Real Costs of Subsidiary Disclosure: Evidence from Corporate Tax Behavior. Scott D. Dyreng Fuqua School of Business Duke University

Transcription

1 Real Costs of Subsidiary Disclosure: Evidence from Corporate Tax Behavior Scott D. Dyreng Fuqua School of Business Duke University Jeffrey L. Hoopes Fisher College of Business Ohio State University Jaron H. Wilde Tippie College of Business University of Iowa May 22, 2014 Abstract: We examine whether an exogenous change in the costs of non-compliance with a mandatory disclosure rule has real effects on firms behavior. In 2010, ActionAid International, a UK-based non-profit, discovered that approximately half of the firms in the FTSE 100 were not in compliance with a U.K. requirement to disclose the name and location of all subsidiaries. The subsequent pressure by ActionAid and the U.K. s Companies House was immediate and severe, resulting in nearly 100 percent compliance with the disclosure requirement. We find that firms that had not disclosed a complete list of their subsidiaries before the ActionAid investigation, but began disclosing a complete list after the ActionAid investigation, subsequently show a rise in corporate effective tax rates (ETRs) and a decrease in tax haven usage relative to other FTSE 100 firms that were in compliance before ActionAid s investigation. The evidence suggests that information made public through compliance with the subsidiary disclosure rules changed the calculus surrounding the costs and benefits of tax avoidance in such a way that the firms altered their real behavior to include less tax avoidance. The result suggests that disclosure can have a significant influence on the real operations of large publicly-traded firms. JEL Codes: H25, H26, H20, G39 We thank Martin Hearson and Chris Jordan of ActionAid International for extensive help understanding ActionAid s investigation of the FTSE 100 firms, and for providing us with data used in our paper. We thank Nick Francis of PwC, U.K. for helping us with technical details of the U.K. tax system. We thank Ken Merkley, Tom Omer, Jeff Pittman, and Steven Savoy for comments on a previous draft of this paper. We also thank Aaron Nelson and for research assistance.

2 1. Introduction Disclosure regulation is often considered a central feature of effective capital markets (Bushee and Leuz 2005; Levitt 1998), and the study of firms information disclosure to capital market participants dates back many decades (e.g., Akerlof, 1970, Demski, 1974; Watts and Zimmerman,1978). Although researchers have long debated the market conditions under which firms will voluntarily provide information to investors versus the conditions under which regulators might need to mandate disclosures (see Beyer et al., 2010 for a review), we have a limited understanding of the actual economic costs and benefits associated with disclosure regulation (Healy and Palepu 2001; Bushee and Leuz 2005). In this study, we use a tax setting to examine real costs associated with mandatory subsidiary disclosure. Specifically, we examine whether an exogenous shock from the enforcement of a U.K. requirement to disclose the location of all firm subsidiaries led to actual changes in firms tax haven subsidiary use and corporate tax avoidance activities. As firms grow in global influence and scope, information related to the location and extent of foreign operations becomes increasingly important for investors as they try to understand firms global activities, opportunities, and risks (e.g., Desai, Foley and Hines, 2009; Dyreng, Hanlon and Maydew, 2012). Because the jurisdiction in which firms choose to locate their operations has significant operating implications and considerable tax consequences (e.g., Dyreng and Lindsay 2009), required subsidiary disclosure can reveal information about firms corporate tax behavior. These subsidiary disclosures are important because the information contained in the disclosures can have significant implications for different stakeholders in the firm, including shareholders, taxing authorities, regulators, and customers. 1

3 In the United States, publicly available information related to geographic operations is limited to data on sales and assets aggregated by arbitrary and broad geographic regions (SFAS 131; Hope et al., 2009; Hope and Thomas, 2008). Firms typically disclose this information in a note to the financial statements related to segment operations, but often aggregate their disclosures in such a way so as to make it difficult to determine the extent of a firm s use of tax havens (Akamah et al, 2014). The Securities and Exchange Commission (SEC) also requires firms listed on U.S. exchanges to disclose the name and country of incorporation of each significant subsidiary in Exhibit 21 of their annual regulatory filing. This information, while valuable, does not paint a complete picture of the nature and scope of firms geographic footprints because only significant subsidiaries must be disclosed and there is evidence that firms selectively disclose subsidiaries (Gramlich and Whiteaker-Poe, 2013). In contrast to the U.S. regulation requiring disclosure of significant subsidiaries, in the United Kingdom the Companies Act of 2006 requires firms filing with the Companies House to disclose the name and location of all subsidiaries, regardless of size or materiality. However, while U.K. law has required U.K. firms to disclose all subsidiaries since 2006, in 2010, ActionAid International, a U.K.-based non-profit dedicated to ending poverty worldwide, discovered that approximately half of the firms in the FTSE 100 did not disclose the name and location of all subsidiaries. 1 ActionAid made this discovery in the process of conducting a study of U.K. firms tax haven usage and ActionAid was not inherently interested in disclosure compliance per se. This finding by ActionAid was prima facie evidence that the Companies House was not enforcing the subsidiaries disclosure requirement from the Companies Act of 1 The FTSE are the 100 largest market cap firms trading on the London Stock Exchange. These 100 firms represent over 75 percent of total market cap traded on the London Stock Exchange. 2

