Credit Ratings and International Tax Planning

I. Introduction

Credit rating agencies are an important source of information for capital market participants, especially given the increased globalization of financial markets. The U.S. corporate bond market reached more than $8 trillion in corporate bonds outstanding in 2015,¹ and many of those bonds were issued by large multinational corporations (MNCs). Prior research shows that credit rating agencies consider overall tax avoidance information (for example, the book-tax difference) when analyzing a company’s credit risk.² To date, however, no study has examined whether there is such an effect for international tax planning activities.³

Unlike purely domestic companies, U.S. MNCs have more opportunities to engage in international tax planning activities, such as locating operations strategically, shifting income between locations, and taking advantage of tax subsidy agreements with host countries.⁴ International tax planning activities, by their nature, can save cash by reducing tax payments. However, these activities may decrease a company’s asset liquidation value, create friction in internal capital markets, and increase the demand for domestic external financing.⁵ International tax planning may also create additional agency and information problems, which makes it different from domestic and general tax avoidance activities. In this study, we investigate whether and to what extent credit analysts consider borrowers’ international tax planning activities when assessing credit risk and assigning a credit rating.

Understanding the effects of international tax planning on bond ratings is timely and relevant. A major component of the Tax Cuts and Jobs Act introduces a one-time repatriation tax reduction for U.S. MNCs, which directly benefits companies with a high degree of international tax planning activities. Anecdotal evidence shows that bond analysts take this into consideration when issuing ratings. For example, Standard & Poor’s changed its rating for Electronic Arts Inc. from BBB- to BBB+, giving the following explanation for the rating upgrade: “Due to the recent U.S. tax reform, the company will be able to repatriate cash to the U.S. more efficiently.”⁶ However, a recent report by S&P states that rating changes caused by the TCJA will not be significant because companies will not use much of the extra cash for permanent debt reduction, leaving current leverage levels unchanged.⁷ It is unclear whether credit analysts use the information from international tax planning, because the rating criteria descriptions provided by rating agencies do not identify it as a credit risk factor.

The TCJA includes several features that change companies’ incentives for tax planning and profit shifting, including the lowered U.S. corporate tax rate (from a maximum of 35 percent to flat 21 percent) and new international provisions. However, as shown by Kimberly Clausing,⁸ the problem of international tax planning will continue to loom large following the passage of the TCJA. For example, a reduction of the tax rate to 21 percent might not be enough to encourage inbound profit shifting because most profit-shifting activity (95 percent in 2015) occurs in countries with tax rates below the global minimum tax rate. Thus, knowing whether and how credit analysts will view the new level of international tax planning — a net increase or decrease to companies’ creditworthiness — is still valuable under the new tax system. Further, the future of corporate taxation and the taxation of foreign profits in the United States has been uncertain leading up to the 2020 presidential election. The incentives to engage in more international tax planning may change with the next administration.

In this study, we empirically investigate the response of credit rating agencies to the international tax planning activities of U.S. MNCs. We use a measure of the difference between the statutory tax rate and the effective tax rate based on the 2017 work of Michelle Hanlon, Rebecca Lester, and Rodrigo Verdi,⁹ as well as a 2017 study by Andrew Bird, Alexander Edwards, and Terry Shevlin,¹⁰ to capture the degree of a borrower’s international tax planning — that is, the money a company has saved by engaging in different types of international tax planning activities. Even though a high degree of international tax planning might result in a company saving more cash, leading to a lower level of default risk, it decreases the liquidation value of the company’s assets and incurs possible future tax cash outflows and liquidity problems. Further, a high degree of international tax planning may also be indicative of high agency costs and low data quality — issues also relevant to bond rating agencies. Together, the relationship between international tax planning activities and credit ratings is unclear and therefore an empirical question.

Using credit ratings for 1,256 companies from 1987 through 2015, we find that credit rating agencies issue lower ratings to companies with a higher degree of international tax planning activities. This effect is mitigated when the credit rating agency has a greater conflict of interest in relation to the company receiving the rating. We identify possible future cash flow effects, agency costs, and information risk as three potential channels through which international tax planning might affect credit ratings, although the specific issues are different from those of domestic or overall tax avoidance activities. We find that the credit rating effect is more pronounced for companies that have a higher likelihood of repatriating earnings, poor governance, and a poor information environment.

In 2004 the United States implemented the American Jobs Creation Act (AJCA), which allowed U.S. MNCs to repatriate their foreign earnings at a reduced tax rate. The AJCA did not affect companies with domestic-only operations, and it provided a larger benefit for companies with high international tax planning activities, leading to variation stemming from international tax differences. We use the AJCA as a quasi-natural experiment of the benefits of international tax planning, and we find that the credit ratings of companies with extensive international tax planning activities increase over the period covered by the AJCA, compared with their peers.

Our results hold after controlling for measures of company-level overall tax avoidance,¹¹ and we find that the effect we document exists in both companies with low and high levels of overall tax avoidance. This result suggests that our measure of borrowers’ international tax planning activities contains information that is incrementally useful to credit rating agencies beyond that in overall company-level tax avoidance. Our results are robust to different international tax planning measures and to controlling for geographic concentration, investment opportunities, liquidity needs, and deferred tax assets and liabilities.

Our study makes several contributions to the literature. First, we add to the understanding of what tax information is used in the credit rating process.¹² Given that 89 percent of S&P 500 companies operate abroad and, on average, report material subsidiaries in 19 countries, with 49 percent of their pretax earnings in foreign subsidiaries,¹³ it is surprising that there is no study examining how bond analysts react to variations in international tax planning. We show that credit rating agencies consider international tax differences when setting the ratings for U.S. MNCs and that this factor is different from overall tax avoidance. We view these results as providing a partial answer to calls for further research — in papers by Hanlon and Shane Heitzman;¹⁴ and John R. Graham, Jana S. Raedy, and Douglas A. Shackelford¹⁵ — to demonstrate how market participants perceive the tax information in companies’ financial statements. These results also contribute to the understanding of how differences in tax laws and financial reporting systems affect the credit ratings of U.S. companies with significant foreign interests.

Second, our study contributes to the literature investigating the consequences of trapped foreign earnings (cash holdings).¹⁶ John L. Campbell et al. find that the value of an incremental dollar of cash holdings is decreasing in the level of foreign cash holdings because of the potential tax liability.¹⁷ Michelle L. Nessa, Shevlin, and Ryan Wilson observe a significant difference in the earnings response coefficient on changes in foreign earnings of companies with low versus high average foreign tax rates,¹⁸ and Lisa De Simone and Lester provide evidence that U.S. MNCs with large foreign cash holdings are also less likely to access domestic lending markets.¹⁹ Our study provides evidence that credit rating agencies impound international tax planning activities into credit ratings and that U.S. MNCs bear additional financing costs, through a lower credit rating, to avoid U.S. tax. We thus document a significant cost of foreign tax planning, which helps us better understand the equilibrium (cost-benefit) levels of tax avoidance.

Third, our study provides insight into the growing literature on the incentive problems that the issuer-pays model creates for credit rating agencies. Investors and the public expect ratings to provide a reliable and timely measure of a debt issuer’s “ability and willingness to meet its financial obligations,” but empirical and anecdotal evidence suggest credit ratings do not always provide an accurate measure of credit risk.²⁰ Moreover, there is growing concern about the incentive problems facing credit rating agencies arising from the issuer-paid model, especially in the period following the financial crisis, as referenced in the Dodd-Frank Act.²¹ We demonstrate that credit agencies indeed ignore some useful information when facing a higher conflict of interest.

