“FDI Determinants in the Southern Euro Area”

(1)

University of Groningen

Faculty of Economics and Business

“FDI Determinants in the Southern Euro

Area”

Master Thesis in Advanced International Economics and Business (IE&B) Groningen, August 2010 Author Riccardo Tresca S1942409 R.Tresca@student.rug.nl University of Groningen

(2)

Table of Contents

1. Introduction 4

2. FDI Dynamics in SEA Countries 5

3. Literature Review 7

3.1 Economic Factors 8

3.2 Institutional Factors 10

4. The Methodology 13

5. Data Description and Sources 15

5.1 Dependent Variable 16

5.2 Economic Variables 17

5.3 Institutional Variables 18

5.4 The Model 19

6. Results 21

7. Implication for SEA 29

9. Conclusion and Limitations 31

References 33

Data Sources 38

(3)

“FDI Determinants in the Southern Euro

Area”

Abstract

This paper analyzes the relative low attractiveness of the Southern Euro Area (SEA): Spain, France, Italy, Greece and Portugal for foreign direct investments. After having defined FDI location determinants common to a narrow group of 15 industrialized countries, this study compares the SEA countries’ endowment of factors affecting FDI. Results of empirical estimates confirm many of the theoretical arguments proposed in the literature section. In the specific, it is showed that Mediterranean countries lag in terms of R&D and institutional background. They invest less in research, are more corrupted, have stricter labour regulations and possess less efficient bureaucracies. In order to attract larger volumes of FDI, higher emphasis should be paid on these issues, and preferential policies should be devoted to strengthen local competitiveness and reduce prominent obstacles to foreign capital.

(4)

1. Introduction

Foreign direct investments (FDI) are considered one of the most striking features of the global economy (Agiomirgianakis et al., 2006). During the last two decades, they increased at a pace of about 20 percent in the 80‟s and 40 percent by the end of the 90‟s, much faster than World GDP and trade flows between countries (Blonigen, 2007). This global upsurge of FDI has been accompanied by a significant number of academic studies seeking to explain their determinants and consequences. A set of stylized facts has been introduced by the literature. In particular, it is proposed that, despite the bulk of investments directed to developing countries is progressing at an outstanding rate, FDI originate to a large extent from developed countries and they go predominantly to developed countries (Barba Navaretti, G. and Venables, J. 2007).

In general terms, among World‟s developed areas, the European Union has become the principal outward investor and inward host area of FDI (UNCTAD, 1993). However, within its borders, some countries have been less successful than others in attracting foreign capital. This especially holds for the Southern Euro Area (SEA) whose countries, France, Italy, Spain, Portugal and Greece, continue to lag behind the internalization process and receive limited volumes of FDI, a percentage much lower than that in many other European neighbors (Figure 1).

Figure 1. FDI Inflows into EU countries. Source: UNCTAD.

(5)

1969; Reuber et al., 1973; Behrman and Wallender, 1976) and raise competitive effects (Barba Navaretti and Venable, 2007). In this context, there are many arguments why a country should enhance its location-specific advantages and attract a higher share of FDI. Multinational firms are indeed very different from local firms, and they bring to host countries a bundle of characteristics that are not necessarily available locally: technologies, management, brands, procedures, market access and so on (Barba Navaretti and Venable, 2007).

A number of studies have focused on motives that make a location more attractive than others. Traditionally, the existing literature considers two broad set of location advantages. The first one pertains to countries‟ economic factors, such as market size, labor costs and R&D. The second one consists of institutional factors, such as fiscal conditions, quality of bureaucracy, labour protection legislation and level corruption (e.g. Wheeler and Mody, 1996; Wei, 1997; Habib and Zuravicki, 2002; Nicoletti et al., 2003; Bénassy-Quéré et al., 2007). By taking into account both set of determinants, the purpose of this paper is to find an answer to the following research question:

“Why do the SEA countries attract a relatively low level of inward FDI?”

This study tries to make some contribution to the current literature. First, it analyzes the impact of institutional factors on bilateral FDI which is relatively a new area of interest. Second, following recent academic developments, e.g. Habib and Zurawicki (2002) on corruption and Görg (2005) on labor market regulation, it implements a complete analysis of institutional determinants. Third, differently from previous research, e.g. De Santis and Vicarelli, (1999) on Italy, Bitzenis et al. (2007) on Greece or Rodríguez and Pallas (2008) on Spain; it covers a wider spectrum of European countries.

While addressing the above research question, the approach followed in the study develops through two steps. First, we estimate a general model of FDI location determinants. Second, we investigate to what extent this model can explain the relatively low levels of FDI inflows into SEA. The remainder of the paper is organized as follows. The next section reviews the inward FDI position of the SEA. Section 3 introduces the theoretical background of the study. Section 4 proposes the methodology. Section 5 describes data and sources we rely on. Section 6 shows the estimation results. Section 7 discusses about results. Section 8 presents conclusions and limitations.

2. FDI Dynamics in SEA Countries

(6)

from US$ 39,174 billion in 1990 to US$ 134,481 in 2005. This amount represented about the 20% of the total European FDI inflows. The percentage in 1992 was much higher and reached the peak of 35%, while the worst FDI inward position has been recorded in 1999 (8%) and 2000 (9%), that is, in the period prior the launch of the single European currency since it was not clear whether they would take part of it.

Taking into account the GDP of the single countries, the weak performance of SEA is even more striking. Data are displayed in Figure 2. The average FDI inflows as a percentage of GDP varied between 1.3% in 1990 and 2.8% in 2005. As a group, the SEA countries are still lagging behind their European neighbors. However, also among the SEA, there seem to be few more successful stories. During the period 2001-2005 the most important receiver of foreign investments has been Spain, with a FDI/GDP ratio equal to 3.6%, exceeding not only FDI inflows of other SEA countries, but also FDI positions of other major European economies, such as Germany, Sweden and UK. On the other hand, also France and Portugal have gradually become more interesting to foreign investors. France has recorded its highest share of FDI inflows (3.9%) in 2005, while Portugal had a peak of foreign activities in 2001 (5.8%).

Figure 2. SEA FDI Inflows (Percentage of GDP). Source: UNCTAD.

At the same time, from the picture above, it clearly emerges the poor performances of the Italian and Greek economies, the lowest among the OECD countries. However, it is worth noting that while FDI into Italy more than doubled from the 1990-1995 to 2001-2005 period1; FDI in Greece decreased in the second half of the 90s and afterwards remained constant. Considering the size of their internal markets, this pattern suggests that there is scope for further attracting foreign capital in these two territories.

1

(7)

3. Literature Review

FDI is generally identified as “investments made by multinational business enterprises in foreign countries to control assets and manage production activities in those countries” (Padma, 1999). In the last decades different theories have been advanced to explain these forms of investments. Unfortunately none of them can boast to be a general theory that includes all aspects of MNEs‟ activities (Dunning, 2008), and the main reason is that there are too many elements that can potentially drive FDI.

