THE GRAVITY OF CORRUPTION ON FOREIGN DIRECT INVESTMENT. Xingwang Qian a SUNY Buffalo State Economics and Finance Department

Transcription

1 VERY PRELIMINARY DRAFT, PLEASE DO NOT CITE THE GRAVITY OF CORRUPTION ON FOREIGN DIRECT INVESTMENT Xingwang Qian a SUNY Buffalo State Economics and Finance Department Jesus Sandoval-Hernandez b College of Charleston School of Business & School of Languages and World Affairs Jinzhuo Zhao c Hampden-Sydney College Department of Economics This version: June 2011 a SUNY Buffalo State College, qianx@buffalostate.edu. b College of Charleston, Charleston South Carolina. sandovalhernandezh@cofc.edu. c Hampden-Sydney College, Hampden Sydney, Virginia. jzhao@hsc.edu. 1

2 THE GRAVITY OF CORRUPTION ON FOREIGN DIRECT INVESTMENT Abstract The empirical literature on the relationship between corruption and FDI has not yet reached a consensus. We study the effect of corruption on bilateral FDI flows from a special perspective. We examine the effect of corruption distance, defined as the difference in corruption level within a country-pair, and its asymmetrical effects on FDI flows. Applying a modified gravity model for bilateral FDI flows, we find that the absolute corruption distance is negatively associated with bilateral FDI flows between countries. For a given FDI home-host-country pair, the smaller absolute corruption distance, the higher bilateral FDI flows. A careful examination upon the asymmetric features of positive and negative corruption distance suggests that the positive corruption distance promotes FDI, while the negative corruption distance hinders FDI. We document that not all investors care the same about corruption. FDI from developing-low corruption countries is not affected by higher corruption in developed countries, but it is deterred by higher corruption in developing countries. FDI outflows from developed-high corruption countries is attracted to developed-low corruption countries but have not effect when host countries are developing-low corruption. These findings are especially relevant, for they shed new light on explanations to increasing FDI activities undertaken by developing countries in recent years. Keywords: Foreign direct investment, corruption, developing countries. JEL Classifications: F21, F23, F59, O5 2

3 1. Introduction Since the mid-1990s and until the irruption of the global financial crisis of 2008, the flows of Foreign Direct Investment (FDI) increased dramatically, reaching in 2007 a record high of $ 2.1 trillion. This figure represents a stunning 332 percent increase compared with the $486 billion recorded in 1997 (UNCTAD STAT). The quest for understanding the factors driving FDI has prompted a myriad of theoretical and empirical literature. Recent research has emphasized the role of corruption in bilateral FDI s flows. Corruption is seen to render a host country unattractive to foreign investors via high cost of entry and uncertainty, distorting incentives to invest (Shleifer and Vishny, 1993; Mauro, 1995, Wei, 2000a; Wei, 2000b). A strand of the empirical literature seems to support the negative effects of corruption on FDI. Bribes pay by firms act as taxes, the rent seeking activities facilitated by corruption result in waste of resources, and there are additional costs due to the inability to enforce contracts that result for the corruption s practices (see for example Wei, 2000a; Habib and Zurawicki, 2002; Lambsdorff, 2003). However, research on the negative effects of corruption on FDI is far from conclusive. Some authors did not find a significant correlation between corruption and FDI (Wheeler and Mody, 1992; Alesina and Weder, 2002, Glass and Wu, 2002). And some other studies find that, under specific circumstances, corruption may even enhance efficiency and stimulate FDI. When companies are willing to pay bribes, corruption acts as a helping hand increasing their revenues (Olson, 1993; Egger and Winner, 2005). Corruption speed up the bureaucratic process to obtain the legal permissions for setting up a foreign plant, the speed money argument (Lui, 1985), and helps to gain access to public funded projects (Tanzi and Davoodi, 2000). 3

4 In spite of perceived lack of improvement, or even an increase, in corruption levels, bilateral FDI activities have been increasing since the late 1990s. Figure 1 plots FDI s inflows and corruption levels. From 1997 to 2000 FDI increased, but corruption worsened (lower corruption index indicates more corruption). For the period, corruption increased even more, reaching its highest level in 2003, and FDI flows dropped. Thereafter, FDI increased sharply even though corruption stayed relatively high. The conflicting results reported in the extant literature strengthen the idea that the relationship between corruption and FDI lacks a consistent pattern. A quick glance at the data indicates that countries with similar size in their economies and similar levels of corruption, such as Malaysia and Colombia (relatively high corruption), or Chile and Portugal (relatively low corruption) receive similar FDI inflows. While Italy is perceived relatively corrupt and receives modest inflows of FDI, France, which is perceived significantly less corrupt, attracts substantial FDI. Brazil, China, Russia, and Mexico attract large flows of FDI despite their high perceived corruption. To address such paradoxes, we study the effects of corruption on FDI flows at the bilateral level. Specifically, we use an augmented gravity model for bilateral FDI flows, using a data sample including 47 developed and developing countries with varying degrees of corruption spanning over a decade ( ). The data set comprises comprehensive country-specific characteristics. We explicitly incorporate into our gravity model the idea of corruption distance (CD henceforth) defined by the difference in corruption level within a country-pair. To the best of our knowledge, only Habib and Zurawicki (2002) have to date studied the impact of CD on 4

