Corporate Inversions: A Case of Having the Cake …...Corporate Inversions: A Case of Having the Cake and Eating it Too? Felipe Cortes, Armando Gomes and Radhakrishnan Gopalan June

Corporate Inversions: A Case of Having the Cake and Eating

it Too?

Felipe Cortes, Armando Gomes and Radhakrishnan Gopalan∗

June 15, 2015

Abstract

The race by American companies to change their incorporation to countries with lower

corporate tax rate has reached fever pitch. Using a comprehensive sample of U.S. com-

panies that reincorporate overseas (“invert”) between 1996 and 2013 and a matched

set of U.S. incorporated multinational firms, we study the benefits and costs of such

transactions. On the benefit side, we find that inverted firms have 7%-8% lower effective

tax rate, mainly on account of the lower marginal tax rate in their country of incorpo-

ration. On the cost side, we find that inverted firms have higher bid-ask spread, their

stock has less institutional ownership and investors put a lower value on the cash on

their balance sheet. We also find inverted firms to have more concentrated institutional

share ownership structure. Overall, our results highlight both the benefits and costs of

inversions.

Keywords: Corporate inversions, corporate governance, taxation, valuation.

∗Cortes is from the D’Amore-McKim School of Business, Northeastern University, Gomes and Gopalanare from the Olin Business School, Washington University in St. Louis. The authors can be reached [email protected], [email protected], and [email protected]. We thank Rebekah McCarty, Adam Rosen-zweig, and Elliot Staffin, and seminar participants at Financial Intermediation Research Society (FIRS)annual conference (2015) for helpful comments and Mariassunta Giannetti for valuable discussion.

Introduction

The race by American companies to change their incorporation to countries with lower

corporate tax rate has reached fever pitch, prompting threats of legislative action to halt

the barrage of corporate inversions. While the benefits of changing a firm’s tax domicile to

outside the U.S. – in terms of lowering the firm’s effective tax rate – are much talked about,

the costs are not. The controversy regarding Walgreen’s recent attempt to reincorporate to

Switzerland, a country with civil law legal origin illustrates this point. While a number of

activist hedge fund shareholders including Jana Partners LLC were attracted to the lower

taxes, CtW Investment Group, another Walgreens investor, opposed the move based on

concerns that it would weaken the company’s corporate governance.1 In this paper, we

collect a large and comprehensive sample of foreign incorporated publicly listed American

companies to understand the benefits and costs of inversions.

Our working definition of an inversion is of an American company that changes its

country of incorporation to outside the U.S. To collect a comprehensive sample of such

inversions, we start with a sample of all foreign incorporated firms listed in an U.S. exchange.

This sample includes both inversions and cross-listed foreign firms. To distinguish between

the two, we first follow the Securities and Exchange Commission (SEC)’s definition of an

“American issuer”.2 SEC defines a foreign incorporated firm as “essentially an U.S. issuer”

if more than 50% of the outstanding voting securities are held by U.S. residents and the

firm has significant business in the U.S., meaning that either more than 50% of its assets

are located in the U.S., or a majority of its executive officers or directors are U.S. citizens

or residents, or its businesses are managed principally from within the U.S. Our analysis

indicates that U.S. firms that invert themselves continue to be classified as an American

issuer by the SEC and enjoy the full benefits of the U.S. securities laws. On the other hand,

most cross-listed firms are classified as a foreign private issuer by the SEC and satisfy less

stringent disclosure and corporate governance rules (???), and they often choose to opt-out

of exchange governance provisions (?).

1See New York Times articled titled “Walgreen Shareholder Opposes Potential Deal to ReincorporateAbroad” dated May 13, 2014 and Wall Street Journal articled titled “Walgreen Weighs Riding Tax-InversionWave”, dated July 14, 2014.

2See Rule 405 of Regulation C under the Securities Act of 1933 and Rule 3b-4 under the Exchange Actof 1934. See also ?.

1

Even within the sample of foreign incorporated “American issuers”, we may have some

cross-listed foreign firms, especially if their primary listing is in the U.S. To exclude these

firms from our analysis, we group the foreign incorporated American issuers into those that

were foreign incorporated from the date of their IPO and those that changed the country

of incorporation sometime after their IPO. We refer to the former as American Foreign

Companies (AFCs) and the latter as inversions.3 In much of the paper, we compare our

sample of inversions to a control sample of U.S. multinationals to conduct our tests.

It is noteworthy that the inverted firms in our sample receive the same treatment under

the Federal Securities Laws and have identical disclosure requirements as U.S. incorporated

firms. Specifically, they are required to file financial statements conforming to U.S. GAAP

denominated in U.S. dollars, must file 10-Ks, 10-Qs, 8-Ks and proxy statements, and must

comply with Regulation FD. Moreover, their executive officers are subject to the same

personal liability as U.S. firms (see ?). Furthermore they cannot exempt themselves from

the corporate governance requirements of U.S. stock exchanges. Thus these firms avoid a

potentially big cost of inverting, namely of not being subject to U.S. securities laws.

There are two important differences between the inversions in our sample and U.S. firms:

the tax rate applicable to the company’s domestic and foreign profits and the applicable

corporate law. For example, the marginal corporate tax rate for firms incorporated in the

Cayman Islands is 0% versus 40% for firms incorporated in the U.S. (federal and state taxes)!

Moreover, inverted firms are subject to the corporate law of their country of incorporation,

while U.S. firms are subject to the corporate law in their state of incorporation, which is

Delaware for many U.S. firms. Our paper is an attempt to understand if inverted firms can

benefit from lower corporate taxation, and to the extent they continue to conform to U.S.

securities laws, be valued and treated like a U.S. firm, effectively have the cake and eat it

too.

We obtain data from four main sources: S&P Capital IQ, Compustat, CRSP and the

SEC. We identify the sample of inversions and AFCs from Capital IQ for the time period

1996-2013. We identify 75 inversions and 186 AFCs. We classify the inversions in our

sample into two subgroups based on how they inverted: Pure inversions and Restructuring

3Note that an alternative method to identify a sample of inversions is by searching through news reports.Our research indicates this to be an imperfect way to identify a comprehensive sample.

2

inversions.

We have 25 Pure inversions and 50 Restructuring inversions in our sample. A Pure

inversion does not involve any change in either a company’s operations or in the identity of

its shareholders. Legally, the American operations of the firm are organized as a subsidiary

of the new foreign parent. There are currently 8 members of the S&P 500 index that have

undergone Pure inversions (see Table 1). In a Restructuring inversion there are material

changes in either the company’s ownership, business, or assets, and they take place as a

result of either a merger, LBO, spin-off or bankruptcy transaction.

We have 186 AFCs including U.S. household names such as Michael Kors Holdings

Ltd., Garmin Ltd., and Carnival Corporation that began operating in the U.S. as a foreign

incorporated entity.4 Note that our sample of AFCs include both U.S. firms that were

foreign incorporated since inception and foreign firms, such as Schlumberger Ltd. and

LyondellBasell Industries N.V. (both S&P 500 members), that arguably have primarily

foreign origins, but over time increased either their business presence or their shareholder

base to become “American”. Given the difficulty (and ambiguity) in distinguishing between

these two sub-groups, we group them together as AFCs. The large number of AFCs in our

sample highlights the importance of taking into account the effect of tax policy on the

incentive of new firms, that want to benefit from U.S. securities laws and list in the U.S.,

to incorporate in the U.S.5

Our univariate comparison of the two groups of inversions shows that firms that un-

dergo pure inversions are larger, have a more liquid stock and slightly higher institutional

ownership than firms that undergo a restructuring inversion. The former also have lower

market to book ratio and are more likely to be rated. The effective tax rate of both sets of

firms is significantly lower than the U.S. marginal corporate tax rate of 40% (federal and

state taxes). We also find that as compared to the median U.S. multinational, our sample

of inversions are larger, have lower market to book ratio, higher profitability and spend less

4Allan Sloan refers to these firms as “never-heres”, in the Fortune Magazine article, “Positively un-American tax dodges”, dated July 7, 2014.

5See also ? which contains the following quote attributed to Intel’s vice-president Robert Perlman inhis testimony to Congress in 1999: “If I had known at Intel’s founding (over thirty years ago) what I knowtoday about international tax rules, I would have advised the parent company be established outside theU.S. This reflects the reality that our Tax Code competitively disadvantages multinationals simply becausethe parent is a U.S. corporation.” Since then the disadvantage between the tax policy of the U.S. and othercountries has only grown.

3

on R&D and advertising.

To highlight the benefits and costs of inverting, our main empirical analysis compares

inversions to a set of U.S. incorporated multinational control firms identified by matching

on observable firm characteristics. Since inverted firms typically have operations in multiple

countries, we confine our control sample to multinationals – firms that report positive sales

in at least one foreign subsidiary.6 For our baseline control sample, we match on industry

– identified at the six-digit Global Industry Classification Standard (GICS) code level–

financial (fiscal) year, Log(Total assets), Market to book and Return on asset (ROA). All

variables we use are defined in Appendix 2. Specifically, for every inverted firm-year in

our sample, we identify up to two unique control firm-year observations involving an U.S.

multinational firm from the same industry GICS code and financial year as the inverted

firm-year and that is closest in terms of the other three covariates. We use the Mahalanobis

distance to identify the closest match. Our matching procedure is effective as the two

samples are indistinguishable in terms of the distribution of the matching variables.

In our multivariate tests we estimate an OLS model within the sample of inverted firms

and control firms after including a number of linear controls including industry and time

fixed effects. Our standard errors are robust to heteroskedasticity and are clustered at the

firm level. We employ a matching method followed by a multivariate regression to ensure

that we adequately control for observable co-variates. We also repeat our estimates using

alternate control samples identified by matching on additional co-variates specific to an

outcome variable. These tests help us estimate bias-corrected Average Treatment Effect on

the Treated (ATT) (?).

Note that while our matching procedure ensures that the inversions and control firms

look similar on observable dimensions, unobserved differences between the two could bias

our estimates. In reality, given that some firms choose to invert, we do expect there to

be unobserved differences between inversions and control firms. In Section 5.3 we perform

tests that will help us evaluate the extent of unobserved heterogenity required to overturn

our estimates. Having said that, to interpret our estimates as the effect of investions on the

outcome variable, the main assumption required is that conditional on the matching covari-

ates employed, the outcome variable in the control group is independent of the treatment,

6We obtain information on sales by foreign subsidiaries from Compustat Segment Data.

4

(see ?).

We begin our empirical analysis by comparing the effective tax rates of inversions and

control firms. Our two measures of tax rates are the effective tax rate as defined under

U.S. GAAP, which we call GAAP ETR, and the firm’s cash taxes paid divided by pre-tax

accounting income, adjusted for special items, Cash ETR (see Appendix 2 for all variable

definitions) (? and ?). Not surprisingly inverted firms have 6.8% lower GAAP ETR and

7.2% lower Cash ETR as compared to control firms. We find that the lower effective tax

rate of inversions is mainly on account of the lower marginal tax rate in their country of

incorporation.

It is often argued that firms may invert so as to be able to distribute internal cash

to shareholders without a tax penalty (?). This would predict lower cash holding among

inversions as compared to the control firms. When we compare the level of cash holding, to

our surprise, we find that inverted firms have 26.49% higher cash balance as a proportion

of total assets as compared to the median control firm. We find this result to be robust

to alternate control samples and is not explained by the marginal tax rate of the inverted

firm’s country of incorporation.

We find evidence consistent with inverted firms having less liquid stock: they have a

higher bid-ask spread, lower turnover and greater dispersion in analyst earnings forecast

as compared to control firms. We also find that the share of institutional ownership is

lower for the inverted firms. As compared to a median firm in our sample which has 73.8%

institutional ownership, inversions have 5.7% lower institutional ownership. This indicates

that although firms follow U.S. securities laws after inversion, investors perceive an extra

element of opaqueness about these firms and are reluctant to hold and trade their shares.

Finally, we investigate if investors value inverted firms differently. To do this, we use the

methodology in ? and compare the value of a dollar of cash with inverted firms and U.S.

firms. To ensure adequate power for these tests, we include all U.S. firms with financial

data in CRSP/Compustat as the control sample.7 Our dependent variable is the abnormal

stock return and we relate it to changes in the amount of cash. We find that on average,

investors put a lower value on cash on the inverted firm’s balance sheet. We also find some

7Since our dependent variable in these tests is abnormal stock return, these tests do not suffer from thesame identification issues as those with firm financial data as the outcome variable.

5

weak evidence consistent with this lower valuation being related to the rule of law in the

inverted firm’s country of incorporation. We employ the rule of law index obtained from

the Worldwide Governance Indicators by the Worldbank to establish this. Our results are

consistent with the evidence in ?.

For our sample countries, rule of law and marginal tax rates are significantly positively

correlated. That is, countries with better rule of law also tend to have higher marginal tax

rates. Since a higher marginal tax rate is likely to decrease the value of cash on the firm’s

balance sheet (due to double taxation of investment income), while better rule of law is

likely to increase the value of cash, we are constained in our ability to isolate the effect of

rule of law on value of cash.

Summarizing, our analysis highlights both the benefits and costs of inversions. On the

benefit size, we find that these firms have a lower effective tax rate. On the cost side, we

find that these firms have higher bid-ask spread, their stock has less institutional ownership

and investors put a lower value on the cash on their balance sheet. Given the documented

costs associated with inversions, firms less dependent on market liquidity and valuation

should invert. Consistent with this we find that inversions have more concerntrated insti-

tutional share ownership – shareholders less dependent on a liquid market to exit – and

access external capital less often. Thus, these firms may be relatively less affected by the

undervaluation in the market.

The rest of the paper is organized as follows. Section 1 discusses the related literature,

Section 2 our hypothesis, Section 3 our empirical methodology. Section 4 describes our data

and provides the summary statistics. Section 5 discusses the results of our empirical tests

while Section 6 concludes. Definitions of empirical variables are in Appendix 2.

