## Karl P. Sauvant and Lisa E. Sachs

Print publication date: 2009

Print ISBN-13: 9780195388534

Published to Oxford Scholarship Online: May 2009

DOI: 10.1093/acprof:oso/9780195388534.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2019. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see www.oxfordscholarship.com/page/privacy-policy). Subscriber: null; date: 15 February 2019

# The Impact of Bilateral Investment Treaties on Foreign Direct Investment *

Chapter:
(p.253) 8. THE IMPACT OF BILATERAL INVESTMENT TREATIES ON FOREIGN DIRECT INVESTMENT *
Source:
The Effect of Treaties on Foreign Direct Investment
Publisher:
Oxford University Press
DOI:10.1093/acprof:oso/9780195388534.003.0008

# Abstract and Keywords

This chapter conducts an empirical assessment of the impact of BITs on FDI stocks. It estimates several variants of the knowledge-capital model of multinational enterprises (MNEs) using the largest available panel of outward FDI stocks provided by the Organization for Economic Cooperation and Development (OECD), which contains FDI of OECD countries into both OECD and non-OECD economies. It shows significant and positive impact of ratified BITs throughout. The estimated effect of BITs on real outward FDI stocks amounts to about 30% in the preferred specification. The chapter also looks at whether simply signing a BIT will have a positive anticipation effect. It finds a positive impact from signing a treaty, although its magnitude is smaller than that associated with the ratification of an existing treaty. However, the estimated anticipation effect is insignificant, in most specifications, leading the conclusion that the advantages to simply signing a BIT are inconsequential.

Keywords:   BITs, FDI, multinational enterprises, MNE, FDI stocks

# INTRODUCTION

The first bilateral investment treaty (BIT) was signed between Germany and Pakistan in 1959 and came into force in 1962. Up to 1999, another 1,856 BITs have been signed and further BITs are expected in the future (United Nations, 2000). BITs are designed to facilitate foreign direct investment (FDI) from economies with abundant capital and skilled labor, that is, mainly Organisation for Economic Co-operation and Development (OECD) countries, to the less developed economies. Many of the existing BITs between the current OECD economies involve one old and one new OECD member. For example, the Czech Republic, Hungary, Poland, and the Slovak Republic concluded BITs with old OECD members in the early 1990s and then joined the OECD afterwards. The theoretical literature on the expected impact of BITs on FDI is not conclusive. Hoekman and Saggi (2000) argue that, due to some differences in national rules, BITs may be the source of higher transaction costs and uncertainty from a firm’s perspective. Although this point would support an argument for a harmonized global BIT, that is, a multilateral investment treaty, these authors concede that differences in cultural, political, and general business climate characteristics are more important determinants of the transaction costs associated with FDI.

Formally, BITs regulate FDI-related issues such as admission, treatment, expropriation, and the settlement of disputes at the bilateral level. Ex ante, they establish transparency about risk and, thus, reduce the risk of investing in a country. Ex post, BITs ensure that firms have certain rights, for example property rights, and preserve them from expropriation.1 According to the Fact Sheet on the U.S. bilateral investment treaty program released by the Office of Investment Affairs of the Bureau of Economic Business Affairs, the program’s basic aims are the following.2 First, BITs should protect U.S. FDI in those countries where (p.254) U.S. investors’ rights are not protected through existing agreements. Second, they should encourage host countries to adopt market-oriented domestic policies that treat private investment fairly. Third, they should support the development of international law standards consistent with these objectives.3 In some sense, BITs extend an investor’s property rights and regulate how host governments must arbitrate disputes covered by the treaty. Further, BITs define what is deemed expropriation, formulate how and under which conditions property may be expropriated, and determine how quickly and comprehensively investors must be compensated.

The United Nations Conference on Trade and Development (UNCTAD 1998) study summarizes the following features of BITs, which are designed to attract FDI. First, BITs facilitate and encourage bilateral FDI between the contracting parties. To achieve this goal, most BITs guarantee foreign investors fair and equitable, non-discriminatory, most-favored-nation, and national treatment in addition to access to international means of dispute resolution. Moreover, BITs usually provide legal protection of both physical and intellectual property under international law and investment guarantees with a special focus on the transfer of funds and expropriation, including the rules of compensation. In this way, they facilitate insurance and reduce insurance premia. Some BITs provide even more reliable and transparent conditions for investors than do national laws. Hence, they allow transition economies to provide guarantees for foreign investors while undertaking national legislative reforms at the same time.4 From this perspective, BITs reduce the costs of investing abroad, including risk premia, so that FDI should increase if new BITs are implemented. In addition, BITs should make new inward investment attractive and also reduce the likelihood of investment outflows.

The theoretical trade literature incorporates multinational enterprises (MNEs) in trade models characterized by increasing returns and considers both horizontal MNEs and vertical MNEs. Horizontal MNEs have production facilities in both the parent and host countries (Markusen 1984; Markusen and Venables 1998, 2000) and tend to be found in the similarly endowed economies, for example within the OECD.5 Vertical MNEs unbundle completely the headquarter services from production to exploit factor cost differentials (Helpman 1984; Helpman and Krugman 1985). Therefore, vertical FDI tends to occur in dissimilar economies, (p.255) for example between the OECD and the developing countries. Carr et al. (2001), Markusen (2002), and Markusen and Maskus (2002) develop knowledge-capital models of MNEs in which both horizontal and vertical activities arise endogenously.