5 bilateral FDI. They find that the absolute value of CD has a negative effect on bilateral FDI 1. Our analysis extends in various dimensions. First, unlike Habib and Zurawicki (2002), who include only seven developed countries as the source countries and a time span of 3 years, we use a broader panel with both developed and developing countries, including transition economies in the latter group, as source countries. Our sample spans eleven years and includes a more comprehensive list of control variables 2. The inclusion of developing countries as source country allows us to analyze the role of corruption in the recent observed phenomena of an ever increasing trend of FDI s outflows from these countries (UNCTAD STAT, 2008). Second, in addition of testing the hypothesis that CD between country pairs is negatively associated with bilateral FDI flows, we scrutinize the asymmetrical impact of corruption on FDI. It is conceivable that the absolute CD values could be masking the possible different effects of positive CD compared to the negative CD effects. To identify the effects of these two types of CD, we extend our analysis by testing the hypothesis that, from the point of view of the source country, the unilateral FDI from low corruption countries to high corruption countries is negatively associated with their CD. Likewise, the unilateral FDI from high corruption countries to low corruption countries is positively associated to their CD. Third, we explore the idea that investors from high/low corruption countries would react differently when they face the choice of allocating investment between developed/developing countries with high or low levels of corruption. 1 In a more recent paper, Bénassy-Quéré et al (2007) generalized Habib and Zurawicki analysis by using not only corruption but a wide variety of institutional characteristics to construct institutional distance. They conclude that institutional distance tends to reduce bilateral FDI. 2 Habib and Zurawicki (2002) data sample yield an expected maximum of 1869 observations. However, their sample size was greatly reduced due to missing values related to a country or a year. 5

6 A typical characteristic of bilateral FDI flow data is the presence of excessive zeros, we deal with this issue in our data sample by implementing a two-stage estimation procedure proposed by Heckman (1979). The two-stage estimation procedure allow us to sequentially analyze the decision making process. The first decision is to invest or not. If the first decision is positive, then the amount of investment has to be determined. To anticipate the results, we find that a country tends to invest more FDI in a host county with smaller corruption distance from its own, and reduces FDI flows to countries with larger corruption distance. A careful examination upon the asymmetric features of positive and negative corruption distance suggests that positive corruption distance promotes FDI, while negative corruption distance hinders FDI. We document that not all investors care the same about corruption. FDI from developing-low corruption countries is not affected by higher corruption in developed countries, but it is deterred by higher corruption in developing countries. FDI outflows from developed-high corruption countries is attracted to developed-low corruption countries but have not effect when host countries are developing-low corruption. These findings are especially relevant, for they shed new light on explanations to increasing FDI activities undertaken by developing countries in recent years. In section 2 we offer an overview of the relevant literature on corruption and FDI. In section 3 we describe the data and methodology. Section 4 presents the results and discussion, and section 5 concludes. 2. Corruption and FDI Corruption, defined by Transparency International (TI), as the abuse of entrusted power for private gain has pervasive and mostly negative effects on the business environment of a country. Corruption represents the need to make additional payments to get things done. It 6

7 indicates a lack of respect for the rules and regulations governing economic interactions in a society (Kaufmann et al., 2003). Theoretically, the impact of corruption on FDI could be ambiguous. To be sure, one has to distinguish between the two opposite effects of corruption, i.e. what Egger and Winner (2005) termed the grabbing hand and the helping hand (see Olson, 1993 and Egger and Winner, 2005). On the negative stance of corruption, many scholars argue that its grabbing hand acts as sand in the wheels of commerce decreasing welfare. Corruption results in the wasteful use of resources allocated to pay bribes or to fight it. In the absence of corruption, these resources could be invested in a more efficient way. Bribes pay by firms act as taxes, the rent seeking activities facilitated by corruption result in waste of resources, and there are additional costs due to the inability to enforce contracts that result for the corruption s practices. From a positive view, the helping hand effect of corruption prompts a process that seems to facilitate transactions and help to speed up procedures that would otherwise be more difficult to attain (Left, 1989). In countries with cumbersome government regulations and effective red tape, corruption speed up the bureaucratic process to obtain the legal permissions for setting up a foreign plant. In this case, investors who value time or access to an input more than others will pay more for it. This is the speed money argument advanced by Lui (1985). Corruption helps to gain access to public funded projects (Tanzi and Davoodi, 2000). However, in this case as Tanzi (1998) suggested, those paying the highest bribes are not necessary the most efficient firms but successful rent-seekers. For the last decade, a plethora of empirical works analyzing the effects of corruption on FDI has flourished. A strand of literature seems to support the negative effects of corruption on 7