1 Related Literature

Our paper is related to the literature on taxes, corporate law and cross-listing. A large

literature in public policy, law, and economics documents the effect of tax havens on firm

behavior. Using financial affiliate data, ? characterize the firms that incorporate subsidiaries

in tax havens. They show that larger, more international firms, and those with extensive

6

intrafirm trade and high R&D intensities are most likely to use tax havens. They further

find that firms with sizeable foreign operations benefit the most from using tax havens. In

comparison, we study instances when the ultimate parent – and not a subsidiary – is incor-

porated in a tax haven. To this extent, the inverted firms in our sample can pay dividends

to shareholders without having to repatriate the earnings to the U.S. and the corporate law

applicable to these firms is the law in the tax havens. ? do a cross-country comparison

of effective tax rates and show that there are significant cross-sectional differences in the

amount of taxes firms pay depending on their tax domicile. ? use financial accounting data

and show that firms that disclose material operations in at least one tax haven country have

a 1.5 percentage points lower tax rate than firms without such operations. In comparison,

our results indicate that inverted firms have on average 7% lower tax rate as compared to

U.S. firms.

For all the firms in our sample the tax-domicile is the same as the country of incorpora-

tion. Note that this is not true always as countries determine the corporation’s tax domicile

either based on the country where the parent is incorporated (e.g., U.S., Bermuda, Cayman,

Islands) or using “real seat” rules (e.g., most European countries). Under real seat rules the

corporation’s tax domicile is the country where the headquarters and significant business

operations/management are conducted from (see Kane and Rock (2008)).

Our paper is also related to the literature in finance that relates firms’ tax rates to their

financing policy. This literature is vast and we refer the reader to the review by ?. In

summary, the literature argues that firms that face lower taxes should, all else equal, have

a higher cash balance (?). Our results are consistent with this literature. Interestingly,

? argue that the high cash balance among U.S. firms may result from their tax planning

activities, i.e., accumulation of cash in overseas subsidiaries.

Our paper is also related to the legal literature that relates corporate law to firm value.

This research is mostly focused on comparing Delaware corporate law (the dominant one in

the U.S.) to that of other states. In an early paper, ? documents a higher value for firms

registered in Delaware as compared to firms registered elsewhere. ? offers some evidence

consistent with Delaware’s advantage narrowing over time. On the other hand, ? argue

that Delaware corporate law favors management over outside shareholders and claim that

its dominant role will overtime result in a progressive loosening of protection for outside

7

shareholders. In comparison, we focus on firms that are incorporated overseas and compare

them to U.S. firms, most of which are registered in Delaware. Thus, in terms of corporate

law, our comparison is between the law in countries outside the U.S. (most of which are tax

havens) to that in Delaware. We employ the World Bank’s Rule of Law index to do this

comparison.

Finally, our paper is also closely related to the large literature on cross-listed firms

(???). Despite apparent similarities, there are important difference between inversions and

cross-listed firms. Most cross-listed firms are classified as foreign private issuers (FPIs)

by the SEC and have weaker disclosure requirements than the inversions in our sample.

Specifically, FPIs need not provide quarterly financial statements, are exempt from provid-

ing detailed information on executive compensation, and insider trading filings. There are

also differences in terms of the applicability of Regulation FD (see Appendix 1 and Kinsey

(2001)). In comparison, the inverted firms in our sample provide the same level of disclo-

sure as U.S. firms. Thus, our comparison between inversions and U.S. firms helps identify

the effect of lower taxes and differences in corporate law on firm behavior. Furthermore,

unlike cross-listed firms, the inverted firms have a majority of shareholders and significant

operations in the U.S. This further enhances the SEC’s ability to enforce its penalties (???;

and Gagnon and Karolyi (2008)). Our tests that compare the marginal value of a dollar of

cash between inverted and U.S. firms is similar to those in ?.

There is also a small literature that studies the causes and consequences of inversions.

? study the determinants of Pure inversions and conclude that they are motivated by

firms trying to reduce U.S. tax on their sizable foreign income; but the savings associated

with the reduction of taxation of foreign income alone cannot account for the increase

in valuation of inverting companies. ? further explore this point and confirm that the

reduction in worldwide taxation post-inversion were also coming from companies reducing

their U.S. taxable income, mostly stripping earnings from their U.S. operations by shifting

expenses like interest expenses. Similar to these papers we also document significant tax

savings from inversions. On the flip side, we also document lower stock liquidity and market

valuation resulting from inversions. Furthermore, unlike these papers, our sample includes

restructuring inversions in addition to pure inversions.

8

2 Hypothesis

Corporations often operate worldwide through various subsidiaries and are subject to the

tax laws and tax rates prevailing in the countries in which they operate and earn income.

However, the country where the ultimate parent company is incorporated plays a pivotal

role in the effective tax rate on the corporation’s worldwide income. The majority of the

inversions in our sample have a parent company incorporated (and with a tax-domicile)

in a tax haven country (see Table 2). In addition, the inverted firms are subject to the

corporate law in their country of incorporation. U.S. firms on the other hand are subject

to the corporate law in their state of incorporation which is Delaware for most U.S. firms.

From the World Bank’s Rule of Law index we find that the countries where AFCs are

incorporated have on average weaker rule of law as compared to the U.S. If this weakness

also applies to their corporate law, then these countries are likely to offer weaker protection

to minority shareholders. In this section we outline the predictions that result from these

differences.

To the extent that the marginal tax rate of their country of incorporation are on average

lower for inversions as compared to for U.S. firms, we expect the inverted firms in our sample

to have a lower effective tax rate.8 We also expect the effective tax rate of the inverted

firms and U.S. firms to be positively related to the marginal tax rate of their country of

incorporation. This forms our first prediction:

Prediction 1: Inversions will have a lower effective tax rate than U.S. firms. The

effective tax rate of the inverted firms and U.S. firms will be positively related to the marginal

tax rate of their country of incorporation.

A firm’s tax rate and quality of rule of law can also affect its cash policy. Investment

income within the firm is taxed twice, once in the hands of the corporation and again in

the hands of the investors – when distributed as dividends. A high corporate tax rate

will increase the penalty from this double taxation and increase the cost of retaining cash

within the firm (?, ?). The lower corporate tax rate of the inverted firms will reduce this

penalty and hence imply that all else equal, inversions should have more cash. On the

8The effective tax rate may differ from the marginal statutory tax rate prevailing in parent country forvarious reasons, including differences in the accounting standards for book and tax income and also due togeographic dispersion of a multinational corporation subsidiaries across different tax regimes.

9

other hand ? argue that an important reason firms invert is to distribute internal cash to

their shareholders without incurring a tax penalty. This will predict lower cash holding

among inverted firms. Weaker protection of outside shareholders in inversions would imply

that investors may anticipate greater agency problems in such firms. Hence they would

want to leave less cash within the firm thereby limiting managerial investment flexibility.

Alternatively, the amount of cash within the firm may itself be a sign of an underlying

agency problem. This would predict that inverted firms will have higher cash balance than

control firms. Summarizing our second prediction reads:

Prediction 2: Inverted firms may have higher or lower cash holding as compared to

U.S. firms. The cash holding of inverted firms and U.S. firms will be negatively related to

their marginal tax rate.

The weaker potention for outside shareholders through corporate law would imply that

inverstors may perceive an additional layer of information and agency costs among inver-

sions. This in turn may translate into a reluctance on the part of investors to invest and

trade in these shares. This would predict that the shares of inversions will have less liquidity

and greater information asymmetry. ? and ? suggest that institutional demand for a stock

is positively related to the amount of public information available about the firm and the

strength of their corporate governance. The weaker protection through corporate law for

inversions would mean less institutional share ownership in these firms. This forms our next

prediction.

Prediction 3: Inverted firms will have lower stock liquidity, greater information asym-

metry and lower institutional investor ownership as compared to U.S. firms.

If investors perceive the inverted firms to have greater agency problems as compared to

U.S. firms then they are likely to put a lower value on the shares of such firms. Testing

this is difficult because of the inability to model the expected stock price. Hence, we focus

on a specific asset namely cash and compare the market value of corporate cash between

inversions and U.S. firms. The lower tax rate of the inverted firms will predict that investors

should put a higher value on the cash held by such as compared to the cash held by U.S.

firms. On the other hand, the greater agency costs among the inverted firms will predict

that investors should put a lower value on cash with such firms. Summarizing our final

10

prediction reads:

Prediction 4: Ceteris paribus, investors may have a higher or lower value on a dollar

of cash on an inverted firm’s balance sheet. The value of cash for inverted firms and U.S.

firms should be positively (negatively) related to the rule of law (marginal tax rate).

3 Empirical Methodology

3.1 Estimation

We test our predictions by comparing the inversions in our sample to a matched set of

U.S. incorporated multinational firms. Since we only identify matches for the treated (in-

versions) observations, our estimates should be interpreted as the average treatment effect

on the treated (ATT). The identification challenge we face is especially complex given the

myriad outcome variables we model. The relevant matching co-variates may vary based on

the outcome variable modeled. On the other hand, employing a different set of matching

covariates for each outcome variable will also mean that the sample for each test will vary.

We overcome this challenge by repeating our estimates using two sets of control samples.

The first is common across the outcome variables while the second is specific to each out-

come variable and employs additional matching covariates. We describe the construction

of the augmented control sample along with the discussion of the results of our tests.

We identify the common control sample by matching the inversions to U.S. multination-

als on industry – identified by the historic GICS industry (six-digit code level)– financial

year, Log(Total assets), Market to book and ROA. Specifically, for every inverted firm-year

in our sample, we identify up to two unique control firm-years involving U.S. multinational

firms from the same GICS industry and financial year as the inverted firm-year and that is

closest in terms of Log(Total assets), Market to book and ROA to the inverted firm-year. We

classify firms that report positive sales in at least one foreign subsidiary as a multinational

and use the Mahalanobis distance to identify the closest match. This constitutes our base

control sample and we refer to it as sample CS1. 9 Since we have a large pool of potential

control observations, we match without replacement to ensure greater power. This does not

9In unreported tests we find that Log(Total assets), Market to book and ROA are statistically correlatedwith whether the firm is an invertion or not. This ensures that these three matching covariates are relevant.

11

appear to be problematic as our treated and control samples are well matched in terms of

covariates.

To control for additional covariates specific to an outcome variable, we include them as

controls in the following regression model we estimate:

yit = β0 + β1 × Treatedit +Xitγ + αi + δt + εit (1)

where the outcome variable yit is one of GAAP ETR, Cash ETR, Log(Cash/Total assets),

Spread, Turnover, Analyst dispersion, Institutional Ownership, Log(HHI Institutional Own-

ership), and Average Institutional Ownership.. Treatedit is a binary indicator that takes

the value of 1 if the firm i at time t is an inverted firm, and it is zero otherwise. γ is a

vector of coefficients and Xit is a set of controls. The specific control variables we include

depend on the outcome variable being studied and includes lagged or contemporaneous

values of one or more of Log(Total assets), Market to book, ROA, Rated, Leverage, Capi-

tal expenditure, Volatility, Stock return, Marginal tax rate, Net operating loss, Advertising

expenditure, Capital expenditure, Foreign operations, Intangible assets, Gross PPE, SG&A,

and RD/Assets. In all our tests we include industry fixed effects (αi), time fixed effects

(δt), and report standard errors that are robust to heteroskedasticity and that are clustered

at the firm level.

Since our matching covariates are all continuous variables, matching may not be exact,

resulting in a conditional bias (?). We employ a regression model in addition to matching to

ensure that we control for discrepancies between the treated and control samples in terms

of covariates. When we estimate the above regression model on the treated and control

sample that we identify by matching on additional covariates, the controls in the regression

coincide with the set of matching covariates. In this case our estimate of β1 coincides with

the bias corrected ATT estimate proposed in ?. In further discussion, we refer to this as

“bias-corrected ATT”.

3.2 Identification Assumptions

We make two sets of assumptions to identify our effects. First, we assume that the treatment

value is stable across units (the SUTVA assumption, see ?). This assumption requires that

12

the treatment status of any unit does not affect the potential outcomes of the other units

and that the treatments for all units are comparable. In our setting, the classification

of one firm as an inversion is unlikely to affect the response of other firms to inverting.

Moreover, since our criteria for identifying our sample, non-U.S. incorporation status and

SEC classification is the same for all units, the treatment is comparable across units.

Second, conditional on the matching variables employed, the assignment between an

inverted firm and control firm status is independent of the outcome variable. This is the

conditional independence or unconfoundedness assumption. Since we only estimate ATT,

this assumption can be relaxed to require only that conditional on the matching covariates,

the outcome variable in the control group is independent of the treatment (see ?). While

there is no direct way to check the validity of this assumption, we perform two indirect tests

to assess the extent of bias due to unobserved heterogenity. First, we perform a placebo

test and estimate ATT where we should not find any. Specifically, we compare the set of

firms in the control sample CS1 to a set of random U.S. incorporated firms. We describe

the results of this test in Section 5.3. Second, we estimate and present the ? bounds that

help understand the extent of unobserved heterogeneity between the treated and control

observations required to overturn our conclusions. We estimate these bounds by repeating

our analysis using a control sample identified using caliper matching. We explain this as

well in Section 5.3.

4 Data

We obtain data from five main sources: S&P Capital IQ, Compustat, CRSP, Thomson

Reuters Ownership Database and the SEC. We identify the sample of inversions from Capi-

tal IQ using the methodology described below. We obtain financial data for these firms and

the control sample from the North-America annual Compustat database and stock price

information from CRSP. The variables we use in our analysis are defined in Appendix 2.

We obtain the corporate tax rates for the countries where our sample firms are incorpo-

rated in from the various sources including the KPMG’s Corporate Tax Rate Surveys, PwC

Worldwide Corporate Tax Summaries, and the Ernst & Young Worldwide Corporate Tax

Guides covering the 1996-2013 period. We use the statutory marginal corporate tax rate

13

of the highest tax bracket including taxes at the federal level and, for some countries, also

state/local taxes that are typically incurred by corporations in our sample. We also obtain

from the corporate tax sources above whether the corporate tax domicile is determined by

the place of incorporation or the main place of management (real seat). For companies

incorporated in countries that use real seat we use the information in the 10-K income tax

footnote to confirm that the country of incorporation is indeed the tax domicile/residence

of the parent company.

