Investment Opportunities, Free-Cash-Flow, and the Market Values of Foreign and Domestic Cash Holdings Richard T. Thakor∗ First Draft: April 2008 This Draft: September 2013 Abstract It has been suggested that firms with foreign operations stockpile large amounts of cash, primarily in their foreign subsidiaries, because bringing the cash home involves paying a repatriation tax on foreign income. This implies that the stock market should value foreign-held cash less than domestically-held cash. But this effect may be moderated by the impact of investment opportunities abroad that would provide an outlet for the foreign cash and affect how the market values it. This paper empirically examines the difference between the market values of on-balance-sheet cash held domestically and that held abroad by US firms, and the impact of the interaction of the repatriation tax and investment opportunities on this difference. The results show that shareholders assign a higher value to cash held abroad than to cash held domestically, and that the marginal value of foreign-held cash is substantially higher than that of domestic cash for US firms with better foreign investment opportunities. This suggests that the effect of the differential investment opportunities for foreign and domestic cash swamps the repatriation-tax disadvantage of foreign cash. As further evidence, this paper also examines the effect of the exogenous shock provided by the tax repatriation holiday in 2004, and finds that firms with better investment opportunities abroad experienced lower abnormal returns following the passage of that legislation.

JEL Classification Numbers: G31, G32, H26 Keywords: Cash holdings, Value of cash, Foreign cash, Repatriation tax, Investment opportunities, Free-cash-flow ∗

MIT Sloan School of Management. E-mail: [email protected]. I would like to thank Bruce Petersen, Mike Faulkender, Sudipto Bhattacharya, Antoinette Schoar, Stew Myers, Becky Lester, and seminar participants at MIT for all of their helpful comments. I would also like to thank Rong Wang and John Graham for their help in providing data. I am grateful to the London School of Economics and Creditflux for financial support from the Creditflux-Cairn Capital Prize, which was awarded to a previous version of this paper. I alone am responsible for all errors.

1

Introduction

Firms in the US hold large amounts of cash on their balance sheets and these amounts have been steadily increasing in recent years. For example, Bates, Kahle, and Stulz (2007) document that the average cash-to-assets ratio for corporations has more than doubled from 10.5% in 1980 to 24% in 2004. And the average firm included in COMPUSTAT held nearly $285 million in cash on its balance sheet in 2005, increasing to more than $1 billion by the end of 2009.1 For many firms, a substantial portfion of the cash is held in foreign subsidiaries (e.g. Drucker (2011)). Why firms hold so much cash and how the market values this cash are interesting questions both theoretically and empirically. In particular, cross-sectional differences in the motivations for holding cash, the locations of the cash holdings, and the market valuation of this cash can be illuminating. This paper focuses on one aspect of these differences by empirically examining how the market valuation of cash held in foreign subsidiaries relative to that held domestically varies in the cross-section of firms and what might account for this variation. In Modigliani and Miller’s (1958) perfect (and frictionless) capital market with no taxes, firms are indifferent between holding a dollar of cash on the balance sheet and paying it out, since the value implications of both choices are identical. However, the introduction of frictions of various sorts, including taxes, can open up a wedge between the net benefit of keeping cash on the balance sheet and the net benefit of distributing it to shareholders. Various papers have focused on these frictions to both explain why firms sometimes hold large amounts of cash on their balance sheets and the market’s valuation of on-balance sheet cash, with the valuation being related to the reason for holding the cash. The justifications offered for holding cash fall into two groups: (i) self-serving behavior by managers that leads to shareholder value destruction (inefficient cash hoarding), and (ii) shareholder-value maximizing behavior that optimally trades off the costs and benefits of holding cash (efficient internal capital markets). 1

Defined in COMPUSTAT as cash and short-term investments.

1

In the first group is Jensen’s (1986) free-cash-flow hypothesis that hoarding cash simply provides managers discretionary resources that can be inefficiently deployed for private benefits. Clearly, these problems will be less severe when corporate governance is better, so the market will value on-balance-sheet cash higher when governance is better, which in turns creates a link between corporate governance and cash holdings. Consistent with this, Dittmar, Mahrt-Smith, and Servaes (2003) and Harford, Mansi, and Maxwell (2008) find that firms with weaker corporate governance have smaller cash reserves. In the second group are papers that have focused on precautionary motives for holding cash (e.g. McLean (2011), Miller and Orr (1966), and Mulligan (1997)) that can be theoretically justified on the basis of the benefits of building internal capital markets (e.g. Stein (1997), and Thakor (1990)) to exploit future real options to invest, in the face of adverse selection (Myers and Majluf (1984)) and other frictions. These benefits are likely to be greater for firms that have greater growth opportunities and riskier cash flows. Evidence consistent with this appears in Opler, Pinkowitz, Stulz, and Williamson (1999).2 Given these two types of justifications for holding cash, the market value of $1 of cash on the balance sheet can be more or less than $1, depending on the types of firms in the sample. With inefficient cash hoarding, we would expect that the market value of $1 of onbalance-sheet cash is less than $1. With efficient capital markets, if investments are infinitely divisible, we would expect the market value of $1 of on-balance-sheet cash to be driven to $1. But if the firm has real options on lumpy investments, we would expect the market value of on-balance-sheet cash to exceed $1. Empirical estimates of the market value of on-balance-sheet cash suggest that both forces may be at work. Pinkowitz and Williamson (2007) estimate that a marginal dollar of cash for the average firm is valued at about $1.20, whereas Faulkender and Wang (2006) use a different methodology and find that the marginal 2 Along these lines, Acharya, Almeida and Campello (2007) develop a theoretical model in which cash is not merely “negative debt”. They show that while cash allows financially-constrained firms to hedge future investment against negative shocks to income, reducing current debt is more effective in increasing income in future high-cash-flow states. Thus, financially-constrained firms prefer higher cash to lower debt if their hedging needs are high, but lower debt to higher cash if their hedging needs are low.

2

value of cash for the average firm is $0.94; the value is less than $1 as a result of taxes (such as dividend taxes) and agency costs. Dittmar and Mahrt-Smith (2007) extend the methodology of Faulkender and Wang (2006) to look at the effects of corporate governance on the value of cash holdings. They find that poorly-governed firms have a lower marginal value of cash: $1.00 of cash in a poorly-governed firm is worth between $0.42 and $0.88, whereas it is twice that value in a well-governed firm. And Liu and Mauer (2011) show that an increase in CEO risk-taking reduces the value of cash to shareholders. While most of this literature has focused on the role of agency problems and informational frictions in explaining cash holdings, Foley, Hartzell, Titman, and Twite (2007) provide an explanation based on taxes. They argue that partly as a result of the tax penalty for repatriating earnings back to the U.S., there is a tax-based incentive for companies to hold cash abroad since doing so reduces/delays the payment of the repatriation tax on foreign income.3 Therefore, the repatriation tax argument implies that a dollar of foreign-held cash should be worth less than a dollar of domestically-held cash, since the foreign-held cash is essentially sitting inefficiently “trapped” abroad in firms that want to avoid the repatriation tax. Other more recent paper provide additional evidence consistent with this argument. Blouin, Krull, and Robinson (2012) find that a higher proportion of permanently reinvested earnings (PRE) is held in the form of cash in low tax jurisdictions relative to high tax jurisdictions. Graham, Hanlon, and Shevlin (2011) provide survey evidence indicating that executives often consider avoiding taxes an important determinant of whether to repatriate foreign earnings. And a working paper by Edwards, Kravet, and Wilson (2012) argues that 3

This tax is equal to the difference between the foreign taxes already paid on earnings and the taxes that would be due if the earnings were taxed at the U.S. rate. As an example, consider $100 earned by a US multinational company in a foreign country with a tax-rate of 10%. If the US tax-rate is 35%, then repatriating the earnings leaves the company with $65 to use, whereas keeping it abroad leaves the company with $90 to use. Thus, the company may have an incentive to keep the cash abroad in order to avoid/delay the additional $25 loss associated with the repatriation tax. A recent Bloomberg article suggests that U.S. firms may have more than $1 trillion of cash sitting offshore (Drucker (2011)). There are various strategies to avoid paying the repatriation tax, such as legal loopholes and transfer pricing, although Foley et al (2007) suggest that these strategies may be cumbersome and costly for firms to use.

3

because cash is trapped abroad due to the repatriation tax, firms are more likely to use the cash to undertake value-destroying acquisitions. However, as explained above, there are other mediating variables as well that affect the value of foreign cash – free-cash-flow problems and growth opportunities. Foreign cash may be worth more than domestic cash if growth opportunities abroad are better than at home, so there is a greater need for internal capital markets in foreign operations. These higher growth opportunities abroad may also mean that free-cash-flow problems (e.g. Jensen (1986)) are less severe in foreign subsidiaries than domestically.4 For example, Doukas (1995) shows that U.S. firms that bid for foreign firms experience higher abnormal returns when they have a higher q, and that the returns are inversely related to free cash flow for low-q bidders but not for high-q bidders. And while Blouin et. al. (2012) argue that firms hold a higher proportion of PRE in cash due to tax incentives, they also recognize that a number of firms appear to hold high levels of cash in low tax jurisdictions, suggesting that growth plays a larger role than taxes for such firms. These considerations lead to the following questions that I empirically address in this paper: First, what is the value of foreign cash relative to domestic cash? Second, what is the impact of investment opportunities on this relative valuation? The methodology of Faulkender and Wang (2006) is extended in order to address the first question and estimate the value of a dollar of foreign cash and a dollar of domestic cash for firms. The results indicate that, for the average firm, $1.00 of cash held abroad is valued at $1.06, while $1.00 of cash held domestically is valued (statistically) significantly lower at $0.93. The higher value of foreign cash relative to domestic cash suggests that factors other than the repatriation tax are impacting the valuation of foreign cash. The analysis then proceeds to address the second question and examine whether growth opportunities abroad explain why foreign cash is valued higher than domestic cash. An 4

The idea is that managers will view the cost of engaging in wasteful activities as being higher when such activities use up cash that has a greater demand for financing investments and growth. Thus, they would be less likely to engage in such waste.

4

extension of the methodology of Dittmar and Mahrt-Smith (2007) is utilized for this purpose. The results indicate that, on average, among the set of firms with high levels of foreign operations, those with better investment opportunities have higher marginal values of cash: $1.42 and $1.30 respectively for $1 of foreign and domestic cash. By contrast, for firms with poor investment opportunities, the marginal values of foreign and domestic cash are $0.92 and $0.80, respectively. This suggests that the effect of foreign investment opportunities dominates that of the repatriation tax in contributing to the higher valuation of foreign cash relative to domestic cash. These findings are consistent with the notion that free-cash-flow problems are worse in domestic operations than in foreign operations of profitable investment opportunities at home. As additional evidence, I examine the effect of an exogenous shock following the enactment of the American Jobs Creation Act of 2004, which provided a one-time tax holiday for firms repatriating cash. The results indicate that firms with poor investment opportunities abroad experienced significant positive abnormal returns following the enactment of the bill, while firms with good investment opportunities abroad did not experience any significant abnormal returns. This is consistent with the hypothesis that firms with good overseas investments are holding cash primarily to fund those investments, so the repatriation tax holiday was of little consequence for them. The rest of the paper is organized as follows. Section 2 explains the baseline empirical methodology used in the paper. Section 3 describes the data set and summary statistics. Section 4 includes the main empirical results and discusses their implication. Section 5 examines additional evidence using an event study methodology to examine the effect of the American Jobs Creation Act of 2004. Section 6 concludes.

5

2

Empirical Methodology

This section describes the empirical methodology used in the analysis. The empirical methodology consists of three main steps. First, the split between foreign-held and domestically-held cash is determined for each firm in the sample. Second, the market values of foreign and domestic cash are estimated. Third, there is an empirical examination of whether the difference in the market values of foreign and domestic cash can be explained by growth opportunity differences.