II. Hypothesis Development

Credit rating agencies play a fundamental role in debt markets. In a survey by Graham and Campbell R. Harvey,²² CFOs state that credit ratings are the second most important determinant of corporate debt policy. Credit ratings are important to borrowers because debt contracts — and most important, interest rates — are frequently determined by a borrower’s credit rating.²³ However, market participants do not fully know the information used by credit rating agencies to determine and issue credit ratings. After the corporate scandals of the early 2000s, section 702(b) of the Sarbanes-Oxley Act of 2002 called for a report on the role and function of credit rating agencies in the operation of the securities market. Soon thereafter, in 2003, the SEC stated that “the marketplace needs to more fully understand the reasoning behind a ratings decision and the type of information relied upon by rating agencies in their analysis.”²⁴

A growing number of studies shows that public and private information, financial information, accrual quality, information quality, accounting conservatism, book-tax difference, earnings quality, and off-balance-sheet financing may all be used by rating agencies in the analysis and rating recommendation.²⁵ However, we know little about how differences in tax rates and tax systems in a global setting will affect the assessment made by credit rating agencies. Shedding light on these issues is important given the increasing prevalence of large companies operating in a variety of foreign countries while also accessing global credit markets.²⁶

In the past few decades, U.S. MNCs have frequently engaged in international tax planning activities such as locating operations strategically, shifting income between locations, and taking advantage of tax subsidy agreements with host countries.²⁷ During our sample period, they also delay the repatriation of foreign earnings because of the requirement to pay any difference between the U.S. statutory tax rate and the foreign tax rate when repatriating.²⁸ Thus, companies engaging in more international tax planning activities will have more foreign cash holdings and might incur more repatriation tax.²⁹ The economic significance of the unrepatriated foreign cash caused by financial reporting and tax incentives of U.S. MNCs is large. According to Capital Economics, U.S. MNCs held approximately $2.5 trillion in cash overseas in 2016, an increase of 20 percent over the prior two years, and the largest amount ever recorded.³⁰

The TCJA included several features that change the incentives for profit shifting, such as lowering the corporate income tax rate from 35 percent to 21 percent, and new international provisions. Theoretically, companies will have incentives to shift income into lower-tax jurisdictions if the foreign tax rates fall below U.S. rates. As shown by Clausing,³¹ the problem of international tax planning continues to loom large after the passage of the TCJA. For example, a reduction of the U.S. corporate tax rate to 21 percent may not be enough to encourage inbound profit shifting because there continue to be lower statutory tax rates in other countries and company-specific tax arrangements with foreign tax authorities. We thus expect our research question to be relevant under the new tax system in the United States.

There are several reasons to expect that international tax planning will affect a company’s credit rating. First, successful international tax planning increases current-period after-tax earnings and cash flows, thus reducing the risk of default. But it also increases the uncertainty of future cash flows, affecting a company’s leverage and liquidity. Under U.S. international tax laws and financial reporting rules in our sample period, there is a cash payment for repatriation taxes and a reduction in reported accounting earnings for companies when they repatriate foreign earnings.³² The increased tax risk caused by the heightened probability of tax authority audits and penalties also increases the uncertainty of future cash flows.³³

Further, engaging in international tax planning activities and leaving foreign earnings abroad also decrease a company’s asset liquidation value, create friction in domestic capital markets, and increase domestic financing needs.³⁴ For example, Cadence Design Systems Inc., a company with significant operations outside the United States, stated in its 2017 Form 10-K, “If our U.S. cash were insufficient to meet our future funding obligations in the United States, we could be required to seek funding sources on less attractive terms, which could negatively impact our results of operations, financial position and the market price of our common stock.”

Second, extensive international tax planning may be indicative of the agency risk between management and company stakeholders. Recent research argues that tax avoidance activities can facilitate managerial opportunism,³⁵ and that the issue is unique and more serious in an international setting because the complex tax structures of U.S. MNCs create opportunities for rent extraction by managers.³⁶ For example, Hanlon, Lester, and Verdi provide evidence that MNCs with greater repatriation tax costs make worse foreign acquisitions.³⁷

Further, to the extent that large foreign cash balances have been built up by MNCs seeking primarily to avoid corporate tax (as opposed to other nontax purposes), the disclosure level and financial statement quality may be lower. Prior research provides mixed evidence on the amount of overall tax avoidance and the quality of disclosure and reporting.³⁸ The relationship is much clearer in the case of international tax planning. For example, companies are required to disclose cumulative foreign earnings designated as permanently reinvested and the associated tax liability. However, Benjamin C. Ayers, Casey M. Schwab, and Steven Utke find that 77 percent of companies either do not disclose this tax liability or state that it is impractical to calculate.³⁹ The post-implementation review report on Statement of Financial Accounting Standards No. 109 concluded that the information provided in the financial statements is not detailed enough for investors to fully understand the tax consequences of repatriating foreign earnings held abroad.⁴⁰ As a result, the Financial Accounting Standards Board began requiring additional disclosure related to foreign earnings in 2015.

Even though credit rating agencies do not provide a comprehensive description of the data they review in assessing financial statement quality, they all state that low data quality would lead to lower ratings, and prior studies find that rating agencies factor risk into the ratings processes.⁴¹ For example, S&P, in its 2006 rating criteria, says that “qualms about data quality (dubbed ‘information risk’) would translate into a lower rating.”⁴² Similar statements can be found in materials available for Moody’s and Fitch.⁴³ In sum, these factors are all likely to be important considerations for a credit rating agency. This leads to the following empirical hypothesis:

Credit ratings will be decreasing with the international tax planning of a firm.

On the other hand, companies may never need to pay taxes if they can access foreign earnings and cash through complex tax planning strategies.⁴⁴ For example, Xiumin Martin, MaryJane Rabier, and Emanuel Zur find that companies brought home approximately $12 billion a year tax-free from 1990 to 2004 by taking advantage of complex tax-advantaged reorganizations.⁴⁵ Companies can also use future tax reform or a new tax holiday to repatriate foreign earnings without incurring significant costs. For example, after enactment of the TCJA, Apple announced its intention to repatriate hundreds of billions of dollars to the United States while paying only approximately $38 billion in taxes on the money, far less than the prevailing tax rate would have implied in taxes owed over our sample period.⁴⁶ If MNCs are able to access foreign cash without paying onerous taxes, it is unclear to what extent credit rating agencies will discount borrowers’ foreign earnings held overseas and how this will map into credit ratings.

III. Research Design and Sample Selection

A. Research Design

To empirically test whether higher degrees of international tax planning activities are associated with less favorable credit ratings, we estimate the following model:

Credit Rating_i,t= β₀ + β₁ * ITP_i,t-1 + ∑ β_i * Controls_i,t-1 + ε (1)

In this equation, Credit Rating is the long-term issuer credit rating, which is the rating typically used in prior work.⁴⁷ S&P’s long-term issuer credit rating is an opinion related to a company’s ability to meet its financial obligations as they become due.⁴⁸ Specifically, the rating is based on the default risk of the company and not on a specific issued instrument. The overall creditworthiness is evaluated. We map letter ratings ranging from AAA to D into a numeric scale Credit Rating ranging from 21 to 1 (for example, AAA = 21, AA+ = 20, and C and below = 1). International tax planning (ITP) is calculated as the maximum of 0 and pretax foreign income (PIFO) multiplied by the U.S. statutory corporate tax rate (35 percent after 1993, 34 percent otherwise) less any foreign tax paid (TXFO) and scaled by total assets (AT).⁴⁹ This variable is calculated using publicly available data from companies’ financial statements, and it captures the incremental U.S. tax due when earnings (cash) are repatriated from foreign subsidiaries during our sample period.

We use this measure as a proxy for a company’s international tax planning activities for several reasons. First, intuitively, this measure captures the money U.S. multinationals save in taxes by delaying (and potentially permanently avoiding) payment to the U.S. government by engaging in international tax planning activities during our sample period. Second, this measure is a company-year-level measure that aggregates the effects of different types of international tax planning activities. Prior literature has shown that this proxy is positively related to cash holdings overseas and highly correlated with measures created using Bureau of Economic Analysis data.⁵⁰ Third, even under the new tax system, the measure, “statutory-minus-effective-tax-rate difference” is still valid to capture the benefit of international tax planning activities, even though the detailed calculation (such as the statutory tax rate) might change.

We control for company-level factors related to credit ratings and other variables of interest from prior literature.⁵¹ All continuous variables are “winsorized” at the 1 percent and 99 percent levels to mitigate any outlier effect.⁵² Regressions include year and industry (Standard Industry Classification two-digit code) fixed effects, and standard errors are heteroscedasticity robust and clustered at the company level. Included variables are detailed in each test, and detailed variable definitions are included in the Appendix.

B. Data Source and Sample Selection

We obtain the full set of S&P long-term issuer credit ratings and financial accounting data from S&P’s Compustat North America Fundamental database.⁵³ We first delete company years without PIFO, without TXFO, or without positive pretax income because their income-shifting incentives are more difficult to reliably estimate.⁵⁴ We delete companies in the financial services and utilities industries because those industries are not analyzed by S&P under the same rating criteria. Lastly, we delete issuers with credit ratings of R and NR, representing companies under regulatory supervision because of their financial condition and companies that are not rated, respectively. After eliminating observations with missing values of our control variables, our final sample includes 9,500 credit ratings over the period 1987 to 2015.

Table 1 shows the distribution of ratings over our sample period. Rating frequency increases from 1994 to 1997 and from 2002 to 2005 related to strong macroeconomic conditions. In contrast, rating frequency decreases are significant in 1998, 2001, and 2008 after major financial crises.⁵⁵

C. Descriptive Statistics

Table 2, Panel A presents the descriptive statistics of companies in our sample. These companies have a mean issuer rating of 13.17, which represents a rating slightly above BBB. The average of ITP in our sample is 0.42, which represents 0.42 percent of average company assets.