We obtain the rule of law index from the Worldwide Governance Indicators from the

Worldbank.10 This index captures the extent to which agents have confidence in and abide

by the rules of society, and in particular the quality of contract enforcement, property rights,

the police, and the courts, as well as the likelihood of crime and violence. The index ranges

from -2.5 (weak) to 2.5 (strong) governance performance. For ease of interpretation, we use

the percentile rank of a country in our tests.

4.1 Construction of the inversions sample

To identify the sample of inversions, we start with all publicly traded firms in the U.S. from

the CRSP/Compustat merged database and the S&P Capital IQ database whose country

of incorporation is not the U.S. during the 1996-2013 period (the foreign-incorporated list).

Our sample starts in 1996 because that is when the SEC EDGAR database starts, and we

use the filings from EDGAR in our data collection process. From this list, we weed out all

foreign private issues (FPIs) that are subject to less stringent disclosure requirements than

the inversions.

Our identification of FPIs takes several steps. First we use the annual list of FPIs that

the SEC publishes on its website for the period 1996-2013 (the FPI lists) and retain firms

in our sample if they are present in the foreign-incorporated list and not in the FPI lists.11

However, our subsequent analysis uncovered a number of inaccuracies in the FPI lists.

According to our conversations with SEC staff lawyers, the type of forms a company files

is the main information the SEC uses to prepare the annual FPI lists. For example, all

“American” issuers are required to file annual reports on forms 10-K, current reports on

10Available at http://info.worldbank.org/governance/wgi/index.aspx#home.11http://www.sec.gov/divisions/corpfin/internatl/companies.shtml

14

form 8-K, and proxy statements on forms DEF-14A. FPIs are not allowed to file proxy

statements on form DEF-14A, even if they voluntarily choose to file annual reports and

current reports on the forms designated for U.S. issuers.12 Companies are not required to

provide notice to the market or to the SEC of their FPI status (there are no specific forms

or notification requirements), even though many companies in our dataset choose to do so

voluntarily.

Hence to ensure that our inversions sample is complete and accurate, we append to our

initial sample all firms in the FPI lists and foreign-incorporated list that file both 10-Ks and

proxy statements (DEF-14A) at any point during their filing history in the SEC EDGAR

database. This results in a sample of 392 firms. We then individually analyze these firms

as follows:

For each firm-year, we obtain the country of incorporation from the first page of the

10-K statement. We perform a keyword search of the term “foreign private issuer” on all

the electronic filings of the firm in the SEC EDGAR database and read around the place

of occurrence of the term to make sure the company’s disclosure regarding its FPI status

agrees with our classification. In instances, when we encounter discrepancies, we overrule

the SEC’s classification only when there is strong evidence to do so. For example, if a

foreign company appears in an FPI list in a given year, but the company filed both forms

10-K and DEF-14A for that fiscal year, and the company explicitly mentions in a previous

filings that it no longer qualifies for FPI status then we include the firm in our sample for

that year (see Appendix 1 for more details).

This manual verification results in a sample of 261 firms that are either an inverted

firm or an AFC for at least one year during the 1996-2013 period. From this sample, we

go through firm’s proxy statements to identify how the firm reached their current status.

We classify the firms that were foreign incorporated from their date of IPO as an AFC

and the firms that changed their country of incorporation sometime after their IPO as an

inversion. Among the inversions in our sample, we have financial data before and after the

inversion for 10 firms. This sample is too small to provide any meaningful results on how

firm characteristics change before and after the inversion, especially if we include firm fixed

12See SEC no-action letter “Proxy Materials of Foreign Private Issuers” (March 10, 1992).

15

effects. Hence we do not include the before inversion period in our sample.13

4.2 How firms become an AFC or an inversion

A contribution of our paper is to provide a classification of foreign incorporated American

firms that uncovers a number of underlying trends in the migration of American firms

overseas. To do this we use a variety of data sources including SEC company filings,

company website, Google search engine, Capital IQ, and Factiva news articles. Based on

this analysis, we group the foreign incorporated American issuers in our sample into those

that are an AFC or reach the status through a corporate inversion transaction. Corporate

inversions can in-turn be either a Pure inversion or a Restructuring inversion– which are

further classified as either a merger, LBO, bankruptcy, or a spin-off inversion. The following

are the definition of these sub-groups:

• Pure inversion: A U.S. company that reincorporates in a new country with no material

change in its business and assets, and the same existing shareholders own the shares

in the new foreign parent company. We have 25 such firms in our sample.

• Restructuring inversion: Usually accompanied by a material change in either the

company’s ownership, business, or assets. It can result from one of the following

transactions (50 such firms in our sample):

– Merger inversion: The origin of the foreign corporation can be traced back to

a merger or acquisition transaction with a U.S. company. More than 50% of

the assets of the merged entity should originate from a former U.S. company, or

more than 50% of the shares of the new entity should be owned by former U.S.

shareholders (this criteria serves to exclude inbound-cross-border mergers deals).

– LBO inversion: A publicly traded U.S. firm or one of its corporate divisions is

taken private in a leverage buyout transaction, followed, a few years later, by an

IPO in which the company emerges with a foreign incorporation.

– Bankruptcy inversion: A new foreign incorporated company emerges with at

least 50% of its assets originating from a bankrupt U.S. corporation.

13The results of the tests we perform comparing the before and after period for this subset of firms isavailable upon request.

16

– Spin-off inversion: A U.S. or an AFC company spins-off a division as an inde-

pendent publicly traded foreign incorporated firm.

• AFC: A company with a significant business presence in the U.S. and primary listing

in the U.S. and is foreign incorporated from before the time of its IPO. These firms

may be started by U.S. entrepreneurs or have foreign origins. We have 186 such firms

in our sample.

We summarize our sample of inversions and AFCs in Tables 1 and 2. Table 1 provides

a listing of our sample categorized into different subgroups that are members of S&P 500

index as of December 31, 2013. We have 28 such firms with largest being Schlumberger

Ltd. with a market capitalization of over $118 billion.

Panel A of Table 2 provides the break-up of our sample into the three subgroups. The

largest subgroup of our sample are AFCs. We have 186 firms in this subgroup. Note that

for much of our analysis we exclude AFCs because of the difficulty in distinguishing the

foreign and domestic firms in the AFC subgroup. The next subgroup involves Restructur-

ing inversions (50 firms) mostly involving merger and LBO inversions. The Restructuring

inversions sample includes 25 merger inversions which have replaced Pure inversions as the

most popular way for American companies to invert since the American Jobs Creation Act

made Pure inversions ineffective in reducing taxes.14 Recent examples of merger inversions

include Actavis Plc., Eaton Corporation Plc., and Perrigo Plc., all S&P 500 firms, which

merged with, respectively, Warner-Chicott, Cooper Industries, and Elan, all Irish compa-

nies. The U.S. shareholders ended-up with between 70% to 80% of the new combined Irish

entity.15 We have a total of 6 LBO/Bankruptcy inversions and 10 spin-off inversions in

our Restructuring inversions sample. Prominent examples of LBO inversions include Avago

Technologies Ltd., a Singaporean company formed when the semiconductor division of Ag-

ilent Technologies was acquired by KKR and Silver Lake Partners, and Seagate Technology

Plc. (which first incorporated in the Cayman Islands, then later Ireland). In the Seagate

14Section 7874 of the Internal Revenue Code, which was part of the American Jobs Creation Act of 2004,established that if after an inversion transaction the percentage ownership by former shareholders of thecorporation is 80% or more, the new foreign parent will be treated as a U.S. corporation for tax purposes.

15The merger inversion frenzy is fueling an M&A boom. Since the beginning of 2014, 14 new mergerinversions have been announced. (Wall Street Journal articled titled “Race to cut taxes fuels urge to merge“dated July 14, 2014). Companies are rushing to do a deal before new legislation closes this window ofopportunity.

17

LBO, the tax savings due to change in incorporation was a big source of value creation.16

Delphi Automotive Plc. (incorporated in Jersey) is a prominent example of a bankruptcy

inversion. Delphi Automotive Plc., once part of GM and the world’s largest auto-parts

maker, emerged from bankruptcy in 2009 as a foreign corporation under the ownership of

JPMorgan Chase, and several hedge funds including Elliot Associates, Silver Point Capital,

and Paulson & Co. The company is currently facing pressure from the IRS to file taxes as

a U.S. based company.17

Panel B of Table 2 provides the distribution of our sample (in terms of firm-years)

during the 1996-2013 period by country of incorporation and transaction type. More than

two-thirds of our sample are incorporated in a tax haven as defined in ? and ?. The top

three tax havens are Bermuda, Cayman Islands, and Ireland. Pure inversions are almost

exclusively incorporated in tax havens (199 of 208 firm-year observations).

Finally, in Tables 3 and 4 we provide a breakdown of our sample by year and subgroup

for the top four tax haven and non-tax haven countries. This panel provides us with

several interesting conclusions. First, there are very few Pure inversions since the passage

of the American Jobs Act of 2004. Second, the number of Restructuring inversions (in

particular merger inversions) is increasing since 2004, with the bulk occurring in tax-havens.

In particular, European tax havens such as Ireland and Switzerland have become very

popular. Third, Canada is becoming an attractive place for U.S. companies to incorporate.

Not only is the corporate tax rate in Canada significantly lower than that in the U.S. (26%

since 2012 as compared to 40%), but Canada also uses a territorial taxation principle versus

the worldwide taxation principle used in the U.S. Given its proximity and close relations

with the U.S., not surprisingly, Canada is also the country with the largest number of AFCs.

Our subsequent analysis is confined to our sample of inversions and control firms.

16Seagate’s average effective tax rate in the last 12 years since the reverse IPO in 2002 was less than 1%.The internal rate of return of over 160% per year of the LBO sponsors in the Seagate LBO can be partlyexplained by the high IPO valuation obtained in anticipation of future sax savings. These tax savings aresubstantially more significant than the savings with interest tax shields associated with high leverage duringthe short two-year period the company stayed private, which is much of the focus of the finance literature(see the Harvard Business School case ?, for more on the Seagate LBO).

17See Wall Street Journal article titled “Delphi Vows to Protect U.K.-Based Status, Fight IRS”, datedAugust 8, 2014. Delphi is incorporated in Jersey, a possession of the British Crown.

18

4.3 Summary statistics

Table 5 provides the summary statistics of the inversions in our sample and for a sample of

all U.S. multinationals. We categorize the variables we use in our analysis into three groups

namely, matching variables, control variables and outcome variables. We find that there are

systematic differences between pure and restructuring inversions and between the inversions

and the sample of U.S. multinationals. Focusing on Log(Total assets), we find that pure

inversions are slightly larger than restructuring inversions and both groups of inversions

are significantly larger than the average U.S. multinational. Firms that invert through

a pure inversion have higher average Market to book ratio and higher ROA as compare

to firms that invert through a restructuring inversion. We also find that the average U.S.

multinational has higher Market to book ratio and lower ROA than the sample of inversions.

The systematic differences between our sample of inversions and the U.S. multinationals

motivates our constraining the control sample to those multinationals that look similar to

the inversions on observable characteristics.

Focusing on mean values, we find that consistent with Pure inversions involving larger

firms, we find that they are more likely to have a bond rating as compared to restructuring

inversions, while both groups of inversions spend less on R&D and SG&A as a proportion

of total assets as compared to the average U.S. multinational.

In terms of outcome variables, interestingly, we find that the mean and median effective

tax rate of inversions is comparable to that of U.S. multinationals. Note that the U.S. multi-

nationals in our sample are less profitable than the inversions. Once we control for this and

other differences, consistent with Prediction 1, we find that the inversions have significantly

lower effective tax rate as compared to the multinationals. We also find that inverted firms

have less cash as a proportion of total assets as compared to the U.S. multinationals, have

slightly higher institutional ownership and greater dispersion in analyst earnings forecast.

We also find that firms that invert through a restructuring inversion have higher bid-ask

spread as compared to both pure inversions and U.S. multinationals.

Table 6 compares the inverted firm-year observations in our sample to the multinational

control firms that we identify by matching on GICS industry, year, Log(Total Assets), Mar-

ket to book and ROA (referred to as control sample CS1 ). Our control sample is very

19

similar to the inverted sample. The differences are not economically or statistically signif-

icant. This is the case not just for matching variables but also for some control variables

such as R&D/Assets and Stock return. Specifically, from the comparison of the distribu-

tion and median of matching variables for our treatment and control firms we accept the

hypothesis that these two groups of variables are statistically equal. From the comparison

of the control variables we find that inverted firms are less likely to have bond ratings (see

the comparison of the 25thpercentile), have less net working capital as a proportion of total

assets, spend less on acquisitions (see the comparison of the 75th percentile), are less likely

to have foreign operations with positive sales (see the comparison of the 25th percentile),

and spend less on advertising (the mean value of the ratio of Advertisement to Sales for

inversions is 0.35% as compared to 0.34% for the control sample). Given these differences,

we control for these variables in our multivariate regressions. In the last column we report

the scaled difference. This is similar to a t-statistic and helps estimate the goodness of

the match. We find that the scaled difference is much smaller than one quarter, a rule of

thumb suggested by ? beyond which linear controls in regression may be problematic. This

offers assurance that misspecification in the linear regression will not significantly bias our

estimates.

5 Empirical Results

5.1 Univariate Results

In Table 7 we report the univariate ATT without any bias correction. The second and third

columns report the mean values of the outcome variables for the inversions and control

sample (CS1 ) while the fourth column reports the difference in means and fifth column

reports the p-values for the difference to be zero. Focusing on the fourth column, we find

that consistent with Prediction 1, the inversions in our sample have significantly lower

effective tax rate than the control firms. We find that the mean GAAP (Cash) ETR is

14.2% (12.5%) for the inverted firms as compared to 21% (19.7%) for control firms. The

lower Cash ETR as compared to GAAP ETR for control firms reflects the fact that many

of these firms may defer paying U.S. taxes on overseas profits by retaining them abroad.

