Купе_The Impact of Policies on FDI flows to Transition Countries

download Купе_The Impact of Policies on FDI flows to Transition Countries

of 37

Transcript of Купе_The Impact of Policies on FDI flows to Transition Countries

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    1/37

    THE EFFECT OF TAX AND INVESTMENT TREATIES ON BILATERAL FDI FLOWS

    TO TRANSITION COUNTRIES

    By Tom Coup, Irina Orlova and Alexandre Skiba

    Tom Coup, Kyiv School of Economics and Kyiv Economics Institute

    Irina Orlova, Center for Social and Economic Research, CASE Ukraine

    Alexandre Skiba, The University of Kansas

    Abstract: This paper estimates the effect of double taxation treaties and bilateral investment

    treaties on FDI flows into transition countries. Using various approaches, including IV, we find

    evidence for the FDI promoting effects of bilateral investment treaties but find no convincing

    evidence for the effect of double taxation treaties.

    1

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    2/37

    INTRODUCTION

    According to the Monterrey Consensus on development financing (2002) rich countries are

    committed to undertake measures to increase foreign direct investment and other private flows to

    low-income and transition economies. The idea behind this political commitment is that all sorts

    of resources must be mobilised for development. Specifically, the Consensus states that a central

    challenge is to create the necessary domestic and international conditions to facilitate direct

    investment flows. The question then is how FDI flows can be promoted by rich countries: what

    policies or measures can enhance FDI out-flows to a given poor country? Among the priority

    areas the Monterrey Consensus mentions are the development of a regulatory framework forpromoting and protecting investments, including the areas of human resource development, the

    avoidance of double taxation and the signing of investment agreements. Also the World

    Investment Report, UNCTAD (2003) focuses on policies that can help to stimulate FDI, focusing

    on such instruments as the adoption of bilateral investment treaties (BITs) and bilateral treaties

    for the avoidance of double taxation (DTTs).

    The Monterrey Consensus was adopted in 2002, when there was a continuing downturn in global

    investment. Already in 2004 the rise in flows to developing countries, South-East Europe (SEE)

    and CIS not only put an end to the downturn that had begun in 2001, but it also represented the

    highest ever level of investment flows to these countries. The question is whether this rise was

    due to the increasing efforts of the international community to enable more foreign investment.

    Even though this particular historical episode escapes our empirical analysis, with this research

    we try to shed light on the relation between specific policies and FDI.

    To contribute to the ongoing debate we focus our study on the effect of international treaties,more specifically, bilateral investment treaties and bilateral tax treaties, on FDI. Bilateral

    investment treaties are agreements between two countries for the reciprocal encouragement,

    promotion and protection of investments in each other's territories by companies based in either

    country. The general purpose of double taxation treaties is to provide a favorable investment

    regime by exclusion of double taxation of an investor. According to UNCTAD (2000), double

    taxation can be defined as the levy of taxes on income/capital in the hands of the same tax payer

    in more than one country in respect to the same income or capital for the same period. DTTs

    arise to avoid such a hardship.

    2

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    3/37

    The existing studies analyzing the effect of treaties on FDI focus either on bilateral investment

    treaties or bilateral tax treaties, but not on both types of treaties simultaneously.1 The studies on

    the effect of BITs provide mixed results. Using a dyadic approach, Hallward-Driemeier (2003)

    finds little evidence for the effect of BITs. Using the same approach, Salacuse and Sullivan

    (2005), however, find a positive effect for BITs signed by the United States, but no effect for

    BITs signed by other OECD countries. Neumayer and Spess (2005) use a non-dyadic approach

    and provide evidence that a greater number of BITs signed by a host country increases the FDI a

    developing country receives. Tobin & Rose-Ackerman (2005) find a positive effect when using a

    non-dyadic approach, but no effect when using a dyadic approach.

    Studies that analyze the effect of tax treaties on FDI are not numerous either. The fundamental

    work on the issue belongs to Blonigen and Davies (2000, 2002, 2004). The basic point of their

    studies is that there is no empirical evidence of the standard view that treaties increase foreign

    direct investment.

    In this study, we empirically examine the effect of both treaties BITs and DTTs on foreign direct

    investment simultaneously. Since previous studies consider either bilateral investment treaties or

    tax treaties, but do not include both in one regression, their results suffer from omitted variable

    bias. This omitted variable bias can be substantial since the correlation between the two treaties

    is quite high: in our sample it is 0.272.

    We further separate tax treaties into three categories: income and capital tax treaties (ICT),

    income tax treaties (IT), and social security treaties (SS). This division is based on the

    categorization by the International Bureau of Fiscal Documentation. Using several dummies,

    rather than one catch-all dummy as is done in previous studies, is important since we thus use

    more information the extent to which a specific treaty influences FDI might depend on its

    specific category - and can get more precise estimates. Some studies also distinguish between

    tax treaties, but based on a time criterion, as Blonigen and Davies (2004) include two treaty

    dummies: one for new treaty and the other for old treaty, differentiating them by the date of

    signature.

    1An exception is the unpublished MA thesis of Goryunov (2004). His results are problematic however because he does notinclude year dummies in his regressions, thus his results confound year effects and treaty effects.2 This correlation is computed between a BIT dummy and a tax treaty dummy which is one if there is any tax treaty and zero

    otherwise (see more about this below).

    3

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    4/37

    Finally, not all studies consider the potential endogeneity issue FDI might cause treaties to be

    signed or not signed rather than vice versa. We provide an extensive treatment of this issue by

    experimenting with several possible instrumental variables.

    Our study provides strong evidence for a positive BITs effect: throughout numerous

    specifications, we get a positive significant estimate of BITs. However, our tax treaty dummies

    do not show such consistency: different specifications give different results. In our preferred

    specifications most tax treaty effects are insignificant.

    This paper is structured as follows: the section describes bilateral investment and tax treaties

    reviews existing empirical studies on both BITs and DTTs. Next we describe the methodology

    and data, followed by estimation technique and results. The final section concludes.

    BITs and DTTs

    Bilateral Investment Treaties (BITs)

    The general purpose of bilateral investment treaties is the promotion and protection of

    investments from one country to another country. While all BITs are very similar in their major

    provisions, some variations exist. These variations are between the two basic models of BITs

    that have emerged so far. First the European model appeared. It was endorsed by the OECD

    Ministers in 1962. Then in the early 1980s the North American model was developed. Both

    models cover the following areas: admission and treatment, transfers, key personnel,

    expropriation and dispute settlement. The two models differ in several aspects. In the

    European model treatment provisions apply only to an investment after establishment, while in

    the North American model treatment provisions concern also the investment at the pre-

    establishment phase (this refers to the entry of investments and investors of a contracting party

    into the territory of another contracting party). Also, the US model has more elaborate

    provisions on some matters (e.g. right of entry) than the European BITs. As to protection of

    established investments, both models contain more or less the same concepts. These include

    national treatment (terms no less favourable than those that apply to domestic investors) and

    most-favoured-nation treatment (terms no less favourable than those that apply to investors from

    third countries), free transfers of funds, adequate and effective compensation in the case of

    expropriation, full protection and security of investments and dispute settlement mechanisms.

    Most BITs have been signed between developed capital exporting countries and developingcapital importing countries. The majority of BITs were signed during the last two decades.

    4

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    5/37

    UNCTAD has been actively monitoring and analysing the increase in the number of BITs. The

    total number of treaties quintupled, rising from 385 at the end of the 1980s to 1,857 at the end of

    the 1990s. During the same period the number of treaties concluded bydeveloping countries and

    by CEE countries has increased from 63 to 833. According to UNCTAD statistics the total

    number of BITs reached 2,265 in 2003 (see Appendix A). They now involve 177 economies.

    The recent proliferation of bilateral investment treaties suggests that they are playing an

    increasingly important role in international investment relations.

    Double Taxation Treaties (DTTs)

    Tax treaties exist between countries on a bilateral basis to prevent double taxation(taxes levied

    twice on the sameincome, profit, capital gain, inheritance or other item). There are a number of

    model tax treaties published by various national and international bodies, such as the United

    Nations and the OECD. The OECD Model Convention on Income and on Capital serves as a

    model used by countries when negotiating bilateral tax agreements. The Convention is dynamic

    in that it is constantly monitored and updated as economies evolve and new tax questions arise.