Despite this gap in the literature, it is widely agreed that MNEs engage in outward FDI when a set of three necessary conditions is satisfied, that is, ownership advantage, location advantage and internalization advantage. This general framework, known as the eclectic (OLI) paradigm has been introduced by John Dunning at the end of the 80s. According to the OLI paradigm, an investing firm has to own something exclusive, unique, which make it recognizable in the foreign market (Ownership). The host market has to offer something to the foreign investor (Location). The MNE has a direct advantage in carrying out the activity by itself instead of licensing it (Internalization). This study focuses on one of the three necessary conditions expressed in the OLI paradigm, viz. the location advantage. Location advantage (L) refers to the host country‟s quality of business environment, and it includes a wide spectrum of elements, such as factor prices, market access, trade barriers, transport costs and institutional environment (Dunning, 1988).

In the broad discussion of FDI location determinants across countries, it is useful to make a theoretical distinction between Horizontal FDI (HFDI) and Vertical FDI (VFDI)2. The first one pertains to market-seeking FDI and duplication of activities in the host market, in order to have a better access to that host market. The second one involves efficiency-seeking FDI since part of the production process is carried out in the host economy to take advantage of its cheap factors of production (Barba Navaretti, G. and Venables, A. J., 2007). Despite there exist practical difficulties in separating the data into HFDI and VFDI3, the dominance of developed economies in both FDI inflows and outflows seems suggesting that the main global strategic intent of firms investing abroad is the penetration of the foreign market4.

Theoretical models employed to explain two phenomena presented above may in some cases produce conflicting predictions. As an example, the VFDI model predicts the opposite than the HFDI model concerning trade and similarity of countries, since it

2

Dunning (1998) has later widened this distinction by adding other motives for which MNEs undertake FDI, namely, resource-seeking, assets-seeking and export-oriented FDI.

3

The distinction is not always clear-cut since some investments may possess characteristics of both kinds of FDI, e.g. investments in China implemented for costs-minimizing reasons, but also a vehicle to serve the foreign market.

4

(8)

predicts that investments decrease when trade costs augment, or when countries converge in similarity (Barba Navaretti and Venables, 2007). This clearly poses serious challenges for applied research and interpretation of results. The existing literature on FDI has therefore developed three approaches to deal with theoretical divergences (Barba Navaretti and Venables, 2007). The first one tries to split the data in investments that are horizontal and investments that are vertical. The second one tries to estimate a statistical model that encompasses both phenomena. The third one, followed by this study and the majority of previous works, accept that data contain all sorts of investments and it does not try to distinguish between their underlying theories.

Theoretically, when assessing the “value” of a location, the literature considers two broad set of location-related variables. The first one pertains to countries‟ economic factors, such as market size, labor costs and R&D expenditures. The second one consists of institutional factors, such as fiscal conditions, quality of bureaucracy, government‟s influence and level corruption (e.g. Wheeler and Mody, 1996; Wei, 1997; Habib and Zuravicki, 2002; Nicoletti et al., 2003; Bénassy-Quéré et al., 2007). Our discussion of FDI location determinants follows this theoretical distinction.

3.1 Economic factors

Market size has been widely considered as the most important FDI determinant. As suggested by the theory, the larger is the host country the higher is the probability that MNEs will prefer to serve locally the foreign market instead of exporting – since it will be easier recouping the fixed costs of setting up a foreign plant (Barba Navaretti and Venables, 2007). Therefore, market size directly affects the return of an investment (Zheng, 2009). In line with the theory, empirical works have extensively supported these predictions (Globerman et al., 2004).

Corporate taxation in the host country is expected to have a clear and unambiguous effect on FDI, despite some studies, such as Brainard (1997), produced abrupt (but insignificant) results, with high corporate taxes increasing affiliate sales relative to exports. Ample evidence is provided by more recent works (e.g Hines, 1999; Basile et al., 2009 and Bénassy-Quéré et al., 2007) which pointed out that corporate tax regimes influence the choice of where to locate investments.

(9)

Trade, as pointed out by Grosse and Trevino (1996) and Zheng (2009), is expected to affect MNEs‟ investment decisions. Even though many studies suggest that trade substitutes rather than complements FDI for individual firms (e.g. Bloningen, 2001), the main bulk of the literature proposes that MNEs would probably prefer to invest more in countries with which there are more intense commercial relationships, since exports may be used to supply foreign affilitiates with inputs or partial products5 (e.g. Barba Navaretti and Venables, 2007; Zheng, 2009; Lipsey and Weiss, 1981).

The macroeconomic stability is proposed to be a determinant of FDI. It is usually proxied by the inflation rate in the host country (Herdal F. and Tatoglu E., 2002). In particular, a higher inflation rate should discourage multinational firms from investing in the country bacause of the volatility and instability of its internal prices (e.g. Nunes L. C. et al., 2006; Ali Al-Sadig, 2009)6.

Since labor costs represent a large component of total production costs, the literature has always brought them into the analysis of factors affecting FDI location. Despite some studies did not find any significant effect (e.g. Owen, 1982; Wheeler and Mody, 1992), empirical results have usually supported the main theoretical expectation that higher wages have an adverse effect on inward FDI (e.g. Fung et al., 2005; Culem, 1988). In the analysis of multinational firms‟ behaviors, Dunning (1998) emphasizes that the location of FDI may be driven by the search of strategic assets. Empirical and anecdotal evidence supports this view that firms may use foreign investments to acquire location-specific assets and set up activities in Research and development-rich countries (Barba Navaretti and Venables, 2007; Fosfuri and Motta, 1999; Siotis, 1999). More recent analysis (Dinning and Narula, 2005) suggest that, along with the ability to generate new technologies and/or products in a foreign location, another reason for MNEs to invest abroad is the need to monitor technological developments. These arguments are particularly relevant for the analysis of SEA‟s investment position. From Table 1 below, it can be clearly recognized the poor R&D performance exhibited by the SEA countries. With the only exception of France, the Mediterranean area confirms its deficiencies in the R&D field: Portugal 3.01, Spain 3.80, Greece 2.84 and Italy 3.11, against an average of 7.23 reported by other 8 OECD countries7 (Table 1).

5

Papers that distinguish between VFDI and HFDI give opposite predictions concerning the two underlying theories.

6

Normally, along with inflation, the exchange rate is expected to play a role. We took it out from our analysis since only Euro currency countries are analyzed.

