North American Journal of Economics and Finance 58 (2021) 101488

Contents lists available at ScienceDirect

North American Journal of Economics and Finance

journal homepage: www.elsevier.com/locate/najef

Tax aggressiveness and idiosyncratic volatility

Neeru Chaudhry
Indian Institute of Technology, Delhi, India

A R T I C L E I N F O A B S T R A C T

Keywords: This study finds that aggressive tax strategies adopted by a firm affect idiosyncratic stock return
Idiosyncratic risk volatility. Aggressive tax strategies, which I measure as tax paid by a firm divided by pretax
Idiosyncratic volatility income (adjusted for special items), are associated with higher levels of idiosyncratic stock
Tax aggressiveness
volatility. Uncertainty associated with tax strategies may result due to several factors, such as
Tax avoidance
Tax planning
penalties, fines, and additional tax payments if particular tax strategies are disallowed by taxation
authorities, or if there are changes in tax rules. Such uncertainty affects the future cash flows of a
firm and is reflected in more volatile stock returns. Financial constraints, corporate governance
mechanisms, and information environments surrounding a firm influence the relation between
idiosyncratic volatility and effective tax rates.

1. Introduction

The taxes paid by a firm represent the share of firm value, which corporations share with the government (Modigliani and Miller
1958). Investing in tax strategies could be considered a value-maximizing activity leading to the transfer of wealth from the gov­
ernment to shareholders. An effective tax rate (ETR), defined as the tax paid by a firm per dollar of pretax income, depends on several
factors such as firm profitability, financial leverage, investment in certain assets, and specific tax-saving strategies adopted by firms to
reduce their tax liability. A firm may end up with a low ETR by taking advantage of tax provisions such as investing in certain assets
promoted by the government. In such cases, there is less uncertainty associated with firms’ tax positions. On the other hand, firms may
take risky tax positions, and given higher chances of tax benefits being claimed under such strategies, such tax strategies may not be
approved by taxation authorities and therefore result in future tax payments and penalties. Such tax strategies result in low ETRs but
increase uncertainty about firms’ future cash flows.
Pástor and Pietro (2003) show that firms with more uncertainty in valuation show higher levels of idiosyncratic volatility. In­
vestors’ valuations of a stock are affected by their assessment of benefits and risks associated with a firm’s tax strategies. Risky tax
activities make it difficult for investors to predict future profitability, introducing uncertainty into the valuation of a firm. Investors
may perceive the information contained in tax rates differently, which will affect stock valuation and the prices at which stock is traded
in the market. An aggressive tax strategy can increase uncertainty in firm valuation by increasing risks of auditing by tax authorities
(Mills 1998); requiring the payment of more taxes, penalties, and fines in the future (Wilson 2009); increasing information asymmetry
(Desai and Dharmapala, 2006; Kim, Li, and Zhang, 2011; Hasan, Hoi, Wu, and Zhang, 2014), and increasing reputational and political
costs (Chen, Chen, Cheng, and Shevlin, 2010; Chyz, Leung, Li, and Rui, 2013; Mills, Nutter, and Schwab, 2013).
An aggressive tax strategy increases the risk of a firm paying increased taxes, penalties and fines in future, thus increasing un­
certainty associated with the future profitability of a firm. Wilson (2009) estimates that firms detected by the IRS as involved in tax
sheltering pay approximately 50 percent of tax savings originally generated by tax sheltering activities in interest and penalties to the

E-mail address: [email protected].

https://doi.org/10.1016/j.najef.2021.101488
Received 27 June 2020; Received in revised form 8 May 2021; Accepted 2 June 2021
Available online 9 June 2021
1062-9408/© 2021 Elsevier Inc. All rights reserved.
N. Chaudhry North American Journal of Economics and Finance 58 (2021) 101488

IRS and pay roughly 8 percent of tax savings as fees to tax shelter purveyors. Moreover, the outcome of any tax strategy is uncertain
because tax laws can change and firms might not be able to claim certain tax benefits in the future. For example, before April 2016,
interest paid on loans from parent firms was tax-deductible for US subsidiaries of foreign companies. To take advantage of this rule,
Pfizer announced a deal to merge with Allergan and move its headquarters to Ireland (Bergin 2016).1 The Pfizer-Allergan merger deal
did not serve any business purpose and was an attempt to reduce Pfizer’s ETR from 25 to 17–18 percent while making it easier for
Pfizer to gain access to its overseas profits. However, as a result of a new rule introduced in April 2016, anticipated tax savings from the
merger could not be realized, and the deal was called off. Thus, the tax strategy adopted by Pfizer introduces uncertainty about future
tax benefits realized by the firm. Furthermore, courts do not always have the same ruling on similar transactions (Graham and Tucker
2006), as it is sometimes difficult to ascertain the legality of tax rules. For example, the IRS has disallowed deductions on certain
leveraged leases made by Wells Fargo and Consolidated Edison.2 The Court of Federal Claims has upheld tax deductions made by
Consolidated Edison while identifying some leasing transactions made by Wells Fargo as tax shelters.
Complex tax strategies lead to a poor information environment for a firm. Accounting disclosure rules do not provide sufficient
information for determining a firm’s tax position (McGill and Outslay 2004). To avoid exposing their confidential tax strategies to their
competitors, firms may not want to (and are not required to) publicly disclose information about their tax strategies (McGill and
Outslay 2004). Kim et al (2011) find that tax avoidance activities allow managers to hoard negative information for long periods and
that when this information is suddenly released to the stock market, a stock price crash results. Beck, Lin, and Ma (2014) find that
financial market development is negatively associated with the incidence and extent of tax evasion. The authors document that it is
more difficult and expensive to receive future loans for firms engaged in tax evasion. Firms suffer additional reputational costs when
news about firms’ involvement in tax avoidance negatively affects investors’ assessments of firm value (Hanlon and Slemrod, 2009;
Chen et al, 2010; Graham, Hanlon, Shevlin, and Shroff, 2014; Hasan et al, 2014).3
I define an effective tax rate (ETR) as cash taxes paid by a firm in a year divided by pretax income (adjusted for special items). ETR
incorporates various tax benefits and deductions that reduce firms’ taxable incomes. I only consider those firms that report positive
pretax income. I estimate IVOL by taking the standard deviation of residuals obtained by regressing daily excess stock returns on daily
Fama and French (1993) three-factors for each stock and for each year and multiplying this by the square root of the number of trading
days in a year. I regress IVOL on ETR while controlling for firm characteristics as well as industry and year fixed-effects.
For a sample of 20,474 firm-year observations for 3226 unique publicly listed US firms and a sample period of 1992–2017, I find
that as ETR decreases (that is, a company pays less taxes per dollar of pretax income), idiosyncratic stock return volatility increases. A
one percentage decrease in ETR is associated with a 0.008 percent increase in IVOL while controlling for variables that may affect the
relationship between ETR and IVOL. These results are robust to using alternative definitions of key variables and model specifications
and to the inclusion of additional control variables in the regression model. Further analyses confirm that main results are robust to
endogeneity. The relation between ETR and IVOL is not driven by loss making firms, firms belonging to any particular industry, or
observations belonging to recessionary periods. The empirical analyses further shows that IVOL shows more sensitivity to ETR for
financially constrained firms and for firms with poor corporate governance and poor information quality, resulting in increased
idiosyncratic stock return volatility for such firms. Additional tests show that stock returns are more volatile when CEOs own less stake
in a firm, when firms belong to a high-technology sector, and when firms are highly levered. However, these variables do not affect the
relation between ETR and IVOL. In another test, I find that the relationship between ETR and IVOL becomes stronger with the volatility
of tax strategies.
This study contributes to the literature by documenting an empirical link between effective tax rates and idiosyncratic stock return
volatility. The effective tax rate paid by a firm can incrementally predict idiosyncratic volatility over and above variables identified in
the prior literature. In contrast to prior studies that relate firm characteristics to idiosyncratic volatility, the present research identifies
investing in tax strategies, as a deliberate action taken by managers, as causing idiosyncratic volatility. The present study’s strong
results linking idiosyncratic volatility to tax rates can help investors better understand cross-sectional variations in idiosyncratic
volatility across firms and thus design their portfolios. The results documented in this work are of importance to managers, investors,
and policy makers, as ETRs may reflect how aggressive firms are in adopting their tax strategies. The study provides an explanation for
the “undersheltering puzzle” in showing why some firms do not invest in tax strategies while others engage in them frequently
(Weisbach 2002). The study results show that adopting risky tax strategies increases idiosyncratic volatility, which discourages
managers from engaging in tax-reducing activities, providing a potential explanation for the under-sheltering puzzle.
The remainder of the paper is organized as follows. Section 2 discusses related literature on tax avoidance and idiosyncratic
volatility. Section 3 describes the data and research methodology used. Section 4 presents descriptive statistics and the study’s main
empirical findings. Sections 5 and 6 present the results of robustness tests. Section 7 explains the role of different factors in the ETR-
IVOL relation. Section 8 concludes this study.

2. Related literature

Tax strategies result in savings but also increase the riskiness of future cash flow, if such strategies result in tax payments and

1
Pfizer is a pharmaceutical company with its headquarters in the US.
2
This example is taken from Lisowsky, Robinson, and Schmidt (2013).
3
The impact of such news can be significant. For example, the revelation of tax shelters employed by Dynegy from September 2000 to April 2002
resulted in a 97 percent of loss in its market value (Desai and Dharamapala 2006).

2
N. Chaudhry North American Journal of Economics and Finance 58 (2021) 101488

penalties in future. Other associated costs include the designing and implementation costs, political costs, and reputational costs
(Graham et al 2014). Mills, Erickson, and Maydew (1998) estimate a saving of $4 for every dollar invested in tax planning. For a sample
of 33 firms, Wilson (2009) notes that tax sheltering activities result in an average saving of $375.5 million in federal taxes (this figure
does not include savings from state and local taxes) and the cost of implementing these strategies is approximately 60 percent of the
resulting tax savings. Gupta and Newberry (1997) find that ETRs are significantly associated with capital structure and asset mix but
not with the firm size. Graham and Tucker (2006) document that firms use tax shelters as a substitute for debt for reducing their taxes,
and the savings from tax shelters are three times more than those generated through interest deductions.
The complex transactions relating to the implementation of tax strategies cause information asymmetry and agency problems. Tax
saving activities are associated with wealth creation for the shareholders but only in well governed firms (Desai and Dharmapala
2009). Market reacts positively to the firms’ involvement in tax sheltering activities when the quality of corporate governance is high
(Wilson 2009). On the other hand, Hanlon and Slemrod (2009) document negative market reaction to the news of firms’ involvement
in tax shelters. Because of the poor information environment surrounding tax avoidance, stock price crash risk (Kim et al 2011) and
cost of debt increases (Hasan et al 2014). Beck et al (2014) show that corporate tax avoidance is lower when financial markets are
developed as that allows better information-sharing and therefore discourages firms from engaging in tax avoidance activities.
Dyreng, Hanlon, Maydew, and Thornock (2017), and Thomsen and Watrin (2018) observe that that the ETRs have been consis­
tently declining in the recent years. The extant literature has identified several factors that affect the level of riskiness of tax strategies
adopted by firms. For example, family firms do not engage in aggressive tax planning because of their long investment horizon and
concerns about potential penalty imposed by the IRS, reputational damage from being involved in a tax related lawsuit, and potential
price discount by minority shareholders (Chen et al 2010). Badertscher, Katz, and Rego (2013) find that because of concentrated
ownership, firms with high managerial stock ownership avoid less income tax than private-equity backed firms. Investments by hedge
funds, who have expertise in tax planning, are associated with greater tax savings, which do not appear to result from high risk or
illegal tax shelters (Cheng, Huang, Li, and Stanfield, 2012). Because of labour unions’ risk preferences and their ability to capture after-
tax cash flows, managers invest less in aggressive tax strategies in the presence of strong unions (Chyz et al 2013). Edwards, Schwab,
and Shevlin (2016), and Law and Mills (2015) document that financially constrained firms engage in more tax avoidance activities.
Several studies examine the role of compensation contracts and corporate governance on firms’ tax aggressiveness. For example,
Desai and Dharmapala (2006), Rego and Wilson (2012), Armstrong, Blouin, and Larcker (2012), and Armstrong, Bloin, and Jagolinzer
(2015) document that risk-taking equity incentives influence managers appetite to invest in risky tax positions. On the other hand, Chi,
Huang, and Sanchez (2017) show that CEO debt holdings (in the form of pension benefits and deferred compensation) constraints
managers to invest in risky tax strategies. Armstrong et al (2015) document that board independence and financial sophistication have
a stronger relation with more extreme levels of tax avoidance. These board characteristics are associated with more (less) investment in
tax strategies at lower (higher) levels of tax avoidance. Dyreng, Hanlon, and Maydew (2010) document that manager fixed-effects
influence the level of tax avoidance. Rego (2003) and Hope, Ma, and Thomas (2013) show that multinational corporations are
associated with more tax avoidance activities. Gallemore and Labro (2015) suggest that high quality internal information environment
helps firms in coordinating operations across geographically dispersed units and identifying tax opportunities with more certain
positive payoffs, leading to better tax planning.
The corporate tax avoidance is negatively related to county-level religiosity (Boone, Khurana, and Raman 2013) and social capital
(Hasan, Hoi, Wu, and Zhang 2017). Politically sensitive firms, which have large contracts from the federal government and the revenue
from federal contracts account for majority of their revenues, pay higher federal taxes than other firms (Mills et al 2013). Tax
avoidance activities reduce because of public pressure following disclosure of subsidiaries, which have been set up in tax haven
countries with the sole purpose of evading taxes and no meaningful business operations (Dyreng, Hoopes, and Wilde 2016). Effective
tax rates are lower in the presence of external auditors with industry-specific tax expertise (McGuire, Omer, and Wang 2012) and when
firms borrow from a bank that also acts as a tax planning intermediary (Gallemore, Gipper, and Maydew 2019).