Table 2, Panel B presents the correlation matrix of the variables in our sample. As expected, many of the control variables are significantly correlated with international tax planning, credit ratings, and each other. Credit Rating is positively correlated with financial transparency, abnormal accrual, company size, interest coverage, return on assets (ROA), capital intensity, and intangibility, and it is negatively correlated with whether the company had a loss, book-to-market ratio, standard deviation of ROA, standard deviation of daily stock returns over the past year, and the level of leverage.

IV. Empirical Results

A. Main Results

Table 3 presents the ordinary least squares (OLS) regression and ordered logit regression results of Equation 1. We regress Credit Rating on international tax planning and a set of control variables. We provide both OLS and ordered logit regression to demonstrate that the results are qualitatively similar as shown by Robert S. Kaplan and Gabriel Urwitz,⁵⁶ although ordered logit models are most appropriate for ordinal variables such as Credit Rating. We estimate each regression using the full sample and also using the subsample of borrowers that have positive international tax planning. Our hypothesis predicts that international tax planning will be viewed by credit analysts as increasing a borrower’s credit risk and, in turn, associated with less favorable ratings. The full sample results reported in columns 1 and 3 provide the statistically significant coefficients of -0.22 and -0.21, respectively, on ITP. The economic significance of the coefficients indicates that one standard deviation of change in ITP will lead to a change in Credit Rating of approximately -0.15, translating into a 0.15-notch-rating decrease in a borrower’s credit rating. Many of the included control variables are statistically significant. Credit ratings are negatively associated with accruals, book-to-market ratio, the standard deviation of ROA over the prior three years, standard deviation of stock returns over the past year, and financial leverage. Credit ratings are positively associated with financial transparency, company size, and ROA.

B. Conflict of Interest

A growing body of theoretical and empirical literature focuses on conflicts of interest of bond rating agencies⁵⁷ and shows that the quality of agency ratings is reduced when adverse incentives are most pronounced (for example, as a result of the issuer-pays operating model, long-term business relationships, or reduced bargaining power). The criticisms center on the conflicts of interest caused by the issuer-pays business model prevailing in the credit rating industry: Because credit rating agencies are paid by the issuers of the securities they are assessing, they have a strong incentive to assign overestimated ratings to retain current clients and attract new ones.

Studies by Han Xia and Günter Strobl⁵⁸ and Jess and Kimberly J. Cornaggia⁵⁹ demonstrate that these conflicts of interest arising from the issuer-pays credit rating model lead to credit rating inflation, and several studies have provided evidence that a variety of factors affect the conflict of interest. For example, Thomas Mählmann;⁶⁰ Jie He, Jun Qian, and Philip Strahan;⁶¹ Sumit Agarwal, Vincent Y.S. Chen, and Weina Zhang;⁶² and Xia and Strobl⁶³ show that credit rating inflation is exacerbated when a company has a longer relationship with a credit rating agency and when a company has a higher proportion of short- versus long-term debt. Their reasoning regarding the proportion of a borrower’s short-term debt is that the higher a company’s short-term liquidity needs, the more likely it will be to issue debt in the near future. The prospect of earning additional fees gives credit rating agencies an incentive to issue favorable credit ratings to obtain the company’s future business. Further, Bo Becker and Todd Milbourn find that the increased competition from Fitch over the past decade has resulted in more issuer-friendly and less informative ratings from S&P and Moody’s.⁶⁴ Consistent with these prior studies, we predict that the effect of international tax planning on credit ratings will be mitigated when a company has a longer history with a credit rating agency and a larger amount of short-term debt, and when Fitch also provides a credit rating for the company.

Motivated by evidence in the credit ratings literature that documents a conflict of interest for bond rating agencies,⁶⁵ we modify our model using indicator variables to partition our sample based on three characteristics that, according to prior research, indicate a higher likelihood of a conflict of interest: the length of the relationship between the credit rating agency and the borrower, the amount of short-term debt, and whether Fitch also provides a credit rating.⁶⁶

In the first model, we add the variable HighRelation to test the effect of a long-standing relationship with S&P, and we analyze its interaction with ITP. We find that a longer relationship between a company and S&P decreases the coefficient on ITP. Specifically, in the full sample result shown in column 1 of Table 4, Panel A, the statistically significant coefficient on the interaction between HighRelation and ITP is 0.18. This result is consistent with the findings of Mählmann,⁶⁷ and Agarwal, Chen, and Zhang,⁶⁸ which show that a longer relationship between a company and S&P leads to more favorable ratings. We interpret our finding as evidence consistent with S&P being less willing to fully impound the effect of international tax planning into credit ratings, but it could also indicate that S&P has private information about a borrower’s true costs of future earnings repatriation.

We also investigate the effect of a higher proportion of short-term debt, which has been documented as an another indicator of a conflict of interest.⁶⁹ We analyze the interaction between a company’s amount of short-term debt (HighSDebt) and international tax planning and report the results in Table 4, Panel B. Similar to the prior result, we find that the coefficient on the interaction is significantly positive (0.27), consistent with a conflict of interest inducing S&P to underweight the effect of international tax planning on credit ratings.

Our third measure of conflict of interest is related to whether Fitch issues a credit rating for the borrower. Becker and Milbourn;⁷⁰ Dion Bongaerts, K.J. Martijn Cremers, and William N. Goetzmann;⁷¹ and John M. Griffin, Jordan Nickerson, and Dragon Yongjun Tang⁷² document that credit ratings are more likely to be inflated when credit rating agency competition increases. We create an indicator variable for whether Fitch also gives a borrower a credit rating (Fitch), and we analyze the interaction of this variable with ITP, reporting the results in Table 4, Panel C. Again, consistent with the prior conflict of interest results, we observe a statistically positive coefficient on this interaction term (0.26). Taken together, we interpret the results in Table 6 as providing evidence that the effect of international tax planning is muted in the presence of a conflict of interest by the credit rating agency.

C. Possible Channels

We propose three possible channels through which rating agencies will regard the information as useful: future possible cash flow effect, agency costs, and information risk. Generally, large companies or companies with fewer financial constraints will be less likely to need to repatriate foreign earnings held abroad, thus incurring the tax cost.⁷³ Companies with foreign bank lending can use the money, paying foreign lending directly, avoiding the repatriation of foreign earnings. Companies with more transparent disclosure or better governance are less likely to face information and agency risks.

Specifically, we first analyze the interaction of international tax planning and measures of a borrower’s financial constraint, company size, and foreign lending in our model. Financial constraint is measured using the HP Index as proposed by Charles J. Hadlock and Joshua R. Pierce.⁷⁴ The sample is classified as HighHP when the HP Index is above the sample year median. HighSIZE is an indicator variable equal to 1 if SIZE for company i in year t-1 is above the sample year median, and zero otherwise. FOREIGN is an indicator variable equal to 1 if there is foreign bank lending for company i in year t-1, and zero otherwise. We present the results in Table 5. The interaction terms are all statistically significant, indicating that companies that are less financially constrained, larger companies, and companies with a foreign lender experience a smaller negative effect of international tax planning on credit ratings. We view these results as being consistent with credit rating agencies’ view of increased risks associated with international tax planning.

We next investigate how analyst coverage and company governance influence the effect of international tax planning. In Table 6, panels A and B, we provide evidence that companies with higher analyst coverage and with more independent directors (a proxy for governance) are also less negatively affected by international tax planning in their credit ratings. In Table 7, panels A and B, we provide evidence that companies that have better financial transparency and that issue management forecasts (another proxy of a company’s information environment) are less negatively affected by international tax planning in their credit ratings. Overall, the results in these tests provide evidence that the negative effect of international tax planning on credit ratings is exacerbated in poor information environments and poor-governance companies. For these types of companies, the effect of international tax planning on creditworthiness is higher, consistent with higher investor uncertainty about the need for and likelihood of repatriating foreign earnings in these types of companies.

D. Endogeneity

As noted, United States implemented the AJCA in 2004. As part of the act, U.S. MNCs were granted the option to repatriate their foreign earnings at a reduced tax rate. The AJCA created a temporary tax holiday that reduced the U.S. tax rate from 35 percent to 5.25 percent on repatriations from foreign subsidiaries. Given that MNCs were allowed to repatriate their permanently reinvested foreign earnings at a reduced tax rate, companies with high international tax planning benefit more during this period (a more recent case being Apple taking advantage of a similar TCJA provision, as discussed earlier). We use the AJCA as a quasi-natural experiment related to the benefits of international tax planning, and we predict that the negative effect of international tax planning on credit ratings will be smaller during the period covered by that act.