2.1

Estimating the Amounts of Foreign and Domestic Cash

In order to estimate the marginal values of foreign-held and domestically-held cash, levels of both variables at the firm level must be obtained. However, while a measure of the level of overall cash for firms is available in COMPUSTAT, separate measures of the cash held abroad and cash held domestically are not readily available and therefore must be estimated.5 While one would ideally like to have access to the split between domestic and foreign cash, there is reason to believe that estimating this split is the appropriate way to proceed. This is because the question of interest is how the market values domestic versus foreign cash, and since investors themselves also lack access to the actual split of total cash between foreign and domestic cash, they need to rely on estimates as well, just as is done in this paper.6 The methodology of Foley et al. (2007) is employed in order to obtain estimates of the 5

Foley et al. (2007) obtain a direct measurement of foreign cash using data from the Bureau of Economic Analysis (BEA) for alternate specifications of their model. However, these are proprietary data which are not readily available, either to investors or to me. In addition, yearly data are needed for the following analysis, and the BEA data are only available at five-year intervals. It appears, therefore, that for the purposes of this paper, it may not be appropriate to use the BEA data. 6 As a test of robustness, results are presented subsequently in the paper using two alternate proxies of the level of cash in firms.

6

amounts of domestic and foreign cash held by firms. The following OLS specification is run:7 

Cash NA



 = β0 + β1

i,t

R&D TA



 + β2

CAP EX TA



+ β3 (Lev)i,t + β4 (CF Std Dev)i,t   BV E +β5 (Div P mt)i,t + β6 (ln (T A))i,t + β7 M V E i,t     F oreign Income Domestic Income + β9 +β8 TA TA i,t i,t i,t

i,t

+β10 (T ax Burden)i,t + εi,t .

(1)

All observations consist of firm-years: the value of the variable for firm i at year t. Several independent variables are scaled by total assets, TA, in order to normalize for firm size. Cash is defined as cash and short-term investments. NA represents net assets, and is defined as Total Assets less cash.8 CAPEX is capital expenditures. R&D represents research and development expenditures, and is set to zero if R&D expenditures is missing in COMPUSTAT. Lev is market leverage, the ratio of total debt to the sum of total debt and the market value of equity, which is defined as the firm’s year-end closing stock price multiplied by the number of shares outstanding. CF Std Dev represents cash flow standard deviation, defined as the standard deviation of the firm’s earnings before interest, taxes, and depreciation as a percentage of total assets.9 Div Pmt is a dummy variable that takes a value of 1 if the firm paid a cash dividend in the year, and a value of 0 otherwise. ln(T A) is the natural log of the firm’s total assets, in millions of dollars.

BV E MV E

is the ratio of the reported

book value of the firm’s equity divided by the market value of the firm’s equity. Domestic Income and Foreign Income represent domestic income and foreign income, respectively, and 7 Qualitatively similar results are obtained by running a lagged rather than a contemporaneous specification. 8 This follows Opler et al. (1999) and Foley et al. (2007). Using Net Assets in the denominator of the dependent variable facilitates interpretation of the results, because a change in cash would only affect the numerator. While Opler et al. (1999) and Foley et al. (2007) use the natural logarithm of cash to net assets as the dependent variable, this paper uses a winsorized ratio of cash to net assets in order to feed into the second stage regression. However, according to Foley et al. (2007), their results are robust to this specification. 9 Foley et al (2007) use operating income standard deviation instead of cash flow standard deviation because of limitations on the duration of their sample. Using cash flow standard deviation follows Opler et al (1999).

7

are both pre-tax earnings. Foreign Income is included as a proxy to measure the magnitude of the repatriation tax effect, as firms that have more foreign income tend to hold more cash in order to avoid the repatriation tax, all else being equal. The inclusion of Domestic Income controls for the possibility that a delay between when earnings are received and used will create a positive relationship between cash holdings and income. Tax Burden is a measure of the taxes a firm would face if it chose to repatriate all of its foreign income. It is defined as:

T ax Burden =

max(U ST axRate × F oreignP reT axEarnings − F oreignT axesP aid, 0) . TA (2)

USTaxRate is the marginal tax rate after interest expense, following Graham (1996).10 The overall Tax Burden variable reflects the repatriation tax faced by firms to repatriate their foreign earnings back to the US. It is equal to the difference between the foreign taxes already paid on earnings and the taxes that would be due if the foreign earnings were taxed at the US tax rate. While all the variables in equation (1) affect the general level of cash for each firm-year, Foreign Income and Tax Burden are the only two variables which specifically affect foreign cash. Since the dependent variable in the regression measures the overall level of cash for the firm, the coefficient for

F oreign Income TA

in the regression can be interpreted as the estimated

incremental increase in foreign cash associated with an increase in foreign income, all other things equal. However, the foreign income variable only accounts for the contribution to foreign cash from foreign earnings. It does not account for the foreign cash that is being held in various locations; the amount of cash held may be affected by the location because tax rates are location-specific. In order to defer paying the costs of repatriation, firms have an incentive to store cash away in low-tax jurisdictions, and the specific tax rates of these 10

The tax rate data were generously provided by John Graham. From Graham (1996), the marginal tax rate is calculated as the present value of current and expected future taxes paid on an additional dollar of income earned today.

8

jurisdictions affect the amount of cash held (since the repatriation tax is based on the amount of foreign taxes paid). Therefore, the Tax Burden variable is included to account for these issues, as an estimate of the amount of cash that has not been brought back to the US. The use of these two variables to determine foreign cash is also supported by Foley et al (2007), who run a regression explaining foreign cash levels and show that Foreign Income and Tax Burden are the two most important variables in terms of magnitude and significance. By combining these two variables from equation (1), an estimate of the amount foreign cash is obtained:11 "



Estimate (F oreign Cash)i,t = T Ai,t β9

F oreign Income TA

#



+ β10 (T ax Burden)i,t . (3) i,t

Since a firm’s total cash is made up of foreign and domestic cash, the estimate of the amount domestic cash is total cash minus the estimate of foreign cash from (3) above: " Estimate (Domestic Cash)i,t = T Ai,t

Cash TA



 − Estimate i,t

F oreign Cash TA

 # .

(4)

i,t

Negative estimates of Foreign Cash and Domestic Cash are truncated at zero.

2.2

Estimating the Market Values of Foreign and Domestic Cash

After obtaining these estimates of domestic and foreign cash, the methodology used by Faulkender and Wang (2006) is used in order to estimate the marginal values to shareholders of cash held domestically and cash held abroad. The following OLS regression is run, in which 11

In (3), Tax Burden reflects the tax burden measurement scaled by total assets, which follows from (2)

9

stock returns over the fiscal year are regressed on changes in various firm characteristics: 

     (4Cash)i,t (4Earnings)i,t (4N A)i,t (Ri,t − Rbi,t ) = λ0 + λ1 + λ2 + λ3 M V Ei,t−1 M V Ei,t−1 M V Ei,t−1       (4Int)i,t (4Div)i,t (4R&D)i,t + λ5 + λ6 +λ4 M V Ei,t−1 M V Ei,t−1 M V Ei,t−1     (Cash)i,t−1 (N F )i,t + λ8 (Lev)i,t + λ9 +λ7 M V Ei,t−1 M V Ei,t−1    (4Cash)i,t +λ10 (Lev)i,t × M V Ei,t−1     (4Cash)i,t (F oreign Cash)i,t−1 × +λ11 M V Ei,t−1 M V Ei,t−1     (Domestic Cash)i,t−1 (4Cash)i,t +λ12 × + εi,t . (5) M V Ei,t−1 M V Ei,t−1 The overall intuition of (5) is to capture the impact of a marginal dollar of cash on the unexpected stock return of the firm, holding constant all firm characteristics that may be correlated with cash. All observations are firm-years, and represent the observation of the variable at firm i in year t. 4X represents the 1-year change: Xt − Xt−1 . The dependent variable is Excess Stock Return, and is defined as the stock’s return over a given year (Ri,t ) minus the stock’s benchmark portfolio return (Rbi,t ) over the same year. Benchmark portfolio returns are calculated for 25 Fama and French (1998) portfolios, constructed based on size and book-to-market value. Returns are calculated by compounding monthly returns over each year. Each stock is then matched at the beginning of each year to one of the portfolios based on its size and book-to-market. With the exception of leverage, all independent variables are standardized for size by the lagged market value of equity. The independent variables include the change and level of cash holdings for the firm, and other controls that may be correlated with changes in cash and firm value, including changes in a firm’s financing policy, profitability, and investment policy. The financing variables include Cash, Int (which represents interest expense), Div (which represents cash dividends), Lev (market leverage), and NF (net financing, defined as total equity issuance minus repurchases plus debt issuance minus debt redemption). The profitability variables include Earnings (defined 10

as earnings before interest and extraordinary items). And the investment variables include R&D (research and development expenditures) and NA (net assets). In addition to these variables, there are three variables with interaction terms. The first i h  , is the interaction of market leverage with the change in cash variable, (Lev) × (4Cash) MV E for the firm; this is included in order to estimate the effect of leverage on the marginal value of h   i Cash) (4Cash) cash for different levels of cash holdings. The second variable, (F oreign × , MV E MV E is the interaction of the previous period’s level of foreign-held cash with the change in cash for the firm; this is included in order to estimate the effect that different levels of foreign cash h   i Cash) (4Cash) holdings have on the value of cash. Similarly, the third variable, (Domestic × , MV E MV E is the interaction of the previous period’s level of domestically-held cash with the change in cash for the firm. These two terms replace the single aggregated cash variable used by Faulkender and Wang (2006). They note that since independent variables are scaled by the lagged market value of equity, and since the stock return in the dependent variable is the spread (M V Et − M V Et−1 ) divided by M V Et−1 , the estimated coefficients can be interpreted as the marginal change associated with a change in the corresponding independent variable. Therefore, the coefficient on the change in cash measures the dollar change in shareholder value stemming from a one-dollar change in the amount of cash held by the firm.12 The overall methodology is similar to a long-term event study, examining abnormal (unexpected) returns within an event window defined to be the fiscal year. The coefficients of the foreign and domestic cash interaction variables (λ11 and λ12 above), in addition to the coefficients of the other variables interacted with the change in cash, provide estimates of the marginal value to shareholders of foreign-held and domesticallyheld cash, respectively, for different levels of cash holdings. Therefore, following (5) above:  Value(F oreign Cash) = γ1 + γ10 Mean(Lev) + γ11 Mean 12

F oreign Cash MV E

 ,

(6)

According to Faulkender and Wang (2006), the marginal value of cash to shareholders is a function of the estimated coefficient on the change in cash, and the interactions with the level of cash holdings and leverage. Therefore, following from (5), the marginal value of cash is: Value(Cash) = γ1 + γ9 Mean(Leverage) + γ10 Mean(Cash) where the above represent the average values of the variables for the sample.

11

and  Value(Domestic Cash) = γ1 + γ10 Mean(Lev) + γ12 Mean

Domestic Cash MV E

 .

(7)

Thus, the value of either foreign-held or domestically-held cash for the average firm is given by the marginal value of $1 of cash not accounting for leverage effects or levels of cash holdings (λ1 above) added to the marginal value of $1 cash accounting for levels of leverage and cash holdings (which, for the average firm, is calculated using the mean values of Lev, Foreign Cash scaled by the market value of equity, and Domestic Cash scaled by the market value of equity for the sample). As discussed earlier, if firms are stockpiling cash abroad in order to avoid or delay payment of the repatriation tax (and the cash is thus inefficiently “trapped” abroad), then we would expect foreign cash to be worth less than domestic cash. However, if factors other than taxes, such as valuable growth opportunities abroad, are the main driver behind firms’ decision to hold cash abroad and these result in smaller free-cash-flow problems abroad, then we would expect foreign cash to be worth more than domestic cash. These two effects pull against each other and lead to the following Hypothesis: Hypothesis 1: The values of foreign and domestic cash may be significantly different from one another. If the repatriation-tax effect is dominant, the value of foreign cash will be lower than the value of domestic cash. If the foreigninvestment-opportunities effect is dominant, the value of foreign cash will be higher than the value of domestic cash.

