−8.420**
	(1.650)	(3.913)	(4.176)	(3.787)	(3.950)
CON		2.287***	3.439***	2.359***	3.499***
		(0.222)	(0.259)	(0.224)	(0.261)
PREDUTB × CON		6.789***	8.682***	6.543***	8.609***
		(0.944)	(1.021)	(0.966)	(1.044)
UNPREDUTB × CON		0.656	1.081	0.433	0.875
		(0.602)	(0.676)	(0.601)	(0.682)
TAX		−0.210	−0.187	−0.253*	−0.158
		(0.131)	(0.123)	(0.145)	(0.137)
PREDUTB × TAX		−0.246	−0.217	1.410*	0.0782
		(0.715)	(0.665)	(0.817)	(0.765)
UNPREDUTB × TAX		1.007**	0.682*	1.127**	0.952**
		(0.407)	(0.400)	(0.495)	(0.466)
ASSETS	0.0604	0.217***	0.139***	0.218***	0.140***
	(0.0509)	(0.0626)	(0.0529)	(0.0630)	(0.0530)
DTA	−0.262	−0.00660	0.206	−0.0436	0.171
	(0.310)	(0.312)	(0.291)	(0.313)	(0.291)
LIAB	−0.0797	−0.221***	−0.102*	−0.222***	−0.104*
	(0.0551)	(0.0675)	(0.0571)	(0.0678)	(0.0571)
DTL	0.656***	0.703***	0.957***	0.701***	0.955***
	(0.223)	(0.217)	(0.203)	(0.217)	(0.203)
EARN	9.343***	7.610***	7.149***	7.626***	7.168***
	(0.439)	(0.445)	(0.420)	(0.449)	(0.422)
LOSS	−0.121	1.740**	2.656***	2.148***	2.870***
	(0.850)	(0.802)	(0.781)	(0.794)	(0.778)
EARN × LOSS	−10.14***	−8.750***	−8.059***	−8.770***	−8.098***
	(0.525)	(0.525)	(0.494)	(0.529)	(0.497)
INTERCEPT	4.516***	−4.135***	−9.942***	−4.045***	−10.27***
	(0.796)	(1.450)	(1.610)	(1.321)	(1.499)
Industry-fixed effects	YES	YES	YES	YES	YES
Year-fixed effects	YES	YES	YES	YES	YES
N	16,167	16,167	16,167	16,167	16,167
Adj. R-squared	0.614	0.660	0.681	0.660	0.680

Note(s): Variables are defined in Appendix. ^*, ^** and ^*** indicates significance at the 10%, 5% and 1% levels. Statistical inferences are made using clustered-robust standard errors (reported in parentheses) at the firm-level to control for heteroscedasticity and serial correlation. All continuous variables are winsorized at the 1st and 99th percentiles