    In recent years, for example, special reports had to be made on tax treatment of software and

    treaty characterization issues arising from e-commerce.

    According to the International Bureau of Fiscal Documentation 2,390 double taxation treaties

    have been signed by the end of 2005. These treaties include income treaties, income and capital

    treaties and social security treaties. Over two thirds of DTTs (1,632 treaties) have been signed

    since 1980 with a notable surge in the activity in the 90s when 731 treaties were signed. A

    sizeable portion of all treaties is signed by Central and Eastern European countries. CEEs

    participate in 41% of all signed treaties. Most of them, about 70%, were signed since 1985.

    As outlined by Blonigen and Davies (2004), tax treaties perform four primary functions. The

    first is to standardize tax definitions of treaty partners. Differing tax definitions can lead to

    double taxation and inefficient capital flows. The second is to reduce transfer pricing and other

    forms of tax avoidance. The third goal of tax treaties is to prevent treaty shopping. Treaty

    shopping is described as the routing of income through particular states in order to take

    advantage of treaty benefits that were designed to be given only to residents of the contracting

    states. The most common rules regarding treaty shopping restrict treaty benefits if more than

    50% of a corporations stock is held by a third, non-treaty countrys residents (Doernberg, 1997).

    Finally, tax treaties affect the actual taxation of multinational corporations. They do so throughthe provisions for double taxation relief and the rules that reduce maximum allowable

    5

    http://en.wikipedia.org/wiki/Double_taxationhttp://en.wikipedia.org/wiki/Double_taxationhttp://en.wikipedia.org/wiki/Taxhttp://en.wikipedia.org/wiki/Incomehttp://en.wikipedia.org/wiki/Incomehttp://en.wikipedia.org/wiki/Profithttp://en.wikipedia.org/wiki/Capital_gainhttp://en.wikipedia.org/wiki/Inheritancehttp://en.wikipedia.org/wiki/United_Nationshttp://en.wikipedia.org/wiki/United_Nationshttp://en.wikipedia.org/wiki/OECDhttp://en.wikipedia.org/wiki/Taxhttp://en.wikipedia.org/wiki/Incomehttp://en.wikipedia.org/wiki/Profithttp://en.wikipedia.org/wiki/Capital_gainhttp://en.wikipedia.org/wiki/Inheritancehttp://en.wikipedia.org/wiki/United_Nationshttp://en.wikipedia.org/wiki/United_Nationshttp://en.wikipedia.org/wiki/OECDhttp://en.wikipedia.org/wiki/Double_taxation
  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    6/37

    withholding taxes on three types of remitted income: dividend payments, interest payments, and

    royalty payments.

    From the above discussion, we can see that theoretically tax treaties can both promote or reduce

    investment. On the one hand, tax treaties can promote investment by reducing uncertainty about

    the tax environment abroad; on the other hand, a tax treaty which reduces the ability to transfer

    price or somehow prevents other types of tax evasion also reduces the incentive to invest (if

    investment activity is purely for tax minimization reasons). Empirically the question about what

    dominates remains open.

    6

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    7/37

    LITERATURE REVIEW

    A dramatic increase of foreign direct investment during the 1990s led to a boom in economic

    research studying the forces affecting FDI. A part of this literature looks at the relation of

    government policies and FDI. In this section we review the literature that deals with

    international treaties and their effect on FDI.

    While most economic texts use the assumption that treaties encourage FDI, one can find little

    evidence in support of this statement in empirical literature.

    We start with the review of studies that deal with the impact of tax treaties on FDI. These are

    not very numerous. Blonigen and Davies (2000) were the first to directly explore the effects of

    tax treaties on foreign direct investment. They use data over the period 1966-1992 on USinbound and outbound FDI. Several approaches are introduced to capture the effect of treaties.

    A first approach uses a simple dummy variable which indicates whether there is a treaty or not

    for a specific country pair. In the second approach the authors use a treaty age variable equal to

    the number of years that a treaty had been in effect. In both cases authors find positive and

    significant effects and conclude that tax treaties have a strong, positive impact on FDI.

    One problem with this approach is that it combines the effects of older treaties in place long

    before their sample period began with more recent ones. Since the old treaty partners of US

    (Europe, Japan, Canada, Australia, New Zealand) are also the largest hosts and homes for US

    FDI; this means that treaty variables may have been capturing some unobserved differences

    between these countries and other nations. Therefore, in the more recent version (Blonigen and

    Davies, 2004) they separate the treaty dummy and include two dummy variables: one for old

    treaties, the other for new treaties. With two dummies they find that the old treaty dummy has a

    positive and significant effect while the new treaty dummy has a negative and significant effect.

    The same approach is used by Blonigen and Davies (2002) in their paper where they use

    inbound and outbound FDI stock and flow data for OECD countries from 1982 to 1992. Their

    findings are similar to those in the previous work: when old and new treaties are not separated,

    the data indicates a positive and significant effect; after separating treaties the old treaty estimate

    is positive and significant whereas the new treaty estimate is negative and significant.

    Louie and Rousslang (2002) use a different approach but also do not find strong support for the

    FDI encouragement story. They use 1990s income tax return data for US MNEs (Multinational

    Enterprises) to calculate the rate of return for foreign subsidiaries. They test whether this rate of

    7

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    8/37

    return changes after a treaty is ratified. They do not separate the effects of old and new treaties,

    but only count those treaties that had been in force prior to 1987. The coefficient on their treaty

    dummy is significant and negative. However, when authors include in the regressions proxies

    for corruption and political instability, the significance of the treaty dummy fades entirely. Thus,

    they attribute their initial result to the omitted variable bias and conclude that good governance

    attracts both FDI and tax treaties but that treaties have no effect on FDI.

    All studies considered above do not account for treaties signed throughout the 1990s, while most

    transition countries emerged during that period. Moreover, CEE countries concluded the

    majority of tax treaties during the 1990s.

    On the contrary, studies that investigate the influence of BITs on FDI flows do cover the 1990s.

    UNCTAD (1998) used a dataset for 133 countries and 200 BITs to show that there is a rather

    weak correlation between the existence of a treaty and an increase in FDI. In this research only

    cross-sectional data was used (year 1995). Hallward-Driemeier (2003) used data on FDI flows

    from 20 OECD countries to 31 developing countries over the period of 1980-2000. She also

    finds little evidence of a strong positive correlation between bilateral investment treaties and

    increase in FDI flows.

    Salacuse and Sullivan (2005) provide three cross-sectional analyses of FDI inflow to 99

    developing countries in the years 1998, 1999, and 2000, as well as a fixed-effects estimation of

    the bilateral flow of FDI from the United States to 31 developing countries over the period of

    1991-2000. They find that signing a BIT with the United States is associated with higher FDI

    inflows, whereas the number of BITs signed with other OECD countries is statistically

    insignificant. But since this analysis is cross-sectional, it cannot detect how a higher number of

    BITs raises the flow of FDI to developing countries over time. In addition, in their fixed-effects

    regression the authors do not include year dummies, which could mean that the treaty dummies

    reflect year effects.

    Neumayer and Spess (2005) use a larger and more representative sample compared to previous

    work. It covers the period from 1970 to 2001 and includes 119 developing countries. Their

    main explanatory variable is the cumulative number of BITs a developing country has signed

    with OECD countries. They find a strong positive effect of BITs on FDI inflows. They use a

    non-dyadic research design, which means that the authors analyze total OECD FDI flows into a

    developing country, instead of looking at country-to-country flows. They criticize Hallward-

    Driemeier (2003) and Salacuse and Sullivan (2005) on the grounds of using a dyadic modeling,

    8

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    9/37

    which cannot capture the potential of BITs to attract more FDI from other developed

    nonsignatory countries as well, and therefore may underestimate the effect of BITs.