2.3

Calculating the Effect of Investment Opportunities on the Value of Cash

In order to further investigate the foreign-growth-opportunities argument implicitly in Hypothesis 1, the final step is to examine whether the difference between the shareholder 12

valuation of cash held abroad and cash held domestically is indeed driven by the difference between foreign and domestic reinvestment opportunities. To see how relative reinvestment/growth opportunities may affect the valuation difference between foreign and domestic cash, note that most U.S. companies pursue foreign markets because these markets offer higher growth opportunities than domestic operations in the U.S.13 Examples are numerous. The emerging markets in Brazil, Russia, India, and China are all attractive foreign destinations for U.S. companies, and all these are high-growth markets. Moreover, even in developed countries like those in Europe, growth potential may be higher as the market for the firm’s products may be less saturated than in the U.S. Higher growth opportunities also mean that the demand for investment is higher, which further implies that internally-generated cash in foreign operations will have a higher shadow price for being used for purposes other than reinvestment compared to internally-generated cash in domestic operations.14 This higher shadow price will tend to lead to a smaller free-cash-flow problem (e.g. Jensen (1986)) in foreign subsidiaries than in domestic operations as managers in foreign subsidiaries will perceive wasteful expenditures on things such as perquisites (e.g. Jensen and Meckling (1976)) and pet projects to be relatively more costly. Conditional on growth opportunities abroad being sufficiently high, U.S. corporations would be able to use up all of their foreign cash in investments outside the U.S. and there would be little that is repatriated to the U.S. If this is true, the repatriation tax would not be significant and its impact on the valuation of foreign cash would be relatively small. Moreover, foreign cash in firms with good growth opportunities should be valued significantly higher than foreign cash in firms with poor growth opportunities, since the latter are more likely to be adversely affected by the repatriation tax burden and free-cash-flow problems stemming from a relatively low shadow price of cash. This further implies that 13

Note that the sample in this paper consists only of firms that are incorporated in the U.S. In part due to the repatriation tax, many foreign companies keep foreign and domestic cash in separate buckets when it comes to allocating it for investments. For example, Emerson Electric, a multinational firm with $21 billion in sales, repatriates very little of its foreign cash back to the US. It consolidates this cash in special subsidiaries in Switzerland and Ireland and uses it to finance various investments abroad. Its investments in the U.S. are financed mainly with cash generated domestically. 14

13

foreign cash for firms with poor investment opportunities should be valued at a discount to $1, while foreign cash for firms with good investment opportunities should be valued at a premium to $1, particularly since real options to invest in the future are likely to involve lumpy investments, so the firm cannot invest all the way up to the point at which the NPV of the investment is zero, i.e. these will be options to invest in positive NPV projects. To summarize, this discussion leads to the following hypothesis: Hypothesis 2: Firms with good investment opportunities abroad should have substantially higher values of foreign cash than firms with poor investment opportunities abroad. In addition, foreign cash for firms with good investment opportunities abroad should be worth more than $1, while foreign cash for firms with poor investment opportunities abroad should be worth less than $1. The tests will focus on firms that derive a relatively high fraction of their income from abroad because the hypothesized effects on the value of cash are expected to be more pronounced for these firms.15 To test these predictions, this paper uses an extension of the methodology used by Dittmar and Mahrt-Smith (2007).16 Thus, the main regression specification below 15

As explained below, extra terms are added in the regression to examine whether good or poor investment opportunities abroad are driving the differential value of cash. So while all firms, including those with a low fraction of foreign income, are included in the regression, the tests focus on whether it is the firms with high levels of foreign operations that are experiencing the largest differences between the values of foreign and domestic cash. That is, the dummy variable that I focus on in my tests is the one associated with the firms that have high foreign operations. 16 This methodology builds on the regression specification of Faulkender and Wang (2006) and examines the effect of differences in corporate governance on the value of cash by including an additional regression term which interacts corporate governance with the change in cash for each firm-year.

14

is run: 

     (4Cash)i,t (4Earnings)i,t (4N A)i,t (Ri,t − Rbi,t ) = τ0 + τ1 + τ2 + τ3 M V Ei,t−1 M V Ei,t−1 M V Ei,t−1       (4Int) (4Div)i,t (4R&D)i,t + τ5 + τ6 +τ4 M V Ei,t−1 M V Ei,t−1 M V Ei,t−1     (Cash)i,t−1 (N F )i,t + τ8 (Lev)i,t + τ9 +τ7 M V Ei,t−1 M V Ei,t−1    (4Cash)i,t +τ10 (Lev)i,t × M V Ei,t−1     (4Cash)i,t (F oreign Cash)i,t−1 × +τ11 M V Ei,t−1 M V Ei,t−1     (Domestic Cash)i,t−1 (4Cash)i,t +τ12 × M V Ei,t−1 M V Ei,t−1    (4Cash)i,t +τ13 (High F or Ops)i,t × (Good Inv Opp)i,t × M V Ei,t−1    (4Cash)i,t + εi,t . +τ14 (High F or Ops)i,t × (P oor Inv Opp)i,t × M V Ei,t−1 (8)

The variables are defined in the same way as in regression (5), except for the last two terms. High For Ops is a dummy variable that indicates whether a firm has a high level of foreign operations. It takes a value of 1 if a firm-year is in the top half (of all the firms reporting foreign income) in terms of foreign operations as a percentage of total operations (defined as

F oreign Income ), T otal Income

and a value of 0 otherwise. Using top and bottom terciles or

quartiles leads to similar results. Good Inv Opp is a dummy variable that indicates whether a firm has good investment opportunities. It takes a value of 1 if a firm-year is in the top half of the sample in terms of Tobin’s q (q-ratio). Poor Inv Opp is defined in a similar way – it indicates whether a firm has poor investment opportunities, and takes a value of 1 if a firm-year is in the bottom half of the sample in terms of q-ratio, and 0 otherwise. The use of the q-ratio as an indicator of investment opportunities for firms follows from Tobin and Brainard (1968, 1977), and various other papers (e.g. Lang and Litzenberger (1988)) that relate Tobin’s q to overinvestment. The calculation of the q-ratio in this paper follows 15

the procedure of Lindenberg and Ross (1981) and the approximation by Chung and Pruitt (1994). It is defined as:

q ratio =

M V E + P ref erred Stock + Debt . TA

(9)

MVE, Debt, and TA are defined above. Preferred Stock is the liquidating value of the firm’s outstanding preferred stock. By defining the dummy variables this way and interacting them, the impact of foreign investment opportunities on the value of cash can be determined. For example, for firms with good investment opportunities and high levels of foreign operations, τ13 will represent the contribution to the value of cash stemming from the good foreign investment opportunities, while the τ14 term will be 0. For firms with poor investment opportunities and high levels of foreign operations, the τ13 term will be 0, and τ14 will represent the deduction from the value of cash stemming from the poor foreign investment opportunities.17 Therefore, in a manner similar to (6) and (7) above, by focusing on the respective coefficients in (8), one can examine the marginal value of cash held abroad and cash held domestically for U.S. firms with different levels of foreign investment opportunities, and thus infer whether foreign investment opportunities are driving the difference between the cash values:18

Value(F oreign Cash | good investment opportunities abroad) =   F oreign Cash , τ1 + τ13 + τ10 M ean(Lev) + τ11 M ean MV E (10) 17

For firms with high levels of foreign operations, it is assumed that most of the investment opportunities will be abroad. This is a necessary assumption because it is not possible with the current data to separately measure Tobin’s q for investment opportunities in foreign subsidiaries. 18 An alternative approach to interacting the two dummy variables would be to conditionally run the regression for only firms with high foreign operations – running separate regressions for the firms with high foreign operations and either good or poor investment opportunities. This is also in line with Faulkender and Wang (2006), who consider separate results for financially constrained vs. unconstrained firms. Similarly, one could also define the dummy variables using terciles or quartiles to denote high values. These approaches are considered separately for the regression in untabulated results, and the results are robust to these alternate specifications.

16

Value(Domestic Cash | good investment opportunities abroad) =   Domestic Cash , τ1 + τ13 + τ10 M ean(Lev) + τ12 M ean MV E (11)

Value(F oreign Cash | poor investment opportunities abroad) =   F oreign Cash τ1 + τ14 + τ10 M ean(Lev) + τ11 M ean , MV E (12)

and

Value(Domestic Cash | poor investment opportunities abroad) =   Domestic Cash τ1 + τ14 + τ10 M ean(Lev) + τ12 M ean . MV E (13)

Equations (10)-(13) above are interpreted in a similar way to equations (6) and (7). By adding the coefficient of the interaction terms between 4Cash, High For Ops, and either Good Inv Opp or Poor Inv Opp, the marginal value of foreign and domestic cash can be estimated specifically for firms with good or poor investment opportunities abroad. Equation (10) gives the marginal value of foreign cash, accounting for a firm’s amount of market leverage and scaled foreign cash holdings, for a firm with a high degree of foreign operations and good investment opportunities. Thus, it serves as a proxy for what the value of foreign cash is for a firm with good investment opportunities abroad. τ1 gives the marginal value of cash for a firm, holding constant the amounts of leverage and cash that the firm holds (ie τ1 would give the marginal value of cash for a firm with no leverage, no cash holdings, and low levels of foreign operations). Adding in the coefficients τ10 and τ11 multiplied by the mean values of Lev and scaled Foreign Cash, respectively, for the sample

17

gives the value of foreign cash for the average firm in the sample, accounting for the levels of leverage and foreign cash held. And adding τ13 gives the portion of the marginal value of cash specifically contributed by a firm’s good foreign investment opportunities. Similarly, equation (11) calculates the value of domestic cash for the average firm with good investment opportunities abroad. Equations (12) and (13) calculate the value of foreign and domestic cash for the average firm with poor investment opportunities abroad, with τ14 representing the portion of the marginal value of cash specifically affected by a firm’s good poor investment opportunities.

3

Data and Summary Statistics

The data in this paper are taken from the CRSP/COMPUSTAT merged database, available through the Wharton Research Data Service (WRDS). The sample includes all U.S. firms incorporated in the U.S. with total assets exceeding $100 million, excluding all financial firms and utilities (SIC codes between 6000 and 6999, and between 4900 and 4999, respectively).19 Financial firms are excluded because they are required to meet certain statutory cash-asset reserve requirements, and they also hold substantial amounts of marketable securities that are included in cash. Utilities are excluded because their cash holdings are subject to regulatory supervision in many cases. Firms that are incorporated outside of the U.S. are excluded from the sample because the U.S. repatriation tax does not apply to them. The final sample includes 10,899 firm-year observations for a total of 1,643 firms. Firms which report no foreign income or foreign taxes are excluded from the sample. The sample period ranges from 1980 to 2006. All data are converted to real 2006 dollar values using the Consumer Price Index (CPI). In all regressions, the independent variables are winsorized at the 1% level in order to reduce the impact of outliers.20 The marginal 19

This sample restriction of firms with total assets exceeding $100 million is to focus on comparably larger firms that are more likely to have foreign operations and hold cash abroad. 20 The exceptions to this are any dummy variables and ln(T A) from regression (1), which already accounts for outliers in size.