7

(10)

Table 1. Competitiveness Indicators (1990-2007 averages)

R&D1 Bureaucracy Start Business Pol. Stability Corruption Labour Reg.2

France 6.32 6.93 5.42 6.69 2.99 7.01 Portugal 3.01 5.74 5.58 7.28 3.71 6.95 Spain 3.80 6.82 5.85 5.90 4.07 7.13 Greece 2.84 5.39 5.18 5.94 5.32 6.29 Italy 3.11 5.73 5.59 6.23 5.56 7.54 Netherlands 4.97 7.36 7.75 7.49 1.14 6.91 Austria 5.11 6.83 5.65 7.48 2.34 5.79 Denmark 6.91 8.22 7.18 7.44 0.56 2.47 Finland 11.02 8.23 8.60 7.83 0.59 6.06 Germany 6.34 7.06 6.23 7.24 2.02 7.18 Ireland 4.67 7.43 7.65 7.35 2.35 5.34 Sweden 8.88 8.18 6.83 7.66 0.82 6.90 United Kingdom 6.07 7.52 8.25 6.71 1.55 3.72 US 8.56 7.38 8.65 6.55 2.38 2.93 Japan 9.82 7.11 5.91 7.16 3.02 5.73

1) Number of researchers per 1000 labour force units 2) Strictness of regulations to hire and fire workers

3.2 Institutional factors

The very first attempt to study the effect of quality of institutions on inward FDI has been made by Wheeler and Mody in 1992. Taking into account 13 different country-related risk factors, such as quality of bureaucracy, level of corruption and political stability, they did not find a significant impact on US firms‟ location choices. However, this result is partly conditioned by the fact that they also considered non-institutional-related factors as inequality and living environment of expatriates (Bénassy-Quéré et al., 2007). Later studies have also tried to identify the relationship between formal institutions and volumes of FDI inflows, but results are often controversial and contradictory due to difficulties of measuring qualitative variables such as institutions and due to the different methodologies and samples employed (Bloningen, 2005).

(11)

al., 2009). As can be seen from Table 1 above (column 2), which compares the quality of bureaucracy8 across 15 developed countries, SEA countries have relatively less efficient public administrations – lower scores denote less efficient national bureaucracies. For that reason, we expect them to receive lower volumes of FDI.

According to the World Bank (2004), a country‟s business climate, identified by specific regulation and policies, can promote or discourage foreign investments. The business climate is measured by four variables, that is, procedures necessary to set up an activity, associated time, cost and minimum capital requirement. The ease of entry in a foreign location is a factor that influences the decision to undertake an investment (Root and Adned, 1978; Habib and Zurawicki, 2002). Bitzenis et al. (2009) argues that the average time to start up a business in OECD countries is 34 days, longer registration time may discourage foreign activities. Also in this case, when looking at Table 1 above, the interpretation is straightforward: SEA economies have the slowest procedures necessary to start-up a business9. Hence we propose this feature to be a major impediment for FDI flows into SEA.

Political stability is another element that strongly conditions MNEs‟ preferences (Kobrin, 1976). Basi (1963), with an extensive survey of international executives, highlights that, along with market potential, political stability is the most important aspect in foreign investments decisions. Aharoni (1966) obtains similar findings interviewing international personnel in more than 30 multinational companies. On the other hand, Bennett and Green (1972), analyzing a sample of 46 countries, find out that political instability is not a decisive FDI obstacle. They do not find any significant correlation and conclude that US firms‟ investment decisions are taken on the basis “other overriding factors”. However, these findings have been methodologically doubted by Kobrin (1976) who points out the contradiction between the emphasis they give to the political climate in the survey responses and in the findings. In line with Kobrin, Aristidis et al. (2009), Jun and Singh (1996) and Habib and Zurawicki (2002) conclude that political stability is an imperative for profitability and long-run success. Accordingly, we propose the poor stability of SEA political systems10 (Table 1) to be an important barrier to foreign capital inflows.

Corruption is generally defined by the literature as the “abuse of public power to obtain private benefits” (World Bank, 2006). Whereas corruption in the public sector is the dominant theme, Coase (1979) sustains that it is a widespread phenomenon also between private parties. Despite some studies (Leff 1964, Leys 1965 and Huntington 1968) have demonstrated that in some cases corruption may confer beneficial effects to the economy by acting as an “efficient grease”, the majority of the literature seems more orientated

8

Measured by the Fraser Institute index with a 0-10 scale. 9

Again, an index from the Fraser Institute is employed. 10

(12)

toward the moral and economic condemn. The consensus is reinforced by the bulk of studies that investigate the economic outcomes of corruption. Effect of corruption on FDI is relatively a new area of interest. Theoretically, it is advanced that it may hinder FDI inflows for several reasons. First, it may be an obstacle when foreign investors consider corruption as morally wrong and decide to stay away from contexts where it is an extensive problem (Habib and Zurawicki, 2002). Second, as corruption hampers the development of a fair market, it does not guarantee equal treatment and equal access among the economic forces which operate in the country (Boatright, 2000). Finally, corruption can be considered as a “tax on profits” since foreign investors have to pay extra costs in the form of bribes in order to get licenses or government permits to conduct investments (Bardhan, 1997). Consequently, as corruption is a severe problem for SEA economies11,we expect it to negatively affect FDI inflows into that region.

Aspects regarding labor legislation are often considered since their effects can more than offset advantages arising from differences in factor endowments and factor costs (Parcon, 2008). According to Görg (2005), labour market regulations played a pivotal role for the location of US firms‟ outward activities during the period 1986-1996. In his research he reveals that nations with less stringent labour regulations receive a higher share of American investments. Benassy-Quere, Coupet and Mayer (2007) include three different measures of the degree of labour regulation in their study12, and conclude that in general terms strict labour market regulations deter FDI inflows. Similar results are also reached by Javorcik and Spatareanu (2005) and Haaland et al. (2002). Nicoletti et al. (2003) suggest that employment protection legislation (EPL) and labour income taxation can affect FDI patterns in the same way than increasing product market regulations do, that is, by augmenting the relative prices of different products or by lowering the expected return of investing in a given country. Parcon (2008) makes a distinction between wage and non-wage costs13. He argues that the mainstream literature frequently measures labour costs using only wage costs; in such a way, the fact that labour costs are formed by both wage and non-wage costs is neglected. Also from Table 1, a striking comparison between the various countries can be obtained. SEA economies are the ones with the strictest employment protections along with Germany, Sweden and the Netherlands14. Thus we suggest it to be an important impediment to inward investments into the Southern Euro area.

11

Table 1 shows up that Spain, Italy, France, Portugal and Greece have the highest CPI (Corruption Perception index) within countries in the dataset.

12

The variables taken into consideration are enforcement of labour laws, formal constraints on hiring and firing, regulation of labour market.

13

Non-wage costs pertain to hiring costs (such as payroll taxes, retirement and contribution funds, health insurance and other obligation) as well as firing costs (such as severance payments and penalties).

14

(13)

4. The methodology

To explain the relatively low attractiveness of SEA countries, we first estimate a general model of FDI location determinants. Subsequently, we investigate to what extent this model can explain the relatively low levels of FDI inflows in SEA‟s. In the general model, we control for the effects of market size, trade openness, corporate tax rate, labour costs differentials and bilateral trade. On basis of the preceding discussion, the following hypotheses are proposed:

H1: There is a positive relationship between FDI inflows and R&D expenditures. H2: There is a negative relationship between FDI inflows and the level of corruption. H3: There is a negative relationship between FDI inflows and strict labour regulations. H4: There is a positive relationship between FDI inflows and political stability.

H5: There is a positive relationship between FDI inflows and quality of bureaucracy. H6: There is a positive relationship between FDI inflows and ease of entry in a foreign location.

A comprehensive analysis of FDI determinants is conducted on a sample size of 15 industrialized countries. Accordingly, after having tested the significance of the variables proposed, a comparison between host countries‟ endowment of factors affecting FDI is carried out to highlight SEA‟s location disadvantages.