In a panel econometric framework, Hallward-Driemeier (2003) finds little evidence of any positive impact of BITs on FDI. A study based on cross-section analysis by UNCTAD (1998) supports only a weak nexus between signing BITs and changes in FDI flows and stocks. By contrast, the United Nations (2000) views BITs as the most important instrument for protecting FDI at the international level. In this paper, we undertake an empirical assessment of the impact of BITs on FDI stocks. We estimate several variants of the knowledge-capital model of MNEs using the largest available panel of outward FDI stocks provided by OECD, which contains FDI of OECD countries into both OECD and non-OECD economies. Information on BITs, both signed and ratified, is available from the World Bank.

We find a significant and positive impact of ratified BITs throughout. The estimated effect of BITs on real outward FDI stocks amounts to about 30% in the preferred specification. Additionally, we look at whether simply signing a BIT will have a positive anticipation effect. We find a positive impact from signing a treaty, although its magnitude is smaller than that associated with the ratification of an existing treaty. However, the estimated anticipation effect is insignificant, in most specifications, leading us to conclude that the advantages to simply signing a BIT are inconsequential. Section A below provides details on the econometric specification and discusses the construction of the variables. In Section B, we present the main estimation results together with extensive sensitivity analysis. The conclusion presents a summary of the empirical findings.

## A. Specification and Data Base

In the estimated empirical models, we focus on variants of the knowledge-capital model estimated by Carr et al. (2001), Egger and Pfaffermayr (2004b), and Markusen and Maskus (2002). Carr et al. (2001) and Markusen and Maskus (2002) use foreign affiliate sales as the dependent variable because their two-factor knowledge-capital model does not include physical capital. Egger and Pfaffermayr (2004a, 2004b) present a three-factor model and derive explicitly the hypotheses for a specification having FDI stocks as the dependent variable. However, Blonigen et al. (2003) illustrate that the key parameters are qualitatively similar if sales instead of FDI are used. This literature supports using four types of variables to explain the stock of outward FDI at the bilateral level, namely, country size, factor endowments, trade and FDI frictions, and interaction terms. Table 1 summarizes the definitions of our explanatory variables and reports their expected signs from the knowledge-capital model, the horizontal models, and the vertical models.

(p.256) Both absolute bilateral country size (ΣGDP) and similarity in bilateral country size (SIMI) affect horizontal FDI positively.6 In contrast to national exporting firms and vertical MNEs, horizontal MNEs run a production plant in each market and, thus, incur higher fixed costs. A larger size of both the home and the host market increases the likelihood that horizontal MNEs cover these fixed costs. The bilateral difference in the endowment ratio of skilled to unskilled labor (ΔSK) increases vertical FDI, because vertical MNEs arise only if countries differ in terms of production costs, that is, if ΔSK > 0. The difference in the skilled to unskilled labor endowment ratio supports vertical FDI to a lesser extent, if the bilateral distance is large, the home country is large, or bilateral size is large. Transport costs impede trade and, thus, sales of vertical MNEs. The positive nexus between distance and transport costs motivates the interaction term, denoted by DIST·ΔSK, which is nonzero if ΔSK>0. The home country size effect supports the inclusion of the interaction term, denoted by ΔGDP·ΔSK, which is nonzero if ΔSK>0.7 The bilateral size-related interaction term is defined as ΣGDP·ΔSK. In Table 1, the last three interaction terms refer to specifications estimated in Markusen and Maskus (2002). Finally, BITs should reduce the impediment to foreign investment and foster FDI, irrespective of whether horizontal or vertical MNEs are considered. We distinguish between the anticipation effect (BITs) and the ratification effect (BITR) of BITs. Hence, the variable BITs is coded 1 after the date of signing and 0 before and the variable BITR is defined analogously for ratification. One set of specifications includes BITR only, while the other one includes both BITs and BITR. By itself, BITR measures the overall effect of a BIT after it has been implemented; however, if BITs is also included in the regressions, BITR reflects the additional impact of ratification. In the latter case, the overall effect equals the sum of the estimated BITs and BITR coefficients.

(p.257)

Table 1. VARIABLES AND THEORETICAL PREDICTIONS

Abbreviation

Definition

KK

HOR

VER

∑GDPijt

ln(GDPit+GDPjt)

+

+

0

SIMIijt

ln{1−[GDPit/(GDPit+GDPjt)]2−[GDPjt/(GDPit+GDPjt)]2}

+

+

0

ΔSKijt

ln(tert. school enr.it)−ln(tert. school enr.jt)

+/−

+/−

+

D+ = 1 if ΔSKijt>0, 0 otherwise

dummy variable

D = 1 if ΔSKijt>0, 0 otherwise

dummy variable

DISTij·D+·ΔSKijt

ln(distanceij)·D+·[ln(tert. school enr.it)−ln(tert. school enr.jt)]

+/−

+/−

ΔGDPijt·D+·ΔSKijt

[ln(GDPit)−ln(GDPjt)]·D+·[ln(tert. school enr.it)−ln(tert. school enr.jt)]

0

∑GDPijt·D+·ΔSKijt

ln(GDPit+GDPjt)·D+·[ln(tert. school enr.it)−ln(tert. school enr.jt)]

+

−∑GDPijt·D·ΔSKijt

−ln(GDPit+GDPjt)·D·[ln(tert. school enr.it)−ln(tert. school enr.jt)]

BITSijt (BIT signed)

1 after the BIT has been signed, 0 otherwise

+

+

+

BITFijt (BIT into force)

1 after the BIT has come into force, 0 otherwise

+

+

+

Notes:

(i) The notation i indicates the parent country, j refers to the host country, and t is the time index.

(ii) The labels KK, HOR, and VER denote the predicted signs from the knowledge capital, the horizontal, and the vertical models of the multinational enterprise, respectively.

(iii) The model predictions are derived in Carr et al. (2001), Egger and Pfaffermayr (2004a and 2004b) and Markusen and Maskus (2002).