Table 8 presents the regression results of the difference-in-differences model.⁷⁵ We report the results using an unbalanced and balanced sample (when we (do not) require observations before and after the event for a particular company) in panels A and B, respectively. AJCA is an indicator variable for the AJCA period, and it is zero for the two years before and after the event period. In both specifications, we find that the coefficient on the interaction between an indicator variable for the 2004 act (AJCA) and ITP is significantly positive. These results are consistent with a smaller effect of international tax planning on credit ratings during the AJCA period, as predicted.

V. Additional Analyses

As suggested by prior research, tax avoidance behavior can be identified and factored into credit ratings.⁷⁶ However, the effect of international tax planning can be different from overall tax avoidance for several reasons.⁷⁷ First, the effect is unique to MNCs. Therefore, credit rating agencies need to pay special attention to this factor when lending to MNCs. The report provided by S&P shows that credit analysts indeed consider it a separate factor. Second, companies may have several strategies for avoiding tax, and there is a trade-off between different methods. A company with high international tax planning need not have a high overall tax avoidance level. Third, as we discussed in the hypothesis development section, the level of foreign earnings held abroad may deliver unique information related to governance or information risk beyond a company’s overall tax aggressiveness.

To differentiate international tax planning from overall tax avoidance, we add three tax avoidance measures to our main regression as additional control variables, and we report the results in Table 9, Panel A.⁷⁸ The results indicate that our findings continue to hold after the inclusion of these overall tax avoidance measures. In Table 9, Panel B, we partition our sample at the median of each of the overall tax avoidance measures. We do not observe a statistically significant difference on the coefficient of ITP across partitions, further suggesting that the international tax planning effect on credit ratings is incremental to overall tax avoidance.

Although our main international tax planning proxy is measured annually, the relationship holds for other measures of international tax planning. Following Edwards, Todd Kravit, and Ryan Wilson,⁷⁹ we use a three-year cumulative measure called ITP_3 Year to measure the difference in U.S. taxes owed on foreign earnings and foreign taxes paid, summed over the prior three years (assuming a 35 percent U.S. tax rate).⁸⁰ The variable is calculated as the sum of PIFO multiplied by 35 percent, minus the sum of TXFO paid, scaled by total assets and multiplied by 100 to be presented as a percentage. We also follow Klassen and Laplante to calculate the tax rate difference between the United States and other countries to proxy the incentive of income shifting,⁸¹ one activity of international tax planning. We use this new measure as our independent variable of interest and rerun our main test. The coefficients on ITP_3 Year and foreign tax rate (FTR) are negative and statistically significant for both the full sample and the subsample.

Last, our results are robust to controlling for geographic concentration, investment opportunities, and liquidity needs, and to including deferred tax assets and liabilities (results untabulated).

VI. Conclusion

In this study, we examine the effect of international tax planning activities on borrowers’ credit ratings. We predict that such activities are negatively associated with credit ratings. Moreover, we predict that the effect on credit ratings will be larger when a borrower is more likely to need to access the foreign earnings held abroad (that is, it is financially constrained), its information environments are poor, and its governance is weak, and we predict that the effect is also related to the conflict of interest of the credit rating agency. Our empirical results support these predictions and are significant in their economic magnitude. Specifically, a one-standard-deviation increase in our measure of international tax planning leads to a 15 percent decrease in a borrower’s credit rating. Given the widespread use of credit ratings in the market, especially by U.S. multinationals, the economic significance is substantial.

Our results are robust to a difference-in-difference research design using the AJCA as an exogenous shock. Our findings are robust to alternative measures of our variable of interest, as well as to controlling for several overall tax avoidance measures from prior research. Overall, our empirical results contribute to our understanding of the association between the U.S. tax system and the credit ratings of U.S. multinationals.

Given the global economy, differing tax regimes, and the expansion of U.S. companies into foreign markets, our results will be of interest to investors, analysts, and regulators. Our findings may be especially pertinent given the changes made by the TCJA. Thus, answers to whether and how credit analysts will view international tax planning, and the effect on the credit ratings of U.S. MNCs, is still valuable and perhaps even more important under the new tax system. Moreover, the public debate about corporate taxation in the United States continued leading up to the 2020 presidential election, and the incentives to engage in more international tax planning may change with the next administration.

VII. Appendix: Variable Definitions

Variables	Definition
RATE	Recoded S&P’s long-term issuer credit ratings in the range of 1-21, where 21 represents AAA and 1 represents default.
ITP	Calculated as the higher of zero or PIFO, multiplied by the U.S. statutory corporate tax rate (35 percent after 1993, 34 percent otherwise) less any TXFO, and scaled by total assets (AT). This variable is multiplied by 100 for ease of interpretation.
TRANSP	Financial transparency, measured as negative 1 times the squared residual from the cross-sectional regression ARET = b0 + b1(NIBX) + b2(LOSS) + b3(NIBX * LOSS) + b4(ΔNIBX) + e, where the regression is estimated for all companies within a three-, two-, or one-digit SIC code (conditional on having at least 10 companies in each SIC group) for a given year, and ARET = the market-adjusted return over the fiscal year, NIBX = net income before extraordinary items scaled by the beginning-of-year market value of equity, LOSS = 1 if NIBX is negative, zero otherwise, and ΔNIBX = change in net income before extraordinary items scaled by the beginning-of-year market value of equity (see Mei Cheng and K.R. Subramanyam, “Analyst Following and Credit Ratings,” 25 Cont. Acct. Res. 1007 (2008)).
ACCRUAL	Abnormal accruals are estimated using the cross-sectional Jones model (Jennifer Jones, “Earnings Management During Import Relief Investigations,” 29 J. Acct. Res. 193 (1991)).
SIZE	The natural log of AT.
LOSS	1 if net income before extraordinary items is negative, zero otherwise.
COVER	Interest coverage, determined as the ratio of operating income before depreciation to interest expense.
ROA	Return on assets measured as NIBX over AT.
BMRATIO	Book-to-market ratio, which is the ratio of the firm’s book value of equity at the end of the year to its market value of equity.
ROASTD3	Standard deviation of ROA over the prior three years.
STDRET	Standard deviation of daily stock returns over past year.
LEV	Ratio of long-term debt plus short-term debt over AT.
CAPINT	Property, plant, and equipment net of depreciation deflated by AT.
INTAN	Research and development expense plus advertising expense scaled by AT.
HP Index	Hadlock and Pierce (supra note 74) financial constraint index, constructed as -0.737 * ln(AT) + 0.043 * ln(at)2 - 0.040 * Age, where Age is the number of years the company is listed with a non-missing stock price on Compustat.

VIII. Tables

Table 1. Sample Distribution by Year

Fiscal Year	Frequency	Percent	Fiscal Year	Frequency	Percent
1987	245	2.58	2002	310	3.26
1988	242	2.55	2003	353	3.72
1989	234	2.46	2004	414	4.36
1990	223	2.35	2005	421	4.43
1991	204	2.15	2006	412	4.34
1992	219	2.31	2007	419	4.41
1993	215	2.26	2008	371	3.91
1994	278	2.93	2009	368	3.87
1995	309	3.25	2010	445	4.68
1996	337	3.55	2011	458	4.82
1997	355	3.74	2012	436	4.59
1998	313	3.29	2013	462	4.86
1999	330	3.47	2014	454	4.78
2000	352	3.71	2015	26	0.27
2001	295	3.11	Total	9,500	100

Table 2. Descriptive Statistics

Variables	N	Mean	Std.	Q1	Median	Q3
Panel A: Descriptive Statistics
RATE	9,500	13.17	3.26	11.00	13.00	16.00
ITP	9,500	0.42	0.70	0.00	0.11	0.56
TRANSP	9,500	-0.10	0.23	-0.09	-0.03	-0.01
ACCRUAL	9,500	0.07	0.74	-0.06	0.00	0.07
SIZE	9,500	8.33	1.37	7.39	8.22	9.23
LOSS	9,500	0.02	0.13	0.00	0.00	0.00
COVER	9,500	17.53	31.21	5.31	9.07	16.03
ROA	9,500	0.06	0.04	0.03	0.06	0.09
BMRATIO	9,500	0.45	0.30	0.25	0.40	0.60
ROASTD3	9,500	0.03	0.03	0.01	0.02	0.03
STDRET	9,500	0.02	0.01	0.02	0.02	0.03
LEV	9,500	0.27	0.15	0.17	0.26	0.36
CAPINT	9,500	0.31	0.21	0.15	0.25	0.42
INTAN	9,500	0.04	0.05	0.00	0.02	0.06

Table 2, Panel A presents year distribution and descriptive statistics for main variables we use in the regressions.