In the specific, we take into consideration 13 host economies, the EU15 with the exception of Luxemburg and Belgium15, and 15 investing economies, that is, the EU13 plus United States and Japan. Our choice is motivated by three different reasons. First, these nations offer consistent data for both FDI and independent variables. Second, they can be considered as a quite homogeneous set of industrialized countries, likely to share similar investment determinants. Third, they are the main investors in our area of interest, the SEA, representing on average more than 80% of the FDI into Spain, Greece, Italy, France and Portugal.

Regarding the time frame, we construct a dataset for the period 1990-2007, considering the availability of institutional data as well as the different definition of FDI employed by countries before the 90s16. In total, a maximum of (13 x 15 x 18) 3510 observations is expected. However, because of missing values related to some countries or years, fewer observations are produced.

In this study, panel data estimation is employed. Using panel data estimation has several advantages which can be summarized in three points. First, it is more informative and

15

Because of the lack of sufficient data. 16

(14)

reduces the multicollinearity among explanatory variables, therefore enhancing efficiencies of econometric estimations. Second, it is able to account for country heterogeneity (or individual differences), whereas cross section estimates cannot – heterogeneity may lead to biased results and generate serious misspecification (Baltagi, 2005). Third, differently from cross section and time series studies, a panel data technique permits to control for factors that vary across entities but do not vary over time, as cultural variables, or factors that could cause omitted variable bias if they are omitted (Hsiao, 2003).

With panel data, a choice between the fixed effects, random effects and pooled OLS model has to be made. Employing the fixed effects model has several advantages compared to both the random effects and pooled OLS model. First of all, the fixed effects specification allows us to control for unobserved country‟s heterogeneity and model misspecification, which could not be possible with the pooled technique. Kimino et al. (2007), comparing the pooled OLS with the other two models, conclude that the pooled regression is the least reliable. Similar conclusions are reached by Bloningen et al. (2005) who point out that pooled OLS may produce inappropriate estimates of parameters because of the omission of unobserved country-specific effects. Also other studies (e.g. Zheng, 2009; Kimino et al., 2007; Garretsen and Peeters 2007) reject the pooled OLS in favour of the fixed effects method. In addition, using the fixed effects allow us to minimize the problem of heteroskedasticity which the pooled OLS technique suffers from.

Normally, for this kind of analysis, the fixed effects model is also preferred to the random effects model for both economic and econometric reasons. Countries are highly complex systems, and a single econometric model is unlikely to fully represent their macroeconomic, social and institutional FDI determinants. The fixed effects technique compensates for this, controlling for the unmeasured explanatory variables or unobserved heterogeneity and, differently from the random effects, it does not assume that individual countries effects and the regressors are not correlated – this is more than justified by economic theory as no macroeconomic variable can be considered perfectly exogenous (Kimino et al., 2007). A further point in favour of the fixed effects model comes from its applicability. As argued by Baltagi (2005), this method is suitable if a defined set of countries is analyzed and inferences are restricted to the behaviors of these countries. In other words, our set of 15 OECD countries cannot be considered to represent a random sample of the nearly 200 sovereign nations in the World.

(15)

on the contrary, these characteristics are retained in the panel regression with random effects. Furthermore, with fixed effects estimations we do lose n degrees of freedom. Statistically, to determine which technique is more appropriate, the Hausman test is usually run. In section 7, results from the Hausman test are presented.

Some related diagnostics checks for econometric models are those for heteroskedasticity, autocorrelation and normality17. Heteroskedasticity arises when “variances for all observations are not the same” (Hill et al., 2001). In presence of heteroskedasticity the usual t statistics or F statistics cannot be used to draw inferences since the standard errors of the estimates are biased. We test for heteroskedasticity examining plots of residuals and Breusch-Pagan test results18. We conclude that heteroskedasticity is probable; therefore, to overcome the problem, when performing the fixed effects estimation, we use White‟s heteroskedasticity-consistent standard errors (or robust standard errors) which allow us to compute correct interval estimates and corrected values.

The autocorrelation problem occurs when residuals are correlated with one another (Hill et al., 2001). If residuals are correlated their estimations are biased and, as a result, the estimated t ratios are unreliable (Gujarati, 2004). We test for this using the Wooldridge (2002) test for autocorrelation in panel data. We do not reject the null hypothesis of no first-order autocorrelation at 5% level of significance. We repeat the test for all other models, and estimations generate similar results, leading us to the conclusion that errors terms are not correlated over time.

Finally, we check for normality with the Jarque-Bera test based on skewness and kurtosis19. Estimation suggests that residuals are not normally distributed. However, as proposed by the same OLS theory, this assumption can be relaxed with large sample sizes, meaning that tests and confidence intervals are still valid (asymptotically) whether the data are normally distributed or not (Hill et al., 2001, Gujarati, 2004). Since our sample is made of 2221 observations, the assumption of normality can be relaxed.

5. Data Description and Sources

As previously stated, this study on FDI determinants employs panel data covering 13 developed host country and 15 developed home country for a total of 18 years

17_{We implement diagnostic checks though our main goal is not estimating a perfect model explaining FDI.} 18

The LM test for heteroskedasticity follows this procedure. It regresses the squared error term with a dependent variable, it generates the test statistics and its corresponding value, and it assess whether the p-value is less than the level of significance. If the p-p-value is lower than the chosen level of significance, we conclude that there is significant heteroskedasticity, otherwise we conclude the opposite.

19

(16)

2007). The dataset built up for the analysis relies on secondary data obtained from different sources (Table 2). Below, we give a description of data used in the empirical part.

Table 2. Data and sources

Variables Description Expected Source

Sign

Log FDI Log of Bilateral FDI flows OECD online database

on FDI

Log GDP/Capita Log of of GDP/capita in US + IMF online database

Dollars at 2005 constant prices

GDP GDP level in the Host country + IMF online database (US$, 2005 constant prices)

Openness Export plus Import divided by + IMF online database

Real GDP (2005 constant prices)

Tax Host country statutory corpora- - OECD online statistics te income tax rate

Trade Export flows from home to host + OECD Stan database

country (US$, 2005 constant prices)

Inflation Host country annual percentage - OECD online statistics

increase in consumer prices

Labour Costs Diff a Difference in annual labour cost + OECD online statistics between home and host country

R&D Diff b Researchers number per 1000 + World Bank

develop-labour force (difference Host- ment indicators Invest economy)

Start Business Ease of entry in a foreign loca- + the Fraser Institute

tion (scale from 1 to 10)

Bureaucracy Quality of host country's bureau- + the Fraser Institute

cratic apparatus (from 1 to 10)

Corruption Corruption Perception Index - Transparency

Interna-(from 1 to 10) tional

Hiring/Firing Regulations Strictness of employment prote- - the Fraser Institute ction legislation (OECD index)

Political Stability Political stability and absence of + World Resources

violence (scale form 1 to 10) Institute

a) Difference in labour costs between investing and host country b) Difference in number of researchers between host and home country

Ec on omi V ar iab le s In sti tu ti on al V ar iab le s 5.1 Dependent Variable

(17)

the natural logarithm of FDI (Log of FDI) in order to “render the distribution nearly normal and the error term homoskedastic”. Source of our data is the OECD database on international direct investments.