(p.258) We deflate nominal outward FDI stocks in U.S. dollars and employ home-country investment deflators from the World Bank, using 1995 as the base year to approximate real stocks of outward FDI. The explanatory variables consist of real GDP, tertiary school enrollment, population figures, and bilateral distance between capitals. Information on signed and ratified BITs is available from the World Bank for the years from 1959 to 1999. Table 1 provides details on the definition of variables and Table A1 in the Appendix identifies the data sources. Due to missing FDI and school enrollment data, we restrict ourselves to the period from 1982 to 1997. The full design matrix for this period would contain 16,017 observations, whereas only 4,291 remain after adjusting for missing values, mainly of FDI stock data.

In Table 2, we report summary statistics for the key variables in the whole sample and also in the two most important sub-samples of the data, namely intra-OECD relations with 2,789 observations and OECD–non-OECD relations with 1,446 observations. Both bilateral stocks of outward FDI and bilateral country size are higher on average for intra-OECD relations. Consistent with stylized facts, OECD countries are more similar in terms of both size (SIMI) and relative factor endowments (ΔSK). Overall, about three times as many BITs are signed and ratified between the OECD and non-OECD countries as between members of the current OECD economies in our sample. Most existing intra-OECD BITs were negotiated and signed in the early 1990s between old members and then– non-members that have since joined the OECD, for example the Czech Republic, Hungary, Poland, or the Slovak Republic. Only a smaller portion of these treaties has been concluded between two states belonging to the OECD at the date of signing the BIT, for example the United States and Turkey in 1990, Mexico and Spain in 1995, and Mexico and Switzerland in 1995. Hence, the typical BIT is negotiated between a highly developed country and a less-developed partner. Table 3 provides information on ratified/signed BITs between 1982 and 1997 for all countries in the regression sample. The Appendix supplies further details on the covered 19 home and 54 host countries. Of course, not all BITs come into force immediately, but rather there are signed BITs which are still ineffective. On average, about 66% of the BITs signed by the countries in the sample indeed entered into force within this 16-year period. Notably, we define the OECD as of 2003 in Table 3.

With this database at hand, we are able to estimate the impact of ratifying a BIT on bilateral outward FDI. Moreover, we can investigate whether there are significant anticipation effects from signing a BIT only. Our choice of controls will be guided by the recent general equilibrium models of trade and FDI summarized above. Hence, the key determinants will be related to economic size, relative factor endowments, and impediments to trade and FDI.

(p.259)

Table 2. DESCRIPTIVE STATISTICS

Variable

Mean

Std. Dev.

Min

Max

Full sample (4,235 observations)

ln real outward FDI stocks (dependent)

5.58

2.69

−4.31

11.60

∑GDPijt

27.84

1.07

23.35

30.13

SIMIijt

−1.67

1.06

−6.27

−0.69

ΔSKijt

16.82

26.34

−69.80

94.40

DISTij·D+·ΔSKijt

30.06

84.26

0.00

544.40

∑GDPijt·D+·ΔSKijt

38.76

73.35

−150.75

389.86

∑GDPijt·D+·ΔSKijt

577.06

590.60

0.00

2,603.92

∑GDPijt·D·ΔSKijt

−106.39

287.31

−1,885.48

0.00

BITSijt (BIT signed)

0.20

0.40

0.00

1.00

BITRijt (BIT ratified)

0.15

0.36

0.00

1.00

Intra-OECD sample (2,789 observations)

ln real outward FDI stocks (dependent)

6.00

2.73

−4.23

11.60

∑GDPijt

27.92

1.03

25.38

30.13

SIMIijt

−1.54

0.96

−6.27

−0.69

ΔSKijt

8.43

24.77

−69.80

83.20

DISTij·D+·ΔSKijt

44.63

100.43

0.00

544.40

∑GDPijt·D+·ΔSKijt

20.23

53.73

−150.75

376.27

∑GDPijt·D+·ΔSKijt

394.79

499.78

0.00

2,280.02

∑GDPijt·D·ΔSKijt

−158.09

341.51

−1,885.48

0.00

BITSijt (BIT signed)

0.11

0.32

0.00

1.00

BITRijt (BIT ratified)

0.10

0.30

0.00

1.00

OECD-to-non-OECD sample (1,446 observations)

ln real outward FDI stocks (dependent)

4.77

2.43

−4.31

10.13

∑GDPijt

27.69

1.11

23.35

29.72

SIMIijt

−1.94

1.18

−6.20

−0.69

ΔSKijt

33.00

21.24

−17.00

94.40

DISTij·D+·ΔSKijt

1.95

12.07

0.00

141.42

ΔGDPijt·D+·ΔSKijt

74.49

90.86

−120.45

389.86

∑GDPijt·D+·ΔSKijt

928.63

593.57

0.00

2,603.92

∑GDPijt·D·ΔSKijt

−6.68

41.74

−448.32

0.00

BITSijt (BIT signed)

0.35

0.48

0.00

1.00

BITRijt (BIT ratified)

0.25

0.43

0.00

1.00

Note: OECD countries are defined according to 2003 membership.