3. Data and methodology

Annual reports and stock return data were obtained from Compustat and the Center for Research in Security Prices (CRSP),
respectively. Data on board independence were collected from Institutional Shareholder Services (ISS). Institutional holdings data
were drawn from Thomson Reuter’s 13f filings. CEO compensation data were taken from Execucomp.
All financial services, real estate, utilities and government regulated firms are excluded (i.e., firms with SIC codes 4900–4949,
6000–6999, greater than9000). These firms are significantly different from other firms in terms of their operating activities, capital
structure, and regulatory environment. Only those stocks that are listed on the NYSE, American Stock Exchange or NASDAQ are
included in the sample. Observation with missing or negative pretax income, negative ETR, zero ETR, or ETR greater than one, and
missing values of effective tax rates, idiosyncratic volatility, and control variables are removed from the sample. The final sample has
20,474 firm-year observations from 3,226 unique firms for the period from 1992 to 2018.
Effective tax rate is computed as taxes paid divided by pretax income (adjusted for special items). Lower values of ETR indicate that
firms are paying less taxes per dollar of pretax income. Several prior studies, such as Chen et al (2010), Chyz et al (2013), and Hasan et
al (2014), use effective tax rate as a measure of tax aggressiveness. Idiosyncratic stock return volatility is estimated as the standard
deviation of residuals obtained by regressing, for each stock i and for each year t, daily excess stock returns on the three Fama and
French (1993) factors, as shown in Model (1) below. To obtain annualized figure, I follow Fu (2009) and multiply this measure by the
square root of the number of trading days in a year for a stock.

3
N. Chaudhry North American Journal of Economics and Finance 58 (2021) 101488

( )
Riτ − rf = αit + bit Rmτ − rf + sit SMBτ + hit HMLτ + εiτ (1)

where, i represents a stock, τ is the subscript for a day, t is the subscript for year (τ ∈ t), rf is the one-month Treasury-bill rate, and bit ,
sit and hit are the factor sensitivities. Firms with at least 120 days of trading data during a year are included in the sample.
Model (2) examines the relationship between effective tax rate and idiosyncratic volatility. The regression is performed using
ordinary least squares with t-statistics computed using standard errors robust to heteroscedasticity and clustered by firm.

IVOLi,t+1 = β0 + β1 LNETRi,t + β2 LNLEVi,t + β3 FIRMSIZEi,t + β4 LNMBi,t + β5 DIVIDENDi,t + β6 FIRMAGEi,t + β7 CFi,t + β8 CF σ i,t

(2)
+ β9 BETAi,t + β10 STKRETi,t + β11 STK TRNOVRi,t + IndustryandYearDummies + εit

where, IVOL is idiosyncratic volatility, LNETR is the natural log of effective tax rate, LNLEV is the natural log of the long-term debt
divided by total assets, FIRMSIZE is the natural logarithm of total assets, LNMB is the natural log of the ratio of market value of equity to
book value of equity, DIVIDEND is an indicator variable that is equal to one if a firm pays dividend and zero otherwise, FIRMAGE is the
number of years for which firm’s data is available in CRSP, CF is the cash flow from operations scaled by total assets, CFσ measures cash
flow volatility, BETA is a measure of systematic risk, STKRET is annual stock returns, and STK_TRNOVR represents stock turnover.
Industry and year dummies control for industry (at the two-digit SIC level) and year fixed-effects, respectively. All continuous variables
are winsorized at the 1st and 99th percentiles.

4. Results

4.1. Descriptive Statistics

Table 1 Panel A presents the sample distribution by industry. Manufacturing accounts for approximately half of the sample.

Table 1
Sample Distribution by Industry and Year. This table presents sample distribution by industry (Panel A) and year (Panel B). Effective tax rate (ETR)
is the tax paid divided by the pretax income adjusted for special items. IVOL is idiosyncratic volatility estimated as the standard deviation of the
residuals obtained by regressing excess stock returns on the three Fama and French (1993) factors multiplied by the number of trading days in a year,
for each stock and for each year. Column (1) reports the number of observations. Columns (2) and (3) report mean IVOL and mean ETR, Columns (4)
and (5) report median IVOL and median ETR. IVOL is winsorized at the 1st and 99th percentile and values of ETR range from 0 to 1. The sample period
is from 1992 to 2017.
(1) (2) (3) (4) (5)
N Mean ETR Mean IVOL Median ETR Median IVOL
(in percent) (in percent) (in percent) (in percent)

Panel A: This panel presents distribution of the sample by industry.

Agriculture 117 21.779 32.144 21.428 28.950
Construction 439 28.108 35.906 29.817 33.422
Manufacturing 12,080 26.356 35.511 25.373 31.182
Mining 1075 20.520 36.929 16.290 33.229
Retail Trade 2133 30.056 36.835 30.420 33.219
Services 3533 26.013 38.365 24.580 34.499
Wholesale Trade 1093 29.179 36.009 30.418 32.675
Panel B: This panel presents distribution of the sample by year.
1992 148 36.493 40.161 33.194 35.308
1993 809 31.899 39.261 31.039 34.517
1994 915 30.822 38.240 31.056 32.842
1995 924 30.688 36.748 30.961 32.231
1996 957 28.598 38.883 29.147 34.050
1997 945 29.360 38.778 28.809 34.549
1998 905 30.346 44.138 29.984 41.397
1999 870 29.186 45.883 28.678 42.511
2000 820 28.022 52.740 27.903 48.762
2001 832 27.407 45.545 25.880 41.942
2002 861 24.216 41.354 23.527 38.131
2003 909 22.960 33.766 21.388 30.468
2004 1004 21.800 32.569 20.173 29.100
2005 1009 24.850 30.696 23.275 27.745
2006 947 26.454 30.998 26.674 29.149
2007 831 27.705 31.344 27.069 28.921
2008 671 26.477 45.636 25.885 43.623
2009 737 26.270 42.151 25.077 38.127
2010 833 23.768 30.253 22.456 27.570
2011 815 22.651 30.456 21.451 28.080
2012 817 23.247 28.228 22.797 25.609
2013 799 24.081 25.809 22.937 23.385
2014 755 24.832 25.689 24.617 22.758
2015 730 24.568 28.250 23.976 25.409
2016 627 23.377 29.484 23.243 26.864

4
N. Chaudhry North American Journal of Economics and Finance 58 (2021) 101488

Industry average (median) ETR varies from 20 (16) to 30 (30) percent. Panel B presents the sample distribution by year. The sample
ranges from a minimum of 148 firms in 1992 to a maximum of 1009 firms in 2005. During the 1992–2004 period, ETR seems to fall
gradually from approximately 33–35 to 20 percent. For the 2005–2017 period, ETR varies within a closer range of 21 to 28 percent. A
downward trend in IVOL is observed during the sample period with the exception of two peaks observed in 2000 and 2008. These two
peaks are attributable to the bursting of the information technology bubble and global financial crisis, respectively. This information is
also presented in graphical form in Fig. 1. Table 2 Panel A reports summary statistics for the variables used in this study. For the full
sample, mean (median) ETR and IVOL are valued at approximately 27 (26) and 36 (32) percent, respectively. Panel B shows that the
Pearson (Spearman) correlation coefficient between ETR and IVOL is negative and statistically significant at the five percent level.
Panel C reports mean and median values of variables for each of the subsamples obtained after dividing the full sample into firm
observations with ETR values of below (low-ETR) and above (high-ETR) sample median ETR. I observe that mean and median IVOL
values are higher for the subsample with low ETR. The difference is statistically significant at the one percent level. This serves as
preliminary evidence that effective tax rates influence stock return volatility. Relative to high-ETR firms, low-ETR firms exhibit high
levels of financial leverage, are larger in size, have more growth opportunities, are less likely to pay dividends, are younger, have
greater cash flow to total asset ratio, exhibit more cash flow volatility, show higher levels of systematic risk, generate more stock
returns, and have high rates of stock turnover. I account for these differences in our regression model by controlling for these variables.

4.2. Regression results

Table 3 presents results obtained by estimating Model (2). Column (1) shows that the coefficients on LNETR are significant at the
one percent level. A one percentage drop in ETR is associated with a − 0.008 percentage change in IVOL. This result is consistent with
the negative relationship hypothesized between IVOL and ETR.
Dyreng, Hanlon, and Maydew (2008) argue that the average value of the effective tax rate over longer periods (e.g., five years)
represents firms’ abilities to pay consistently less in taxes over longer periods. Therefore, I check my results using three-year (ETR3)
and five-year ETR (ETR5). For year t, ETR3 is calculated by taking the sum of taxes paid by a firm over a period of three years (i.e., taxes
paid in years t, t + 1, and t + 2) and dividing it by a sum of pretax income (adjusted for special items) for the same period. Similarly, I
calculate five-year average ETR (ETR5). Accumulating cash taxes paid over three/five years minimizes the mismatching problem.
Columns (2) and (3) show results obtained from the estimation of Model (2) where I use log forms of ETR3 and ETR5 as explanatory
variables, respectively. The coefficients on these variables are similar in sign to those obtained for ETR in Column (1). The results show
that measuring ETR from taxes and income for longer periods generates similar relations to IVOL as the annual measure of ETR used in
Column (1).
Consistent with prior literature, the results show that idiosyncratic volatility decreases with firm size, the market-to-book ratio,
firm age and cash flows from operations and increases with leverage, cash flow volatility, market beta, stock returns and stock
turnover. Relative to non-dividend paying firms, dividend-paying firms have low levels of idiosyncratic volatility.