5.2 Economic Variables

Concerning the economic variables, the following measures are employed. The variables GDP and GDP per capita capture the attractiveness of a market in terms of its internal demand. The higher the internal demand is, the more a country is able to attract foreign investors. Like FDI flows, market size is measured by the natural logarithm of the GDP per capita, expressed in US Dollars at 2005 constant prices, plus its Gross Domestic Product in US trillions. Data are obtained from the IMF online database. As previous studies did (e.g. Habib and Zurawicki, 2002; Zheng, 2009; Ang J. B., 2008), we also try to assess the impact of the host market growth rate. However, the estimation output suggests this variable not to be significant; consequently we only used GDP and Log GDP/capita as a proxy for market size.

Also from the IMF online database, we include data on countries‟ openness to trade. As in Ang J.B. (2006), Openness proxies the extent to which a country is economically open. It is defined as export plus import divided by the real GDP at 2005 constant prices.

The variable Tax reflects the basic central government statutory corporate income tax rate recorded in the host economies. Conform the literature (Barba Navaretti and Venable, 2007), we expect it to play a significant role in affecting MNEs‟ location decisions. Data on corporate tax rates are gathered from the OECD online statistics.

In order to assess whether direct investments are complement or substitute of trade, we brought the variable Trade into our analysis. This measure represents exports from the home country to the host country, expressed in US Dollars (billions) at 2005 constant prices. As pointed out in section 4, we expect MNEs to invest more in countries which they have more intense commercial relationships with. Trade flows data are obtained from the OECD Stan database.

Inflation intends to capture the macroeconomic stability of the host economy. It is measures as the annual percentage increase in consumer prices, and as previously discussed we expect it to be negatively related to the dependent variable. Data are obtained from the OECD online database.

(18)

Countries with a high number of researchers are a potentially more attractive location for firms seeking skilled labour force. The R&D Diff variable proxies the differential in Research and Development expenditure between the host and home country. As in De Santis and Vicarelli (1999), it is expressed in terms of the number of researchers per 1000 labour force – difference host-home economy. For this measure, we relied on the World Bank development indicators.

5.3 Institutional Variables

Regarding the institutional variables, different indexes are employed to test our hypothesis. Start Business represents the ease of entry in a foreign location. This measure is collected from the Fraser Institute, an independent and international research organization based in Canada and US, whose prime goal is to assess the impact of markets and government interventions on individuals‟ welfare20. In the specific, the Start Business index is based on another index, that is, the World Bank‟s Doing Business measure of the amount of time and money necessary to set up a new activity. It assesses the complexity of the regulatory framework a foreign individual has to face when trying to start up a business in a given country. It ranges from 0 to 10, with a score close to 10 indicating a well efficient national regulatory framework.

The same scale and the same source are used for the variable Bureaucracy, which considers the amount of time the management of a company usually spends in dealing with government bureaucratic practices/red tapes. It measures the quality of national bureaucracies. The higher the score is, the better organized is the bureaucracy of a country.

Hiring/Firing Regulations is the variable used to assess the effects of labour regulations on FDI inflows. Again, data relies on the Fraser Institute database. In the specific, this indicator measures costs and procedures necessary to hire and dismiss individuals or groups of workers with fixed or temporary contracts. We employed the inverse of the scale used by the Fraser Institute, so that the higher the score is, the stricter the internal labour regulations are21.

The host country level of corruption (Corruption) is measured with the Transparency International Corruption Perception Index (CPI). Transparency International is an influential NGO which ranks the level of corruption in more than 180 countries. The CPI index is drawn upon different surveys from ten independent institutions, such as Freedom House, Columbia University and World Economic Forum. It scores the freedom from corruption with a 10-points scale. In this study we employ the inverse of the scale used

20

Many studies relied on the Fraser Institute indexes. Examples are Sciantarelli F. (2005), Bénassy-Quéré et al. and Nyström K. (2007).

21

(19)

by Transparency International so that, a value of 0 indicates a corruption-free society, while a value of 10 indicates a highly corrupted environment22.

Finally, Political Stability refers to the extent to which the host countries have a stable political government. We relied on the World Resources Institute political stability and violence index, computed taking into consideration the percentage of politicians who drops from the government. Very low scores in this index indicate that citizens cannot count upon continuity of national policies or possibility to democratically select and replace persons in charge. The variable proposed ranges from 1 to 10, the higher is the score, the more stable is the host country23.

5.4 The Model

In the empirical analysis, we control for the effects of market size, trade openness, corporate tax rate, labour costs differentials and bilateral trade. Drawing upon previous studies24, the following model is put forth in order to test our hypotheses25:

jt 12 jt 11 jt 10 jt 9 jt 8 ijt 7 ijt 7 jt 6 ijt 5 jt 4 jt 3 jt 2 jt 1 0 ijt Business Start β y Bureaucrac β Regulation ing Hiring/Fir β Stability Political β Corruption β Diff D & R β Diff Costs Labour β Inflation β Trade β Openness β Tax β GDP β ) pita Log(GDP/Ca β α ) Log(FDI

We mainly follow the approach of Habib and Zurawicki (2002) 26 when testing institutional variables. Despite no pair of institutional variables move in systematic way (Table 3, Correlation matrix), we perform 5 separate models to test them since it is proposed that, in some cases, collinear relationships may involve more than two explanatory variables and it may not be directly detected examining pairwise correlations (Hill et al., 2001). Therefore, each of our 5 models is formed by all economic variables plus an institutional one27. As a consequence, the R&D variable is tested in all the regressions.

22

The original scale gives a value of 0 to highly corrupted countries, and 10 to corruption-free countries. 23

We rescaled the original index ranging from -2.5 to 2.5. 24

In particular Habib and Zurawicki (2002). 25

Terms i and j denote the investing and host country respectively, while the term t denotes the year. Variable Lab Cost Diffijt=Lab Costit-Lab Costjt, while R&D Diffijt=R&Djt-R&Dit.

26

See Appendix B. 27

(20)

(21)

6. Results

Table 4 below presents the results for the full sample of 13 host countries and 15 home countries during the period 1990-2007.