18

tax rate data is estimated following the method of Graham (1996). Firm-years for which marginal tax rate data are not available are dropped from the regressions. The benchmark portfolio returns are taken from Kenneth French’s website. The sample period ends in 2006 to exclude the potential effect of the sub-prime crisis which began in late 2007. Table 1 includes the summary statistics for the variables in regression (1), which is used to estimate the amounts of foreign and domestic cash. The table indicates that firms in the sample held on average 12.11% of their net assets in the form of cash (an actual cash amount of $951.77 million on average). This is striking because it is roughly comparable in magnitude to the average annual capital expenditure, which is 7.32% of total assets (an actual amount of $643.16 million). Domestic Income is about 8.09% of total assets for the average firm (an actual amount of $710.82 million), and Foreign Income is about 3.70% of total assets for the average firm (an actual amount of $325.10 million). Roughly half of the firms in the sample do not report any foreign income. [Insert Table 1 here] Table 2 includes the summary statistics for regression (8), which is used to estimate the value of foreign and domestic cash. Excess Stock Return over the sample period is slightly negative, on average. The level of cash holdings is on average 11.77% of the market value of equity for firms. Lev indicates that, over the sample period, firms on average have 24.03% debt in their capital structures. Foreign Operations indicates that the average firm in the sample derives roughly 36.76% of its total income from its foreign operations. The numbers for Foreign Cash, however, indicate that in general the distribution of foreign cash holdings is skewed – a few firms hold large amounts of foreign cash, as a percentage of the market value of equity of the firm. This may also be because not all of the firms with foreign cash are being picked up due to noise in the estimation of the level of foreign cash, and also since firms are not required to report foreign income. [Insert Table 2 here] 19

4

Main Results

In this section, the results of the empirical tests are discussed using the approach described in Section 2. Section 4.1 details the estimation of the amounts of foreign and domestic cash for firms. Section 4.2 estimates the market values of foreign and domestic cash for the average firm. Section 4.3 examines whether differences in the quality of investment opportunities explains the difference between the values of foreign and domestic cash. Section 4.4 provides robustness checks, by estimating the results using alternative proxies for the levels of foreign and domestic cash. Section 4.5 conducts an event study which exploits an exogenous change in the tax code, in order to lend further support for the results.

4.1

Estimating the Amounts of Foreign and Domestic Cash

Table 3 contains the results of regression (1), which are used to estimate the amounts of foreign and domestic cash. The regression results are similar to the results obtained by Foley et al. (2007). The Adjusted R2 of 0.33 is the same as the Foley et al. (2007) regression R2 of 0.33. R&D/T A, CAP EX/T A, Lev, CF Std Dev, Div Pmt, ln(T A), and BV E/M V E all share the same signs as the original regression, and are significant. Tax Burden is significant and has a positive sign, and this corroborates the results of Foley et al. (2007) – an increase in the repatriation tax burden implies an increase in cash holdings for firms. Domestic Income/T A has the same sign as in the original regression, but is significant at the 0.01 level in this regression as opposed to insignificant in the original regression. [Insert Table 3 here] The coefficient of F oreign Income/T A is significant at the 0.10 level, and has a negative sign, which means that an increase in foreign income will decrease overall firm cash holdings. While this finding is at odds with the results of Foley et al. (2007), it is in line with the investment-opportunities-abroad Hypothesis posited in this paper. If a firm’s foreign operations are experiencing high growth, the firm will have to make sizable reinvestments 20

to sustain this growth, and these reinvestments may well exceed the contemporaneous cash flows being generated abroad. This would not only lead to all of the foreign cash being reinvested, but also some domestic cash being siphoned off to finance the reinvestment, leading to the relationship documented here. This is an economically plausible explanation since an important reason why many US firms expand internationally is to pursue higher growth prospects abroad.21 The regression coefficients on and Tax Burden, β9 and β10 , respectively, are used in equation (3) to obtain an estimate of the amount of foreign cash. The amount of domestic cash is obtained by deducting the estimated amount of foreign cash from total cash, as shown in equation (4).

4.2

Estimating the Market Values of Foreign and Domestic Cash: Testing Hypothesis 1

Using these estimates, the marginal values of foreign and domestic cash are estimated from regression (5). The results of this regression are included in Table 4. Overall, the regression results are similar to those in Faulkender and Wang (2006). The Adjusted R2 is 0.17, which is slightly lower than 0.20 found by Faulkender and Wang (2006). The coefficients for (4Cash)t , (4Earnings)t , (4N A)t , (4Int)t , (4Div)t , (Cash)t−1 , (Lev)t , and (Lev)t × 4(Cash)t are all significant, have the same sign, and are of a comparable magnitude to the Faulkender and Wang (2006) regression. Compared to Faulkender and Wang (2006), the (N F )t coefficient has the opposite sign and is insignificant at the 0.10 level, and (4R&D)t is also insignificant and has a much smaller coefficient. Differences related to these two variables may be due to the difference between the samples in this regression and the original regression. In their original regression, Faulkender and Wang (2006) have a sample size of more than 80,000 observations, while this regression has approximately 10,900 because of 21

Since the sample in this paper is larger than that in Foley et al. (2007), it is possible that Foley et al. (2007) had a sample which excluded some of the high-growth firms in the current sample.

21

the need to drop observations due to missing data while running regression (1) to estimate the levels of cash. In addition, this regression includes data up to and including 2006, while Faulkender and Wang (2006) include data only up to 2001; thus, events after 2001 may have had an effect on the regression results. These differences notwithstanding, these insignificant variables are included as controls, and are not central to the interpretation of the results. [Insert Table 4 here] The coefficient for (∆Cash)t × (Domestic Cash)t−1 is negative and significant, as expected. As explained above, since the stock return is the excess equity return divided by the lagged market value of equity and the independent variables are scaled by the lagged market value of equity, the estimated coefficients can be interpreted as the marginal value associated with a change in the corresponding independent variable. The original regression indicates a negative coefficient for the interaction of the change in cash and lagged holdings of total cash, indicating that as firms’ cash positions improve, the marginal value of that cash declines, which is economically intuitive. The negative coefficient for the interaction of the change in cash and lagged holdings of domestic cash concurs with the original findings, and indicates that this declining marginal value of cash holds for domestic cash as well. This is the case because cash-starved firms are more likely than others to raise external funds that may be costly, and thus would receive a greater benefit from having more internal funds (cash). However, the coefficient for (∆Cash)t × (F oreign Cash)t−1 is positive and insignificant. This is somewhat surprising, but it may be because there are two forces pulling in opposite directions. On the one hand, the higher the amount of foreign cash, the lower is its marginal value, as I have argued in the case of domestic cash. On the other hand, a higher amount of foreign cash may also signify higher growth opportunities abroad, which would imply a higher marginal value of cash. This could lead to a positive but insignificant coefficient as these two opposing forces partially negate each other. In addition, as can be seen from the summary statistics, the amount of foreign cash held is driven by a small number of

22

firms. This might be a reflection of missing data, since firms are not required to report their foreign operations. This may also be a factor leading to the insignificant coefficient. With these coefficient values, the marginal value of cash held abroad and cash held domestically can now be estimated and examined for the average firm. Plugging the coefficient values above into equations (6) and (7) gives a marginal value of cash held abroad of $1.06 and a marginal value of cash held domestically of $0.93. Thus, cash held abroad is valued at a premium to $1, and cash held domestically is valued at a discount to $1. These numbers generally fall in line with the results obtained by Faulkender and Wang (2006), who find that the marginal value of overall cash is approximately $0.94 for the average firm. The results also show that the average marginal value of foreign cash is statistically significantly higher than the average marginal value of domestic cash.22 Moreover, these results are also robust to an alternate estimate of the amounts of foreign and domestic cash, which is detailed in Section 4.4. Overall, the results provide support for the prediction in Hypothesis 1 – the values of foreign and domestic cash for the average firm are significantly different from one another. In addition, foreign cash is valued higher than domestic cash, indicating that the effect of the repatriation tax is more than offset by the greater reinvestment opportunities and smaller free-cash-flow problems in foreign operations than in domestic operations. There are two possible reasons for this result. First, the repatriation tax may have a rather small influence on the value of foreign cash because, as suggested by Foley et. al. (2007), strategies to avoid or indefinitely defer this tax may not be too costly. Moreover, the effects of the tax break from the American Jobs Creation Act of 2004 may also be playing a role. As Albring, Dzuranin, and Mills (2005) point out, this tax break was sizable. Second, as explained by Faulkender and Petersen (2012), if a firm is able to finance domestic investments from internal funds or a capital market with relatively small frictions, then it does not have an incentive to repatriate foreign cash as long as the free-cash-flow problems in holding the cash abroad are not too 22

In a one and two-sided paired t-test between means rejects the null hypothesis of no difference between the values of foreign and domestic cash at the 0.01% level.

23

large. Section 2.3 discussed how more profitable reinvestment opportunities abroad could increase the shadow price of cash and thereby deter wasteful deployment of cash. But there may also be an additional reason, related to career concerns, that may lead to smaller freecash-flow problems abroad than domestically. U.S. companies tend to send relatively young (high-potential) managers abroad as expatriates to lead foreign subsidiaries. As argued by Chevalier and Ellison (1999), younger managers may have stronger career concerns, so free-cash-flow problems abroad are likely to be smaller.

4.3

Determining if Differences in Investment Opportunities Drive the Difference in the Values of Foreign and Domestic Cash: Testing Hypothesis 2

To examine whether differences in investment opportunities are driving the higher value of cash held abroad versus cash held domestically, regression (8) is estimated. The results are presented in Table 5. All coefficients from regression (5) remain mostly unchanged in terms of magnitude, sign, and significance. The coefficient of (∆Cash)t × (High F or Ops)t × (Good Inv Opp)t is positive and significant at the 5% level, and the coefficient of (∆Cash)t × (High F or Ops)t × (P oor Inv Opp)t is negative, although insignificant. The signs of the two coefficients indicate that a firm with good foreign investment opportunities will have a higher marginal value of cash relative to a firm with poor investment opportunities, which is in line with what was expected. To obtain the marginal values of cash for firms with good foreign investment opportunities and poor foreign investment opportunities, the coefficient values of regression (8) are plugged into (10)-(13). For firms with high levels of foreign operations and good investment opportunities, the marginal values of foreign and domestic cash are $1.42 and $1.30, respectively. For firms with high levels of foreign operations and poor investment opportunities, the marginal values of foreign-held and domestically-held cash are $0.92 and $0.80, respectively.

24

[Insert Table 5 here] Overall, the results provide strong support for Hypothesis 2: firms use their cash held abroad to finance investment opportunities, and firms with better growth opportunities abroad have higher values of domestic and foreign cash. In addition, the results indicate that cash held domestically and abroad are both valued at a premium to $1.00 in high-growth firms. This suggests that, for these firms, the repatriation tax disadvantage of holding cash abroad is swamped by the valuable investment opportunities of these firms. However, for firms with overall poor growth opportunities, cash held domestically and abroad are both valued at a discount to $1.00. Moreover, the value of foreign cash for firms with poor investment opportunities abroad ($0.92) is substantially lower than the value of foreign cash for firms with good investment opportunities abroad ($1.42), likely reflecting the potential impact of the repatriation tax burden as well as free-cash-flow problems.

