Why Should I Believe You? The Costs and Treaties Andrew Kerner

advertisement
International Studies Quarterly (2009) 53, 73–102
Why Should I Believe You? The Costs and
Consequences of Bilateral Investment
Treaties
Andrew Kerner
Emory University
Bilateral Investment Treaties (BITs) are the primary legal mechanism
protecting foreign direct investment (FDI) around the world. BITs are
thought to encourage FDI by establishing a broad set of investor’s rights
and by allowing investors to sue a host state in an international tribunal
if these rights are violated. Perhaps surprisingly, the empirical literature
connecting BITs to FDI flows has produced conflicting results. Some
papers have found that BITs attract FDI, while others have found no
relationship or even that BITs repel FDI. I suggest in this paper that
these results stem from statistical models that do not fully capture the
causal mechanisms that link BITs to FDI. Extant literature has often
suggested that BITs may encourage investment from both protected
and unprotected investors, yet the literature has not allowed for a full
evaluation of this claim. This paper explores the theoretical underpinnings and empirical implications of an institution that works in these
direct and indirect ways, and offers a statistical test that is capable of
distinguishing between the two. The results indicate that: (1) BITs
attract significant amounts of investment; (2) BITs attract this investment from protected and unprotected investors; and (3) these results
are obscured by endogeneity unless corrected for in the statistical
model.
One of the most striking aspects of recent financial globalization has been the
proliferation of Bilateral Investment Treaties (BITs). BITs are interstate agreements designed to protect foreign direct investment (FDI) by establishing a
broad set of investors’ rights and, in most cases, by allowing investors to sue host
states in an international tribunal if these rights are violated. Over 2,000 BITs
have been ratified, including treaties ratified by every OECD country and the vast
majority of developing countries (UNCTAD 2005).1 In recent years BITs have
been justified and embraced on the grounds that, by protecting FDI, they
encourage capital flows into signatory countries (UNCTAD 1998). But do BITs
actually increase FDI inflows? If so, how is this achieved?
Author’s note: The author can be reached via email at akerner@emory.edu. The author would like to thank the
editors and three anonymous reviewers, as well as Terrence Chapman, Mark Hallerberg, Yoram Hoftel, Jeff Kucik,
Lisa Martin, Eric Reinhardt, Scott Wolford, Jason Yackee, seminar participants at Emory University, and participants
at the 2007 International Studies Association Conference for their comments and suggestions. A replication data
set and is available at http://isanet.ccit.arizona.edu/data_archive.html, and the ISQ Dataverse Network page at
http://dvn.iq.harvard.edu/dvn/.
1
The mean number of BITs in force per country was roughly 15 in 2005, which was up from roughly 6.5 in
1995 and 1.5 in 1985.
! 2009 International Studies Association
74
Bilateral Investment Treaties
Commonly used theories in international relations make different predictions
on this matter. Realist arguments, notably those of Downs, Rocke, and Barsoom
(1996), contend that international treaties are ‘‘shallow’’ in the sense that they
make formal commitments to actions that would take place regardless of treaty
obligations. Alternatively, institutionalist arguments suggest that international
agreements can be designed to alter future state behavior (e.g., Abbott and Snidal 2000; Keohane 1984; Simmons 2000). Naturally, and importantly for the purposes of this paper, these two theoretical paradigms make different predictions
for the effects of BITs on FDI inflows. The institutionalist paradigm suggests that
if BITs are effective, it is because they present significant ex post costs to signatory states that violate the agreement. It follows that we should look for evidence
of BITs’ effectiveness in the behavior of investors from signatory states. For realists, if BITs work at all, they work by identifying cooperative states. In order for
BITs to be effective identifiers, there needs to be significant ex ante costs to ratifying a BIT; otherwise states that were not committed to protecting FDI would
‘‘pool’’ on a strategy of signing and ratifying BITs. Under this view, the effects of
BITs should not be limited to investors that are specifically protected by the BIT.
Rather, any investor that observes a BIT being ratified will be able to update
their beliefs over a state’s intentions.
Extant empirical work does not clearly indicate whether BITs work according to
realist expectations, according to institutionalist expectations, or even whether
BITs work at all. Consistent with both realism and institutionalism, several studies
have found that signing more BITs correlates with greater aggregate FDI flows
(e.g., Büthe and Milner 2008; Neumayer and Spess 2005). On the other hand,
studies have generally been unable to show that signing a BIT with a specific
country correlates with more FDI from that country (e.g., Hallward-Driemeyer
2003; Tobin and Rose-Ackerman 2005). Taken together, these results present a
puzzle. Why do we find evidence that BITs encourage aggregate FDI flows, but fail
to find evidence of increased investment by investors who are actually protected?
This article is an attempt to address two holes in the literature that may
account for these otherwise curious results. One hole is an under-appreciation of
the different mechanisms through which BITs might impact FDI flows. While it
is generally accepted that BITs might work by ‘‘tying the hands’’ of ratifying host
states or by sending a broadly received ‘‘signal’’ that a country is trustworthy, the
extant literature does not fully consider the theoretical distinctions between
these hypotheses. Perhaps for this reason, extant empirical tests do not differentiate between the two mechanisms. The second hole in the literature is that it
generally fails to appreciate the endogeneity between BITs and FDI. Developing
countries are more likely to pursue BITs with developed countries when they
believe doing so will significantly increase FDI inflows. This is tantamount to saying that BITs are more likely to be ratified when a country’s FDI inflows are significantly below what they would be were it not for fears of expropriation. This
endogeneity can bias results if it is left un-addressed or if it is addressed insufficiently.2
This paper makes three innovations in its attempt to resolve these issues. First,
I explore some of the mechanisms through which BITs may be able to generate
ex ante and ex post costs to probe the applicability of institutionalist and realist
theories. I argue that there is ample evidence of ex post costs to violating BITs
and ex ante political costs that are borne by politicians who pursue them. Second, I employ a statistical model that is able to differentiate between these channels of influence. To do so I use a directed dyad data set to examine the effects
2
Much recent work in IPE has noted similarly endogenous relationships with other agreements. For an example using preferential trade agreements (PTAs) see Baier and Bergstrand (2007) and Magee (2003); for examples
using the IMF see Knight and Santaella (1997).
Andrew Kerner
75
of BITs, as well as BITs ratified outside of the dyad, on bilateral FDI flows.3 This
allows for an evaluation of whether FDI flows are correlated with being protected
by a BIT, observing that a host state has ratified numerous other BITs, or both.
Third, I correct for endogeneity by using a two-stage least squares estimator.
The results of the statistical models suggest that BITs increase FDI flows from
protected and unprotected investors, and that the former relationship is
obscured unless endogeneity is corrected. These results have several implications.
First, they contribute to a wider debate on the efficacy of international economic
institutions, particularly in the context of endogenous institutional choice. Second, the results of this paper comment on the nature of treaty compliance.
Downs, Rocke, and Barsoom (1996) have argued that international agreements
are entered into by countries that are most likely to observe the terms of the
treaty. The results of this paper indicate the opposite. The countries that are
most likely to ratify BITs are the ones that, in the eyes of investors, are most in
need of the treaty’s assurances. Third, the results of this paper point to an
under-recognized irony of globalization. By actively opposing international investment agreements, actors opposed to BTIs increase the ex ante costs of ratifying
BITs, making BITs more credible and, perhaps, more attractive to governments
in need of FDI.
Background Information on BITs
Foreign direct investment is an investment by foreign nationals that is large
enough to grant the investor a significant amount of corporate control (United
States Department of Commerce, Bureau of Economic Analysis 1998). Unlike
portfolio investment, which can easily be pulled out and reinvested elsewhere,
FDI ‘‘has long time horizons and is generally not done for speculative purposes,
but rather to serve domestic markets, exploit natural resources, or provide platforms to serve world markets through exports’’ (Jensen 2003, 588). Countries
seek to attract FDI for a variety of reasons. First and foremost, FDI, similar to
other forms of investment, is a source of capital and jobs. Second, the typically
knowledge-intensive nature of FDI creates ‘‘knowledge spillovers.’’ These spillovers can arise through firm-to-firm emulation or through the sharing of employees, who bring sophisticated production techniques from foreign-owned firms to
domestic firms. Similarly, successful multinationals are often subject to an ‘‘obsolescing bargain,’’ whereby the skills and resources developed over time in the
host state lead to greater negotiating leverage (Vernon 1971). Third, FDI is often
concentrated in exporting industries and can generate foreign exchange. Not
surprisingly, many developing countries have made the promotion of FDI a key
part of their development strategies.4
Despite the priority that governments place on attracting FDI, these same governments often present political risks that prevent investment from taking place.
The need to protect FDI from host governments arises from time inconsistency
problems (Simmons 2000). Foreign investors favor host countries that will not
expropriate from them in the future (Büthe and Milner 2008; Henisz 2000; Jensen 2003, 2006; Li and Resnick 2003; Neumayer and Spess 2005). Prior to an
investment, potential host states have strong incentives to assure investors that
they will not expropriate; however, because FDI typically involves high-sunk costs,
the state may decide ex post to forgo its earlier commitments, knowing the investor will have little legal or practical recourse (Hallward-Driemeyer 2003). Absent
3
A similar research design can be found in Salacuse and Sullivan (2005).
In a 2000 survey of FDI related policies, UNCTAD noted that nearly all surveyed countries reported using various tax incentives to attract investment (UNCTAD 2000).
4
76
Bilateral Investment Treaties
some remedy, the investor is left insecure and may be less likely to invest as a
result.
Bilateral Investment Treaties are designed to reduce the risk of state-led
expropriation.5 BITs define a minimum standard of behavior toward foreign
investment. These protections generally include: (1) equal and fair treatment
in accordance to that received by any domestic or third-party firm; (2) protection against arbitrary or discriminatory polices; (3) flexibility with respect to
staffing; (4) protection against performance requirements of any kind; and,
most importantly, (5) ‘‘that investments shall not be expropriated or nationalized either directly or indirectly through measures tantamount to expropriation or nationalization except for a public purpose; in a non-discriminatory
manner; upon payment of prompt, adequate and effective compensation; and
in accordance with due process of law’’ (Article 6 of the United States Model
BIT).6 In the event of a perceived violation, an aggrieved investor (or their
home state) typically has the right to adjudicate his or her allegation in an
international tribunal, most often to the International Centre for Settlement
of Investment Disputes (ICSID).7 The resulting awards are binding and failure
to accommodate a ruling can lead to retaliation by the aggrieved party or
third-party actors.
Why Might BITs Attract Investment?
If BITs attract FDI, it is because they generate a credible expectation that a host
state will not expropriate. But how is this credibility generated? Fearon (1997)
suggests that credible commitments can broadly be seen as serving either
‘‘hands-tying’’ or ‘‘cost-sinking’’ functions. A hands-tying mechanism works by
creating costs that an actor will suffer ex post if they do not follow through on a
commitment. A cost-sinking mechanism attaches irretrievable ex ante costs to an
action such that it separates sincere and insincere actors. It is not clear a priori
where BITs fit within these distinctions. Extant work suggests protected investors
are motivated to invest by BITs’ capacities to shape future behaviors, but also
suggests that unprotected investors are motivated to invest by BITs’ capacities to
signal states’ intentions. Despite these assertions, the literature has not fully
established the plausibility of these mechanisms theoretically or empirically. In
the following two sections I argue that either of these possibilities is at least plausible, if not likely.
The Direct Effect: BITs as Hands-Tying Devices
Institutionalist authors have argued that treaties can be designed to induce compliance. An effective treaty is costly enough to shirk that the commitments made
in the treaty are necessarily credible. BITs could function as an effective handstying mechanism if, and only if, the ex post cost of expropriating (and not paying the resulting award) is high enough that countries will refrain from doing
so. Elkins, Guzman, and Simmons (2006) suggest several ways that BITs tie the
hands of host states through ex post costs. First, BITs’ dispute settlement procedures make the expectations of the host state explicit: any action deemed expropriation by the adjudicating body must be compensated with the prescribed
award. This disallows states from claiming that no expropriation has taken place
5
For a review of FDI related policies before the BIT, see Elkins, Guzman, and Simmons (2006).
The US Model BIT is available at: http://www.ustr.gov/assets/Trade_Sectors/Investment/Model_BIT/
asset_upload_file847_6897.pdf.
