Advertising and tax avoidance

Abstract

We examine the link between firms’ advertising and their tax avoidance. By generating customer awareness, advertising helps shape firm image and reputation among customers. Such benefits of advertising would diminish if the firm is viewed as a greedy tax dodger. Greater customer awareness generated by higher advertising spending also increases the likelihood that customers would find out tax-aggressive behaviors of the firm. Thus, firms that spend more on advertising may want to be less tax-aggressive. Consistent with this argument, we find that firms with a greater extent of advertising spending have fewer tax-sheltering activities, smaller book-tax differences, and higher cash-effective tax rates. The negative effect of advertising on tax avoidance is stronger for firms that are less known, more opaque, or that have lower institutional holdings. We control for other factors affecting tax avoidance, including corporate governance and social responsibility ratings. We also use the instrumental variable method, propensity score matching, and change regressions to address endogeneity concerns. Our results remain statistically and economically significant.

Appendix 1

Variable definitions

Variable	Definition
	Measures of tax aggressiveness and advertising
SHELTERi,t	Wilson’s (2009) sheltering probability is computed using the following regression model (Table 5 Column 3): $\begin{aligned} {\text{SHELTER}}\_{\text{PROB}}_{\text{it}} & = - 4.86 + 5.20 \times {\text{BTD}}_{{{\text{i}},{\text{t}}}} + 4.08 \times {\text{DA}}_{{{\text{i}},{\text{t}}}} - 0.41 \times {\text{LEV}}_{{{\text{i}},{\text{t}}}} \\ & \quad + 0.76 \times {\text{AT}}_{{{\text{i}},{\text{t}}}} + 3.51 \times {\text{ROA}}_{{{\text{i}},{\text{t}}}} + 1.72 \times {\text{FI}}_{{{\text{i}},{\text{t}}}} + 2.43 \times {\text{RD}}_{{{\text{i}},{\text{t}}}} \\ \end{aligned}$ The above ${\text{SHELTER}}\_{\text{PROB}}_{\text{it}}$ is the sheltering probability from firm i in year t. ${\text{BTD}}_{{{\text{i}},{\text{t}}}}$ is the book-tax difference ${\text{KIMBTD}}_{{{\text{i}},{\text{t}}}}$ used in Kim et al. (2011). ${\text{DA}}_{{{\text{i}},{\text{t}}}}$ is discretionary accruals from the performance-adjusted modified cross-sectional Jones Model. ${\text{LEV}}_{{{\text{i}},{\text{t}}}}$ is the long-term debt scaled by lagged asset (#9/#6). ${\text{AT}}_{{{\text{i}},{\text{t}}}}$ is the Log of total asset (#6). ${\text{ROA}}_{{{\text{i}},{\text{t}}}}$ is the return on assets measured as operating income (#170 minus #192) scaled by lagged assets (#6). ${\text{FI}}_{{{\text{i}},{\text{t}}}}$ (#273) is a dummy variable, which equals one for a firm-year that reports foreign income and zero otherwise. ${\text{RD}}_{{{\text{i}},{\text{t}}}}$ is research and development expenses scaled by lagged total assets (#46/#6). Following Rego and Wilson (2012), we rank ${\text{SHELTER}}\_{\text{PROB}}_{{{\text{i}},{\text{t}}}}$ each year and create a dummy variable to capture firms that have a high sheltering probability. In particular, SHELTERi,t is an indicator variable that equals one if the firm’s estimated sheltering probability is in the top quartile in that year, and zero otherwise
CETRi,t	Defined as Cash Taxes Paid/Pretax Income (#317 divided by #170). CETRi,t is set to missing when the denominator is zero or negative. It is truncated to the range of [0, 1]
KIMBTDi,t	Defined as book income (#170) minus taxable income over lagged assets (#6). Taxable income is computed as the sum of current federal tax expense (#63) and current foreign tax expense (#64) divided by the statutory tax rate, and then subtracted the change in net operating loss carryforwards (#52). If current deferral tax expense is missing, total current tax expense is calculated by subtracting deferred taxes (#50), state income taxes (#173), and other income tax (#211) from total income taxes (#16) in year t
DDKIMBTDi,t	Desai and Dharmapala’s (2006) residual book-tax difference, which equals the residual of the firm fixed effect regression: ${\text{KIMBTD}}_{{{\text{i}},{\text{t}}}} =\upbeta_{1} {\text{TACC}}_{{{\text{i}},{\text{t}}}} +\upmu_{\text{i}} +\upvarepsilon_{{{\text{i}},{\text{t}}}}$, where ${\text{TACC}}_{{{\text{i}},{\text{t}}}}$ is total accruals measured using the cash flow method of Hribar and Collins (2002). Both variables are scaled by lagged total assets (#6) and are winsorized at the 1% and 99% levels