5. Robustness tests

Several robustness tests are performed to verify these findings. First, I replace the raw idiosyncratic volatility measure with its log
transformation. This test addresses the fact that the skewness of the idiosyncratic volatility measure may influence statistical inferences
of the analysis (Goyal and Santa-Clara, 2003; Xu and Malkiel, 2003; Gasper and Massa, 2006). Second, I perform a Fama and MacBeth
(1973) regression to address concerns that within-firm autocorrelation in the pooled-OLS regression might bias the standard errors. In
the presence of cross-sectional correlation, Fama and Macbeth’s (1973) method provides unbiased standard errors (Petersen 2009).
Third, I include the lagged values of idiosyncratic volatility as an additional control variable. This variable controls for the effect of past
idiosyncratic volatility on effective tax rates, which addresses the endogeneity problem to a certain extent. Fourth, I control for
profitability variables to address concerns that the findings may be driven by the omitted profitability variables. Specifically, in the
regression model, I include an indicator variable that is equal to one when a firm reports a net operating loss (NOL) and zero otherwise,
the change in net operating loss (DNOL) scaled by total assets, and the return on assets (ROA). The regression results are presented in
the first four columns of Table 4. In all tests, the coefficient on the ETR variable is negatively and significantly related to idiosyncratic
volatility, suggesting that the findings are robust to varying regression specifications.
To remove influence of recessionary years on the main findings, I modify the sample by removing all observations belonging to
years identified as recessionary periods by the National Bureau of Economic Research.4 During an economic downturn, a government
may offer certain tax incentives to corporations to encourage them to invest in specific assets and stimulate the economy. Table 1 Panel
B (and Fig. 1) shows that during recessionary years, very high IVOL is observed while no sharp change is observed in ETR. Perhaps
during a recession, firms’ profits and taxes paid by them are low, causing the ratio of pretax income to paid taxes to be at a level similar
to that in a non-recession period. Results for this modified sample are similar to those observed for the full sample and are shown in
Column (5). The results confirm that periods of economic downturn do not influence the main results obtained for the full sample.
Next, I check the robustness of the findings by using alternative models to calculate idiosyncratic volatility, including the Carhart
(1997) four-factor model, value-weighted market model, and equal-weighted market model. As is shown in Columns (6)-(8), the

4
This information is obtained from www.nber.org/cycles.html.

5
N. Chaudhry North American Journal of Economics and Finance 58 (2021) 101488

Fig. 1. This figure plots the mean and median values of idiosyncratic stock return volatility and effective tax rate variables over the sample period.
Idiosyncratic volatility is estimated as the standard deviation of the residuals obtained by regressing excess stock returns, for each stock and for each
year, on the three Fama and French (1993) factors and effective tax rate is the tax paid divided by pretax income (adjusted for special items).

coefficient on the tax variable (LNCETR) is negative and statistically significant, suggesting that the findings are robust to using
alternative models of idiosyncratic volatility. As shown in Column (9), I find similar results when I use total stock return volatility as
the dependent variable.
Table 1 shows that manufacturing firms account for more than half of the sample. The observed relationship between ETR and IVOL
may exist due to the specific features of industries to which firms belong. For example, some firms may have low tax rates because they
belong to industries requiring considerable capital investment, and such firms may also have more volatile stock returns. In the main
analysis, I control for industry fixed-effects as a robustness test, and I estimate Model (2) separately for each industry (i.e., Agriculture,
Mining, Construction, Manufacturing, Wholesale Trade, Retail Trade, and Services). The results are presented in Table 5.5 The relation
between ETR and IVOL holds for all industries except for Agriculture, Mining, and Construction, for which the coefficient on LNETR is
not statistically significant. This analysis accounts for differences in industry while also confirming that the main results are not driven
by any one industry.
I construct a new tax variable, IND_SIZE_ETR, adjusted for industry affiliation and firm size (the book value of total assets). I
compare firms’ ETR values with those of similarly sized firms of the same industry. Specifically, for each industry and year, I divide
firms into quartiles by firm size. For each size quartile, I divide observations into quartiles based on ETR. I define the most (least) tax
aggressive firms as those belonging to the first (fourth) quartile. IND_SIZE_ETR takes a value of one for tax aggressive firms and a value
of zero otherwise. Results obtained from this measure of tax aggressiveness are presented in Column (8) of Table 5. The coefficient on
IND_SIZE_ETR is positive and statistically significant at the one percent level.
Taken together, there is a strong and robust evidence supporting the hypothesis that the tax strategies of a firm affect the idio­
syncratic volatility of stock returns.

6. Endogeneity

If the tax strategies adopted by firms vary endogenously with some unobservable firm characteristics that are responsible for
idiosyncratic volatility, then using pooled OLS results in biased and inconsistent estimates. There could be a potential reverse causality
from idiosyncratic volatility to tax aggressiveness, for example, if higher idiosyncratic volatility incentivizes managers to undertake
aggressive tax strategies so that they can reduce tax payments and build financial slack to absorb unexpected shocks to the cash flow.
To address these endogeneity concerns, I use first-difference, fixed-effect, and instrument variable estimation. In addition I exploit a
quasi-experiment setting to establish that aggressive tax strategies cause higher levels of idiosyncratic stock volatility.

5
As few observations pertain to the Agriculture and Construction sectors (117 and 439, respectively), the results for these industries could not be
reliably estimated.

6
 Table 2
Descriptive Statistics. This table presents descriptive statistics. ETR is the tax paid divided by the pretax income adjusted for special items. IVOL is idiosyncratic volatility estimated as the standard

N. Chaudhry
deviation of the residuals obtained by regressing excess stock returns on the three Fama and French (1993) factors multiplied by the number of trading days in a year, for each stock and for each year. LEV
is long term debt divided by total assets, FIRMSIZE is the natural log of the book value of total assets, MB is market value of equity divided by book value of equity, DIVIDEND is an indicator variable that is
equal to one if a firm pays dividend and zero otherwise, FIRMAGE is the number of years a firm exists in the Compustat database, CF is cash flow from operating activities scaled by total assets, CFσ is
standard deviation of CF, using data for the trailing five years, BETA is market beta, STKRET is annual stock returns, and STK_TRNOVR is average monthly trading volume to shares outstanding over a year.
All continuous variables are winsorized at the 1st and 99th percentile. The sample period is from 1992 to 2017.
Panel A: This panel reports summary statistics for all variables.
(1) (2) (3) (4) (5)
Mean Median 25th Percentile 75th Percentile Std Dev

ETR (in percent) 26.538 25.770 15.069 34.985 17.037

IVOL (in percent) 36.233 32.173 24.002 44.331 16.848
LEV 0.189 0.172 0.075 0.278 0.140
FIRMSIZE 6.780 6.795 5.532 8.019 1.786
MB 2.902 2.264 1.533 3.461 2.276
DIVIDEND 0.542 1 0 1 0.498
FIRMAGE 21.991 18 10 30 16.236
CF 0.109 0.105 0.068 0.147 0.066
CFσ 0.048 0.039 0.024 0.060 0.034
BETA 1.035 0.992 0.645 1.365 0.604
STKRET 0.224 0.150 − 0.066 0.417 0.461
STK_TRNOVR 1.407 1.062 0.527 1.911 1.209

Panel B: This panel reports Pearson (Spearman) correlation coefficients in the above (below) the diagonal. Significant coefficients (at the 5 percent level) are shown in bold.
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)

ETR (1) ¡0.037 ¡0.096 ¡0.081 ¡0.055 0.136 0.064 ¡0.065 ¡0.025 ¡0.085 ¡0.098 ¡0.152
IVOL (2) ¡0.096 ¡0.095 ¡0.559 ¡0.232 ¡0.412 ¡0.328 ¡0.070 0.385 0.066 − 0.009 0.005
7

LEV (3) ¡0.050 ¡0.123 0.230 0.003 0.053 0.001 ¡0.178 ¡0.188 0.019 ¡0.039 0.010
FIRMSIZE (4) 0.003 ¡0.565 0.230 0.252 0.318 0.240 0.047 ¡0.376 0.149 ¡0.044 0.395
MB (5) ¡0.018 ¡0.223 ¡0.016 0.252 0.084 0.009 0.352 ¡0.058 − 0.002 0.235 0.242
DIVIDEND (6) 0.162 ¡0.395 0.092 0.330 0.087 0.376 0.080 ¡0.254 ¡0.102 ¡0.048 ¡0.166
FIRMAGE (7) 0.082 ¡0.288 0.078 0.310 0.056 0.367 0.006 ¡0.204 ¡0.069 ¡0.027 ¡0.097
CF (8) 0.046 0.350 0.069 − 0.008 0.002 0.082 0.093

North American Journal of Economics and Finance 58 (2021) 101488

¡0.054 ¡0.061 ¡0.157 ¡0.055
CFσ (9) ¡0.085 0.347 ¡0.182 ¡0.350 ¡0.023 ¡0.247 ¡0.196 0.011 0.061 0.048 0.022
BETA (10) ¡0.088 0.040 0.002 0.132 − 0.002 ¡0.111 ¡0.058 ¡0.039 0.063 0.023 0.257
STKRET (11) ¡0.110 0.130 ¡0.048 ¡0.102 0.201 ¡0.097 ¡0.061 0.071 0.087 0.030 0.013
STK_TRNOVR (12) ¡0.116 0.059 ¡0.038 0.287 0.180 ¡0.168 ¡0.096 0.115 0.079 0.249 0.032

Panel C: The full sample is divided into two subsamples based on whether the ETR is below or above the sample median ETR (for each year). This panel reports mean and median values of variables for the two subsamples and
reports results from test of difference-in-means and difference-in-medians. Statistical significance at the one, five, and ten percent level is indicated by ***, **, and *, respectively.
Low-ETR Firms High-ETR Firms Difference-in-Means Difference-in-Medians
(Below sample median) (Above sample median)
(1) (2) (3) (4) (5) (6) (7) = (2) – (5) (8) = (3) – (6)

ETR 10,245 14.502 15.082 10,225 38.597 34.997 − 24.095*** − 19.915***

IVOL 10,149 37.380 33.273 10,120 35.077 31.119 2.304*** 2.154***
LEV 10,232 0.202 0.184 10,220 0.176 0.160 0.025*** 0.024***
FIRMSIZE 10,245 6.842 6.904 10,225 6.718 6.691 0.124*** 0.213***
MB 10,244 2.991 2.296 10,224 2.813 2.235 0.178*** 0.061***
DIVIDEND 10,245 0.490 0 10,225 0.594 1 − 0.104*** − 1***
FIRMAGE 10,245 21.538 17 10,225 22.445 19 − 0.907*** − 2***
CF 10,245 0.111 0.106 10,225 0.108 0.104 0.003*** 0.001***
CFσ 10,245 0.049 0.039 10,225 0.047 0.038 0.002*** 0.001***
BETA 10,245 1.083 1.038 10,225 0.986 0.947 0.097*** 0.091***
STKRET 10,245 0.259 0.170 10,225 0.188 0.130 0.071*** 0.040***
STK_TRNOVR 10,245 1.478 1.127 10,225 1.335 1.002 0.142*** 0.125***
N. Chaudhry North American Journal of Economics and Finance 58 (2021) 101488

Table 3
Effective Tax Rate and Idiosyncratic Stock Return Volatility. This table presents OLS results, where dependent variable is idio­
syncratic volatility (IVOL), which is estimated as the standard deviation of the residuals obtained by regressing excess stock returns on
the three Fama and French (1993) factors multiplied by the number of trading days in a year, for each stock and for each year. LNETR
is the natural log of the tax paid divided by the pretax income adjusted for special items. LNETR3 (LNETR5) is the natural log of the
sum of taxes paid over a three (five) year period divided by the pretax income (adjusted for special items) over the same period. LNLEV
is the natural log of the long term debt divided by total assets, FIRMSIZE is the natural log of the book value of total assets, LNMB is the
natural log of the market value of equity to book value of equity ratio, DIVIDEND is an indicator variable that is equal to one if a firm
pays dividend and zero otherwise, FIRMAGE is the number of years a firm exists in the Compustat database, CF is cash flow from
operating activities scaled by total assets, CFσ is standard deviation of CF, using data for the trailing five years, BETA is market beta,
STKRET is annual stock returns, and STK_TRNOVR is average monthly trading volume to shares outstanding over a year. All
continuous variables are winsorized at the 1st and 99th percentile. The sample period is from 1992 to 2017. Statistical significance at
the one, five, and ten percent level is indicated by ***, **, and *, respectively. t-Statistics (in parentheses) are calculated based on
heteroscedasticity-robust standard errors clustered by firm.
(1) (2) (3)

LNETR − 0.008
(-6.730)***
LNETR3 − 0.014
(-6.347)***
LNETR5 − 0.013
(-3.314)***
LNLEV 0.002 0.002 0.002
(2.206)** (1.666)* (1.436)
FIRMSIZE − 0.040 − 0.040 − 0.039
(-34.428)*** (-31.058)*** (-21.694)***
LNMB − 0.019 − 0.021 − 0.019
(-8.522)*** (-8.397)*** (-5.963)***
DIVIDEND − 0.041 − 0.041 − 0.048
(-13.064)*** (-12.070)*** (-10.513)***
FIRMAGE − 0.000 − 0.000 − 0.000
(-4.626)*** (-3.606)*** (-1.570)
CF − 0.154 − 0.140 − 0.138
(-8.157)*** (-6.749)*** (-5.230)***
CFσ 0.446 0.420 0.355
(10.637)*** (9.128)*** (5.677)***
BETA 0.011 0.009 0.009
(4.972)*** (4.019)*** (2.753)***
STKRET 0.029 0.032 0.033
(13.521)*** (14.029)*** (10.960)***
STK_TRNOVR 0.018 0.016 0.015
(17.348)*** (13.916)*** (9.783)***
Constant 0.574 0.566 0.568
(23.588)*** (21.441)*** (14.403)***
Year Effects Yes Yes Yes
Industry Effects Yes Yes Yes
N 20,474 16,961 9,712
R2 0.569 0.570 0.566

6.1. First-Difference and Fixed-Effects model

If the omitted variable does not change over time then first-difference or fixed-effect method can be used to address endogeneity.
Both these methods help in mitigating concerns related to the potential bias that may occur from the presence of time-invariant
unobservable heterogeneity. First-difference method considers how changes in the tax strategies over time affect the change in
idiosyncratic volatility over the same time period. To remove the unobserved effect, it requires first-differencing of all variables in
adjacent time periods and then use a standard OLS analysis. The fixed-effects method involves time-demeaning of all variables to
remove the unobserved fixed-effect from the model. In both the methods, any variable that is constant over time drops out of the
analysis. The explanatory variables are strictly exogenous after taking out the unobserved effect.
Table 6 present results for the first-difference method. The coefficient on the tax variable is negative and statistically significant,
suggesting that with the decrease in ETR, idiosyncratic stock return volatility increases. Column (1) of Table 7 report results obtained
from the fixed-effect model. The coefficient on the effective tax rate is negative and statistically significant at the one-percent level.
These results are consistent with the main findings reported earlier and mitigate concerns regarding the omitted variable bias which
may result from the correlation between time-constant variables in the error term and explanatory variables.