(p.260)

(p.261)

Table 3. NEW BITS FROM 1982 TO 1997

Country

New BITs signed

New BITs ratified

overall

thereof with OECD

overall

thereof with OECD

Algeria

5

4

3

2

Argentina

38

20

25

16

Australia

12

3

11

3

Austria

19

6

14

6

Belgium-Luxembourg

27

4

14

4

Brazil

10

9

10

9

Bulgaria

33

18

20

14

17

4

9

4

Chile

29

13

7

4

China

71

24

53

22

Colombia

3

2

3

2

Costa Rica

3

3

0

0

Czech Republic

43

22

31

20

Denmark

30

6

22

6

Egypt

21

10

7

6

Finland

27

5

20

5

France

48

4

29

4

Germany

66

4

41

5

Greece

18

4

12

4

Hong Kong

11

11

6

6

Hungary

44

22

33

20

Iceland

1

0

1

0

India

11

6

3

3

Indonesia

18

10

10

9

Israel

12

3

5

3

Italy

43

5

24

3

Japan

3

1

3

1

Korea

38

12

28

10

Kuwait

15

7

7

5

Malaysia

35

9

15

11

Mexico

2

2

2

2

Morocco

15

12

9

8

Netherlands

44

5

36

5

Norway

15

4

14

4

Panama

6

6

5

5

Philippines

10

8

4

4

Poland

53

23

50

23

Portugal

19

4

9

3

Romania

53

21

40

18

Russia

32

22

11

10

Saudi Arabia

1

1

1

1

Singapore

6

2

4

1

Slovak Republic

25

17

23

16

Slovenia

9

5

3

2

South Africa

9

8

9

8

Spain

37

7

27

5

Sweden

25

4

22

3

Switzerland

43

6

32

5

Thailand

14

6

7

4

Turkey

36

15

20

11

Ukraine

26

15

12

7

United Arab Emirates

8

6

3

2

United Kingdom

69

5

58

4

United States

37

4

21

4

Venezuela

15

10

7

4

Notes:

(i) Data are from the International Centre for Settlement of Investment Disputes.

(ii) OECD economies as of 2003 are printed in bold face.

## B. Empirical Results

In this section, we estimate the following basic specification:

$Display mathematics$
where Fijt denotes the log of real stocks of outward FDI of home country i in host country j for year t. The explanatory variables are defined in Table 1. In all regressions, we include fixed country-pair effects (μij) to correct for omitted, time-invariant, geographical or cultural variables, and fixed time effects (λt) to control for omitted time-variant effects that affect all country-pairs in the same way. The usual error term is denoted by εijt.

(p.262) Table 4 reports the estimated coefficients measuring the impact of the explanatory variables on outward FDI. We use tertiary school enrollment as a proxy for a country’s skilled to unskilled labor ratio. In models I and II, we include only BITR, that is, β8=0; in models III and IV, we consider additional anticipation effects captured by the coefficient of BITs. Models I and III that exclude the distance interaction term are closest to the specifications estimated by Markusen and Maskus (2002). Models II and IV both include ΔSKijt and the interaction term DISTij·D+·ΔSKijt as suggested by Egger and Pfaffermayr (2004b). In all cases, we exclude extreme outliers defined as falling in the outer two percentiles of the distribution of residuals and we report only heteroskedasticity-robust t- statistics. The country-pair and time effects always enter significantly and the random effects model is rejected firmly when compared to its fixed effects counterpart. The high adjusted R2 statistics corroborate the appropriateness of our specification. Moreover, the jointly significant size variables underscore the importance of horizontal FDI, whereas the parameters of the skill difference variable and the interaction terms indicate that vertical MNEs are also present. Hence, the signs of these knowledge-capital model coefficients are consistent with the theoretical predictions and with Markusen and Maskus (2002).

Regarding the impact of BITs, several findings are worth emphasizing. First, in all the models in Table 4, the coefficient of BITR is significantly different from zero and it ranges from 0.21 to 0.26. Hence, the estimated impact is relatively unaffected by the choice of specification.8 Second, we find weak evidence for an anticipation effect from the coefficients of BITs in models III and IV. Third, the estimated effect of signing a treaty (BITs) is always smaller than that of ratifying it (BITR). Since FDI stocks are measured in logs, we must transform the BITs effect and its standard deviation to percentage figures. Following Kennedy (1983) and van Garderen and Shah (2002), the overall effect of implementing a treaty is calculated as 100·exp(β9–0.5Var(β9)–1). According to Table 4, this estimated impact amounts to about 30% in models I and II. However, the point estimate of the combined effect of BITS and BITR is even higher. Nonetheless, since the coefficient for BITs is not significant, the parsimonious models I and II are preferred.

In Table 5, we check the robustness of our findings. First, we include additional control variables to ensure that the estimated BITs coefficients do not include effects from other omitted variables. We take account of European Union (EU) and North American Free Trade Agreement (NAFTA) membership for two reasons: Daude et al. (2002) argue that FDI regressions should consider the relevance of trading bloc effects for MNEs, and both the EU and NAFTA are multilateral investment treaties that might generate different effects from their (p.263)

Table 4. THE IMPACT OF BITS

Explanatory variables

Model I

Model II

Model III

Model IV

∑GDPijt

3.692***

3.718***

3.639***

3.666***

(11.92)

(12.00)

(11.72)

(11.80)

SIMIijt

0.203

0.230

0.212

0.239

(1.17)

(1.32)

(1.23)

(1.38)

ΔSKijt

-

0.052

-

0.051

-

(1.17)

-

(1.16)

DISTij·D+·ΔSKijt

-

−0.003

-

−0.003

-

(−1.07)

-

(−1.04)

ΔGDPijt·D+·ΔSKijt

−0.004***

−0.004***

−0.004***

−0.004**

(−3.88)

(−3.42)

(−3.93)

(−3.47)

∑GDPijt·D+·ΔSKijt

−0.000***

−0.002

−0.000***

−0.002

(−3.17)

(−1.35)

(−3.28)

(−1.35)

−∑GDPijt·D·ΔSKijt

−0.000**

−0.003*

−0.000***

−0.003*

(−2.51)

(−1.77)

(−2.57)

(−1.75)

BITSijt (BIT signed)

-

-

0.126

0.124

-

-

(1.62)