    Tobin and Rose-Ackerman (2005) analyze the impact of BITs using non-dyadic FDI inflows, in

    a panel from 1980 to 2000. Their general conclusion is that a higher number of BITs lowers the

    FDI a country receives at high levels of political risk and raises the FDI only at low levels of

    risk. In an additional dyadic analysis of 54 countries, they fail to find any statistically significant

    effect of BITs on FDI flows from the Unites States to developing countries.

    Summarizing the above studies on BITs, different approaches tend to lead to different

    conclusions authors that use a dyadic approach find no effect, while authors that use a non-

    dyadic approach do find an effect.

    In our research we adhere to the dyadic design, since it controls for country pair specific effects,

    and hence allows us to control better for omitted variables. In contrast to previous studies that

    use a dyadic design and despite the fact that we work with a smaller sample, include both BIT

    and DTTs and tackle the endogeneity issue, we do find robust empirical evidence of a positive

    effect of BITs on FDI.

    9

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    10/37

    METHODOLOGY AND DATA

    Methodology

    The empirical specification is based on the so-called gravity model. The gravity model was

    developed by Tinbergen (1962) and Poyhonen (1963). Originally, it was used to analyze only

    trade flows between countries. Theoretical work on the gravity model for FDI however is scant.

    One prominent study is Bergstrand and Egger (2005). The general idea is the following: the

    amounts of bilateral resource flows will positively depend on size of source/destination countries

    (usually represented by GDP, sometimes by population size or land area, or even all mentioned

    factors simultaneously), which just reflects potential supply/demand, and negatively by

    transportation costs (that is inversely proportional to physical distance between countries).

    Usually, the gravity equation takes a log-linear form:

    ij

    k

    kijkijjiij DDISTGDPGDPFDI +++++= )ln()ln()ln()ln( 321

    where FDIij FDI inflows from country i to country j

    GDPi- GDP of country i

    GDPj- GDP of country j

    DISTij- distance between the capitals of countries i and j

    Dij- dummy variables (treaty dummy, common border, common language dummy)

    k indicates the number of dummy variables included in the regression

    In this study, we will focus on transition countries. The reasons for this focus are multiple.

    First, a lot of newly concluded treaties involved transition countries hence, a sample of

    transition countries is onee where there is a lot of variation in the treatment variable over time.

    Such variability is important to get precise (more efficient) estimates. Second, by focusing on

    transition countries we get a relatively homogenous sample while we sacrifice degrees of

    freedom by restricting our sample, we decrease the extent of omitted variables. Third, for the

    transition countries, the EBRD reform ratings are available which provide us with a good proxy

    for a wide range of home policies3.

    Previous studies either include a BIT dummy or a tax treaty dummy as an explanatory variable.

    In this study, we include both dummies simultaneously to avoid omitted variable bias. In

    3The only other study that focuses on transition countries is the above mentioned unpublished MA thesis of Goryunov (2004).

    10

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    11/37

    addition, instead of including just one tax treaty dummy variable as is usually done, we classify

    tax treaties on the basis of their content and include three dummies. The International Bureau of

    Fiscal Documentation categorizes tax treaties in the following manner: income/capital tax

    treaties, social security treaties, administrative assistance, inheritance/gift, and transport tax

    treaties. We pick up three major categories out of this classification, which constitute the vast

    majority of treaties. These are: 1) income and capital tax treaty; 2) income tax treaty; 3) social

    security treaty. Income and income and capital tax treaties can be differentiated on the bases of

    taxes covered. Thus, income and capital tax treaties cover taxes on both income and capital,

    while income tax treaties cover taxes on income only. For instance, Ukraine has an agreement

    with the Federal Republic of Germany for the avoidance of double taxation with respect to taxes

    on income and capital (income and capital tax treaty); this treaty covers the following taxes in

    Germany: income tax, corporation tax, capital tax, and trade tax, and in Ukraine: tax on profitof enterprises, income tax on citizens, tax on property of enterprises, and tax on immovable

    property of citizens. Whereas with Sweden Ukraine has concluded only an income tax treaty,

    which covers tax on profit of enterprises and income tax on citizens in Ukraine; and State

    income tax, special income tax for non-residents, special income tax for non-resident artistes, the

    communal income tax in Sweden.

    As to social security agreements, they are often signed with tax treaty partners. For example,

    United States has social security agreements with many of its tax treaty partners. Under these

    agreements, many people who work or have worked for both countries can receive credit for

    work performed in both countries under the social security system of one country.

    We allow for more differentiation between tax treaties because it does matter what treaty or

    treaties among the three mentioned above has/have been signed. Different treaties imply

    different degrees of integration between the countries. For instance, the existence of an income

    tax treaty between two countries implies a lower degree of integration as opposed to the casewhere an income and capital tax treaty is in place. Moreover, a country with an income and

    capital tax treaty might be more attractive than a country with just an income tax treaty. In our

    sample there can be at most two tax treaties in place between any two countries, since income

    tax treaties and income and capital tax treaties are mutually exclusive. We incorporate this

    information in our research design in the following way. We include three dummies: income

    and capital tax treaty dummy, income tax treaty dummy, and social security treaty dummy.

    The signs and magnitudes of correlations between these tax treaty dummies vary substantially.

    The ICT dummy and IT dummy are negatively correlated (-0.68) which makes sense since these

    11

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    12/37

    treaties are mutually exclusive; IT and SS treaties are negatively correlated (-0.17) while there is

    a positive correlation between ICT and SS treaties (0.19).

    Variables

    Our dependent variable is the annual bilateral flow of FDI, converted to constant US$ of 2000

    with the help of the US Implicit Price Deflator. Since we take the natural log of the dependent

    variable, we lose all zero FDI flows. In our analysis we disregard zero flows (259 out of 1,224

    observations). But if instead we set all zero FDI flows to positive FDI flows of US$1, as do

    Neumayer and Spess (2005), our results are hardly affected. We only include flows from OECD

    countries to transition countries since we are interested in how FDI streams to developing

    countries can be stimulated (the list of countries can be found in appendix C).

    Our main explanatory variables are four bilateral treaty dummies: (1) BIT dummy capturing the

    effect of bilateral investment treaties; (2) ICT dummy capturing the effect of income and capital

    tax treaties; (3) IT dummy capturing the effect of income tax treaties; (4) SS dummy capturing

    the effect of social security treaties. All four variables are constructed as zero/one dummies.

    Thus, we have one bilateral investment treaty dummy and three bilateral tax treaty dummies.

    Since income tax treaty (IT) and income and capital tax treaty (ICT) are mutually exclusive, the

    maximum number of tax treaties per country pair is two. The possible combinations are: (1)

    social security treaty and income tax treaty and (2) social security treaty and income and capital

    tax treaty. Each dummy takes on the value of one for every year starting from the ratification

    year. To have an effect on FDI attraction, a treaty needs to be ratified. However, there can be

    initial mover advantages and investors might start investing long before the actual date of

    ratification. Therefore, as a robustness check we also run a specification with a dummy based on

    the date of signature.

    Our control variables are those conventionally considered as the determinants of FDI in a gravity

    model. We include the natural logs of host and source GDPs; host GDP (GDPj) is used as an

    indicator of market size, while source GDP (GDPi) is proportional to the pool of potential

    investors from source country. Natural log of hosts GDP per capita (GDPCAPj) proxying for

    production costs is also included. However, as discussed by Globerman and Shapiro (2002), the

    sign of the impact of GDP per capita is ambiguous, since this variable both reflects the level of

    development (encouraging FDI) and the level of wages which can discourage inward FDI if not

    compensated by productivity.

    12

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    13/37

    The distance variable (DISTij) is calculated as a mean distance between the main towns of each

    country. These basic variables are complemented with a dummy for contiguity (CONTij) or

    common border and another one for common language (COMLGij). Like distance, these two

    variables account for various transaction costs incurred when investing abroad. Common border

    and common language data are taken from CIA World Fact book.

    Since transition countries that are members of the World Trade Organization (WTO) might

    receive more FDI as it is easier to export goods back to the home country, we account for this

    with a dummy variable (WTOj) indicating whether a host country is a member of the WTO or

    not.