4 2006, and suggests that for some firms the cost of disclosing the detailed information on subsidiary location was greater than the benefit of a more complete information environment. 2 In order to get the data regarding tax haven usage they desired, ActionAid requested that the U.K. Companies House enforce the regulation (see the letter in Appendix A), and pressured firms individually to comply with the regulation. This pressure involved suggesting to firms the possibility of negative publicity of not complying with the disclosure rules and reminding firms of ActionAid s previous success in garnering negative media attention for firms. 3 The increased pressure was sufficient to bring nearly all FTSE 100 firms into compliance with the disclosure rule within two years of the report. The fact that a large proportion of the FTSE 100 firms initially failed to disclose suggests that these firms likely perceived that subsidiary disclosure could be costly. Recent survey evidence suggests that disclosure related to a firm s tax function is a significant challenge because publicity and reputation concerns can change interactions with tax authorities, invite customer retaliation and organized boycotts, affect relationships with government clients, and require firms to reassure key shareholders (EY 2014). 4 The survey found that 89 percent of the largest respondent firms indicate that they are somewhat or significantly concerned about media 2 While unenforced, the Companies Act of 2006 did provide for explicit monetary fines for non-compliance. However, these fines, set at a maximum of 1000 for initial noncompliance (for the company, and each officer in default), would have been inconsequential for the FTSE 100 firms we examine (Companies Act of 2006, Section 410 (4) and 410 (5)). 3 In correspondence with the authors ActionAid explained that they used these tactics. For one example, see For more recent examples of ActionAid s work in this area, see noting Barclay s use of tax havens in Africa, and how it deprives African school children of tax money. This particular attack on Barclays was part of a larger campaign, #taxpaysfor, by ActionAid, in which people posted on Twitter what vital services taxes paid for, and how tax haven use would deprive the government of funds to be put to those purposes. Over 500 tweats reference #taxpaysfor (see 4 ActionAid s 2011 report titled Addicted to Tax Havens garnered significant media coverage, resulting in members of the U.K. Parliament sponsoring and signing two early day motions to press U.K. government officials to confront tax haven use, as well as resulting in abnormal trading market returns for firms targeted in the report (Choy et al., 2014). 3

5 coverage of firms tax activities, with one survey participant stating, There is a far higher threshold for public approval of a tax position than there is when you are dealing with a tax auditor (EY 2014, 6). In this study, we use the change in enforcement and compliance generated by the ActionAid investigations as a natural experiment. We examine whether enhanced subsidiary disclosures rules required by the Companies Act of 2006, and made salient by ActionAid s activities starting in 2010, made tax avoidance more costly by exposing to the public a channel through which tax avoidance takes place, namely the use of subsidiaries located in foreign countries commonly considered tax havens. Specifically, we ask whether the requirement to disclose the location of all subsidiaries sufficiently changed the net costs and benefits of tax avoidance and using tax haven operations and thus altered firms tax avoidance behavior. Using a difference in differences research design, we find that FTSE 100 firms that were forced to make additional subsidiary disclosures (noncompliant firms) avoid less tax after the enforcement change relative to FTSE 100 firms that were not affected by the change (compliant firms). Specifically, our estimates suggest a 3.8 percentage point increase in the effective tax rates (ETRs) of noncompliant firms relative to the tax rates of compliant firms in the years following the enforcement change. We verify this finding in a number of ways. First, we conduct a placebo analysis in which we arbitrarily change the date of the ActionAid event to one of the two years before it actually occurred, and find no evidence of significantly higher ETRs in the pre-enforcement period. Second, to the extent that the enforcement shock increases the costs of tax avoidance, firms can respond by changing their tax positions or by decreasing tax haven use. Thus, the tax avoidance results potentially reflect changes in firms tax positions, but given the heightened scrutiny of tax 4

6 haven subsidiaries, they like reflect decreases in tax haven use. We expect that firms would initially decrease haven use in locations where they would incur high disclosure costs and where it would be relatively easy to close subsidiaries without generating significant operating costs. Our evidence is consistent with these arguments. In particular, we find that noncompliant firms decreased the proportion of their subsidiaries located in tax havens relative to compliant firms, consistent with the increase in ETRs being driven by affected firms forgoing tax avoidance techniques involving tax haven subsidiaries. Similarly, we provide evidence that relative to compliant firms, the decrease in tax avoidance for noncompliant firms in the postenforcement period is stronger in the subsample of firms with a decrease in the percentage of total subsidiaries located in tax haven countries. We also find the decrease in tax avoidance for noncompliant firms in the post-enforcement period is most pronounced in the subsample of firms that experience a decrease in the percentage of total subsidiaries located in small ( dot ) tax haven countries, countries where subsidiaries are unlikely to have operational substance (Desai et al. 2006). 5 Finally, we conduct a number of additional tests to confirm that the results are robust to alternative model specifications, estimation techniques, and outlier correction procedures. All of this leads us to conclude that firms incurred real costs in the form of increased taxes as a result of the increased disclosure requirements. Our research contributes to the literature in a number of ways. First, it responds to calls for studies that investigate the real effects of tax avoidance activities (Hanlon and Heitzman 2010). While much of the work in the tax literature examines economic and managerial determinants of tax avoidance (e.g., Badertscher et al. 2013; Brown and Drake 2014; Dyreng et 5 Desai, Foley, and Hines (2006) delineate between tax haven countries associated with tax benefits, but few operational benefits (i.e., dot havens) and tax haven countries that provide some tax benefits, but also have a sufficiently large workforce to potentially provide operational benefits (Big 7 tax havens). Typically, tax haven subsidiaries located in dot havens suggest a tax-focused motive for the location decision. 5

