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 Gopalan are from the Olin Business School, Washington University in St. Louis. The authors can be reached at [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.
55
Embed
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
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
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
∗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,
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:
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
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
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
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
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.
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:
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
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:
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.
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:
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.
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:
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)
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.
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.
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.
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.