7
Of the 219 cases that are reported as of November 2005, 132 of them were arbitrated through ICSID, 65
through UNCITRAL, and the remainder through various standing and ad hoc bodies (UNCTAD 2005).
6
Andrew Kerner
77
when it has, or that proper compensation has been rendered when it has not.
Second, by involving the home state of the investor in a dispute, BITs raise the
stakes of non-compliance by potentially tying it to diplomatic relations. Home
states have an interest in having their treaties honored, and host states have an
interest in maintaining good relations with their trade and investment partners.
Third, the consequences of not abiding by the decision of a ‘‘neutral authoritative third party with which it has voluntarily pre-committed to comply’’ (Elkins,
Guzman, and Simmons 2006, 15) can be severe. These potential consequences
include: (1) reputational costs in the eyes of other countries and investors; (2)
the possibility of diminished credit rating8; and (3) the (as yet never exercised)
possibility of being referred to the International Court of Justice for non-compliance with an ICSID ruling (International Centre for Settlement of Investment
Disputes 2006).
If BITs truly serve as a hands-tying mechanism, it should be evident in
cases in which compliance can only be explained by a state’s treaty obligations. Furthermore, any such evidence would have to show that it was the
BIT, and not related or accompanying changes in domestic law, that led a
country to protect FDI or to comply with unfavorable rulings stemming from
an investment dispute. Unfortunately, we currently lack any comprehensive
data on compliance rates. Complicating matters, the decisions rendered in
these cases are often kept secret (Peterson 2003). In substitute I rely on anecdotal evidence from Lauder v. The Czech Republic. Except where otherwise
cited, the following borrows generously from Desai and Moel (2008).
In 1992, the Czech Republic began issuing broadcasting licenses for its newly
privatized television airways. One such license was granted to a joint venture
between Central European Television for the 21st Century (CET 21) and Central
European Media Enterprises (CME). CET 21 was led by a group of prominent
Czechs, including businessman Vladimir Železný, and was meant to bring local
knowledge and political connectivity to the partnership. CME, which provided
much of the capital for the project, was founded by Ronald Lauder in order to
acquire and operate a portfolio of broadcast licenses that became available during the privatizations of the 1990s. Having both the capital and the license, this
partnership began broadcasting TV Nova in the Czech Republic, with considerable success.
The original ownership structure called for the license to be issued to CET 21,
in which CME was to acquire a 49 percent share. The notion of such extensive
foreign ownership of the broadcasting license created a political backlash, and
so the Czech Media Council, along with CET 21 and CME, agreed to issue the
license to the Czech-owned CET 21, and allowed CET 21 and CME to create a
new company—Ceska Nezavisla Televizni Spolecnost (Czech Independent Television Company, or CNTS)—that would operate the license. Originally, CME held
a 66 percent share in CNTS, with CET 21 holding a 12 percent share and Czech
Savings Bank (CSB) holding a 22 percent share. By 1996, however, CME bought
out CSB to take an 88 percent stake in CNTS. In May 1997, CNTS altered its
agreement with CET 21 so that CET 21 was no longer granted exclusive control
of the broadcasting license. In 1997 and 1998, Vladimir Železný bought out the
remaining shareholders of CET 21 and then sold most of those shares to CME,
under the agreement that Železný would not compete with CNTS, leaving CME
with a 99 percent stake in CNTS. CET 21, which at this point was owned entirely
by Železný, held the remaining 1 percent.
8
Omar E. Garcı́a-Bolı́var, a member of the panel of arbiters of ICSID, offers that failure to adhere to the
awards stipulated by an ICSID ruling ‘‘may also decrease a country’s credit rating especially regarding funds from
multilateral organizations such as the World Bank, with which ICSID is affiliated and diminish the value of its commercial instruments such as government bonds’’ (Garcı́a-Bolı́var and Schmid 2004).
78
Bilateral Investment Treaties
In 1999, CME entered merger negotiations with Scandanavian Broadcasting
Services (SBS), a company with holdings in the Scandinavian broadcasting market and elsewhere in Europe. Železný, perhaps sensing an opportunity,
demanded a renegotiation of existing agreements between CNTS and CET 21,
arguing that CNTS’ exclusive right to broadcast on CET 21’s license ran counter
to the Media Council’s intention. Železný, whose control over TV Nova made
him a household name and political player in the Czech Republic, refused buyout offers by CME of over $100 million. Instead Železný withdrew CET 21’s
broadcasting license from CNTS and began operating the license using programming made available through Železný’s other media holdings. Lacking access to
a broadcasting license, CNTS went out of business. CME impending merger with
SBS was scrapped and CME’s stock tumbled.
Lauder and CME objected that Železný had broken his contract and appealed
their case to the media council and to the Czech courts, neither of which upheld
CME’s contracts with CET 21 or Železný. They also tried alternative routes to
pressure the Czech government. In November 1999 Lauder took out full page
ads in The Washington Post and The New York Times that were timed to coincide
with a visit by the Czech Prime Minister to Washington, urging investors to
‘‘think twice’’ before investing in the Czech Republic. Lauder also persuaded
Czech-born Secretary of State Madeline Albright to plead his case with the Czech
Prime Minister during that visit. Neither bad press nor diplomatic intervention
was enough to compel cooperation by the Czechs. Commenting on the situation,
the head of the American Chamber of Commerce in Prague noted that ‘‘No
Czech politicians who has any desire to continue being a politician is going to
cross swords with Železný’’ (Desai and Moel 2008, 235).
Having exhausted the aforementioned remedies, Lauder filed two cases in
international tribunals relating to the Czech Republic’s BIT obligations.9 In the
first case Lauder sued the Czech Republic personally, claiming that the government’s failure to uphold his contracts violated the Czech Republic-United States
1991 BIT. The London-based tribunal hearing the case found that the government of the Czech Republic ‘‘did not take any measure of, or tantamount to,
expropriation of the claimant’s property rights’’ (Ronald S. Lauder v. the Czech
Republic, UNCITRAL Arbitration Proceedings Final Award 2001, 43).10 The second case was filed by CME as a corporate entity claiming that CME’s treatment
violated the 1991 Netherlands-Czech Republic BIT. CME sought damages of
$556 million and was awarded $353 million, an amount roughly equal to the
annual budget of the Czech Ministry of Health. Outraged at the conflicting decisions and the size of the award, the Czech Government appealed (in vain) and
eventually sought a thorough review of its treaty obligations (Bouc 2005). Despite
their obvious dismay with the decision, the force of the treaty was enough to
compel the Czechs to pay because, as then Foreign Minister Cyril Svoboda said,
‘‘prompt payment is a must in order to safeguard the nation’s reputation
abroad’’ (Desai and Moel 2008, 239).
The Lauder v. Czech Republic case highlights BITs’ abilities to protect investors
above and beyond what can be achieved with domestic law. The politically influential Železný (who in 2002 was elected to the Czech Senate) was able to use his
influence to prevent Lauder and CME from having sufficient recourse to a
domestic remedy. Furthermore, this anecdote demonstrates that BITs can influence state behavior beyond what can be achieved through publicity alone. The
use of the international tribunal was able to prompt payment from the Czechs
even after the facts of the dispute had already become public knowledge. Recall
9
Lauder also filed a case against Železný at the International Chamber of Commerce that was unrelated to
any BIT.
10
The final award in Lauder v. Czech Republic is available at: http://ita.law.uvic.ca/documents/LauderAward.pdf.
Andrew Kerner
79
that Lauder placed his anti-Czech Republic advertisements in The New York Times
and The Washington Post in November 1999, almost 4 years before the Czechs
paid their fine.
Despite the impact that BITs can have (and did have in the preceding case),
they do have at least one clear limitation: only protected investors can bring BITrelated suits. Unprotected foreign investors can be expropriated from without
triggering any of the ex post costs that are critical to BITs’ credibility. It should
follow that if BITs only work through ex post costs, they should only cause an
increase in investment from protected investors. From this we can deduce a first
hypothesis.
Hypothesis 1: Bilateral Investments Treaties work by amassing ex post costs into an
effective hands tying mechanism. They should encourage investment from investors that are
protected by BITs.
The Indirect Effect: BITs as Sunk Costs
Political economists commonly assert that BITs constitute a globally observed signal that a country is serious about protecting foreign investment (Neumayer and
Spess 2005, 1571; Tobin and Rose-Ackerman 2005, 14). The rationale typically
given is that third parties can observe that a host state has tied its hands with
respect to other investors and infer that their investments will be safe as well.
But why should BITs have this effect? The logic of hands tying implies that, without the BIT, the host state cannot fully be trusted not to expropriate. Unprotected investors do not have recourse to an international tribunal. The host state
does not necessarily have its hands tied in any way with respect to how it treats
them.11
But we need not dismiss this hypothesis. If BITs have a global effect, then the
agreements must be credible for reasons beyond the ex post costs of violation.
BITs can also present enough ex ante costs, or, to use Fearon’s term, sunk costs,
that ratifying a BIT credibly signals that a state is predisposed against expropriating from foreign investors. Any investor can observe the signal sent by a ratified
BIT, regardless of whether they are protected. To the extent that ex ante costs
effectively convey credibility, any investor should be more willing to invest in a
state that signs and ratifies BITs.
In what follows I will argue that there are, in fact, good reasons to suspect the
BITs can create enough ex ante costs to send a credible signal. The main drivers
behind these ex ante costs are twofold. First, BITs often limit a host state’s regulators and lawmakers. All manner of policy has been challenged by BIT-related
suits and these limitations can be politically costly for the politicians that pursue
BITs. The South African policy of post-apartheid property reallocation has been
challenged by foreign mining companies under the terms of BITs (Peterson
2004). Environmental laws are particularly vulnerable. A Spanish company, Tecnicas Medioambientales Tecmed S.A., successfully challenged the Mexican government
after its hazardous waste confinement plant had its permit revoked out of environmental concerns (Peterson 2004). The mere threat of litigation can also be
enough to affect national policy. The Canadian government has found that,
because of its BIT-related obligations, it may not be able to extend publicly provided health care into services currently provided by foreign-based companies
11
There are, of course, other mechanisms for protecting FDI besides BITs or BIT equivalents embedded in
free trade agreements. To the extent that these are located in domestic law, however, they lack the hands tying
capacity of a BIT (Fearon 1997).
80
Bilateral Investment Treaties
(Canadian Centre for Policy Alternatives Consortium on Globalization and
Health 2002).
Second, BITs put domestic and preexisting foreign investors at a disadvantage.
Domestic and unprotected foreign investors must face a legal system that is often
slower, more capricious, and less investor-friendly than their protected foreign
competition. In the South African mining case mentioned above, Peterson
(2004) notes that BIT-protected, foreign-controlled mines could be granted
awards by an international tribunal, while domestically controlled mines faced a
legal system much more likely to accept the legality of South African policy. Foreign and domestic firms that have invested in knowledge of the local legal system
will see that competitive advantage disappear as they compete with firms that can
operate outside of it. To the extent that BITs attract new investors, domestic and
foreign investors that operated prior to a BIT may find themselves facing greater
competition consequences.
Despite their potential to generate opposition from public and private interest
groups, BITs held a low profile for much of their history. This is not to say that
they were costless, but neither did they present serious political obstacles.12
Today, and increasingly over the past decade, international investment agreements hold a higher profile position in the debate on globalization and politicians must account for this when entering into these treaties. I attribute much of
this change in prominence to the dramatic failure of the negotiations over the
Multilateral Agreement on Investment (MAI), after which BITs’ profile—and the
associated political costs—rose dramatically.
The MAI was an attempt by developed countries to establish a unified global framework to protect FDI. Following the failure to reach a multilateral
agreement during the Uruguay round of the WTO, developed countries felt
that a more expeditious strategy would be to limit negotiations to themselves.
Major provisions of the MAI called for an expansive definition of investment,
strict limitations on performance requirements such as local content requirements or mandated technological transfer, a prohibition on expropriation or
any tantamount action without public purpose and prompt compensation, free
transfer and repatriation of assets, and an international dispute settlement
mechanism to enforce compliance with the MAI, to be conducted in accordance with, if not actually in, the ICSID or UNCITRAL (Kobrin 1998; Walter
2001). In many ways the draft agreement they developed resembled a multilateral BIT.
In February of 1997, Public Citizen, an advocacy group founded by Ralph
Nader, obtained a draft copy of the MAI and posted it on its Web site. News of
the proposed agreement sparked controversy among increasingly internet-savvy
political activists. The MAI, if it had succeeded, would have had dramatic consequences for a variety of actors in the developed world and the developing world.