Consistent with the inverted firms facing a lower tax rate, they have lower Log(Cash/TA),

20

although the difference is not significant. We find that once we control for the covariates

in a linear regression, the difference is significant. We find that inversions have higher

bid-ask spread, lower share turnover and higher dispersion in analyst earnings forecast.

All these indicate greater information asymmetry about this firms and consequently lower

stock liquidity. We also obtain some weak evidence for lower institutional ownership among

inversions, although the result is not statistically significant at conventional levels. Finally,

we find that inverted firms have more concentrated institutional ownership, as represented

by a lower value of Log(HHI Inst. own).

5.2 Multivariate Results

In Table 8 we provide the results of our multivariate tests comparing the effective tax rate

of inversions and control firms. In column (1), we report the estimate of equation (1); the

dependent variable is GAAP ETR and the control sample is CS1. To control for residual

difference between the AFCs and control firms, we follow ?, and include contemporaneous

values of Log(Total assets), Net operating loss, Advertising expenditure, Capital expenditure,

Foreign operations, Intangible assets, Leverage, Gross PPE, SG&A, and RD/Assets, as

controls. From the coefficient on Treated we find that inversions have 7.1% lower effective

tax rate as compared to the control firms. This estimate is very similar to our ATT from

Table 7. The coefficients on the control variables indicate that firms with higher leverage

have lower effective tax rates. In column (2) we repeat our tests using a control sample that

we identify by matching on Log(Total assets), Net operating loss, Advertising expenditure,

Capital expenditure, Foreign operations, Intangible assets, Leverage, Gross PPE, SG&A,

and RD/Assets. In this column the set of control variables and the set of matching covariates

are the same. Thus column (2) reports the estimate of biased-corrected ATT. We find

this estimate to be slightly smaller than the ATT but still economically and statistically

significant.

In column (3), in addition to the other control variables, we also include the marginal tax

rate of the country of incorporation as an additional regressor. We find that once we include

the marginal tax rate, the coefficient on Treated turns insignificant. This indicates that the

lower effective tax rate of inversions as seen in columns (1) & (2) is mainly due to the lower

marginal tax rate in their country of incorporation. We also find that the coefficient on

21

the Marginal tax rate is positive and significant. The coefficient on the Marginal tax rate

indicates that a one percentage point change in the marginal tax rate is associated with a

0.367 percentage point change in the GAAP ETR. One reason for the less than one to one

correspondence between the marginal and effective tax rates is because most of our sample

consists of multi-national firms. The effective tax rates of such firms will depend on the tax

rate in all their jurisdictions of operation, not just on the tax rate in the country where the

parent is incorporated in.

In columns (4) - (6) we repeat our tests with the Cash ETR as our dependent variable.

One main potential difference between GAAP ETR and Cash ETR is in the inclusion

of deferred taxes on foreign profits. While the U.S. tax code allows deferral of taxes on

unrepatriated foreign profits, accounting rules may cause them to be included as tax expense

in the GAAP ETR measure. However, the Cash ETR would not include these deferral as

this measure uses the actual cash amount paid in taxes. We find that inversions have

lower Cash ETR as compared to control firms (column (4)), this result is robust to bias

correction (column (5)) and is partly due to the lower marginal tax rate in their country of

incorporation (column (6)).

In Table 9 we test Prediction 2 by comparing the cash balance of inversions and control

firms. In column (1) the control firms are from CS1 and the control variables include

lagged values of Log(Total assets), ROA, Leverage, R&D/Total assets, Dividends, and Net

working capital along with industry and time fixed effects as in ?. We find that while the

coefficient on Treated is positive, it is not significant at conventional levels. The coefficient

on the control variables indicates that smaller firms, those with lower leverage, higher

R&D expenditure, those that pay dividends have more cash as a proportion of total assets.

In columns (2) and (3) we repeat our tests using a control sample that we identify by

matching on Leverage, R&D/Total assets and Acquisitions in addition to those employed

in identifying CS1, and include the matching covariates as regressors. We find that the

coefficient on Treated in column (2) continues to be positive and is now significant, consistent

with inverted firms having higher cash balance than the control firms. We find the size of

the coefficient to be economically significant. The median control firm in our sample has

Log(Cash/TA) of -2.401. In comparison, the coefficient estimate indicates that inversions

have 26.49% (exp(0.235)-1) higher cash balance as a proportion of total assets as compared

22

to the control firms. In column (3) we include the Marginal tax rate as an additional

regressor and find that its coefficient is positive but statistically insignificant. Interestingly

inclusion of the Marginal tax rate actually makes the coefficient on Treated larger. Thus the

difference in marginal tax rates between inversions and control firms is unable to explain

the difference in cash holdings amongst these firms. An alternate reason for the higher

cash balance among inversions could be a precautionary savings motive. Our subsequent

results indicate that inversions have lower stock liquidity. This in turn make increase the

cost of outside capital for these firms, prompting them to rely on internal cash to a greater

extent. Given the small sample size, we are not able to design any convincing tests of this

hypothesis.

In Tables 10 we test Prediction 3 by comparing the stock liquidity and dispersion in

analyst earnings forecast of inversions and control firms. In column (1) the dependent

variable is Spread and the control sample is CS1. We also include contemporaneous values

of Log(Total assets), Leverage, Rated, Capital expenditure, and Volatility as controls along

with industry and time fixed effects. We find that the coefficient on Treated is positive

and significant. This indicates that inversions, on average, have higher bid-ask spread than

control firms. We also find our estimates to be economically significant. In comparison

to the average control firm, inverted firms in our sample have 75% higher bid-ask spread

(.49/.656). Consistent with prior literature, the coefficient on the control variables indicate

that smaller firms and those with more volatile stock returns have higher bid-ask spread.

We also find that rated firms have higher bid-ask spread, which is surprising. In column

(2) we repeat our tests using a control sample that we identify by matching on Rated, and

Volatility in addition to those employed in identifying CS1. We find that the coefficient on

Treated continues to be positive and significant and of the same magnitude as in column

(1). Thus our results in columns (1) and (2) show inverted firms have higher bid-ask spread

as compared to control firms.

In columns (3) - (4) we repeat our tests with stock Turnover as the dependent variable

and again find that inversions have lower share turnover as compared to control firms. The

estimates are statistically significant when we employ the augmented control sample in

column (4).

In columns (5) - (6) we compare the level of dispersion in analyst earnings forecast for

23

inversions and control firms. In column (5) our control sample is CS1 and we find that firms

that invert have higher dispersion in analyst earnings forecast. This is consistent with these

firms have greater information asymmetry. From the control variables we find that firms

with higher leverage and rated firms have higher dispersion in analyst earnings forecast. In

column (6) we repeat our tests with the augmented control sample after we match on Rated

and Volatility in addition to the variables employed in identifying CS1. We find that while

the coefficient on Treated is marginally smaller and is marginally not statistically significant

at conventional levels. The p-value for the coefficient is 10.6%. Summarizing, our results in

Table 10 indicate that inversions on average have higher bid-ask spread, lower turnover and

greater dispersion in analyst earnings forecast as compared to control firms. This highlights

that the difference in the corporate law between inversions and control firms may add an

extra layer of opacity and deter investors from investing in these shares. These results are

consistent with Prediction 3.

In Table 11 we compare the level of institutional shareholding in inverted firms and

control firms. We include contemporaneous values of Log(Total assets), Volatility, Spread,

Market-to-Book, Stock Return, Leverage, and Cash, as controls. Column (1) reports the

estimates when using the control sample CS1. We find that the coefficient on Treated is

negative and significant. This indicates that after controlling for cross-sectional variations,

inverted firms have 7.7% lower institutional ownership than control firms. In column (2) we

repeat our tests using a control sample that we identify by matching on Log(Total assets),

Volatility, Spread, Market-to-Book, Stock Return, Leverage, and Cash, and include these

matching covariates as regressors to correct for matching discrepancies. We find that the

coefficient on Treated continues to be negative and significant. As compared to the median

control firm in our sample with a institutional ownership of 73.8%, this difference represents

a 10.4% (7.7/73.8) lower value.

An interesting question, given the lower stock liquidity for inverted firms, is whether

there is something unique about their ownership structure that limits the costs to share-

holders from lower liquidity. If inverted firms have more concentrated ownership structures

then their shareholders may not mind the fall in stock liquidity. To test this in columns (3)

& (4) we compare the herfindahl index of institutional ownership, Log(HHI Inst. Own), of

inversions and control firms. Our control sample is CS1 in column (3) and the augmented

24

control sample in column (4). We find that the coefficient on Treated is positive and signif-

icant in both columns. This indicates that the herfindahl index of institutional ownership

is higher for inversions as compared to for control firms. Since a higher herfindahl index

implies a more concentrated structure, this is consistent with inverted firms having a more

concentrated institutional ownership structure.18 In columns (5) and (6) we repeat our

tests with the Average institutional ownership as the dependent variable and again find the

coefficient on Treated to be positive and significant. Thus the average institutional investor

in an inverted firm holds a larger percentage of the outstanding shares as compared to the

average institutional investor in a control firm. Summarizing, our evidence in Table 11

shows that while inverted firms have lower institutional share ownership, they have more

concentrated institutional ownership. Furthermore the average institutional investor in an

inverted firm holds more shares.19

In Table 12 we test Prediction 4 by comparing the value of corporate cash using the

methodology in ?. Our dependent variable is the size and book to market adjusted an-

nual abnormal return. In these tests we include all U.S. firms with financial data in

CRSP/Compustat as the control sample. That is, we do not confine the sample to inverted

firms and matched control firms. We do this for two reasons. First is to increase the power

of our tests and second because the identification issues are less severe in these tests given

our dependent variable is stock returns. Our main independent variable is ∆Cash/Mkt.Cap

the coefficient on which measures the market value of a dollar of cash on the firm’s balance

sheet. To test if cash with an inverted firm is differentially valued, we include the interaction

term Inverted × ∆Cash/Mkt.Cap along with Inverted. We include contemporaneous values

of Leverage along with changes in Earnings, Non-cash assets, Interest expense, Total divi-

dends, all normalized by the Market capitalization. We also include the interaction of the

change in Cash with Leverage, lagged values of Cash/Market Capitalization, and Inverted.

Column (1) reports the estimates with firm and year fixed effects along with standards

errors clustered at the industry level. Column (2) reports the estimates that include within

18A possible concern with this result is the extent to which the lower stock liquidity of inversions is dueto their more concentrated institutional share ownership. While this is a legitimate concern, it is importantto note that inversions on average have lower institutional ownership and the effect of ownership structureon stock liquidity will depend on the level of concentration of both institutional and non-institutional share-holders. Since we are not able to get measures of concentration levels of non-institutional shareholders, weare not able to evaluate the effect of ownership structure on liquidity.

19In unreported tests we find that inversions are less likely to raise outside capital as compared to thecontrol firms. This may be another reason why they may not mind the lower stock liquidity.

25

industry year fixed effects and standard errors clustered at the industry level, Column (3)

presents the results of the model where the standard errors are clustered both at the year

and industry level (?) while Column (4) reports the estimates from the Fama-Macbeth

regressions. We find that the coefficient on Inverted×∆Cash/Mkt.Cap is negative and signif-

icant in all columns. Our results indicate that ceteris paribus, investors put a lower value

on cash on an inverted firm’s balance sheet as compared to that on a control firm’s balance

sheet. Our estimates are economically significant. As shown in the last two rows, our esti-

mates in column (1) shows that while the average value of the marginal dollar of cash on

a control firm’s balance sheet is $1.16, investors only assign a $.70 value on the marginal

dollar of cash on an inverted firm’s balance sheet. The higher than $1 value of internal cash

for the average control firm during our sample period is reasonable given that our sample

spans the financial crisis when internal liquidity was quite valuable.

As mentioned before there are two opposing factors that affect the value of cash on

an inverted firm’s balance sheet. The first is the lower marginal tax rate, which is likely

to increase the value of cash on an inverted firm’s balance sheet, and the second is the

weaker corporate law, which is likely to reduce the value. The negative coefficient on

Inverted × ∆Cash/Mkt.Cap in Table 12 indicates that the effect of the weaker rule of law

dominates the effect of the lower tax rate. In the next table we explore this further.

In Table 13 we analyze the interplay between the rule of law in the country of incor-

poration and the marginal value of corporate cash. We use the rule of law index from

the Worldwide Governance Indicators from the Worldbank. To allow for ease of interpre-

tation of the coefficients, we use 100 minus a country’s percentile rank, [100-Percentile

rank] as our main variable to capture a country’s rule of law. Thus, a one unit increase

in [100-Percentile rank] indicates a percentile fall in the ranking of the country in terms

of rule of law. We repeat our tests in Table 12 after including [100-Percentile rank] and

[100 − Percentilerank] × ∆Cash/Mkt.Cap. We find that while the coefficient on the inter-

action term is uniformly negative, it is not statistically significant. Thus we obtain weak

evidence consistent with inverstors attributing a lower valuation on cash on the balance

sheet of firms incorporated in countries with weaker rule of law. Given the strong correla-

tion between [100−Percentilerank] and Inverted, we find that when we include interaction

terms between both variables and ∆Cash/Mkt.Cap, the coefficient on both interaction terms

26

is insignificant.

An important factor that is likely to bias our estimates on [100 − Percentilerank] ×∆Cash/Mkt.Cap is the fact that in our sample there is significant positive correlation between a

country’s rule of law and marginal tax rate (53%). Countries that rank high in the rule of law

index also have higher marginal tax rates. To the extent a higher marginal tax rate reduces

the value of internal cash, this is likely to bias our estimates downward. Furthermore, the

correlation between [100-Percentile rank] and Tax rate is sufficiently strong that it does not

provide meaningful estimates when we include both at the same time.

5.3 Robustness tests

In this section we discuss the results of a number of robustness tests that we perform.