4.4

Robustness and Alternate Measures of Cash

As the results in sub-sections 4.1 to 4.3 are based upon an estimation of the amounts of foreign-held and domestically-held cash that firms carry, a concern is whether the results are robust to alternative estimations of these cash amounts. An alternative method is presented to proxy for the amounts of cash, and the values of foreign and domestic cash are calculated using these estimates. The estimation method is to use coefficient estimates from Foley et al. (2007) in their foreign cash regression, which uses BEA data on levels of foreign cash, to create estimates of foreign and domestic cash for firms. Specifically, Foley et al. (2007) run the following OLS

25

specification:  ln

F oreign Cash NA

 i,t



Domestic Income TA



= γ0 + γ1 (Eff Repat T ax Rate)i,t + γ2 i,t   F oreign Income +γ3 + γ4 (ln (T A))i,t + γ5 (Div P mt)i,t TA i,t     R&D BV E + γ7 (Std Dev Op Income)i,t + γ8 +γ6 M V E i,t T A i,t   CAP EX +γ9 + γ10 (Lev)i,t + εi,t . (14) TA i,t

The variables are defined analogously to regression (1). Eff Repat Tax Rate is a firm’s effective repatriation tax rate, and is defined as the maximum of zero and the difference between a firm’s weighted foreign tax rate and its marginal effective tax rate. The sample period Using the coefficient estimates for (14) from Foley et al. (2007), and rearranging (14) yields the following estimates for foreign and domestic cash:23 " 0

Est(F oreign Cash)i,t = (N A)i,t

 exp −5.608 + 8.802(Eff Repat T ax)i,t



 F oreign Income −1.576 + 14.088 TA i,t i,t   BV E +0.038(ln(T A))i,t − 0.020(Div P mt)i,t + 0.167 M V E i,t     R&D CAP EX − 2.139 +0.875 (CF Std Dev)i,t + 4.382 T A i,t TA i,t # Domestic Income TA





−1.140 (Lev)i,t

(15)

and 0

0

Est (Domestic Cash)i,t = (Cash)i,t − (F oreign Cash)i,t . 23

(16)

Foley et al (2007) use operating income standard deviation instead of cash flow standard deviation because of limitations on the duration of their sample. In (15), cash flow standard deviation is used. This facilitates calculations, and follows Opler et al (1999).

26

Negative estimates of cash are truncated at zero. In (16), Eff Repat Tax is a proxy for Eff Repat Tax Rate, since the latter is measured using BEA data for the exact foreign tax that firms face. It is defined as:

(Eff Repat T ax)i,t =

n max (U ST axRate)i,t ×

F oreign Income TA i,t




o − (F oreignT axesP aid)i,t , 0

(T A)i,t (17)

Summary statistics for (15), (16), and (17) when applied to the data are included in Table 6. The sample period runs from 1980-2006, which is not too different from the sample period of 1982-2004 in Foley et al (2007).24 The mean and standard deviation of these estimates of foreign and domestic cash are close to the estimates obtained earlier in the paper. [Insert Table 6 here] Column (1) of Table 7 includes regression results for equation (5) for estimating the values of foreign and domestic cash. The magnitudes and signs of the regression coefficients are very similar to those obtained using the original estimates of foreign and domestic cash. Overall, the results indicate that for the average firm a dollar of cash held domestically is valued at $0.95, and a dollar of cash held abroad is valued at $1.11. These numbers are very close to those obtained in Section 4.2 (domestic cash valued at $0.93 and foreign cash valued at $1.06), and once again indicate that the value of foreign cash is valued at a premium to $1, and is significantly higher than the value of domestic cash, which is valued at a discount to $1. Column (2) of Table 7 includes regression results for equation (8). The magnitudes and signs of the regression coefficients are also very similar to those obtained in Section 4.3. The results indicate that for firms with high levels of foreign operations and good investment opportunities, the marginal values of foreign and domestic cash are $1.43 and $1.29, respectively. For firms with high levels of foreign operations and poor investment opportunities, the marginal values of foreign-held and domestically-held cash are $0.95 and 24

With the caveat being that the data in this paper also include post-repatriation-tax-holiday data, whereas Foley et al’s (2007) analysis does not.

27

.

$0.81, respectively. These results are similar to those obtained earlier, and once again provide strong support for the hypotheses developed in Section 2.3. Thus, the results seem robust to an alternative estimation of the amount of cash held by firms. [Insert Table 7 here]

5

Additional Evidence from the American Jobs Creation Act of 2004

As a further test of the hypothesis that investment opportunities are a key driver of the results documented here, an exogenous change in the tax code can be exploited to run an additional test. In 2004, Congress passed the American Jobs Creation Act which, among other things, provided a one-time break in the repatriation tax for multinational firms. Thus, any firms that had stockpiled cash abroad because of the repatriation tax had an opportunity to bring the cash back to the U.S. without incurring the penalty of the tax. According to Albring et al. (2005), this created an opportunity for significant tax savings for multinational firms. If investment opportunities are indeed playing a significant role in the results, then the tax holiday should affect firms differently depending on their investment opportunities. In particular, firms with poor investment opportunities abroad are likely to be stockpiling cash abroad to avoid the repatriation tax rather than to reinvest the cash abroad, so these firms should experience a significant positive effect from the passage of the tax holiday. However, firms with good investment opportunities abroad are holding cash abroad mainly in order to invest in projects and to a lesser extent to avoid the repatriation tax, so the tax break should not have a significant effect for these firms.

28

5.1

Event Study Methodology and Results

In order to test this prediction, an event study methodology is used to examine the cumulative abnormal returns (CARs) of firms around when the American Jobs Creation Act of 2004 was passed. The CARs are calculated following Campbell, Lo, and MacKinlay (1997) using both the market model and the Fama and French (1993) three-factor model. Denote event time by τ . In this case, τ = 0 represents June 17th – the day that the Act was passed in the U.S. House of Representatives.25 The market model can be represented as

Ri,τ = αi + βi Rm,τ + εi,τ ,

(18)

where Ri,τ is the return of stock i at time τ , and Rm,τ is the market return at time τ (proxied by the value-weighted . The Fama-French three-factor model can be represented as

Ri,τ = αF F,i + βm,i Rm,τ + βhml,i Rhml,τ + βsmb,i Rsmb,τ + i,τ ,

(19)

where Rhml and Rsmb are the returns of the HML and SMB portfolios, respectively. For each security i, regressions (18) and (19) are estimated in order to produce parameter estimates for the regression coefficients. Denoting the parameter estimates via OLS from (18) and (19) ˆ the estimated abnormal returns of security i at time τ for the market model by α ˆ and β, and the Fama-French model are:

m ARi,τ = Ri,τ − α ˆ i − βˆi Rm,τ , 25

(20)

At the time, the Republican party controlled both the House and the Senate. Thus the passing of the legislation through the House likely resolved a good deal of uncertainty about the prospects of the bill, as the Senate was likely to follow suit (apart from disagreements surrounding reconciliation of the House and Senate versions of the bill). That being said, any remaining uncertainty about the passage of the bill, as well as which firms would be most affected, was likely to be resolved over a period of time rather than over one or two days, as is explained below.

29

and FF ARi,τ = Ri,τ − α ˆ F F,i − βˆm,i Rm,τ − βˆhml,i Rhml,τ − βˆsmb,i Rsmb,τ .

(21)

Using the individual firm abnormal returns, average abnormal returns for each event date FF τ (denoted by ARm τ and ARτ ) are calculated separately for firms with for firms with good

investment opportunities abroad (high degrees of foreign operations and good investment opportunities) and for firms with poor investment opportunities abroad (high degrees of foreign operations but poor investment opportunities). As before, a firm is defined as having a high degree of foreign operations if it is in the top half of firms in the sample reporting F oreign Income TA

as of the end of the previous year (here: 2003), and a firm is defined as having

good investment opportunities if it is in the top half of firms in the sample for Tobin’s q, defined in (9), as of the end of 2003.26 Finally, average cumulative abnormal returns (CARs) between two dates τ1 and τ2 are defined as the sum of the average abnormal returns between those dates: CARτ1 ,τ2 ≡

τ2 X

ARτ ,

(22)

τ =τ1

where (22) is calculated separately for firms with good investment opportunities abroad and for firms with poor investment opportunities abroad. Table 8 below gives the average abnormal returns and cumulative average abnormal returns, calculated using both the market model and the Fama-French model, for firms with good and poor investment opportunities abroad. The returns are shown for a window of 15 days surrounding June 17th (τ = 0), which was the day the Act passed in the U.S. House. The reason for focusing on this interval is that the bill was introduced to Congress and the public on June 4th, a few days before it won passage in the House. During this period (and even after the bill passed the House), we would expect to see a gradual revelation 26

Because these variables are only reported on a yearly basis, the values as of the previous fiscal year-end (i.e. 2003) are used. As before, for firms with high levels of foreign operations, the assumption is that most of the investment opportunities will be abroad. This is because it is not possible with the data to separately measure investment opportunities that are specifically based abroad. The results are also robust to defining high foreign operations or good investment opportunities as being in the upper tercile or quartile of the sample rather than the upper half.

30

of information regarding how likely the bill was to be passed, as well as an evolution of the market’s assessment of the foreign cash positions and investment opportunities of the affected firms. If this were the case, we would expect to see a steady positive increase in the abnormal returns for firms with poor foreign investment opportunities (rather than a large discontinuous jump over just one or two days), but flat/insignificant abnormal returns for firms with good foreign investment opportunities. The sample represents a total of 4,867 returns total of 157 firms – 97 firms with good investment opportunities abroad, and 60 firms with poor investment opportunities abroad. [Insert Table 8 Here] Figure 1 below plots the CARs as a function of event time for firms with good foreign investment opportunities, and for firms with poor foreign investment opportunities. The top graph does this with CARs calculated using the market model, and the bottom graph does this with CARs calculated using the Fama-French model. [Insert Figure 1 Here] As Figure 1 indicates, firms with good investment opportunities abroad did not seem to experience any significant abnormal returns either immediately before or after the American Jobs Creation Act was passed in the House – the dashed line for CARs seems to hover around 0. By contrast, firms with poor investment opportunities abroad seemed to experience positive cumulative abnormal returns before the Act was passed (i.e. around the time when it was introduced), and the CARs steadily increase around and past the date when the bill was passed. This seems to support the interpretation that firms with poor investment opportunities abroad are stockpiling cash abroad to avoid the repatriation tax rather than to reinvest the cash abroad, so thus these firms experience a significant positive effect from the passage of the tax holiday. However, firms with good investment opportunities abroad are holding cash abroad mainly in order to invest in projects and to a lesser extent to avoid the repatriation tax, so the tax break does not have a significant effect for these firms. 31

A formal statistical test buttresses the graphical evidence. Following Campbell, Lo, and MacKinlay (1997), under the null hypothesis that the expectation of the abnormal returns
¯τ21 ,τ2 , is zero, we can draw inferences about the CARs using the fact that CARτ1 ,τ2 ∼ N 0, σ where σ ¯τ21 ,τ2 is the variance of CARτ1 ,τ2 . We can then test the null hypothesis using CARτ1 ,τ2 a J1 =   ∼ N (0, 1), ˆ¯τ21 ,τ2 1/2 σ

(23)

where σ ¯ˆτ21 ,τ2 is the sample variance of CARτ1 ,τ2 . When using CARs calculated through the Fama-French three-factor model, applying (23) to firms with good foreign investment opportunities yields J1 = 0.71, and applying it to firms with poor foreign investment opportunities yields J1 = 3.29.27 These results provide statistical evidence that firms with poor investment opportunities abroad experienced positive and significant abnormal returns as a result of the American Jobs Creation Act, but firms with good investment opportunities abroad did not experience any significant abnormal returns, indicating that investment opportunities are a significant driver of multinational firms’ decision to hold cash abroad.28

5.2

Robustness of Event Study Results

As noted by Kolari and Pynn¨onen (2010) and others, when event dates are clustered, resulting in cross-sectional correlation of abnormal returns, standard errors can be severely biased in terms of over-rejecting the null hypothesis of zero average abnormal returns. Since there is only one common event date of interest amongst the firms in the sample, this cross-sectional correlation is a concern in this setting. As an alternative robustness check in order to account for this potential bias, the portfolio method following Jaffe (1974) is used to conduct the event study. 27

The significance of the results remains unchanged when using the market model to calculate abnormal returns. 28 These results are also consistent with Blouin and Krull (2009), who find that firms that repatriated cash under the American Jobs Creation Act of 2004 had poorer investment opportunities and higher free cash flows than non-repatriating firms.

