Discussion of Dhaliwal, Goodman, Hoffman, and Schwab (2019): Revisiting Tax-Related Reputational Costs

Abstract

In the broadest sense, Dhaliwal, Goodman, Hoffman, and Schwab (2019; hereafter, DGHS) examine whether firms incur tax-related reputational costs. While such costs are often discussed in the literature as a factor dissuading managers from engaging in some types of corporate tax avoidance (see Dhaliwal et al. 2019, footnote 9), prior research generally fails to find evidence consistent with firms or managers bearing these costs (e.g., Hanlon and Slemrod 2009; Gallemore, Maydew, and Thornock 2014; Austin and Wilson 2017; Chen, Schuchard, and Stomberg 2019). DGHS re-examine this question in the context of an interesting setting where tax-related reputational costs, if present, should be most discernable. This setting is the year 2011—a period of U.S. socioeconomic unrest characterized by widespread protests of income inequality, corporate greed, and regulatory capture.DGHS test their research question through three sets of analyses: (1) whether tax-avoiding firms experience greater negative media sentiment during 2011; (2) whether tax-avoiding firms experience lower market valuations during 2011; and (3) whether firms that incur reputational costs during 2011 increase their reported taxes in subsequent years. The authors document a positive association between corporate tax avoidance and negative media sentiment during 2011 and find that a hedge portfolio that takes a short (long) position in high (low) tax-avoidance firms generates positive abnormal returns during 2011. These findings are consistent with the media and investors imposing tax-related reputational penalties on firms. The authors also document that firms experiencing the most negative media sentiment in 2011 exhibit less future tax avoidance, consistent with firms responding to and managing tax-related reputational costs. Collectively, these findings can be interpreted as firms incurring tax-related reputational costs during a period of social protest.I find this paper thought-provoking and insightful. DGHS contribute to the tax-avoidance literature by furthering our understanding of when, and to what extent, tax-related reputational costs exist. Demonstrating that such costs are present only during periods characterized by high societal scrutiny of socioeconomic concerns suggests that future researchers should think carefully before motivating their research questions with time-invariant arguments regarding tax-related reputational costs. The authors also contribute to the literature by introducing important aspects of sociology theory and research to an accounting audience. Understanding when and how societal movements affect perceptions of corporate economic activity and corporate financial reporting is an under-developed research stream rife for further examination.DGHS use observational data to identify and measure tax-related reputational penalties. When evaluating studies that rely on observational data to identify and measure complex social constructs, I am particularly interested in the validity of the setting and the empirical measures. Thus, my discussion focuses on DGHS' ability to (1) identify a setting where tax-related reputational costs are particularly salient, and (2) measure these tax-related reputational costs with reasonable accuracy. My comments are based on the version of DGHS presented at the 2019 Journal of the American Taxation Association Conference. I first offer an overview of the 2011 social protest movement to provide context for my comments.The year 2011 was a period of intense socioeconomic unrest. In the United States, the seasonally adjusted unemployment rate immediately prior to the 2008 financial crisis was approximately 4.5 percent. This rate rose steadily during the financial crisis and peaked at 10 percent in late 2009 (its highest since the early 1980s). The rate remained high at 8.5 percent as of the end of 2011 (Bureau of Labor Statistics 2012). Growing frustrations with U.S. economic inequality, which was perceived to be fueled by policies that disproportionately benefited corporations and wealthy individuals at the expense of ordinary Americans, led hundreds of thousands of individuals to the streets in protest during 2011.The most notorious social protest in the United States was Occupy Wall Street (OWS), in which protestors physically occupied an area near Wall Street in the financial district of New York City in Fall 2011. Here is how OWS describes itself:Frustration with the richest 1 percent inspired OWS' signature rallying cry: “We are the 99 percent!” OWS began with approximately 1,000 individuals (Zerbisias 2011), and grew over time to include tens of thousands of people (including nearly 40 different unions and community organizations) marching across