7 al. 2010; McGuire et al. 2014; Rego and Wilson 2012), this study documents evidence of real costs associated with disclosures that can provide information related to firms corporate tax avoidance activities. In addition, our evidence suggests that activist groups can have a meaningful influence on firm outcomes, improving our understanding of the role of non-traditional monitors in overseeing firms behavior (e.g., Dyck et al., 2010; Miller 2006) This evidence is consistent with other research indicating that firms respond to pressure from external stakeholders (Smith, 1996) and is similar in spirit to research finding that firms respond to non-binding shareholder votes (Ertimur et al., 2011). By examining the effect of exogenous changes in enforcement of mandatory subsidiary disclosure, we provide evidence that builds on recent research by Hope et al. (2013), who find changes in ETRs surrounding voluntary changes in geographic segment reporting requirements in the U.S. We also contribute to the empirical literature examining the economic consequences of disclosure regulation (e.g., Bushee and Leuz 2005; Coffee 1984; Kanodia, 2007) by providing evidence of real economic costs associated with increased mandatory disclosure. Consistent with research examining the role of enforcement in the effectiveness of financial reporting or securities laws (Bushman et al. 2005; Christensen et al. 2013), this study highlights how external party scrutiny can facilitate and encourage enforcement efforts that have significant economic implications for firms. Our results suggest that firms do not always comply with written laws, something the disclosure literature often implicitly assumes. Finally, our study is informative to policy debates regarding the granularity of firm disclosures related to geographic operations and the taxes paid to the governments in which these 6

8 operations are located (e.g., the OECD s initiative on country by country reporting). 6 With growing revenue shortfalls, countries are interested in reducing corporate tax avoidance in order to increase government revenue. Several countries, including the U.S., have increased tax-related disclosure requirements for both public and private information provided by firms (e.g., accounting for uncertain tax positions, Schedule UTP, Schedule M-3, additional 1099 informational reporting, etc.). The growing importance of the location of a firm s activities has given rise to calls for increased geographic disclosures that allow investors to better understand firm activities (Jaworski 2012). Our paper speaks to potential effects of increased geographic disclosure and should be useful as policymakers consider the implications of proposals to expand disclosure requirements for multinational corporations. 2. Motivation and Hypothesis Despite notable advances in our understanding of the benefits of disclosure in terms of lower cost of capital and higher liquidity (e.g., Botosan and Plumlee, 2002; Easley and O hara, 2004), there is still considerable debate surrounding the conditions under which firms will voluntarily disclose information to external parties versus the settings in which it is optimal for regulators to mandate disclosures (e.g., Beyer et al. 2010). Although we have a limited understanding of the economic consequences of disclosure regulation and enforcement (Healy and Palepu 2001; Bushee and Leuz 2005), recent research suggests that disclosing information can change firms real activities. For example, Edmans et al. (2013) model the effect of mandatory disclosure of hard versus soft information and find that disclosure requirements can affect firms decisions, sometimes perversely in that more disclosure potentially leads to underinvestment. Empirically, Link et al. (2011) and Gigler et al. (2012) find that the 6 To view the OECD s most recent report, see 7

9 requirement to disclose quarterly results of corporate activities in form 10-Q induce managerial myopia, leading managers to change real behaviors in order to meet earnings targets at the expense of long-term profitability. Most of the research investigating behavioral responses to mandatory disclosures describes management responses to the additional disclosure as unintended consequences (e.g., Kanodia, 2007). Regulators likely intend additional disclosure rules to provide information to capital market participants and other stakeholders about firms actions but they do not necessarily intend to change the actions of the firm. There are, however, notable exceptions, such as recent mandated disclosures intended to curb corporate tax abuse by forcing firms to be more public about their tax planning strategies. 7 One such case recently developed in Australia when the Australian government passed Tax Laws Amendment (2013 Measures No. 2) Bill 2013, which mandates public disclosure of certain tax return information for firms reporting total income in excess of AUS $100 million (Dummond, 2013). Likewise, French president François Hollande recently promised to introduce new corporate disclosure rules, including a requirement for French banks to disclose all foreign subsidiaries each year (Carnegy and Boxell 2013). The explicit purpose of this requirement is to curb tax planning that involves tax havens. Despite these recent episodes in mandated disclosures focused on reducing tax abuses, little is known about the effect of these disclosures on the tax avoidance activities of the firms they are intended to constrain. One exception is Hasegawa et al (2013), who examine mandated disclosures related to corporate taxes in Japanese public and private firms. Hasegawa et al (2013) find that firms attempted to avoid disclosing tax information, but find no evidence of changes in firms tax avoidance activities. A lack of evidence indicating a change in firm behavior resulting 7 One recent non-tax example is the requirement of the disclosure of conflict mineral usage, mandated under Section 1502 of the Dodd Frank Act. The intent of the disclosure is dissuading companies from continuing to engage in trade that supports regional conflicts (EY 2012), potentially in ways that do not add to shareholder value. 8