32

Specifically, firm returns are aggregated into equal-weighted portfolios for each date in event time, and the event study analysis is applied to the portfolios. Thus, the following model is estimated over all trading days in 2004:

Rp,τ = αp + βm,p Rm,τ + (βhml,p Rhml + βsmb,p Rsmb ) + φp Iτ + εp,τ

(24)

In (24), p represents the equal weighted portfolio of all firms in the sample, which is considered for each trading day τ in 2004. Firms which reported both domestic and foreign income as of the end of 2003 are considered. Rp,τ is the equal-weighted average of returns for all firms in portfolio p on day τ . Rm,τ is the market return on date τ , which is the return of the value-weighted CRSP market index. Rhml and Rsmb are the returns on the Fama and French (1993) value and size portfolios, and are included as additional factors for robustness. Iτ is a dummy variable which takes a value of 1 if the date τ falls in the event window, which is between τ = −15 and τ = 15, and a value of 0 otherwise.29 The regression (24) is run separately by forming portfolios either for firms with good investment opportunities abroad (high degrees of foreign operations and good investment opportunities) or for firms with poor investment opportunities abroad (high degrees of foreign operations but poor investment opportunities), as defined earlier. In specification (24), the coefficient on the dummy variable, φp , thus corresponds to the average abnormal returns surrounding the enactment of the bill. Based on the previous discussion, it is predicted that φp will be positive and significant for firms with poor investment opportunities abroad, but insignificant for firms with good investment opportunities abroad. [Insert Table 9 here] The results of regression (24) are included in Table 9. The specifications in columns (1) and (2) form the portfolios using only firms with poor investment opportunities abroad, and 29

Thus this event window captures the period in time from when the bill was introduced and when it passed the House, during which time information about the probability of passage and the foreign cash holdings of firms was likely slowly revealed to the market.

33

the specifications in columns (3) and (4) forms the portfolios using only firms with good investment opportunities abroad. The results indicated that the passage of the American Jobs Creation Act of 2004 resulted in positive and significant average abnormal returns of 0.2% for firms with poor investment opportunities abroad. However, for firms with good investment opportunities abroad, the enactment of the bill resulted in comparatively smaller average abnormal returns that are not significantly different from zero. Moreover, the results are robust to using the Fama and French (1993) factors in addition to the market factor in calculating the abnormal returns. These results are consistent with those in Section 5.1, and support the prediction that firms with good investment opportunities abroad experienced a negligible stock price reaction following enactment of the bill, while firms with poor investment opportunities abroad experienced a significant positive stock price reaction following the enactment of the bill.

6

Conclusion

This paper examines the average marginal values of cash held domestically in the U.S. and cash held abroad, and the relative roles of the repatriation tax and investment opportunities in driving the difference in the values of foreign and domestic cash. The two main results are as follows. First, cash held abroad is valued more highly by shareholders than cash held domestically – the marginal value of cash held abroad is $1.07 on average, and the marginal value of cash held domestically is $0.94 on average. Second, investment opportunities significantly impact the value of cash. For firms with high growth opportunities, the values of cash held domestically and cash held abroad exceed $1, whereas for firms with low growth opportunities, the values of cash held domestically and that held abroad are less than $1. The results indicate that investment opportunities are an important reason for why firms are stockpiling cash abroad, and that for many firms (and on average), it is likely to be a more important reason than the repatriation tax. Additional evidence from the change in

34

the tax code provided by the enactment of the American Jobs Creation Act of 2004 provides further support for these results. There is a caveat about one limitation of the analysis. The Faulkender and Wang (2006) approach does not deal with the issue that there is an endogeneity problem that arises from the fact that more profitable firms tend to generate more cash, so it may not be that cash drives the value but also that value determines how much cash the firm holds. Although market-based measures of value are forward looking, the fact that they may be correlated with contemporaneous profitability makes the interpretation of the regression coefficients tricky. The focus of this paper has been on how the determinants of the demand for cash within firms – as represented by investment opportunities that firms have and the taxes in the domains in which they operate – are correlated with how investors value the cash these firms hold on their balance sheets. An obvious factor that is not accounted for is the supply side. One can imagine that the market value of on-balance-sheet cash is also affected by the cost and availability of liquidity for firms through, say, the commercial paper market and bank lines of credit. It would be interesting to include these factors in future research.

35

References [1] Acharya, Viral V., Heitor Almeida, and Murillo Campello. ”Is cash negative debt? A hedging perspective on corporate financial policies.” Journal of Financial Intermediation 16, no. 4 (2007): 515-554. [2] Albring, Susan M., Ann C. Dzuranin, and Lillian F. Mills. ”Tax savings on repatriations of foreign earnings under the Jobs Act.” Tax Notes, Feb 15 (2005): C3. [3] Bates, Thomas W., Kathleen M. Kahle, and Ren´e M. Stulz. ”Why do US firms hold so much more cash than they used to?.” The Journal of Finance 64, no. 5 (2009): 1985-2021. [4] Blouin, Jennifer, and Linda Krull. ”Bringing it home: A study of the incentives surrounding the repatriation of foreign earnings under the American Jobs Creation Act of 2004.” Journal of Accounting Research 47, no. 4 (2009): 1027-1059. [5] Blouin, Jennifer, Linda Krull, and Leslie Robinson. ”Where in the world are “permanently reinvested” foreign earnings?.” University of Pennsylvania, University of Oregon and Dartmouth College working paper (2012). [6] Brown, Stephen J., and Jerold B. Warner. ”Measuring security price performance.” Journal of Financial Economics 8, no. 3 (1980): 205-258. [7] Campbell, John Y., and W. Andrew. ”Lo, and A. Craig MacKinlay, 1997, The econometrics of financial markets.” (1997). [8] Chevalier, Judith, and Glenn Ellison. ”Career concerns of mutual fund managers.” The Quarterly Journal of Economics 114, no. 2 (1999): 389-432. [9] Chung, Kee H., and Stephen W. Pruitt. ”A simple approximation of Tobin’s q.” Financial Management (1994): 70-74. [10] Dittmar, Amy, and Jan Mahrt-Smith. ”Corporate governance and the value of cash holdings.” Journal of Financial Economics 83, no. 3 (2007): 599-634. [11] Dittmar, Amy, Jan Mahrt-Smith, and Henri Servaes. ”International corporate governance and corporate cash holdings.” Journal of Financial and Quantitative Analysis 38, no. 01 (2003): 111-133. [12] Doukas, John. ”Overinvestment, Tobin’s q and gains from foreign acquisitions.” Journal of Banking & Finance 19, no. 7 (1995): 1285-1303. 36

[13] Drucker, Jesse. ”Dodging repatriation tax lets US companies bring home cash.” Bloomberg Businessweek 28 (2010). [14] Edwards, Alexander, Todd Kravet, and Ryan Wilson. ”Permanently reinvested earnings and the profitability of foreign cash acquisitions.” Rotman School of Management Working Paper 1983292 (2012). [15] Fama, Eugene F., and Kenneth R. French. ”Common risk factors in the returns on stocks and bonds.” Journal of Financial Economics 33, no. 1 (1993): 3-56. [16] Fama, Eugene F., and Kenneth R. French. ”Taxes, financing decisions, and firm value.” The Journal of Finance 53, no. 3 (1998): 819-843. [17] Faulkender, Michael, and Mitchell Petersen. ”Investment and capital constraints: repatriations under the American Jobs Creation Act.” Review of Financial Studies 25, no. 11 (2012): 3351-3388. [18] Faulkender, Michael, and Rong Wang. ”Corporate financial policy and the value of cash.” The Journal of Finance 61, no. 4 (2006): 1957-1990. [19] French, Kenneth R. Data Library. [20] Fritz Foley, C., Jay C. Hartzell, Sheridan Titman, and Garry Twite. ”Why do firms hold so much cash? A tax-based explanation.” Journal of Financial Economics 86, no. 3 (2007): 579-607. [21] Graham, John R. ”Debt and the marginal tax rate.” Journal of Financial Economics 41, no. 1 (1996): 41-73. [22] Graham, John R., Michelle Hanlon, and Terry Shevlin. ”Real effects of accounting rules: Evidence from multinational firms’ investment location and profit repatriation decisions.” Journal of Accounting Research 49, no. 1 (2011): 137-185. [23] Harford, Jarrad, Sattar A. Mansi, and William F. Maxwell. ”Corporate governance and firm cash holdings in the US.” Journal of Financial Economics 87, no. 3 (2008): 535-555. [24] Jaffe, Jeffrey F. ”The effect of regulation changes on insider trading.” The Bell Journal of Economics and Management Science (1974): 93-121.

37

[25] Jensen, Michael C. ”Agency costs of free cash flow, corporate finance, and takeovers.” The American Economic Eeview (1986): 323-329. [26] Jensen, Michael C., and William H. Meckling. ”Theory of the firm: Managerial behavior, agency costs and ownership structure.” Journal of Financial Economics 3, no. 4 (1976): 305-360. [27] Kolari, James W., and Seppo Pynn¨onen. ”Event study testing with cross-sectional correlation of abnormal returns.” Review of Financial Studies 23, no. 11 (2010): 3996-4025. [28] Liu, Yixin, and David C. Mauer. ”Corporate cash holdings and CEO compensation incentives.” Journal of Financial Economics 102, no. 1 (2011): 183-198. [29] Lindenberg, Eric B., and Stephen A. Ross. ”Tobin’s q ratio and industrial organization.” The Journal of Business 54, no. 1 (1981): 1-32. [30] David McLean, R. ”Share issuance and cash savings.” Journal of Financial Economics 99, no. 3 (2011): 693-715. [31] Miller, Merton H., and Daniel Orr. ”A Model of the Demand for Money by Firms.” The Quarterly Journal of Economics 80, no. 3 (1966): 413-435. [32] Modigliani, Franco, and Merton H. Miller. ”The cost of capital, corporation finance and the theory of investment.” The American Economic Review (1958): 261-297. [33] Mulligan, Casey B. ”Scale economies, the value of time, and the demand for money: Longitudinal evidence from firms.” Journal of Political Economy 105, no. 5 (1997): 1061-1079. [34] Myers, Stewart C., and Nicholas S. Majluf. ”Corporate financing and investment decisions when firms have information that investors do not have.” Journal of Financial Economics 13, no. 2 (1984): 187-221. [35] Opler, Tim, Lee Pinkowitz, Ren´e Stulz, and Rohan Williamson. ”The determinants and implications of corporate cash holdings.” Journal of Financial Economics 52, no. 1 (1999): 3-46. [36] Pinkowitz, Lee, and Rohan Williamson. ”Bank Power and Cash Holdings: Evidence from Japan.” Review of Financial Studies 14, no. 4 (2001). [37] Pinkowitz, Lee, and Rohan Williamson. ”What is the market value of a dollar of corporate cash?.” Journal of Applied Corporate Finance 19, no. 3 (2007): 74-81. 38

[38] Stein, Jeremy C. ”Internal capital markets and the competition for corporate resources.” The Journal of Finance 52, no. 1 (1997): 111-133. [39] Thakor, Anjan V. ”Investment” myopia” and the internal organization of capital allocation decisions.” Journal of Law, Economics, & Organization (1990): 129-154. [40] Brainard, William C., and James Tobin. ”Pitfalls in financial model building.” The American Economic Review (1968): 99-122. [41] Tobin, James, and William C. Brainard. ”Asset markets and the cost of capital.” Economic Progress, Private Values, and Public Economic Journal 96 (1977): 120-138. [42] White, Halbert. ”A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity.” Econometrica: Journal of the Econometric Society (1980): 817-838.