    To control for host country policies and quality of institution, we include a composite transition

    index (INDEXj) compiled from the EBRD Transition Indicators. These are nine qualitative

    country-by-country indicators with the scale from 1 (little/no progress) to 4.33 (standards of

    advanced market economy) covering first phase reforms (small scale privatization, price and

    trade liberalization) and second phase reforms concerning institutional development (large scale

    privatization, governance, competition, infrastructure, and financial institutions). The transition

    index is constructed as the first principal component of the EBRD Transition Indicators for the

    host countries over the whole sample period. The first component accounts for 80.55% of the

    variance in the nine EBRD Transition Indicators.

    Thus, our baseline equation is the following:

    ijjjijijijijij

    ijijjjiij

    INDEXWTOCOMLGCONTSSITICT

    BITDISTGDPCAPGDPGDPFDI

    ++++++++

    +++++=

    13121110987

    654321 )ln()ln()ln()ln()ln(

    Data Description

    We have created a database of tax treaties involving transition countries. The data includes the

    dates of signature, ratification, and termination. We find a few cases4 when a treaty was

    terminated, but in the same year a new treaty was ratified. In these cases since the treaties were

    re-signed within a year of termination the dummies do not have an interruption.

    We have also considered the difference between the old/inherited and new treaties. Inherited

    treaties are ones which were in place before our sample began.5 Inherited treaties may

    4Renewed tax treaty dummies constitute less than 3% of observations.5These are SS, ICT, and IT, which were mostly signed during the 1960s and 1970s, as well COMECON or CMEA tax treaties.The Council for Mutual Economic Assistance (CMEA) was dissolved in 1990. However, the tax authorities of the signatories tothe multilateral CMEA treaties have agreed to observe the provisions of the treaties amongst themselves until new bilateraltreaties are in place. This agreement took place before the dissolution of the USSR. Original parties to this treaty were: Bulgaria,

    13

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    14/37

    complicate identification of the treaty effect. If we get a positive correlation between the tax

    treaty variable and FDI, it will not be clear whether other unobservable characteristics may be

    leading to increased FDI activity. This occurs because the tax treaty variable will pick up any

    residual effects on FDI which are not measured by other regressors. However, the majority of

    treaties in our sample were signed after our sample data began. We will call them new treaties.

    Inherited treaties comprise only about 6% of our data. Thus, in this respect our treaties can be

    divided into two categories: inherited have only cross-country variation and no time-series

    variation, however, cross-country variation is sufficient; and new treaties have both cross-

    country and time-series variation. These new treaties have a better opportunity to measure the

    impact of a tax treaty, as we have data on FDI activity before and after the treaty.

    The data on tax treaties is taken from International Bureau of Fiscal Documentation tax treaties

    database.As to bilateral investment treaties, this data is drawn from UNCTAD database on

    Bilateral Investment Treaties.

    FDI data is drawn from OECD Direct Investment Database. The source of OECD data are

    Central Banks and Statistical Offices of the FDI home countries, which follow the

    recommendations of the IMF Balance of Payments Manual and the OECD Benchmark

    Definition of Foreign Direct Investment. OECD data is considered more reliable compared to

    other databases compiled on the basis of host country statistics, since it is based on IMF

    methodological guidelines and therefore tends to be more uniform. Host country source of data

    in case of FDI is less reliable because of large discrepancies. Even though new EU member

    states and the candidate countries now follow the IMF definition of foreign direct investment,

    deviations were frequent in the past. For instance, most Western Balkan countries still fail to

    report all the forms of FDI (equity investment, reinvested profits, other investment). There are

    discrepancies due to the very definition of FDI as well. According to IMF Balance of Payments

    Manual, Revison 5, capital investment abroad is regarded as foreign direct investment if thepurpose is to establish and maintain permanent equity relations with a foreign company and at

    the same time to exercise a noticeable influence on the management of that company; the share

    of the foreign investor must make up at least 10 per cent of the target firms equity capital.

    However, not all the countries apply the 10% equity threshold.

    Czechoslovakia, Hungary, German Democratic Republic, Mongolia, Poland, Romania and the USSR. Following the dissolutionof the USSR, the Members of the Commonwealth of Independent States (CIS) have, in principle, agreed to honour theinternational treaties, including income tax treaties, concluded by the USSR until new treaties have been negotiated in their ownname. However, some exceptions took place. For example, Estonia, Latvia, and Lithuania have announced their generaldisapproval of USSR treaties.

    14

    http://www.oecd.org/dataoecd/10/16/2090148.pdfhttp://www.oecd.org/dataoecd/10/16/2090148.pdfhttp://www.oecd.org/dataoecd/10/16/2090148.pdfhttp://www.oecd.org/dataoecd/10/16/2090148.pdf
  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    15/37

    Thus, as mentioned above we consider OECD data to be more reliable. Also, it is observed that

    most of the outward FDI emerges from developed countries and about 87 per cent of this comes

    from the OECD countries, where the top three positions are taken by United States, United

    Kingdom and France (see Appendix B). Our dataset covers 17 OECD (home) countries and 9

    transition (host) countries over the period of 1990-2001. We should mention here as well that

    OECD data does not contain negative values of FDI (disinvestment).

    The OECD database contains FDI data in nominal terms; in local currencies. With the help of

    annual average exchange rates and Implicit Price Deflator (IPD) we convert flow data to US

    dollars 2000.

    Data on GDP is borrowed from the World Development Indicators (WDI) database. Information

    about common borders and common language is available in the CIA World Fact book.

    Common border may affect the amounts of capital flows. Distances between countries capitals

    are calculated using online source http://www.indo.com/cgi-bin/dist/ .

    Data on WTO dates of membership is taken from the World Trade Organization official website

    http://www.wto.org/english/thewto_e/whatis_e/tif_e/org6_e.htm .

    The composite transition index is compiled on the basis of EBRD Transition Indicators borrowed

    from EBRD website.

    Information on treaties is summarized in figures 1 through 3. Figure 1 with the cumulative

    number of bilateral investments treaties by FDI host country demonstrates the differences in the

    treaty signing patterns among transitioning countries. We will exploit these differences to

    identify the effect of treaties on FDI. Hungary, Poland, Czech Republic come into our data

    having investment treaties with most source countries while during the same period there are

    some country pairs that do not sign a treaty. Bulgaria, Romania, Slovenia, and Ukraine show themost activity and sign most of the treaties between 1990 and 2002. Even among the most

    dynamic countries there is some variation in the timing of activity. Romania and Ukraine sign

    most of the treaties in the first half of the 90s while Slovenia and Bulgaria are still signing

    treaties towards the end of the decade. Figure 2 shows the cumulative number of investment

    treaties and cumulative number of the double taxation treaties (IT and ICT). The figure suggests

    that the periods of signing of both types of treaties often coincide and both exhibit a fair amount

    of variation over time. Figure 3 describes components of the total number of BITs by groupings

    of the host countries. Countries experience growth from different sources. Slovenia starts

    signing treaties with the non-G7 European countries without a common border and does not sign

    15

    http://www.indo.com/cgi-bin/dist/http://www.wto.org/english/thewto_e/whatis_e/tif_e/org6_e.htmhttp://www.indo.com/cgi-bin/dist/http://www.wto.org/english/thewto_e/whatis_e/tif_e/org6_e.htm
  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    16/37

    any treaties outside Europe. Romania signs treaties with G7 European countries before

    expanding to other countries in Europe. Ukraine starts with G7 Europeans, continues to other

    European countries before signing outside the continent.

    16

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    17/37

    EMPIRICAL ESTIMATION AND RESULTS

    Estimation procedures

    Our sample can be grouped in two different ways: by country pairs (153 groups) and by hostcountries (9 groups). We do the estimation for both groupings. When we do the host country

    grouping we keep host characteristics constant and look across source countries. When we do

    country pair grouping we take a pair and look within the pair trying to identify whether the

    timing of the effect played a role.