10 from the additional disclosure is consistent with the arguments made against tax related disclosures by Blank (2009) and Evers et al (2014), who argue that expanded corporate disclosure calling attention to tax avoidance would be costly and ineffective. 8 Mandated information disclosure related to the income tax function is important to study because the information in the disclosures could have important implications for different stakeholders in the firm, including investors, tax authorities, customers, and policymakers (e.g. Hanlon and Slemrod 2009; Choy et al. 2014). For example, the additional information in tax related disclosures might be useful to shareholders because it helps them understand actions management is taking to reduce tax payments to the government. This additional information may reduce information asymmetry between management and shareholders, thereby reducing the cost of capital or increasing liquidity. On the other hand, that same information may prove useful to the taxing authority in determining where to allocate scarce enforcement resources (Mills, 1998). If the taxing authority can use the information to more efficiently audit firms tax positions, disclosure of information related to the tax function may be costly to shareholders. Other stakeholders may also find the information valuable. For example, customers may prefer to purchase goods and services from providers that they perceive as a good corporate citizen paying its fair share of taxes (e.g., Austin and Wilson 2013). Although there is little empirical evidence to support the claim that reputational concerns influence tax activities (Gallemore et al. 2013), recent survey evidence from tax executives at multinational firms suggests that reputational concerns are a significant determinant of corporate tax planning 8 While little evidence exists regarding corporate tax disclosure and tax avoidance, Slemrod et al (2014) find that the wider dissemination of tax information about the citizens of Norway caused business owners subject to the broadened disclosure to report, on average, three percent more income tax relative to others not subject to the enhanced disclosure. 9

11 decisions (Graham et al. 2014). In addition, consumer boycotts generally have negative implications for firm value (Pruitt and Friedman, 1986). Two recent events in the U.K. illustrate this point. First, Starbucks experience in the U.K. provides anecdotal support for the notion that reputational concerns matter. In 2012, numerous articles in the financial press revealed that Starbucks paid no corporate income tax to the U.K., despite strong growth in U.K. sales. The negative publicity resulted in customer boycotts of Starbucks stores and prompted the company to voluntarily pay future taxes and eventually relocate physical offices to the U.K. in 2014, even though the firm has always allegedly been in compliance with U.K. tax law. 9 Another recent example involves Ethical Consumer magazine s proposed boycott of Amazon, supported by 8 members of the U.K. parliament. Ethical Consumer demanded a boycott based on the fact that Amazon paid only 0.1% of their U.K. sales in U.K. taxes and noted that if enough people boycott Amazon then we will damage their business. Amazon's market share and reputation matters. 10 Much of the public outcry related to corporate tax avoidance is centered on the use of subsidiaries located in tax haven countries. Large multinational firms can strategically structure intra-company transfers, intra-company financing, and ownership of assets so that profits are reported in countries with relatively low tax rates, even though profits may have been economically generated by assets and activities located in countries with high tax rates. Indeed, Markle and Shackelford (2011) and Dyreng and Lindsey (2009) show that firms with significant disclosed subsidiaries located in tax haven countries have lower ETRs than other firms. Despite evidence showing a correlation between operations located in tax havens and ETRs, the conditions under which shareholders value that avoidance are not well understood. 9 See, for example, 10 See 10

12 Desai and Dhharmapala (2006) provide one possible scenario in which shareholders do not value tax avoidance. In the study, they assume that the mechanisms that firms install to hide income from taxing authorities can also be used to hide income from shareholders. Thus, tax avoidance is only desired by shareholders if governance mechanisms prevent managers from expropriating shareholder resources. Hanlon et al (2014) provide evidence consistent with the notion that these tax-related governance failures may manifest themselves in accounting choices. Another possibility is that shareholders only want firms to avoid taxes if the tax avoidance is not made public. If customers believe firms should be paying taxes as part of its responsibility to society, then news of tax avoidance could result in lower sales as customers reduce their purchases from firms perceived as avoiding taxes. Thus, shareholders may not demand that firms voluntarily provide disclosure of information related to tax avoidance. If regulators believe firms disclosures are necessary to both provide information to investors and provide information to the public regarding firms contributions to and use of publicly provided resources (e.g., natural resources, national infrastructure, etc.), then regulators might mandate disclosures related to firms tax functions. Notably, many countries have recently proposed or adopted financial statement disclosures that are directly or indirectly related to the tax function. As mentioned, Australia recently enacted a bill that mandates public disclosure of certain tax return information for firms reporting total income in excess of AUS $100 million (Dummond, 2013). In late 2006, the U.S. expanded its required tax disclosures through the promulgation of FIN 48, giving investors in public firms more information about a firms uncertain tax positions. In addition, the IRS has demanded expanded disclosure and third party information reporting for both large and small firms, through Schedule UTP (Towery, 2014) and 1099-K reporting (Slemrod et al, 2014). Finally, as recently as January 30, 2014, the OECD 11