		(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)
Panel B: Correlation Matrix
(1)	RATE
(2)	ITP	0.012
(3)	TRANSP	0.156***	-0.054***
(4)	ACCRUAL	0.011	0.033***	0.008
(5)	SIZE	0.529***	0.080***	0.105***	0.060***
(6)	LOSS	-0.123***	-0.046***	-0.040***	-0.009	-0.016
(7)	COVER	0.202***	0.218***	-0.035***	-0.006	0.138***	-0.050***
(8)	ROA	0.363***	0.349***	0.010	0.029***	0.071***	-0.220***	0.352***
(9)	BMRATIO	-0.245***	-0.124***	0.041***	-0.027***	-0.089***	0.003	-0.114***	-0.421***
(10)	ROASTD3	-0.264***	0.141***	-0.174***	-0.009	-0.157***	0.069***	0.021**	0.055***	-0.024**
(11)	STDRET	-0.455***	0.068***	-0.323***	-0.005	-0.324***	0.100***	-0.038***	-0.140***	0.215***	0.309***
(12)	LEV	-0.367***	-0.113***	-0.021**	0.020**	-0.181***	0.096***	-0.431***	-0.267***	-0.097***	0.038***	0.114***
(13)	CAPINT	0.071***	-0.077***	0.046***	-0.037***	0.045***	-0.006	-0.104***	-0.066***	0.133***	0.013	-0.010	0.116***
(14)	INTAN	0.238***	0.207***	-0.065***	0.024**	0.067***	-0.014	0.215***	0.322***	-0.256***	0.066***	-0.029***	-0.164***	-0.265***

Table 2, Panel B reports Pearson correlations for main variables in the regressions. *** and ** denote significance at the 1 percent and 5 percent levels, respectively.

Table 3. The Effect of International Tax Planning on Credit Ratings

Dependent Variable =	Full Sample	ITP > 0 Sample	Full Sample	ITP > 0 Sample
	Rate	Rate	Rate	Rate
ITP	-0.22*** (-4.01)	-0.22*** (-4.00)	-0.21*** (-3.83)	-0.24*** (-4.03)
TRANSP	0.45*** (4.31)	0.38*** (2.92)	0.58*** (4.81)	0.58*** (3.77)
ACCRUAL	-0.04* (-1.85)	-0.05* (-1.87)	-0.05* (-1.95)	-0.05* (-1.68)
SIZE	1.16*** (24.87)	1.20*** (23.82)	1.22*** (21.94)	1.32*** (20.99)
LOSS	-0.23 (-1.30)	-0.14 (-0.55)	-0.36* (-1.86)	-0.46 (-1.49)
COVER	0.00 (0.58)	-0.00 (-0.04)	0.00 (0.04)	-0.00 (-0.58)
ROA	15.90*** (14.66)	15.34*** (12.32)	17.16*** (14.39)	17.34*** (12.48)
BMRATIO	-1.44*** (-9.22)	-1.42*** (-7.43)	-1.58*** (-9.37)	-1.64*** (-7.91)
ROASTD3	-9.42*** (-10.58)	-9.50*** (-9.93)	-10.72*** (-11.49)	-11.05*** (-10.35)
STDRET	-96.44*** (-18.05)	-96.42*** (-16.42)	-109.34*** (-18.63)	-112.00*** (-17.28)
LEV	-4.87*** (-14.41)	-4.17*** (-11.37)	-5.53*** (-14.20)	-4.89*** (-11.35)
CAPINT	0.23 (0.63)	0.53 (1.45)	0.20 (0.50)	0.50 (1.19)
INTAN	1.52 (1.28)	2.58** (1.99)	1.70 (1.34)	3.28** (2.26)
Year FE	Yes	Yes	Yes	Yes
Ind FE	Yes	Yes	Yes	Yes
Constant	8.83*** (11.01)	7.61*** (9.58)
Observations	9,500	6,003	9,500	6,003
R-squared	0.716	0.724	0.255	0.264

Table 3 presents the results from the estimation of the following model:

Credit Rating_i,t = β₀ + β₁ * ITP_i,t-1 + ∑ β_i * Controls_i,t-1 + ε

All continuous variables are winsorized at the 1 percent and 99 percent levels. Regressions include year and SIC two-digit industry fixed effects, and standard errors are heteroscedasticity robust and clustered at company level. T- statistics are reported in parentheses. ***, **, and * denote significance at the 1 percent, 5 percent, and 10 percent levels, respectively.

Table 4. Conflict of Interest for Bond Rating Agency

Dependent Variable =	Full Sample	ITP > 0 Sample
	Rate	Rate
Panel A: Effect of S&P Relationship History
ITP	-0.25*** (-4.17)	-0.28*** (-4.21)
HighRelation	0.14 (1.22)	0.01 (0.08)
HighRelation * ITP	0.18* (1.73)	0.23** (2.28)
TRANSP	0.45*** (4.34)	0.38*** (2.96)
ACCRUAL	-0.05** (-2.02)	-0.05** (-2.05)
SIZE	1.14*** (23.18)	1.18*** (21.94)
LOSS	-0.24 (-1.40)	-0.16 (-0.64)
COVER	0.00 (0.70)	0.00 (0.05)
ROA	15.82*** (14.68)	15.21*** (12.25)
BMRATIO	-1.43*** (-9.12)	-1.42*** (-7.37)
ROASTD3	-9.46*** (-10.60)	-9.50*** (-9.89)
STDRET	-94.92*** (-17.67)	-95.24*** (-16.20)
LEV	-4.84*** (-14.24)	-4.15*** (-11.24)
CAPINT	0.23 (0.64)	0.52 (1.41)
INTAN	1.54 (1.29)	2.66** (2.04)
Year FE	Yes	Yes
Ind FE	Yes	Yes
Constant	8.97*** (10.64)	7.75*** (9.35)
Observations	9,500	6,003
R-squared	0.717	0.725
Panel B: Effect of the Proportion of Short-Term Debt
ITP	-0.33*** (-5.11)	-0.33*** (-5.02)
HighSDebt	0.31*** (3.39)	0.20** (1.99)
HighSDebt * TP	0.27*** (2.98)	0.26*** (2.87)
TRANSP	0.44*** (4.31)	0.36*** (2.85)
ACCRUAL	-0.05** (-2.06)	-0.05** (-2.09)
SIZE	1.07*** (20.95)	1.11*** (20.21)
LOSS	-0.22 (-1.27)	-0.14 (-0.57)
COVER	0.00 (0.80)	0.00 (0.15)
ROA	15.75*** (14.68)	15.20*** (12.26)
BMRATIO	-1.45*** (-9.32)	-1.43*** (-7.53)
ROASTD3	-9.11*** (-10.36)	-9.15*** (-9.66)
STDRET	-95.18*** (-18.01)	-95.40*** (-16.40)
LEV	-5.08*** (-15.18)	-4.38*** (-15.18)
CAPINT	0.21 (0.60)	0.52 (1.44)
INTAN	1.16 (0.97)	2.26* (1.72)
Year FE	Yes	Yes
Ind FE	Yes	Yes
Constant	9.47*** (11.51)	8.26*** (9.96)
Observations	9,500	6,003
R-squared	0.720	0.727
Panel C: The Effect of the Presence of a Fitch Rating
ITP	-0.29*** (-4.79)	-0.27*** (-4.47)
Fitch	-0.43*** (-4.71)	-0.33*** (-3.09)
Fitch * ITP	0.26*** (3.26)	0.19** (2.11)
TRANSP	0.45*** (4.36)	0.37*** (2.91)
ACCRUAL	-0.04* (-1.84)	-0.05* (-1.86)
SIZE	1.18*** (25.43)	1.21*** (25.43)
LOSS	-0.22 (-1.26)	-0.13 (-0.49)
COVER	0.00 (0.44)	-0.00 (-0.15)
ROA	15.88*** (14.77)	15.29*** (12.35)
BMRATIO	-1.47*** (-9.45)	-1.44*** (-7.55)
ROASTD3	-9.53*** (-10.67)	-9.58*** (-9.98)
STDRET	-95.82*** (-17.96)	-96.24*** (-16.35)
LEV	-4.87*** (-14.50)	-4.17*** (-11.40)
CAPINT	0.23 (0.65)	0.52 (1.44)
INTAN	1.47 (1.23)	2.52* (1.95)
Year FE	Yes	Yes
Ind FE	Yes	Yes
Constant	8.77*** (10.77)	7.60*** (9.42)
Observations	9,500	6,003
R-squared	0.718	0.725

Table 4 presents the effect of international tax planning under the conflict of interest of bond rating agencies.