Table 4. Results for the full samplea (Dependent variable = Log FDI)

Variable Model 1 Model 2 Model 3 Model 4 Model 5

Log GDP/Capita 4.507*** 4.345*** 4.409*** 4.580*** 4.559*** (0.532) (0.529) (0.559) (0.527) (0.525) GDP 0.154 0.154 0.301 0.158 -0.021 (0.147) (0.147) (0.188) (0.147) (0.160) Tax -0.040*** -0.036*** -0.040*** -0.037*** -0.035*** (0.006) (0.006) (0.007) (0.006) (0.006) Openness 0.016*** 0.014*** 0.017*** 0.016*** 0.014*** (0.005) (0.005) (0.005) (0.005) (0.005) Trade 0.005 0.003 0.004 0.002 0.004 (0.010) (0.010) (0.010) (0.010) (0.010) Inflation 0.112*** 0.123*** 0.117*** 0.122*** 0.123*** (0.015) (0.015) (0.015) (0.015) (0.016)

Labour Costs Diff b 0.001*** 0.001*** 0.001*** 0.001*** 0.001***

(0.000) (0.000) (0.000) (0.000) (0.000) R&D Diff c 0.072*** 0.082*** 0.073*** 0.082*** 0.069*** (0.023) (0.023) (0.023) (0.023) (0.023) Corruption -0.066* (0.039) Hiring/Firing Regulations -0.201*** (0.059) Political Stability 0.154 (0.124) Bureaucracy 0.098*** (0.034) Start Business 0.086*** (0.023) Constant -41.240*** -38.384*** -41.063*** -42.291*** -41.759*** (5.146) (5.201) (5.194) (5.138) (5.123) Observations 2221 2221 2221 2221 2221 R-Squared 0.7410 0.7426 0.7411 0.7419 0.7426 Adjusted R-Squared 0.7167 0.7185 0.7168 0.7178 0.7185 F-Statistics 42.78 43.33 43.08 44.32 43.53 Prob F-Statstics 0.000 0.000 0.000 0.000 0.000

a) Using White heteroskedasticity-consistent standard errors b) Difference in labour costs between investing and host country c) Difference in number of researchers between host and home country d) Standard errors in parentheses:

(22)

Model 1 includes all the economic variables plus the Corruption variable. As expected, the difference in research expenditures between host and home country (R&D Diff) has a significant positive impact on FDI. This confirms Hypothesis 1. Concerning Hypothesis 2, we find weak significant evidence (p-value < 0.10) that more corrupted countries receive lower volume of investments flows. Furthermore, consistent with previous studies, market size, openness to trade, labor costs differential (difference between investing and host country) and bilateral trade positively influence direct investments, while the host country corporate income tax rate is negatively related to the dependent variable. Against our expectations, the coefficient of Inflation is significant and with the unexpected positive sign. Overall, the model is statistically significant (F = 42.78, p-value < 0.001) and able to explain more than the 71% of the variance (adjusted R-squared = 0.7167). With Model 2 the impact of strict regulations to hire and fire workers is assessed. The proposed variable is significant at 1% level and with the expected negative sign. The equation specified in the model is able to explain 71.85% (adjusted R-squared) of the variability across countries and over time.

Model 3 includes the political stability index. Despite having the expected sign, the Political Stability variable is not found to be significant at any conventional level. However, importantly, the effect of GDP per capita, Tax, Openness, Inflation, Labour Costs and R&D Diff remain significant at p < 0.01 level.

In Model 4, a second index drawn upon the Fraser Institute, Bureaucracy, is included and tested. Results show that countries with a more efficient bureaucratic apparatus receive higher volumes of foreign capitals. This confirms Hypothesis 5 at 1% level of significance.

Finally, the last model propose that the ease of entry in a foreign location does matter and it positively affects FDI28 (p < 0.001), as proposed in hypothesis 6. In Model 5, nearly 72% of the variance is explained.

As previously stated, the Hausman test for fixed vs. random effects suggests that using a methodology with fixed effects is appropriate. Therefore, some indications of home and host countries‟ characteristics can be drawn from the analysis of the constant terms29in the models we presented – the constant terms embody countries‟ peculiarities which are not controlled for by the variables included in the statistical model. Taking into account all the 182 bilateral FDI flows30, we note that coefficients related to pair of countries sharing a common language31 generally exhibit higher level. Thus, among the non-included explanatory variables, linguistic and cultural ties might explain those higher

28

A higher score means less time and money necessary to start up a new activity in the foreign country. 29

In the fixed effects model only the intercept parameter varies across entities. However, it remains constant over time.

30

We are analyzing bilateral FDI; hence the fixed effects concern pair of countries. 31

(23)

values. Then, to detect to what extent the general model can explain the relatively low attractiveness of SEA economies, we check for the existence of region-specific effects by introducing a dummy variable relative to SEA32.

Table 5. Results with SEA dummya (Dependent variable = Log FDI)

Log GDP/Capita 3.831*** 4.064*** 4.231*** 3.920*** 3.963*** (0.364) (0.381) (0.384) (0.381) (0.379) GDP 0.320*** 0.206** 0.311*** 0.281*** 0.246*** (0.082) (0.086) (0.086) (0.085) (0.084) Tax -0.032 -0.034*** -0.037*** -0.029*** -0.033*** (0.006) (0.006) (0.006) (0.007) (0.007) Openness 0.007** 0.001 0.002 0.005** 0.003 (0.002) (0.003) (0.002) (0.002) (0.002) Trade 0.086*** 0.085*** 0.085*** 0.085*** 0.085*** (0.005) (0.005) (0.005) (0.005) (0.005) Inflation 0.100*** 0.088*** 0.098*** 0.103*** 0.087*** (0.017) (0.017) (0.018) (0.018) (0.017) Labour Costs Diff b 0.001*** 0.001*** 0.001*** 0.001*** 0.001***

(0.000) (0.000) (0.000) (0.000) (0.000) R&D Diff c -0.020*** 0.003 -0.001 0.001 0.005 (0.013) (0.013) (0.013) (0.013) (0.013) Corruption -0.319*** (0.041) Hiring/Firing Regulations -0.059* (0.032) Political Stability 0.242*** (0.093) Bureaucracy 0.171*** (0.044) Start Business 0.018 (0.028) SEA -0.941*** -0.008 -0.328** -0.333*** -0.142 (0.145) (0.131) (0.130) (0.113) (0.109) Constant (non-SEA) -37.299*** -36.078*** -39.938*** -36.703*** -35.630*** (3.594) (3.754) (4.029) (3.749) (3.758) Observations 2221 2221 2221 2221 2221 R-Squared 0.4137 0.3997 0.4005 0.4030 0.3989 F-Statistics 120.07 115.16 114.36 112.27 114.66 Prob F-Statstics 0.000 0.000 0.000 0.000 0.000

* significant at 10%, ** significant at 5%, *** significant at 1%

32

(24)

As can be seen from Table 5, dummy coefficient for SEA is negative and statistically significant in almost all models. This result suggests that, along with proposed variables, there is also a still unexplained, region-specific negative performance effect with respect to FDI. This can be due to cultural factors or distance to main markets.

We further check for specific host country effects (Table 6). We introduce 12 dummy variables representing each host country in the dataset33, apart from the Germany which is therefore considered as the benchmark34. Each dummy variable scores 1 if the FDI is directed to the country, 0 otherwise (for example, the dummy variable Greece assumes value 1 when bilateral FDI flows go to Greece, 0 if not).

Table 6. Host country effects1

Coefficient Stand. Error t-statistic P> | t |

France -0.0189 0.1970 -0.10 0.923 Portugal -1.3368 0.4177 -3.20 0.001 Spain -0.2270 0.2936 -0.77 0.440 Greece -3.2459 0.4584 -7.08 0.000 Italy -1.2113 0.2726 -4.44 0.000 Constant (Germany) -51.8765 6.3523 -8.17 0.000

1) Germany is the benchmark. Results obtained from Model 1.