Table A1

Variable Definitions

Variable name	Definition (COMPUSTAT data items in parentheses)
PRICE	Stock price per share (CRSP PRC) one day after the Form 10-K filing date
UTB	Ending gross unrecognized tax benefits (TXTUBEND) scaled by shares outstanding (CRSP SHROUT) on the date after Form 10-K filing date. When missing, replaced by zero
MTB	Three-year average of market to book ratio calculated as (CSHO*PRCC_F) divided by CEQ
CON-MB	Deciles of ranks of MTB
Easton_Pae	Three-year average of the unconditional conservatism measure estimated following Easton and Pae (2004)
CON-EP	Deciles of ranks of Easton_Pae
CashETR-IND	Three-year average of industry-adjusted cash effective tax rate
TAX-IND	Deciles of ranks of CashETR-IND
TaxAvoid-HS	Three-year average of tax avoidance measure developed by Henry and Sansing (2018)
TAX-HS	Deciles of ranks of TaxAvoid-HS
ASSETS	Total assets (AT) net of deferred tax assets (TXNDBA) scaled by common shares outstanding on the date after Form 10-K filing date
DTA	Deferred tax assets (TXNDBA) scaled by common shares outstanding on the date after Form 10-K filing date
LIAB	Total liabilities (LT) net of deferred tax liabilities (TXNDBL) and unrecognized tax benefits (TXTUBEND) scaled by common shares outstanding on the date after Form 10-K filing date
DTL	Deferred tax liabilities (TXNDBL) scaled by common shares outstanding on the date after Form 10-K filing date
EARN	Net income before extraordinary item (IB) scaled by common shares outstanding on the date after Form 10-K filing date
LOSS	Indicator variable that is equaled to one if EARN is negative
Estimation of UTB Components
PT_ROA	Pretax income (PI) scaled by the beginning of year total assets (AT)
SIZE	Natural log of total assets (AT)
FOR_SALE	Sum of foreign sales scaled by total sales
R&D	Research and development expense (XRD) scaled by the beginning of year total assets
LEV	Total debt (DLTT + DLC) scaled by the beginning of year total assets
DISC_ACCR	Discretionary accruals estimated using performance adjusted modified Jones model
SG&A	Selling, general and administrative expenses (XSGA) scaled by the beginning of year total assets
MTB	Market to book ratio calculated as (CSHO*PRCC_F) divided by CEQ
SALES_GR	The three-year average change in sales (SALE)
PPE	Property, plant and equipment (PPENT) scaled by the beginning of year total assets (AT)
NOL	An indicator variable equals to one when net operating loss carry forwards
	(TLCF) is positive and zero otherwise

Notes

1. Similarly, greater conservatism would result in UTBs that are less fully informative about potential tax benefits to the firm (e.g. cases where firms adopt more strict requirements for current deductions) as suggested in Robinson et al. (2016).

2. Goh et al. (2017) show that accounting conservatism plays a positive role in equity markets by reducing information asymmetry between firms and shareholders. They further show that the positive impact of accounting conservatism on equity markets is more pronounced for firms with more severe information problems. In the sense of their findings, UTBs for firms with more accounting conservatism can receive more positive value to the extent that investors associate the accrued amount with greater transparency and more sustainable tax positions.

3. Specifically, we calculate this unconditional conservatism measure as the market value of equity (Compustat PRCC_F × CSHO) minus the book value of financial assets (CHE + IVAO), divided by the book value of operating assets (CEQ + DLTT + DLC).

4. We appreciate anonymous reviewer's comments and suggestions regarding this issue.

5. Our treatment of missing values is based on FIN48, para.23, which states that “The provisions of this Interpretation shall be applied to all tax positions upon initial adoption of this Interpretation” (emphasis added), which suggests that firm-year observation with missing UTBs are those for which there are zero uncertain tax positions.

6. We also use the rank of conservatism and tax aggressiveness measures for our sensitivity tests and find virtually identical results. We report the results from using their decile rank for the ease of interpretation.

7. Because the regression coefficient of UTB, financial reporting conservatism, and tax aggressiveness would reflect their conditional relationships, the interpretation on the relationship between UTB and stock price becomes conditional to the value of conservatism and tax aggressiveness, impeding the articulation of linear relationship in the OLS regression (Jaccard and Turrisi 2003).

8. We appreciate anonymous reviewer's suggestion of this approach.

9. This might occur because, unlike Koester et al. (2015), we replaced missing observations of UTBs with zeros on the assumption that firms with missing UTBs are those that do not have uncertain tax position to establish UTB reserves. To check the sensitivity of our treatment, we remove the observations with missing UTBs from the sample and re-estimate our models; the results (omitted for brevity) are similar to those reported in the text.

10. Blouin and Robinson (2014) contend that either overstating UTBs (thereby overstating tax expenses and understating earnings) or pricing of the current cash tax saved (through risky tax avoidance) may explain the positive relationship.

11. All variables in equation (2) are defined in Appendix.