6.2. Instrument variable regression

One of the disadvantages of first-difference and fixed-effect methods is that along with unobserved effect, they also remove

8
N. Chaudhry North American Journal of Economics and Finance 58 (2021) 101488

Table 4
Robustness Tests. This table presents OLS regression results, where dependent variable is idiosyncratic volatility (IVOL). In Columns (1)-(5), IVOL is
estimated as the standard deviation of the residuals obtained by regressing excess stock returns on the three Fama and French (1993) factors
multiplied by the number of trading days in a year, for each stock and for each year. In Column (1), dependent variable is natural log of IVOL. In
Columns (6)-(8), IVOL is estimated using Carhart (1997) four-factor model, value-weighted market model and equal-weighted market model,
respectively. In Column (9), dependent variable is the standard deviation of the daily returns (annualized). LNETR is the natural log of the tax paid
divided by the pretax income adjusted for special items. LNLEV is the natural log of the long term debt divided by total assets, FIRMSIZE is the natural
log of the book value of total assets, LNMB is the natural log of the market value of equity to book value of equity ratio, DIVIDEND is an indicator
variable that is equal to one if a firm pays dividend and zero otherwise, FIRMAGE is the number of years a firm exists in the Compustat database, CF is
cash flow from operating activities scaled by total assets, CFσ is standard deviation of CF, using data for the trailing five years, BETA is market beta,
STKRET is annual stock returns, and STK_TRNOVR is average monthly trading volume to shares outstanding over a year. LAGGED_IVOL is one-year
lagged IVOL. ROA is return-on-assets, TLCF is equal to one if a firm reports tax loss carryforward, ΔTLCF is the change in tax loss carryforward. All
continuous variables are winsorized at the 1st and 99th percentile. Column (2) results obtained from Fama and McBeth (1973) regression. The sample
period is from 1992 to 2017. Statistical significance at the one, five, and ten percent level is indicated by ***, **, and *, respectively. t-Statistics (in
parentheses) are calculated based on heteroscedasticity-robust standard errors clustered by firm.
(1) (2) (3) (4) (5) (6) (7) (8) (9)
IVOL estimated using
y = Log Fama- Controlling Controlling for Excluding Carhart Value- Equal- y = Stock
(IVOL) MacBeth for lagged profitability Recessionary (1997) weighted weighted return
Regression IVOL variables years Model market market volatility
model model

LNETR − 0.021 − 0.009 − 0.002 − 0.006 − 0.008 − 0.011 − 0.011 − 0.010 − 0.010
(-6.877) (-7.649)*** (-2.204)** (-3.654)*** (-4.918)*** (-6.589) (-6.509)*** (-6.350)*** (-5.666)
*** *** ***
LNLEV 0.005 0.001 0.001 0.001 0.002 0.002 0.002 0.001 0.001
(2.018)** (2.589)** (1.739)* (0.750) (2.202)** (1.954)* (1.793)* (1.431) (1.217)
FIRMSIZE − 0.107 − 0.038 − 0.013 − 0.037 − 0.040 − 0.042 − 0.042 − 0.040 − 0.037
(-38.218) (-25.061) (-21.939)*** (-27.739)*** (-21.151)*** (–32.975) (–33.106) (-31.438) (-28.913)
*** *** *** *** *** ***
LNMB − 0.058 − 0.022 − 0.001 − 0.009 − 0.019 − 0.020 − 0.020 − 0.018 − 0.017
(-10.086) (-7.548)*** (-0.549) (-3.076)*** (-5.008)*** (-8.570) (-8.317)*** (-7.678)*** (-6.495)
*** *** ***
DIVIDEND − 0.109 − 0.043 − 0.015 − 0.036 − 0.041 − 0.043 − 0.045 − 0.044 − 0.045
(-13.323) (-7.508)*** (-9.107)*** (-9.723)*** (-8.006)*** (-12.526) (-12.760) (-12.540) (-11.234)
*** *** *** *** ***
FIRMAGE − 0.002 − 0.000 − 0.000 − 0.001 − 0.000 − 0.000 − 0.000 − 0.000 − 0.000
(-6.850) (-6.931)*** (-5.675)*** (-5.303)*** (-3.869)*** (-4.065) (-3.944)*** (-3.524)*** (-3.628)
*** *** ***
CF − 0.372 − 0.090 − 0.089 − 0.106 − 0.137 − 0.166 − 0.168 − 0.159 − 0.183
(-8.242) (-3.476)*** (-7.491)*** (-4.195)*** (-4.783)*** (-7.451) (-7.493)*** (-7.090)*** (-7.803)
*** *** ***
CFσ 1.097 0.451 0.176 0.477 0.428 0.460 0.456 0.450 0.415
(10.792) (15.139)*** (7.399)*** (8.560)*** (10.515)*** (9.730)*** (9.557)*** (9.373)*** (7.747)***
***
BETA 0.039 0.011 0.004 0.010 0.011 0.010 0.011 0.010 0.016
(7.809) (5.156)*** (2.908)*** (3.982)*** (3.777)*** (4.338)*** (4.634)*** (4.329)*** (6.330)***
***
STKRET 0.079 0.021 0.005 0.029 0.030 0.026 0.028 0.028 0.033
(16.341) (3.043)*** (2.352)** (10.306)*** (4.754)*** (10.877) (11.290) (11.313) (12.782)
*** *** *** *** ***
STK_TRNOVR 0.054 0.022 0.001 0.020 0.020 0.017 0.018 0.018 0.023
(20.059) (8.751)*** (1.082) (15.758)*** (7.111)*** (14.799) (15.504) (15.290) (17.465)
*** *** *** *** ***
ROA − 0.159
(-3.918)***
NOL 0.007
(2.072)**
DNOL 0.013
(0.247)
LAGGED_IVOL 0.578
(68.880)***
Constant − 0.503 0.588 0.210 0.553 0.580 0.588 0.592 0.580 0.554
(-9.399) (30.294)*** (15.686)*** (21.203)*** (27.061)*** (22.207) (22.859) (22.127) (22.551)
*** *** *** *** ***
N 20,474 20,474 20,273 12,076 18,361 20,778 20,778 20,778 20,778
R2 0.612 0.465 0.719 0.567 0.575 0.533 0.534 0.523 0.495

9
N. Chaudhry North American Journal of Economics and Finance 58 (2021) 101488

Table 5
Industrywise Analysis. This table presents OLS results, where dependent variable is idiosyncratic volatility (IVOL), which is estimated as the
standard deviation of the residuals obtained by regressing excess stock returns on the three Fama and French (1993) factors multiplied by the number
of trading days in a year, for each stock and for each year. LNETR is the natural log of the tax paid divided by the pretax income adjusted for special
items. LNLEV is the natural log of the long term debt divided by total assets, FIRMSIZE is the natural log of the book value of total assets, LNMB is the
natural log of the market value of equity to book value of equity ratio, DIVIDEND is an indicator variable that is equal to one if a firm pays dividend
and zero otherwise, FIRMAGE is the number of years a firm exists in the Compustat database, CF is cash flow from operating activities scaled by total
assets, CFσ is standard deviation of CF, using data for the trailing five years, BETA is market beta, STKRET is annual stock returns, and STK_TRNOVR is
average monthly trading volume to shares outstanding over a year. IND_SIZE_ETR is obtained after sorting firms within an industry first on firm size
and then on ETR; it is defined as one if a firm belongs to first quartile of ETR and zero if firm belongs to the fourth quartile. All continuous variables are
winsorized at the 1st and 99th percentile. The sample period is from 1992 to 2017. Statistical significance at the one, five, and ten percent level is
indicated by ***, **, and *, respectively. t-Statistics (in parentheses) are calculated based on heteroscedasticity-robust standard errors clustered by
firm.
(1) (2) (3) (4) (5) (6) (7) (8)
Agriculture Mining Construction Manufacturing Wholesale Retail Services Dependent variable is
Trade Trade IND_SIZE_ETR

LNETR − 0.017 0.001 − 0.008 − 0.008 − 0.020 − 0.009 − 0.009

(-1.362) (0.402) (-1.066) (-4.815)*** (-3.151)*** (-2.151)** (-3.365)
***
IND_SIZE_ETR 0.022
(7.291)***
LNLEV 0.021 − 0.000 0.008 0.000 0.006 0.007 0.003 0.003
(1.866)* (-0.011) (1.172) (0.175) (1.698)* (3.019)*** (1.749)* (2.153)**
FIRMSIZE − 0.006 − 0.030 − 0.047 − 0.039 − 0.047 − 0.038 − 0.047 − 0.043
(-0.447) (-7.714) (-5.100)*** (-25.047)*** (-9.416)*** (-12.972) (-19.387) (-31.529)***
*** *** ***
LNMB − 0.014 − 0.027 0.001 − 0.018 − 0.019 − 0.023 − 0.019 − 0.023
(-0.637) (-2.247) (0.050) (-5.907)*** (-1.917)* (-4.268) (-3.723) (-8.412)***
** *** ***
DIVIDEND − 0.046 − 0.041 − 0.013 − 0.048 − 0.038 − 0.033 − 0.028 − 0.040
(-1.052) (-3.659) (-0.756) (-11.048)*** (-3.568)*** (-4.334) (-3.704) (-10.796)***
*** *** ***
FIRMAGE − 0.000 − 0.000 0.000 − 0.000 − 0.001 − 0.000 − 0.001 − 0.000
(-0.712) (-1.524) (0.639) (-3.279)*** (-2.086)** (-1.239) (-2.990) (-3.907)***
***
CF − 0.404 − 0.205 0.013 − 0.181 − 0.047 − 0.131 − 0.146 − 0.099
(-2.605)** (-3.141) (0.147) (-7.360)*** (-0.653) (-2.355)** (-2.798) (-4.224)***
*** ***
CFσ 0.715 0.452 0.103 0.425 0.468 0.392 0.353 0.380
(1.190) (2.537)** (0.471) (7.758)*** (2.214)** (3.139)*** (3.873)*** (8.058)***
BETA 0.097 0.005 0.009 0.011 0.028 0.006 0.005 0.010
(3.519)*** (0.674) (0.901) (3.604)*** (2.789)*** (1.005) (1.163) (3.756)***
STKRET 0.040 0.051 0.027 0.033 0.012 0.016 0.019 0.026
(1.449) (5.799) (1.973)* (11.040)*** (1.290) (3.139)*** (3.763)*** (8.860)***
***
STK_TRNOVR − 0.015 0.010 0.002 0.019 0.012 0.012 0.023 0.017
(-0.510) (2.755) (0.417) (13.199)*** (2.113)** (4.681)*** (9.339)*** (13.182)***
***
Constant 0.322 0.763 0.629 0.599 0.566 0.641 0.712 0.634
(3.864)*** (9.614) (9.755)*** (32.160)*** (15.496)*** (16.139) (20.749) (21.998)***
*** *** ***
Year Effects Yes Yes Yes Yes Yes Yes Yes Yes
Industry Yes Yes Yes Yes Yes Yes Yes Yes
Effects
N 117 1,075 439 12,083 1,093 2,134 3,533 10,198
R2 0.654 0.598 0.552 0.574 0.583 0.620 0.594 0.573

explanatory variables that are constant over time for all firms. The instrumental variable approach recognizes the presence of the
omitted variables in the error term. It provides better estimates of the ceteris paribus effect of the explanatory variable on dependent
variable when explanatory variable and error are correlated. The 2SLS estimator is less efficient than OLS when the explanatory
variables are exogenous. Both Durbin and Wu–Hausman test statistics are highly significant, rejecting the null hypothesis of exogeneity
and confirming that effective tax rate be treated as endogenous.
Addressing endogeneity issue using instrument variable two-stage-least-square (2SLS) method require instrument variable, which
is significantly related to the effective tax rate but is not correlated with other factors that affect idiosyncratic volatility. In the first-
stage, ETR is regressed on the instrument variable and all control variables as in Model (2). In the second stage, Model (2) is estimated
after replacing ETR with its fitted value from the first-stage regression.
I use state-level religiosity and mean industry ETR as instrument variables in the 2SLS regression. Boone, Khurana, and Raman
(2013) find that religiosity is an important determinant of corporate tax avoidance. I measure religiosity as the fraction of population