8 FDI. One of the first empirical investigations on corruption and FDI was undertaken by Mauro (1995), who use a sample of 67 countries and a corruption index provided by Business International (BI). He finds that corruption impacts negatively on the ratio of investment to GDP. The author claims that if Bangladesh were to improve its level of integrity to that of Uruguay, its investment rate would increase by almost five percent of GDP. In a latter paper, Mauro (1997) provides further evidence of his earlier findings by using a larger data sample of 94 countries and the corruption index from the Political Risk Services group (PRS). Focusing on bilateral flows between 12 source and 45 host countries in 1990 and 1991, Wei (2000a) detects a significant negative impact of corruption on FDI. He finds that an increase in the corruption level that of Singapore to that of Mexico is equivalent to raising the tax rate by over twenty percentage points. Wei (2000b) confirmed the negative relationship between corruption in the host country and FDI after considering government policies towards FDI. This result was challenged by Stein and Daude (2001) who argue that high collinearity between corruption and GDP per capita could lead to spurious results when GDP per capita was not included in the equation. Hines (1995) examines the effect of the U.S. anti-bribery legislation, the Foreign Corrupt Practices Act of 1977, on the operation of U.S. firms in countries where corruption is high. He uses the growth rate of U.S. FDI flows into 35 host countries over the period 1977 to 1982 as the dependent variable and the Business International Index as a measure of corruption. His finding suggests that the Corrupt Practices Act significantly reduced U.S. FDI flows into more corrupt host countries after

9 Some other authors have found the same negative relation between FDI and corruption for more specific group of countries. For example, Lambsdorff and Cornelius (2000) show an adverse impact of corruption on FDI for African countries. Smarzynska and Wei (2000) provide evidence for corruption to reduce firm-level assessments of FDI in Eastern Europe and the former Soviet Union. Using a single source country, Voyer and Beamish (2004) use crosssectional regressions to investigate the effects of the level of corruption on Japanese FDI in 59 (developed and emerging) host countries. They find that Japanese FDI is negatively related to the level of corruption especially in emerging countries. Further, their results show that in emerging countries where a comprehensive legal system is underdeveloped or does not exist to effectively reduce illegal activities, corruption serves to reduce Japanese FDI inflows. However, research on the negative effects of corruption on FDI is far from conclusive. Some researchers did not find a significant correlation between corruption and FDI. In a study of foreign investment by U.S. firms, Wheeler and Mody (1992) use the first principal component of 13 risk factors, including bureaucratic red tape, political instability, corruption, and the quality of the legal system; they did not find a significant effect of institutions of US foreign affiliates. Wei (2000a), however, argues that the reason why Wheeler and Mody (1992) failed to find a significant relationship between corruption and FDI is that corruption is not explicitly incorporated into their model. Wheeler and Mody (1992) combined corruption with 12 other indicators to form one variable, but some of these indicators may be marginally important for FDI. Abed and Davoodi (2000) use a cross-sectional as well as a panel data analysis to examine the effects of levels of corruption on per capita FDI inflows to transition economies. They find that countries with a low level of corruption attract more per capita FDI. However, 9

10 once they control for the structural reform factor, corruption becomes insignificant. They conclude that structural reform is more important than reducing the level of corruption in attracting FDI. Alesina and Weder (2002) also reported insignificant results. But they make use of a variable that determines, not levels of corruption but the political instability due to corruption. Glass and Wu (2002), using a general equilibrium model to study corruption and FDI reached similar conclusion, the effects of corruption on FDI were ambiguous. Focusing on only developing countries, Akçay (2001) uses cross-sectional data from 52 developing countries with two different indices of corruption to estimate the effects of the level of corruption on FDI inflows. He fails to find evidence of a negative relationship between FDI and corruption. He concludes that the most significant determinants of FDI are market size, corporate tax rates, labor costs, and openness. In contrast, some other studies find that, under specific circumstances, corruption may even enhance efficiency and stimulate FDI. In the presence of a rigid regulation and an inefficient bureaucracy, corruption may increase bureaucratic efficiency by speeding up the process of decision making (Bardhan, 1997). Egger and Winner (2005) corroborated that corruption can be beneficial in circumventing regulatory and administrative restrictions concluding that, indeed, corruption encourages FDI. The authors use a data set of 73 developed and developing countries. The existence of the helping type of corruption, they argued, is supported by the findings of a positive short run and long run effects of corruption. They concluded that since the short run impact is considerably smaller than the long run counterpart, this constitutes an indication of a short run grabbing hand effect. Cuervo-Cazurra (2006) examined the relationship between corruption and FDI 10