39

Table 1: Summary Statistics of Data used in Regression (1) This table provides summary statistics of data used in regression (1), which is used to estimate the amounts of foreign and domestic cash. Mean, standard deviation, and quartile values are provided for each variable. Indicated variables are scaled by total assets, TA, in order to normalize for firm size. NA is net assets, defined as total assets less cash. Cash/N A is cash and short-term investments as a percentage of net assets. R&D/TA is the amount of research and development expenses scaled by total assets for each firm, and is set to zero if R&D expenditures is missing in COMPUSTAT. CAPEX/TA is capital expenditures as a percentage of total assets for each firm. Lev is defined as the ratio of total debt to the sum of total debt and the market value of equity. CF Std Dev is defined as the standard deviation of the firm’s earnings before interest, taxes, and depreciation as a percentage of total assets. Div Pmt is a dummy variable that takes a value of 1 if the firm paid a dividend in the year, and a value of 0 otherwise. ln(T A) is the natural log of total assets, in millions of dollars. BV E/M V E is the ratio of the reported book value of the firm’s equity, divided by the market value of the firm’s equity, which is defined as defined as the firm’s year-end closing stock price multiplied by the number of shares outstanding. Both Domestic Income and Foreign Income are pre-tax earnings, and are included as a percentage of total assets. The overall Tax Burden variable reflects the repatriation tax faced by firms to repatriate their foreign earnings back to the U.S. It is defined as a percentage of total assets of the firm. All variables except for Div Pmt and ln(T A) have been winsorized at the 1% level. Variable Cash/N A R&D/T A CAP EX/T A Lev CF Std Dev Div Pmt ln(T A) BV E/M V E Domestic Income/T A F oreign Income/T A Tax Burden

Mean 0.1211 0.0156 0.0732 0.2913 0.0335 0.0479 7.7663 0.6480 0.0809 0.0370 0.0013

SD 0.2747 0.0346 0.0545 0.2463 0.0343 0.4336 1.9963 0.4711 0.0776 0.0357 0.0033

40

p25 0.0068 0.0000 0.0315 0.0799 0.0066 0.0000 6.0599 0.3146 0.0204 0.0072 0.0000

Median 0.0257 0.0000 0.0587 0.2356 0.0250 1.0000 7.3728 0.5338 0.0505 0.0352 0.0000

p75 0.1096 0.0130 0.1084 0.4621 0.0479 1.0000 9.3955 0.8808 0.1331 0.0609 0.0002

Table 2: Summary Statistics of Data used in Regression (8) This table provides summary statistics of data used in regression (8). Mean, standard deviation, and quartile values are provided for each variable. Excess Stock Return is defined as the stock’s return over a given year minus the stock’s benchmark portfolio over the same year. With the exception of Mkt Leverage, all independent variables are standardized for size by the lagged market value of equity. 4X represents 1-year change: Xt − Xt−1 . Cash is defined as cash holdings plus marketable securities. Earnings is earnings before interest and extraordinary items. NA is net assets, defined as total assets excluding cash. R&D represents research and development expenditures for the firm. Dividends are total dividends, and Interest is interest expense, as reported by the firm. Mkt Leverage is the ratio of total debt to the sum of total debt and the market value of equity. NF is net financing, defined as total equity issuance minus repurchases plus debt issuance minus debt redemption. F oreign Casht−1 is the lagged estimate of foreign cash. Domestic Casht−1 is the lagged estimate of domestic cash. High For Ops is a dummy variable which takes a value of 1 if a firm-year is in the top half of the sample reporting foreign income in terms of amount of foreign operations as a percentage of total operations, and a value of 0 otherwise. Good Inv Opp is a dummy variable which takes a value of 1 if a firm-year is in the top half of the sample in terms of Tobin’s q (q-ratio). Poor Inv Opp is a dummy variable which takes a value of 1 if a firm-year is in the bottom half of the sample in terms of q-ratio, and 0 otherwise. The q-ratio is defined as the market value of equity (defined as the product of its share price and the number of common shares outstanding) plus preferred stock (the liquidating value of the firm’s outstanding preferred stock) plus debt (the value of short-term liabilities net of short-term assets, plus the book value of long term debt), all divided by the book value of the total assets of the firm. Foreign Operations is defined as F oreign Income/T A. The q-ratio and Foreign Operations are used for the calculation of the dummy variables, but are not directly included in the regression. All variables have been winsorized at the 1% level.

Variable Excess Stock Return (4Cash)t (4Earnings)t (4N A)t (4R&D)t (4Interest)t (4Dividends)t (Cash)t−1 (M kt Leverage)t (N F )t (F oreign Cash)t−1 (Domestic Cash)t−1 (High F or Ops)t (Good Inv Opp)t (P oor Inv Opp)t (q-ratio)t (F oreign Operations)t

Mean -0.0063 0.0089 0.0064 0.0232 0.0009 0.0000 0.0004 0.1177 0.2403 0.0154 0.0061 0.1128 0.4471 0.5592 0.4408 1.2331 0.3676

SD 0.4269 0.0970 0.1427 0.3750 0.0121 0.0209 0.0068 0.1490 0.2092 0.1685 0.0183 0.1503 0.4972 0.4965 0.4965 1.1426 0.3574

41

p25 -0.2641 -0.0171 -0.0237 -0.0541 -0.0001 -0.0029 -0.0002 0.0257 0.0738 -0.0370 0.0000 0.0219 0.0000 0.0000 0.0000 0.5215 0.0475

Median -0.0480 0.0015 0.0056 0.0205 0.0000 -0.0000 0.0000 0.0649 0.1905 -0.0017 0.0000 0.0601 0.0000 1.0000 0.0000 0.8813 0.2504

p75 0.1862 0.0267 0.0319 0.1079 0.0024 0.0031 0.0009 0.1464 0.3564 0.0380 0.0003 0.1394 1.0000 1.0000 1.0000 1.5105 0.6254

Table 3: Results of Regression (1): Estimating the Amounts of Foreign and Domestic Cash This table presents the results of regression (1), which is used to calculate foreign and domestic cash levels. The dependent variable is Cash/N A. NA is net assets, defined as total assets less cash. Indicated variables are scaled by total assets, TA, in order to normalize for firm size. R&D/TA is research and development expenses scaled by total assets for each firm, and is set to zero if R&D expenditures is missing in COMPUSTAT. CAPEX/TA is capital expenditures as a percentage of total assets for each firm. Lev is defined as the ratio of total debt to the sum of total debt and the market value of equity. CF Std Dev is defined as the standard deviation of the firm’s earnings before interest, taxes, and depreciation as a percentage of total assets. Div Pmt is a dummy variable that takes a value of 1 if the firm paid a dividend in the year, and a value of 0 otherwise. ln(T A) is the natural log of total assets, in millions of dollars. BV E/M V E is the ratio of the book value of the firm’s equity to the market value of the firm’s equity, which is defined as defined as the firm’s year-end closing stock price multiplied by the number of shares outstanding. Both Domestic Income and Foreign Income are pre-tax earnings, and are included as a percentage of total assets. The overall Tax Burden variable reflects the repatriation tax faced by firms to repatriate their foreign earnings back to the U.S. It is defined as a percentage of total assets of the firm. All variables are winsorized at the 1% level, with the exception of Div Pmt and ln(T A). Specification includes year and industry fixed effects. Robust standard errors are in parentheses. Standard errors are corrected for correlation across observations of a given firm (White (1980)). *, **, and *** indicate significance at the 10%, 5%, and 1% level, respectively.

Dependent Variable: Cash/N A Independent Variables R&D/T A

Coefficients 1.780*** (0.156) -0.720*** (0.083) -0.282*** (0.025) 1.363*** (0.190) -0.080*** (0.010) -0.017*** (0.004) -0.021** (0.010) 0.227*** (0.082) -0.216* (0.121) 5.470*** (1.429) 0.485*** (0.067)

CAP EX/T A Lev CF Std Dev Div Pmt ln(T A) BV E/M V E Domestic Income/T A F oreign Income/T A Tax Burden Intercept

Observations Number of Firms Adjusted R2

15,660 2,059 0.33

42

Table 4: Results of Regression (5): Market Value of Foreign and Domestic Cash This table presents the results of regression (5). The dependent variable is Excess Stock Return, defined as the return of a stock’s benchmark portfolio over a given year subtracted from the return of that stock during the same year: Rbi,t is the benchmark portfolio return of a stock over a given year, and Ri,t is the return of that same stock over the year. All variables except for Lev and Excess Stock Return are scaled by the lagged market value of equity (M V Ei,t−1 ). ∆X represents 1-year change: Xt − Xt−1 . Cash is defined as cash holdings plus marketable securities. Earnings is earnings before interest and extraordinary items. NA represents net assets, and is defined as total assets excluding cash. R&D are research and development expenditures for the firm. Div is cash dividends, and Int is interest expense, as reported by the firm. Lev is the ratio of total debt to the sum of total debt and the market value of equity. NF is net financing, defined as total equity issuance minus repurchases plus debt issuance minus debt redemption. (F oreign Cash)t−1 is the lagged estimate of foreign cash. (Domestic Cash)t−1 is the lagged estimate of domestic cash. Robust standard errors are in parentheses. Standard errors are corrected for correlation across observations of a given firm (White (1980)). *, **, and *** indicate significance at the 10%, 5%, and 1% level, respectively.

Dependent Variable: Excess Stock Return (Ri,t − Rbi,t ) Independent Variables Coefficients (4Cash)t 1.361*** (0.138) (4Earnings)t 0.614*** (0.044) (4N A)t 0.133*** (0.022) (4R&D)t 0.369 (0.494) (4Int)t -2.079*** (0.340) (4Div)t 2.049*** (0.737) (Cash)t−1 0.341*** (0.040) (Lev)t -0.397*** (0.024) (N F )t -0.011 (0.045) (Lev)t × (4Cash)t -1.316*** (0.303) (4Cash)t × (F oreign Cash)t−1 3.832 (6.259) (4Cash)t × (Domestic Cash)t−1 -0.933*** (0.364) Intercept 0.031*** (0.007) Observations Number of Firms Adjusted R2

10,899 1,643 0.17

43

Table 5: Results of Regression (8): Investment Opportunities and the Values of Cash This table presents the results of regression (8). The dependent variable is Excess Stock Return, defined as the return of a stock’s benchmark portfolio over a given year subtracted from the return of that stock during the same year: Rbi,t is the benchmark portfolio return of a stock over a given year, and Ri,t is the return of that stock over the same year. All variables except for Lev and Excess Stock Return are scaled by the lagged market value of equity (M V Ei,t−1 ). ∆X represents 1-year change: Xt − Xt−1 . Cash is defined as cash holdings plus marketable securities. Earnings is earnings before interest and extraordinary items. NA is net assets, defined as total assets excluding cash. R&D represents research and development expenditures for the firm. Div is cash dividends, and Int is interest expense, as reported by the firm. Lev is the ratio of total debt to the sum of total debt and the market value of equity. NF is net financing, defined as total equity issuance minus repurchases plus debt issuance minus debt redemption. (F oreign Cash)t−1 is the lagged estimate of foreign cash. (Domestic Cash)t−1 is the lagged estimate of domestic cash. High For Ops is a dummy variable which takes a value of 1 if a firm-year is in the top half of the sample in terms of amount of foreign operations as a percentage of total operations, and a value of 0 otherwise. Good Inv Opp is a dummy variable which takes a value of 1 if a firm-year is in the top half of the sample in terms of Tobin’s q (q-ratio). Poor Inv Opp is a dummy variable which takes a value of 1 if a firm-year is in the bottom half of the sample in terms of q-ratio, and 0 otherwise. Robust standard errors are in parentheses. Standard errors are corrected for correlation across observations of a given firm (White (1980)). *, **, and *** indicate significance at the 10%, 5%, and 1% level, respectively.