In the following months over 600 organizations stretching across 70 countries
voiced their opposition to the MAI, including such wide ranging groups as the
World Development Movement, the AFL-CIO, the Western Governors Association, and Oxfam (Kobrin 1998). In April 1998 negotiations over the MAI were
suspended for 6 months so that negotiators could consult with ‘‘interested parts
of their societies’’ (Kobrin 1998, 98). In November of that year, France, followed
soon after by Australia and Canada, pulled out of negotiations altogether, leading to the MAI’s collapse. The French government specifically cited political
pressure, arguing that ‘‘...more than any other international agreement of an
economic nature, the MAI has raised objections and tensions at the heart of civil
12
Perhaps indicative of these costs, many legislatures in developing countries refused to ratify the BITs that
their executives have signed, well before BITs became a popularly recognized political issue. Brazil, for example,
signed 14 BITs during the mid-nineties, none of which have been ratified.
Andrew Kerner
81
society. The opposition to it was surprising in its scale, strength, and the speed
with which it appeared and developed’’ (Lalumière and Landau 1998).13
Anti-globalization activists ‘‘drew blood’’ by helping to sink the MAI (de Jonquières 1998). In the period since there has been a greater awareness of investment agreements, including BITs, and more assertive action against them by
concerned parts of civil society. Often these groups have forced governments to
negotiate less stringent investment protections and occasionally they have been
able to scuttle BIT negotiations altogether. The ill-fated BIT between South
Korea and the United States stands as a particularly striking example of how
effective the opposition to a BIT can be.
Negotiations for a U.S.-South Korea BIT began in 1998. In many ways, it should
have been an easy BIT to pass. In 1998 U.S.-sourced FDI in South Korea totaled
$7.3 billion, amounting to 38.7 percent of Korean FDI stock (Choi and Schott
2004). South Korean President Kim Dae-Jung made a BIT with the United States
(as well as a BIT with Japan) a cornerstone of his economic policy. The following
Korean President, Roh Moo Hyun, appointed Jeffery D. Jones, five-time President
of the American Chamber of Commerce in Korea and a vocal supporter of the
U.S.-Korean BIT, as a member of the Regulatory Reform Commission.
Despite this, the U.S.-Korea BIT proved to be elusive. The root of the difficulty
was Korean resistance to the BIT’s impact on Korea’s screen quota for domestically
made films. The Korean screen quota was initially installed in 1967 to help promote the domestic film industry against foreign (principally American) competition. At the time that negotiations began, the screen quota required Korean movie
theaters to show domestically produced films for 146 days out of the year. Faced
with the prospect of losing this protection, the Korean film industry, in concert
with organizations such as the Coalition for Cultural Diversity in Moving Pictures and
Korean People’s Action against BITs and the WTO (now called Korean People’s Action
against FTAs and the WTO) organized a campaign against the abandonment of the
screen quota and against the U.S.-Korea BIT more generally.14 The political resistance to the BIT was so great that public officials refrained from even mentioning
the screen quota publicly for fear of creating a backlash (Kim 2006).
The campaign against the U.S.-Korean BIT hobbled government initiatives to
sign a major international treaty with its primary foreign investor.15 The activation of public resistance is not limited to Korea. In another recent example, public pressure forced the Australian government to dilute the investment chapter
of the Australia-U.S. FTA (Capling and Nossal 2004). More dramatically, Evo
Morales was able to unseat the ruling party in the 2005 Bolivian presidential election, in large part by harnessing public discontent over the BIT-protected rights
of foreign firms operating in Bolivia. The Cochabamba ‘‘Water Wars’’ of 2000,
which helped propel then congressman Morales into the national spotlight is
perhaps the most significant expression of these sentiments.16 Morales, along
13
The reasons for the MAI’s failure were not likely due solely to the work of anti-MAI activists, though these
activists undoubtedly played a significant role. See Graham (2000) for a more complete evaluation of the topic.
14
The Web site for the Coalition for Cultural Diversity in Moving Pictures, http://www.screenquota.com,
includes letters solicited and then sent to President Roh from around the world in support of the screen quota and
against the U.S. Korea BIT and liberal economic policy more generally. Interestingly, as of November 19, 2007 you
can also download a copy of Hallward-Driemeyer (2003) from their Web site, available at http://www.screenquota.com/home2/news/research_list.asp.
15
Temporarily, anyway. While the U.S.-Korea BIT was shelved, it would eventually be adopted as part of the
U.S.-Korea FTA, including a requirement that screen quotas be reduced to 73 days a year.
16
The Cochabamba water crisis arose after a Bechtel subsidiary, Aguas del Tunari, was granted control over the
water distribution network in Cochabamba, Bolivia. Following a rate hike and suspension of service for non-paying
customers, large scale protests broke out in Cochabamba and, later, nationally. The Bolivian government was forced
to revoke its contract with Aguas del Tunari and handed control of water distribution to a consortium of domestic
groups. Aguas del Tunari filed a suit against Bolivia for $40 million under the Netherlands-Bolivia BIT, which it
later withdrew.
82
Bilateral Investment Treaties
with President Correa in Ecuador, have continued to tap into this discontent,
most recently by pulling their countries out of ICSID entirely. As the ramifications of BITs continue to become better appreciated by civil society, BITs have
been given a more prominent place in the anti-globalization movement. Opposition to BITs has become a cause celebre in and of itself.17 Investors all over the
world can observe host states taking on the costs of this opposition.
Of course, BITs are not the only mechanism for protecting the rights of foreign investors. States can and do alter the legal status and obligations of foreign
investors through changes in domestic law. However, there are several reasons
why BITs are particularly well positioned to generate these ex ante costs. First,
the infringement on national sovereignty that is implied by BITs’ use of international courts has particular resonance with, and draws the ire of, anti-globalization activists. Second, in order to be effective, a signal must be observed. BITs,
by taking the form of an international treaty, should be more visible than equivalent changes in domestic law. If these costs are substantial enough and visible
enough to make a BIT credible, the effect that it has should not be limited to
protected investors. From this we can deduce a second hypothesis:
Hypothesis 2: Bilateral Investment Treaties work by marshaling ex ante costs into a
widely received, credible signal that a country will not expropriate from foreign investors.
They should encourage investment regardless of whether or not the investors are protected by
the treaty.
Whether BITs sink costs or tie hands, they involve significant costs that imply
an endogenous relationship between FDI and BITs. Host states have the greatest
incentives to pursue BITs when their expected gain in FDI is highest. Put
another way, the more FDI a country receives without the use of a BIT, the less
motivation there is for that country to sign and ratify them and vice versa.18
Because this relationship implies that BITs should be negatively correlated with a
shock to FDI, failure to account for this should lead to a systematic underreporting of BITs’ effectiveness. Moreover, this endogeneity should be particularly troublesome in dyadic research designs. Following similar logic as above, countries
should be most likely to enter BITs with a specific country when that BIT’s
expected impact on FDI flows is highest.
Hypothesis 3: Correcting for endogeneity should reveal a stronger relationship between
FDI and BITs than would otherwise be evident.
Research Design
Sample and Dependant Variable
To test my hypotheses, I construct a data set with one observation per
ordered pair (‘‘directed dyad’’) of countries per year. The coverage of countries and years is constrained by the availability of data taken from the
OECD’s International Direct Investment by Country-Volume 2005-Release 1.19
This data set measures the size of FDI flows in millions of U.S. dollars (which
17
In 2004 several activist groups including Asia Pacific Research Network, GRAIN, GATT Watchdog, Global Justice Ecology Project, IBON Foundation, XminY Solidariteitsfonds collaboratively launched bilaterals.org, a Web site
solely dedicated to opposing BITs and similar agreements.
18
This expectation is consistent with Elkins, Guzman, and Simmon’s (2006) observation that less reliable governments sign and ratify BITs more often than more reliable governments.
19
Those familiar with this data set may note that there is a significant amount of missing data. This is dealt
with in the robustness checks section.
Andrew Kerner
83
I deflate into constant 2000 dollars) from OECD countries into a broad cross
section of developing countries. FDI flows often fluctuate greatly, taking on
very large values some years, zero other years. In order to account for this
distribution I use the natural log of FDI flows as my dependent variable. This
is made somewhat problematic by the prevalence of zeros and negative values,
neither of which have a log.20 To get around this, I adopt the technique used
by Blonigen and Davies (2004). To accommodate the zeros, this strategy calls
for adding one dollar to the value of FDI flows, changing observations that
would have called for taking the log of zero (which does not exist) to the
log of 1, which is 0. To accommodate negative values, Blonigen and Davies
suggest logging the absolute value of FDI flows, and then reintroducing
the original sign after the transformation. Thus my dependent variable is
given by:
DV ¼ ð#Þ log jðFDI % 1; 000; 000Þ þ 1j
There is some controversy in the literature on whether using the logged
values is preferable to the unlogged values of FDI flows. Yackee (2007) demonstrates that Neumayer and Spess’ (2005) results using a logged dependent
variable are not robust to using the unlogged form of the dependent variable.
The statistical findings I report are robust to either functional form. I report
my baseline models (models 1 and 2, in Table 1) with both the logged and
unlogged dependent variables.21 Otherwise, I report results using the logged
values of FDI flows.
My sample of source countries includes the original OECD countries.22 plus
Japan, Australia, and New Zealand. Late-joiners to the OECD such as Mexico,
Korea, and the Central European countries are not included as home countries.
I also exclude Turkey from the sample of home countries, despite its being an
original member of the OECD.23 The sample of host countries includes all countries not included as home countries. The 127 developing countries that are
included as host countries in the main model are listed in Appendix A.24 All
models cover the years 1982–2001.
Independent Variables
My model has two primary independent variables. The first, BITijt, is a dichotomous variable coded 1 if country pair ij is protected by a ratified BIT in year t
and 0 if otherwise (taken from UNCTAD’s Country-specific Lists of BITs).25 The
second independent variable that I use, Other BITsit is a moving average of the
number of new BITs that a host country i has ratified with OECD countries outside of the dyad in years t ) 1, t, and t + 1. I operationalize the variable in this
way for a variety of reasons. In its level form, a simple count of extra-dyadic BITs
contains a unit root and thus cannot be used for statistical analysis (Enders
2004). Unit roots are typically addressed by first differencing the variable containing the unit root. In this case, however, a first differenced variable would not
20
Each data point represents the balance of capital moving into a country and received earnings or other capital moving back to the home country. The dependent variable takes on negative values in years where more capital
was repatriated than sent into a host nation. Thus, the difference between negative and positive values of the
dependent variable is one of scale, not of type, and we cannot dismiss these observations in statistical analysis.
21
The unlogged dependent variable is expressed in millions of dollars.
22
The original twenty members of the OECD are Austria, Belgium, Canada, Denmark, France, Germany,
Greece, Iceland, Ireland, Italy, Luxembourg, the Netherlands, Norway, Portugal, Spain, Sweden, Switzerland,
Turkey, Great Britain, and the United States.
23
The decision to exclude or include these countries as home states turns out to be inconsequential in most
models.
24
The 127 figure refers to the sample used in model 1–7. Other models have slightly different samples.
25
Available at http://www.unctad.org/Templates/Page.asp?intItemID=2344&lang=1.
Log Source
GBP (lag)
log GBP (lag)
PTA
Trade (lag)
Savings (lag)
· Other BITs
Polcon
Savings (lag)
1997 · Other BITs
1997
Democracy · Other BITs
Other BITs
(3 year avg of
new BITs)
Democracy
BIT
FBI (lag)
Variable
)136.098**
(56.638)
0.024
(0.276)
185.795***
(66.801)
1136.020***
(275.890)
153.749***
(47.124)
0.692
(1.342)
)0.004
(0.010)
0.934
(1.115)
9.334*
(5.168)
)1.050
(1.203)
0.052**
(0.024)
0.008
(0.010)
0.912***
(0.341)
)1.433***
(0.543)
65.180***
(0.411)
620.144**
(274.635)
250.539***
(0.074)
)0.008
(0.016)
0.288
(0.615)
0.413**
(0.200)
0.402***
(0.073)
)22.364
(26.992)
9.627
(12.699)
)371.988***
(100.887)
1.077**
(0.440)
)139.767
(141.378)
1334.151***
(293.640)
64.809
(60.776)
)0.062
(0.891)
1.475***
(0.420)
Model 2a
2SLS
(Std. Err.)
