THE EFFECTIVENESS OF LAWS AGAINST BRIBERY ABROAD *

THE EFFECTIVENESS OF LAWS AGAINST BRIBERY ABROAD * Alvaro CUERVO-CAZURRA Sonoco International Business Department Moore School of Business University of South Carolina 1705 College Street, Columbia, SC 29208 USA. Tel.: 1-803-777-0314, Fax: 1-803-777-3609, acuervo@moore.sc.edu August 31, 2007 Forthcoming: Journal of International Business Studies Abstract I analyze the effectiveness of laws against bribery abroad in inducing foreign investors to reduce their investment in corrupt countries. The laws are designed to reduce the supply of bribes by foreign investors by increasing the costs of bribing abroad. Such increase in costs will make foreign investors more sensitive to corruption and induce them to reduce their investments in corrupt countries. However, I argue that these laws need to be implemented and coordinated in multiple countries to become effective. Otherwise, investors in a country will have incentives to bypass them when competitors from other countries are not bound by similar legal constrains. The empirical analysis shows that investors from countries that implemented the OECD Convention on Combating Bribery of Foreign Public Officials in International Business Transaction of 1997 reduced their investments in corrupt countries. Investors from the US, which were bound by the Foreign Corrupt Practices Act of 1977, also reduced investments in corrupt countries, but only after the OECD Anti-Bribery Convention was in place. Key words: corruption, foreign direct investment, law, institutions JEL: F21, F23 * I thank anonymous referees, the Special Issue Editor Witold Henisz, Kendall Roth, and participants at the Academy of International Business annual meeting for useful suggestions, and the Center for International Business Research and Education at the University of South Carolina for financial support. All errors are mine.

2 I analyze the effectiveness of laws against bribery abroad in inducing foreign investors to reduce their investments in corrupt countries. Corruption, the abuse of public power for private gain, is detrimental for the country (Eliott, 1997; Kaufmann, 1997; Rose-Ackerman, 1999; Svensson, 2005; Tanzi, 1998; Wei, 2000a). It acts like an irregular tax, increasing costs and uncertainty and distorting incentives to investment (Shleifer & Vishny, 1993). As a result, countries with high corruption show lower levels of economic growth (Mauro, 1995) and distorted government expenditure (Mauro, 1998). Hence, reducing corruption is important for the development of the country. All countries have laws that punish bribery to reduce the demand for bribes by politicians and government officials. However, in many cases such legislation is not effective in countries with high corruption because judges may be open to accept a bribe to alter the application of the anti-corruption law. Additionally, the government official demanding the bribe may be a politician with the power to alter the legislation or its implementation. An alternative is to reduce the supply of bribes by foreign investors by introducing laws against bribery abroad in countries with better institutions. Such laws reduce the incentives for corruption by increasing the cost and risk of detection of the multinational enterprise (MNE) that pays a bribe to a foreign government official 1. Such laws may be effective in combating corruption abroad because the largest foreign investors come from countries with low corruption and effective judicial systems. MNEs would not be able to bribe politicians and judges at home not to implement the laws. However, studies of the effectiveness of the US Foreign Corrupt Practices Act (FCPA) of 1977 have cast 1 In addition to increasing the costs and risks of corruption, there are other alternatives to reduce corruption such as reducing government intervention and discretion, increasing competition, reducing incentives for bribes, improving transparency in government activities, or government decentralization (Ades & Di Tella, 1997, 1999; Fisman & Gatti, 2002; Shleifer & Vishny, 1993; Rose-Ackerman, 1999). I do not discuss these in the paper. doubts about the effectiveness of such laws. On the one hand, Hines (1995) finds that after the introduction of the FCPA US investors reduced investments in countries with corruption. On the other hand, Wei (2000b) and Smarzynska and Wei (2000) find no additional sensitivity to corruption in the host country by US investors in comparison to investors from other countries. In this paper I argue that to reduce the supply of bribes by foreign investors, countries need not only to implement laws against bribery abroad, but also coordinate this implementation with other countries. If only one country implements laws against bribery abroad, the law may not be effective in dealing with bribery. The corrupt government official may just demand a bribe from competitors from other countries that do not have a legal constraint. As a result, the firm from the country where there are laws against bribery abroad faces a prisoners dilemma because although it does not want to pay the bribe, it has an incentive to find ways to get around the law to remain competitive. However, when not only the foreign firm but also its main foreign rivals are bound by similar laws, the laws may be effective because they provide a level playing field with all foreign firms facing similar increases in costs if they bribe. I test this idea by analyzing the effectiveness of the Organization for Economic Cooperation and Development (OECD) Convention on Combating Bribery of Foreign Public Officials in International Business Transactions in inducing investors from the countries that implemented it to reduce their investment in corrupt countries. The analysis of FDI flows reveals that investors from countries that implemented the Convention have become more sensitive to host country corruption. These investors have reduced their FDI in corrupt countries after their home country had laws against bribery abroad in place. I also analyze the behavior of US investors, who have been bound by the FCPA since 1977. These investors reduced their FDI in corrupt countries, but only after the OECD Anti-Bribery Convention was in place. Hence, laws against bribery abroad appear to be effective in making investors more sensitive to host country corruption, but only