13 released a discussion draft to tax authorities discussing the possibility of country-by-country reporting standards that would help countries better eliminate aggressive transfer pricing schemes used by international firms (OECD, 2014). While this OECD discussion is only in the early stages, it highlights a belief that increased disclosure may have real effects on both firms tax avoidance behavior and tax authorities ability to enforce the tax law. Whether these disclosure requirements will have a real effect on firms tax avoidance behavior is uncertain, and there is little academic evidence to shed light on the issue. One very useful exception is Hope et al. (2013), who find that firms voluntarily discontinuing disclosure of geographic earnings after the implementation of SFAS 131 had lower ETRs than other firms. Although SFAS 131 still requires sales and assets reporting by geographic segment, it relaxed the requirement for firms to report earnings from geographic segments, (Collins et al., 1997; Klassen and Laplante, 2012; Dyreng and Markle, 2014). Whereas Hope et al. (2013) examine voluntary disclosure changes related to segment reporting, we examine a shift in the enforcement of mandatory subsidiary disclosure. It is unclear how the Hope et al. (2013) evidence is applicable to changes in the enforcement of mandatory subsidiary disclosure, our topic of inquiry, for at least two reasons. First, geographic segment information rarely corresponds to legal jurisdictions that are meaningful for tax purposes. For example, Akamah et al. (2014) note that most firms disclosures offer very limited (if any) information useful in understanding specific geographic operations and the use of structured transactions in foreign countries to avoid taxes. 11 Second, the disclosure of earnings by geographic segment and the definition of the scope of the geographic segment are firm choices, thus, whether exogenous shocks in the enforcement of 11 Indeed, only 0.87 percent of all segments listed in the Compustat Segment data can be associated with a tax haven country, and only 0.45 percent can be associated with dot havens. 12

14 mandatory subsidiary disclosures alters the net costs and benefits of subsidiary disclosure sufficiently to affect tax avoidance behavior is an open empirical question. In this study, we use a change in the enforcement of the Companies Act of 2006 to examine the question of whether mandatory disclosure can affect real firm tax behavior. Section 409 and 410 of the Companies Act of 2006 mandate that all U.K. companies, either in their annual report or annual return, disclose a complete list of all the company s subsidiaries. 12 In 2009 and 2010, ActionAid International, a non-profit dedicated to ending poverty worldwide, examined the annual returns and the annual reports of firms included on the FTSE 100 and found that nearly half of them did not disclose their full subsidiary list. 13 ActionAid began scrutinizing and communicating with FTSE 100 firms in 2010 and subsequently petitioned the Companies House of the U.K. to compel these firms to comply with the legally mandated disclosure requirements (see the Appendix for a copy of the letter, provided to us by ActionAid). These efforts resulted in nearly 100 percent compliance by The widespread non-compliance among the largest firms in the U.K. attracted media coverage and led to an investigation by U.K. Business Secretary Vince Cable (Holms 2011). Moreover, Choy et al. (2014) report that ActionAid s 2011 report, titled Addicted to Tax Havens, attracted significant media coverage and led multiple groups in the U.K. Parliament to sponsor and sign two early day motions to press HMRC, the U.K. s tax authority, to confront U.K. multinationals on their haven use. While disclosing the existence of tax haven subsidiaries is not likely to reveal illegal or distasteful tax avoidance activity per se, it may increase the likelihood that the firms are scrutinized for tax avoidance activities in the public arena, potentially increasing the non-tax 12 The annual report refers to the financial statement report and other information similar to the annual reports of publicly-traded U.S. firms. U.K. law also requires firms to file an annual return, which refers to a specific form that lists information about a firm s directors, address, shares and shareholders. It is not the firm s tax return. The annual return form can be accessed, for a fee, at 13 ActionAid was examining the returns in an effort to gather data that would be used to publically shame U.K. firms for their use of tax shelters, which it subsequently, and successfully, did, after obtaining the data. 13

15 costs of tax avoidance. The fact that firms were not complying with the disclosure requirement before the scrutiny by ActionAid suggests that some firms perception of the cost of disclosing the detailed information on subsidiary location was greater than the benefits of a more complete information environment. These costs could come in many forms, including shareholder penalties, tax authority audits, reputational damage,) and political backlash (Hanlon and Slemrod 2009; Graham et al. 2014; Choy et al. 2014). After the ActionAid scrutiny, increased enforcement by the Companies House effectively removed firms noncompliance. Accordingly, we compare the tax activities of whether firms that were forced to comply with the subsidiary disclosure in 2010 with the tax avoidance activities of firms that were already in compliance with the requirement before the ActionAid report. 3. Research Design, Sample Selection, and Data 3.1 Research Design: Tax Avoidance Tests The exogenous shock in subsidiary disclosure enforcement described in the previous section provides a powerful setting to investigate our research question. Approximately half of FTSE 100 firms did not comply with the required subsidiary disclosure laws prior to the third party scrutiny beginning in Using compliant and non-compliant firms and the pre- and post- scrutiny treatment, we employ a difference-in-differences estimation to identify the effect of mandatory subsidiary disclosures on firms tax planning activities. The difference-indifferences research design allows us to control for time-invariant differences between treatment firms (non-compliant subsidiary disclosure firms) and control firms (compliant subsidiary disclosure firms) as well as for economic trends common to both treatment and control firms. We estimate the following difference-in-differences model: 14