Model 1b
OLS
(Std. Err.)
Model 1a
OLS
(Std. Err.)
)6.847***
(2.496)
0.031**
(0.015)
)8.769***
(3.154)
16.917***
(6.325)
)4.268**
(1.746)
0.096***
(0.033)
0.027**
(0.012)
)0.032*
(0.019)
17.061**
(6.821)
8.565***
(2.026)
Model 2b
2SLS
(Std. Err.)
TABLE 1. Effects of BITs on FBI
)6.167***
(2.289)
0.029**
(0.015)
)7.396***
(2.674)
17.311***
(6.244)
)4.791***
(1.815)
0.105***
(0.034)
0.001
(0.016)
0.049*
(0.029)
)0.029
(0.019)
14.582**
(5.921)
4.618***
(1.590)
Model 3
2SLS
(Std. Err.)
)5.272**
(2.549)
0.029*
(0.017)
)17.524***
(6.330)
11.710
(7.255)
)7.268***
(2.651)
)18.386***
(6.398)
17.603***
(5.964)
0.163***
(0.048)
0.016
(0.014)
)0.049**
(0.024)
13.886*
(7.889)
5.229***
(1.703)
Model 4
2SLS
(Std. Err.)
)0.069
(0.098)
0.362*
(0.219)
)7.600***
(2.792)
0.054**
(0.024)
)8.509***
(3.254)
15.873**
(6.617)
)4.044**
(1.872)
0.028**
(0.013)
)0.034*
(0.020)
18.887**
(7.718)
1.931
(3.475)
Model 5
2SLS
(Std. Err.)
84
Bilateral Investment Treaties
21.547***
(4.928)
4.719
(4.059)
7691
17.72
Significance levels : *: 10% **: 5% ***: 1%.
N
F Statistic
first stage F BIT
first stage F Other
BIT
First stage interaction
term
Hansen J p-value
CAOI (lag)
World Growth Rate
Variable
Model 1a
OLS
(Std. Err.)
0.080
(0.135)
)0.117
(0.108)
7691
4.890
Model 1b
OLS
(Std. Err.)
0.101
(0.162)
)0.784***
(0.211)
7691
4.660
26.380
48.090
0.384
0.720
Model 2b
2SLS
(Std. Err.)
21.452***
(5.627)
)15.144**
(7.352)
7691
13.140
26.440
49.080
Model 2a
2SLS
(Std. Err.)
Table 1. Continued
0.230
36.990
0.138
(0.159)
)0.723***
(0.192)
7691
4.890
24.720
47.110
Model 3
2SLS
(Std. Err.)
0.758
9.920
0.160
(0.201)
)0.683***
(0.230)
7691
3.350
18.860
57.190
Model 4
2SLS
(Std. Err.)
0.585
25.820
0.079
(0.171)
)0.795***
(0.222)
7691
4.090
20.340
40.460
Model 5
2SLS
(Std. Err.)
Andrew Kerner
85
86
Bilateral Investment Treaties
capture the underlying concept. If a country ratifies several BITs in 1 year, but
for some reason fails to ratify any in a given year, it is unlikely that investors
would only look at the lack of BITs ratified in that year and not recognize the
larger trends. Using a 3-year moving average avoids both pitfalls. Choosing a
length of 3 years for the moving average is arbitrary, but the results do not
change substantially if I use a 5-year moving average or even a simple differenced
variable. I limit this variable to include only extra-dyadic BITs ratified with OECD
countries because so called ‘‘strange BITs’’—BITs signed between poor, often
non-contiguous states, neither of which is a significant source of FDI—are unlikely to lead to increased FDI and thus should not generate the sorts of costs that
motivate the theory of this paper (Elkins, Guzman, and Simmons 2006). If
Hypothesis 1 is correct, FDI flows should be responsive to whether a specific pair
of countries have ratified a BIT, even when controlling for how many other
countries with which the host state has concluded a BIT. If Hypothesis 2 is correct, FDI should be responsive to the amount of BITs ratified with other states,
even when controlling for whether or not a specific dyad has ratified a BIT.
Control Variables
My control variables include several macroeconomic measures taken from the
World Bank Development Indicators.26 I include the log of the source country’s
GDP, Log Source GDPjt, because I suspect that home countries with higher GDPs
send out more FDI and, in anticipation of that, developing countries will be
more likely to pursue BITs with these countries. For similar reasons, I include
Log GDPit, which measures the log of the host country’s GDP. I include a measure of the domestic savings rate, savingsit, which is equal to domestic savings as
a percentage of GDP. A low domestic savings rate could provide foreign investors
with an opportunity for investment, but might also provide politicians with a
motivation to take proactive steps to attract FDI, presumably including BITs. All
of the aforementioned economic control variables are lagged 1 year to avoid
potential reverse causality.
I include a measure of the world growth ratet, also taken from WDI, which is
meant to capture year to year variation in the world economy that could trigger
increases or decreases in FDI. This variable is meant to serve a similar purpose
as year-specific dummy variables, which is precluded because of the de-trending
process used on the data (see below). I control for the number of veto players in
a political system using Witold Henisz’s polconiiiit variable. As the number of veto
players in a political system increases, it becomes harder for a government to
make new policy, including signing and ratifying a BIT. As Henisz (2000) has
shown, the policy stasis that accompanies a high number of veto players might
also affect the investment decisions of foreign investors.
As many observers have noted, trade may have a powerful, positive effect on
investment flows (Büthe and Milner 2008; Neumayer and Spess 2005). I include
(the lag of) Trade as a percentage of GDPit, which is taken from WDI and equal to
the sum of exports and imports as a percentage of GDP, because it is also possible that countries with a high volume of trade would be more likely to participate in international trade and investment agreements, including BITs. Similarly,
I control for the existence of a trade agreement between host and source countries (PTAijt).27 The predicted sign for the PTA variable is ambiguous, however.
On one hand, to the extent that adopting a PTA increases trade it should also
lead to an increase in bilateral FDI flows. On the other hand, restrictions on
26
Downloaded March, 2006.
The preferential trade agreement data was taken from the World Trade Organization’s list of regional trade
agreements. This list is available at http://www.wto.org/english/tratop_e/region_e/region_e.htm.
27
Andrew Kerner
87
trade may promote FDI if they make producing inside a country a cheaper way
to reach a market than importing into it. Under this possibility, ratifying a trade
agreement may actually decrease FDI flows by allowing firms to import their
goods more cheaply into the host market. This possibility is particularly pronounced in the dyad fixed-effects specification that I use.
The variable democracyit measures a country’s polity scores from the Polity IV data
set (Marshall and Jaggers 2002). I use a standardized version of this measure (ranging from 0 to 100) made available by Pippa Norris.28 I include this control because
it is possible that democratic countries are more likely to receive FDI as well as
being more likely to ratify BITs (Jensen 2003; Li and Resnick 2003). Last, I include
a measure of (the lag of) capital account openness, CAOIit, using the capital
account openness index used by Brune (2004) and coded from the IMF’s Annual
Report on Exchange Arrangements and Exchange Restrictions. This measure runs
from 0 to 11 with higher values signifying a freer capital account. This measure is
included to guard against the possibility that BITs are simply proxies for broader
efforts that countries take to encourage inward capital flows.29
Instrumental Variables
As alluded to earlier, I suspect that the relationship between FDI and BITs is
endogenous. This is to say that the error term from the structural equation is
correlated with the independent variables, violating an assumption of ordinary
least squares. More formally,
FDIi;j;t ¼ b0 þ b1FDIi;j;t'1 þ b2BITi;j;t þ b3RBITi;t þ di;t þ Uj;t þ Wijt þ li;j þ ei;j;t
ð1Þ
Where i refers to the host country, j refers to the home country, di,t is a vector
of control variables specific to the host country, Fj,t is a vector of control variables specific to the home country, Wijt is a vector of control variables specific to
the country pair and li,j are dyad fixed effects. ei,j,t is the error term such that:
and
COVðei;j;t ; BITi;j;t Þ 6¼ 0
ð2Þ
COVðei;j;t ; RBITi;t Þ 6¼ 0
ð3Þ
In order to capture the endogenous relationship between BITs and FDI I use
a two-stage least squares regression. I use the percentage of a host state’s neighbors that have ratified a BIT with the home state in question as an instrument
for the existence of a ratified BIT.30 I use the 3-year moving average of new BITs
that a host state’s neighbors have ratified as an instrument for the 3-year moving
average of the number of BITs that a host country has ratified with countries
outside of the dyad.31 I define neighbor using the Correlates of War data set’s
coding for type 1 or type 2 contiguity, which includes countries that share a land
border or are separated by 12 miles of water or less (Stinnett et al. 2002). I also
include a third instrument capturing the 3-year average of the number of BITs
28
This variable can be downloaded at http://ksghome.harvard.edu/~pnorris/Data/Data.htm.
While controlling for capital account policy reduces the potential for omitted variable bias, it does not erase
it entirely. It does not, for example, capture the multitude of tax breaks and other inducements that countries
often offer multinational firms to attract their investment. However, such inducements are often firm-specific and
belie accurate coding at the country-year unit of observation (Jensen 2007). I leave it to future research to more
completely account for these efforts in statistical models of FDI flows.
30
This variable is labeled ‘‘instrument 1’’ in the correlation matrix and descriptive statistics that appear in the
Appendix Tables A1 and A2.
31
This variable is labeled ‘‘instrument 2’’ in the correlation matrix and descriptive statistics that appear in the
Appendix Tables A1 and A2.
29
88
Bilateral Investment Treaties
ratified by other countries worldwide in a given year.32 More formally, I estimate
the following reduced form equations.
BITi;j;t ¼ b0Zi;j;t þ b1Pi;t þ b2ki;t þ b3di;t þ b4Uj;t þ b5Wijt þ li;j þ ei;j;t
ð4Þ
RBITi;j;t ¼ b0Zi;j;t þ b1Pi;t þ b2ki;t þ b3di;t þ b4Uj;t þ b5Wijt þ li;j þ ei;j;t
ð5Þ
Where Fi,j,t is the percentage of a host state’s neighbors that have ratified a
BIT with the home state in question, Pi,t is the 3-year average number of new
BITs that a host state’s neighbors have ratified and ki,t is the 3-year average of
the number of BITs ratified by other countries worldwide.
An ideal instrument should be uncorrelated with the error term of the equation but correlated with the endogenous independent variable. As for the latter
requirement, there are a variety of reasons why a country might be moved to
sign and ratify BITs when its neighbors do. First, I expect that as more and more
BITs are signed and ratified, doing so will increasingly become a norm (in the
sense of Finnemore and Sikkink 2005) among developing states in need of foreign capital. Norm diffusion would likely accelerate if leaders observed that BITs
were successful in neighboring countries or globally. Similarly, as more and more
states ratify BITs, states may fear (perhaps unjustifiably, but certainly speculatively, given the paucity of data confirming BITs’ effectiveness) that they must
pursue BITs in order to maintain their competitiveness with other countries in
their drive for foreign capital (Elkins, Guzman, and Simmons 2006; Tobin and
Rose-Ackerman 2005). This property is easily established statistically as well as
theoretically. In my main model, as well as in all of my robustness checks, I
report first stage F statistics that indicate my instruments are more than sufficiently correlated with the endogenous variables.
The strict exogeneity assumption is more troublesome. One could argue that
there is a limited amount of FDI available to the developing world or to certain
regions and that FDI flows can therefore be thought of as a zero-sum game. In
that situation, BITs ratified by neighbors may deprive a country of FDI through
‘‘FDI diversion.’’ If FDI diversion were the norm, then BITs in other countries
(assuming those BITs are correlated with other countries’ FDI inflows) would
likely be negatively correlated with the error term of my structural equation and
therefore would not make an adequate instrument.33 On the contrary, one could
argue that FDI flows to the developing world are better thought of as an expanding pie, rather than a static amount of funds that are diverted from one developing country into another. Under this view, improving legal protections for
foreign investors creates new investment opportunities rather than simply spurring the re-allocation of investment from one country to another. If this were
the case, then BITs in neighboring countries would not have a negative correlation with FDI inflows, and my instrument set would be justified on theoretical
grounds. Recent empirical work on FDI diversion, much of which focuses on the
impact of rising FDI flows to China, tends to corroborate the latter view. Chantasasawat et al. (2004) finds that increases in China-bound FDI has not reduced
FDI flows to Latin America, even as it has reduced Latin America’s share of
worldwide FDI flows. Weiss (2004) finds that increasing FDI flows into China
does not diminish FDI flows into its South-East Asian neighbors, though he does
suggest that increased investment in China may spur FDI in these countries
32
This variable is labeled ‘‘instrument 3’’ in the correlation matrix and descriptive statistics that appear in the
Appendix Tables A1 and A2.