In Table 14 we redo our tests using caliper matching. In the table we provide estimates

of the Hodges- Lehmann median difference for the different outcome variables along with the

? bounds. In the table we present the median comparison to complement our mean analysis

so far. From Table 14 we find that when we compare medians using caliper matching, we

no longer find inversions to have higher Spread, and lower Turnover and Institutional own-

ership. Thus these results appear to be the least robust of our results. We continue to find

inversions to have lower tax rates, higher cash balance, higher analyst dispersion and higher

average institutional ownership. Note that the results from caliper matching and nearest

neighbor matching are not entirely comparable because when we do caliper matching, we

put more weight on treated observations that look similar to control observations – since

such observations are likely to have more control observations for a given caliper size. On

the otherhand, with nearest neighbor match, we have two control observations for every

treated observation.

The ? bounds allow us to understand the robustness of our results to unobserved

heterogenity and outliers. The bound provides an estimate of the amount of unobserved

heterogeneity required to overturn our conclusions. The bound in the last column relate to

a comparison of the mean value for the treated and control sample (using caliper matching),

with a Γ of one indicating that the means are not significantly different from one another

and a larger value of Γ indicating a more robust result. For example Γ of 1.7 for GAAP ETR

27

indicates that unobserved heterogeneity should be strong enough to increase the odds ratio

of being treated by 70% to overturn our conclusion of a lower GAAP ETR among inversions.

We find that the bound is lowest for Turnover and Institutional owership implying that

they are the least robust of our findings. Overall the high values of the Rosenbaum bounds

indicates our results are reasonably robust to unobserved heterogeneity.

In unreported tests we also repeat our analysis without constraining the sample to

selected control firms. That is we compare inversions to all U.S. incorporated multinational

firms in Compustat. Note that the main disadvantage of this approach is that the treated

and control samples can be very different along covariates and thus any misspecification in

the OLS model, in terms of assuming a linear relationship between covariates and outcome

variables when the true relationship is non-linear can bias our estimates. Notwithstanding

this, we find our results are qualitatively similar to the ones reported.

Finally in Table 15 we report the results of our tests that include AFCs along with

inversions. We perform these tests in a sample that includes all inversions and AFCs along

with control observations for both identified by matching on industry, year, and Log(Total

assets), Market to book and ROA i.e., the variables used to identify CS1. We implement

these tests using a model similar to (1) after including two dummy variables, Inversions

and AFC that identify the inversions (and their control observations) and AFCs (and their

control observations) respectively, along with interaction terms between the two variables

and Treated. In 15 we report the coefficients on the two interaction terms with the different

outcome variables. Focusing on the coefficient on AFC × Treated we find that as compared

to the control sample, AFCs have lower GAAP ETR, lower institutional ownership, and

higher concentration of institutional ownership. We find that the other coefficients although

of the same size as those on Inversions × Treated are not statistically significant. This

indicates that while there are similarities between AFCs and inversions, there are also some

important differences.

6 Conclusion

There is flurry of activity among American companies to change their incorporation to

countries with lower corporate tax rate. Just since the beginning of 2014, 14 new merger

28

inversions have been announced, prompting legislative action to stop the move abroad.

In this paper, we collect a large and comprehensive sample of publicly traded American

companies that are incorporated overseas to understand the benefits and costs of changing

a firm’s country of incorporation. Our analysis brings forth a number of interesting results.

Our dataset has over 75 inversions and 186 American foreign companies publicly traded

in U.S. stock exchanges. These firms are classified as an U.S. issuer by the SEC and satisfy

all securities laws that an U.S. firm would. Some of these are large firms as seen from the fact

that over 28 of them are part of the S&P 500 index. We also find that a large percentage

of these firms start out being foreign incorporated since their inception. This highlights

the importance of considering the effect of U.S. tax policy on young firms’ incentives to

incorporate overseas.

Our multivariate analysis indicates a number of differences between inversions and U.S.

firms. Not surprisingly, inversions have significantly lower effective tax rate than U.S. multi-

national firms. Inversions also hold more cash than U.S. incorporated firms. Highlighting

an important cost of foreign incorporation, we find that inversions have less liquid stock

and have less institutional ownership. Finally, we find that investors put a lower value on

corporate cash for firms incorporated overseas. Overall, our analysis highlights both the

benefits and costs of inversions.

The gap between the corporate tax rate in the U.S. and other OECD countries is in-

creasing. The average corporate tax rate among OECD countries reduced from 33% in 2000

to 25% in 2013. Furthermore unlike most OECD countries, the U.S. uses the worldwide

taxation system as opposed to the territorial taxation system. For example, countries like

the U.K., Canada, and Switzerland have now adopted the territorial taxation system, and

their corporate tax rate is now 15% less than the U.S. corporate tax rate. Moreover, they

all rank well in terms of rule of law. Given the costs and benefits of foreign incorporation

documented in this paper, these factors combined with competitive pressures are likely to

prompt more U.S. corporations to explore the possibility of a foreign incorporation. We

hope our work will contribute to the current debate on this growing important phenomenon.

29

Appendix 1: American Foreign Corporations and Foreign

Private Issuers

This appendix, based on the Federal Securities Laws, explains the definition of foreign pri-

vate issuers and American foreign corporations (the terminology we are using in this paper).

It also helps better understand more detailed aspects of our data collection procedure.

A key consideration for a foreign corporation is whether it qualifies as a foreign private

issuer as defined in Rule 405 of Regulation C under the Securities Act and Rule 3b-4 under

the Exchange Act. If a company does not qualify as a foreign private issuer (FPI), it is

subject to the same registration and disclosure requirements applicable to domestic U.S.

entities. The SEC considers that if a foreign company has sufficient contacts with the U.S.,

that is the company is essentially an U.S. issuer, there is an important U.S. public interest in

the company that justifies treating it the same as a U.S. company for regulatory purposes.

A foreign corporation is classified as a foreign private issuer if it meets the following

conditions:

• More than 50 percent of the outstanding voting securities of such issuer are directly

or indirectly owned of record by residents of the U.S.; and

• Any of the following:

– The majority of the executive officers or directors are U.S. citizens or residents;

– More than 50 percent of the assets of the issuer are located in the U.S.; or

– The business of the issuer is administered principally in the U.S.

Foreign private issuers are granted special status and the disclosure and governance rules

that applies to them are less stringent than the ones that applies to U.S. issuers and AFCs.

In particular:

• Foreign private issuers may present financial statements pursuant to U.S. generally

accepted accounting principles (GAAP), International Financial Reporting Standards

(IFRS) as issued by the International Accounting Standards Board (IASB), or home

country accounting standards with a reconciliation to U.S. GAAP.

30

• Foreign private issuers are exempt from the proxy rules under Rule 3a12-3(b) of the

Exchange Act.

• Insiders of foreign private issuers are exempt from filing beneficial ownership reports

required by Section 16(a) of the Exchange Act and are not subject to the short-swing

trading rules under Section 16(b) of the Exchange Act.

• Foreign private issuers are exempt from the disclosure requirements of Regulation FD.

• Foreign private issuers may use particular registration and reporting forms designed

specifically for them.

• Foreign private issuers may use a special exemption from registration under the Ex-

change Act.

A foreign company must determine its status as an FPI on an annual basis, as of the end

of its second fiscal quarter (this particular aspect of the law changed a few years ago; the

test had to be performed at the end of every fiscal quarter). Companies doing an IPO, and

filing a registration statement under the Securities Act or the Exchange Act for the first

time, may make a determination as to FPI/AFC status up to 30 days before filing its initial

registration statement. Several companies in our dataset, such as Seagate Technologies

Plc. and Avago Technologies Ltd., file as U.S. domestic issuers since their IPOs. Upon

registration, a foreign company would determine its status on an annual basis, as of the end

of its second fiscal quarter.

If the company determines that it no longer meets the FPI definition, it must transition

to a domestic reporting status and it becomes subject to the reporting requirements for a

domestic company beginning on the first day of the next fiscal year. A company that no

longer qualifies as FPI as of the end of its second fiscal quarter in 2012, for example, would

file a form 10-K in 2013 for its 2012 fiscal year. The company would also begin complying

with the proxy rules and Section 16, and become subject to reporting on forms 8-K and

10-Q, on the first day of its 2013 fiscal year. (Note though that the company can voluntarily

start using forms 8-K and 10-Q before the beginning of the next fiscal year).

Alternatively, if a company qualifies as an FPI on the last business day of its second

fiscal quarter, it can immediately avail itself of the FPI accommodations, including the use

31

of FPI forms under the Securities Act and reporting requirements under the Exchange Act.

For example, if an AFC that switches to FPI status as of the end of its second fiscal quarter

would no longer need to continue reporting on form 8-K and form 10-Q for the remainder

of that fiscal year. Instead, it could immediately begin furnishing reports on form 6-K and

would file an annual report on form 20-F or form 40-F (for Canadian companies) for the

current fiscal year.

A company need not provide notice to the market nor the SEC of its FPI/AFC status

(there are no specific forms nor notification requirements). However, many companies in our

dataset do provide notice to the market of changes in their FPI/AFC status. Practically,

however, by furnishing a current report on form 6-K rather than 8-K, or form 10-K rather

than 20-F or 40-F, and by filing a proxy statement on form DEF-14A, the company will in

essence be providing clues of its FPI/AFC status. Note that FPIs can voluntarily file forms

8-K and 10-K, and several firms indeed choose to do so. However, the proxy statement on

form DEF-14A may not be filed by FPIs (although we did encounter a few instances of

FPIs filing both forms 10-Ks and forms DEF-14A).20

According to our conversations with SEC staff lawyers from the Division of Corporation

Finance, the type of forms the company is filing is the main information the SEC itself uses

when producing the annual FPI lists, which is available in the SEC website.21 Notice that

these annual lists (say the December 2012 list) are produced by the SEC around July of

the next year (July 2013), after the SEC staff has had the time to observe the forms most

companies used to file their annual reports for the 2012 fiscal year (most companies file

their annual reports around April).

This may potentially lead to inaccuracies, and thus we individually double checked

every AFC/FPI classification in our sample. For all the foreign companies in our dataset,

we performed a keyword search of the term “foreign private issuer” on all their electronic

filings in the SEC EDGAR database and analyzed the results of the matches (the EDGAR

database started in 1996, the date our sample starts). In instances, when we encountered

discrepancies from the company reported information and the SEC list, we overruled the

SEC classification only when there was very strong evidence to do so. For example, if a

20SEC no-action letter “Proxy Materials of Foreign Private Issuers” (March 10, 1992).21http://www.sec.gov/divisions/corpfin/internatl/companies.shtml

32

foreign company appeared in an SEC list in a given year, say December 2012, and the

company filed both forms 10-K and DEF-14A for the 2012 fiscal year, and the company

explicitly mentioned in some of its previous filings that it no longer qualified as an FPI,

then we did change the SEC classification from FPI to AFC.

33

Appendix 2: Variable Definitions

• AFC: A dummy variable that identifies firm observations that are incorporated abroad

but are classified as an U.S. issuer (i.e., not a foreign private issuer) by the SEC.

• Treated: A dummy variable that identifies observations for which there is at least one

control firm in the matched sample.

Compustat Variables

• GAAP ETR: The ratio of total income tax expense to pre-tax income before special

items, with extreme values truncated at zero and one.

• Cash ETR: The ratio of cash tax paid to pre-tax income before special items, with

extreme values truncated at zero and one.

• Cash/TA: The ratio of cash and short-term investments and the book value of total

assets.

• ROA: The ratio of earnings before interest, depreciation, and taxes to the book value

of total assets.

• Market-to-Book: The ratio of the sum of the market value of equity and the book

value of debt to the book value of total assets.

• Leverage: The ratio of the book value of total debt to the book value of book value

of total assets.

• R&D/Assets: The ratio of R&D expenditures to the book value of total assets.

• Acquisitions: The ratio of acquisition expenditures to the book value of total assets.

• Dividends: The ratio of total dividends paid to the book value of total assets.

• Net Working Capital: The ratio of accounts receivable plus inventories minus accounts

payables to total sales.

• Capital expenditure: The ratio of capital expenditures to the book value of total assets.

• Earnings: Operating income after depreciation.

34

• Non-Cash assets: Book value of total assets minus the book value of cash and short-

term investments.

• Rated: A dummy variable that identifies borrowers that have an unsecured long-term

credit rating.

• Interest: Interest expense from the income statement.

• Dividends: The ratio of total dividends paid to the book value of total assets.

• Net Operating Loss: An indicator if the firm has a non-missing value of tax loss

carry-forward.

• SG&A: Selling, general, and administrative expense divided by net sales.

• Advertising: Advertising expense divided by net sales.

• Foreign Operations: An indicator variables that takes the value of one if the firm has

a non-missing, non-zero value for pre-tax income from foreign operations.

• Gross PPE: The ratio of gross property, plant and equipment divided by total assets.

Stock Performance Variables

• Spread: The ratio of the closing ask minus the closing bid to the closing price.

• Share Turnover: The yearly average of daily ratio of share volume to the number of

shares outstanding.

• Analysts dispersion: Standard deviation of the analysts earnings forecast.

• Abnormal returns: Difference between the stock return and the benchmark return of

the 25 size and book-to-market portfolio.

Institutional Ownership Variables:

• Institutional Ownership: The ratio of total 13-F institutional ownership to the number

of shares outstanding.

35

• Log(HHI Institutional Ownership): Log of the Herfindahl-Hirschman Index of owner-

ship concentration.

• Log(Average Institutional Ownership): Log of the ratio of total 13-F institutional

ownership to the number of shares outstanding divided by the number of 13-F Insti-

tutional Owners

Country specific characteristics

• ROL: Rule of law index. The index ranges from approximately -2.5 (weak) to 2.5

(strong) governance performance.

• Maginal Tax Rate: Statutory marginal corporate tax rate of the highest tax bracket

prevailing in the country of incorporation.

• Percentile rank: Ranking of the country’s of incorporation rule of law.