(1.60)

BITRijt (BIT ratified)

0.264***

0.262***

0.214**

0.212**

(2.62)

(2.60)

(2.05)

(2.03)

BITS-effect in %

-

-

13.223

13.050

(Kennedy, 1981)

-

-

(1.51)

(1.49)

BITR-effect in %

30.231**

29.914**

23.521*

23.302*

(Kennedy, 1981)

(2.32)

(2.31)

(1.83)

(1.82)

Observations

4,235

4,235

4,235

4,235

0.97

0.97

0.97

0.97

Hausman test:

329.80

418.28

269.77

346.33

p-value

0.00

0.00

0.00

0.00

F-tests:

Country-pair effects

88.08

87.21

87.02

86.15

p-value

0.00

0.00

0.00

0.00

Time effects

29.56

28.54

29.29

28.32

p-value

0.00

0.00

0.00

0.00

Notes:

(i) The figures below the coefficients in parentheses are t-statistics.

(ii) The symbols ***, **, and * denote significance levels of 1%, 5%, and 10%, respectively.

(iii) The t-statistics in bold face for the BITs and BITR coefficients are based on the approximated standard errors suggested by van Garderen and Shah (2002).

(p.264)

Table 5. SENSITIVITY ANALYSIS

BITS (BIT signed)

BITR (BIT ratified)

BITS (BIT signed)

BITR (BIT ratified)

Model I

0.254***

Model I

0.259***

(4.46)

(4.52)

Model II

0.209**

Model II

0.254***

(2.31)

(4.46)

Model III

0.139*

0.156*

Model III

0.143**

0.204***

(1.87)

(1.67)

(2.55)

(3.21)

Model IV

0.138*

0.155*

Model IV

0.140**

0.201***

(1.85)

(1.66)

(2.50)

(3.18)

As in Table 4, but excluding Hungary, Mexico, and Poland

Model I

0.204***

Model I

0.172**

(2.61)

(2.36)

Model II

0.221***

Model II

0.171**

(2.81)

(2.34)

Model III

0.140**

0.147*

Model III

0.070

0.144*

(2.02)

(1.77)

(0.79)

(1.66)

Model IV

0.114

0.173**

Model IV

0.069

0.142*

(1.60)

(2.05)

(0.77)

(1.64)

As in Table 4, but excluding transition countries(b)

As in Table 4, but including an interaction term of BITR with DC-LDC(c)

Model I

0.218**

Model I

0.209*

(2.15)

(1.81)

Model II

0.220**

Model II

0.203*

(2.13)

(1.75)

Model III

0.100

0.182**

(1.51)

(2.41)

Model I

0.252**

Model IV

0.097

0.180*

(2.37)

(1.20)

(1.67)

Model II

0.228**

(2.14)

Notes:

1. ((i)) The figures below the coefficients in parentheses are t-statistics.

2. ((ii)) The symbols ***, **, and * denote significance levels of 1%, 5%, and 10%, respectively.

3. ((iii)) The superscript (a) indicates that additional controls are EU and NAFTA dummies, a dummy capturing the political change in the CEEC after the fall of the iron curtain (1 after 1989 for CEEC and 0 else), the average corporate tax rate, and three variables capturing infrastructure endowments in host countries (road network in km, telephone main lines per 1000 people, and electricity production in kWh).

4. ((iv)) The superscript (b) indicates that Bulgaria, Czech Republic, Hungary, Poland, Romania, Russia, Slovak Republic, Slovenia, and Ukraine are excluded from the sample.

5. ((v)) In the specifications denoted by superscript (c), the main BITR effect is the basis and a dummy capturing OECD outward FDI to non-OECD countries is interacted with BITR and included as well. The corresponding parameter estimates are 0.14 with a t-statistic of 1.05 and 0.15 with a t-statistic of 1.10, respectively.

6. ((vi)) The Difference-in-difference matching approach under superscript (d) proceeds in three steps. In the first step, a probit model is estimated for each year with all exogenous variables plus two outside instruments, the home and host country overall number of BITs, excluding the respective bilateral agreement. In the second step, the Mills ratio is computed for each year. In the third step, a fixed effects model is estimated, including the Mills ratio derived in the second step.

7. ((vii)) In the matching approach under superscript (d), we include the same controls as listed in (a) except for the potentially endogenous EU and NAFTA dummies and the corporate tax rate. An alternative specification, which also excludes the three infrastructure variables mentioned in (a), yields very similar but slightly higher BITR parameter estimates. We obtain virtually the same results from an alternative model that does not use any outside instruments.

(p.265) bilateral counterparts.9 As suggested by Hallward-Driemeier (2003), we account for change in the political system in Central and Eastern European (CEE) countries in the 1990s by introducing a dummy variable equal to 1 in 1990 or later if the host country belongs to this group. Furthermore, we include the host country’s corporate tax rate and three host country infrastructure variables, namely, telephone mainlines per 1,000 people, the size of the road network in kilometers, and the electricity supply in kilowatt hours. Tables A1 and A2 in the Appendix provide details on data sources and summary statistics. To sum up the results, the estimated BITR parameters do not change much in the augmented specifications and are slightly lower compared to their counterparts in Table 4. However, the BITs parameters are now estimated more precisely.

Second, we run this augmented specification but use real GDP per capita instead of tertiary school enrollment to measure the skill difference. This change does not alter the point estimates of BITs and BITR substantially, but the t-statistics are higher than in Table 3 because we now have more observations. Third, we estimate the augmented specification measuring the skill difference by secondary rather than tertiary school enrollment. The corresponding BITs and BITR estimates are somewhat smaller than in Table 4, but BITR is always positive and significant. Fourth, we exclude the new OECD members, namely, Hungary, Mexico, and Poland, from the sample to investigate to what extent the (p.266) estimation results are driven by these countries. In this regression, we obtain a slightly smaller BITR point estimate, but it remains positive and significant.