33
An identical concern would arise even if BITs were ineffective, but correlated with other policies that attract
FDI inflows. The effect on my instruments set’s adequacy would be identical; FDI diversion poses significant problems in either case.
Andrew Kerner
89
because of the effects of a global supply chain. Eichengreen and Tong (2007)
find no evidence that rises in Chinese-bound FDI is diverting FDI from other
countries in the developing world, but does find some evidence of diversion
from OECD countries. Outside of China, Kalotay (2004) argues that EU enlargement has not adversely affected FDI flows into the original EU 15. Similarly,
Buch, Kokta, and Piazolo (2001) find that the then impending expansion of the
EU was not diverting FDI flows away from Southern Europe and into Central
and Eastern Europe. Based on these findings I am confident in the adequacy of
my instrument set on theoretical grounds, though, for the reasons noted above,
some skepticism may be warranted.
The exogeneity of my instrument set can also be established statistically using
the Hansen J statistic to test the regression equation’s over-identifying restrictions. The Hansen J statistic is distributed chi-squared with degrees of freedom
equal to the number of over-identifying restrictions and carries the null
hypothesis that the instrument set is exogenous. Thus, a rejection of this
hypothesis casts doubt on the validity of the statistical model. I report the p-values associated with the Hansen J statistic for each model that uses a two-stage
least squares specification. In all but one model (out of 15 models presented
in the following sections), these p-values do not indicate that my instrument
set is endogenous.
Other Econometric Issues
As with all time series cross sectional analyses, the stationarity of the data is a
major concern. Stationarity, which is a necessary condition to make statistical
inferences in a time series context, simply states that the expected values of a series’ mean and variance are not dependent on time. As mentioned earlier, a
Maddala-Wu panel unit root test34 of the variables indicates that a simple count
of BITs ratified with countries outside of the dyad contains a unit roots. The
same is true for the number of BITs ratified by other countries worldwide and
the average number of BITs ratified by a country’s contiguous neighbors, both
of which are used as instruments. Using the 3-year moving average of the first
difference of each variable, as I do, avoids this problem. The remaining trends
in the data, which are substantial, are accounted for by the inclusion of a linear
time trend and its square, which was the highest order polynomial that was
found to be statistically significant (Enders 2004).
I also use dyad fixed effects and include a lagged dependent variable in my
analysis. Dyad fixed effects are increasingly viewed as the most appropriate technique in analyzing dyadic trade flows (Subramanian and Wei 2007; Tomz, Goldstein, and Rivers 2007). Dyad fixed effects effectively control for distance, as well
as historical or cultural ties, between states. I include the lagged dependent variable for two reasons. First, I expect FDI to be ‘‘sticky’’ in that investment flows
in 1 year should be highly correlated with flows in previous years. Second,
including the lagged dependent variable corrects for first order autocorrelation.
In all models I estimate my regressions with robust standard errors to correct for
unspecified forms of heteroskedasticity. Formally, the model that I estimate is
given by
FDIi;j;t ¼ b0 þ b1FDIi;j;t'1 þ b2ZBITi;j;t þ b3ZRBITi;t
þb4di;t þ b5Uj;t þ b6Wijt þ li;j þ ei;j;t
ð6Þ
Where ZBITi,j,t and ZRBITi the exogenous products of equations (4) and (5).
34
Implemented using the xtfisher command in Stata 9.
90
Bilateral Investment Treaties
Results
Models 1a and 1b shows the results of my basic model using OLS, and thus not
accounting for endogeneity. Model 1a uses the unlogged form of the dependent
variable and model 1b uses the logged form, as described above. Neither of the
dyadic BIT variables are significant in these models. The dyadic BIT variable in
model 1a even carries a negative sign, which is consistent with previous results
obtained by Hallward-Driemeyer (2003) and Tobin and Rose-Ackerman 2005.
The count of extra dyadic BITs is found to be significant in model 1b, using the
logged dependent variable, while it is insignificant in model 1a, using the unlogged dependent variable. This finding mirrors those described by Yackee
(2007) and Neumayer and Spess (2005).
Models 2a and 2b are identical to models 1a and 1b, except that both of the
BIT variables have been instrumented using a two stage least squares specification. The first stage F statistics measure the strength of my instruments’ correlation with the endogenous variables. As a rule-of-thumb, an F statistic at or above
10 is acceptable for an equation with two endogenous variables (Baum, Schaffer,
and Stillman 2003). The insignificant p-values for the Hansen J statistic reveal no
evidence that the instrument set is endogenous. The contrast between models 1
and 2 is striking. Both of the BIT variables are statistically significant and positively signed in models 2a and 2b. The coefficients on each are considerably larger than in models 1a and 1b. In other words, investors appear to invest more
when they are protected by a BIT, but also invest more in countries that have ratified more BITs, even though these BITs do not offer the investor any additional
protection from expropriation. Furthermore, correcting for endogeneity reveals
a significantly larger effect than is otherwise evident.
To get a better sense of what these results mean, Table 2 shows the substantive
effect of a 1 standard deviation change in each independent variable on the
dependent variable, or in the case of dichotomous independent variables a shift
from 0 to 1. The substantive effects in Table 2 are based on model 2a, simply
because the unlogged form of the dependent variable makes them easier to
interpret. Table 2 shows that, perhaps unsurprisingly, the most important predictor of dyadic FDI flows is the size of the source country’s economy. The BIT variable also has a large substantive effects. A 1 standard deviation increase in extradyadic BITs yields an increase of $218,671,000 in annual dyadic flows of FDI,
which is equal to an increase of 0.45 standard deviations of the dependent variable. Moving from zero to one in the dyadic BIT variable yields an increase of
TABLE 2. Substantive Effects of BITs on FDI
Variable
FDI (lag)
BIT
Other BITs (3 year avg of new BITs)
Savings (lag)
Polcon
Democracy
Trade (lag)
PTA
Log Source GDP (lag)
Log GDP (lag)
World Growth Rate
CAOI (lag)
Significance levels: *: 10% **: 5% ***: 1%.
Dollar change (millions)
Percentage of SD
184.726***
620.144**
218.671***
)0.760
)80.269***
56.819***
41.127**
)55.285
1688.551***
121.296
20.544***
)42.329**
0.382
1.283
0.452
)0.002
)0.166
0.118
0.085
)0.114
3.493
0.251
0.042
)0.088
Andrew Kerner
91
$620,114,000 in annual dyadic flows of FDI, which is an increase of 1.23 standard
deviations of the dependent variable.
The results described in Tables 1 and 2 are at odds with previous findings that
have modeled the effect of BITs on FDI using a two-stage least squares model.
To recapitulate these results, Tobin and Rose-Ackerman (2005) and HallwardDriemeyer (2003) found that BITs have insignificant or negative effects on FDI
flows between ratifying states. There are several possible reasons why our results
differ. First, my model employs a larger cross section of host countries (127)
than either Tobin and Rose-Ackerman (54) or Hallward-Driemeyer (31). I also
include 22 source countries in my sample, whereas Hallward-Driemeyer uses a
comparable 20 source countries but Tobin and Rose-Ackerman use only one, the
United States.
A more pressing issue is that both Hallward-Driemeyer and, in their bilateral
analysis, Tobin and Rose Ackerman, use the number of BITs ratified by the host
state with countries outside of the dyad as an instrument for BITs ratified within
the dyad. The results of my analysis, as well as those reported by others, indicate
that their instrument is likely invalid. Far from being exogenous, it is a statistically significant and theoretically important predictor of FDI flows. If this instrument is correlated with the error term of the structural equation, as it seemingly
must be, the resulting estimator is inconsistent and the resulting bias can be
large for even modestly correlated instruments (Wooldridge 2003).
The control variables yield interesting results as well. Unsurprisingly, the GDP
of the source country has a large, positive, and statistically significant impact on
FDI flows. The coefficients for democracy are positive and consistently significant,
lending support to arguments given by Jensen (2003) and others. Polcon takes on
consistently negative values, which ostensibly contradicts the logic that foreign
investors are wary of investing where governments can easily change policies (Henisz 2000; Li 2006).
Perhaps the most interesting findings among the control variables are the coefficients on PTA and CAOI. As mentioned earlier, the negative coefficient on PTA
may suggest that FDI acts as a substitute for trade when such trade is made artificially costly by high tariffs. Under this view, ratifying a PTA could actually cause
capital flight by allowing firms to import into a country what had previously been
produced there. Similarly, a liberalization of the capital account may trigger capital
outflows as foreign firms move to allocate their assets more efficiently. It should be
noted, however, that both of these findings are specific to regressions using a dyad
fixed effects specification. Unreported analyzes using a pooled model yield different results. In the case of PTA, the coefficient becomes positive and highly significant, while CAOI becomes insignificant at the 0.1 percent level.
Last, the coefficient for lnGDP is strangely insignificant when the dependent
variable is expressed in its unlogged form and, even more strangely, consistently
takes on negative values when the dependent variable is expressed in logs. Like
PTA and CAOI, however, this appears to be specific to the fixed effects specification. In the unreported pooled models, ln GDP takes on positive and significant
values regardless of the functional form taken by the dependent variable.
Further Tests of Hypothesis 2
Of the three hypotheses tested in this paper, Hypothesis 2 is potentially the most
controversial. Skeptical readers may not accept that BITs are sufficiently costly to
the host state or that these costs are translated into a credible signal of intent in
the way theorized by this paper. As further evidence in support of Hypothesis 2,
models 3, 4, and 5 test some of the less obvious implications of Hypothesis 2.
According to the theory presented in this paper, BITs provide unprotected
investors with a credible signal due to the ex ante political costs with which it is
92
Bilateral Investment Treaties
associated. While this suggests Hypothesis 2, it also suggests a conditional relationship. When the political costs are high, BITs should send a strong message to investors. When political costs are low, BITs should send a weaker message, if they send
a message at all. To test this conditional relationship, I interact extra-dyadic BITs
with three proxies of political costs.
First, the political costs of ratifying a BIT should be greater when the political
opposition can challenge the incumbent government in free and fair elections,
and when political activism on the part of civil society is a protected feature of
political discourse. Both of these qualities are among the hallmarks of political
democracy. I expect that the signal sent by extra-dyadic BITs should be stronger
when they are being sent by relatively more democratic governments that are, by
assumption, more sensitive to pressure from civil society. In order to test this
hypothesis, I interact the extra-dyadic BITs variable with the democracy measure
that already appears in the regression.
Similarly, the political costs of ratifying a BIT should be greater when BITs are
better known to civil society and affected actors. Earlier I pointed to the breakdown of the MAI as a watershed moment in the political opposition to BITs. It
follows that BITs should be costlier after 1997, when a draft form of the MAI was
released and widespread public opposition to the agreement began. As above,
this increased costliness should therefore send a stronger signal of government
intent and attract more investment from unprotected investors. In order to test
this hypothesis, I interact the extra-dyadic BITs variable with 1997, which is a
dummy variable that is set to zero for the period up to and including 1996, and
one for the post-1996 period.
The final operationalization of political costs is through the variable Savings.35 At
low levels of domestic savings, a country becomes more reliant on foreign capital,
while at higher levels of domestic savings it could more plausibly afford to forgo FDI.
Political opposition against BITs should be higher when foreign investment comes
at the expense of a real or potential domestic alternative. If there is no plausible
domestic alternative to foreign investment, the political costs of ratifying a BIT
should be relatively low. As the act of ratification becomes more costly—i.e., at high
rates of domestic savings—the signal being sent to unprotected investors should be
stronger and should accordingly lead to more investment.36
The results from models 3, 4, and 5 provide support (though somewhat limited
support) for these conditional hypotheses. In order to fully evaluate these results,
it is helpful to examine the conditional coefficients for Other BITsit, which can be
found in Tables 3–5. Table 3 contains the conditional coefficients for model 3.