Panel A presents the results from the estimation of the following model:

Credit Rating_i,t = β₀ + β₁ * ITP_i,t-1 + β₂ * HighRelation_i,t-1 + β₃ * HighRelation_i,t-1 * ITP_i,t-1 + ∑ β_i * Controls_i,t-1 + ε

Panel B presents the results from the estimation of the following model:

Credit Rating_i,t = β₀ + β₁ * ITP_i,t-1 + β₂ * HighSDebt_i,t-1 + β₃ * HighSDebt_i,t-1 * ITP_i,t-1 + ∑ β_i * Controls_i,t-1 + ε

Panel C presents the results from the estimation of the following model:

Credit Rating_i,t = β₀ + β₁ * ITP_i,t-1 + β₂ * Fitch_i,t-1 + β₃ * Fitch_i,t-1 * ITP_i,t-1 + ∑ β_i * Controls_i,t-1 + ε

Where HighRelation_i,t-1 is an indicator variable equal to 1 if company i in year t has a relationship with S&P longer than the sample year median, and zero otherwise. HighSDebt_i,t-1 is an indicator variable equal to 1 if short-term debt for company i in year t-1 is above the sample year median, and zero otherwise. Fitch_i,t-1 is an indicator variable equal to 1 if there is a Fitch rating for company i in year t-1, and zero otherwise. All continuous variables are winsorized at 1 percent and 99 percent levels. Regressions include year and SIC two-digit industry fixed effects, and standard errors are heteroscedasticity robust and clustered at company level. t-statistics are reported in parentheses. ***, **, and * denote significance at the 1 percent, 5 percent, and 10 percent levels, respectively.

Table 5. Future Cash Flow Effects of International
Tax Planning

Dependent Variable =	Full Sample	ITP > 0 Sample
	Rate	Rate
ITP	0.02 (0.18)	0.03 (0.40)
HighHP	-0.12 (-0.94)	-0.03 (-0.26)
HighHP * ITP	-0.35***	-0.39***
	(-3.43)	(-3.67)
TRANSP	0.45*** (4.35)	0.38*** (2.97)
ACCRUAL	-0.05* (-1.91)	-0.05* (-1.89)
SIZE	1.09*** (20.00)	1.12*** (18.83)
LOSS	-0.24 (-1.36)	-0.18 (-0.71)
COVER	0.00 (0.57)	-0.00 (-0.07)
ROA	15.75*** (14.55)	15.09*** (12.15)
BMRATIO	-1.43*** (-9.05)	-1.41*** (-7.30)
ROASTD3	-9.25*** (-10.50)	-9.24*** (-9.72)
STDRET	-94.83*** (-17.73)	-94.67*** (-16.08)
LEV	-4.88*** (-14.20)	-4.19*** (-11.14)
CAPINT	0.30 (0.83)	0.62* (1.71)
INTAN	1.62 (1.35)	2.74** (2.09)
Year FE	Yes	Yes
Ind FE	Yes	Yes
Constant	9.30*** (10.40)	8.10*** (9.13)
Observations	9,500	6,003
R-squared	0.718	0.727
Panel B: Size Interaction
ITP	-0.29*** (-4.28)	-0.32*** (-4.54)
HighSIZE	1.89*** (14.63)	1.80*** (12.65)
HighSIZE * ITP	0.33*** (2.90)	0.38*** (3.24)
TRANSP	0.42*** (3.70)	0.33** (2.39)
ACCRUAL	-0.03 (-0.98)	-0.04 (-1.18)
LOSS	-0.12 (-0.65)	-0.11 (-0.45)
COVER	0.00 (0.67)	0.00 (0.17)
ROA	14.02*** (12.20)	13.67*** (10.76)
BMRATIO	-1.48*** (-8.76)	-1.51*** (-7.23)
ROASTD3	-11.41*** (-11.94)	-11.73*** (-11.45)
STDRET	-127.55*** (-22.76)	-127.18*** (-21.09)
LEV	-5.62*** (-14.15)	-4.88*** (-10.67)
CAPINT	0.42 (1.13)	0.55 (1.44)
INTAN	3.41*** (1.44)	4.26*** (2.92)
Year FE	Yes	Yes
Ind FE	Yes	Yes
Constant	17.60*** (24.30)	16.92*** (26.25)
Observations	9,500	6,003
R-squared	0.652	0.657
Panel C: Foreign Bank Interaction
ITP	-0.46*** (-5.24)	-0.38*** (-4.16)
FOREIGN	-0.55*** (-6.10)	-0.39*** (-3.51)
FOREIGN * ITP	0.41*** (4.32)	0.26*** (2.60)
TRANSP	0.45*** (4.26)	0.36*** (2.78)
ACCRUAL	-0.05* (-1.90)	-0.05* (-1.91)
SIZE	1.17*** (25.62)	1.20*** (24.03)
LOSS	-0.24 (-1.40)	-0.16 (-0.62)
COVER	0.00 (0.43)	-0.00 (-0.11)
ROA	15.65*** (14.46)	15.24*** (12.29)
BMRATIO	-1.47*** (-9.48)	-1.44*** (-7.52)
ROASTD3	-9.45*** (-10.58)	-9.51*** (-9.89)
STDRET	-95.55*** (-17.95)	-95.93*** (-16.44)
LEV	-4.81*** (-14.47)	-4.14*** (-11.34)
CAPINT	0.16 (0.47)	0.47 (1.29)
INTAN	1.23 (1.04)	2.37* (1.82)
Year FE	Yes	Yes
Ind FE	Yes	Yes
Constant	8.80*** (10.98)	7.66*** (9.61)
Observations	9,500	6,003
R-squared	0.720	0.726

Table 5 presents the results related to the channel of future cash flow.

Panel A presents the results from the estimation of the following model:

Credit Rating_i,t = β₀ + β₁ * ITP_i,t-1 + β₂ * HighHP_i,t-1+ β₃ * HighHP_i,t-1 * ITP_i,t-1 + ∑ β_i * Controls_i,t-1 + ε

Panel B presents the results from the estimation of the following model:

Credit Rating_i,t = β₀ + β₁ * ITP_i,t-1+β₂ * HighSIZE_i,t-1+ β₃ * HighSIZE_i,t-1* ITP_i,t-1+ ∑ β_i * Controls_i,t-1 + ε

Panel C presents the results from the estimation of the following model:

Credit Rating_i,t = β₀ + β₁* ITP_i,t-1+β₂ * FOREIGN_i,t-1+ β₃ * FOREIGN_i,t-1* ITP_i,t-1+ ∑ β_i * Controls_i,t-1 + ε

Where HighHP_i,t-1 is an indicator variable equal to 1 if HP Index for company i in year t-1 is above the sample year median, and zero otherwise. HighSIZE_i,t-1 is an indicator variable equal to 1 if SIZE for company i in year t-1 is above the sample year median, and zero otherwise. FOREIGN_i,t-1 is an indicator variable equal to 1 if there is foreign bank lending for company i in year t-1, and zero otherwise. All continuous variables are winsorized at 1 percent and 99 percent levels. Regressions include year and SIC two-digit industry fixed effects, and standard errors are heteroscedasticity robust and clustered at company level. t-statistics are reported in parentheses. ***, **, and * denote significance at the 1 percent, 5 percent, and 10 percent levels, respectively.