10
N. Chaudhry North American Journal of Economics and Finance 58 (2021) 101488

Table 6
First-Difference Method. This table presents OLS re­
sults after first-differencing all variables. Δ represents
change in variable from year (t – 1) to year t. The
dependent variable is the change in idiosyncratic vola­
tility (IVOL), which is estimated as the standard devia­
tion of the residuals obtained by regressing excess stock
returns on the three Fama and French (1993) factors
multiplied by the number of trading days in a year, for
each stock and for each year. LNETR is the natural log of
the tax paid divided by the pretax income adjusted for
special items. LNLEV is the natural log of the long term
debt divided by total assets, FIRMSIZE is the natural log
of the book value of total assets, LNMB is the natural log
of the market value of equity to book value of equity
ratio, DIVIDEND is an indicator variable that is equal to
one if a firm pays dividend and zero otherwise, FIR­
MAGE is the number of years a firm exists in the Com­
pustat database, CF is cash flow from operating
activities scaled by total assets, CFσ is standard devia­
tion of CF, using data for the trailing five years, BETA is
market beta, STKRET is annual stock returns, and
STK_TRNOVR is average monthly trading volume to
shares outstanding over a year. All continuous variables
are winsorized at the 1st and 99th percentile. The
sample period is from 1992 to 2017. Statistical signifi­
cance at the one, five, and ten percent level is indicated
by ***, **, and *, respectively. t-Statistics (in paren­
theses) are calculated based on heteroscedasticity-
robust standard errors clustered by firm.
(1)

ΔLNETR − 0.002
(-1.790)*
ΔLNLEV 0.007
(6.366)***
ΔFIRMSIZE − 0.037
(-7.644)***
ΔLNMB − 0.068
(-18.366)***
ΔDIVIDEND − 0.013
(-2.823)***
ΔFIRMAGE − 0.003
(-6.142)***
ΔCF 0.011
(0.679)
ΔCFσ − 0.041
(-0.723)
ΔBETA − 0.011
(-3.574)***
ΔSTKRET 0.014
(6.510)***
ΔSTK_TRNOVR 0.065
(33.639)***
Constant − 0.007
(-6.816)***
N 18,095
R2 0.178

(in a state where firm is incorporated) that claims affiliation with an organized religious group. To calculate the religiosity measure,
data is obtained from the Churches and Church Membership files of the 1990, 2000, and 2010 surveys conducted by the Association of
Religion Data Archives. I use interpolation to get religiosity measure for all the years in my sample. The second instrument variable is
the mean industry ETR (MEAN_IND_ETR) obtained by taking all firms in the same three-digit SIC industry and excluding the firm itself.
It could be expected that the tax practices of peer firms from the same industry may influence the tax strategies adopted by a particular
firm but is unlikely to have any direct effect on the stock return idiosyncratic volatility. I conduct a number of diagnostic tests to check
the validity of the chosen instrument variables. The null hypothesis that the instruments are weak is rejected with the F-statistics of
35.431 significant at the one percent level. Wooldridge’s robust score test statistics of 0.126, which is statistically insignificant, confirm

11
N. Chaudhry North American Journal of Economics and Finance 58 (2021) 101488

Table 7
Endogeneity Tests. Column (1) presents results from the fixed-effect regression model. In Columns (1), (3), and (4), the dependent variable is
idiosyncratic volatility (IVOL), which is estimated as the standard deviation of the residuals obtained by regressing excess stock returns on the three
Fama and French (1993) factors multiplied by the number of trading days in a year, for each stock and for each year. LNETR is the natural log of the
tax paid divided by the pretax income adjusted for special items. LNLEV is the natural log of the long term debt divided by total assets, FIRMSIZE is the
natural log of the book value of total assets, LNMB is the natural log of the market value of equity to book value of equity ratio, DIVIDEND is an
indicator variable that is equal to one if a firm pays dividend and zero otherwise, FIRMAGE is the number of years a firm exists in the Compustat
database, CF is cash flow from operating activities scaled by total assets, CFσ is standard deviation of CF, using data for the trailing five years, BETA is
market beta, STKRET is annual stock returns, and STK_TRNOVR is average monthly trading volume to shares outstanding over a year. Columns (2) and
(3) report 2SLS regression results. Column (2) reports first-stage regression results in which dependent variable is LNETR, and RELIGIOSITY and
IND_MEAN_ETR are instrument variables. RELIGIOSITY is fraction of the state population that claims affiliation with a religious group based on the
surveys conducted by the Association of Religion Data Archives. MEAN_IND_ETR is the mean industry effective tax rate obtained by excluding the firm
itself. Column (3) reports results from the second-stage of the 2SLS regression. Column (4) presents results from the quasi-experiment. TREATMENT is
a dummy variable that equals one if a firm reports unrecognized tax benefit in year 2007 and zero for matched-control firms. POSTFIN48 is one for
year 2007 and zero for 2005. All continuous variables are winsorized at the 1st and 99th percentile. The sample period is from 1992 to 2017. Sta­
tistical significance at the one, five, and ten percent level is indicated by ***, **, and *, respectively. t-Statistics (in parentheses) are calculated based
on heteroscedasticity-robust standard errors clustered by firm.
(1) (2) (3) (4)

Panel A: Endogeneity Tests

Fixed-Effects Regression 2SLS Regression Quasi-Experiment
First-Stage Second-Stage
LNETR − 0.005 − 0.049
(-3.598)*** (-3.193)***
TREATMENT × POSTFIN48 0.043
(2.081)**
TREATMENT − 0.052
(-2.186)**
POSTFIN48 − 0.034
(-2.122)**
LNLEV 0.004 − 0.049 0.001 0.002
(3.678)*** (-6.840) (1.246) (0.347)
FIRMSIZE − 0.042 0.005 − 0.042 − 0.053
(-12.461)*** (0.54) (–32.810)*** (-6.506)***
LNMB − 0.008 0.026 − 0.014 − 0.022
(-2.462)** (1.390) (-5.519)*** (-1.048)
DIVIDEND − 0.026 0.229 − 0.033 − 0.021
(-5.740)*** (9.970) (-6.677)*** (-1.186)
FIRMAGE − 0.000 0.000 − 0.000 − 0.001
(-0.355) (0.610) (-1.720)* (-1.131)
CF − 0.150 − 0.599 − 0.191 − 0.169
(-7.425)*** (-3.74) (-8.441)*** (-1.242)
CFσ 0.112 − 1.374 0.372 0.532
(2.353)** (-4.330) (7.452)*** (1.809)*
BETA − 0.003 − 0.064 0.008 0.016
(-1.298) (-3.960) (3.275)*** (1.381)
STKRET 0.017 − 0.176 0.021 0.069
(7.745)*** (-10.100) (5.860)*** (3.230)***
STK_TRNOVR 0.014 − 0.028 0.019 0.028
(9.777)*** (-2.680) (15.011)*** (3.637)***
RELIGIOSITY 0.050 –
(1.700)
MEAN_IND_ETR 0.191
(8.350)
Constant 0.589 0.268 0.524 0.678
(21.975)*** (7.660) (14.394)*** (9.791)***
Year Effects
Industry Effects
Observations 20,474 18,714 17,938 196
R-squared 0.767 0.108 0.538 0.583
Panel B: Two-Stage Least Square Regression: Diagnostic Tests
Test of Endogeneity
H0 = Variables are exogenous
Robustregression F (1, 2855) = 7.879 (p = 0.005)

Test of Weak Instruments

H0 = Instruments are weak
F (2, 2855) = 35.431 (p = 0.000)

Test of Over-Identifying Restrictions

H0 = Instruments are valid
Score chi2 (1) = 0.126 (p = 0.723)

12
N. Chaudhry North American Journal of Economics and Finance 58 (2021) 101488

that the instruments are valid and model is specified correctly.

In Table 7, Columns (2) and (3) present the first-stage and second-stage results obtained from the 2SLS estimation, respectively. The
coefficients on both the instrument variables are statistically significant, confirming that the instruments are correlated with the
endogenous variable ETR. The coefficient on the fitted ETR is negative and statistically significant. These results further confirm that
the results reported earlier in this paper are robust to endogeneity.

6.3. Quasi-Experiment: Evidence from change in accounting rule

In this subsection I exploit a change in the accounting rule, which requires firms to disclose information related to their tax
strategies. Financial Interpretation No. 48 (FIN48) became effective for fiscal years beginning after December 15, 2006. This new rule
mandates firms to disclose unrecognized tax benefit (UTB) if there is more than 50 percent chance that benefit resulting from a tax
position will be disallowed by the tax authorities. FIN48 only affects tax aggressive firms (treatment firms) while other firms which do
not undertake risky tax strategies (control firms) do not have uncertain tax positions to disclose and therefore were not affected by the
new rule. FIN48 could cause an exogenous increase in the treatment firms’ UTB disclosures. This argument is consistent with Lisowsky
et al (2013) who document that disclosure of UTB reflect tax aggressiveness and Hasan et al (2014) who use similar approach to
address endogeneity of tax measures in their study.
I define dummy variable, TREATMENT, as one if a firm discloses a positive UTB in 2007 and zero for matched-control firm, which
does not report UTB in that year. To find matched-control firms, I estimate logistic regression model with TREATMENT as the
dependent variable and factors that can explain variations in the UTB levels as the independent variables. Rego and Wilson (2012) note
that profitability, firm size, foreign operations, research and development expenditure, discretionary accruals, financial leverage,
selling, general, and administrative expenses, market-to-book ratio and sales growth affect UTB levels. Using the predicted propensity
score from the logistic regression, I match each treatment firm with a control firm using the closest propensity score, without
replacement and within a caliper of 10 percent. I obtain a sample of 196 observations, which consist of 49 pairs of treatment and
matched control firms, which are present both before (2005) and after (2007) FIN48 became effective.
In Model (2), I replace ETR with TREATMENT, POST_FIN48, and an interaction term between the two, where POST_FIN48 is one for
year 2007 and zero for 2005. The coefficient on the interaction term measures the effect of the new accounting rule on idiosyncratic
volatility. It compares difference in average IVOL between treatment and control firms after FIN48 became effective relative to the
difference in average IVOL between treatment and control firms before FIN48 became effective. The coefficient on the interaction term
is positive and statistically significant as shown in Column (4) of Table 7. This result suggests that as a result of FIN48 implementation,
treatment firms have significantly higher increases in idiosyncratic volatility compared with matched-control firms. Investors interpret
UTB disclosures as indicative of riskiness of firms’ tax strategies, which introduces uncertainty in the valuation of treatment firms
resulting in the increased idiosyncratic stock return volatility. This finding could be viewed as providing evidence of a causal effect of
aggressive tax strategies on idiosyncratic volatility.
Overall, the results discussed in this section demonstrate that the main findings of this paper are robust to endogeneity. These
results confirm that aggressive tax strategies adopted by a firm cause idiosyncratic stock return volatility.

7. Additional tests

7.1. Nature of tax strategies

Tax strategies adopted by firms may vary from objectively legal strategies (e.g., investments in municipal bonds) to illegal stra­
tegies (e.g., tax shelters). Lisowsky et al (2013) describe “tax aggressiveness” as a subset of tax avoidance whereby underlying positions
likely have weak legal support, and they refer to the most extreme subset of tax aggressiveness as “tax sheltering,” which tests the
bounds of legality. The authors place various measures of tax avoidance commonly used in the literature on a continuum of least
aggressive to most aggressive tax positions. Both GAAP ETR (GETR) and book-tax difference (BTD) measures of tax avoidance capture
tax positions along the continuum and may reflect benign tax positions such as differences in book and tax depreciation methods.6
These tax measures do not explicitly capture tax sheltering and indeed to some extent perfectly capture legitimate tax positions. They
also do not capture conforming tax positions, whose effects on taxable and financial income are the same. Discretionary permanent
book-tax differences (DTAX) capture risky tax positions (Rego and Wilson 2012). DTAX is considered to represent a subset of tax
aggressive strategies, as it measures permanent book-tax difference, which cannot be explained by legitimate tax positions (Frank,
Lynch, and Rego 2009). Unrecognized tax benefits (UTB) are a contingent liability reflecting the dollar amount of tax benefits related
to all open tax positions that may ultimately be disallowed. Firms started disclosing UTB in footnotes from 2006. Lisowsky et al (2013)
find that UTB to increase with uncertainty about a tax position and to be positively associated with the use of tax shelters.
In addition to annual GETR, I calculate three-year and five-year GETR (GETR3 and GETR5) and replace ETR in regression Model (2)
with GETR measures BTD, DTAX and UTB. The results are reported in Table 8. Columns (1)-(3) show that the coefficient on GETR
variables is negative and significant, consistent with my findings that a lower tax rate is associated with higher IVOL levels. The
coefficient on BTD is positive and significant, suggesting that less taxable income than book income is associated with more

6
GAAP ETR is calculated by dividing tax expenses (accrued) by pretax income, which is adjusted for special items. BTD is the difference between
book and taxable income. Lower values of GETR indicate more tax saving by a firm and increasing BTD levels indicate tax aggressiveness.