DTAXi,t	Frank et al.’s (2009) discretionary permanent book-tax difference for firm i in year t, ${\text{DTAX}}_{{{\text{i}},{\text{t}}}}$, is the $\upvarepsilon_{\text{it}}$ from the following regression estimated by 2-digit SIC code and fiscal year: ${\text{PERMDIFF}}_{\text{it}} =\upbeta_{0} +\upbeta_{1} {\text{INTANG}}_{\text{it}} +\upbeta_{2} {\text{UNCON}}_{\text{it}} +\upbeta_{3} {\text{MI}}_{\text{it}} +\upbeta_{4} {\text{CSTS}}_{\text{it}} +\upbeta_{5} \Delta {\text{NOL}}_{\text{it}} +\upbeta_{6} {\text{LAGPERM}}_{\text{it}} +\upvarepsilon_{\text{it}}$ where ${\text{PERMDIFF}}_{\text{it}} = {\text{BI}}_{\text{it}} - \left[ {({\text{CFTE}}_{\text{it}} + {\text{CFOR}}_{\text{it}} )/{\text{STR}}_{\text{it}} } \right) - ({\text{DTE}}_{\text{it}} /{\text{STR}}_{\text{it}} )$ ${\text{BI}}_{\text{it}}$ = pre-tax book income (#170) for firm i in year t ${\text{CFTE}}_{\text{it}}$ = current federal tax expense (#63) for firm i in year t ${\text{CFOR}}_{\text{it}}$ = current foreign tax expense (#64) for firm i in year t ${\text{STR}}_{\text{it}}$ = statutory tax rate in year t ${\text{DTE}}_{\text{it}}$ = deferred tax expense (#50) for firm i in year t ${\text{INTANG}}_{\text{it}}$ = good will and other intangibles (#33) for firm i in year t ${\text{UNCON}}_{\text{it}}$ = income (loss) reported under the equity method (#55) for firm i in year t ${\text{MI}}_{\text{it}}$ = income (loss) attributable to minority interest (#49) for firm i in year t ${\text{CSTE}}_{\text{it}}$ = current state income tax expense (#173) for firm i in year t $\Delta {\text{NOL}}_{\text{it}}$ = change in net operating expense (#52) for firm i in year t ${\text{LAGPERM}}_{\text{it}}$ = 1-year lagged PERMDIFF firm i in year t. Following Frank et al. (2009), we handle the missing value problems as follows: if ${\text{MI}}_{\text{it}}$,${\text{CFOR}}_{\text{it}}$,${\text{UNCON}}_{\text{it}}$, or ${\text{CSTE}}_{\text{it}}$ is missing, it is set to zero. If ${\text{CFTE}}_{\text{it}}$ (#63) is missing, then ${\text{CFTE}}_{\text{it}}$ is computed as total tax expense (#16) less current foreign tax expense ($\# 64)$, less current state tax expense (#173), and less deferred tax expense (#50). If ${\text{INTANG}}_{\text{it}}$ (#33) is missing, then it is set to zero. If ${\text{INTANG}}_{\text{it}}$ (#33) = “C”, then it equals that for good will (#204)
Log(ADV)i,t	Natural logarithm of one plus advertising expenditure (#45) times 1,000,000
ADVGPi,t	Ratio of advertising expenditure (#45) to gross profits (#12-#41)
HADVDUMi,t	Dummy variable that equals one if advertising expenditure is above the median, and zero otherwise
	Firm specific variables
High_Neg_CSRi,t	Following Hoi et al. (2013), we construct a CSR (corporate social responsibility) weakness index based on seven categories: corporate governance, employee relations, environment, community, diversity, human rights, and product quality and safety. High_Neg_CSRi,t is an indicator variable on negative CSR activities, which equals one when firm i has four or more irresponsible CSR activities in year t, and zero otherwise
OPACITYi,t	Following Anderson et al. (2009), we compute opacity by aggregating the decile ranks from four variables: bid-ask spread, trading volume, analyst coverage, and analyst forecast errors, and then dividing the sum by 40. Trading volume is the nature logarithm of average daily trading volume during a fiscal year. Bid-ask spread is ask-price minus bid-price over the average of bid and ask prices, computed by averaging all trades for each firm from the third Wednesday of each month and then calculated a yearly average based on these 12 observations. Analyst coverage is the natural logarithm of the number of analysts following each firm, and analyst forecast error is the square of difference between the mean analysts’ earnings forecast and actual firms’ earnings scaled by the firm’s stock price
SP1500i,t	Dummy variable that equals one if a firm is listed in the S&P 1500 firms over a firm year
INST.OWNi,t	Percentage of shares owned by institutional investors scaled by common shares outstanding
ROAi,t	Return on assets measured as operating income (#170 minus #192) scaled by lagged assets (#6)
LEVi,t	Leverage ratio measured as long-term debt (#9) scaled by lagged asset (#6)
ΔNOIi,t	Change in loss carry forward (#52) scaled by lagged asset (#6)
NOIi,t	Dummy variable that equals one if loss carry forward (#52) is positive as of the beginning of the year
FIi,t	Foreign income (#273) for firm i, year t, scaled by lagged assets (#6)
PPEi,t	Property, plant, and equipment (#8) scaled by lagged assets (#6)
INTANGi,t	Intangible assets (#33) scaled by lagged assets (#6)
EQINCi,t	Equity income in earnings (#55) scaled by lagged assets (#6)
SIZEi,t	Natural logarithm of firm i’s total assets (#6)
MTBi,t−1	Market-to-book ratio for firm i, at the beginning of year t, measured as market value of equity [(#199) × (#25)], scaled by book value of equity (#60)
RDi,t	Research and development ratio measured as research and development expenses (#46) scaled by lagged total assets (#6). Missing values are set to zero
CASHi,t	Cash and cash equivalents (#1) scaled by lag of total assets (#6) net of cash
LAGEi,t	Natural logarithms of firm age in Compustat
DIVi,t	Dummy variable that equals one if a firm pays dividends (#201) at year t, and zero otherwise
EMPi,t	Number of employees in the firm (#29) in thousands at year t
High_Eindexi,t	Following Bebchuk et al. (2008), we construct an entrenchment index, Eindex, based on six categories: staggered board, limitation on amending bylaws, limitation on amending the charter, supermajority to approve a merger, golden parachute, and poison pill. High_Eindexi,t is an indicator variable on strong entrenchment or weak governance, which equals one if firm i has at least three potential red flags on governance issues at year t, and zero otherwise