11 using data on bilateral FDI inflows from 183 home to 106 host countries. They find that corruption results in relatively higher FDI from countries with high levels of corruption. In the literature on management, the Uppsala model has described the internationalization of a firm as a process of experiential learning and incremental commitments which leads to an evolutionary development in a foreign market (Johanson and Wiedersheim-Paul, 1975; Johanson and Vahlne, 1977, 1990). Within the Uppsala model, Johanson and Vahlne (1977) categorized the Psychic distance as a major impediment to the decision of companies to enter foreign markets. Companies enter markets perceived to be psychologically closer before considering the remote ones. Psychic distance is the difference between countries in terms of language, culture, education, business practices, industrial development, and regulation (Johanson and Vahlne, 1977). Using this similarity argument, we argue that FDI from countries with high corruption may not only be undeterred by host country corruption, but it may even attract by it. Firms from high-corruption countries face lower costs of doing business abroad when they enter other countries with similar high levels of corruption. Recently, Ghemawat (2001) suggested that four dimensions of distance namely cultural, administrative, geographic and economic, influence companies considering global expansion. In this context, difference in corruption levels between the host and home countries is part of the administrative distance that creates significant barriers for foreign investors. In this paper we apply the similarity approach to corruption, which closely resembles the psychic distance idea. Our approach implies that the difference in corruption between home and host countries is part of the administrative distance, which can be conceptualized as a form of Psychic distance. Investors from less corrupt countries are less suited to handle corruption; this 11

12 disadvantage would deter companies from these countries to invest in countries with higher levels of corruption. A closely related paper to our exercise is Habib and Zurawicki (2002), who analyze the effects of corruption on bilateral FDI flows using a sample of seven source countries and 89 host countries. Habib and Zurawicki (2002) regressed bilateral FDI on a set of control variables including the absolute difference between the corruption levels in the source and the host countries. They find that foreign firms tend to avoid situations where corruption is visibly present because corruption is considered immoral and might be an important cause of inefficiency. 3. Data and Methodology 3.1 Data We use data on bilateral flows of FDI gathered by the Economist Intelligence Unit (EIU) for 47 developed and developing countries with varying degrees of corruption and spanning over a decade ( ). Appendix A1 presents the complete list of countries in our sample. The size of our panel is determined by data s availability. The explanatory variable of interest is CD, which is constructed using the corruption perception index compiled by the International Country Risk Group (ICRG). In the ICRG s index lower scores indicate high levels of corruption. The minimum and maximum rating any country receives is 1 and 6, respectively (for a detailed description see Knack and Keefer, 1995). The dataset comprises comprehensive country-specific characteristics, compiled from different sources. Tables A.2 and A.3 in the appendix provides the definitions and sources of these and other variables used in the study and their correlation coefficients. 12

13 11 t 1 With 47 pair countries and 11 years, the data set is an unbalanced panel with n ( 1) observations percent are positive FDI observations, 7.7 percent are t n t negative values and 57.7 percent are not registered or zero values 3. Table 1 presents summary statistics of the variables used in the study. 3.2 Methodology A gravity model and the Heckman two stages procedure The gravity model has been widely used in the literature for explaining FDI (Eaton and Tamura, 1995; Wei, 2000; Wei and Wu, 2001). Gravity models used in international economics theoretically rely on the proximity-concentration hypothesis (Horstmann and Markusen, 1992; Brainard, 1993; Markusen and Venables, 2000). These models postulate that bilateral international flows (goods, FDI, etc.) between any two economies are positively related to the size of the two economies (e.g., population, GDP), and negatively to the distance and a set of variables accounting for relative costs (tariffs barriers, information asymmetries, etc.) A standard practice in the literature is to log-linearize the FDI gravity models and estimate the parameters of interest by OLS. There are several reasons for doing this. First, the log specification provides a useful normalization that reduces the weight of pairs with very large FDI flows. This helps to ensure the homoscedasticity of the error term (Wei, 2000). Second, it allows interpreting the coefficients of the continuous variables as elasticities. Third, it has typically provided the best fit in gravity equations. There is a problem with this procedure, however, since not all countries receive direct investment from all source countries at all time periods, we have 3 Unlike on what it happens with trade flows, negative FDI inflows may exist. This situation takes place when foreign firms disinvest in host countries, for example when a foreign subsidiary is liquidated or when financial crises erupt. We treat both not registered and negative values as zeros. 13

14 the presence of excessive zeros. Working on the logarithm of FDI requires dropping these zero observations. By disregarding pair-countries with zero or negative FDI inflows, the researcher gives up important information contained in the data, and produce biased estimates as a result. Several approaches to circumvent this problem have been adopted. A popular one is to work with ln(a+fdi) instead of ln(fdi), with a relative small constant a (Stein and Daude, 2001; Bénassy-Quéré and Mayer, 2007). With a=1, the dependent variable is equal to zero when FDI is zero 4. Another approach is to use Tobit instead of OLS, which can be justified either by assuming that zero values are due to the presence of fixed costs of investing abroad, or by assuming that flows below certain threshold value are incorrectly recorded as zeros (Eaton and Tamura, 1995; Wei, 2000; Cuervo-Cazurra, 2006). Santos Silva and Tenreyro (2006) proposed the Poisson Quasi-Maximum Likehood (QMLE) as a method to deal with zeros in trade gravity equations. The QMLE was used in a bilateral FDI equation by Head and Reis (2006). To overcome the selection bias problem, in this paper we implement the two-stage estimation procedure proposed by Heckman (1979). This method accommodates measurement errors and it has been employed to analyze the impact of set up costs in FDI decisions (Razin et al, 2004; Davis and Kristjánsdόttir, 2010) and to analyze the investment decision process (e.g. Cheung et al, 2011). The estimation procedure allow us to sequentially analyze the decision making process. In the first stage, we estimate the influence of locational variables on the probability of FDI. We use Probit to regress a dummy variable equal to one if there is positive FDI from country i to country j in year t on a set of controls. The probit estimation includes random effects with year dummies. In the second stage, conditional on being selected, the 4 This specification has been used extensively in gravity models of trade (Eichengreen and Irwin, 1995; Redding and Venables, 2000). 14