Table 6. Agency Risks and International Tax Planning

Dependent Variable =	Full Sample	ITP > 0 Sample
	Rate	Rate
Panel A: Analyst Coverage Interaction
ITP	-0.32*** (-4.89)	-0.34*** (-4.89)
HighANA	0.08 (0.75)	0.01 (0.04)
HighANA * ITP	0.17* (1.84)	0.20** (2.18)
TRANSP	0.45*** (4.29)	0.38*** (2.92)
ACCRUAL	-0.04* (-1.83)	-0.05* (-1.86)
SIZE	1.13*** (21.56)	1.17*** (20.82)
LOSS	-0.23 (-1.35)	-0.14 (-0.58)
COVER	0.00 (0.49)	-0.00 (-0.14)
ROA	15.71*** (14.51)	15.17*** (12.18)
BMRATIO	-1.42*** (-9.01)	-1.41*** (-7.24)
ROASTD3	-9.27*** (-10.36)	-9.32*** (-9.68)
STDRET	-96.34*** (-17.96)	-96.06*** (-16.35)
LEV	-4.85*** (-14.47)	-4.16*** (-11.41)
CAPINT	0.21 (0.58)	0.51 (1.41)
INTAN	1.29 (1.09)	2.32* (1.80)
Year FE	Yes	Yes
Ind FE	Yes	Yes
Constant	9.11*** (11.35)	7.94*** (9.82)
Observations	9,500	6,003
R-squared	0.717	0.725
Panel B: Number of Independent Directors Interaction
ITP	-0.46*** (-4.48)	-0.57*** (-5.45)
HighINDE	-0.22 (-1.64)	-0.32* (-1.88)
HighINDE * ITP	0.55*** (3.44)	0.58*** (3.47)
TRANSP	0.27 (1.62)	0.23 (1.08)
ACCRUAL	-0.06 (-0.78)	-0.02 (-0.21)
SIZE	1.06*** (13.75)	1.06*** (12.17)
LOSS	-0.16 (-0.45)	-0.08 (-0.18)
COVER	-0.00 (-1.54)	-0.00* (-1.90)
ROA	17.16*** (9.36)	17.57*** (8.86)
BMRATIO	-1.36*** (-4.82)	-1.50*** (-4.14)
ROASTD3	-10.71*** (-6.09)	-12.90*** (-6.39)
STDRET	-104.90*** (-10.53)	-98.18*** (-8.79)
LEV	-4.98*** (-8.22)	-3.81*** (-5.96)
CAPINT	-0.44 (-0.84)	-0.26 (-0.46)
INTAN	0.96 (0.55)	2.53 (1.38)
Year FE	Yes	Yes
Ind FE	Yes	Yes
Constant	9.31*** (10.12)	8.41*** (5.33)
Observations	2,619	1,653
R-squared	0.688	0.713

Table 6 presents the results on the channel of agency risk.

Panel A presents the results from the estimation of the following model:

Credit Rating_i,t = β₀ + β₁ * ITP_i,t-1+ β₂ * HighANA_i,t-1+ β₃ * HighANA_i,t-1* ITP_i,t-1+ ∑ β_i * Controls_i,t-1 + ε

Panel B presents the results from the estimation of the following model:

Credit Rating_i,t = β₀ + β₁ * ITP_i,t-1+ β₂ * HighINDE_i,t-1+ β₃ * HighINDE_i,t-1*ITP_i,t-1+ ∑ β_i * Controls_i,t-1 + ε

Where HighANA_i,t-1 is an indicator variable equal to 1 if analyst coverage for company i in year t-1 is above the sample year median, and zero otherwise. HigINDE_i,t-1 is an indicator variable equal to 1 if the number of independent directors for company i in year t-1 is above the sample year median, and zero otherwise. All continuous variables are winsorized at 1 percent and 99 percent levels. Regressions include year and SIC two-digit industry fixed effects, and standard errors are heteroscedasticity robust and clustered at company level. t-statistics are reported in parentheses. ***, **, and * denote significance at the 1 percent, 5 percent, and 10 percent levels, respectively.

Table 7. Information Risk and International Tax Planning

Dependent Variable =	Full Sample	ITP > 0 Sample
	Rate	Rate
Panel A: Financial Transparency
ITP	-0.27*** (-4.68)	-0.28*** (-4.59)
HighTRANSP	0.07 (1.44)	0.08 (1.29)
HighTRANSP * ITP	0.12** (2.19)	0.11* (1.82)
ACCRUAL	-0.05* (-1.94)	-0.05* (-1.96)
SIZE	1.16*** (24.81)	1.19*** (23.82)
LOSS	-0.24 (-1.38)	-0.17 (-0.65)
COVER	0.00 (0.55)	-0.00 (-0.07)
ROA	15.86*** (14.60)	15.27*** (12.25)
BMRATIO	-1.42*** (-9.14)	-1.41*** (-7.42)
ROASTD3	-9.56*** (-10.76)	-9.54*** (-10.03)
STDRET	-99.17*** (-18.56)	-98.32*** (-16.91)
LEV	-4.86*** (-14.38)	-4.16*** (-11.38)
CAPINT	0.24 (0.65)	0.55 (1.49)
INTAN	1.53 (1.29)	2.62** (2.03)
Year FE	Yes	Yes
Ind FE	Yes	Yes
Constant	8.83*** (10.93)	7.57*** (9.51)
Observations	9,500	6,003
R-squared	0.716	0.724
Panel B: Management Forecasts
ITP	-0.32*** (-3.60)	-0.38*** (-3.88)
MF	-0.14 (-1.18)	-0.25* (-1.69)
MF * ITP	0.22** (2.23)	0.24** (2.29)
TRANSP	0.43*** (4.14)	0.35*** (2.73)
ACCRUAL	-0.06** (-2.52)	-0.06** (-2.36)
SIZE	1.18*** (25.11)	1.20*** (23.38)
LOSS	-0.25 (-1.39)	-0.20 (-0.78)
COVER	0.00 (1.07)	0.00 (0.45)
ROA	14.79*** (12.89)	14.78*** (12.26)
BMRATIO	-1.37*** (-8.42)	-1.39*** (-7.34)
ROASTD3	-8.99*** (-9.81)	-9.52*** (-9.54)
STDRET	-88.43*** (-16.08)	-91.18*** (-15.43)
LEV	-4.04*** (-11.36)	-3.64*** (-9.54)
CAPINT	0.24 (0.69)	0.42 (1.18)
INTAN	1.22 (0.98)	2.15 (1.60)
Year FE	Yes	Yes
Ind FE	Yes	Yes
Constant	6.86*** (9.54)	6.29*** (8.18)
Observations	7,331	5,022
R-squared	0.719	0.728

Table 7 presents the results on the channel of information risk.

Panel A presents the results from the estimation of the following model:

Credit Rating_i,t = β₀ + β₁ * ITP_i,t-1+ β₂ * HighTRANSP_i,t-1+ β₃ * HighTRANSP_i,t-1* ITP_i,t-1+ ∑ β_i * Controls_i,t-1 + ε

Panel B presents the results from the estimation of the following model:

Credit Rating_i,t = β₀ + β₁ * ITP_i,t-1+ β₂ * MF_i,t-1+ β₃ * MF_i,t-1* ITP_i,t-1+ ∑ β_i * Controls_i,t-1 + ε

Where HighTRANSP_i,t-1 is an indicator variable equal to 1 if TRANSP_i,t-1 for company i in year t-1 is above the sample year median, and zero otherwise. MF_i,t-1 is an indicator variable equal to 1 if there is at least one management forecast for company i in year t-1, and zero otherwise. (The sample began after 1995.) All continuous variables are winsorized at 1 percent and 99 percent levels. Regressions include year and SIC two-digit industry fixed effects, and standard errors are heteroscedasticity robust and clustered at company level. t-statistics are reported in parentheses. ***, **, and * denote significance at the 1 percent, 5 percent, and 10 percent levels, respectively.

Table 8. Difference-in-Differences Design: Using the AJCA as a Shock

Dependent Variable =	Full Sample	ITP > 0 Sample
	Rate	Rate
Panel A: DID Unbalanced Sample
ITP	-0.33*** (-3.32)	-0.41*** (-3.93)
AJCA	-0.67*** (-8.45)	-0.63*** (-6.21)
AJCA * ITP	0.17** (2.22)	0.16* (1.85)
TRANSP	-0.09 (-0.54)	-0.23 (-1.13)
ACCRUAL	-0.08 (-1.23)	-0.08 (-0.95)
SIZE	1.23*** (17.93)	1.26*** (16.74)
LOSS	-0.79*** (-2.84)	-0.37 (-1.09)
COVER	0.00 (0.27)	-0.00 (-0.07)
ROA	13.06*** (6.64)	14.86*** (7.02)
BMRATIO	-1.47*** (-5.43)	-1.19*** (-4.12)
ROASTD3	-14.33*** (-8.31)	-14.21*** (-7.32)
STDRET	-49.52*** (-6.01)	-54.27*** (-5.87)
LEV	-4.06*** (-6.76)	-3.25*** (-5.30)
CAPINT	1.00* (1.95)	1.11** (2.17)
INTAN	2.37 (1.23)	3.58* (1.73)
Ind FE	Yes	Yes
Constant	4.36*** (4.70)	3.66*** (3.79)
Observations	2,624	1,842
R-squared	0.660	0.683
Panel B: DID Balanced Sample
ITP	-0.40*** (-2.84)	-0.48*** (-3.29)
AJCA	-0.58*** (-6.69)	-0.59*** (-5.69)
AJCA * ITP	0.15* (1.72)	0.20** (2.24)
TRANSP	-0.10 (-0.43)	-0.24 (-0.83)
ACCRUAL	0.05 (0.68)	0.05 (0.63)
SIZE	1.20*** (13.26)	1.21*** (12.42)
LOSS	-1.15*** (-3.23)	-1.07* (-1.83)
COVER	0.00 (0.12)	-0.00 (-0.17)
ROA	12.68*** (4.85)	15.44*** (5.84)
BMRATIO	-2.01*** (-4.60)	-1.77*** (-3.97)
ROASTD3	-12.98*** (-4.32)	-14.93*** (-4.02)
STDRET	-49.47*** (-4.36)	-49.87*** (-3.87)
LEV	-3.51*** (-4.19)	-2.61*** (-3.20)
CAPINT	1.15 (1.64)	1.46** (2.08)
INTAN	2.35 (0.96)	3.79 (1.52)
Ind FE	Yes	Yes
Constant	4.26*** (3.91)	4.00*** (3.40)
Observations	1,808	1,281
R-squared	0.629	0.661