16 ETR = β FE + β 1 Incomplete Subs List + β 2 Post Enforcement + β 3 Incomplete Subs (1) List Post Enforcement + β 4 Size + β 5 Leverage + β 6 Intangibles + β 7 Inventory Intensity + β 8 RD Intensity + β 9 Capital Intensity + β 10 Capex + β 11 Return on Assets + β 12 % Havens + ε, where ETR equals a firm s ETR for the period and Incomplete Subs List is an indicator variable equal to one for FTSE 100 firms that changed their disclosure as a result of the ActionAid request to the Companies House. β 1 reflects the average pre-enforcement difference in ETR between noncompliant and compliant firms. Post Enforcement is an indicator variable equal to one for firm-year observations ending during 2010 (i.e., the period of the initial increase in enforcement) or later and equal to zero, otherwise. β 2 represents compliant firms average difference in ETR between the pre- and post-enforcement periods. We describe all other variables below and in Table 1. The variable of interest in the model is Incomplete Subs List Post Enforcement, and β 3 represents the difference-in-differences estimator, which reflects the effect of subsidiary disclosure scrutiny on the ETRs of non-compliant (i.e., treatment) FTSE 100 firms relative to control firms already compliant with subsidiary disclosure requirements. We also control for various determinants of ETRs documented in prior research to help alleviate concerns that correlated omitted variables are confounding our inferences. In particular, we control for firm size (Size) using the natural log of assets because larger firms potentially have higher political costs (Zimmerman 1983) or greater tax planning opportunities (Rego 2003). Following prior research, we also control for other firm attributes that are potentially associated with firms tax planning activities, including firm leverage (Leverage; equal to long-term debt, scaled by total assets); intangibles (Intangibles; equal to intangible assets, scaled by total assets; inventory intensity (Inventory Intensity; equal to inventory, scaled by total assets); research and 15

17 development intensity (RD Intensity; equal to research and development expense, scaled by total assets); capital intensity (Capital Intensity; equal to beginning net property, plant, and equipment, scaled by beginning total assets); and capital expenditures (Capex; equal to capital expenditures, scaled by lagged assets) (e.g., Chen et al. 2010; Hoopes et al. 2012; McGuire et al. 2012). Finally, we control for pre-tax profitability (Return on Assets; pre-tax income, scaled by total assets) and haven intensity (% Havens; defined as the percentage of total subsidiaries ultimately reported that are located in tax havens, as provided by ActionAid International) given prior work that suggests that firm performance and tax haven usage are associated with ETRs (Rego 2003; Dyreng and Lindsay 2009). We winsorize all continuous variables at the 1 st and 99 th percentiles and we only retain firms with positive pre-tax income, given the confounding effects of negative denominators in ETR regression models (Dyreng et al. 2008). 14 In our tests, we report results with alternative specifications using industry, year, and firm fixed effects and compute t-statistics with standard errors clustered by firm (Petersen 2009). 3.2 Research Design: Changes in Subsidiaries Tests The focus of ActionAid s investigation was to examine FTSE 100 firms subsidiary disclosures in order to publicize their use of tax havens. Given that noncompliant firms did not disclose the full list of their subsidiaries in the pre-enforcement period, it is likely that full disclosure would be costly to these firms. Accordingly, we investigate whether mandatory disclosure enforcement changes result in firms real subsidiary location decisions that are significantly different in the post-enforcement period from firms that are already compliant with 14 For some of the variables that we include in Model (1), we set missing values equal to zero in order to retain observations for our tests (i.e., Intangibles, RD Intensity, and Capex). Results are qualitatively similar (i.e., the coefficient on Incomplete Subs List Post Enforcement retains the same sign and significance) if we estimate Model (1) after separately omitting the variables for which we set missing values to zero or include separate indicator variables equal to one for observations for which we set missing values of equal to zero. 16

18 the disclosure requirements. To investigate this possibility, we examine how noncompliant firms changes in tax haven subsidiaries relative to changes in total subsidiaries compare to compliant firms in the year following enforcement. Specifically, we estimate the following difference-in-differences model: ΔHavens = β FE + β 1 ΔSubs + β 2 Incomplete Subs List + β 3 Incomplete Subs List ΔSubs + (2) Σβ j Controls + ε, 3.3 Sample Selection Our sample consists of all FTSE 100 firm-year observations from 1997 through 2012, which provides a sufficiently long time series to allow us to examine the effects of scrutiny and accompanying enforcement of mandatory subsidiary disclosures. 15 Because subsidiary disclosure scrutiny focused on FTSE 100 firms in 2010, we use the list of FTSE 100 firms in 2010 to identify the relevant firms for the study. We obtain the list of FTSE 100 firms that were compliant (not compliant) with subsidiary disclosure requirements from ActionAid International, and we obtain financial statement data for our model variables from Compustat Global. We require firms to have data for Model (1) in the year prior to, initial year of, and the first year after subsidiary disclosure scrutiny (i.e., for years 2009, 2010, and 2011, respectively). 16 Because our sample consists of a relatively small set of firms, we take various precautions to ensure that a few influential observations are not driving our results. Specifically, in addition to winsorizing model variables at the 1 st and 99 th percentiles, we estimate Model (1) after removing any observation 15 We do not include the small number of firm-year observations have fiscal years ending in The results are qualitatively similar if we include these observations (i.e., the coefficient on Incomplete Subs List Post Enforcement remains positive and significant, p-value < 0.05). Our results are also qualitatively similar if we shorten the sample period, as we report later in the paper. 16 When we relax this requirement and only require observations to have non-missing data for the model, our inferences are unchanged (i.e., the coefficient on Incomplete Subs List Post Enforcement remains positive and significant, p-value < 0.05). 17