36

Table 1: S&P 500 index membership of American foreign corporations

This table shows the 28 American foreign corporations (AFCs) included in the S&P 500 index as ofDecember 31, 2013. Country/State of Incorporation denotes, in chronological order, all the countriesand U.S. states where the company has been incorporated during the 1996-2013 period. AFC originrepresents how the firm became an AFC (see Section 4.2 for the classification system). Market cap.is the company’s stock market capitalization (in millions of U.S. dollars) as of December 31, 2013.

Corporation Name Country/State of

Incorporation

AFC Origin Market

Cap.

Schlumberger Ltd. Netherlands Antilles AFC 118,670

LyondellBasell Ind. N.V. Delaware, Netherlands AFC 44,392

Accenture Plc. Illinois, Bermuda, Ireland AFC 52,373

ACE Ltd. Cayman Islands, Switzerland AFC 35,224

Carnival Corp. Panama AFC 23,777

Michael Kors Hldgs Ltd. British Virgin Islands AFC 16,550

Invesco Ltd. U.K., Bermuda AFC 16,135

Garmin Ltd. Taiwan, Cayman Islands,

Switzerland

AFC 9,017

XL Group Plc. Cayman Islands, Ireland AFC 9,004

Aon Plc. Delaware, U.K. Pure Inversion 25,254

Tyco Intl. Ltd. Massachusetts, Bermuda,

Switzerland

Pure Inversion 19,096

Transocean Ltd. Texas, Cayman Islands,

Switzerland

Pure Inversion 17,820

Ingersoll-Rand Plc. New Jersey, Bermuda, Ireland Pure Inversion 17,746

Ensco Plc. Delaware, U.K. Pure Inversion 17,565

Noble Corp. Delaware, Cayman Islands,

Switzerland, U.K.

Pure Inversion 9,495

Nabors Ind. Ltd. Delaware, Bermuda Pure Inversion 5,014

Rowan Companies Plc. Texas, U.K. Pure Inversion 4,392

Eaton Corp. Plc. Ohio, Ireland Merger Inversion 36,118

Actavis Plc. Nevada, Ireland Merger Inversion 29,240

Perrigo Plc. Michigan, Ireland Merger Inversion 20,514

Pentair Plc. Minnesota, Switzerland Merger Inversion 15,482

Nielsen N.V. Netherlands Merger/LBO Inv. 17,359

Seagate Tech. Plc. Delaware, Cayman Islands, Ireland LBO Inversion 18,316

Avago Tech. Ltd. Delaware, Singapore LBO Inversion 13,172

Delphi Automotive Plc. Michigan, U.K., Jersey Bankruptcy Inversion 18,503

Covidien Plc. Bermuda, Ireland Spin-off Inversion 30,808

TE Connectivity Ltd. Bermuda, Switzerland Spin-off Inversion 22,615

Allegion Plc. Ireland Spin-off Inversion 4,242

37

Table 2: The number of inversions and American Foreign Corporations (AFCs)

Panel A: The number of unique inversions and AFCs and the number of firm-year observationswithin each group during the 1996-2013 period.

# Companies # Firm-year Observations

Pure Inversions 25 (33.33%) 208 (53.1%)

Restructuring Inversions 50 (66.67%) 184 (46.9%)

Inversions- Total 75 (100%) 392 (100%)

AFCs 186 1017

Total 261 1409

Panel B: Number of inversions and AFC firm-year observations during the 1996-2013 period bycountry of incorporation. The tax-haven classification follows ?.

Country of

Incorporation

Pure

Inversions

Restructuring

Inversions

Total-

Inversions

AFCs Total

Tax haven countries

Bermuda 98 46 144 335 479

Cayman Islands 43 30 73 91 164

Ireland 12 15 27 35 62

Marshall Islands 0 0 0 43 43

Switzerland 28 8 36 8 44

British Virgin Islands 0 7 7 32 39

Panama 18 3 21 18 39

Netherlands Antilles 0 2 2 26 28

Singapore 0 11 11 17 28

Luxembourg 0 0 0 15 15

Bahamas 0 0 0 28 28

Jersey 0 2 2 5 7

Liberia 0 0 0 9 9

Tax Havens - Subtotal 199 124 323 662 985

Non-Tax haven countries

Canada 0 16 16 254 270

Netherlands 2 43 45 16 61

Israel 0 0 0 38 38

United Kingdom 7 0 7 29 36

France 0 0 0 12 12

Australia 0 1 1 3 4

Curacao 0 0 0 3 3

Non-Tax Havens - Subtotal 9 60 69 355 424

Total - All countries 208 184 392 1017 1409

38

Table 3: Number of inversions and AFC firms by year and by country of incorporation for the top four countries

This table shows, for each year during the 1996-2013 period, the number of inversions and AFCs by country of incorporation for the four countries

with the largest total number of inversions/AFCs.

Year ’96 ’97 ’98 ’99 ’00 ’01 ’02 ’03 ’04 ’05 ’06 ’07 ’08 ’09 ’10 ’11 ’12 ’13 Total

Bermuda

Pure Inv. 1 1 2 2 3 6 8 8 9 9 9 9 10 6 4 4 4 3 98

Restructuring Inv. 0 0 0 0 1 3 2 2 2 3 5 4 6 6 4 3 3 2 46

AFC 13 13 13 14 11 10 12 15 18 23 26 26 25 23 25 23 23 22 335

Total - Bermuda 14 14 15 16 15 19 22 25 29 35 40 39 41 35 33 30 30 27 479

Cayman Islands

Pure Inv. 2 3 3 3 3 2 2 3 3 3 3 3 3 3 1 1 1 1 43

Restructuring Inv. 0 0 0 0 0 0 1 1 2 2 2 3 3 4 4 2 3 3 30

AFC 3 2 2 3 3 4 5 5 5 5 6 6 4 8 8 7 8 7 91

Total - Cayman Islands 5 5 5 6 6 6 8 9 10 10 11 12 10 15 13 10 12 11 164

Ireland

Pure Inv. 0 0 0 0 0 0 0 0 0 0 0 0 0 2 3 3 2 2 12

Restructuring Inv. 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 3 4 7 15

AFC 2 2 2 1 1 2 2 2 2 2 2 2 1 2 2 3 3 2 35

Total - Ireland 2 2 2 1 1 2 2 2 2 2 2 2 1 4 6 9 9 11 62

Canada

Pure Inv. 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

Restructuring Inv. 2 2 2 2 0 0 0 0 0 0 1 0 0 0 1 2 2 2 16

AFC 9 11 11 11 11 13 14 16 13 18 18 20 17 18 15 13 14 12 254

Total - Canada 11 13 13 13 11 13 14 16 13 18 19 20 17 18 16 15 16 14 270

39

Table 4: Number of inversions and AFC firms by year and by country of incorporation for tax haven and non-tax-haven countries

This table shows, for each year during the 1996-2013 period, the total number of inversions and AFCs by country of incorporation for all tax-haven

and non-tax-haven countries.

Year ’96 ’97 ’98 ’99 ’00 ’01 ’02 ’03 ’04 ’05 ’06 ’07 ’08 ’09 ’10 ’11 ’12 ’13 Total

All Tax Havens

Pure Inv. 4 5 6 6 7 9 11 12 13 13 13 13 15 15 15 15 14 13 199

Restructuring Inv. 1 1 1 0 2 4 3 4 5 6 9 9 12 13 12 11 14 17 124

AFCs 24 25 25 27 35 24 28 33 36 44 48 47 42 48 50 47 49 10 662

All Tax Havens 29 31 32 33 34 37 42 49 54 63 70 69 69 76 77 73 77 70 985

All Non-Tax Havens

Pure Inv. 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 2 4 9

Restructuring Inv. 3 3 5 5 3 3 3 3 2 2 3 2 2 2 3 5 5 6 60

AFCs 13 15 14 16 19 25 26 22 20 23 22 23 19 20 19 18 20 21 355

Non-Tax Havens 16 18 19 21 22 28 29 25 22 25 25 25 21 23 23 24 27 31 424

All Countries

Pure Inv. 4 5 6 6 7 9 11 12 13 13 13 13 15 16 16 16 16 17 208

Restructuring Inv. 4 4 6 5 5 7 6 7 7 8 12 11 14 15 15 16 19 23 184

AFCs 37 40 39 43 44 49 54 55 56 67 70 70 61 68 69 65 69 61 1017

Total - All Countries 45 49 51 54 56 65 71 74 76 88 95 94 90 99 100 97 104 101 1409

40

Table 5: Summary Statistics for inversions

This table presents the descriptive statistics based on the nature of the transaction that preceded the firms the Corporate Inversions (Pure Inversions, andRestructuring Inversion). Pure inversions include U.S. companies that reincorporate in a new country, and the same previous shareholders own shares in thenew foreign parent company with no material change in the company’s business and assets. Restructuring Inversion include Merger inversions, LBO inversions,Bankruptcy inversions, and Spin-off inversions (refer to section 4.2 for a complete description of these categories). U.S. multinationals are the U.S. incorporatedfirms that report positive foreign sales in the Compustat Segments data. All variables are winsorized at the 1st and 99th percentile. All variables are defined inthe Appendix 2.

Corporate Inversions Restructuring Inversions U.S. Multinationals

(Potential control candidates)

N Mean Median N Mean Median N Mean Median

Matching variables

Log(Total Assets) 209 8.681 9.197 109 7.978 8.185 17758 5.984 5.802

Market to book 209 1.657 1.285 103 1.383 1.249 17261 2.291 1.7

ROA 209 0.083 0.077 109 0.034 0.061 17758 0.016 0.057

Control variables

RD/Assets 209 0.005 0 109 0.024 0.003 17758 0.083 0.05

Rated 209 0.914 1 109 0.147 0 17758 0.727 1

Acquisition 209 0.015 0 109 0.034 0 17758 0.026 0

Stock return 204 0.175 0.141 88 0.126 0.149 15835 0.168 0.043

Volatility 206 0.026 0.023 101 0.031 0.025 16957 0.039 0.034

Foreign Operations 209 0.742 1 109 0.661 1 17758 0.655 1

Capital Expenditures/Assets 209 0.045 0.02 109 0.031 0.018 17758 0.046 0.029

Advertisement Expenditure 209 0.005 0 109 0.002 0 17758 0.011 0

Gross PPE/Assets 209 0.419 0.303 109 0.29 0.183 17758 0.383 0.273

SG&A Expenditures/Assets 209 0.116 0.087 109 0.166 0.113 17758 0.419 0.339

Outcome variables

GAAP ETR 205 0.188 0.186 108 0.129 0.171 17592 0.143 0.215

Cash ETR 198 0.143 0.16 107 0.136 0.109 16034 0.153 0.124

Log(Cash/Assets) 209 -2.384 -2.274 108 -2.382 -2.320 17729 -1.887 -1.580

Spread 202 0.664 0.109 92 1.443 0.133 15925 1.303 0.446

Turnover 202 9.331 7.656 92 7.378 6.148 15929 9.332 6.780

Analyst Dispersion 185 0.296 0.16 83 0.227 0.17 12550 0.158 0.07

Institutional Ownership 121 0.706 0.756 56 0.62 0.714 14266 0.579 0.626

Log (Inst. Own Herfindahl Index) 121 -2.904 -3.057 57 -2.838 -2.967 14278 -2.560 -2.762

Log(Average Institutional Ownership) 121 -5.699 -5.619 56 -5.414 -5.343 14266 -5.182 -5.086

41

Table 6: Summary comparison of inversions and control sample

This table presents descriptive statistics that compare treatment firms and control firms. The sample comprises 443 firm-year Corporate Inversion observations,and up to twice the number of control firms matched by industry, Log(Total Assets), Market to Book, and ROA (sample CS1). Both groups of firms are publiclytraded operating firms. The last column reports the scaled difference statistic proposed by ?.

T =X̄1 − X̄0√S2

1 + S20

All variables are scaled by total assets. All variables are winsorized at the 1st and 99th percentile. All variables are defined in the Appendix 2.

25th Percentile 50th Percentile 75th Percentile P-values for median

comparison

P-values for

distribution

comparison

Scaled difference

Inversions Control Inversions Control Inversions Control

Matching variables

Log(Total Assets) 6.754 6.830 8.434 8.309 9.588 9.112 0.234 0.131 0.037

Market to book 1.029 1.076 1.317 1.375 2.038 1.869 0.264 0.310 0.078

ROA 0.030 0.032 0.072 0.069 0.113 0.115 0.341 0.937 -0.020

Control variables

RD/Assets 0.000 0.000 0.000 0.000 0.012 0.019 0.907 0.539 0.004

Rated 0.000 1.000 1.000 1.000 1.000 1.000 . 0.000 -0.505

Net working capital 0.127 0.133 0.196 0.207 0.351 0.332 0.031 0.721 0.101

Acquisition 0.000 0.000 0.000 0.000 0.010 0.013 0.043 0.074 0.008

Stock return -0.122 -0.130 0.143 0.113 0.407 0.359 0.153 0.334 0.030

Volatility 0.017 0.017 0.025 0.024 0.035 0.034 0.533 0.676 0.045

Foreign Operations 0.000 1.000 1.000 1.000 1.000 1.000 . 0.000 -0.187

Capital Expenditures/Assets 0.004 0.010 0.026 0.027 0.063 0.076 0.552 0.009 -0.122

Intangibles 0.002 0.007 0.049 0.070 0.283 0.242 0.061 0.652 0.027

Net Operating Loss Balance 0.000 0.000 1.000 1.000 1.000 1.000 . 0.132 0.064

Advertisement Expenditure 0.000 0.000 0.000 0.000 0.000 0.000 0.077 0.077 -0.026

Gross PPE/Assets 0.096 0.097 0.284 0.315 0.609 0.770 0.284 0.020 -0.107

SG&A Expenditures/Assets 0.042 0.007 0.092 0.103 0.194 0.223 0.234 0.783 -0.006

42

Table 7: ATT without bias correction

This table presents mean comparison between treatment firms and control firms. The sample comprises443 Corporate Inversions firm-year observations, and up to twice the number of control firms matched byindustry, Log(Total Assets), Market to Book, and ROA (sample CS1). Both groups of firms are publiclytraded multinational operating firms. All corporate policy variables are scaled by total assets. All variablesare winsorized at the 1st and 99th percentile. All variables are defined in the Appendix 2.