15 amount of investment has to be determined. We regress the log of positive values of FDI on a set of control variables. At this stage, we use pooled ordinary least squares regression method with year, source and host country dummies with correction for selectivity bias. The information used to calculate the correction term, the Mills ratio, is obtained from the probit estimate in stage one. 3.2 The effects of CD on FDI flows In this section we perform regressions to test the claim that in pair countries with a similar high/low level of corruption, or equivalently low absolute CD, FDI increases. The inclusion of the controls is guided by economic theory and previous findings in the literature on corruption and FDI. The first stage Heckman procedure The decision to invest or not is estimated using the following regression specification. D ij, t dcor ij, t X ij, t 1 ij, t (1) where D ij, t is an indicator taking the value D ij, t 1 if FDI ij, t 0 and 0 otherwise. FDI ij, t are FDI flows from the source country, i to the host country, j; dcor ij, t is the similarity of corruption level, or corruption distance, measured as the absolute value of the difference between home country and host country s corruption index. X ij, t 1 is a vector containing all other standard control variables that affect FDI flows, β and γ are parameters and ij, t is the error term. With the exception of time invariant variables, but including the geographical distance, the control variables are lagged one period to address endogeneity issues (Razin et al, 2004). 15

16 A typical motivation for FDI is the search for new markets. We incorporate a group of gravity variables to capture the market-seeking motive, LGDP ijt 1, and LGGDP ijt 1. Larger countries are more likely to attract FDI, because firms can achieve economies of scale in the country. LGDP ijt 1 represents the market size and is the log of GDP from source country i to host country j, measured in current US dollars (Kravis and Lipsey, 1982; Wheeler and Mody, 1992). To capture market growth potential, we use the pair-country real growth rate, LGGDP ijt 1 (Lipsey, 1999; Lee, 2000). The market seeking motive implies that these variables have a positive coefficient. A second set of gravity variables are incorporated into our analysis to capture the effects of geographic and cultural distance. Geographical distance indicates the existence of transportation costs, and higher administrative and operational costs between parent companies and their subsidiaries. The natural log of distance between the economic centers in the countrypairs is used to measure the impact of geographical distance. We expect this coefficient to be negative. Geographical distance measure is traditionally complemented by indicators for contiguity and for whether the country is landlocked or an island (Feenstra, et al, 2001; Frankel and Rose, 2002; Wei, 2000). We use a dummy equal to 1 if source and host countries share a common border. In early regressions, we included dummies for island and landlocked. We decide to drop these variables due to their lack of significance and poor contribution to the explanatory power of the models. The effects of cultural distance are assessed using dummies for common language, colonial ties, and common legal system. We expect positive coefficients here. Similarities in 16

17 cultural backgrounds facilitate FDI, because investors benefit from a reduced psychic distance between home and host countries (Johanson and Vahlne, 1977; Ghemawat, 2001). To explain the effects of countries international orientation and economic ties on FDI, we include a dummy for World Trade Organization (WTO) membership and the ratio of total trade over country s GDP as a measure of openness. Rose (2000) finds that membership to WTO has not significant effect on trade. However, Subranian and Wei (2007) demonstrated that WTO membership has a positive and significant effect on trade between developed countries but negative and significant effect between developing countries. They attribute the latter result to problems of sampling and measurement error for the developing countries data. If WTO membership increases trade s volumes and trade acts as a complement of FDI, we expect the sign of WTO to be positive. A country s international orientation, measured by openness, reflects its competitiveness. Therefore, we expect a positive coefficient for openness. Foreign investors care about the quantity and quality of labor in host countries. A location becomes more attractive for FDI when the labor force is more abundant and with lower costs. This criterion is at the core of the vertical integration decision of firms. We use countries unemployment rates to proxy for labor availability. Under high unemployment, workers value their current job higher and are willing to accept lower wages to keep the jobs (Habib and Zurawicki, 2002). Hence, we expect unemployment to be positively related with FDI. An important motivation for multinational firms to invest abroad is to take advantage of the availability of natural resources. We examine the role of the resources seeking motive by incorporating into our regressions a composite variable, natural resources depletion (natl_depl), constructed by adding energy depletion (engy_depl) and minerals depletion (min_depl), both 17