various parts of Manhattan (Boyle, Sher, Mullany, and Kennedy 2011). As the movement grew in size and scope, it attracted significant attention and spread across the U.S. and around the globe. By mid-November 2011, 951 cities in 82 countries had hosted some sort of “Occupy” event (Rogers 2011).Time Magazine's decision to declare “The Protester” as its Person of the Year summarizes how important protests as a social movement were during 2011. The magazine cover headline, “The Protestor: From the Arab Spring to Athens, from Occupy Wall Street to Moscow,” highlights the global reach of the 2011 protest movement (http://content.time.com/time/covers/0,16641,20111226,00.html). Thus, there is little question in my mind that 2011 was a tipping point for intense socioeconomic frustration that led to public displays of dissent and increased awareness of economic inequality. Having set the stage for 2011 as a year of social protest, I now turn to my two primary comments regarding DGHS.DGHS focus on 2011 because “this period of social protest represents an exogenous shock that temporarily increased scrutiny of corporate tax behavior” (Dhaliwal et al. 2019, 1). While it seems clear that 2011 was a period of social protest, it is less clear as to whether this social protest increased scrutiny of corporate income tax avoidance. To consider this question, I first turn to OWS' self-description noted in the prior section. The movement's rallying cry “we are the 99 percent!” reflects that the people in the bottom 99 percent of the income/wealth distribution are frustrated with the people in the top 1 percent. This rally cry is about individual income inequality, not corporate tax avoidance. Tax policy is often tied to income inequality, but more so in the context of a progressive tax system and using the tax code to redistribute wealth. OWS' self-description also notes that it is “fighting back against the corrosive power of major banks and multinational corporations over the democratic process” (http://occupywallst.org/about/). This statement is consistent with perceptions of crony capitalism and frustration with corporate lobbying and special interests. Corporate tax breaks could be the outcome of such activities, but so could many other things (e.g., lax environmental regulations, weakening protections for workers' rights, preferential regulatory approval, lucrative government contracts). These points question whether the 2011 protest period is a setting where tax-related reputational costs are expected to be particularly salient.The authors acknowledge this concern on several occasions, noting that “tax avoidance was not the primary focus of these protests” (Dhaliwal et al. 2019, 12) and “taxes may not have been the protests' primary target” (Dhaliwal et al. 2019, 13). However, they go on to note that “taxes and tax avoidance activities did garner substantial attention as important components of those larger problems and the proposed solutions” (Dhaliwal et al. 2019, 12). DGHS validate their assertion that corporate tax avoidance received significant attention during the protest period through an extensive Factiva article search. The authors compile a list of three high-level issues being protested (income inequality, corporate greed, and regulatory capture), 24 grievances related to these three high-level issues, and 107 terms related to these 24 grievances. One of the 24 grievances and four of the 107 terms relate to taxation (“fair tax,” “corporate tax,” “tax loophole,” and “progressive tax”).1 The authors find a 77 percent increase in frequency (from the pre-protest to protest period) of articles mentioning one of these four tax terms (Dhaliwal et al. 2019, Table 7). This statistic suggests that there was an increase in the scrutiny of taxation during the 2011 protest period. However, it is difficult to know if a 77 percent increase in article frequency is small or large absent a relevant benchmark.The online appendix (Dhaliwal et al. 2019) provides numerous potential benchmarks and shows that many other corporate-related grievances/terms experienced qualitatively similar increases in frequency from the pre-protest to protest period. For example, the term “corporate bribery” experienced a 73 percent increase. If 2011 is a good setting in which to examine tax-related reputational costs, then it should also be a good setting in which to examine bribery-related reputational costs. Extant research finds that “the revelation of bribery, by itself, has little long-term impact [on corporate reputation]” (Karpoff, Lee, and Martin 2017, 4). Could examining the reputation effects of corporate bribery during 2011 yield a different insight?Relatedly, the term “Citizens United” (which relates to a U.S. Supreme Court case regarding corporate money in politics) experienced a 69 percent increase from