Mean Difference p-value of the difference

Inversion Control

(1) (2) (3) (4) (5)

GAAP ETR 0.142 0.210 -0.068∗∗∗ 0.008

Cash ETR 0.125 0.197 -0.072∗∗∗ 0.001

Log(Cash/TA) -2.328 -2.507 0.179∗∗∗ 0.007

Spread 1.042 0.656 0.386∗∗∗ 0.002

Turnover 8.940 9.588 -0.649 0.169

Analyst dispersion 0.298 0.246 0.052∗∗ 0.026

Institutional Ownership 0.680 0.738 -0.057 0.185

Log(HHI Inst. Own.) -2.806 -2.929 0.123∗∗ 0.033

Log(Average Institutional Ownership) -5.506 -5.615 0.109∗ 0.056

43

Table 8: Effect of inversions on effective tax rate

This table reports the results of regressions investigating the impact of Corporate Inversions on the effectivetax rate. The sample in columns (1) and (4) comprises 443 Corporate Inversions firm-year observations andup to twice the number of control firm-year observations matched by industry, Log(Total Assets), Marketto Book, and ROA (sample CS1 ). The sample in columns (2), (3), (5), and (6) comprises 443 CorporateInversion firm-year observations and up to twice the number of control firm-year observations matchedby industry, Log(Total assets), Net operating loss, Advertising expenditure, Capital expenditure, Foreignoperations, Intangible assets, Leverage, Gross PPE , SG&A, and RD/Assets. In each column, we estimatethe regression:

Tax Rateit = β0 + β1 ∗ Treatedit +Xit · γ + αi + δt + εit

We estimate this regression on all the firm-year treated and control firms in our sample from 1996 to 2012.Columns (1)-(3) estimate the effect of Corporate Inversion on the (GAAP) Effective Tax Rate. Columns(4)-(6) present the effect of Corporate Inversion on the (Cash) Effective Tax Rate. Columns (2) and (4)include as regressors the set of matching covariates and report the bias-corrected average treatment effect onthe treated. All regressions include industry and year fixed effects and standard errors clustered at the firmlevel. All variables are winsorized at the 1st and 99th percentile. All variables are defined in the Appendix2. For brevity, we suppress the coefficients on the fixed effects.

GAAP ETR Cash ETR

Control sample CS1 Augmented control sample CS1 Augmented control sample

(1) (2) (3) (4) (5) (6)

Treatedt -.071 -.068 .039 -.080 -.080 -.083(.035)∗∗ (.031)∗∗ (.059) (.024)∗∗∗ (.023)∗∗∗ (.051)

Marginal tax ratet .367 -.027(.192)∗ (.152)

Log(Total assets)t .021 .006 .009 .006 .011 .011(.016) (.014) (.014) (.012) (.010) (.009)

Net Operating Losst .021 -.006 -.009 -.004 -.005 -.009(.032) (.037) (.037) (.027) (.030) (.031)

Advertising Expendituret -.172 -1.195 -.681 -.256 -.521 -.705(.809) (3.247) (3.407) (.669) (2.086) (2.151)

Capital Expendituret -.136 -.451 -.411 -.073 -.288 -.328(.454) (.638) (.622) (.321) (.381) (.372)

Foreign Operationst -.022 .025 .028 -.024 -.032 -.030(.043) (.044) (.043) (.032) (.033) (.033)

Intangible Assetst -.049 -.015 -.026 .255 .195 .202(.117) (.148) (.146) (.079)∗∗∗ (.090)∗∗ (.088)∗∗

Leveraget -.389 -.383 -.382 -.241 -.157 -.163(.141)∗∗∗ (.158)∗∗ (.157)∗∗ (.102)∗∗ (.108) (.107)

Gross PPEt .026 -.027 -.031 .005 -.041 -.047(.080) (.090) (.090) (.063) (.066) (.065)

SG&At .026 .063 .070 -.072 .045 .044(.055) (.083) (.087) (.106) (.061) (.062)

RD/Assetst -.138 .039 .157 -.456 -.039 .027(.406) (.443) (.455) (.456) (.323) (.316)

Const. .132 .260 .085 .208 .166 .184(.140) (.130)∗∗ (.164) (.102)∗∗ (.098)∗ (.111)∗

Obs. 1176 1230 1220 1111 1170 1161

R2 .058 .046 .049 .065 .068 .067

44

Table 9: Effect of inversions on cash holdings

This table reports the results of regressions investigating the impact of Corporate Inversion on corporate cashholdings for Corporate Inversions. The sample in columns (1) and (4) comprises 443 Corporate Inversionsfirm-year observations and up to twice the number of control firm-year observations matched by industry,Log(Total Assets), Market to Book, and ROA (sample CS1 ). The sample in columns (2), (3), (5), and(6) comprises 443 Corporate Inversion firm-year observations and up to twice the number of control firm-year observations matched by industry, Log(Total Assets), Market to Book, ROA , Leverage, R&D andAcquisitions expenditures scaled by total assets. In each column, we estimate the regression:

Yit = β0 + β1 ∗ Treatedit +Xit−1 · γ + αi + δt + εit

Columns (2) and (4) include as regressors the set of matching covariates and report the bias-corrected averagetreatment effect on the treated. We estimate this regression on all the firm-year treated and control firmsin our sample from 1996 to 2013. All variables are winsorized at the 1st and 99th percentile. All variablesare defined in the Appendix 2. For brevity, we suppress the coefficients on the fixed effects.

Log(Cash/TA)

Control sample CS1 Augmented control sample

(4) (5) (6)

Treatedt .113 .235 .334(.089) (.089)∗∗∗ (.181)∗

Marginal tax ratet .335(.499)

Log(Total assets)t−1 -.102 -.064 -.061(.041)∗∗ (.036)∗ (.036)∗

ROAt−1 .269 -.277 -.264(.405) (.480) (.469)

Leveraget−1 -1.262 -1.400 -1.392(.259)∗∗∗ (.325)∗∗∗ (.318)∗∗∗

R&D/Assetst−1 1.591 1.192 1.288(.702)∗∗ (.937) (.936)

Dividendst−1 5.436 1.884 1.935(1.651)∗∗∗ (2.476) (2.489)

Acquisitionst−1

Net working capitalt−1 .073 .020 .027(.112) (.127) (.124)

Const. -1.037 -1.269 -1.462(.406)∗∗ (.319)∗∗∗ (.477)∗∗∗

Obs. 1042 1063 1057

R2 .423 .414 .41

45

Table 10: Effect of inversions on analyst coverage and stock liquidity

This table reports the results of regressions investigating the impact of Corporate Inversion on analystcoverage and stock liquidity. The sample in columns (1), (3), and (5) comprises 443 Corporate Inversionfirm-year observations and up to twice the number of control firm-year observations matched by industry,Log(Total Assets), Market to Book, and ROA (sample CS1 ). The sample in columns (2), (4), and (6)comprises 443 Inversions firm-year observations and up to twice the number of control firm-year observationsmatched by industry, Log(Total Assets), Market to Book, ROA , Stock volatility and a binary indicator thattakes the value of 1 if the firm has a credit Rating and zero otherwise. In each column, we estimate theregression:

Yit = β0 + β1 ∗ Treatedit +Xit · γ + αi + δt + εit

Columns (2), (4), and (6) include as regressors the set of matching covariates and report the bias-correctedaverage treatment effect on the treated. We estimate this regression on all the firm-year treated and controlfirms in our sample from 1996 to 2013. All variables are winsorized at the 1st and 99th percentile. Allvariables are defined in the Appendix 2. All regressions include year and industry fixed effects and standarderrors clustered at the firm level. For brevity, we suppress the coefficients on the fixed effects.

Spread Turnover Analyst dispersion

CS1 Augmented CS1 Augmented CS1 Augmented

control sample control sample control sample

(1) (2) (3) (4) (5) (6)

Treatedt .490 .485 -.583 -1.465 .120 .091(.223)∗∗ (.215)∗∗ (.618) (.641)∗∗ (.064)∗ (.056)

Log(Total assets)t -.287 -.261 1.358 2.029 -.010 -.008(.071)∗∗∗ (.053)∗∗∗ (.189)∗∗∗ (.279)∗∗∗ (.020) (.020)

Leveraget -.277 -.362 3.890 2.803 .252 .280(.557) (.589) (1.766)∗∗ (2.236) (.151)∗ (.134)∗∗

Ratedt .516 .622 1.781 -.213 .144 .094(.302)∗ (.310)∗∗ (.691)∗∗∗ (.810) (.056)∗∗ (.061)

Capital expendituret -.490 .390 -5.638 5.410 -.211 .022(1.258) (1.420) (5.927) (7.689) (.256) (.355)

Volatilityt 36.823 52.840 148.557 171.254 1.110 1.222(18.105)∗∗ (20.239)∗∗∗ (28.105)∗∗∗ (32.388)∗∗∗ (1.469) (1.489)

Const. 3.754 3.128 -11.059 -14.508 .018 -.029(.653)∗∗∗ (.645)∗∗∗ (1.864)∗∗∗ (2.665)∗∗∗ (.164) (.144)

Obs. 1164 1197 1167 1199 1019 1064

R2 .569 .587 .46 .484 .255 .211

46

Table 11: Effect of inversions on institutional ownershipThis table reports the results of regressions investigating the impact of Corporate Inversion on institutionalownership and ownership characteristics. The sample in columns (1), (3), and (5) comprises 443 CorporateInversions firm-year observations and up to twice the number of control firm-year observations matched byindustry, Log(Total Assets), Market to Book, and ROA (sample CS1 ). The sample in columns (2), (4), and(6) comprises 443 Corporate Inversions firm-year observations and up to twice the number of control firm-year observations matched by industry, Log(Total assets), Volatility, Spread, Market-to-Book, Stock Return,Leverage, and Cash. In each column, we estimate the regression:

Yit = β0 + β1 ∗ Treatedit +Xit−1 · γ + αi + δt + εit

Columns (2) includes as regressors the set of matching covariates and report the bias-corrected averagetreatment effect on the treated. We estimate this regression on all the firm-year treated and control firms inour sample from 1996 to 2013. All variables are winsorized at the 1st and 99th percentile. All variables aredefined in the Appendix 2. All columns include fixed effects at year and industry level along with standarderrors clustered at industry level. For brevity, we suppress the coefficients on the fixed effects.

Institutional Ownership Log(HHI Inst. Own ) Log(Avg. Institutional Ownership)

CS1 Augmented CS1 Augmented CS1 Augmented

control sample control sample control sample

(1) (2) (3) (4) (5) (6)

Treatedt -.077 -.057 .127 .164 .130 .113(.041)∗ (.025)∗∗ (.062)∗∗ (.062)∗∗∗ (.068)∗ (.066)∗

Log(Total assets)t .022 .037 -.187 -.182 -.302 -.342(.026) (.011)∗∗∗ (.022)∗∗∗ (.021)∗∗∗ (.041)∗∗∗ (.035)∗∗∗

Volatilityt -.354 .184 -.394 -.762 -.886 -.481(.420) (.097)∗ (.250) (.272)∗∗∗ (.325)∗∗∗ (.251)∗

Stock Returnt -.008 .005 -.048 -.037 -.084 -.105(.029) (.012) (.025)∗ (.031) (.033)∗∗ (.032)∗∗∗

Spreadt -.442 -1.485 4.163 2.354 2.455 .655(1.552) (.957) (1.878)∗∗ (2.024) (3.351) (3.298)

Market to bookt -.035 -.003 -.0006 .007 .065 .159(.035) (.015) (.038) (.046) (.035)∗ (.053)∗∗∗

Tangibilityt -.069 -.061 .091 .164 -.151 -.156(.042) (.016)∗∗∗ (.026)∗∗∗ (.045)∗∗∗ (.090)∗ (.085)∗

Debt/Assetst -.365 -.115 .327 .294 .090 .174(.279) (.074) (.246) (.301) (.181) (.236)

Dividend payert .187 .079 .289 -.013 .469 .694(.210) (.083) (.161)∗ (.224) (.234)∗∗ (.288)∗∗

Cash flow/Total Assetst .097 .006 .062 .101 .025 .101(.109) (.026) (.076) (.120) (.052) (.067)

Const. .546 .541 -1.620 -2.029 -3.082 -2.780(.218)∗∗ (.111)∗∗∗ (.275)∗∗∗ (.258)∗∗∗ (.413)∗∗∗ (.371)∗∗∗

Obs. 788 802 760 776 716 730

R2 .082 .461 .504 .456 .574 .633

47

Table 12: Effect of inversions on the value of corporate cash holdings

This table reports the results of regressions investigating the impact of Corporate Inversions on the marginalvalue of cash holdings. The sample comprises all the Corporate Inversions and U.S. incorporated firms.Column (1) reports the estimates that include firm and year fixed effects along with standards errors clusteredat the firm level. Column (2) reports the estimates that include within-industry year fixed effects andindustry clustered standard errors. Column (3) presents the results where the standard errors are clusteredsimultaneously at the industry and year level while Column (4) reports the estimates from the cross-sectionalregression for each year in the data using the Fama-Macbeth procedure. In each column, we estimate theeffect of foreign incorporation on the marginal value of cash using a procedure similar to ?. Similar to ?,the marginal value for the average firm is the coefficient on the change in cash plus the sample average forall variables that are interacted with the change in cash times the respective regression coefficient from themodel. We estimate this regression on all the firm-year observations in our sample from 1996 to 2013. Allvariables are winsorized at the 1st and 99th percentile. All variables are defined in the Appendix 2. Forbrevity, we suppress the coefficients on the fixed effects.