Dependent Variable: Excess Stock Return (Ri,t − Rbi,t ) Independent Variables Coefficients (4Cash)t

1.252*** (0.157) 0.612*** (0.044) 0.131*** (0.021) 0.341 (0.494) -2.067*** (0.340) 1.990*** (0.732) 0.343*** (0.039) -0.397*** (0.024) -0.010 (0.045) -1.066*** (0.333) 2.830 (6.221) -0.885** (0.366) 0.405** (0.178) -0.094 (0.139) 0.031 (0.007)

(4Earnings)t (4N A)t (4R&D)t (4Int)t (4Div)t (Cash)t−1 (Lev)t (N F )t (Lev)t × (4Cash)t (4Cash)t × (F oreign Cash)t−1 (4Cash)t × (Domestic Cash)t−1 ∆(Cash)t × (High F or Ops)t × (Good Inv Opp)t ∆(Cash)t × (High F or Ops)t × (P oor Inv Opp)t Intercept Observations Number of Firms Adjusted R2

44

10,893 1,642 0.17

Table 6: Summary Statistics of Alternative Measures of Foreign and Domestic Cash, and the Effective Repatriation Tax Rate This table provides summary statistics of data defined in (15), (16), and (17). Mean, standard deviation, and quartile values are provided for each variable. Eff Repat Tax is defined    Income − (F oreignT axesP aid)i,t , 0 / (T A)i,t , and is a proxy as max (U ST axRate)i,t × F oreign TA i,t

0

for a firm’s effective repatriation tax rate. (F oreign Cash)t−1 is the lagged estimate of foreign cash defined using coefficient estimates from the Foley et al (2007) regression results. (DomesticCash)t−1 is the lagged estimate of domestic cash, defined as total cash less the 0 0 estimate of foreign cash. (F oreign Cash)t−1 and (Domestic Cash)t−1 have been standardized for size by the lagged market value of equity. All variables have been winsorized at the 1% level. Variable (Eff Repat T ax)t 0 (F oreign Cash)t−1 0 (DomesticCash)t−1

Mean 0.0117 0.0063 0.1116

SD 0.0420 0.0084 0.1577

45

p25 0.0000 0.0012 0.0196

Median 0.0000 0.0038 0.0571

p75 0.0000 0.0075 0.1365

Table 7: Results of Regressions (5) and (8) Using Alternate Measures of Cash Column (1) includes results for regression (5), and column (2) presents the results of regression (8) using the alternate measures of foreign and domestic cash defined by (16) and (17).The dependent variable is Excess Stock Return, defined as the return of a stock’s benchmark portfolio over a given year subtracted from the return of that stock during the same year: Rbi,t is the benchmark portfolio return of a stock over a given year, and Ri,t is the return of that stock over the same year. All variables except for Lev and Excess Stock Return are scaled by the lagged market value of equity (M V Ei,t−1 ). ∆X represents 1-year change: Xt − Xt−1 . Cash is defined as cash holdings plus marketable securities. Earnings is earnings before interest and extraordinary items. NA is net assets, defined as total assets excluding cash. R&D represents research and development expenditures for the firm. Div is cash dividends, and Int is interest expense, as reported by the firm. Lev is the ratio of total debt to the sum of total debt and the market value of equity. NF is net financing, defined as total equity issuance minus repurchases 0 0 plus debt issuance minus debt redemption. (F oreign Cash)t−1 is the lagged estimate of foreign cash. (Domestic Cash)t−1 is the lagged estimate of domestic cash. High For Ops is a dummy variable which takes a value of 1 if a firm-year is in the top half of the sample in terms of amount of foreign operations as a percentage of total operations, and a value of 0 otherwise. Good Inv Opp is a dummy variable which takes a value of 1 if a firm-year is in the top half of the sample in terms of Tobin’s q (q-ratio). Poor Inv Opp is a dummy variable which takes a value of 1 if a firm-year is in the bottom half of the sample in terms of q-ratio, and 0 otherwise. Robust standard errors are in parentheses. Standard errors are corrected for correlation across observations of a given firm (White (1980)). *, **, and *** indicate significance at the 10%, 5%, and 1% level, respectively.

Dependent Variable: Excess Stock Return (Ri,t − Rbi,t ) Independent Variables (1) (4Cash)t 1.354*** (0.148) (4Earnings)t 0.610*** (0.044) (4N A)t 0.134*** (0.021) (4R&D)t 0.397 (0.494) (4Int)t -2.030*** (0.337) (4Div)t 2.038*** (0.743) (Cash)t−1 0.336*** (0.040) (Lev)t -0.400*** (0.024) (N F )t -0.016 (0.045) (Lev)t × (4Cash)t -1.274*** (0.303) 0 (4Cash)t × (F oreign Cash)t−1 9.929 (13.665) 0 (4Cash)t × (Domestic Cash)t−1 -0.874*** (0.288) ∆(Cash)t × (High F or Ops)t × (Good Inv Opp)t

0.032*** (0.007)

(2) 1.263*** (0.161) 0.607*** (0.044) 0.132*** (0.021) 0.371 (0.494) 2.019*** (0.337) 1.981*** (0.738) 0.338*** (0.040) -0.400*** (0.024) -0.015 (0.045) -1.029*** (0.332) 6.900 (14.179) -0.834*** (0.290) 0.369** (0.179) -0.109 (0.144) 0.032*** (0.007)

10,849 1,640 0.17

10,843 1,639 0.17

∆(Cash)t × (High F or Ops)t × (P oor Inv Opp)t Intercept

Observations Number of Firms Adjusted R2

46

Table 8: Abnormal Returns Surrounding Passage of the American Jobs Creation Act of 2004 This table displays abnormal returns and CARs for firms with good and poor investment opportunities abroad. Results are given separately using the market model and using the Fama and French (1993) factors. τ = 0 represents June 17th, when the Act won passage in the U.S. House of Representatives. The sample covers represents a total of 4,867 returns for 157 firms, comprised of 97 firms with good investment opportunities abroad, and 60 firms with poor investment opportunities abroad.

Event Time τ -15 -14 -13 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Market Good Foreign Invest. Opps. AR CAR 0.0004 0.0004 0.0013 0.0017 -0.0059 -0.0042 0.0027 -0.0015 0.0026 0.0012 -0.0092 -0.0080 -0.0050 -0.0130 0.0014 -0.0116 -0.0015 -0.0132 0.0023 -0.0109 -0.0010 -0.0119 0.0003 -0.0116 -0.0019 -0.0135 0.0048 -0.0087 0.0014 -0.0073 -0.0064 -0.0137 -0.0027 -0.0164 0.0019 -0.0144 0.0062 -0.0082 0.0035 -0.0047 -0.0021 -0.0068 0.0112 0.0044 -0.0031 0.0013 0.0056 0.0069 0.0009 0.0078 -0.0050 0.0027 -0.0044 -0.0017 -0.0118 -0.0134 -0.0012 -0.0146 0.0008 -0.0138 0.0020 -0.0117

Model Poor Foreign Inv. Opps. AR CAR 0.0039 0.0039 -0.0032 0.0006 -0.0005 0.0001 0.0041 0.0042 0.0072 0.0114 -0.0018 0.0096 -0.0047 0.0048 -0.0004 0.0044 -0.0009 0.0035 0.0006 0.0041 0.0091 0.0132 -0.0022 0.0110 -0.0009 0.0100 0.0027 0.0127 0.0023 0.0150 0.0063 0.0213 -0.0027 0.0186 0.0032 0.0219 0.0015 0.0234 0.0041 0.0275 0.0005 0.0280 0.0059 0.0338 0.0008 0.0347 0.0064 0.0411 0.0054 0.0465 -0.0001 0.0464 0.0046 0.0510 0.0041 0.0551 -0.0021 0.0530 -0.0033 0.0497 0.0007 0.0504 47

Fama-French Model Good Foreign Poor Foreign Inv. Opps. Inv. Opps. AR CAR AR CAR 0.0017 0.0017 0.0035 0.0035 0.0012 0.0030 -0.0034 0.0002 -0.0022 0.0008 0.0027 0.0029 0.0034 0.0042 0.0039 0.0068 -0.0022 0.0019 0.0040 0.0108 -0.0066 -0.0047 0.0009 0.0117 -0.0011 -0.0058 -0.0009 0.0108 0.0010 -0.0048 -0.0010 0.0097 0.0036 -0.0012 0.0001 0.0098 0.0036 0.0024 0.0020 0.0118 -0.0007 0.0017 0.0096 0.0214 0.0050 0.0068 -0.0012 0.0202 -0.0025 0.0043 0.0017 0.0219 0.0005 0.0048 -0.0019 0.0200 0.0005 0.0053 0.0013 0.0213 -0.0060 -0.0007 0.0056 0.0269 0.0002 -0.0005 -0.0019 0.0251 -0.0003 -0.0008 0.0014 0.0265 0.0056 0.0048 0.0011 0.0275 0.0032 0.0081 0.0022 0.0297 -0.0037 0.0044 -0.0003 0.0294 0.0009 0.0053 -0.0017 0.0277 -0.0043 0.0010 0.0013 0.0290 0.0021 0.0031 0.0040 0.0330 0.0026 0.0057 0.0049 0.0379 -0.0027 0.0030 -0.0012 0.0367 -0.0054 -0.0025 0.0018 0.0385 -0.0050 -0.0075 0.0036 0.0421 0.0024 -0.0052 -0.0005 0.0415 0.0065 0.0013 -0.0008 0.0408 0.0015 0.0028 0.0009 0.0417

Table 9: Results of Regression (24): Abnormal Returns from the American Jobs Creation Act This table presents the results of regression (24). The dependent variable, Rp,τ , is the equalweighted average of returns for portfolio p as of event time τ . Rm,τ is the market return on date τ , which is the return of the value-weighted CRSP market index. Rhml and Rsmb are the returns from the Fama and French (1993) value and size portfolios. Iτ is a dummy variable which takes a value of 1 if the date τ falls in the event window, which is τ = −15 to 15, and a value of 0 otherwise. Specifications (1) and (2) form the portfolio using only firms with poor investment opportunities Income abroad, defined as firms that are in the upper half of the sample in terms of F oreign , but are TA in the lower half of the sample in terms of q-ratio. Specifications (3) and (4) include only firms with good investment opportunities abroad, defined as firms that are in the upper half of the sample in Income terms of F oreign , and are also in the upper half of the sample in terms of q-ratio. Robust TA standard errors are in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% level, respectively.

Portfolio:

Iτ Rm,τ

(1) Poor Investment Opportunities Abroad 0.002*** (0.001) 1.284*** (0.042)

Rhml Rsmb Constant

0.0004 (0.0003)

Observations Adjusted R2

252 0.78

Dependent Variable: Rp,τ (2) Poor (3) Good Investment Investment Opportunities Opportunities Abroad Abroad 0.002*** -0.0002 (0.0006) (0.0004) 1.080*** 1.505*** (0.045) (0.049) 0.334*** (0.071) 0.586*** (0.050) 0.0003 -0.0002 (0.0003) (0.0004) 252 0.84

252 0.81

48

(4) Good Investment Opportunities Abroad 0.0001 (0.0007) 1.139*** (0.040) -0.272*** (0.063) 0.834*** (0.055) -0.0002 (0.0002) 252 0.90

Figure 1: Plot of CARs Surrounding Enactment of American Jobs Creation Act These graphs plot the average cumulative abnormal returns (CARs) for firms with good foreign investment opportunities (blue dashed line) and firms with poor foreign investment opportunities (solid red line). The top graph displays CARs calculated using the market model, while the bottom displays CARs calculated using the Fama-French factors.

.0 2 -.0 2

-.0 1

0

.0 1

CARs

.0 3

.0 4

.0 5

.0 6

CARs Calculated Using Market Model

-15 -14 -13 -12 -11 -10

-9

-8

-7

-6

-5

-4

-3

-2

-1

0

1

2

3

4

5

6

7

8

9

10

11

12 13

14

15

10

11

12 13

14

15

Event Time Good Foreign Inv. Opps.

Poor Foreign Inv. Opps.

.0 2 .0 1 0 -.0 1

CARs

.0 3

.0 4

CARs Calculated Using Fama-French Factors

-15 -14 -13 -12 -11 -10

-9

-8

-7

-6

-5

-4

-3

-2

-1

0

1

2

3

4

5

6

7

8

9

Event Time Good Foreign Inv. Opps.

Poor Foreign Inv. Opps.

49