18 measured as % of GNI. Energy depletion is equal to the product of unit resource rents and the physical quantities of energy extracted. It covers crude oil, natural gas, and coal. Minerals depletion is equal to the product of unit resource rents and the physical quantities of minerals extracted. It covers bauxite, copper, iron lead, nickel, phosphate, tin, zinc, gold, and silver. Higher natural resources depletion in host countries implies less availability of these resources to be used as inputs in new projects. Investors seeking for natural resources would find less attractive to invest in countries with high natural resources depletion.. Hence, we expect the sign of nat_depl to be negative. The significance of these variables should shed light on countries drive for natural resources and FDI (Cheung and Quian, 2009; Cheung et al, 2011). The incentive to invest could be adversely affected by the presence of other risk factors related to the quality of institutions besides corruption (Méon, and Sekkat 2004; Bénassy-Quéré et al, 2007, Cheung et al, 2011). We include the political system risk index, pol_risk, to proxy for risk. A higher value of the index indicates a lower level of risk. This index is constructed from the 12 country risk indexes from the International Country Risk Guide (ICRG). The sign of its coefficient is expected to be positive. The second step Heckman procedure In the second stage of the Heckman procedure, conditional on countries being selected for investment, the amount of investment is determined. In doing that, an ordinary least squares regression model with correction for selectivity bias is estimated. We assess the impact of CD on FDI using only positive FDI based on the regression equation LFDI ij, t dcor ij, t X ij, t 1 IMRij, t ij, t (2) 18

19 where the dependent variable LFDI ij, t is the natural log of FDI flows from the source country, i to the host country j at time t. The explanatory variables included in (2) are the same as in equation (1) plus IMR,. The selection bias problem, caused by the presence of excessive zeros ij t in the data, is controlled for by using the inverse Mills ratio (IMR). The ratio that contains information about the unobserved factors that determine FDI from source to host countries is retrieved from equation (1) and included in the second stage of the Heckman regression. The significance of the inverse Mills ratio reflects the importance of selection bias 5.,, and are parameters to be estimated and ij, t is the error term. 3.3 Asymmetrical effects of CD on FDI Flows Conceivably, absolute CD could be masking different effects that either positive or negative CD values may have on FDI flows. For example, if the corruption index for Australia is 5 and the corresponding corruption index for China is 2, then, according to our calculations in section 3.2, the absolute value of CD between Australia and China should be 3. However, this 3 could come from +3 or -3 depending on whether Australia is the source or the host country. In the case of +3, FDI is allocated from a less corrupt country, Australia, into a country relatively more corrupt, China. Our conjecture is that when investors from the source country face a positive differential in corruption, higher corruption abroad, they would be less eager to invest in the host country, reducing FDI flows (in our example) from Australia to China. Likewise, from the point of view of China as a source country, the differential will be -3. In this case, Chinese 5 The inverse Mills ratio is given by the probability density function over the cumulative distribution function estimated in the first stage, which includes both zero and non-zero observations. Intuitively, the ratio captures the effect of truncating the sample and is included to control for selection biases in the second stage regression, which uses only positive (but not zero ) FDI observations. 19

20 investors would find Australia more attractive as FDI destination; this should increase FDI from China to Australia. In this sub-section we analyze the asymmetrical effect of corruption distance to disentangle the different effects of positive and negative CD in FDI flows. To conduct our analysis, we conduct regressions on the following equations D ij, t 1dcor ( ) ij, t 2dcor ( ) ij, t X ij, t 1 ij, t (3) LFDI ij, t 1dcor ( ) ij, t 2dcor ( ) ij, t X ij, t 1 IMRij, t ij, t (4) Where D ij, t is defined as in equation (1). is the positive corruption distance, and corresponds to the negative corruption distance. We use the same set of controls and the two stages Heckman procedure as in equations (1) and (2). Equation (3) is the corresponding probit of the Heckman first stage and equation (4) is the OLS regression model with correction for selectivity bias. The different estimates for and should provide with information about the asymmetrical effects of CD on FDI flows. 3.4 The Effects of CD on FDI between developed and developing countries Most of the FDI flows take place among developed countries. For example in 2007, this group was the recipient of 69 percent of total FDI inflows (UNCTAD STAT). However for the last decade, developing countries have been experiencing a boom of FDI s inflows. Developing and transition economies FDI inflows grew by an astonishing 69 percent over a decade, from $201 billion in 1997 to $656 billion in 2007 (UNCTAD STAT). A related recent phenomenon is the dramatic increase of FDI s outflows from developing to both developing and developed countries. Figure 2 plows the annual percentage changes of FDI outflows from the twenty seven 20

21 developing countries in our sample since FDI outflows from these countries increased eight-fold with a 48.3 net percentage points change from the year 2000 to Part of the conventional wisdom is that corruption in developing countries is generally higher and more widespread. Against this backdrop, in this part of the study, we seek to shed light on the role of CD on FDI acknowledging countries development standing and relative corruption. Our analysis is based on the conjecture that investors would react differently when they face the choice of allocating investment between developed and developing countries and whether the host is more corrupt or less corrupt than their own country. In particular, if the source of FDI is a country with relative low corruption, regardless of being in the developed or developing category, and investors have the option of investing, say in either Italy or Turkey, two countries with similar levels of relative high corruption, they would feel more comfortable investing in the former since Turkey is a developing country. To test our conjecture, we split out the sample into four categories of FDI flows. (i) Outflows from developed to developed countries; (ii) outflows from developing to developing; (iii) Outflows from developed to developing; and (iv) Outflows from developing to developed. Additionally, we separate the positive and negative effects of corruption distance, i.e. dcor(+) and dcor(-). This procedure would allow us to better identify not only the effects of similarity in corruption, but in what direction of corruption level FDI is flowing. Similar to the empirical strategy adopted in subsections 3.2 and 3.3, we present the Heckman two-stage regression results based on equations (1)-(4). 21