Table 8 presents the results from the estimation of the following model around the AJCA:

Credit Rating_i,t = β₀ + β₁ * ITP_i,t-1+ β₂ * AJCA + β₂ * AJCA * ITP_i,t-1+ ∑ β_i * Controls_i,t-1 + ε

All continuous variables are winsorized at 1 percent and 99 percent levels. Regressions include year and SIC two-digit industry fixed effects, and standard errors are heteroscedasticity robust and clustered at company level. t-statistics are reported in parentheses. ***, **, and * denote significance at the 1 percent, 5 percent, and 10 percent levels, respectively.

Table 9. The Incremental Effect of International Tax Planning Above General Tax Avoidance

Dependent Variable =	Full Sample			ITP > 0 Sample
	Rate	Rate	Rate	Rate	Rate	Rate
Panel A: Controlling for General Tax Avoidance Measures
ITP	-0.29*** (-5.16)	-0.19*** (-3.42)	-0.23*** (-3.51)	-0.30*** (-5.27)	-0.19*** (-3.31)	-0.23*** (-3.32)
BT	-3.71*** (-4.61)			-3.85*** (-4.19)
TA_CETR		-0.71*** (-3.85)			-1.04*** (-4.40)
DTAX			-0.08 (-0.40)			-0.02 (-0.09)
TRANSP	0.45*** (4.26)	0.42*** (4.01)	0.63*** (4.47)	0.38*** (2.83)	0.37*** (2.91)	0.60*** (3.64)
ACCRUAL	-0.04* (-1.71)	-0.05** (-1.96)	-0.01 (-0.37)	-0.04* (-1.71)	-0.05* (-1.92)	-0.02 (-0.58)
SIZE	1.16*** (24.81)	1.16*** (24.26)	1.20*** (22.57)	1.19*** (23.86)	1.20*** (23.57)	1.22*** (21.76)
LOSS	-0.30* (-1.76)	-0.34* (-1.87)	-0.09 (-0.37)	-0.20 (-0.80)	-0.19 (-0.76)	-0.12 (-0.32)
COVER	0.00 (0.57)	0.00 (0.72)	-0.00 (-0.15)	0.00 (0.02)	0.00 (0.11)	-0.00 (-0.94)
ROA	17.16*** (14.65)	15.42*** (13.92)	15.51*** (12.36)	16.61*** (12.59)	14.83*** (11.83)	14.37*** (10.39)
BMRATIO	-1.39*** (-8.86)	-1.45*** (-9.02)	-1.78*** (-9.76)	-1.39*** (-7.32)	-1.43*** (-7.40)	-1.81*** (-8.08)
ROASTD3	-8.83*** (-10.15)	-8.86*** (-10.00)	-7.92*** (-7.27)	-9.03*** (-9.63)	-8.93*** (-9.46)	-8.14*** (-7.26)
STDRET	-96.24*** (-17.78)	-97.59*** (-17.84)	-93.66*** (-14.63)	-97.34*** (-16.41)	-97.37*** (-16.31)	-95.09*** (-13.77)
LEV	-4.69*** (-13.87)	-4.63*** (-13.62)	-4.99*** (-13.28)	-4.04*** (-11.11)	-3.98*** (-11.02)	-4.29*** (-10.90)
CAPINT	0.33 (0.93)	0.12 (0.32)	0.46 (1.12)	0.62* (1.71)	0.47 (1.25)	0.77* (1.87)
INTAN	1.16 (0.98)	1.12 (0.94)	-0.09 (-0.06)	2.13* (1.65)	2.12 (1.65)	1.63 (1.12)
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
Ind FE	Yes	Yes	Yes	Yes	Yes	Yes
Constant	7.83*** (9.90)	9.46*** (10.95)	7.68*** (8.95)	6.78*** (8.25)	7.68*** (8.58)	7.06*** (7.71)
Observations	9,221	8,627	6,208	5,879	5,572	3,928
R-squared	0.717	0.715	0.727	0.727	0.728	0.740
Panel B: Partition on Tax Avoidance Measures
ITP	-0.26*** (-4.20)	-0.23*** (-2.93)	-0.16* (-1.84)	-0.15** (-2.52)	-0.19** (-2.24)	-0.27*** (-3.73)
TRANSP	0.37*** (2.73)	0.55*** (3.57)	0.38** (2.38)	0.44*** (3.26)	0.46*** (2.65)	0.75*** (3.84)
ACCRUAL	-0.07** (-1.99)	-0.01 (-0.24)	-0.07* (-1.94)	-0.03 (-1.08)	-0.02 (-0.54)	0.01 (0.14)
SIZE	1.21*** (20.73)	1.10*** (21.93)	1.19*** (20.02)	1.16*** (22.79)	1.24*** (20.35)	1.16*** (20.16)
LOSS	-0.39** (-2.19)	0.12 (0.18)	-0.21 (-0.99)	-0.27 (-0.90)	-0.11 (-0.36)	-0.12 (-0.41)
COVER	0.00 (0.75)	-0.00 (-0.12)	0.00 (1.46)	-0.00 (-1.49)	-0.00 (-0.06)	-0.00 (-0.38)
ROA	15.59*** (10.37)	17.95*** (13.48)	15.51*** (10.28)	14.82*** (11.08)	14.58*** (8.47)	16.48*** (11.48)
BMRATIO	-1.18*** (-6.64)	-1.65*** (-8.26)	-1.42*** (-7.32)	-1.46*** (-7.62)	-1.89*** (-8.71)	-1.75*** (-8.36)
ROASTD3	-5.98*** (-5.26)	-12.41*** (-10.10)	-8.79*** (-6.85)	-8.44*** (-8.24)	-6.67*** (-4.27)	-8.60*** (-6.63)
STDRET	-98.59*** (-15.27)	-91.13*** (-13.07)	-100.45*** (-14.48)	-93.10*** (-14.37)	-93.98*** (-11.38)	-92.96*** (-12.66)
LEV	-4.62*** (-11.59)	-4.85*** (-11.65)	-4.91*** (-12.20)	-4.29*** (-10.18)	-5.00*** (-11.10)	-5.10*** (-12.33)
CAPINT	0.44 (0.98)	0.05 (0.14)	0.15 (0.30)	-0.08 (-0.19)	0.31 (0.64)	0.68 (1.51)
INTAN	2.82** (2.14)	-0.57 (-0.39)	1.57 (1.10)	1.27 (0.89)	-0.16 (-0.10)	-0.28 (-0.18)
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
Ind FE	Yes	Yes	Yes	Yes	Yes	Yes
Constant	6.66*** (8.57)	9.46*** (13.52)	9.94*** (10.47)	9.07*** (8.79)	7.50*** (7.55)	7.72*** (8.59)
Observations	4,621	4,600	4,319	4,308	3,111	3,097
R-squared	0.725	0.720	0.722	0.716	0.710	0.751

Table 9 presents the results differing the effect of international tax planning from the effect of overall tax avoidance.

Panel A reports results with overall tax avoidance measures in the regressions. Regressions are partitioned on sample year median of overall tax avoidance measures in Panel B.

Both panels presents the results from the estimation of the following model:

Credit Rating_i,t = β₀ + β₁ * ITP_i,t-1 + ∑ β_i * Controls_i,t-1 + ε

BT is book-tax difference for a company in a specific year (Manzon and Plesko, supra note 78). TA_CETR is the permanent book-tax difference for a company in a specific year (Frank, Lynch, and Rego, supra note 78). DTAX is minus 1 times the cash effective tax rate of a company in a specific year. All continuous variables are winsorized at 1 percent and 99 percent levels. Regressions include year and SIC two-digit industry fixed effects, and standard errors are heteroscedasticity robust and clustered at company level. t-statistics are reported in parentheses. ***, **, and * denote significance at the 1 percent, 5 percent, and 10 percent levels, respectively.