    Our estimation starts with several tests, which allow us todecide whether fixed effects, random

    effects, or simple OLS on pooled data should be preferred. First, we look at panel versus simple

    OLS on pooled data estimation. Since we suspect country-specific effects in the data, we

    perform fixed-effects, and the F-test applied after carrying out the fixed-effects estimation

    answers the question of fixed-effects vs. pooled OLS. Second, if F-test suggests that panel

    estimation should be preferred, the Hausman specification test is applied to discriminate between

    fixed and random effects estimation of the panel data. The tests mentioned above suggest fixed

    effects specification for host country grouping and random effects specification for country pair

    grouping.

    There is potentially a problem of endogeneity: countries with larger bilateral FDI flows are also

    more likely to enter into an agreement. One way to tackle endogeneity is to use instrumental

    variables estimation. Successful implementation of this approach hinges on availability of valid

    instruments. What we need are variables that affect the probability that two countries conclude a

    treaty but are not directly related to the volumes of FDI between these two countries. We will

    refer to these factors as countrys propensity to conclude treaties.

    One possibility is to use the number of other tax/investment treaties a host has entered into with

    countries other than the source country being considered. We will further refer to these treaties

    as outside treaties. This is exactly what Hallward-Driemeier (2003) does. She writes: The

    willingness of a host to ratify a BIT, as measured by the number of outside BITs, should be

    correlated with the probability it signs with this particular host country, but shouldnt affect the

    amount of FDI that particular source country would send. Thus, when US investors are

    considering investing in India, their decision would not be affected by whether India has ratified

    treaties with the UK or France. However, that India has entered other treaties would be expected

    to influence their willingness to enter such a treaty with the US. Note that those who claim the

    17

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    18/37

    non-dyadic approach is better, like Neumayer and Spess (2005) believe that BITs do have

    positive spillover effects because of what they call the signaling effect. The signing of BITs

    (especially with major capital exporting countries) sends out a signal to potential investors that

    the developing country is generally serious about the protection of foreign investment. The

    authors assume that it is difficult to say how important the signaling effect is (which benefits

    investors from all countries), compared to the commitment effect (which only relates to investors

    from BIT partner countries). Still, if the signaling effect is important, how can one explain, for

    example, Ukraine signing 66 tax treaties and 58 BITs with other countries. Indeed if spillovers

    were important we would observe that countries would sign only treaties with the most important

    capital exporting nations. Therefore, in our opinion, the dyadic approach is better because it

    allows us to control for omitted variables (unlike the non-dyadic approach) and the propensity of

    signing treaties can be used as a valid instrument.

    Instead of using just the number of outside treaties a host country has, we construct another

    instrument the sum of outside treaties a host and a home country have. A pair of countries is

    more likely to conclude an agreement, when both countries, not just the host country, have the

    habit of signing such agreements. Our instruments are intended to reflect the overall propensity

    of countries to sign such types of treaties. The coefficients of correlation between treaty

    dummies and suggested instruments range from 42% (SS) to 64% (ICT). We have tried other

    possible options for instruments. Among them the conventionally used number of outside

    treaties a host country has, as well as the number of outside treaties a home country has, and

    possible combinations between the two. We have also tested the option of using the number of

    treaties signed during the last five years for both home and host countries. The logic behind this

    is that those countries that have been more active signing treaties in the last five years are more

    likely to conclude such agreements now. The highest correlations between the instruments and

    instrumented variables are found for the sum of outside treaties that both host and home

    countries have for all years. Previous studies based on IV methodology, use the number of

    outside treaties a host country has signed. To test whether the instruments we use (sums) are

    stronger than the traditional alternative (outside treaties of hosts) we compare first-stage F-

    statistics6. They are significantly higher for the instrument we use, especially for ICT and IT.

    It is worth noting that we cannot use measures of openness or the transition index to instrument

    for investment treaties because those measures inevitably correlate with the FDI flows.

    6 F-statistics reported in the Appendix D.

    18

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    19/37

    Another potential problem is a statistically significant relationship between two upward trending

    variables that is spurious. We deal with this problem by including year dummies to account for

    any year-to-year variation in FDI flows unaccounted for by our other explanatory variables.

    Finally, we explore the timing of the treaty effects on FDI. A dummy for each of the three years

    prior to and after the ratification of the treaties is included. Another dummy takes on the value

    of unity for years after the third year from ratification. We estimate the effects of timing in

    specifications with host and pair random effects, and pooled OLS. The coefficients on the

    timing dummies should be interpreted in terms of difference with the reference period. In our

    case the reference period comprises all years up to 4 years prior to ratification. The timing of the

    treaty effect has been investigated in a similar fashion by Hallward-Driemeier (2003). The

    difference is that we include a separate dummy for all years after three years from ratification of

    each treaty. The usefulness of our approach lies in separating the periods before and after the 3

    year window, which otherwise are lumped together in the reference category and forces the

    effect of treaties to die out. Our approach allows the coefficient on a treaty timing dummy to be

    interpreted as the difference between the level of FDI in that year and the level of FDI from

    years earlier than 3 years prior to ratification.

    We expect that the effect of the third year prior and after the ratification will be estimated less

    precisely because there are fewer observations for these dummies.

    19

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    20/37

    Results

    Table (1) reports pooled OLS estimation results for the logged amount of FDI in US$ of 2000

    flowing from OECD country to a transition country as the dependent variable. As a starting date

    for our treaty dummies in columns (1) and (2) we use the date of ratification and then the date of

    signature as a robustness check in column (3). Almost all control variables are significant and of

    expected sign. The larger the home country and the host country, the larger is the FDI flow.

    Flows are also higher to host countries with smaller GDP per capita, which makes sense if we

    think of GDP per capita as a proxy for production costs.

    20

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    21/37

    Table 1. OLS estimation results (logged FDI flows in $US 2000)

    Variables OLS (rat)

    Interactionwith

    transition

    index

    OLS (sign)

    lnGDP host 0.71068(0.000)***

    0.73006(0.000)***

    0.7603(0.000)***

    lnGDP home 0.91832(0.000)***

    0.9161(0.000)***

    0.9478(0.000)***

    lnGDPCAP host -0.22465(0.099)

    -0.2749(0.044)**

    -0.3485(0.007)**

    lnDistance -0.79725(0.000)***

    -0.83469(0.000)***

    -0.88202(0.000)***

    Common language 2.26368

    (0.000)***

    2.2405

    (0.000)***

    2.1495

    (0.000)***

    Contiguity -0.19366(0.327)

    -0.28437(0.152)

    -0.2832(0.161)

    Transition Index 0.47526(0.000)***

    0.6187(0.000)***

    0.54209(0.000)***

    WTO -0.05892(0.741)

    0.7427(0.681)

    -0.0322(0.860)

    BIT 0.44207

    (0.001)***

    0.70463

    (0.000)***

    0.2829

    (0.052)***

    ICT 0.93916

    (0.004)***

    0.9908

    (0.002)***

    0.2482

    (0.380)

    IT 0.66218

    (0.037)**

    0.6652

    (0.034)**

    -0.10102

    (0.719)

    SS -0.5422

    (0.001)***

    -0.5934

    (0.000)***

    -0.5446

    (0.000)***

    BIT*Transition Index -0.22456(0.000)***

    N. Obs 962 962 962

    R2 0.4878 0.4832 0.4788

    Note: p-values in parentheses; * - significance at 10%; ** - 5%; *** - 1%

    Robust standard errors; year dummies not reported.

    21

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    22/37

    Our distance variable is negative, while having a common language is positive, both are

    significant and in accordance with theoretical expectations. Contiguity (common border)

    dummy is insignificant. This might be because few neighboring countries are present in our

    sample. The transition index as a measure of institutional quality in host countries shows a

    positive significant effect. Our variables of interest are significant at least in columns (1) and

    (2); in column (3) only the bilateral investment treaty dummy and the social security treaty

    dummy remain significant. Since column (3) presents the results based on the date of signature,

    this supports the hypothesis that to have an effect on FDI a treaty needs to be ratified. All treaty

    dummies but SS have a positive effect. Thus, according to the results in column (1), the

    existence of a BIT between two contracting parties increases the FDI flow by 44%. If ICT or IT

    is in place we get a 94% or 66% increase correspondingly, while SS treaty decreases FDI flow

    by 54%. There could be a possible explanation of this negative effect based on the tax evasion

    provisions, which can serve as disincentives to engage in FDI activity.7

    In column (2) we interact the existence of a BIT with the institutional quality (proxied by the

    Transition Index). We do so to test whether BITs are only valuable within a country with a

    certain level of overall institutional development. Two possibilities exist: BITs can act either as

    complements or as substitutes for strong domestic protection of property rights. A positive

    interaction term on institutional quality and the ratification of a BIT would favor the former

    interpretation, while a negative interaction term would favor the latter. The results show a

    negative interaction, thus revealing that BITs per se will have a stronger effect in a weak

    institutional setting. Since tax treaties are not really aimed at complementing or substituting

    domestic protection of property rights, we do not interact them with the institutional quality.