Fifth, we exclude all of the transition countries in the sample, namely, Bulgaria, Czech Republic, Hungary, Poland, Romania, Russia, Slovak Republic, Slovenia, and Ukraine, and find the BITs and BITR parameter estimates to be almost unchanged. Sixth, we construct an interaction term between BITR and a dummy variable equal to 1 for FDI between OECD and non-OECD countries. We re-estimate models I and II including both the BITR main effect and this interaction term.10 If the parameter estimate of the interaction term were significant, ratified BITs would exert a different impact on FDI between OECD and non-OECD countries than on intra-OECD FDI. Most of the intra-OECD BITs were signed and ratified between old and new OECD members before the latter joined the OECD. Accordingly, the coefficient of the interaction term indicates whether fast growing, recently joining OECD members are affected differently by BITs than are old OECD members. The parameter estimates for this interaction term are 0.14 and 0.15, respectively. Since the corresponding t-statistics are only 1.05 and 1.10 and the coefficients of BITR in models I and II do not change much, we conclude that no differences of this kind exist.

Finally, we check for possible endogeneity of the BITR effect in model I of Table 4. The fixed effects estimator provides an unbiased estimate of BITR only if the selection into treatment, that is, ratifying a BIT, is dominated by time- invariant variables (Wooldridge 2002). If there is endogenous selection and the time-varying treatment is correlated with time-variant unobservables, the fixed-effects estimate is biased. Since we have no fixed-effects panel estimator for endogenous dummy variables available, we follow Blundell and Costa Dias (2002) and apply a matching estimator in a difference-in-difference framework. The proposed estimator compares the change in bilateral FDI stocks between the periods before and after a BIT was ratified with the change in a properly defined control group. Hence, we obtain the average treatment effect of the treated, which is defined as the counterfactual impact of abolishing an earlier-ratified BIT in the presence of self-selection. Furthermore, by taking the difference of differences approach, we eliminate all unobserved time-invariant effects.

The estimation procedure involves four steps. First, we calculate averages of log FDI stocks for the two-year period before and after a BIT has been ratified. The second two-year period includes the ratification year itself. Second, we calculate the change in log FDI between these two periods for both the country pairs that ratified a BIT and the control group. The control group is constructed for each year separately and consists of all country pairs that did not ratify a BIT until the end of the two years defined as the ratification period. Third, we estimate a probit model using the controls of model I, with GDP per capita as a (p.267) measure of relative factor endowments, to obtain more observations. In addition, we apply several external controls, that is, the home and host country overall number of BITs, excluding the respective bilateral agreements, the telephone main lines per 1,000 people of the importing country, the dummy variable for CEE countries in the 1990s, and time dummies.11 Fourth, based on the scores derived in step three, we apply a simple one-to-one matching estimator and a radius matching estimator. The corresponding standard errors are derived by bootstrapping with 100 replications.

The additional controls in the probit model are highly significant but they do not differ significantly between the group of countries that have ratified BITs and the matched control group. Thus, the balancing property is satisfied and, conditional on the explanatory variables in the probit regression, all country pairs are equally likely to ratify a BIT. The estimated BITR-effect remains highly significant, and it is even slightly higher when the endogeneity of BITR is taken into account. The point estimate equals 0.457 with a t-statistic of 2.44 in the one-to-one matching and 0.373 with a t-statistic of 4.28 in the radius matching. The matching estimates indicate that the standard fixed-effects estimates in Tables 4 and 5 may be downward biased. Hence, they represent lower bound estimates of the impact of BITs on FDI stocks.

To summarize, our findings of a positive and significant impact of BITs on bilateral FDI stocks are robust to the inclusion of infrastructure variables, corporate tax rates, and trading bloc effects, to the use of alternative measures of relative factor endowment differences, to the exclusion of new OECD members, and to an endogenous treatment of BITs. Furthermore, BITs do not affect FDI between OECD and non-OECD economies differently than intra-OECD FDI. By taking explicit account of the endogeneity of BITs, we estimate a slightly higher effect of the counterfactual abolishing of an earlier-ratified BIT.

# CONCLUSION

In the last four decades, the number of BITs signed has been large and the pace is accelerating. We investigate whether these treaties are signed only for political reasons or whether they remove important economic obstacles. To do so, we concentrate on FDI and analyze whether bilateral real outward FDI stocks rise as new treaties are signed or implemented. Hence, we hypothesize that bilateral investment treaties reduce barriers to FDI. Embedded in alternative specifications of the knowledge-capital model, we estimate the ratification effect of BITs. In addition, we look at potential anticipation effects after signing and before (p.268) ratifying a BIT. We use the largest available set of bilateral outward FDI stock data from the OECD and a comprehensive data set on bilateral investment treaties from the World Bank. We control for the importance of time-invariant and common-cycle fixed effects as well as for the potential influence of outliers and use heteroskedasticity-robust estimates. We find an overall BIT ratification effect on FDI of about 30% in the preferred specifications. Moreover, if it has an impact at all, signing a treaty exerts a significantly lower impact on real FDI stocks. Our results are robust to alternative measures of relative factor endowment differences, to the impact of trading blocs such as EU or NAFTA, and to infrastructure endowments. In all our estimated specifications, BITs exert a positive and significant effect on real stocks of outward FDI, with a lower bound of 15%. (p.269) (p.270)

(p.271) Appendix List of Countries in the Sample

A. Home countries: Austria, Australia, Canada, Denmark, Finland, France, Germany, Iceland, Italy, the Republic of Korea, Netherlands, New Zealand, Norway, Poland, Portugal, Sweden, Switzerland, United Kingdom, and United States.