the pre-protest to protest period.2 The year 2011 should also be a good setting in which to examine the reputation effects of corporate political connections. A long line of research finds a positive association between political connections and firm value (e.g., Fisman 2001; Faccio 2006; Cooper, Gulen, and Ovtchinnikov 2010). Is this relation less positive (or potentially negative) during 2011? Examining bribery-related reputational penalties in 2011 seems beyond the scope of DGHS. However, examining political connections-related reputational penalties in 2011 could be within scope given the documented association between political connections and corporate income tax avoidance (Brown, Drake, and Wellman 2015; Kim and Zhang 2016). Thus, it is possible that corporate political connections is a correlated omitted variable; what DGHS label as tax-related reputational costs could in part reflect political-connectedness reputational costs.Finally, other corporate-related grievances/terms experienced qualitatively greater increases in frequency from the pre-protest to protest period. For example, the term “American jobs” experienced a 118 percent increase. Williams (2018) finds some evidence that corporations off-shoring U.S. jobs report lower income taxes. Thus, it is possible that what DGHS label as tax-related reputational costs in part capture the reputational costs of offshoring American jobs.The authors discuss several settings within 2011 where tax-related reputational costs might be easier to isolate. Here are some thoughts on three of those settings. First, DGHS could look at firms specifically targeted by the protests for tax-avoidance activities. The analyses in Dhaliwal et al. (2019, Table 9) are in the spirit of this suggestion. Dhaliwal et al. (2019, Table 9) show the differential market response to the release of Citizens for Tax Justice (CTJ) reports in 2011 versus 2004 for high versus low tax-avoidance firms. However, documenting that tax-avoiding firms experienced lower market returns surrounding the 2011 CTJ report release date only tests DGHS' second hypothesis. I expected the authors to use the CTJ list of nearly 300 targeted firms to also examine whether the higher tax-avoidance firms experienced greater negative media sentiment in the month the 2011 report was released (DGHS' first hypothesis). The list of firms could also be used to determine whether the tax-avoidance firms experiencing the lower market returns or negative media scrutiny during the protest period subsequently changed their tax-avoidance behavior (DGHS' third hypothesis). Relatedly, the manuscript notes that “U.S. Uncut, which was created to protest corporate tax avoidance and cuts to social spending and public sector jobs, targeted a number of corporations directly” (Dhaliwal et al. 2019, 11). Are tax-related reputational costs greatest (or present only) for the firms targeted by U.S. Uncut?3Second, DGHS could look at an industry targeted by the protests—the financial services industry. The authors write that the socioeconomic unrest of 2011 was rooted in the 2008 financial crisis, and open their paper with a quote noting that “ongoing revelations of double dealing by banks” (Dhaliwal et al. 2019, 1) was a contributing factor in mobilizing protestors. OWS notes it is “fighting back against the corrosive power of major banks,” and the organization's decision to occupy Wall Street—the financial district of New York City and home to the U.S. financial markets—reflects protestors' disdain for the financial services industry. The increased focus on this industry is further supported by financial services-related terms materially increasing in article frequency during 2011—the online appendix (Dhaliwal et al. 2019) notes that “underwater mortgage” increased by 177 percent, “student debt” increased by 267 percent, and “Prudent Banking Act” increased by 625 percent.4 Thus, I was surprised to see DGHS exclude the financial services industry (i.e., firms with SIC 6000–6999) from their analyses.5 Focusing on the industry most targeted by the protests would be particularly insightful in helping us understand corporate reputational penalties (and whether these penalties are tax related).Finally, DGHS could explore cross-sectional variation in their full sample as a function of income inequality—one of the central tenants of the 2011 protest movement. Are tax-related reputational costs greater (or present only) for firms headquartered or operating in places where income inequality is the greatest? See Sommeiller and Price (2018) for how to access county-state-year measures of income inequality from various publicly available data sources.In sum, I think that examining specific firms targeted for tax avoidance, a specific industry targeted by the largest U.S. protest movement, and cross-sectional variation