The coefficients indicate several things. First, as a country becomes more and more
democratic, extra-dyadic BITs have a greater and greater effect on FDI flows, which
is what the theory predicts. Second, the change in effect is substantively significant.
Moving from the 25th percentile of democracy to the 75th percentile of democracy
almost doubles the coefficient for Other BITsit. That said, the interaction term is significant only at the 0.1-level, rather than the more conventional 0.05-level.
The conditional coefficients for model 4 are reported in Table 4. These conditional coefficients strongly support the hypothesis that BITs are able to translate
increased political costs following the breakdown of the MAI into a stronger signal
of state intentions. The coefficient for Other BITsit is almost 3.3 times as large for
the period after the MAI increased the political salience of international investment agreements. What is perhaps surprising is that extra-dyadic BITs are effective
35
The author wishes to thank an anonymous reviewer for suggesting this test.
I am not suggesting that FDI acts as a substitute for domestic savings in developing countries. Indeed there
is good reason to believe, as I have argued earlier, that FDI can be a valuable complement to domestic investment.
The regressions reported in this paper suggest that high levels of domestic savings attracts, rather than repels FDI. I
am suggesting, however, that the political costs of ratifying a BIT will be less salient when a country has less domestic capital to draw on to finance its economic growth, and is thus more dependent on foreign capital.
36
Andrew Kerner
93
TABLE 3. Conditional Coefficient Other BITs Model 3
25th percentile of Democracy
50th percentile of Democracy
75th percentile of Democracy
4.618***
(1.590)
5.592***
(1.384)
9.000***
(2.277)
Significance levels: *: 10% **: 5% ***: 1%.
TABLE 4. Conditional Coefficient Other BITs Model 4
Pre-1997
5.229***
(1.703)
22.832***
(6.807)
Post-1997
Significance levels: *: 10% **: 5% ***: 1%.
TABLE 5. Conditional Coefficient Other BITs Model 5
25th percentile of Savings
50th percentile of Savings
75th percentile of Savings
6.588***
(1.827)
8.812***
(2.184)
11.188***
(3.210)
Significance levels: *: 10% **: 5% ***: 1%.
at all during the period before 1997. It is worth remembering, however, that BITs
often languished for years in developing country legislatures before being ratified,
indicating that BITs were troublesome agreements even before the collapse of the
MAI brought these issues to the forefront of political activism.37
The coefficients reported in Table 5 also support Hypothesis 2. These results
indicate that as the domestic savings rate gets higher, extra-dyadic BITs have a
greater effect on FDI flows. This effect is substantial: moving from the 25th percentile of domestic savings to the 75th percentile increases the coefficient by
roughly 70 percent. Similarly to the interaction term in model 3, however, the
interaction term in model 5 is only significant at the 0.10-level.
Robustness Checks
Several of the modeling choices made in this paper might be controversial for
some readers. In the following section I reestimate my models using different
specifications, each time arriving at similar results. The results of these robustness checks are recorded in Table 6. In each of the robustness checks, I use
37
While I locate 1997 as the watershed year in opposition to the MAI, one could reasonably argue that resistance to the MAI was widespread in the years immediately preceding or that it did not begin in earnest until 1998.
Indeed, without a theory about how long it would take for the business community to appreciate changes in the
anti-globalization movement, it is not a priori clear which year should be considered the pivotal year for investment
decisions, even if 1997 is considered the watershed year for anti-globalization activists. In practice, the results of
model 4 are generally the same, though the interaction term is only significant at the 0.1 level, if either 1996 or
1998 is used as the cut-point. On the other hand, there is no statistical support for using 1995 as the pivotal year.
Thus, while my operationalization is admittedly imprecise, the data makes clear that a shift happened in the late
1990s and appears to isolate 1997 as the year this shift took place. Both of these are consistent with the idea that
the break down of the MAI and the associated activism played a significant role in increasing the ex ante costs of
BITs.
4.860
19.020
38.350
0.258
)8.776
(42.077)
16.139**
(6.529)
7.545***
(2.421)
0.039
(0.024)
0.095***
(0.032)
)5.630*
(2.953)
0.028*
(0.015)
)6.253
(4.730)
13.743*
(8.082)
)4.650***
(1.813)
0.072
(0.165)
)0.792***
(0.206)
7691
4.560
45.800
0.223
8.087***
(1.988)
0.018
(0.012)
0.111***
(0.035)
)6.161**
(2.422)
0.031**
(0.015)
)11.858***
(4.116)
16.843***
(6.489)
)4.866***
(1.864)
0.111
(0.166)
)0.887***
(0.240)
7691
4.480
22.850
)0.030
(0.019)
18.286**
(7.478)
)0.027
(0.020)
Significance levels: *: 10% **: 5% ***: 1%.
N
F Statistic
first stage F BIT
first stage F BIT (pre-arbitration)
first stage F BIT (post arbitration)
first stage F Other BIT
Hansen J p- value
CAOI (lag)
World Growth Rate
log GBP (lag)
Log Source GBP (lag)
PTA
Trade (lag)
Polcon
Savings (lag)
Other BITs
(3 year avg of new BITs)
Democracy
BIT post arbitration clause
BIT pre arbitration clause
BIT
FBI (lag)
Variable
Model 7
2SLS
(Std. Err.)
Model 6
2SLS
(Std. Err.)
43.930
0.057
6.606***
(2.003)
0.010
(0.014)
0.134***
(0.037)
)3.736
(2.329)
0.015
(0.015)
)10.615***
(3.630)
16.872**
(6.927)
)2.410
(1.772)
0.176
(0.177)
)0.551***
(0.184)
4931
5.160
)0.018
(0.022)
Model 8
2SLS
(Std. Err.)
24.150
0.505
4.430*
(2.657)
0.014
(0.021)
)0.074
(0.058)
)1.973
(3.032)
0.026
(0.028)
0.922
(1.840)
)3.052
(11.748)
)1.164
(3.036)
)0.007
(0.260)
)0.686**
(0.346)
2696
1.970
)0.102***
(0.028)
Model 9
2SLS
(Std. Err.)
TABLE 6. Effects of BITs on FBI
306.550
0.000
1.365***
(0.302)
0.006***
(0.002)
0.000
(0.004)
)0.715**
(0.299)
0.001
(0.002)
0.145
(0.435)
0.739*
(0.398)
0.593**
(0.236)
)0.030
(0.024)
)0.084***
(0.025)
43762
56.440
266.020
0.169***
(0.014)
5.226***
(0.926)
Model 10
2SLS
(Std. Err.)
29.850
0.982
6.603***
(2.135)
0.030**
(0.015)
0.106***
(0.041)
)7.217***
(2.812)
0.011
(0.016)
)9.137**
(3.580)
6.565
(6.713)
)6.921***
(2.027)
)0.332
(0.226)
)0.682***
(0.237)
3753
3.530
19.870
)0.087***
(0.030)
13.287**
(6.696)
Model 11
2SLS
(Std. Err.)
21.760
0.814
2.717
(1.736)
0.056***
(0.021)
0.035
(0.054)
)8.263**
(3.660)
)0.006
(0.026)
)2.739
(3.534)
24.037**
(11.608)
)3.062*
(1.783)
)3.216*
(1.929)
)0.498***
(0.194)
1455
3.890
12.180
)0.128***
(0.043)
13.402**
(5.873)
Model 12
2SLS
(Std. Err.)
18.310
0.523
509.142**
(246.434)
1.573
(1.273)
3.907
(2.496)
)505.710
(315.951)
5.725**
(2.479)
)378.832
(313.788)
2427.589***
(824.391)
)183.677
(235.667)
16.050
(13.489)
)35.183
(22.449)
4936
86.460
17.470
0.972***
(0.054)
903.875
(633.925)
Model 13
2SLS
(Std. Err.)
94
Bilateral Investment Treaties
Andrew Kerner
95
model 2b as my baseline model to be altered in the specified way. Much recent
work has emphasized the distinction between BITs that contain arbitration
clauses and BITs that do not (Peinhardt and Allee 2007; Yackee 2008). This distinction is critical for my argument: without a pre-consent to third-party arbitration, it is not clear how BITs could effectively tie the hands of ratifying states. A
treaty by treaty coding of which BITs contain an arbitration clause and which do
not regrettably falls outside of the scope of this paper. Instead, I use a shorthand, which is to code ratified BITs signed after 1985 (and 1991 for Ex-Communist countries) as having an arbitration clause and ratified BITs signed before
this date as lacking an arbitration clause.38 Consistent with expectations, the
results of model 6 indicate that there is no evidence that BITs signed in the period before arbitration clauses became standard increase bilateral FDI flows.
There is strong evidence, however, that BITs signed after arbitration clauses
became the norm have a significant effect on FDI flows.39
Recently, investment protection has been included in PTAs as well as BITs.
Chapter 11 of the North American Free Trade Agreement, for example, is not
counted as a BIT but offers essentially the same protections. Many more PTAs
include protection for trade in the provision of goods and services, which arguably serve the same purpose as a BIT. Model 7 recodes the BIT measures to
include PTAs that are notified to the WTO under GATS Article V.40 Recoding
the variable in this way does not affect the results reported in model 2.
Model 8 restricts the sample to only include dyads that have not concluded a
BIT (or a PTA with an investment protection clause). Some readers may be skeptical that the specification given in model 2 is truly able to distinguish between
the roles of protected and unprotected investors. By limiting the sample to only
those dyads that have not concluded a BIT, I can be more certain that I am truly
measuring the investment decisions of unprotected investors. The results from
model 8 echo those found in model 2. Model 9 restricts the sample to only
include dyads that are covered by a BIT (or a PTA with an investment protection
clause). Some readers may prefer this specification because, as is evident from
the discussion of Lauder v. the Czech Republic, some corporations stretch across
borders and can take advantage of several BITs that may be in effect. This possibility could result in misleading results if what appears to be an attracting effect
of extra-dyadic BITs is actually the result of seemingly unprotected investors
investing under the knowledge that they can repatriate their assets in order to
take advantage of a BIT.41 By restricting the sample to only include dyads in
which a BIT (or PTA with an investment chapter) exists, we know for certain
38
This suggestion was made by Jason Yackee in an e-mail exchange on December 10, 2007. Note the distinction
between the date of signature and the date of ratification. Following this coding rule, if a BIT was signed in 1980,
but languished for a decade and a half before being ratified, that BIT would be coded as pre-dating the omnipresence of arbitration clauses.
39
Needless to say, the results of model 6 are only as reliable as the coding rule used, which is inexact by construction. I defer to other work for any authoritative statements on the impact of arbitration clauses on BITs’ effectiveness.
40
A list of trade agreements organized by their provisions can be found at http://www.wto.org/english/tratop_e/region_e/provision_e.xls.
41
It is worth noting that, while it does happen, there are provisions in place meant to discourage this behavior.
Indeed, it would be hard to explain the continued proliferation of BITs if MNCs could easily sue under the auspices of preexisting agreements. There are at least two reasons why firms cannot consistently repatriate themselves
to take advantage of other country’s BITs. First, some countries conscientiously word their BITs to disallow firms
from claiming to be of one nationality when a controlling interest in the firm lies in another country. As an example, Article 17 of the United States Model BIT allows the denial of benefits to any party that ‘‘has no substantial
business activities’’ in a signatory country. Second, dubious claims of nationality can be denied legitimacy by the relevant judicial body by ‘‘piercing the corporate veil’’ of the claimant during a hearing on jurisdiction. The ICSID
retains the precedent set in Banro American Resources, Inc and Societé Aurifére du Kivu et du Maniema S.A.R.L.
v. Democratic Republic of the Congo (2000) that it use ‘‘flexibility’’ and not be ‘‘limited by formalities, and rather
they [the ICSID panel] rule on their competence based on a review of the circumstances surrounding the case,
and, in particular, the actual relationships among the companies involved.’’
96
Bilateral Investment Treaties
that the investor is protected in their home state. Any effect of extra-dyadic BITs
on this sample cannot be explained by opportunistic repatriation of assets. The
results from model 8 are similar to those reported earlier, though the coefficient
on Other BITsijt only reaches the 0.1 level of significance.
The advantages of using the OECD data set are that the data set offers unparalleled breadth of coverage and disaggregates the data by host country as well as by
source country. To my knowledge this is the only available data set that is disaggregated in this way and provides the necessary breadth. The disadvantage of using
the OECD data set is that there is a significant amount of missing data, and that its
‘‘missingness’’ is not random.42 Compared with non-missing observations, missing
observations on the dependent variable tend to cluster in host countries that are
poorer, less democratic and smaller in terms of population. Dyads that are farther
away from each other and that do not share a colonial relationship are more likely
to have missing data. Most of the missing data are clustered in earlier time periods,
though missing data are a significant issue even in the later years of the sample.
