More and more international corporations, such as Vodafone, Google and Starbucks are being accused in the media of avoiding taxes. Starbucks managed to pay no corporate tax in the UK, in 2012, while the company actually had a sales revenue of ''400 million (BBC 2013). Starbucks managed to pay no tax by making optimal use of the existing loopholes in the tax law. An important thing to note, is that the activities of these firms are within legal boundaries of the tax law. Those firms are participating in tax avoidance, not tax evasion. Dyreng, Hoopes and Wilde (2016) suggest that the media's accusation of a firm's tax avoidance activities may have a reputational cost on the firm, due to for example, consumer boycotts and reputational damage. The fact that Starbucks voluntarily paid more tax than it was legally obligated to do suggest that increased media attention indeed has a reputational effect.
Simultaneously, advertising expenses are generally seen as a reputational cost and as a proxy for potential consumer backlash (Gallemore, Maydew, and Thornock 2014; Hanlon and Slemrod 2009). For example, the announcement that Starbucks did not pay any corporate tax led to negative media attention resulting in consumer boycotts, store closures and drops in their reputation rating (Christensen, Dhaliwal, Boivie, and Graffin 2015; The Guardian 2012). Increased advertising expenses may be used to counter this reputational damage (Gallemore et al. 2014).
Despite the fact, that there has been a lot of research on tax avoidance, little is actually known about the way in which corporations respond to media scrutiny on their tax avoidance behaviour (Dyreng et al. 2016). The results of recent studies are mixed. Several studies show that the media's accusation of tax avoidance activities have a negative effect on corporate reputation, consumer purchase intention and consumer willingness to pay (Hardeck and Hertl 2014; Trudel and Cotte 2009; Creyer and Ross 1996). The former indicates a reputational cost for the firm, which is more severe in the retail sector, according to Hanlon and Slemrod (2009). This sheds light on the idea that these costs may be caused by a consumer backlash. Several other studies, on the other hand, show that the media's accusation exert no reputational cost on the firm. For instance, Gallemore et al. (2014) discover no major change in advertising expenses.
This paper seeks to answer the following research question: Do firms spend more on advertising after they are accused of tax avoidance in the news media? If a firm increases its advertising expenses, after the media revealed its tax avoidance activities, this might indicate an attempt to try and reverse the reputational damage caused by a consumer backlash. In addition, this paper will research the following hypothesis, H1: Firms spend more on advertising after they are accused of tax avoidance in the news media.
The sample of this paper consists of the combined sample of Graham and Tucker (2006), Wilson (2009) and the sample resulting from self-conducted research. The final sample contains 43 firms with 164 firm-year observations from 1975-2013.
The regression results of this sample demonstrate that there is no significant relationship between the change in advertising expenses and the media's accusation of firms' tax avoidance activities. This follows from the fact that the coefficient '' of the regression analysis is insignificant. Nevertheless, the regression analysis shows that R&D has a significantly positive effect on the change in advertising expenses.
This research question is important because it answers the growing call for an investigation into the corporation's responses to the media's accusation of their tax avoidance activities. According to Graham, Hanlon, Shevlin, and Schroff (2012) more than two-thirds of the tax executives they surveyed, express that their firm's reputation is a very important factor in deciding whether or not to avoid taxes. Therefore, it is important to study these related reputational cost.
This study contributes to current research by investigating the company's response to the media's revelation of their tax avoidance activities and the effect of these responses on their advertising expenses. Moreover, it might be possible that the revelation of the major accounting scandals (e.g. Enron and WorldCom) in the early 2000s changes consumers' attitude towards tax avoidance activities. Therefore, it is possible that the media's accusation of tax avoiders leads to more severe reputational damage in the post-2000 period and therefore, also to higher advertising expenses to reverse the damage.
The rest of this paper is structured as follows. Section two gives a review of relevant literature on this topic, which will lead to the hypothesis of this study. Section three discusses the methodology of this study, while section four describes the sample selection. Furthermore, the results are presented in section five and the final section of this study contains the conclusion.