At least in the case of Italy, Greece and Portugal, dummy coefficients are negative and statistically significant at 1% compared to the coefficient of Germany. These results confirm that those countries possess characteristics, other than those used in the model, which negatively influence FDI. Examples of these characteristics can be the lack of infrastructure, high crime rate, low labour productivity, inefficient legal system or, for Portugal and Greece, being far away and isolated from the core of Europe.

To additionally test the robustness of results presented in Table 4 and 5, other regressions are run. In Table 7, 8 and 9 three more models are presented. The first one removes US and Japan from the sample of investing countries, therefore considering only FDI flows between European economies. The second and the third one assess the consistency of our results across the 18-years period. In the specific, they check whether the effects of economic and institutional variables before (1990-1998) and after (1999-2007) the Euro introduction are the same.

Looking at the coefficients of the second model, there is indication that, despite Corruption loses its statistical significance, the proposed institutional variables have a

33

We carried out this procedure for each of the five models. However, results appear to be similar. Hence we report only those ones obtained from Model 1.

34

(25)

larger impact when only the European subsample is considered. At the same time, interesting differences emerge when comparing results obtained in Table 8 and 9. For the period prior the single European currency debut, the majority of the indexes have in general a larger effect. On the other hand, Political Stability becomes significant at 5% level only in the second lapse of time, whereas Bureaucracy and Start Business exhibit lower coefficients but higher statistical validity35.

35

(26)

Table 7. Results for EU subsamplea (Dependent variable = Log FDI)

Log GDP/Capita 4.542*** 4.360*** 4.501*** 4.614*** 4.594*** (0.582) (0.578) (0.611) (0.578) (0.574) GDP 0.145 0.145 0.258 0.157 -0.046 (0.167) (0.168) (0.210) (0.164) (0.179) Tax -0.047*** -0.042*** -0.046*** -0.044*** -0.041*** (0.007) (0.007) (0.007) (0.007) (0.007) Openness 0.017*** 0.016*** 0.018*** 0.018*** 0.016*** (0.005) (0.005) (0.005) (0.005) (0.005) Trade 0.005 0.003 0.005 0.002 0.003 (0.011) (0.011) (0.011) (0.011) (0.011) Inflation 0.120*** 0.131*** 0.124*** 0.129*** 0.132*** (0.016) (0.016) (0.016) (0.016) (0.016)

Labour Costs Diffb 0.001*** 0.001*** 0.001*** 0.001*** 0.001***

(0.000) (0.000) (0.000) (0.000) (0.000) R&D Diffc 0.084*** 0.095*** 0.084*** 0.095*** 0.082*** (0.024) (0.024) (0.024) (0.024) (0.024) Corruption -0.069 (0.045) Hiring/Firing Regulations -0.225*** (0.064) Political Stability 0.111 (0.139) Bureaucracy 0.098*** (0.037) Start Business 0.099*** (0.026) Constant -41.601*** -38.445*** -41.627*** -42.6365*** -42.174*** (5.616) (5.639) (5.657) (5.594) (5.587) Observations 1868 1868 1868 1868 1868 R-Squared 0.7283 0.7302 0.7282 0.7291 0.7304 Adjusted R-Squared 0.7022 0.7042 0.7020 0.7030 0.7044 F-Statistics 41.06 40.97 42.07 44.18 40.87 Prob F-Statstics 0.000 0.000 0.000 0.000 0.000 a) Using White heteroskedasticity-consistent standard errors

b) Difference in labour costs between investing and host country c) Difference in number of researchers between host and home country d) Standard errors in parentheses:

(27)

Table 8. Results for the full sample (1990-1998)a (Dependent variable = Log FDI)

Log GDP/Capita 3.964*** 2.639** 4.227*** 3.404*** 2.793** (1.246) (1.228) (1.278) (1.259) (1.280) GDP 1.006** 0.677 1.061** 1.128*** 1.322*** (0.460) (0.454) (0.445) (0.430) (0.427) Tax -0.019 -0.018 -0.024 -0.017 -0.007 (0.014) (0.013) (0.015) (0.014) (0.014) Openness 0.017 0.010 0.017 0.012 0.014 (0.010) (0.010) (0.010) (0.010) (0.010) Trade 0.019 0.010 0.013 0.017 0.013 (0.025) (0.025) (0.025) (0.025) (0.025) Inflation 0.070*** 0.069*** 0.071*** 0.074*** 0.075*** (0.022) (0.022) (0.022) (0.022) (0.023)

(0.000) (0.000) (0.000) (0.000) (0.000) R&D Diff c 0.111** 0.116** 0.113** 0.139** 0.137** (0.053) (0.051) (0.053) (0.056) (0.053) Corruption -0.087* (0.047) Hiring/Firing Regulations -0.557*** (0.135) Political Stability -1.095 (0.689) Bureaucracy 0.259** (0.123) Start Business 0.172*** (0.064) Constant -37.365*** -20.403* -31.340*** -32.726*** -26.329** (11.972) (11.929) (11.815) (11.954) (12.166) Observations 1078 1078 1078 1078 1078 R-Squared 0.7780 0.7821 0.7781 0.7784 0.7795 Adjusted R-Squared 0.7341 0.7389 0.7342 0.7345 0.7359 F-Statistics 12.53 13.61 12.34 11.91 12.35 Prob F-Statstics 0.000 0.000 0.000 0.000 0.000

(28)

Table 9. Results for the full sample (1999-2007)a (Dependent variable = Log FDI)

Log GDP/Capita 4.441*** 4.494*** 4.865*** 4.368*** 3.591*** (1.339) (1.349) (1.370) (1.336) (1.341) GDP 0.092 0.252 0.490** 0.243 -0.041 (0.210) (0.207) (0.232) (0.206) (0.218) Tax -0.035* -0.025 -0.032 -0.031 -0.016 (0.020) (0.020) (0.021) (0.020) (0.020) Openness -0.012 -0.010 -0.008 -0.014 -0.012 (0.010) (0.011) (0.011) (0.010) (0.010) Trade 0.005 0.005 0.004 0.002 0.005 (0.009) (0.009) (0.009) (0.009) (0.009) Inflation 0.023 0.022 0.005 0.074 0.119* (0.064) (0.068) (0.067) (0.064) (0.068)

(0.000) (0.000) (0.000) (0.000) (0.000) R&D Diff c 0.087 0.084 0.075 0.094 0.076 (0.069) (0.069) (0.069) (0.069) (0.069) Corruption -0.433*** (0.154) Hiring/Firing Regulations -0.047 (0.106) Political Stability 0.356** (0.172) Bureaucracy 0.117*** (0.038) Start Business 0.119*** (0.028) Constant -41.286*** -38.975*** -45.481*** -38.251*** -30.801** (13.484) (13.766) (14.279) (13.607) (13.655) Observations 1078 1078 1078 1078 1078 R-Squared 0.7780 0.7821 0.7781 0.7784 0.7795 Adjusted R-Squared 0.7341 0.7389 0.7342 0.7345 0.7359 F-Statistics 12.53 13.61 12.34 11.91 12.35 Prob F-Statstics 0.000 0.000 0.000 0.000 0.000

(29)

7. Implication for SEA

Statistical results presented in the previous section confirm that variables introduced are able to explain a relevant share of the variability across countries and over time. The majority of theoretical arguments taken into consideration in section 3 are therefore confirmed. However, an unexpected result occurred and it needs to be explained.