(1) (2) (3) (4)

Inverted×∆Cash/Mkt.Cap -.454 -.679 -.591 -.308(.126)∗∗∗ (.229)∗∗∗ (.128)∗∗∗ (.147)∗∗

∆Cash/Mkt.Cap 1.310 1.142 1.124 1.024(.133)∗∗∗ (.060)∗∗∗ (.168)∗∗∗ (.109)∗∗∗

Inverted t .169 .006 -.017 -.009(.248) (.028) (.040) (.025)

∆Earningst .831 .925 .984 1.124(.172)∗∗∗ (.066)∗∗∗ (.193)∗∗∗ (.169)∗∗∗

∆Non cash assetst .130 .161 .136 .163(.048)∗∗∗ (.018)∗∗∗ (.050)∗∗∗ (.020)∗∗∗

∆Interestt -1.583 -2.390 -2.336 -2.647(.689)∗∗ (.471)∗∗∗ (.391)∗∗∗ (.508)∗∗∗

∆Dividendst 1.254 1.371 1.662 1.732(.673)∗ (.491)∗∗∗ (.694)∗∗ (.549)∗∗∗

(Cash/Mkt.Cap)t−1 .819 .230 .160 .065(.103)∗∗∗ (.032)∗∗∗ (.104) (.074)

(Cash/Mkt.Cap)t−1 ×∆Cash /Mkt. Capt -.524 -.575 -.599 -.916(.106)∗∗∗ (.104)∗∗∗ (.116)∗∗∗ (.184)∗∗∗

Leveraget -.236 -.135 -.094 -.113(.113)∗∗ (.027)∗∗∗ (.055)∗ (.055)∗∗

Leverage ×∆Cash /Mkt. Capt -.814 -1.022 -.978 -.872(.302)∗∗∗ (.191)∗∗∗ (.261)∗∗∗ (.210)∗∗∗

Const. -.185 -.072 -.076 -.057(.039)∗∗∗ (.009)∗∗∗ (.026)∗∗∗ (.032)∗

Obs. 13313 13313 13314 13174

R2 .38 .244 .162 .149

Marginal Value of Cash U.S. incorporated firms $1.16 $0.95 $0.94 $0.86

Marginal Value of Cash Corporate Inversions $0.70 $0.28 $0.35 $0.55

48

Table 13: Effect of the rule of law in the country of incorporation on the value of

corporate cash holdings

This table reports the results of regressions investigating the effect of the quality of rule of law in a countryon the marginal value of cash holdings. The sample comprises all the Corporate Inversions and U.S. incor-porated firms .Column (1) reports the estimates that include firm and year fixed effects along with standardserrors clustered at the firm level. Column (2) reports the estimates that include within-industry year fixedeffects and industry clustered standard errors. Column (3) presents the results where the standard errorsare clustered simultaneously at the industry and year level while Column (4) reports the estimates from thecross-sectional regression for each year in the data using the Fama-Macbeth procedure. In each column, weestimate the effect of rule of law of the parent’s country of incorporation on the marginal value of cash usinga procedure similar to ?. We estimate this regression on all the firm-year observations in our sample from1996 to 2013. All variables are defined in the Appendix 2. For brevity, we suppress the coefficients on thefixed effects.

(1) (2) (3) (4)

[100-Percentile rank ROL] × ∆Cash/Mkt.Cap -.040 -.056 -.047 -.014(.033) (.040) (.034) (.037)

∆Cash/Mkt.Cap 1.325 1.163 1.140 1.008(.076)∗∗∗ (.064)∗∗∗ (.169)∗∗∗ (.119)∗∗∗

[100-Percentile rank ROL]t .001 .003 .001 .0001(.005) (.003) (.005) (.004)

∆Earningst .831 .925 .985 1.129(.075)∗∗∗ (.066)∗∗∗ (.193)∗∗∗ (.172)∗∗∗

∆Non cash assetst .131 .161 .136 .163(.024)∗∗∗ (.018)∗∗∗ (.050)∗∗∗ (.020)∗∗∗

∆Interestt -1.585 -2.389 -2.336 -2.643(.648)∗∗ (.471)∗∗∗ (.393)∗∗∗ (.510)∗∗∗

∆Dividendst 1.245 1.381 1.671 1.693(.579)∗∗ (.491)∗∗∗ (.696)∗∗ (.550)∗∗∗

(Cash/Mkt.Cap)t−1 .820 .230 .160 .065(.061)∗∗∗ (.032)∗∗∗ (.104) (.075)

(Cash/Mkt.Cap)t−1 ×∆Cash /Mkt. Cap -.525 -.578 -.601 -.910(.126)∗∗∗ (.104)∗∗∗ (.118)∗∗∗ (.183)∗∗∗

Leveraget -.235 -.135 -.094 -.113(.077)∗∗∗ (.027)∗∗∗ (.055)∗ (.055)∗∗

Leverage ×∆Cash /Mkt. Cap -.809 -1.015 -.973 -.866(.241)∗∗∗ (.191)∗∗∗ (.260)∗∗∗ (.207)∗∗∗

Const. -.194 -.094 -.049 -.061(.055)∗∗∗ (.030)∗∗∗ (.032) (.036)∗

Obs. 13308 13307 13308 13168

R2 .38 .244 .162 .149

49

Table 14: Robustness: Caliper matching and Rosenbaum (2002) bound

This table presents the estimation of the difference in median between treated and control firms, when using caliper matching with 0.25 as threshold. The sample443 Corporate Inversions firm-year observations and a similar number of control firm-year observations. For each outcome variable, we include the covariatesincluded in the multivariate regressions. The second and third column reports the Hodge-Lehman point estimate of the difference in the median between treatedand control firm, and the p-value of such point estimate, respectively. lThe fourth column reports the covariates included in the matching. The last column reportsthe result from Rosenbaum (2002) sensitivity bound that measures the maximum impact an omitted/unobserved variable must exert to change the inferenceregarding the treatment effect. We estimate the difference in the median between treated and control firms on all the firm-year treated and control firms in oursample from 1996 to 2013. All variables are winsorized at the 1st and 99th percentile. All variables are defined in the Appendix 2.

Hodges-

Lehmann

median

difference

point

estimate

p-value HL

statistic

Covariates Γ

Outcome Variable

GAAP ETR −0.066∗∗∗ 0.001 Log(assets), Net Operating Loss, Advertising Expenditure, Capex,

Foreign Operations, Gross PPE, R&D expense and Intangible Assets.

1.7

Cash ETR −0.052∗∗∗ 0.001 1.6

Log(

CashTA

)0.102∗ 0.059 Lagged values of Log(Assets), ROA, Leverage, ,R&D expense,

Acquistions, and Net Working Capital

1.3

Spread 0.019 0.160

Leverage, Log(Assets), Capital Expeditures, Stock Return Volatility.

1.2

Share Turnover -0.154 0.367 1.1

Analyst Dispersion 0.035∗∗∗ 0.004 1.4

Institutional Ownership -0.006 0.387Log(Assets), Stock Return Volatility, Spread, Market-to-Book,

Leverage, Dividend Payer, and ROA.

1.1

Log(HHI Institutional Ownership) 0.062 0.148 1.2

Log(Average Institutional Ownership) 0.114∗ 0.055 1.3

50

Table 15: Effect of foreign incorporation on corporate outcomes by AFCs and inversions

This table reports the results of regressions similar to (1) that we implement in a sample that includes both inversions and AFCs. The sample also includesup to twice the number of control firm-year observations for every inversion and AFC firm-year observation identified by matching on industry, year, Log(TotalAssets), Market to Book, and ROA (sample CS1 ). To distinguish between inversions and AFCs, we estimate (1) after including four terms instead of Treated .These are Inversions, AFC and interaction terms Inversions × Treated and AFC × Treated . Inversions (AFC ) is a dummy variable that takes a value one forall inversion (AFC) firm-year observations and their respective control firm-year observations. Below we report just the coefficients on the interaction terms. Thesample extends from 1996-2013 and all variables are winsorized at the 1st and 99th percentile. All variables are defined in the Appendix 2. All regressions includestandard errors clustered at the firm level and year and ndustry fixed effects. For brevity, we suppress the coefficients on the fixed effects.

GAAP

ETR

Cash ETR Log(

CashTA

)Spread Turnover Analysts

Dispersion

Inst.

Ownership

Log(HHI

Inst.

Own.)

Log(Avg.

Inst. Own)

(1) (2) (3) (4) (5) (6) (7) (8) (9)

Inversions× Treated -.071 -.077 .130 .439 -.554 .113 -.063 .154 .114(.035)∗∗ (.025)∗∗∗ (.095) (.204)∗∗ (.634) (.064)∗ (.037)∗ (.092)∗ (.079)

AFC× Treated . -.048 -.022 .067 .177 1.038 .031 -.050 .240 -.017(.025)∗ (.021) (.082) (.126) (.773) (.045) (.024)∗∗ (.059)∗∗∗ (.082)

Obs. 3242 2890 2822 3167 3170 2519 2214 2214 2214

R2 .053 .063 .465 .58 .272 .238 .254 .598 .451

51

References

Abadie, A. and G. W. Imbens (2006). Large sample properties of matching estimators for average treatment

effects. Econometrica 74 (1), 235–267.

Abadie, A. and G. W. Imbens (2011). Bias-corrected matching estimators for average treatment effects.

Journal of Business & Economic Statistics 29 (1).

Andrade, G., S. Gilson, and T. Pulvino (2001). Seagate technology buyout. Case study, Harvard Business

School .

Bebchuk, L. and A. Cohen (2003). Firms decisions where to incorporate. Journal of Law and Economics 46,

383–425.

Bushee, B. J. and C. F. Noe (2000). Corporate disclosure practices, institutional investors, and stock return

volatility. Journal of Accounting Research, 171–202.

Chen, S., X. Chen, Q. Cheng, and T. Shevlin (2010). Are family firms more tax aggressive than non-family

firms? Journal of Financial Economics 95 (1), 41–61.

Chung, K. H. and H. Zhang (2011). Corporate governance and institutional ownership. Journal of Financial

and Quantitative Analysis 46 (01), 247–273.

Daines, R. (2001). Does delaware law improve firm value? Journal of Financial Economics 62, 525–558.

Desai, M. A., F. C. Foley, and J. R. Hines (2006). The demand for tax haven operations. Journal of Public

Economics 90, 513–531.

Desai, M. A. and J. R. Hines (2002). Expectations and expatriations: Tracing the causes and consequences

of corporate inversions. National Tax Journal 55 (3), 409–40.

Dharmapala, D. and J. R. Hines (2009). Which countries become tax havens? Journal of Public Eco-

nomics 93, 1058–1068.

Doidge, C., A. Karolyi, K. Lins, D. Miller, and R. M. Stulz (2009). Private benefits of control, ownership,

and the cross-listing decision. Journal of Finance 64, 425–466.

Doidge, C., A. Karolyi, and R. M. Stulz (2004). Why are foreign firms listed in the U.S. worth more?

Journal of Financial Economics 71, 205–238.

Doidge, C., A. Karolyi, and R. M. Stulz (2007). Why do countries matter so much for corporate governance?

Journal of Financial Economics 86, 1–39.

Dyreng, S. and B. P. Lindsey (2009). Using financial accounting data to examine the effect of foreign

operations located in tax havens and other countries on U.S. multinational firms tax rates. The Journal

of Accounting Research 47, 1283–1316.

52

Dyreng, S. D., M. Hanlon, and E. L. Maydew (2010). The effects of executives on corporate tax avoidance.

The Accounting Review 85 (4), 1163–1189.

Faulkender, M. and R. Wang (2006). Corporate financial policy and the value of cash. The Journal of

Finance 61 (4), 1957–1990.

Foley, C. F., P. Goldsmith-Pinkham, J. Greenstein, and E. Zwick (2014). Opting out of good governance.

Technical report, National Bureau of Economic Research.

Foley, C. F., J. Hartzell, S. Titman, and G. Twite (2007). Why do firms hold so much cash? A tax-based

explanation. Journal of Financial Economics 86, 579–607.

Graham, J. R. (2003). Taxes and corporate finance: A review. Review of Financial Studies 16, 1074–1128.

Harford, J., S. Mansi, and W. Maxwell (2008). Corporate governance and a firm’s cash holdings. Journal

of Financial Economics 87, 535–555.

Hines, J. R. and E. M. Rice (1994). Fiscal paradise: Foreign tax havens and american business. The

Quarterly Journal of Economics 109, 149–182.

Imbens, G. W. and D. B. Rubin (1997). Bayesian inference for causal effects in randomized experiments

with noncompliance. Annals of Statistics 25 (1), 305–327.

Kinsey, S. (2001). Foreign private issuers. Review of Securities and Commodities Regulation 34, 31–8.

Licht, A. N. (2003). Cross-listing and corporate governance: Bonding or avoiding? Chicago Journal of

International Law 4, 141–163.

Markle, K. S. and D. A. Shackelford (2012). Cross-country comparisons of corporate income taxes. National

Tax Journal 65, 493–527.

Opler, T., L. Pinkowitz, R. Stulz, and R. Williamson (1999). The determinants and implications of corporate

cash holdings. Journal of financial economics 52 (1), 3–46.

Petersen, M. A. (2009). Estimating standard errors in finance panel data sets: Comparing approaches.

Review of Financial Studies 22, 435–480.

Pinkowitz, L., R. M. Stultz, and R. Williamson (2006). Does the contribution of corporate cash holdings

and dividends to firm value depend on governance? A cross-country analysis. Journal of Finance 61,

2725–2751.

Rosenbaum, P. R. (2002). Observational studies. Springer.

Seida, J. A. and W. F. Wempe (2004). Effective tax rate changes and earnings stripping following corporate

inversion. National Tax Journal , 805–828.

53

Shnitser, N. (2010). A free pass for foreign firms? An assessment of SEC and private enforcement against

foreign issuers. Yale Law Journal 119, 1638–1701.

Siegel, J. I. (2005). Can foreign firms bond themselves effectively by renting U.S. securities laws? Journal

of Financial Economics 75, 319–359.

Subramanian, G. (2004). The disappearing Delaware effect. Journal of Law, Economics, and Organiza-

tion 20, 32–59.

Webber, S. (2011). Escaping the U.S. tax system: From corporate inversions to re-domiciling. Tax Notes

International 63 (4), 273–295.

Wooldridge, J. M. (Ed.) (2002). ”Econometric analysis of cross-section and panel data”. MIT Press.

54