II. LITERATURE REVIEW AND HYPOTHESIS DEVELOPMENT
The Cone Corporate Citizenship Study (2002) documents that almost 90% of the surveyed Americans say that it is highly important for companies to be socially responsible, especially given the recent corporate scandals. High corporate social responsibility (CSR) namely leads to a better view on the company (Brown and Dacin 1997). This better evaluation of the company ensures that the company's product is viewed more favourably and it also has a positive impact on consumer's purchase intentions (Mohr and Webb 2005). This led to the idea that the evaluation of the company and consumers' purchase intentions are influenced by a more intangible characteristic of the company, namely a company's reputation (Brown and Dacin 1997). A company's reputation can occur from their philanthropic responsibilities. These philanthropic responsibilities are part of the CSR pyramid of Carroll (1991) which reflects the corporate actions that need to be undertaken to be viewed as a good corporate citizen by the general public.
Being viewed as a good corporate citizen and ethical company has many advantages. For example, consumers are willing to pay more for products of an ethical firm (Trudel and Cotte 2009; Bhattacharya and Sen 2005). Moreover, they also demand lower prices for firms that are seen as unethical (Trudel and Cotte 2009; Creyer and Ross 1996; Hardeck and Hertl 2014). This lowers willingness to pay and makes it difficult for a corporation to compete with its rivals (Creyer and Ross 1996). Consequently, reputation is a tool which can be used to achieve and sustain a competitive advantage. Another advantage according to Bhattacharya and Sen (2003) is the fact that consumers are more loyal to companies viewed as being ethical.
According to Carroll (1991) ethical responsibilities can be reflected by the stakeholders' (e.g. consumers, employees, shareholders) concern of what they regard as fair, such as norms or particular expectations. If corporations do not comply with their philanthropic responsibilities, they can be viewed as a bad corporate citizen by the general public and hence as unethical. Consumers are more sensitive and react more strongly towards negative CSR than to positive CSR information, when it comes to their evaluation of the company (Sen and Bhattacharya 2001; Mohr and Webb 2005; Creyer and Ross 1996; Hardeck and Hertl 2014). Due to this reason, unethical behaviour is punished more than social responsible behaviour is rewarded (Trudel and Cotte 2009). Bhattacharya and Sen (2005) find that consumers are more likely to avoid companies held in bad regard, than they would go out of their way to purchase products from a socially responsible company. Hardeck and Hertl (2014) view an aggressive corporate tax strategy (e.g. tax avoidance) as a negative CSR activity. Lisowsky (2010) defines tax aggressiveness as a behaviour in which the most important thing is to reduce the taxes that has to be paid. The non-tax effects are seen as the second most important thing. Tax shelters are considered illegal because they are created solely for the purpose of tax evasion and they do not address any other economic need/risk of the business (Lisowsky 2010).
Firms that are engaged in tax shelter activities are labelled as socially irresponsible (Erle 2008; Sch''n 2008). Due to the fact that the general public has very little tolerance towards aggressive tax strategies, these strategies can impose immense reputational risks on a corporation (Shulman 2009; Merkl-Davies and Brennan 2007). This is especially the case when these aggressive tax strategies are being revealed in the media. The media report on corporate tax avoidance news makes sure that the complex tax issues are being translated into simpler issues, which can easily be understood by the general public and can therefore become part of a social concern (Lee 2015).
Reputational risk comes in different forms, for instance consumer backlash and stock price declines (Hanlon and Slemrod 2009). If a firm gets caught engaging in tax avoidance, investors expect the firm value to decline because of expected fines, penalties and lost sales revenues (Davidson and Worrel 1988). Consequently, corporate reputation is also very important for corporate profits. Lost sales revenues after the accusation of tax avoidance can be explained by consumers' lower willingness to pay and consumers' decreased purchase intentions (Hardeck and Hertl 2014). Hence, tax avoidance is seen as a negative CSR activity and will therefore be value destroying, looking at it from a consumer's perspective. German and U.S. executives consider corporate reputation to be the most fundamental measure of corporate success (Dunbar and Schwalbach 2001). These findings show that tax avoidance has a negative effect on corporate reputation and corporate success.
A commonly used test to address reputational concerns is the Wall Street Journal test. In this test, the company's activities are assessed based on how these activities would look on the front page. If the activities appear to look bad on the front-page, looking at it from the general public's perspective, than the company should not engage in these activities. The fact that corporations look at this test indicates that corporations are aware of the reputational damage that stems from tax avoidance. Furthermore, Austin and Wilson (2013) expect managers to be aware of the potential reputational damage that comes from negative consumer reactions to the accusation of tax avoidance activities in the news. Nearly 70% of the executives surveyed by Graham et al. (2012) admit that potential harm to corporate reputation is one of the most important reasons not to engage in aggressive tax strategies. Another 60% say that media attention also plays a significant role in their decision. Their reasons seem legit since accusation of tax avoidance behaviour can lead to revenge from consumers (e.g. boycotts) (Pruitt and Friedman 1986).