    When significant the per capita GDP of the host country negatively affects the investments

    inflows. This finding is expected if FDI is carried out to take advantage of the low labor cost,

    (vertical FDI) rather than to access the market (horizontal FDI). Producers seeking lower cost

    locations are more likely to channel their investments into countries with lower incomes.

    Another possible explanation comes from the fact that the low income counties experience

    relatively fast growth in the FDI inflows because they start from a lower stock of FDI. In that

    case the coefficient on per capita GDP captures the effect of catching up by the lower income

    economies.

    7 A Detailed explanation can be found below.

    22

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    23/37

    Table (2) reports panel estimation results. Hausman test fails to reject the random-effects

    assumption (time-invariant factors are uncorrelated with the explanatory variables) for country

    pair grouping; but rejects this assumption for host country grouping. We therefore present the

    random-effects estimation results for pair grouping in column (1) and fixed-effects for host

    grouping in column (2). For both of these specifications we interact BIT with Transition Index

    in columns (3) and (4). For the random effects pair specification, as a robustness check, we use

    the date of signature instead of the date when a treaty was ratified (column 5).

    Throughout all the specifications control variables stay significant and of the expected sign with

    the exception to host GDP p.c, contiguity, and WTO variables, which are insignificant. The

    WTO dummy turns negative significant in host fixed effects specifications. This is rather a

    strange result, since the WTO dummy is expected to have a positive effect. But since all

    countries in our sample entered WTO in 1995 (except for Bulgaria - 1996; Russia and Ukraine

    did not enter WTO so far), variation in this variable is low, and we probably should not pay too

    much attention to this. Among the variables of interest only BIT and SS show consistency in the

    signs and significance throughout all specifications8.

    BITs encourage FDI flows and this effect ranges from 42% to 83%, which is close to what we

    get in pooled OLS estimation. All specifications give a negative significant SS result, which is

    puzzling because SS should increase FDI if they remove the possibility of paying social security

    contributions twice. However, there could be a possible explanation originating from legal

    literature on tax treaties, which assumes that treaties are intended to reduce tax evasion by

    MNEs, not reduce double taxation. In particular, the matter concerns transfer pricing provisions

    and the exchange of tax information between governments. These measures can offset the FDI-

    promoting effects of tax treaties.

    The ICT dummy displays sufficient consistency too. It is positive and significant in the first four

    specifications, and turns insignificant only in the last column, where we do a specification based

    on the date of signature. This again supports the hypothesis that to have an effect on FDI a treaty

    needs to be ratified. The IT dummy is significant (with positive sign) only in host fixed effects

    specifications, but not in country pair random effects specifications.

    Including a traditional tax treaty dummy (one if there is any tax treaty and zero if there is no

    treaty) like in previous works shows no significant effect.

    8If we use pair Fixed effects, BIT and SS remain significant, while the other tax treaties are insignificant.

    23

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    24/37

    The interacted BIT*Transition Index gives negative significant result which is consistent with

    what we get for pooled OLS specification.

    24

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    25/37

    Table 2. Panel estimation results (logged FDI flows in $US 2000)

    Variables Pair RE (rat) Host FE (rat)Pair RE

    (interaction)Host FE

    (interaction)Pair RE(sign)

    lnGDP host 0.68945(0.000)***

    1.09312(0.023)**

    0.7193375(0.000)***

    1.276357(0.007)***

    0.7201158(0.000)***

    lnGDP home 0.93828

    (0.000)***

    0.93966

    (0.000)***

    0.9390046

    (0.000)***

    0.9360072

    (0.000)***

    0.9555236

    (0.000)***

    lnGDPCAP host 0.20578(0.300)

    -0.739452(0.468)

    0.1945107(0.321)

    -0.3901564(0.699)

    0.1449517(0.466)

    lnDistance -0.84545(0.000)***

    -0.8528(0.000)***

    -0.8865569(0.000)***

    -0.8995257(0.000)***

    -0.8962245(0.000)***

    Common language 2.4291(0.007)***

    1.87444(0.000)***

    2.387142(0.007)***

    1.881151(0.000)***

    2.421931(0.009)***

    Contiguity -0.49278(0.232)

    -0.04306(0.833)

    -0.5718995(0.157)

    -0.1636568(0.421)

    -0.5446089(0.194)

    Transition Index 0.24028(0.001)***

    -0.0848(0.562)

    0.3283385(0.000)***

    0.0599106(0.685)

    0.2686834(0.000)***

    WTO -0.25549(0.135)

    -0.5515(0.052)*

    -0.1929713(0.262)

    -0.5970304(0.034)**

    -0.2463965(0.150)

    BIT 0.42502

    (0.005)***

    0.50338

    (0.000)***

    0.5234234

    (0.001)***

    0.8333026

    (0.000)***

    0.4527339

    (0.006)***

    ICT 0.45236

    (0.087)*

    0.79268

    (0.001)***

    0.4807682

    (0.067)*

    0.8302537

    (0.000)***

    -

    0.0689651

    (0.770)

    IT 0.17418

    (0.524)

    0.55490

    (0.018)**

    0.1947564

    (0.474)

    0.5394726

    (0.020)**

    -

    0.3384541

    (0.197)

    SS -0.62424

    (0.012)**

    -0.50398

    (0.004)***

    -0.6498642

    (0.008)***

    -0.5165181

    (0.003)***

    -

    0.5626095

    (0.011)**

    BIT*Transition Index -0.140099(0.012)**

    -0.294981(0.000)***

    N. Obs 962 962 962 962 962

    R2within=0.35

    between=0.52

    overall=0.47

    within=0.41between=0.0

    2overall=0.22

    within=0.35between=0.5

    3overall=0.47

    within=0.43between=0.0

    7overall=0.25

    within=0.35between=0.

    50

    overall=0.45

    Note: p-values in parentheses; * - significance at 10%; ** - 5%; *** - 1%Year dummies not reported.

    25

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    26/37

    Table (3) presents Instrumental Variables estimation results. It starts with IV (2SLS) regressions

    followed by fixed-effects IV regression. Hausman test suggests country pair random effects, as a

    robustness check we also do pair fixed effects (column 5). In column (4) the interacted

    BIT*Transition Index is included. We simultaneously instrument four treaty dummies: ICT, IT,

    SS, and BIT. As discussed above, the instruments used are the sums of outside treaties that both

    host and home countries have corresponding to each treaty type (ICT, IT, SS, BIT). Tax treaty

    dummies are insignificant throughout IV specifications, except for SS treaty, which is negative

    significant in columns (1) and (2). BIT is the only treaty that stays significant throughout all

    specifications, and the effect of BIT is unchangeably positive. As can be seen from the table

    below, instrumenting treaties produces a very large coefficient in case we do not control for

    country pair effects, while the pair fixed effects and random effects estimates remain fairly

    reasonable. Given that the size of the coefficient increases, high FDI flows would decrease the

    chances of having a treaty. Indeed, if there is already a high FDI inflow, why would one want to

    sign a treaty?