13
N. Chaudhry North American Journal of Economics and Finance 58 (2021) 101488

idiosyncratic volatility. However, I do not observe DTAX and UTB to be significantly related to idiosyncratic volatility. These results
suggest that firms adopting aggressive tax strategies have more volatile stock returns, but these strategies do not involve illegal
transactions, such as those involved in tax shelters.

7.2. Long-Term Effects

Aggressive tax strategies may result in increased future taxes, penalties, and fines from the IRS, thus distorting a firm’s future
profitability. The IRS takes, on average, four years to audit tax returns filed by large corporations and an additional two–three years to
resolve any appeals before final tax assessments can be determined (White (2013)). This lengthy audit process leads to years of un­
certainty about firms’ tax liabilities. Thus, obtaining certainty about taxes filed by a firm for any given year may take up to six or seven
years. I investigate how far in the future tax strategies affect idiosyncratic volatility by lagging all independent variables in Model (2)
by three, five, seven, and ten years. The results are presented in Table 9.
The coefficients on ETR are − 0.007, − 0.007, − 0.007, and − 0.004 (significant at the five percent level or better) when independent
variables are lagged by three, five, seven, and ten years, respectively. These findings indicate that tax strategies affect firms’ future
idiosyncratic volatility. This evidence is consistent with the prediction that most uncertainty about tax claims is resolved by roughly
the end of the seventh year.

7.3. Role of financial constraints

Firms subject to financial constraints have limited financing sources. These firms are most affected by decreased cash flows
resulting from fines, penalties, interest, and tax payments made to the IRS when claimed tax benefits are rejected by taxation au­
thorities. Compared to financially-unconstrained firms, financially-constrained firms rely more on tax strategies to save cash and use
such approaches for their business activities (Edwards et al, 2016). A failure to retain benefits claimed under a tax strategy may have
more severe effects on the cash flows of a financially-constrained firms than on those of financially-unconstrained firms. A tax strategy
is expected to influence the idiosyncratic volatility of a financially-constrained firm more than that of a financially-unconstrained firm.
I examine the relation between effective tax rates and idiosyncratic volatility by keeping the value of financial constraints constant. For
this purpose, I divide the sample into financially-constrained and financially-unconstrained firms and estimate Model (2) separately for
these two subsamples. Doing so allows us to vary the slope of all variables and to account for the differential nature of financially
constrained and financially unconstrained firms.
I use five measures of financial constraints. The first measure is based on whether a firm pays dividends. Dividend-paying firms can
counteract the effects of an unexpected cash flow shock on their business activities by reducing dividend payments. Therefore,
dividend-paying firms are expected to be more financially-unconstrained than non-dividend paying firms. I partition the full sample

Table 8
Nature of Tax Strategies. This table presents OLS results, where dependent variable is idiosyncratic volatility (IVOL), which is estimated as the
standard deviation of the residuals obtained by regressing excess stock returns on the three Fama and French (1993) factors multiplied by the number
of trading days in a year, for each stock and for each year. LNGETR is the natural log of the accrued tax expense divided by the pretax income adjusted
for special items. LNGETR3 (LNGETR5) is the natural log of the sum of accrued tax expense over a three (five) year period divided by the pretax
income (adjusted for special items) over the same period. BTD is the difference between book income and taxable income, where taxable income is
calculated by grossing up the sum of the current federal tax expense and the current foreign tax expense and subtracting tax-loss carry forward; if the
current federal tax expense is missing then total current tax expense is calculated by subtracting deferred taxes, state income taxes, and other income
taxes from total income taxes. DTAX is discretionary permanent book-tax difference developed by Frank, Lynch, and Rego (2009). UTB is unrec­
ognized tax benefit. LNBTD, LNDTAX, and LNUTB, are natural log of BTD, DTAX, and UTB all scaled by total assets. All control variables are same as in
the main analysis. All continuous variables are winsorized at the 1st and 99th percentile. The sample period is from 1992 to 2017. Statistical sig­
nificance at the one, five, and ten percent level is indicated by ***, **, and *, respectively. t-Statistics (in parentheses) are calculated based on
heteroscedasticity-robust standard errors clustered by firm.
(1) (2) (3) (4) (5) (6)

LNGETR − 0.014
(-6.442)***
LNGETR3 − 0.019
(-5.148)***
LNGETR5 − 0.023
(-3.528)***
LNBTD 0.002
(1.912)*
LNDTAX − 0.000
(-0.033)
LNUTB − 0.001
(-0.480)
N 20,175 16,218 10,372 8,069 4,728 5,081
R2 0.569 0.569 0.562 0.585 0.582 0.555

14
N. Chaudhry North American Journal of Economics and Finance 58 (2021) 101488

Table 9
Long-Lasting Effect. This table presents OLS results, where dependent variable is idiosyncratic volatility (IVOL), which is estimated as the standard
deviation of the residuals obtained by regressing excess stock returns on the three Fama and French (1993) factors multiplied by the number of trading
days in a year, for each stock and for each year. LNETR is the natural log of the tax paid divided by the pretax income adjusted for special items. LNLEV
is the natural log of the long term debt divided by total assets, FIRMSIZE is the natural log of the book value of total assets, LNMB is the natural log of
the market value of equity to book value of equity ratio, DIVIDEND is an indicator variable that is equal to one if a firm pays dividend and zero
otherwise, FIRMAGE is the number of years a firm exists in the Compustat database, CF is cash flow from operating activities scaled by total assets, CFσ
is standard deviation of CF, using data for the trailing five years, BETA is market beta, STKRET is annual stock returns, and STK_TRNOVR is average
monthly trading volume to shares outstanding over a year. All continuous variables are winsorized at the 1st and 99th percentile. The sample period is
from 1992 to 2017. Statistical significance at the one, five, and ten percent level is indicated by ***, **, and *, respectively. t-Statistics (in parentheses)
are calculated based on heteroscedasticity-robust standard errors clustered by firm.
(1) (2) (3) (4)
3-Year Lagged independent 5-Year Lagged independent 7-Year Lagged independent 10-Year Lagged independent
variables variables variables variables

LNETR -0.007 -0.007 -0.007 -0.004

(-4.535)*** (-4.509)*** (-4.153)*** (-2.269)**
LNLEV 0.003 0.003 0.003 0.006
(2.569)** (3.144)*** (2.553)** (4.452)***
FIRMSIZE -0.034 -0.031 -0.030 -0.028
(-26.476)*** (-24.229)*** (-22.785)*** (-18.533)***
LNMB -0.019 -0.022 -0.024 -0.027
(-7.729)*** (-8.266)*** (-9.079)*** (-8.377)***
DIVIDEND -0.032 -0.026 -0.017 -0.012
(-9.011)*** (-6.872)*** (-4.437)*** (-2.721)***
FIRMAGE -0.000 -0.000 -0.000 0.000
(-2.160)** (-1.469) (-0.451) (0.133)
CF -0.132 -0.114 -0.134 -0.085
(-6.199)*** (-5.433)*** (-5.587)*** (-3.146)***
CFσ 0.330 0.307 0.241 0.365
(7.269)*** (6.534)*** (5.141)*** (5.763)***
BETA 0.011 0.007 0.004 -0.000
(4.162)*** (2.553)** (1.328) (-0.030)
STKRET 0.007 0.006 0.004 -0.001
(3.063)*** (2.382)** (1.459) (-0.391)
STK_TRNOVR 0.010 0.007 0.006 0.007
(8.222)*** (5.070)*** (4.380)*** (3.960)***
Constant 0.514 0.560 0.661 0.454
(21.843)*** (24.890)*** (33.394)*** (18.433)***
Year Effects Yes Yes Yes Yes
Industry Effects Yes Yes Yes Yes
N 13,143 Yes Yes Yes
R2 0.555 0.534 0.491 0.479

into dividend-paying and non-dividend paying firms and estimate Model (2) separately for the two groups. The results, which are
presented in Table 10 Panel A, show that the effect of ETR on IVOL is stronger for financially-constrained firms. I obtain similar results
when I classify a firm as financially-constrained when it does not have a long-term debt rating and as financially-unconstrained when it
has a long-term debt rating.
In addition, I use measures developed by Kaplan and Zingales (1997), Whited and Wu (2006), and Hadlock and Pierce (2010) to
assess whether a firm is financially constrained. A larger value for these measures indicates a higher degree of financial constraint. For
each of these proxies, I divide the sample into financially constrained and financially unconstrained firms and estimate Model (2)
separately for each of the two subsamples. I obtain similar results when using these alternative proxies of financial constraints. The
results reveal a potential mechanism (overcoming financing constraints) through which tax aggressiveness increases firm volatility.

7.4. Role of corporate governance

Managers tend to invest more than what is optimal for a firm for personal gain at the expense of shareholders when monitoring
mechanisms are weak (Jensen 1986). Free cash leads to agency problems as managers get access to funds and avoid monitoring by
capital markets (Jensen, 1986; Harford, 1999). High quality information can deter opportunistic managerial behavior by increasing
transparency and facilitating more efficient contracting to prevent value-destroying corporate decisions (Jensen and Meckling 1976).
Managers are more likely to waste resources especially when the information environment is of poor quality.

15
N. Chaudhry North American Journal of Economics and Finance 58 (2021) 101488

Table 10
Role of Financial Constraints and Corporate Governance. This table presents OLS results, where dependent variable is idiosyncratic volatility
(IVOL), which is estimated as the standard deviation of the residuals obtained by regressing excess stock returns on the three Fama and French (1993)
factors multiplied by the number of trading days in a year, for each stock and for each year. LNETR is the natural log of the tax paid divided by the
pretax income adjusted for special items. All control variables are same as in the main analysis. All continuous variables are winsorized at the 1st and
99th percentile. Full sample is divided into two subsamples based on different measures of financial constraints and corporate governance. The
sample period is from 1992 to 2017. Statistical significance at the one, five, and ten percent level is indicated by ***, **, and *, respectively. t-Statistics
(in parentheses) are calculated based on heteroscedasticity-robust standard errors clustered by firm.
(1) (2)

Panel A: The sample is divided into financially constrained and financially unconstrained firms using various measures of financial constraints.
Non-Dividend Paying Firms Dividend-Paying Firms
LNETR − 0.008 − 0.005
(-4.973)*** (-3.295)***
Firms with no long-term rated debt Firms with long-term rated debt
LNETR − 0.008 − 0.007
(-4.825)*** (-4.661)***
Financially constrained using Kaplan and Zingales (1997) index Financially unconstrained using Kaplan and Zingales (1997) index
LNETR − 0.009 − 0.006
(-5.652)*** (-3.275)***
Financially constrained using Whited and Wu (2006) index Financially unconstrained using Whited and Wu (2006) index
LNETR − 0.010 − 0.006
(-5.016)*** (-4.517)***
Financially constrained using Hadlock and Pierce (2010) index Financially unconstrained using Hadlock and Pierce (2010) index
LNETR − 0.010 − 0.006
(-4.807)*** (-4.357)***
Panel B: The sample is divided into two subsamples based on the quality of corporate governance and information environment surrounding a firm.
Less independent board More independent board
LNETR − 0.007 − 0.006
(-3.009)*** (-2.760)***
Less institutional ownership More institutional ownership
LNETR − 0.008 − 0.005
(-4.164)*** (-3.850)***
Less industry competition More industry competition
LNETR − 0.010 − 0.006
(-5.858)*** (-3.470)***
Small firms Large firms
LNETR − 0.008 − 0.006
(-4.262)*** (-4.671)***

A reduction in a firm’s explicit tax liability increases a firm’s after-tax cash flow and therefore increases the internal resources
available to managers. Tax savings are often substantial and represent resources that may facilitate empire building. For example,
Lisowsky et al (2013) estimate that the reportable transactions of 48 firms reduced taxable income in 2007 by $10.7 billion, repre­
senting tax savings of $3.7 billion. In addition to increasing the internal resources of a firm, tax avoidance intensifies information
asymmetry between managers and investors (Balakrishnan, Blouin, and Guay, 2019; Desai and Dharmapala, 2009). Tax avoidance
introduces opacity by facilitating the hoarding and accumulation of bad news (Kim et al 2011). Managers may exploit an increase in
information asymmetry from tax avoidance and use cash savings from tax avoidance to invest beyond what is optimal for shareholders.
Due to increased agency and information risk, I expect tax strategies to introduce more uncertainty into stock prices for firms with
weak corporate governance and for those operating in poor information environments. I test this prediction by partitioning the sample
based on the quality of corporate governance and information environments surrounding a firm and analyze Model (2) separately for
each of the subsamples. I use board independence, institutional ownership, industry competitiveness, and firm size as proxies for the
quality of corporate governance and information. I estimate board independence (BINDEP) as the ratio of the number of independent
board members to the total number of board directors. To estimate institutional ownership, I summarize the number of shares held by
institutional investors for each quarter and divide this sum by the total number of outstanding shares. The average of these quarterly
estimates for a year is used as a measure of institutional ownership.
A competitive product market may have a disciplining effect on managers and serve as a substitute for high-powered incentives
(Panousi and Papanikolaou 2012). Kubick, Lynch, Mayberry, and Omer (2015) argue that cash flow shocks due to penalties/fines and
increased tax payments resulting from an aggressive tax strategy restrict firms operating in competitive environments from investing in
new projects or from developing new and better products than their competitors, thus affecting their profitability even more. This
situation is further aggravated when news about a firm’s involvement in tax avoidance generates negative publicity about that firm.