High advertising expenses indicate potential consumer backlash (Hanlon and Slemrod 2009; Gallemore et al. 2014). For example, the announcement that Starbucks did not pay any corporate tax led to negative media attention resulting in consumer boycotts, store closures and drops in their reputation rating (Christensen et al. 2015; The Guardian 2012). These potential risks can require the corporation to spend additional resources (e.g. increased advertising expenses) in order to reinsure key shareholders (Ernst & Young 2014). Graham et al. (2012) find that corporations which operate in the public eye consider reputation and negative media attention important reasons with regard to their decision of tax avoidance. High advertising firms are less likely to avoid taxes because they hope to avoid negative views or comments towards their corporation (Dyreng, Hanlon, and Maydew 2008; Hanlon, Mills, and Slemrod 2005; Lee 2015). Corporations with higher advertising expenses are more likely to be in the retail-industry since this is more visible to the consumers' eye (Hanlon et al. 2005; Hanlon and Slemrod 2009; Austin and Wilson 2013). The reputational damage might stem from the fact that the general public views corporations participating in aggressive tax strategies as bad corporate citizens (Hanlon and Slemrod 2009; Lee 2015). Increased advertising expenses may be used to counter the reputational damage resulting from the media's accusation of tax avoidance activities (Gallemore et al. 2014). Therefore, the hypothesis will be as following: H1: Firms spend more on advertising after they are accused of tax avoidance in the news media.
III. METHOD
This study will answer the research question by empirically testing the hypotheses by the use of a regression model. In particular, this study will make use of the difference-in-differences methodology to study the company's response to the accusation of tax avoidance activities in the news media and the effect of these responses on their advertising expenses. The regression model considers the relationship between the change in advertising expenses and the media's accusation of a firm's tax avoidance activities. The regression model of this study is based on the model used in Gallemore et al. (2014).
Model: 'AdExp=''0+''1DCAUGHTFIRM+''2DCAUGHTYEAR +''3DCAUGHTFIRM*DCAUGHTYEAR +''4SIZE+''5LEV+''6ROA+''7R&D + ''
Dependent variable
The change in advertising expenses ('AdExp) represents the variable that needs to be explained. 'AdExp is measured by current advertising expenses minus last year's advertising expenses divided by last year's advertising expenses.
Independent variable
The explanatory variables in this paper are DCAUGHTFIRM, DCAUGHTYEAR and the interaction term DCAUGHTFIRM*DCAUGHTYEAR. DCAUGHTFIRM is a dummy variable, equal to one for firms which are accused of tax avoidance in the media, and zero for control firms. DCAUGHTYEAR is another dummy variable which is set to one for both the treatment and control firms, in the year the treatment firm was accused of tax avoidance in the media and zero otherwise. The interaction term DCAUGHTFIRM*DCAUGHTYEAR is the most important variable in the model since it indicates the change in advertising expenses in the year the treatment firms were accused of tax avoidance in the media, compared to the matched control firm in the same year.
Control variables
In addition, several control variables are included in the model, just like in prior research studies, namely size, leverage, profitability and R&D expenses. Wilson (2009) finds that tax avoidance is positively related to firm size (SIZE) and profitability (ROA). SIZE is calculated as the natural logarithm of total assets, while profitability (ROA) is calculated by dividing net income by total asset. Next, Leverage (LEV) is calculated as long-term debt divided by total assets. According to Lisowsky (2010), Graham and Tucker (2006), and Wilson (2009) firms that use aggressive tax strategies use less debt since they can both be seen as an instrument for lowering income taxes, so there is a negative relationship. Further, Research & Development expense (R&D) is calculated by dividing the R&D expenditures by total assets. Gupta and Newberry (1997) find that R&D has a positive relationship with tax avoidance, which can be explained by the tax-deductible R&D expenses. Finally, '' is the error term.