    26

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    27/37

    Table 3. IV estimation results (logged FDI flows in $US 2000)

    Variables 2SLS (rat) 2SLS (sign)Pair RE IV

    (rat)Pair RE IV

    (interaction)Pair FE IV (rat)

    lnGDP host 0.0318697

    (0.905)

    0.3938

    (0.481)

    1.013625

    (0.130)

    0.4605846

    (0.058)*

    1.4177

    (0.071)*

    lnGDP home 1.233196(0.000)***

    1.1777(0.000)***

    0.6852966(0.614)

    1.089217(0.003)***

    0.2858(0.850)

    lnGDPCAP host 2.3945(0.014)**

    1.4212(0.272)

    -0.1580569(0.916)

    0.43603(0.374)

    -1.1032(0.408)

    lnDistance -0.150128(0.653)

    -0.1538(0.723)

    -0.3545127(0.874)

    -0.5040456(0.511)

    dropped

    Common language 0.50637(0.738)

    1.7379(0.194)

    3.785506(0.611)

    2.006902(0.466)

    dropped

    Contiguity 2.0292

    (0.022)**

    1.7283

    (0.107)

    -0.4026737

    (0.879)

    0.1156395

    (0.912) dropped

    Transition Index -.52427(0.152)

    -0.0874(0.791)

    -0.1123104(0.354)

    -0.1554021(0.483)

    -0.18403(0.173)

    WTO -0.09522(0.816)

    0.2946(0.685)

    -0.4881483(0.144)

    -0.6211766(0.025)

    -0.45214(0.119)

    BIT 6.4553

    (0.003)***

    6.8703

    (0.001)***

    1.672409

    (0.009)***

    1.677447

    (0.002)***

    1.4616

    (0.061)*

    ICT -0.18744

    (0.947)

    -3.6337

    (0.690)

    0.8162004

    (0.946)

    -1.605046

    (0.716)

    0.9662

    (0.920)

    IT -1.55113

    (0.617)

    -5.0019

    (0.566)

    -0.9587139

    (0.932)

    -3.282159

    (0.456)

    -1.0424

    (0.901)

    SS -5.5825

    (0.011)**

    -3.8844

    (0.075)*

    0.0011532

    (0.999)

    -1.0321

    (0.496)

    -0.51178

    (0.760)

    BIT*Transition Index 0.3204584(0.219)

    N. Obs 962 962 962 962 962

    Note: p-values in parentheses; * - significance at 10%; ** - 5%; *** - 1%Year dummies not reported.

    27

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    28/37

    Timing of the Treaty Effect

    There is no clear theoretical prediction about the timing of the treaty effect. A treaty can foster

    investment or itself be a response to an increase in the investment activity between two

    countries. The exact length of the lag or lead in the treaty effect is also an empirical question.

    In order to explore the timing of the treaty effect we augment the parsimonious specification

    with the dummies for each of the 3 years prior and each of the 3 years after the treaty was

    ratified. All years after the third year are designated with a separate dummy. The three year

    window is chosen based on the limited length of our panel because for a wider window the

    estimates of the effects closer to the beginning and end of the window become less precise. In

    this setup our reference category corresponds to years prior to the third year before the earliest

    treaty was ratified. In other words, a positive and significant coefficient on "BIT Yr ratify +1"

    implies that in the year after a BIT was ratified there was an increase in annual FDI flows

    relative to the period more than 3 years prior to the ratification. The results are presented in table

    4.

    We estimate three specifications: pair random effects, host random effect, and a pooled OLS

    specification. Our findings are catalogued below along with possible explanations.

    First, generally our results do not reveal an increase in investment before ratification. If anything,

    in one case we observe just the opposite and the levels of FDI drop before ratification, which is

    consistent with investors delaying their investments till ratification.

    Second, only the BIT's encourage FDI. The FDI flows increase in the year of ratification and

    remain higher afterwards. "BIT Yr ratify +3" is significant only at the 10.2% level in the pair

    random effect specification. This lower significance could be partially due to lack of variation

    because treaties ratified in or after 1999 will not have a Year+3 dummy by construction.

    Third, ICT and IT have a similar effect on FDI. Pair RE estimates reveal no effect at

    conventional levels of significance in all but one case. There seems to be a drop in FDI 2 years

    prior to a ratification of an ICT. The host RE (and OLS) suggest that ICT and IT both have a

    positive long term effect on FDI even though the exact timing is not statistically distinguishable.

    Fourth, there are indications that the effect of the social security treaty (SS) is negative. The

    preferred pair RE estimate picks up a decrease in the year of ratification. All specifications show

    a decrease in the level of FDI for the period of 4 years or more after the ratification.

    28

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    29/37

    Table 4. Timing of the Treaty Effect

    Variables Pair Random Effects Host Random Effects OLS

    lnGDPhost 0.712 (0.000) 0.684 (0.000) 0.684 (0.000)

    lnGDPhome 0.954 (0.000) 0.886 (0.000) 0.886 (0.000)

    lnGDPPChost 0.199 (0.324) -0.206 (0.147) -0.206 (0.147)

    EBRD Index 0.259 (0.000) 0.460 (0.000) 0.460 (0.000)

    lnDIST -0.797 (0.000) -0.723 (0.000) -0.723 (0.000)

    Language 2.467 (0.004) 2.170 (0.000) 2.170 (0.000)

    Contiguity -0.356 (0.362) -0.057 (0.788) -0.057 (0.788)

    WTO -0.161 (0.399) 0.024 (0.906) 0.024 (0.906)

    BIT Yr ratify -3 0.350 (0.203) 0.027 (0.935) 0.027 (0.935)

    BIT Yr ratify -2 0.213 (0.423) -0.032 (0.919) -0.032 (0.919)

    BIT Yr ratify -1 0.353 (0.166) 0.162 (0.58) 0.162 (0.581)

    BIT Year ratify 0.855 (0.000) 0.665 (0.01) 0.665 (0.010)

    BIT Yr ratify +1 0.487 (0.044) 0.425 (0.075) 0.425 (0.075)

    BIT Yr ratify +2 0.601 (0.011) 0.433 (0.057) 0.433 (0.057)

    BIT Yr ratify +3 0.402 (0.102) 0.255 (0.261) 0.255 (0.261)

    BIT After +3 0.591 (0.017) 0.538 (0.002) 0.538 (0.002)

    ICT Yr ratify -3 -0.549 (0.336) 0.226 (0.741) 0.226 (0.741)

    ICT Yr ratify -2 -0.965 (0.042) -0.476 (0.383) -0.476 (0.383)

    ICT Yr ratify -1 0.303 (0.497) 0.412 (0.405) 0.412 (0.405)

    ICT Year ratify 0.421 (0.350) 0.721 (0.136) 0.721 (0.136)

    ICT Yr ratify +1 0.486 (0.270) 0.535 (0.235) 0.535 (0.235)

    ICT Yr ratify +2 0.181 (0.679) 0.500 (0.264) 0.500 (0.264)

    ICT Yr ratify +3 0.064 (0.876) 0.367 (0.365) 0.367 (0.365)

    ICT After +3 0.359 (0.359) 1.226 (0.000) 1.226 (0.000)IT Yr ratify -3 -0.830 (0.176) 0.025 (0.972) 0.025 (0.972)

    IT Yr ratify -2 -0.762 (0.161) -0.535 (0.394) -0.535 (0.394)

    IT Yr ratify -1 0.052 (0.918) 0.119 (0.835) 0.119 (0.835)

    IT Year ratify 0.367 (0.426) 0.574 (0.255) 0.574 (0.255)

    IT Yr ratify +1 -0.278 (0.548) 0.104 (0.832) 0.104 (0.832)

    IT Yr ratify +2 0.358 (0.424) 0.679 (0.152) 0.679 (0.153)

    IT Yr ratify +3 -0.177 (0.673) 0.321 (0.444) 0.321 (0.444)

    IT After +3 -0.160 (0.683) 0.830 (0.007) 0.830 (0.007)

    SS Yr ratify -3 -0.090 (0.834) 0.218 (0.685) 0.218 (0.685)

    SS Yr ratify -2 -0.141 (0.745) 0.153 (0.777) 0.153 (0.777)

    SS Yr ratify -1 -0.220 (0.613) 0.171 (0.751) 0.171 (0.751)

    SS Year ratify -0.962 (0.024) -0.639 (0.213) -0.639 (0.213)

    SS Yr ratify +1 -0.645 (0.216) -0.071 (0.909) -0.071 (0.909)

    SS Yr ratify +2 -0.506 (0.243) 0.044 (0.930) 0.044 (0.930)

    SS Yr ratify +3 -0.220 (0.632) 0.126 (0.805) 0.126 (0.805)

    SS After +3 -0.769 (0.010) -0.652 (0.001) -0.652 (0.001)

    N.obs 962 962 962

    R-sq within = 0.369between = 0.520overall = 0.470

    within = 0.412between = 0.927overall = 0.502

    0.502

    Notes: p-values are in parentheses. All specifications include year dummies. The choice between fixed and

    random effects was made based on a Hausman test of null hypothesis that there is no systematic differencebetween coefficients.