B. Host countries: Algeria, Argentina, Australia, Austria, Belgium-Luxembourg, Brazil, Bulgaria, Canada, Chile, People’s Republic of China, Colombia, Costa Rica, Czech Republic, Denmark, Egypt, Finland, France, Germany, Greece, Hong Kong (China), Hungary, Iceland, India, Indonesia, Ireland, Israel, Italy, Japan, Republic of Korea, Kuwait, Malaysia, Mexico, Morocco, Netherlands, New Zealand, Norway, Panama, Philippines, Poland, Portugal, Romania, Russian Federation, Saudi Arabia, Singapore, Slovakia, Slovenia, South Africa, Spain, Sweden, Switzerland, Thailand, Turkey, Ukraine, United Arab Emirates, United Kingdom, United States, and Venezuela.

Table a1. DATA SOURCES

Variable

Source

Bilateral stocks of outward FDI in current US$Foreign Direct Investment Statistics Yearbook 2002 (OECD) Investment deflators in constant 1995 US$

World Development Indicators (World Bank)

GDP in constant 1995 US\$

World Development Indicators (World Bank)

Population

World Development Indicators (World Bank)

Secondary school enrollment share

World Development Indicators (World Bank)

Tertiary school enrollment share

World Development Indicators (World Bank)

Bilateral distance in miles

Own calculations of greater circle distance

Bilateral investment treaties signed (BITs)

World Bank

Bilateral investment treaties ratified (BITR)

World Bank

Host country corporate tax rates

World Development Indicators (World Bank)

Host country telephone main lines (per 1,000 people)

World Development Indicators (World Bank)

Host country roads, total network (km)

World Development Indicators (World Bank)

Host country electricity production (kWh)

World Development Indicators (World Bank)

Table a2. DESCRIPTIVE STATISTICS FOR VARIABLES USED IN THE SENSITIVITY ANALYSIS

Variable

Mean

Std. Dev.

Minimum

Maximum

Endowment variables (GDP/capita)

−SKijt

0.83

1.29

−2.79

4.86

DISTij·D+·ΔSKijt

0.92

2.20

0.00

19.87

ΔGDPijt·D+·ΔSKijt

1.75

3.46

−16.62

18.33

∑GDPijt·D+·ΔSKijt

26.34

32.32

0.00

131.91

∑GDPijt·D·ΔSKijt

−3.35

8.05

−74.64

0.00

Endowment variables (secondary school enrolment)

ΔSKijt

15.48

28.11

−55.57

102.60

DISTij·D+·ΔSKijt

35.06

71.62

0.00

513.88

ΔGDPijt·D+·∑SKijt

28.74

61.46

−234.59

342.48

∑GDPijt·D+·ΔSKijt

560.14

620.86

0.00

2,747.15

∑GDPijt·D·ΔSKijt

−132.91

266.70

−1506.17

0.00

Dummy variables

EU membership

0.05

0.21

0.00

1.00

NAFTA membership

0.00

0.07

0.00

1.00

Central and Eastern  Europe after 1989

0.05

0.22

0.00

1.00

Other variables

Host country corporate tax  rates

30.98

14.67

0.00

76.90

Host country telephone main  lines (per 1,000 people)

340.62

226.38

3.00

1,147.00

Host country roads, total  network (1,000 km)

372.04

977.80

0.00

6,300.00

Host country electricity  production (in billion kWh)

263.00

587.00

2.25

3,560.00

Note: The number of observations is 3,904.

References

Bibliography references:

Blonigen, Bruce A. and Ronald B. Davies. (2003). “The effects of bilateral tax treaties on US FDI activity,” International Tax and Public Finance, forthcoming.

Blonigen, Bruce A., Ronald B. Davies and Keith C. Head (2003). “Estimating the knowledge-capital model of the multinational enterprise: comment,” American Economic Review, 93, pp. 980–994.

Blundell, Richard and Monica Costa Dias (2002). “Alternative approaches to evaluation in empirical microeconomics,” Portuguese Economic Journal, 1, pp. 91–115.

Carr, David, James R. Markusen and Keith E. Maskus (2001). “Estimating the knowledge-capital model of the multinational enterprise,” American Economic Review, 91, pp. 693–708.

Carr, David, James R. Markusen and Keith E. Maskus (2003). “Estimating the knowledge-capital model of the multinational enterprise: reply,” American Economic Review, 93, pp. 995–1001.

Chisik, Richard and Ronald B. Davies (2004). “Gradualism in tax treaties with irreversible foreign direct investment,” International Economic Review, forthcoming.

Dagan, Tsilla (2000). “The tax treaties myth,” New York University Journal of International Law and Politics, 32, pp. 939–996.

(p.272) Daude, Christian, Ernesto Stein and Eduardo L. Yeyati (2002). “Regional integration and the location of FDI,” Working Paper No. 492 (Washington, D.C.: Inter-American Development Bank).

Davies, Ronald B. (2004). “Tax treaties, renegotiations, and foreign direct investment,” Economic Analysis and Policy, forthcoming.

Drabek, Zdenek and Warren Payne (2002). “The impact of transparency on foreign direct investment,” Journal of Economic Integration, 17, pp. 777–810.

Egger, Peter and Michael Pfaffermayr (2004a). “Trade, multinational sales and FDI in a three-factors model,” Review of International Economics, forthcoming.

Egger, Peter and Michael Pfaffermayr (2004b). “Distance, trade, and FDI: a Hausman-Taylor SUR approach,” Journal of Applied Econometrics, 19, pp. 227–246.