19 with a Cook s D outlier statistic in the top two percent of observations, leaving a final sample of 921 firm-year observations, representing 72 unique firms Results 4.1 Sample Descriptive Statistics Table 1 reports the summary statistics for the variables in Model (1). We present the aggregate statistics for the full sample and also report the statistics separately for the firms in the non-compliant and compliant subsamples. During the sample period, non-compliant firms are associated with average ETRs of 28.5 percent, which is significantly lower than compliant firms average ETR of 30.8 percent (p-value < 0.01). Consistent with non-compliant firms engaging in significantly greater tax avoidance activities in the pre-enforcement period, non-compliant firms are associated with average ETRs of 29.3 percent during the pre-enforcement period, which is significantly lower than compliant firms average ETR of 32.5 percent (p-value < 0.01). However, the pattern reverses during the post-enforcement period such that non-compliant firms are associated with higher ETRs relative to compliant firms, average ETRs of 25.5 percent and 24.2 percent, respectively, although the difference is insignificant. We note that in 2009, the year prior to ActionAid s investigation, the U.K. changed the tax treatment of dividends received from non-u.k. firms, which likely gives rise to lower average ETRs for large multinational firms. Given that the adoption of this system is close to the period of enforcement change in our study (2010 and later), we consider the potential effect of such a shift on our results. As discussed in Section 5, the nature of our sample coupled with the evidence from our tests mitigates concerns that the change in the tax treatment of dividends from non-u.k. firms drives our results. Non-compliant firms are different from compliant firms on 17 In Section 7 we report the results of various alternative estimation techniques that we conduct to assess the robustness of our results to influential observations, alternative time periods, etc. 18

20 various other dimensions. In light of these differences, we are careful to take various steps to ensure these differences are not driving our results, including difference-in-differences estimation, placebo analyses in the pre-enforcement period, examination of the parallel trend assumption, and the use of a differences-in-differences specification using firm fixed effects. Table 2 presents correlation coefficients for the Model (1) variables. Consistent with the summary statistics reported in Table 1, non-compliant firms are smaller in size, have a lower percentage of tax havens actually reported, and the incidence of non-disclosure is negatively associated with the other variables in Model (1). 18 We report the results of the multivariate analyses in the next section. 4.2 Multivariate Results Analysis of Subsidiary Disclosure Enforcement and ETRs In Table 3, we present the results of our difference-in-differences regression (Model 1), which estimates the effect of subsidiary disclosure enforcement on firms ETRs. We estimate Model (1) using four specifications. In Column (1), we report results of estimating Model (1) with no fixed effects. To verify that our results are not the result of some industry phenomena or the effect of a simple time trend, we report the results of estimating Model (1) with industry and industry and year fixed effects in Columns (2) and (3), respectively. Finally, we report the results of estimating Model (1) with year and firm fixed effects in Column (4). In Columns 1-4 of Table 3, we find that the coefficient on Incomplete Subs List Post Enforcement is positive and significant (p-value < 0.01), indicating that subsidiary disclosure scrutiny of non-compliant (i.e., treatment) FTSE 100 firms is associated with significantly higher ETRs relative to control firms which were already compliant with subsidiary disclosure requirements. This suggests that even after controlling for industry or time-invariant firm-effects, 18 In untabulated collinearity diagnostics, we do not find evidence to suggest that Model (1) suffers from degrading collinearity problems (e.g., variance inflation factors are no larger than 2.01). 19

21 increased subsidiary disclosure regulation is associated with significantly higher ETRs for noncompliant firms relative to control firms. The most conservative model, the firm fixed effects model in Column 4, estimates a treatment effect of 3.8%. We include firm fixed effects in this difference-in-differences specification to verify that that time-invariant firm-level attributes do not drive our results. The results of this test suggest that the effect of the enforcement of the subsidiary disclosure requirements on the non-compliant firms was to reduce their corporate tax planning such that their ETRs increased by 3.8%. This is an economically large effect. The 34 firms subject to this treatment in 2010 had median pre-tax book income of 618 million pounds. Using a simple calculation, a 3.8% increase in ETR suggests increased tax expense, and associated revenues accruing to HMRC, of 23 million (about $40 million) per firm. 4.3 Parallel Trends Assumption, Placebo Effects, and By-Year Model Estimates A key identifying assumption behind a difference-in-differences estimation is that in the period prior to the treatment both treatment and control firms exhibit parallel trends in the outcome (Roberts and Whited 2012). In our setting, increased subsidiary disclosure scrutiny is the treatment, and the outcome is ETRs. Although this assumption is not formally testable, we conduct falsification tests to examine whether there is evidence inconsistent with the parallel trends assumption. Specifically, we estimate Model (1) after disaggregating the sample firm-year observations surrounding the increased scrutiny period (2010) in order to assess whether noncompliant firms have significantly different ETRs in individual years prior to the enforcement in addition to the years following the increased subsidiary disclosure scrutiny. We estimate the following difference-in-differences model using individual year indicator and interaction variables: 20