22 4. Estimation results 4.1 The effects of CD on FDI flows We postulate that the likelihood of FDI from source to host countries is negatively related with the absolute CD. This claim is validated by the results obtained from the regressions on equations (1) and (2) reported in table 2. We report the results for two specifications. Model 1 shows the analysis with the variable of interest, dcor (absolute CD), and the traditional gravity variables, GDP, geographical distance, common legal system, border and language. In model 2, we augment model one by incorporating GDP growth, political risk, natural resources depletion, unemployment, and openness. We focus our analysis on model 2, the complete model. The results from the first and second stages Heckman procedure in table 2 are displayed under de headings first stage and second stage respectively. The likelihood of FDI decreases with greater absolute CD the coefficients of dcor are negative and highly significant. The implication is straightforward, when faced with the decision of in which countries to invest firms are less willing to undertake foreign operations in countries with less similar levels of corruption. The regression results in the second stage show that the more different two countries are in corruption level, the smaller the amount of FDI invested in the host country. These findings are in line with the results reported by Habib and Zurawicki (2002) and Cuervo-Cazurra (2006) among others. The magnitude of the reduction is quite large, for each additional point of CD, FDI drops by around 11 percent. The likelihood to invest is positively affected by the source and host country GDP, common legal system, common language, colonial ties, host GDP growth, unemployment rate and openness, and negatively by geographical distance, and host natural resources depletion. Once the decision to invest is taken, the amount of investment is roughly influenced by the same 22

23 factors affecting the likelihood to invest. Host countries with higher GDP, low political risk, lower natural resources depletion, greater openness, and sharing with the source country common legal system, common language, and colonial ties receive higher amounts of FDI. Host GDP growth does not influence the amount of investment, and sharing common border does not have an impact on neither the likelihood to invest nor the amount of investment. Overall, the estimates for the control variables provide support for the market seeking and natural resources seeking motives and indicate the importance of political risk, cultural distance and international orientation of countries on the allocation and amount of FDI flows. 4.2 Asymmetrical effects of CD on FDI Flows In this part of the paper we analyze the asymmetrical effects of CD on FDI. As mentioned in subsection 3.3, it is conceivable investors may react differently when faced with the decision to invest in countries with higher or lower corruption levels compared with their own country s corruption. While FDI flows from relatively less to more corrupt countries will be deterred, investors from more corrupt countries will find attractive to invest in less corrupt hosts. We capture these features by incorporate into our analysis measures of corruption based on positive and negative CD. We find that positive CD promotes FDI, while negative CD hinders FDI 6. Similar to table 2, the results from the first and second stages Heckman procedure are presented in table 3 under de headings first stage and second stage respectively. The estimates of dcor(+) and dcor(-), displayed under first stage in table 3, are highly significant 6 Given our definition of CD and since a higher number in the corruption index indicates low corruption, and viceversa, a positive CD implies lower corruption in host country, and a negative CD results from higher corruption in a source country. 23

24 and with the expected signs. The likelihood to invest is negatively related to positive CD and positively affected by negative CD. It is noteworthy to point out that the likelihood to invest in more corrupt countries drops almost twice as much as the increase of the likelihood to invest in less corrupt countries. Turning our attention to the column headed second stage in model 2, the estimates suggest that the amount of FDI allocated is adversely influenced by positive CD and positively related with negative CD. The estimates of the rest of controls from the two-stage analysis are very similar in both sign and size to those reported in table 2. These results provide support to our hypothesis that investors react differently depending upon whether they are from a relatively high or low corrupt country. 4.3 The Effects of CD on FDI between developed and developing countries In this sub-section we extend our analysis to test the hypothesis that investors would react differently to the risk of corruption when they face the choice of allocating investment between developed and developing countries, taking into consideration the stage of development of their own country and the asymmetry of corruption between the source and host countries. We split out the data set into four country-pair groups based on development stage and sourcehost status. Again, we focus our discussion on the analysis of the full model specification results. Table 4 pertains to absolute CD and FDI. It displays the results of regressing equations (1) and (2). Recall that the lower absolute CD, the more similar in corruption level country-pairs are. The likelihood to invest is only affected by absolute CD between developing-developing country pairs. The coefficient is negative, as expected, but with a level of significance of 10 percent. CD does not have an effect on the amount invested, however. Investors from developing countries are deterred to invest in other less similar, in terms of corruption, developing countries. 24