Ekholm, Karolina (1998). “Proximity advantages, scale economies, and the location of production,” in Braunerhjelm Pontus and Ekholm Karolina, eds., The Geography of Multinational Firms (Dordrecht: Kluwer Academic Publishers).

Hallward-Driemeier, Mary (2003). “Do bilateral investment treaties attract FDI? Only a bit…and it might bite,” World Bank Policy Research Working Paper Series No. 3121 (Washington, D.C.: World Bank).

Hausman, Jerry A. (1978). “Specification tests in econometrics,” Econometrica, 46, pp. 1251–1271.

Helpman, Elhanan (1984). “A simple theory of international trade with multinational corporations,” Journal of Political Economy, 92, pp. 451–472.

Helpman, Elhanan and Paul R. Krugman (1985). Market Structure and Foreign Trade (Cambridge, MA: MIT Press).

Hoekman, Bernard and Kamal Saggi (2000). “Multilateral disciplines for investment-related policies?” in Paolo Guerrieri and Hans-Eckart Scharrer, eds., Global Governance, Regionalism, and the International Economy (Baden-Baden: Nomos Verlagsgesellschaft).

Kennedy, Peter E. (1981). “Estimation with correctly interpreted dummy variables in semilogarithmic equations,” American Economic Review, 71, p. 801.

Markusen, James R. (1984). “Multinationals, multi-plant economies, and the gains from trade,” Journal of International Economics, 16, pp. 205–226.

Markusen, James R. and Keith E. Maskus (2002). “Discriminating among alternative theories of the multinational enterprise,” Review of International Economics, 10, pp. 694–707.

Markusen, James R. and Anthony J., Venables (1998). “Multinational firms and the new trade theory,” Journal of International Economics, 46, pp. 183–203.

Markusen, James R. and Anthony J., Venables (2000). “The theory of endowment, intra-industry, and multinational trade,” Journal of International Economics, 52, pp. 209–234.

Maskus, Keith E. (2000). “Intellectual property rights and foreign direct investment,” Policy Discussion Paper No. 0022 (Adelaide: University of Adelaide, Centre for International Economic Studies).

United Nations Conference on Trade and Development (UNCTAD) (1998). Bilateral Investment Treaties in the Mid 1990s (New York: United Nations).

United Nations (2000). Bilateral Investment Treaties: 1959–1999. (New York: United Nations).

van Garderen, Kees J. and Chandra Shah (2002). “Exact interpretation of dummy variables in semilogarithmic equations,” Econometrics Journal, 5, pp. 149–159.

Wooldridge, Jeffrey M. (2002). Econometric Analysis of Cross Section and Panel Data (Cambridge, MA: MIT Press).

# Notes

## Notes:

(*) This chapter was reprinted with permission from the Journal of Comparative Economics. The chapter was originally published as “The impact of bilateral investment treaties on foreign direct investment,” 32 Journal of Comparative Economics 788 (2004). Egger acknowledges financial support by the Fonds zur Förderung der wissenschaftlichen Forschung through the Erwin Schroedinger Auslandsstipendium Grant J2280-G05.

(1.) Maskus (2000) addresses the issue of protecting intellectual property. Drabek (2002) considers the importance of the risk of expropriation.

(2.) Hallward-Driemeier (2003) provides further details.

(3.) Hoekman and Saggi (2000) remark that, with the notable exception of those negotiated by the U.S., BITs do not usually address the question of market access liberalization.

(4.) Whereas BITs aim to avoid additional fixed costs by reducing these types of risk, bilateral tax treaties deal with the repatriation of profits. Davies (2004) and Chisik and Davies (2004) present a thorough theoretical treatment of tax treaties. Blonigen and Davies (2003) provide an empirical assessment of the impact of bilateral tax treaties on FDI.

(5.) Horstmann and Markusen (1987 and 1992) model horizontal MNEs in a somewhat different framework.

(6.) Carr et al. (2001) and Markusen and Maskus (2002) use the squared difference in bilateral GDP instead of SIMI. Whereas SIMI rises if two countries are similar with respect to GDP, the squared difference in GDP declines. Therefore, we expect a positive sign on the coefficient for SIMI (Egger and Pfaffermayr 2004a).

(7.) A large difference between the parent and the host countries’ skilled to unskilled labor endowment ratio (ΔSK) is associated with both more horizontal and more vertical FDI, because skilled-labor-abundant countries have a comparative advantage in inventing blue prints and setting up firms or multinational networks. However, this effect applies less to large parent economies (ΔGDP·ΔSK). Blonigen et al. (2003) argue that (ΔGDP)2·|ΔSK| should be used instead. Ekholm (1998) motivates a similar, although more parsimonious, specification. In a reply to Blonigen et al. (2003), Carr et al. (2003) verify that the simple difference rather than the absolute difference in factor endowments should be used as a regressor. Moreover, Carr et al. (2003) and Markusen and Maskus (2002) recommend a specification that allows a positive skill difference to exert a different effect than a negative skill difference. In our case, the specification issue has an impact neither on the sign of the dummies for BITs nor on their significance.

(8.) UNCTAD (1998) and Hallward-Driemeier (2003) do not identify any important, significant effect of BITs, probably because UNCTAD relies on cross-section analysis and Hallward-Driemeier focuses on developing economies and FDI flows rather than stocks.

(9.) We do not report these effects in Table 5. The EU dummy tends to be positive but insignificant; the NAFTA effect is almost always negative and significant, indicating that NAFTA favored trade at the expense of FDI.

(10.) Due to high correlations, BITs and a similar interaction term cannot both be used.

(11.) Hallward-Driemeier (2003) uses similar variables to instrument her dummy variable for BITs in a two-stage least-squares framework.