IV. SAMPLE SELECTION
The sample of this paper consists of the combined sample of Graham and Tucker (2006), Wilson (2009) and the sample resulting from self-conducted research. This led to 75 firms with 299 firm-year observations. To maximize the sample, the variables advertising expenses, R&D expenses and long-term debt are set to zero if missing, following Dyreng and Lindsey (2009). Removed from the sample are firms with insufficient data (69 firm-year observations), those with insufficient data for matching (7), those for which no matched control firm with advertising expenses could be found (52) and one outlier (7). The final sample therefore contains 43 firms with 164 firm-year observations from 1975-2013. All financial accounting data is obtained from North America Compustat.
Accordingly with Graham and Tucker (2006) the tax avoidance years of the accused firm are collapsed into one single year t, which is done by averaging the data. Data across firms can therefore easily be compared regardless of the length of the tax avoidance period because t-1 represents the year before the tax avoidance occurred and t+1 represents the year it ended. Control firms are matched based on closeness in firm size and on the same Standard Industrial Classification (SIC) code as the accused tax avoidance firm in year t-1. Hence, matched control firms are those with total assets (ROA) within +/- 25% (50%) of the accused firms' total assets (ROA). Each treatment firm is matched to one control firm.
V. RESULTS
Table 1 shows the descriptive statistics of the variables for the full sample of treatment and matched control firms (with N = 164 firm-year observations). The descriptive statistics of the variable 'AdExp is the most important to consider in this model. The mean of the change in advertising expenses is 20.6521%, this means that the advertising expenses went up with roughly 20% compared to the year before.
Table 1: Descriptive Statistics
Variable N Mean Minimum Median Maximum Standard deviation
'AdExp (%) 270 20.6521 -100.0000 0.9870 493.7500 66.2367
DCAUGHTFIRM 270 0.5000 0.0000 0.5000 1.0000 0.5009
DCAUGHTYEAR 270 0.3333 0.0000 0.0000 1.0000 0.4722
DCAUGHTFIRM*DCAUGHTYEAR 270 0.1667 0.0000 0.0000 1.0000 0.3734
SIZE 270 8.2542 1.6557 8.0779 14.6975 2.5867
LEV 270 0.1314 0.0000 0.1109 0.7583 0.1308
ROA (%) 270 6.3719 -50.3770 6.7794 31.7622 8.2473
R&D 270 0.0389 0.0000 0.0093 0.2199 0.0526
In Table 2 the Difference-in-Differences methodology is used to compare the means of the change in advertising expenses before and after the firms were accused of tax avoidance in the news media. The sample is divided into two groups based on the dummy variable DCAUGHTFIRM, which is one for firms accused of tax avoidance in the news media and zero for control firms. Although it is indicated in Table 2 that the advertising expenses of the treatment firms increase after the accusation of tax avoidance in the news media, this result is insignificant and therefore we cannot use this results to provide an answer to the hypothesis. Moreover, the insignificant result of the difference-in-differences estimator indicates that control firms increased their advertising expenses more than the treatment firms. This difference may be due to the fact that all control firms do not have any missing data for advertising expenses (as this was required to be included in the sample) while treatment firms did have missing data which was set to zero. Likewise, this results in lower advertising expenses for treatment firms than those of the control group.
Table 2 : Difference-in-Differences
Variable 'AdExp before 'AdExp after Difference
Treatment
(DCAUGHTFIRM=1) 7.679%
(1.670) 14.158%
(0.753) 6.479%
(0.665)
Control
(DCAUGHTFIRM=0) 9.317%
(1.615) 28.241%
(1.700)** 18.924%
(1.510)
Difference-in-differences -12.445%
(-0.784)
*, **, *** denote statistical significance at 10%, 5% and 1%, respectively
Table 3 presents the Pearson correlation matrix for the dependent and independent variables used in this study. The matrix shows a weak positive correlation between 'AdExp and DCAUGHTFIRM*DCAUGHTYEAR, however this coefficient is insignificant (p = 0.103). Nevertheless, there is a significant positive correlation between 'AdExp and DCAUGHTYEAR (p = 0.021) and R&D (p = 0.002). Furthermore, the matrix also show a significant negative correlation between 'AdExp and SIZE (p = 0.066), where a positive relationship was expected. Moreover, it can be concluded from this matrix that there are no problems of multicollinearity among the independent variables, since the correlation coefficients do not exceed 0.8.