25 For the other three groups of country pairs, the likelihood to invest is not affected by CD, but the amount of investment from developed to developed countries is negatively impacted by greater absolute CD. Table 5 present the regression results for equations (3) and (4) for the four country-pair groups. The likelihood to invest is positively affected by negative CD when developed countries are the source, regardless of the stage of development of the host countries, and negatively impacted by positive CD in developing-developing country-pairs. The probability to invest is neither affected by positive CD in developed-developing and developing-developed nor by negative CD in developing-developed and developing-developing country-pairs. The results from the second stage columns indicate that only in the group of countrypairs developed-developed the negative CD has a positive impact on the amount of investment. From these results, a clearer image emerges. Investors from low-corruption, developed countries decide to invest in other countries with similar low-corruption levels, regardless of the host country s stage of development. The similarity in low-corruption has an impact on the amount of investment allocated into other developed countries. Investors from developing countries are discouraged to invest in other developing countries with higher corruption levels. But their investment decisions are not affected by higher corruption in developed countries. 5. Concluding Remarks This paper is concerned with the effects of corruption on FDI flows at the bilateral level. We explicitly incorporate into an augmented gravity model the idea of corruption distance to analyze the impact of corruption and its asymmetrical effects on FDI. We examine a data set of 47 developed and developing countries with different levels of corruption over the time span of 25

26 1997 to Our empirical strategy deals with the presence of excessive zeros in the data by adopting the two-stage Heckman estimation procedure. This method accommodates for sample selection bias and allow us to sequentially analyze the decision making process. We find that in countries pairs with a similar high/low level of corruption, or equivalently low absolute corruption CD, FDI flows increase. When faced with the decision of in which countries to invest, firms are less willing to undertake foreign operations in countries with less similar levels of corruption. Once investors decide where to invest, we find that the more different two countries are in corruption level, the smaller the amount of FDI invested in the host country. The regression result is consistent with the hypothesis that, overall, the more different two countries are in corruption level, the amount of FDI is smaller. It is found that FDI respond positively to market opportunities, availability of natural resources, unemployment, and international orientation, and negatively to political risk, and geographic and cultural distance. To better identify in what direction of corruption level FDI flows, we separate the positive and negative effects of corruption distance. While a positive corruption distance implies source country has better institutional environment and less corruption, a negative corruption distance means host country is relatively less corrupt. We find that the likelihood to invest and the amount of investment are negatively impacted by positive corruption distance and positively related to negative corruption distance. Our empirical findings provide support to our hypothesis that investors react differently depending upon whether they are from a relatively high or low corrupt country. Investors from less corrupt countries are less suited to handle corruption; this 26

27 disadvantage would deter companies from these countries to invest in countries with higher levels of corruption. We document that not all investors care the same about corruption. For example, investors from developing countries are deterred to invest in other less similar, in terms of corruption, developing countries. But higher levels of corruption in developed countries do not have effect on FDI outflows from developing countries. FDI outflows from developed-high corruption countries is attracted to developed-low corruption countries but have not effect when host countries are developing-low corruption. These findings are especially relevant, for they shed new light on explanations to increasing FDI activities undertaken by developing countries in recent years. 27

28 References Abed, G., and Davoodi, H., 2000, Corruption, Structural Reforms and Economic Performance in the Transition Economies, IMF Working Paper No Akçay, S. (2001) Is Corruption an Obstacle for Foreign Investors in Developing Countries? Cross-Country Evidence, Yapi Kredi Economic Review 12 (2): Alesina, A. and B. Weder, (2002), "Do Corrupt Governments Receive Less Foreign Aid?" American Economic Review, 92 (4): Bénassy-Quéré, A., Coupet, M., and Mayer, T., 2007, Institutional Determinants of Foreign Direct Investment, The World Economy, Brainard, L., 1993, A simple theory of multinational corporations and trade with a trade-off between proximity and concentration, NBER working paper No 4269, National Bureau of Economic Research, Cambridge, MA. Brenton, P., Di Mauro, F. and Lucke, M., 1999, Economic integration and FDI: an empirical analysis of foreign investment in the EU and in central and eastern Europe, Empirica, 26(2): Cheung, Y.W., and Qian, X., 2009, Empirics of China's Outward Direct Investment, Pacific Economic Review, 14: Cheung, Y.W., de Haan, J., Qian, X., and Shu, Y., 2011, China s Outward Direct Investment in Africa, HKIMR Working Paper No. 13/2011, Hong Kong Institute for Monetary Research. Cuervo-Cazurra, A., 2006, Who Cares about Corruption? Journal of International Business Studies, 37: Davis, R.B., and Kristjánsdόttir, H., 2010, Fixed Costs, Foreign Direct Investment, and Gravity with Zeros, Review of International Economics, 18(1): Eaton, J. and Tamura, A., 1995, Bilateralism and regionalism in Japanese and US trade and direct foreign investment patterns, NBER working paper No 4758, National Bureau of Economic Research, Cambridge, MA. Eichengreen, B. and Irwin, D., 1995, Trade Blocks, Currency Blocs and the Reorientation of Trade in the 1930s, Journal of International Economics 38(1-2): Egger, P. and Winner, H., 2005, Evidence on Corruption As An Incentive For Foreign Direct Investment, European Journal of Political Economy, 21(4):