16
N. Chaudhry North American Journal of Economics and Finance 58 (2021) 101488

Consumers may then boycott the firm and shift their demand to their competitors within an industry (The Week 2012). I estimate the
level of competition in an industry by taking the reciprocal of the Herfindahl-Hirschman index (HHI), which is the sum of squares of the
market share of each firm in a two-digit SIC code industry.
The results reported in Panel B Table 10 show that a decrease in ETR is associated with a more positive change in idiosyncratic
volatility when a firm experiences weak corporate governance and a poor information environment. That is, when boards are less
independent, levels of institutional ownership are high, firms belong to competitive industries, and firms are large, tax strategies are
associated with more volatile stock returns.

7.5. Alternative explanations

I investigate alternative explanations by examining how the ETR-IVOL relation varies with managerial ownership and risk-taking
incentives, industry affiliations, foreign operations, financial leverage, and varied tax strategies.
When managers own a substantial stake in their firms, they are reluctant to invest in projects that increase idiosyncratic risk, as
managers can hedge exposure to systematic risk but cannot reduce their exposure to idiosyncratic risk (Panousi and Papanikolaou
2012). The negative effect of managerial risk aversion on investment when using risky tax strategies can be mitigated when a CEO is

Table 11
Additional Tests. This table presents OLS results, where dependent variable is idiosyncratic volatility (IVOL), which is estimated as the standard
deviation of the residuals obtained by regressing excess stock returns on the three Fama and French (1993) factors multiplied by the number of trading
days in a year, for each stock and for each year. ETR is the natural log of the tax paid divided by the pretax income adjusted for special items. All
control variables are same as in the main analysis. All continuous variables are winsorized at the 1st and 99th percentile. CEOOWN is one if the
percentage of shares held by the CEO are greater than the sample median, CEODELTA and CEOVEGA measure sensitivity of CEO’s wealth to stock
price and stock price volatility, HITECH is one if a firm belongs to high-technology sector and zero otherwise, MNC is one if a firm has foreign
operations and zero otherwise, LEV is one if the leverage is more than the sample median and zero otherwise, and CFσ is one if ETR volatility is greater
than the sample median and zero otherwise, where ETR volatility is measured as the standard deviation of CF, using data for the trailing five years.
The sample period is from 1992 to 2017. Statistical significance at the one, five, and ten percent level is indicated by ***, **, and *, respectively. t-
Statistics (in parentheses) are calculated based on heteroscedasticity-robust standard errors clustered by firm.
(1) (2) (3) (4) (5) (6) (7)
CEO Ownership CEO Delta CEO Vega Technology Firms Multinational Corporations Leverage CETR Volatility

LNETR − 0.007 − 0.008 − 0.008 − 0.008 − 0.011

(-2.743)*** (-6.064)*** (-4.850)*** (-5.292)*** (-3.931)***
LNETR × CEOOWN 0.013
(2.199)**
CEOOWN 0.002
(0.650)
LNETR × CEODELTA − 0.003
(-1.050)
CEODELTA − 0.011
(-1.958)*
LNETR × CEOVEGA 0.004
(1.131)
CEOVEGA − 0.008
(-1.355)
LNETR × HITECH 0.028
(2.996)***
HITECH 0.001
(0.332)
LNETR × MNC − 0.008
(-1.531)
MNC 0.000
(0.139)
LNETR × LEV 0.010
(2.305)**
LEV 0.000
(0.182)
LNETR × ETRσ 0.015
(2.659)***
LNETRσ 0.006
(1.788)*
Control Variables Yes Yes Yes Yes Yes
Constant Yes Yes Yes Yes Yes
Year Effects Yes Yes Yes Yes Yes
Industry Effects Yes Yes Yes Yes Yes
N 6,251 20,474 20,474 25,900 10,400
R2 0.636 0.571 0.570 0.562 0.560

17
N. Chaudhry North American Journal of Economics and Finance 58 (2021) 101488

compensated with options rather than with shares, as the value of an option increases with both stock prices and stock price volatility
(Phillips, 2003; Rego and Wilson, 2012). The sensitivity of a CEO’s wealth to stock prices (CEO delta) encourages a CEO to take risks
that are expected to generate a sufficient increase in stock prices. I test whether the observed relationship between ETR and IVOL is
driven by manager ownership and risk-taking incentives.
Following Core and Guay (2002), I define CEO delta as the dollar change in a CEO’s equity portfolio value with a one percent
increase in the stock price and CEO vega as the dollar change in a CEO’s equity portfolio value with a one percent increase in stock
return volatility. I define a dummy variable indicating whether CEO ownership in a firm is more/less than the sample median.
Similarly, I define dummy variables indicating risk incentives (CEO delta and CEO vega). I introduce an interaction term between ETR
and dummy variables for CEO ownership and risk-taking in regression Model (2). The results are presented in Columns (1)-(3) of
Table 11. The coefficient on the tax variable (LNCETR) is significant in all columns. The coefficient on the interaction term is not
significant in any of the three regressions, suggesting that CEO ownership and risk-taking incentives do not seem to have an incre­
mental effect on the relation between ETR and IVOL.
Schwert (2002) shows that stocks listed on the NASDAQ exchange are more volatile (compared with S&P 500 firms) and that this
high volatility is driven by firms belonging to the technology sector (Computer, Biotechnology, and Telecommunications). The author
argues that in recent years, economic growth has concentrated in the technology sector and that any negative news about future
growth has much stronger effects on these stocks. To confirm that the main results are not driven by technology firms, I define a
variable (HITECH) indicating whether a firm belongs to the technology sector and introduce an interaction term between HITECH and
ETR in regression Model (2). The results presented in Column (4) show that the sign and significance of the coefficient on ETR are
consistent with my main results. Idiosyncratic volatility is higher for technology firms than for non-technology firms. However, the
interaction term, HITECH × ETR, is not significant, suggesting that stock return volatility introduced as a result of aggressive tax
strategies does not vary across technology and non-technology firms.
Multinational corporations have more opportunities to save taxes, opportunities which are not available to firms with only do­
mestic operations (domestic-only firms). For example, multinationals can avoid taxes by locating operations/income in countries with
low tax rates or by exploiting differences in the tax rules of different countries. To check whether multinational firms drive the main
findings, in the regression model, I introduce an interaction term between ETR and a dummy variable, MNC, which is equal to one for
multinational firms and equal to zero for domestic-only firms. The results, which are reported in Column (5), show that both the
coefficients on MNC and the interaction term are insignificant. This result does not suggest that there are any significant differences
between multinationals and domestic-only firms in terms of their effects on idiosyncratic volatility directly or through their tax
strategies.
To test whether leverage magnifies the effect of ETR on idiosyncratic volatility, I calculate leverage as the ratio of long-term debt to
the book value of total assets and partition the sample based on median sample leverage. The coefficient on ETR has the expected
negative sign and is significant at the one percent level while leverage is positively associated with idiosyncratic volatility. However, I
do not find that the relation between ETR and IVOL varies across low and high levered firms.
Finally, I test whether the relation between ETR and idiosyncratic volatility is stronger for firms with volatile tax rates than for firms
with less volatile tax rates. Volatile tax rates might be suggestive of tax strategies that increase the risk of cash flows being more
uncertain. Firms with volatile tax rates are likely to suffer shortfalls in cash flows related to their tax strategies and are likely to rely less
on tax avoidance to fund their business activities. I measure the volatility of tax avoidance strategies (ETRσ ) by taking the standard
deviation of annual cash ETRs based on the past five years of data. I interact ETR with the volatility of tax avoidance in the regression
model and report these results in the last column. The coefficients on ETR remains negative and significant at the one percent level.
ETRσ is positively associated with idiosyncratic volatility. A positive coefficient on the interaction term is consistent with the
conjecture that firms with more volatile tax avoidance exhibit more uncertainty about future cash flows and therefore high levels of
idiosyncratic volatility.

8. Conclusion

This paper finds strong and consistent evidence in support of the hypothesis that effective tax rates influence the volatility of stock
returns. It demonstrates that as effective tax rates drop, idiosyncratic stock return volatility increases. These results are robust to
various sensitivity checks. The tax strategies of a firm affect stock return idiosyncratic volatility over long periods up to 7–10 years. The
relation between tax strategies and idiosyncratic volatility is stronger when firms are financially constrained, have weak corporate
governance and operate in poor information environments. I find that levels of idiosyncratic volatility are high for firms in which levels
of CEO ownership are high, for firms belonging to the technology sector, and for firms with high degrees of financial leverage. Variance
in tax strategies leads to greater increases in idiosyncratic volatility with lower effective tax rates. The findings documented in this
work will be of use to investors, managers as well as policy makers.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to
influence the work reported in this paper.

18
N. Chaudhry North American Journal of Economics and Finance 58 (2021) 101488

Appendix. Variable Definitions

Compustat variables are shown in BOLD and estimated variables are shown in ITALICS
Variable Definition

BETA Market beta obtained from the Fama and French (1993) model using past 60 months of monthly returns data
BTD/LNBTD Difference between book income and taxable income, where taxable income is calculated by grossing up the sum of the current
federal tax expense and the current foreign tax expense and subtracting tax-loss carry forward; if the current federal tax expense is
missing then total current tax expense is calculated by subtracting deferred taxes, state income taxes, and other income taxes from
total income taxes
LNBTD is the natural log of BTD scaled by total assets
CEOOWN Percentage of shares held by CEO
CEODELTA Sensitivity of CEO wealth to stock price estimated following Guay (1999)
CEOVEGA Sensitivity of CEO wealth to stock price volatility estimated following Guay (1999)
CF Cashflow from operating activities (OANCF) scaled by total assets (AT)
CFσ Standard deviation of cash flow (CF), using data for the trailing five years
Competition Reciprocal of Herfindahl-Hirschman index, which is estimated by summing squares of the market shares of firms in the industry (at
the 3-digit SIC level)
Credit Rating Indicator variable that is equal to one for firms with public rated debt and zero otherwise
DIVIDEND Indicator variable that is equal to one if a firm pays dividend (DVC > 0) and zero otherwise (DVC = 0 or DVC is missing)
DNOL Change in tax loss carryforward (TLCF) divided by total assets (AT)
DTAX Discretionary permanent book-tax difference developed by Frank, Lynch, and Rego (2009)
ETR/LNETR Tax paid (TXPD) divided by pretax income (PI) adjusted for special items (SPI)
LNETR is the natural log of ETR
ETR3/ETR5 Tax paid (TXPD) over leading three/five years divided by pretax income (PI) adjusted for special items (SPI) over the same period
LNETR3/LNETR5 LNETR3/LNETR5 is the natural log of LNETR3/LNETR5
ETRσ One if standard deviation of ETR (using past five years of values) is more than the sample median ETR volatility and zero otherwise
FIRMAGE Number of years for which a firm exists in the Compustat database
FIRMSIZE Natural log of the book value of total assets (AT)
GETR/LNGETR Reported tax expense (TXT) divided by pretax income (PI) adjusted for special items (SPI)
LNGETR is the natural log of GETR
GETR3/GETR5 Tax expense (TXT) over leading three/five years divided by pretax income (PI) adjusted for special items (SPI) over the same period
LNGETR3/LNGETR5 LNGETR3/LNGETR5 is the natural log of LNGETR3/LNGETR5
HITECH One if a firm belongs to high-technology sector (Computer, Biotechnology, and Telecommunications) and zero otherwise
IND_SIZE_ETR Obtained after sorting firms within an industry first on firm size and then within each size quartile on ETR; IND_SIZE_ETR is defined as
one if a firm belongs to first quartile of ETR and zero if firm belongs to the fourth quartile.
Institutional Shares held by institutional investors divided by shares outstanding at the end of each quarter. I take average of this quarterly
Ownership estimates
IVOL Standard deviation of the residuals obtained from the Fama and French (1993) model multiplied by the square root of the number of
LAGGED_IVOL trading days in a year
LAGGED_IVOL is one-year lagged IVOL
LEV/LNLEV Long term debt (DLTT) divided by total assets (AT)
LNLEV is the natural log of LEV
MB/LNMB Market value of equity (CSHO × PRCC_F) divided by book value of equity (CEQ)
LNMB is the natural log of MB
MEAN_IND_ETR Mean industry (3-digit SIC) ETR obtained by excluding the firm itself
MNC One if firm reports foreign-income (PIFO > 0) and zero otherwise
NOL Indicator variable if tax-loss-carryforward is positive (TLCF > 0) and zero otherwise
RELIGIOSITY Fraction of the state population that claims affiliation with a religious group based on the surveys conducted by the Association of
Religion Data Archives
ROA Net income (NI) divided by total assets (AT)
STKRET Annual stock returns
STK_TRNOVR Average of monthly trading volume to shares outstanding over a year
Technology Firms Firms belonging to industries - Computers, Biotechnology and Telecommunications
LNUTB Natural log of unrecognized tax benefit (TXTUBEND)
Whited and Wu (2006) -0.091OANCF/AT – 0.062DIV + 0.021DLTT/AT – 0.044ln(AT) + 0.102INDGROWTH – 0.025SGRTH
Index where, INDGROWTH is the industry sales growth rate, SGRTH is the firm’s annual sales growth rate, and DIV as described earlier in
(WW) this table