as a function of income inequality could help solidify 2011 as a setting where tax-related reputational costs are particularly salient.My next comment considers challenges in measuring tax-related reputational costs. The authors use RavenPack data to construct various measures of firm-month negative media sentiment. These media data are firm specific and sentiment specific, but not tax specific.6 My concern is whether the negative media sentiment about a particular firm relates to its income tax-avoidance practices, or to something else. It seems that a media coverage measure that is firm, sentiment, and tax specific would best match the construct DGHS seek to capture.Chen et al. (2019) create such a measure, but for a smaller sample of firms. Similar to DGHS, Chen et al. (2019) employ a Factiva article search to identify media mentions of tax issues. The authors begin their hand-collection efforts with a sample of 500 publicly traded firms (the 250 largest firms from the S&P 500 Index and the 250 largest firms from the S&P 600 Small Cap Index), and search for tax-specific media coverage of these firms in four popular news outlets (The New York Times, Wall Street Journal, The Washington Post, and USA Today) from 1993 through 2015. This search identifies firm- and tax-specific media coverage for 400 firm-years. The authors assess the extent of negative tax connotation in these articles through an iterative process of reading the articles and identifying a custom dictionary of tax-specific words. This process culminates in a media coverage measure that is firm, sentiment, and tax specific.Chen et al. (2019) provide time-series plots of their data by type of tax and sentiment tone. In my view, these plots fail to indicate that 2011 was a stand-out year in terms of firm- and tax-specific media coverage frequency (Chen et al. 2019, Figure 1, Panel A) or negative sentiment regarding this firm- and tax-specific media coverage (Chen et al. 2019, Figure 2). In addition, the most frequent type of tax mentioned in 2011 is sales tax (Chen et al. 2019, Figure 1, Panel A), not income tax, which is the focus of DGHS. Collectively, these figures suggest that the media was not particularly focused on corporate income tax avoidance during 2011.7I acknowledge that a similarly detailed article search for the large sample of firms that DGHS study is impracticable. However, I believe that DGHS could strengthen the internal validity of their study by taking additional steps to ensure that the negative media sentiment used as their primary measure of tax-related reputational costs is tax specific. Otherwise, it is possible that firms avoiding more taxes in 2011 are receiving negative media attention for factors other than tax avoidance. This would mean that a sort on tax avoidance is also a sort on these other factors, and what DGHS label as tax-related reputational costs could in part reflect reputational costs for these other activities.Considering the economic magnitude of the hedge portfolio returns could shed light on whether this concern is warranted. The results presented in Dhaliwal et al. (2019, Table 4) indicate that the hedge portfolio generates an annualized abnormal return between 12 and 15 percent (depending on the specification) during the protest period. Given the average firm's market capitalization during the protest period, this translates to a $1.0 to $1.3 billion decline in market value on average. Is such a large decline in market value as a function of investors' negative perception of tax avoidance during the protest period reasonable? Sorting firms on a media coverage measure that is firm, sentiment, and tax specific would help ensure that this large decline in market value is from tax-related reputational costs and not another type of reputational cost.DGHS examine whether firms incur tax-related reputational costs by focusing on the year 2011—a period of U.S. socioeconomic unrest characterized by widespread protests. The authors present findings consistent with the media and investors imposing tax-related reputational costs on firms, and firms responding to and managing these tax-related reputational costs. My discussion of DGHS focuses on two primary observations: the validity of 2011 as a setting for identifying tax-related reputational costs, and challenges associated with measuring tax-related reputational costs. DGHS contribute to the tax-avoidance literature by furthering our understanding of whether and when tax-related reputational costs exist. More broadly, the paper highlights the need to consider whether, and more importantly why, time-series variation in the relation between dependent and independent variables exists. I applaud the authors for tackling an important research question and introducing key aspects of sociology research into the accounting literature.