22 ETR = β FE + β 1 Incomplete Subs List + β 2 Yr β 3 Yr β 4 Yr β 5 Yr (3) β 6 Yr β 7 Incomplete Subs List β 8 Incomplete Subs List β 9 Incomplete Subs List β 10 Incomplete Subs List β 11 Incomplete Subs List β j Controls + ε, where ETR and Incomplete Subs List are as defined above and in Table 1. We include year fixed effects for and the interaction of these five year fixed effects with Incomplete Subs List in order to examine the trend for these two sets of firms in the years just before the treatment by ActionAid. In Model (2), the interactions of Incomplete Subs List and each of the postenforcement years (i.e., 2010, 2011, and 2012, respectively) comprise the variables of interest. One advantage of estimating Model (2) is it allows us to assess whether there appears to be an inconsistent trend in ETRs between compliant and non-compliant firms in the period prior to the actual enforcement. To the extent that the coefficients Incomplete Subs List 2008 (2009) are positive and significant, it suggests that an alternative factor distinct from subsidiary disclosure scrutiny is driving our results. Similar to our expectations in Model (1), we expect that β 9, β 10, and/or β 11 should be significantly positive to the extent that subsidiary disclosure scrutiny of noncompliant (i.e., treatment) FTSE 100 firms has a persistent effect on the ETRs of non-compliant firms relative to control firms. Table 4 reports the results of our estimates of Model (2). We begin by examining the coefficients on Incomplete Subs List 2008 and Incomplete Subs List 2009, which reflect placebo treatment year observations prior to the actual subsidiary disclosure scrutiny. We find that the coefficients on the coefficients Incomplete Subs List 2008 (2009) are insignificant (pvalue > 0.10), indicating that we find no evidence that non-compliant FTSE 100 firms have a trend in ETRs that is different from the trend exhibited by compliant control firms prior to the 21

23 shock in subsidiary disclosure enforcement. In addition, we find that the coefficients on Incomplete Subs List 2010, Incomplete Subs List 2011, and Incomplete Subs List 2012 are positive and significant (p-value < 0.05), indicating that increased subsidiary disclosure regulation is associated with significantly higher ETRs for non-compliant firms relative to control firms. The fact that we observe positive and significant coefficients of 0.049, 0.048, and on the interaction variables for 2010, 2011, and 2012, respectively, suggests that the effect of increased subsidiary disclosure enforcement on non-compliant firms ETRs relative to control firms is a relatively persistent shift in tax avoidance activities rather than an immediate shift in ETRs that then reverts to pre-enforcement levels. Figure 1 presents the coefficients on the interaction terms for 2008 through Graphing the interaction coefficients from Model (2) allows us to visually observe how differences between non-compliant and compliant FTSE 100 firms differs in the pre- and postenforcement periods after controlling for other factors, industry, and year fixed effects. The parallel trends assumption of difference-in-differences estimation suggests that the trend in the difference-in-differences coefficient estimates should be relatively flat in the pre-enforcement period. Then, to the degree that subsidiary disclosure enforcement is associated with a significantly different effect on ETRs for non-compliant versus compliant firms, one would expect to observe a shift in the trend from the pre-enforcement period to the enforcement period. Consistent with our expectations, Figure 1 shows a relatively flat trend in difference-indifferences coefficients in the pre-enforcement period, with a noticeable positive shift in the post-enforcement periods, reflecting a significant increase in the difference between noncompliant and compliant firms ETRs in the post-enforcement period. 22

24 4.4 Real Changes in Subsidiary Locations Table 6 reports the results of tests examining how noncompliant firms changes in tax haven subsidiaries relative to changes in total subsidiaries compare to compliant firms in the year following enforcement. We focus our tests of Model (2) on changes in the post enforcement period. We use the complete subsidiary disclosure data in 2010 relative to that for 2011, as provided by ActionAid. In Column (1) (Column (2)), we report estimates of Model (2), without controls and without (with) industry fixed effects. In Column (3), we report estimates of Model (2) with controls and industry fixed effects. In each of the tests, we find that the coefficient on Incomplete Subs List ΔSubs is negative and significant (p-value < 0.05), indicating that in the post-enforcement disclosure period, for each additional subsidiary, noncompliant firms are associated with a significantly lower proportion of tax havens relative to compliant firms. 4.5 Changes in Subsidiary Locations and Tax Avoidance Outcomes The results reported in Tables 4-6 indicate that subsidiary disclosure enforcement is associated with tax avoidance activities and real subsidiary location decisions. The relative decrease in tax avoidance for noncompliant firms relative to compliant firms could reflect noncompliant firms responding to enforced subsidiary disclosure requirements by changing their tax positions in existing havens, or by decreasing tax haven use altogether. The results of our tests suggest that noncompliant firms began reporting higher ETRs relative to compliant firms in the first year following enforcement. In order for the decrease in tax avoidance to be associated with the disclosure enforcement, firms would need to be able to change some tax positions within the first year following the enforcement. A recent survey conducted in collaboration with the Tax Executives Institute suggests that firms can unwind a majority of their tax positions within one year. Tax executives of large corporations indicated that: 12 percent of tax positions 23