References

Armstrong, C. S., Blouin, J. L., & Larcker, D. F. (2012). The incentives for tax planning. Journal of Accounting and Economics, 53(1-2), 391–411.
Armstrong, C. S., Blouin, J. L., Jagolinzer, A. D., & Larcker, D. F. (2015). Corporate governance, incentives, and tax avoidance. Journal of Accounting and Economics, 60
(1), 1–17.
Badertscher, B. A., Katz, S. P., & Rego, S. O. (2013). The separation of ownership and control and corporate tax avoidance. Journal of Accounting and Economics, 56(2-
3), 228–250.
Balakrishnan, K., Blouin, J. L., & Guay, W. R. (2019). Tax aggressiveness and corporate transparency. The Accounting Review, 94(1), 45–69.
BECK, THORSTEN, LIN, CHEN, & MA, YUE (2014). Why do firms evade taxes? The role of information sharing and financial sector outreach. The Journal of Finance, 69
(2), 763–817.

19
N. Chaudhry North American Journal of Economics and Finance 58 (2021) 101488

Bergin, T. (2016). U.S. plans to curb tax ‘inversions’ could hit foreign companies. (April 11) Available at: http://www.reuters.com/article/us-usa-tax-inversions-
subsidiaries-idUSKCN0X81P4.
Boone, J. P., Khurana, I. K., & Raman, K. K. (2013). Religiosity and tax avoidance. Journal of the American Taxation Association, 35(1), 53–84.
Carhart, M. M. (1997). On persistence in mutual fund performance. The Journal of Finance, 52(1), 57–82.
Chen, S., Chen, X., Cheng, Q., & Shevlin, T. (2010). Are family firms more tax aggressive than non-family firms? Journal of Financial Economics, 95(1), 41–61.
Cheng, C. S. A., Huang, H. H., Li, Y., & Stanfield, J. (2012). The effect of hedge fund activism on corporate tax avoidance. The Accounting Review, 87(5), 1493–1526.
Chi, S., Huang, S. X., & Sanchez, J. M. (2017). CEO inside debt incentives and corporate tax sheltering. Journal of Accounting Research, 55(4), 837–876.
Chyz, J. A., Ching Leung, W. S., Zhen Li, O., & Meng Rui, O. (2013). Labor unions and tax aggressiveness. Journal of Financial Economics, 108(3), 675–698.
Core, J., & Guay, W. (2002). Estimating the value of employee stock option portfolios and their sensitivities to price and volatility. Journal of Accounting Research, 40
(3), 613–630.
Desai, M. A., & Dharmapala, D. (2006). Corporate tax avoidance and high-powered incentives. Journal of Financial Economics, 79(1), 145–179.
Desai, M. A., & Dharmapala, D. (2009). Corporate tax avoidance and firm value. Review of Economics and Statistics, 91(3), 537–546.
Dyreng, S. D., Hanlon, M., & Maydew, E. L. (2008). Long-run corporate tax avoidance. The Accounting Review, 83(1), 61–82.
Dyreng, S. D., Hanlon, M., & Maydew, E. L. (2010). The effects of executives on corporate tax avoidance. The Accounting Review, 85(4), 1163–1189.
Dyreng, S. D., Hanlon, M., Maydew, E. L., & Thornock, J. R. (2017). Changes in corporate effective tax rates over the past 25 years. Journal of Financial Economics, 124
(3), 441–463.
Dyreng, S. D., Hoopes, J. L., & Wilde, J. H. (2016). Public pressure and corporate tax behavior. Journal of Accounting Research, 54(1), 147–186.
Edwards, A., Schwab, C., & Shevlin, T. (2016). Financial constraints and cash tax savings. The Accounting Review, 91(3), 859–881.
Fama, E. F., & French, K. R. (1993). Common risk factors in the returns on stocks and bonds. Journal of Financial Economics, 33(1), 3–56.
Fama, E. F., & MacBeth, J. D. (1973). Risk, return, and equilibrium: Empirical tests. The Journal of Political Economy, 81(3), 607–636.
Frank, M., Lynch, L. J., & Rego, S. O. (2009). Tax reporting aggressiveness and its relation to aggressive financial reporting. The Accounting Review, 84(2), 467–496.
Fu, F. (2009). Idiosyncratic risk and the cross-section of expected stock returns. Journal of Financial Economics, 91(1), 24–37.
Gallemore, J., Gipper, B., & Maydew, E. (2019). Banks as Tax Planning Intermediaries. Journal of Accounting Research, 57(1), 169–209.
Gallemore, J., & Labro, E. (2015). The importance of the internal information environment for tax avoidance. Journal of Accounting and Economics, 60(1), 149–167.
Gaspar, J.-M., & Massa, M. (2006). Idiosyncratic volatility and product market competition. Journal of Business, 79(6), 3125–3152.
Goyal, A., & Santa-Clara, P. (2003). Idiosyncratic risk matters. The Journal of Finance, 58(3), 975–1007.
Graham, J. R., & Tucker, A. L. (2006). Tax shelters and corporate debt policy. Journal of Financial Economics, 81(3), 563–594.
Graham, J. R., Hanlon, M., Shevlin, T., & Shroff, N. (2014). Incentives for tax planning and avoidance: Evidence from the field. The Accounting Review, 89(3),
991–1023.
Guay, W. R. (1999). The sensitivity of CEO wealth to equity risk: An analysis of the magnitude and determinants. Journal of Financial Economics, 53(1), 43–71.
Gupta, S., & Newberry, K. (1997). Long-run corporate tax avoidance. The Accounting Review, 83(1), 61–82.
Hadlock, C. J., & Pierce, J. R. (2010). New evidence on measuring financial constraints: moving beyond the KZ index. The Review of Financial Studies, 23(5),
1909–1940.
Hanlon, M., & Slemrod, J. (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.
Harford, J. (1999). Corporate cash reserves and acquisitions. Journal of Finance, 54(6), 1969–1997.
Hasan, I., Hoi, C. K., Wu, Q., & Zhang, H. (2014). Beauty is in the eye of the beholder: The effect of corporate tax avoidance on the cost of bank loans. Journal of
Financial Economics, 113(1), 109–130.
Hasan, I., Hoi, C. K., Wu, Q., & Zhang, H. (2017). Does social capital matter in corporate decisions? Evidence from corporate tax avoidance. Journal of Accounting
Research, 55(3), 629–668.
Hope, O.-K., Ma, M.(., & Thomas, W. B. (2013). Tax avoidance and geographic earnings disclosure. Journal of Accounting and Economics, 56(2-3), 170–189.
Jensen, M. C. (1986). Agency costs of free cash flow, corporate finance and takeovers. American Economic Review, 76(2), 323–329.
Jensen, M. C., & Meckling, W. H. (1976). Theory of the firm: Managerial behaviour, agency costs and ownership structure. Journal of Financial Economics, 3, 305–360.
Kaplan, S. N., & Zingales, L. (1997). Do investment-cash flow sensitivities provide useful measures of financing constraints? Quarterly Journal of Economics, 112(1),
169–215.
Kim, J.-B., Li, Y., & Zhang, L. (2011). Corporate tax avoidance and stock price crash risk: Firm-level analysis. Journal of Financial Economics, 100(3), 639–662.
Kubick, T. R., Lynch, D. P., Mayberry, M. A., & Omer, T. C. (2015). Product market power and tax avoidance: Market leaders, mimicking strategies, and stock returns.
The Accounting Review, 90(2), 675–702.
Law, K. K. F., & Mills, L. F. (2015). Taxes and financial constraints: evidence from linguistic cues. Journal of Accounting Research, 53(4), 777–819.
Lisowsky, P., L. A. Robinson, and A. Schmidt (2013). Do publicly disclosed tax reserves tell us about privately disclosed tax shelter activity? Journal of Accounting
Research, 51, 583-629.
McGill, G. A., & Outslay, E. (2004). Lost in translation: Detecting tax shelter activity in financial statements. National Tax Journal, 57(3), 739–756.
McGuire, S. T., Omer, T. C., & Wang, D. (2012). Tax avoidance: does tax-specific industry expertise make a difference? The Accounting Review, 87(3), 975–1003.
Mills, L. F. (1998). Book-tax differences and internal revenue service adjustments. Journal of Accounting Research, 36(2), 343–356.
Mills, L., Erickson, M. L., & Maydew, E. L. (1998). Investments in tax planning. The Journal of the American Taxation Association, 20(1), 1–20.
Mills, L. F., Nutter, S. E., & Schwab, C. M. (2013). The effect of political sensitivity and bargaining power on taxes: Evidence from federal contractors. The Accounting
Review, 88(3), 977–1005.
Modigliani, F., & Miller, M. H. (1958). The cost of capital, corporation finance and the theory of investment. American Economic Review, 48(3), 261–297.
Panousi, V., & Papanikolaou, D. (2012). Investment, idiosyncratic risk, and ownership. The Journal of Finance, 67(3), 1113–1148.
Pástor, Ľ., & Pietro, V. (2003). Stock valuation and learning about profitability. The Journal of Finance, 58(5), 1749–1789.
Petersen, M. A. (2009). Estimating standard errors in finance panel data sets: Comparing approaches. The Review of Financial Studies, 22(1), 435–480.
Phillips, J. D. (2003). Corporate tax-planning effectiveness: The role of compensation-based incentives. The Accounting Review, 78(3), 847–874.
Rego, S. O. (2003). Tax avoidance activities of US multinational corporations. Contemporary Accounting Research, 20(4), 805–833.
Rego, S. O., & Wilson, R. (2012). Equity risk incentives and corporate tax aggressiveness. Journal of Accounting Research, 50(3), 775–810.
Schwert, G. William (2002). Stock volatility in the new millennium: How wacky is Nasdaq? Journal of Monetary Economics, 49(1), 3–26.
Thomsen, M., & Watrin, C. (2018). Tax avoidance over time: A comparison of European and US firms. The Journal of International Accounting, Auditing, and Taxation,
33, 40–63.
Whited, T. M., & Wu, G. (2006). Financial constraints risk. The Review of Financial Studies, 19(2), 531–559.
Weisbach, D. A. (2002). Ten truths about tax shelters. Tax Law Review, 55, 215–254.
White, J. R. (2013). Corporate tax compliance: IRS should determine whether its streamlined corporate audit process is meeting its goals (GAO-13-662). Report to
Congressional Requesters: US Government Accounting Office. Available at http://www.gao.gov/assets/660/657092.pdf.
Wilson, R. J. (2009). An examination of corporate tax shelter participants. The Accounting Review, 84(3), 969–999.
Xu, Y., & Malkiel, B. (2003). Investigating the behavior of idiosyncratic volatility. The Journal of Business, 76(4), 613–645.

20