In the literature review, we have highlighted that macroeconomic stability (captured by the inflation rate) should be negatively related to the level of FDI inflows. But contrary to usual findings, our regressions displayed the variable Inflation with a significant positive sign, suggesting that the instability of a system is an element fostering FDI inflows. Despite being against conventional results, previous studies have already debated on this point. As pointed out by Ang J.B. (2008), this result emerges because “foreign investors perceive a higher level of uncertainty as a greater potential return”. Similarly, Zheng (2009) argues that high inflation encourages inward FDI since it appreciates the host currency and the same amount invested has a higher value in terms of the home currency. This logic finds support in our statistical results when the time frame is split in two parts. The variable Inflation is in fact significant and with a higher coefficient only in the first lapse of time, that is, prior the lunch of the Euro currency when inflation rate differentials across EU were wider.

Besides the unexpected result produced by variable Inflation, with Model 1 through 5 six hypotheses are tested, and five out of six have the significant expected effect on bilateral FDI. Hence, variables proposed may contribute explaining much of the FDI gap of SEA‟s. Variable R&D Diff is found to be relevant in all our statistical regressions with the only exception in the period 1999-2007. Taken together, these results highlight the importance of R&D as a vehicle of attracting MNEs‟ activities. From Table 1 in Section 3, we recognized the poor R&D performance exhibited by the SEA countries. According to statistical results, this variable partly describes why SEA countries receive relatively low level of FDI. Compared to the average of 8 non-SEA economies, the level of R&D in France explains the 6.5 percent of its annual FDI lag during the period 1990-200736. For Portugal, Spain, Italy and Greece, the relative percentages are much higher and approximately equal to 30.4%, 24.7%, 29.7% and 31.6% (Table 10). In this context, different actions could be advanced by SEA nations in order to promote scientific activities and technical innovations. Two feasible solutions would be a more efficient use of public and private funding, and the establishment of tighter relationships between universities, research institutes and enterprises. For instance, if SEA countries move to German level of R&D, they would increase their annual foreign investments flows by approximately 1% for France, 24% Portugal, 18.3% Spain, 25.2% Greece and 23.3% Italy.

36

(30)

Table 10. FDI lag explained by variablesa (1990-2007 averages)

Model 2 Model 4 Model 5

R&D Corruption Labour Reg. Bureaucracy Start Business

France 6.5% 8.7% 34.4% 5.8% 16.0%

Portugal 30.4% 13.4% 33.1% 17.5% 14.5%

Spain 24.7% 15.8% 36.7% 6.9% 12.2%

Greece 31.6% 24.0% 19.9% 20.9% 18.0%

Italy 29.7% 25.6% 45.0% 17.6% 14.5%

a) Relatively to the average FDI inflows received by 8 non-SEA countries

Model 1

Concerning the institutional variables, great consensus emerges when comparing the level of corruption among countries (Table 10). France, Spain, Portugal, Italy and Greece display the highest CPI37 during the whole period. As suggested by estimations of Model 1 (Table 4), more corrupted regions enjoy lower volumes of FDI. Precisely, a one-point increase in the Transparency International index reduces a country‟s FDI inflows by approximately 6.6%. This finding partly reveals to which extent SEA economies are relatively less attractive locations for foreign investments (Table 10). Furthermore, despite the effect of the variable Corruption is significant only at 10% level, we do believe that a complex phenomenon such as that of corruption may invade different economic, social and political spheres, thus exercising an indirect negative influence on bilateral FDI. As an example, it is advanced that corruption diverts public expenditures allocated for infrastructure projects and consequently reduces the attractiveness of a country in terms of its physical assets. Following the approach developed by the World Bank, the set-up of anti-corruption task forces would probably restrain corrupted behaviors, thus improving the location appeal of SEA countries. Drawing upon results in Table 4, if Italy, Greece, Spain and Portugal succeed in reducing their level of corruption to that one of France38, the least corrupted country in the SEA, they would increase their annual FDI inflows by approximately 17%, 15%, 7% and 5%, respectively.

When analyzing the impact of strict labour regulations, a clear picture emerges. Conform the literature, strict labour market arrangements deter inward FDI. According to Model 2 (Table 4), the variable Hiring/Firing Regulations is the institutional variable explaining the largest part of the FDI gap of SEA‟s. In particular, because of its strict regulations, it is Italy that discloses the highest share of FDI lag39, 45%, followed by Spain (36.7%), France (34.4%), Portugal (33.1%) and Greece (19.9%). Consequently, despite successful reforms have been recently introduced into SEA countries to encourage a more flexible

37

Corruption perception index. 38

2007 value. 39

(31)

labour market, institutional reforms need to move forward and start implementing a sustainable reduction of non-wages costs40.

Two indexes that have always provided expected significant results in all estimations are those ones representing the ease of entry in a foreign country and the quality of national bureaucracy. Also in this case, the interpretation is straightforward. All five countries have relatively less efficient public administrations which, as debated in the literature section, negatively affect multinational firms‟ location choices. In the 1990-2007 period, compared to the mean of the other 8 economies, the average annual FDI lag explained by the variable Bureaucracy (Model 4) is 20.9% for Greece, 17.6% for Italy, 17.5% for Portugal, 6.9% for Spain and 5.8% for France. For that reason, SEA countries should pay greater attention to regulatory reforms aimed to modernize and make more efficient their bureaucratic apparati. According to estimation results, improving national bureaucracy to Germany level would increase annual FDI by approximately 13% for France, 12.9% for Portugal, 2.4% for Spain, 16.4 for Greece and 13% for Italy.

Manoeuvres directed to speed up and upgrade the set of mechanisms necessary to start-up a business would also raise national competitiveness. Using estimation output obtained in Model 5, the variable Start Business explains an important part of the low performance of Mediterranean area. With the only exception of Austria (0.48), coefficients for Spain (0.50), Italy (0.48), Portugal (0.48), France (0.46) and Greece (0.44) are found to be the lowest. Setting up faster mechanisms such as those ones observed in other EU countries41 would increase the annual flows of FDI by approximately 18% for Greece, 16% France, 15% Portugal, 14% Italy, and 12% Spain.

We conclude this section drawing some inferences on the last variable, Political Stability, which is not found significant in any of our models. Two points can be debated. First, like corruption, also in this case we do believe in the possibility of indirect effects on FDI flows, for example when the continuous change in political powers becomes an impediment to accomplish structural reforms, such as liberalizations or privatizations, which make a country more attractive to foreign capital. Second, the effect of political stability might be better captured when using a more specific index than the one we relied on, e.g. the International Country Risk Group Index.

8. Conclusions and Limitations

Although the European Union represents the largest inward host area of FDI, the SEA countries receive limited volumes of foreign investments. The purpose of this study is to

40

That is, hiring costs (such as payroll taxes, retirement and contribution funds, health insurance and other obligation) as well as firing costs (such as severance payments and penalties).

41