    29

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    30/37

    CONCLUSION

    The majority of economic texts stress the intuitive notion that bilateral tax treaties and bilateral

    investment treaties should promote FDI. However, not all the empirical studies support either

    hypothesis. The most prominent works on tax treaties, on the contrary, even find negative

    impacts on FDI. Whether bilateral investment treaties have an impact is more ambiguous.

    Papers using a dyadic approach do not find any evidence of BITs effect on FDI, in contrast to

    papers that use a non-dyadic approach which tend to find a positive effect on FDI.

    Using OECD data we find that transition countries that have BITs with developed countries

    receive more FDI inflows from these countries. The effect is robust to various model

    specifications. On the contrary, tax treaties do not reveal any robust effect on FDI flows.

    This finding is particularly valuable for three main reasons. First, to our knowledge, this is the

    first study examining both bilateral investment treaties and tax treaties simultaneously in one

    model. Second, it focuses on transition countries, which allows for more homogeneity in the

    sample. This is especially important, since then we can account for possible omitted variable

    bias. Thirdly we tackle the endogeneity issue.

    This study also provides evidence that BITs function to some extent as substitutes forinstitutional quality. In the majority of estimations that we do with interaction, the interaction

    term between the BIT dummy and a proxy for institutional quality is negative and statistically

    significant, implying that the net effect of a BIT is smaller (but still positive) if the quality of a

    host countrys institutions is better.

    We do not find any robust effect of tax treaties on FDI. When we include a traditional tax treaty

    dummy, which is one if there is any tax treaty and zero if there is no treaty, the result is

    consistent with what we get in general when doing tax treaty categorization no significant

    effects are found. This can be explained by the two effects of tax treaties on FDI that might

    offset each other. One of them reduces double taxation, which should encourage FDI; the other

    prevents tax evasion through setting constraints on transfer pricing by MNEs, which might

    discourage FDI.

    Summing up, the main message we have for policy makers of developed and developing

    countries alike is that signing and ratifying BITs with developed countries does have the desiredpayoff of higher FDI flows.

    30

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    31/37

    Figure 1. Cumulative Number of Bilateral Investment Treaties.

    0

    5

    10

    15

    0

    5

    10

    15

    0

    5

    10

    15

    1990 1995 2000 1990 1995 2000 1990 1995 2000

    Bulgaria Czech Republic Hungary

    Poland Romania Russia

    Slovakia Slovenia Ukraine

    CumilativenumberofsignedBIT's(by

    host)

    YearGraphs by FDI host country

    Figure 2. Cumulative Number of Investment and Double Taxation Treaties.

    0

    5

    10

    15

    20

    0

    5

    10

    15

    20

    0

    5

    10

    15

    20

    1990 1995 2000 1990 1995 2000 1990 1995 2000

    Bulgaria Czech Republic Hungary

    Poland Romania Russia

    Slovakia Slovenia Ukraine

    IT and ICT BIT

    Cumulativ

    enumberofsignedtreaties

    Year

    Graphs by FDI host country

    31

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    32/37

    Table 3. Cumulative Number of Bilateral Investment Treaties by Groupings of FDI Source

    Countries.

    0

    5

    10

    15

    0

    5

    10

    15

    0

    5

    10

    15

    1990 1995 2000 1990 1995 2000 1990 1995 2000

    Bulgaria Czech Republic Hungary

    Poland Romania Russia

    Slovakia Slovenia Ukraine

    total with G7 countries

    with adjacent countries with European countries

    CumulativenumberofsignedBIT's

    Year

    Graphs by FDI host country

    Notes: 1) Canada is not among G7 countries due to availability of FDI flows data; 2) non-European countries are Korea, Japan, and the USA.

    32

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    33/37

    Appendices

    Appendix A. Number of BITs and DTTs concluded, cumulative and year to year, 1990-2004

    Source: WIR 2005

    33

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    34/37

    Appendix B. Direct investment cumulative outflows from OECD countries 1990-2003.

    34

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    35/37

    Appendix C. List of countries included in sample.

    Home countries Host countries

    Austria, Belgium, Denmark, Finland, France,

    Germany, Italy, Japan, Korea, Netherlands,

    Poland, Portugal, Spain, Sweden, Switzerland,

    United Kingdom, United States of America

    Bulgaria, Czech Republic, Hungary, Poland,

    Romania, Russia, Slovakia, Slovenia, Ukraine

    Appendix D. First-stage F-statistics

    Instruments BIT ICT IT SS

    sums 36.21 186.86 132.10 41.95

    outside treaties of hosts 23.55 7.99 6.63 31.87

    35

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    36/37

    References

    Bergstrand, J., Egger, P. 2005. A-Knowledge-and-Physical-Capital Model of InternationalTrade, Foreign Direct Investment, and Outsourcing: Part I Developed Countries. Universityof Notre Dome Working paper.

    Blonigen, B., Davis, R. 2000. The Effects of Bilateral Tax Treaties on U.S. FDI Activity. NBERWorking Paper #7929.

    Blonigen, B., Davis, R. 2004. The Effects of Bilateral Tax Treaties on U.S. FDI Activity.International Tax and Public Finance, 11 (5), 601-622.

    Blonigen, B., Davis, R. 2002. Do Bilateral Tax Treaties Promote Foreign Direct Investment?NBER Working paper # 8834.

    Davis, R. 2004. Tax Treaties and Foreign Direct Investment: Potential versus Performance.International Tax and Public Finance, 11 (5), 775-802.

    Doernberg, R. 1997. International Taxation in a Nutshell, 3rd edition. St. Paul: West Publishing.

    Globerman, S., Shapiro, D. 2002. Global Foreign Direct Investment Flows: The Role ofGovernance Infrastructure. World Development, 30, 1898-1919

    Goryunov, D., 2004, The Effectiveness of FDI Promotion in Transition Countries, EERC MAThesis

    Hallward-Driemeier, M. 2003. Do Bilateral Investment Treaties Attract FDI? Only a bit andthey could bite. World Bank, DECRG

    Louie, H., Rousslang D. 2002. Host Country Governance, Tax Treaties, and American DirectInvestment Abroad. Mimeo.

    Neumayer, E., Spess, L. 2005. Do Bilateral Investment Treaties Increase Foreign DirectInvestment to Developing Countries? World Development Vol. 33, No. 10, 1567-1585.

    Monterrey Consensus 2002, UN International Conference on Financing for Development, 18-22

    March 2002, Monterrey

    Salacuse, J. , Sullivan N. 2005. Do BITs really work? An Evaluation of Bilateral InvestmentTreaties and their Grand Bargain. Harvard International Law Journal, 46 (1).

    Tobin J., Rose-Ackerman, S. 2005. Foreign Direct Investment and the Business Environment inDeveloping Countries: The Impact of Bilateral Investment Treaties. Yale Law School Center,Economics and Public Policy Research Paper No. 293

    UNCTAD. 1998. Bilateral Investment Treaties in the Mid-1990s, United Nations Conference onTrade and Development, New York.

    UNCTAD. 2003. World Investment Report 2003 FDI Policies for Development: National andInternational Perspectives, New York and Geneva: United Nations.

    36

  • 8/7/2019 _The Impact of Policies on FDI flows to Transition Countries

    37/37

    UNCTAD. 2005. World Investment Report 2005 Transnational Corporations and theInternationalization of R&D, New York and Geneva: United Nations.