'AdExp DCAUGHTFIRM DCAUGHT
YEAR DCAUGHT
FIRM*
DCAUGHT
YEAR SIZE LEV ROA R&D
'AdExp 1
DCAUGHT
FIRM -0.041 1
DCAUGHT
YEAR 0.124** 0.000 1
Table 3 : Correlation Matrix
DCAUGHT
FIRM*
DCAUGHT
YEAR 0.077 0.447*** 0.632*** 1
SIZE -0.092* 0.265*** 0.017 0.131** 1
LEV -0.074 -0.024 -0.011 -0.015 0.128** 1
ROA -0.008 0.220*** -0.051 0.074 -0.160*** -0.084* 1
R&D 0.171*** 0.092* -0.016 0.040 -0.310*** -0.230*** 0.176*** 1
*, **, *** denote statistical significance at 10%, 5% and 1%, respectively. Significance levels are one-tailed.
Table 4 shows the results of the regression analysis with the dependent variable 'AdExp. The most important variable in our model, namely the interaction term DCAUGHTFIRM*DCAUGHTYEAR, has a positive coefficient ('' = 6.964) which would indicate that firms respond to the negative publicity that results from being accused of tax avoidance in the news media by increasing their advertising expenses. However this result is insignificant (p = 0.680) and therefore we cannot use this result to confirm the hypothesis (H1) of this paper. The insignificance of the coefficient '' is consistent with the result of Gallemore et al. (2014) as they also did not find any significant relationship between being accused of tax avoidance in the news media and increased advertising expenses.
Although the model is significant at a 90% confidence level (p = 0.051), the fact that most of the coefficients of the model are insignificant (except for R&D) indicates that there is no convincing evidence that the independent variables have an effect on our dependent variable. The positive and significant relationship for R&D is consistent with Gupta and Newberry (1997), which can be explained by the tax-deductible R&D expenses.
Table 4: Regression Analysis
Variable Predicted sign Coefficient P-value
INTERCEPT ? 21.383 0.208
DCAUGHTFIRM ? -7.771 0.452
DCAUGHTYEAR ? 14.108 0.239
DCAUGHTFIRM*
DCAUGHTYEAR + 6.964 0.680
SIZE + -0.841 0.629
LEV – -17.111 0.586
ROA + -0.241 0.638
R&D + 206.248 0.014**
N
Adjusted R''
F-statistic 270
0.026
2.039*
Dependent variable is 'AdExp.
*, **, *** denote statistical significance at 10%, 5% and 1%, respectively
VI. CONCLUSION
The main purpose of this study is to investigate whether firms would increase their advertising expenses after they have been accused of tax avoidance in the news media. This study contributes to current research by answering the growing call for an investigation into the corporation's responses to the media's accusation of their tax avoidance activities. It was expected that firms would increases their advertising expenses after they were accused of tax avoidance in the news media. This study however, did not find any empirical evidence to support the formulated hypothesis. This result is consistent with the result of Gallemore et al. (2014).
Some limitations of the study can be mentioned. Due to the fact that the sample is partly hand selected from public records makes it likely that the sample suffers from self-selection issues. Furthermore, this study only examines firms whose tax avoidance activities were revealed in the news media. The effect of tax avoidance news on the change in advertising expenses might be underestimated, because of the fact that firms who might suffer a significant reputational loss restrain themselves from avoiding taxes due to the fact that they have the most to lose (Graham et al. 2012). As a result this could potentially affect the inferences of this study since firms who might suffer a significant reputational loss are not included in the sample. Additionally, this study takes a one-year time horizon by looking at t-1, t and t+1. However, it might well be possible that the reputational effect caused by the news media's accusation of tax avoidance will take a longer time horizon before it affects the firm (Gallemore et al. 2014). Lastly, the size of the sample used in this study is relatively small (43 firms with 164 firm-year observations) which decreases the significance of the results.
In terms of future research, the focus should be on broadening the area of tax avoidance research that barely covers the relationship between tax avoidance and reputational effects such as the change in advertising expenses. Future studies could examine this relationship by also including firms who might suffer a significant reputational loss due to the fact that they have the most to lose. This could lead to interesting results and eliminates one of the limitations of this study. Moreover, by using a bigger sample size the contribution and significance of the results of the study will be higher.