Ideally this problem could be overcome by modeling the missingness directly
through a Heckman type of model (Heckman 1979). This procedure, however,
requires the use of an instrument that is correlated with the missingness of the
data, but not with FDI and I have been unable to produce an instrument that plausibly fits this requirement. In models 1–9 I report regression results that use the
sample as it is. Model 10 addresses the missing data in the sample by using
data interpolation where possible and recoding missing data as zero where interpolation is not possible.43 This recoding scheme is plausible for the vast majority of
missing observations (given the likelihood that there is not a significant amount of
bilateral FDI between dyads that report little, if any, data) and is similar to the
approach taken by Hallward-Driemeyer (2003). Addressing missing data in this way
does not diminish the statistical significance of the BIT variables nor does it alter
the directionality of the coefficients. However, the p-value of the Hansen J statistic
indicates that my instrument set does not meet the exogeneity requirement in this
sample. Thus, these results should be viewed with some skepticism.
Finally, in the preceding models I have addressed the possibility of extreme
year to year fluctuations that the dependent variable can take by estimating my
models using the log of FDI flows. In the following models I address this further
by assembling the data into 2-year averages (model 11) and 5-year averages
(model 12). The directionality and significance of the dyadic BIT variable is
unchanged by arranging the data in this way. The extra-dyadic BITs variable is
unchanged when the data is arranged into 2-year averages, but loses significance
when the data is arranged into 5-year averages. Model 13 uses dyadic FDI stock
as the dependent variable rather than dyadic flows.44 Given the high sunk cost
involved in FDI, I suspect that changes in the legal architecture would affect FDI
flows more than FDI stocks. However, and largely for that reason, FDI stock data
has the advantage of being less susceptible to large, year-by-year fluctuations.
The results for model 13 are similar to those using FDI flows, except that dyadic
BITs does not reach statistical significance. Unsurprisingly, the explanatory
power of this model is dominated by the lagged dependent variable.
Conclusions
Extant literature on BITs has had only intermittent success linking BITs with
FDI. The results of this paper go some way toward understanding how these
42
Roughly 71 percent of observations on the dependent variable are missing.
In order to fill in missing values using data interpolation there must be non-missing values before and after
the missing value. I use the ipolate command in Stata 9.2 to perform this procedure.
44
This data is also taken from OECD (2005).
43
Andrew Kerner
97
results were obtained. I find consistent evidence that BITs do attract FDI, and
that they attract it through direct and indirect channels. Furthermore, I find that
the impact of BITs is obscured by endogeneity unless it is corrected for statistically. As alluded to earlier, these results also have implications for broader literatures. First, this paper adds to a growing literature on international economic
institutions and their ability to meet their stated goals. The results reported in
this paper complement similar findings in previously cited work that other international economic institutions are effective. More pertinently, the approach
taken in this paper adds to a growing appreciation of institutional endogeneity
and the complications it can cause for statistical models.45
Second, these results speak to the literature on treaty compliance. Downs, Rocke,
and Barsoom (1996) and others have argued that treaties do not have the capacity
to constrain state behavior, but rather that they serve to identify states that are
already committed to a certain action. Others argue that international agreements
can be designed to actually change state behavior (Abbott and Snidal 2000; Simmons 2000). The empirical results that are presented here do not bear on whether
treaty obligations will affect state behavior per se. However, they do capture investors’ expectations over how states will behave given their treaty obligations. The
finding that BITs attract investment even when the investor is not protected indicates that investors view BITs, in part, as a signal of state preferences that does not
rely on constraints on future state behavior. The finding that investors invest more
where they are specifically protected indicates that the constraining capacities of
BITs are also relevant to investors’ expectations of host state behavior. Contrary to
Downs, Rocke, and Barsoom’s (1996) claim that international agreements are
entered into by the countries that are already predisposed to the prescribed
actions, my results support the idea that BITs are ratified by countries because they
are not viewed as being predisposed to protecting foreign investment.
Third, by focusing on how BITs use their ex ante costliness to induce investment from unprotected investors, this paper hopes to begin a debate on the role
of domestic opposition in the international political economy. In his 2002 essay,
‘‘Feasible Globalizations,’’ Rodrik (2002) notes several barriers to globalization.
One of these barriers is the difficulty in enforcing contracts across borders and
the difficulty in doing business where the social networks that keep businessmen
honest do not apply. BITs are, or have become, a tool for governments to overcome this barrier to foreign investment. Another barrier noted by Rodrik is that
full globalization implies an erosion of national sovereignty—Thomas Friedman’s
(1999) ‘‘Golden Straitjacket’’—that democratic citizens will not willingly tolerate.
This paper addresses both of these concepts and notes a somewhat counterintuitive relationship between the two.
Like Rodrik, we typically think of public opposition to globalization as a factor
slowing down the process. We rarely think of public opposition as a factor that can
spur more and deeper economic integration, but the results of this paper imply
exactly that. By energetically opposing BITs, opposition activists attach ex ante
costs to investment treaties and increase the amount of foreign investment that will
take place after the treaties have been ratified. In doing so, public opposition may
increase the incentives that governments have for ratifying the treaties in the first
place. This is not to suggest that if the opposition to globalization would go away
then so too would globalization. Nor am I suggesting that domestic opposition
does more to justify BITs than they do to dissuade governments from signing and
ratifying them. A more thorough examination of this effect would have to account
for the BITs that never happen or whose scope is limited by the threat of politically
45
It should be noted that not all international agreements are revealed to be effective when modeled properly.
Dreher (2006) and Przeworski and Vreeland (2000), among others, find that accounting for endogeneity reveals
IMF programs to be ineffective.
98
Bilateral Investment Treaties
damaging opposition. These results do, however, point to a nuanced relationship
between national economic policy and domestic activism. The ideal outcome, from
point of view of a government that seeks to negotiate and ratify BITs, is not complete acquiescence from civil society. Rather, the ideal outcome is that there be
political costs that are visible to all, but that fall short of actually taking a toll on the
government at the polls. Exploring the implications of this dynamic on government decisions to pursue BITs lies outside the boundaries of this paper, but may
be a promising avenue for future research.
Appendix A
Albania, Algeria, Angola, Argentina, Armenia, Azerbaijan, Bahrain, Bangladesh,
Belarus, Benin, Bolivia, Botswana, Brazil, Bulgaria, Burkina Faso, Burundi, Cambodia, Cameroon, Central African Republic, Chad, Chile, China, Colombia, Comoros, Congo, Dem. Rep., Congo, Rep., Costa Rica, Cote d’Ivoire, Croatia, Cyprus,
Czech Republic, Djibouti, Dominican Republic, Ecuador, Egypt, El Salvador, Equatorial Guinea, Eritrea, Estonia, Ethiopia, Gabon, Gambia, Georgia, Ghana, Greece,
Guatemala, Guinea, Guinea-Bissau, Guyana, Haiti, Honduras, Hungary, India,
Indonesia, Iran, Israel, Jamaica, Jordan, Kazakhstan, Kenya, Korea, Kuwait, Kyrgyz
Republic, Lao PDR, Latvia, Lesotho, Libya, Lithuania, Macedonia, Madagascar,
Malawi, Malaysia, Mali, Mauritania, Mauritius, Mexico, Moldova, Mongolia, Morocco, Mozambique, Namibia, Nepal, Nicaragua, Niger, Nigeria, Oman, Pakistan,
Panama, Papua New Guinea, Paraguay, Peru, Philippines, Poland, Portugal, Romania, Russian Federation, Rwanda, Saudi Arabia, Senegal, Sierra Leone, Slovak
Republic, Slovenia, South Africa, Sri Lanka, Sudan, Swaziland, Syrian Arab Republic, Tajikistan, Tanzania, Thailand, Togo, Trinidad and Tobago, Tunisia, Turkey,
Turkmenistan, Uganda, Ukraine, United Arab Emirates, Uruguay, Uzbekistan,
Venezuela, Vietnam, Yemen, Zambia, Zimbabwe.
1
0.277
0.752
0.013
0.028
0.001
0.048
0.084
0.047
0.106
0.012
)0.019
0.024
0.171
0.150
)0.012
0.064
0.253
0.189
0.152
0.144
0.138
0.109
0.154
0.132
0.144
)0.030
0.180
0.282
0.173
)0.008
0.047
1
BIT
1
0.420
0.240
0.119
0.052
0.113
0.005
0.171
0.091
0.192
)0.020
0.060
Other BITs inst1
1
)0.017
1
)0.007
0.267 1
)0.041
0.353
0.182
)0.011
0.178
0.419
0.026
0.181
0.312
0.020
0.124
0.250
0.158
0.084
0.119
)0.054
0.140
0.278
0.026
0.027
0.024
)0.022
0.209
0.227
0.240
0.173
0.299
0.283
0.163
0.027
0.008 )0.013 )0.017
0.112
0.055
0.154
LnFDI FDI(la,2a) InFDI(l0) FDI Stock
Log FDI
1
FDI(1a,2a)
0.277
InFDI(10)
1
FDI Stock
0.226
BIT
0.114
Other BITs
0.137
inst1
0.050
inst2
0.150
inst3
0.173
polity
0.196
savings
0.154
polcon
0.176
tradegdp
)0.032
pta
0.141
Ingdp
0.295
Insourcegdp 0.126
worldgrowth )0.026
CAOI
0.060
Variable
polity savings polcon tradegdp
pta
Ingdp
1
0.459 1
0.274 0.251
1
0.094 )0.016 0.112
1
0.258 0.258 0.785 0.093 1
0.040 0.112 0.085 0.094 0.039
1
0.272 0.104 0.331 0.055 0.303 0.075
1
0.224 0.051 0.266 0.499 0.238 )0.288 0.192 1
0.039 0.081 0.025 0.000 0.025 0.010 )0.013 0.007
0.001 )0.088 )0.042 )0.004 )0.010 )0.043 )0.005 )0.012
0.098 0.165 0.151 0.139 0.079 0.311 0.014 0.084
inst2 inst3
TABLE A1. Correlation Matrix
1
)0.005
0.019
1
)0.013
1.00
Insourcegdp worldgrowth CAOI
Andrew Kerner
99
100
Bilateral Investment Treaties
TABLE A2. Descriptive Statistics
Variable
Log FDI
FDI (models 1a, 2a)
InFDI (model 10)
FDI Stock
BIT
Other BITs
instrument 1
instrument 2
instrument 3
polity
savings (lag)
polcon
tradegdp (lag)
pta
Ingdp (lag)
Insourcegdp (lag)
worldgrowth
CAOI (lag)
N
mean
sd
min
max
12699
12699
67720
8232
67720
67720
67720
67720
67720
52914
57490
59294
56720
61912
58942
67720
67720
58699
6.72
87.89
1.38
759.36
0.11
0.31
0.11
0.76
157.51
34.56
14.68
0.17
77.14
0.06
22.61
26.3
2.98
2.17
11.5
489.99
5.81
2446.35
0.31
0.66
0.23
1.28
102.23
38.27
15.45
0.2
43.73
0.23
1.97
1.56
1.07
3.25
)22.36
)5127.1
)22.36
1.15E-015
0
0
0
0
12
0
)84.41
0
1.53
0
17.4
22.36
0.23
0
23.63
18285.14
23.63
51312.96
1
5.33
1
11.83
317
100
65.94
0.67
294.65
1
27.78
29.92
4.63
11
References
Abbott, Kenneth W., and Duncan Snidal. 2000. ‘‘Hard and Soft Law in International Governance.’’ International Organization 54 (3): 421–456.
Baier, Scott L., and Jeffrey. H. Bergstrand. 2007. ‘‘Do Free Trade Agreements Actually Increase
Members’ International Trade?’’ Journal of International Economics 71 (1): 72–95.
Banro American Resources, Inc and Societé Aurifére du Kivu et du Maniema S.A.R.L. v.
Democratic Republic of the Congo. 2000. International Centre for Settlement of Investment
Disputes. No. ARB ⁄ 98 ⁄ 7, September 1.
Baum, Christopher F., Mark E. Schaffer, and Steven Stillman. 2003. ‘‘Instrumental Variables
and GMM: Estimation and Testing.’’ Stata Journal 3 (1): 1–31.
Blonigen, Bruce A., and Ronald B. Davies. 2004. ‘‘The Effects of Bilateral Tax Treaties on US FDI
Activity.’’ International Tax and Public Finance 11 (5): 601–622.
Bouc, Frantisek. 2005. ‘‘Changes Worry Investors from U.S.’’ Prague Post, November 23.
Brune, Nancy. 2004. In Search of Credible Commitments: The IMF and Capital Account Liberalization in the
Developing World. New Heaven, CT: Typescript, Yale University.
Buch, Claudia, Robert M. Kokta, and Daniel Piazolo. 2001. Does the East Get What Would Otherwise
Flow to the South? FDI Diversion in Europe. Kiel Working Paper No. 1061, Kiel: Kiel Institute of
World Economics.
Büthe, Tim, and Helen V. Milner. 2008. ‘‘The Politics of Foreign Direct Investment into
Developing Countries: Increasing FDI through International Trade Agreements?’’ American
Joutrnal of Political Science 52 (4): 741–762.
Canadian Centre for Policy Alternatives Consortium on Globalization and Health.
2002. Putting Health First: Canadian Health Care Reform, Trade Treaties and Foreign Policy.
Available at http://www.policyalternatives.ca/documents/National_Office_Pubs/putting_health_
first.pdf (Accessed May 2008).
Capling, Ann, and Kim R. Nossal. 2004. ‘‘The Rise and Fall of Chapter 11: Investor-State Dispute
Mechanisms in the North American Free Trade Agreement and the Australia–United States Free
Trade Agreement.’’ Paper presented at the Oceanic Conference on International Studies, Australian National University, Canberra, July 14–16.
Chantasasawat, Busakorn, K.C. Fung, Hitomi Iizaka, and Alan Siu. 2004. ‘‘Foreign Direct
Investment in East Asia and Latin America: Is there a People’s Republic of China Effect?’’ Asian
Development Bank Working Paper Series No.28, Tokyo, Japan, November 16.
Choi, Inbom, and Jeffrey J. Schott. 2004. ‘‘Korea-US Free Trade Revisited.’’ In Free Trade Agreements: US Strategies and Priorities, ed. Jeffrey J. Schott. Washington: Peter J. Peterson Institute for
International Economics.
Andrew Kerner
101
Desai, Mihir A., and Alberto Moel. 2008. ‘‘Czech Mate: Expropriation and Investor Protection in
a Converging World.’’ Review of Finance 12 (1): 221–251.
Downs, George W., David M. Rocke, and Peter N. Barsoom. 1996. ‘‘Is the Good News about
Compliance Good News about Cooperation?’’ International Organization 50 (3): 379–406.
Dreher, Axel. 2006. ‘‘IMF and Economic Growth: The Effects of Programs, Loans, and Compliance
with Conditionality.’’ World Development 34 (5): 769–788.
Eichengreen, Barry, and Hui Tong. 2007. ‘‘Is China’s FDI Coming at the Expense of Other Countries?’’ Journal of the Japanese and International Economies 21 (2): 153–172.
Elkins, Zachary, Andrew. T. Guzman, and Beth A. Simmons. 2006. ‘‘Competing for Capital: The Diffusion of Bilateral Investment Treaties, 1960–2000.’’ International Organization 60 (4): 811–846.
Enders, Walter. 2004. Applied Econometric Time Series, 2nd Edition. New York: John Wiley and Sons.
Fearon, James D. 1997. ‘‘Signaling Foreign Policy Interests: Tying Hands versus Sinking Costs.’’ Journal of Conflict Resolution 41 (1): 68–90.
Finnemore, Martha, and Kathryn Sikkink. 2005. ‘‘International Norm Dynamics and Political
Change.’’ International Organization 52 (4): 887–917.
Friedman, Thomas L. 1999. The Lexus and the Olive Tree: Understanding Globalization. New York: Farrar,
Strauss and Giroux.
Garcı́a-Bolı́var, Omar E., and Jon Schmid. 2004. ‘‘The Rise of International Investment Arbitration
in Latin America.’’ Latin American Law and Business Report 12 (12): 4–5, 24–25.
Graham, Edward M. 2000. Fighting the Wrong Enemy: Antiglobal Activists and Multinational Enterprises.
Washington: Institute for International Economics.
Hallward-Driemeyer, Mary. 2003. ‘‘Do Bilateral Investment Treaties Attract FDI? Only a Bit… and
They Could Bite.’’ World Bank Policy Research Paper WPS 3121, Washington.
Heckman, James J. 1979. ‘‘Sample Selection Bias as a Specification Error.’’ Econometrica 47 (1): 153–61.
Henisz, Witold J. 2000. ‘‘The Institutional Environment for Multinational Investment.’’ Journal of
Law, Economics, and Organization 16 (2): 334–364.
International Centre for Settlement of Investment Disputes. 2006. ICSID Convention, Regulations And Rules. Washington: ICSID.
Jensen, Nathan M. 2003. ‘‘Democratic Governance and Multinational Corporations: Political
Regimes and Inflows of Foreign Direct Investment.’’ International Organization 57 (3): 587–616.
Jensen, Nathan M. 2006. Nation-states and the Multinational Corporation: A Political Economy of Foreign
Direct Investment. Princeton, NJ: Princeton University Press.
Jensen, Nathan M. 2007. ‘‘Where Do Multinationals Pay Taxes?’’ Paper presented at the 2nd Annual
Meeting of the International Political Economy Society, Palo Alto, CA, November.
de Jonquières, Guy. 1998. ‘‘Network Guerrillas’’. Financial Times, April 30.
Kalotay, Kalman. 2004. UNCTAD, Geneva, Switzerland. ‘‘EU Enlargement: No Massive FDI Diversion
Seen.’’ Press Release, April 30. Available at http://www.unctad.org/Templates/webflyer.
aspdocid=4754&intItemID=2807&lang=1 (Accessed May 2008.)
Keohane, Robert O. 1984. After Hegemony: Cooperation and Discord in the World Political Economy,
Princeton NJ.: Princeton University Press.
Kim, Sung-jin. 2006. ‘‘Korea Seeks Screen Quota Reduction.’’ Korea Times, January 20.
Knight, Malcom, and Julio A. Santaella. 1997. ‘‘Economic Determinants of IMF Financial
Arrangements.’’ Journal of Development Economics 54 (2): 405–436.
Kobrin, Stephen J. 1998. ‘‘The MAI and the Clash of Globalizations.’’ Foreign Policy 112: 97–109.
Lalumière, Catherine, and Jean-Pierre Landau. 1998. ‘‘Report on the Multilateral Agreement on
Investment (MAI).’’ Ministry of the Economy, Finance, and Industry Report, Paris.
Li, Quan. 2006. ‘‘Democracy, Autocracy, and Expropriation of Foreign Direct Investment.’’ Philadelphia, PA:
Typescript, Pennsylvania State University.
Li, Quan, and Adam Resnick. 2003. ‘‘Reversal of Fortunes: Democracy, Property Rights and Foreign
Direct Investment Inflows in Developing Countries.’’ International Organization 57 (1): 175–214.
Magee, Christopher. 2003. ‘‘Endogenous Preferential Trade Agreements: An Empirical Analysis.’’
Contributions to Economic Analysis and Policy 2 (1): 1–17.
Marshall, Monty G., and Keith Jaggers. 2002. Polity IV Project: Political Regime Characteristics and Transitions, 1800–2002. University of Maryland, Centre for International Development and Conflict
Management. Available at http://www.bsos.umd.edu/cidcm/inscr/polity (Accessed May 2008.)
Neumayer, Eric, and Laura Spess. 2005. ‘‘Do Bilateral Investment Treaties Increase Foreign Direct
Investment to Developing Countries?’’ World Development 33 (10): 1567–1585.
OECD. 2005. International Direct Investment by Country Vol. 2005 release 01. Available at http://
www.sourceoecd.org (Accessed May 2008.)
102
Bilateral Investment Treaties
Peinhardt, Clint, and Todd Allee. 2007. ‘‘The Effects of International Investment Arbitration on Foreign
Direct Investment.’’ Paper presented at the 2nd Annual Meeting of the International Political
Economy Society, Palo Alto, CA, November.
Peterson, Luc Eric. 2003. ‘‘Emerging Bilateral Investment Treaty Arbitration and Threats to Sustainable Development.’’ In Allies or Antagonists: Investment, Sustainable Development and the WTO,
ed. Liane Schalatek. Washington DC: Heinrich Böll Foundation North America.
Peterson, Luc Eric. 2004. Bilateral Investment Treaties and Development Policy-Making. Winnipeg: International Institute for Sustainable Development.
Przeworski, Adam, and James R. Vreeland. 2000. ‘‘The Effect of IMF Programs on Economic
Growth.’’ Journal of Development Economics 62 (2): 385–421.
Rodrik, Dani. 2002. ‘‘Feasible Globalizations.’’ National Bureau of Economic Research Working
Paper w9129, Cambridge, MA.
Salacuse, Jeswald W, and Nicholas P. Sullivan. 2005. ‘‘Do BITs Really Work? An Evaluation of Bilateral Investment Treaties and Their Grand Bargain.’’ Harvard International Law Journal 46 (1): 67–
130.
Simmons, Beth A. 2000. ‘‘International Law and State Behavior: Commitment and Compliance in
International Monetary Affairs.’’ The American Political Science Review 94 (4): 819–835.
Stinnett, Douglass, Jaroslav Tir, Philip Schafer, Paul F. Diehl, and Charles Gochman. 2002.
‘‘The Correlates of War Project Direct Contiguity Data, Version 3.’’ Conflict Management and
Peace Science 19 (2): 58–66.
Subramanian, Arvind, and Shang-Jin Wei. 2007. ‘‘The WTO Promotes Trade, Strongly but
Unevenly.’’ Journal of International Economics 72 (1): 151–175.
Tobin, Jennifer, and Rose-Ackerman, Susan. 2005. ‘‘Foreign Direct Investment and the Business
Environment in Developing Countries: The Impact of Bilateral Investment Treaties.’’ Yale Law
& Economics Research Paper No. 293. New Haven, CT.
Tomz, Michael, Judith Goldstein, and Douglas Rivers. 2007. ‘‘Membership Has Its Privileges:
The Impact of GATT on International Trade.’’ American Economic Review 97 (5): 2005–2018.
United Nations Conference on Trade and Development (UNCTAD). 1998. Bilateral Investment
Treaties in the Mid-1990s. Geneva: United Nations.
United Nations Conference on Trade and Development (UNCTAD). 2000. Bilateral Investment
Treaties 1959–1999. New York: United Nations.
United Nations Conference on Trade and Development (UNCTAD). 2005. Research Note: Recent
Developments in International Investment Agreements. IIA Monitor No. 2. UN Doc. No. UNCTAD ⁄
WEB ⁄ ITE ⁄ IIT ⁄ 2005 ⁄ 1. Available at http://www.unctad.org/en/docs/webiteiit20051_en.pdf
(Accessed January 15, 2009.)
United States Department of Commerce, Bureau of Economic Analysis. 1998. U.S. Direct Investment Abroad: 1994 Benchmark Survey, Final Results. Washington: Government Printing Office.
Vernon, Raymond. 1971. Sovereignty at Bay: The Multinational Spread of US Enterprises. New York: Basic
Books.
Walter, Andrew. 2001. ‘‘Unravelling the Faustian Bargain: Non-State Actors and the Multilateral
Agreement on Investment.’’ In Non-State Actors in World Politics, ed. Daphne Josselin, and William
Wallace. Basingstoke: Palgrave.
Weiss, John. 2004. ‘‘Peoples’s Republic of China and its Neighbours: Partners or Competitors for
Trade and Investment?’’ ADB Institute Research Paper Series No. 59, Asian Development Bank,
Tokyo, Japan. January 21.
Wooldridge, James M. 2003. Introductory Econometrics. New York: South-Western.
Yackee, Jason W. 2007. ‘‘Do BITs Really Work?’’ University of Wisconsin Legal Studies Research
Paper No. 1054, Madison, WI.
Yackee, Jason W. 2008. ‘‘Conceptual Difficulties in the Empirical Study of Bilateral Investment Treaties.’’ Brooklyn Journal of International Law 33 (2): 405–462.
Download