3 when the laws are in place and coordinated in multiple countries. The paper contributes to two areas of research, one theoretical and another thematic. First, the paper contributes to the literatures on institutions and firms, which has discussed how institutions affect companies (e.g. Djankov, La Porta, Lopez-de-Silanes & Shleifer, 2002; La Porta, Lopez-de-Silanes, Shleifer & Vishny, 1998; North, 1990), and on institutions and international business, which has discussed how institutions in the host country affect the behavior of foreign investors (e.g. Bevan, Estrin & Meyer, 2004; Boddewyn, 1988; Delios & Henisz, 2003a, 2003b; Henisz, 2000). It extends this line of thinking by arguing that institutions in the host country interact with institutions in the home country to affect the behavior of foreign investors. Second, the paper contributes to studies of corruption, which have discussed how corruption in the host country reduces FDI and changes the behavior of the foreign investor (e.g. Cuervo-Cazurra, 2006; Habib & Zurawicki, 2002; Lambsdorf, 2003; Rodriguez, Uhlenbruck & Eden, 2005; Smarzynska & Wei, 2000; Svaleryd, Hakkala, & Norback, 2005; Uhlenbruck, Rodriguez, Doh & Eden, 2006; Voyer & Beamish, 2004; Wei, 2000b). It helps settle the debate about the effectiveness of laws against bribery abroad (e.g. Hines, 1995; Wei, 2000b; Smarzynska & Wei, 2000) by indicating that these laws are effective in discouraging foreign investors from investing in corrupt countries, but only when multiple countries implement them. LAWS AGAINST BRIBERY ABROAD Although all countries have laws banning bribery at home, only a few countries have laws against bribery abroad. The first country to implement them was the US in 1977. Twenty years later, the 30 members of the OECD and another 6 non-member countries (Argentina, Brazil, Bulgaria, Chile, Estonia and Slovenia) agreed to implement them. International agencies like the United Nations, World Bank, World Trade Organization, Organization of American States, and the Council of Europe have also encouraged the development of such laws. US Foreign Corrupt Practices Act of 1977 A 1976 investigation by the Securities and Exchange Commission (SEC) revealed that more than 400 companies, including 117 of Fortune 500, disclosed paying questionable or illegal payments. Congress viewed this as negatively affecting foreign relations, damaging the image of the US, and reducing public trust on corporations. The initial legislative effort considered only requiring the disclosure of bribes and criminalizing the failure to disclose. However, Congress deemed that rather than legalize through disclosure, it was better to criminalize bribery. This would be easier to implement and it would solve the paradox that domestic bribes were illegal whereas foreign ones were not. The FCPA that was created as a result has three main requisites 2 : accurate record keeping, effective internal accounting control systems, and prohibition of corrupt payment to foreign officials, politicians, and political candidates. The FCPA provides for penalties of up to US$1mn for a firm and US$10,000 and 5 years of prison for an employee. It makes it illegal not only the direct payment to foreign officials or politicians but also the payment to other people -facilitating agents- who would pay on the firm s behalf. However, the FCPA is explicit in not penalizing grease payments, or payments to foreign officials to expedite procedures, such as customs procedures or permits, that would otherwise occur without a bribe, but more slowly. Although grease payments are illegal in the US, the legislators considered that they were an accepted business practice in other countries. In case of doubt about the legality of a payment, the firm can ask the opinion of the Attorney General regarding whether the payment conforms to the FCPA. At the time of its passing, the legislation was perceived to be beneficial to US firms as it would bind them not to pay bribes. One executive argued that such legislation would make it easier not to pay bribes because the firm could claim that it could not because of the legislation. 2 The legislative history appears in US Congress (1977) and US Senate (1977) and is summarized in Hines (1995) and Seitzinger (1999).

4 However, after its implementation the FCPA received numerous criticisms, which led to its reform in 1988. Critics contended that the FCPA placed US firms at a disadvantage against foreign competitors that were not constrained in paying bribes (Kaikati & Label, 1980). Others argued that the FCPA was too vague in its separation of grease payments from corrupt payments. Hence, the FCPA was amended as Title V of the Omnibus Trade and Competitiveness Act of 1988. One of the key changes was that the firm no longer had to have reason to know that payments given to an agent would be passed on as a bribe. Instead, it had to know that such illegal payments were done. The amendment increased the maximum criminal fines for violating the FCPA to US$2mn for a firm and to US$100,000 for an individual, added a new civil penalty of $10,000, and maintained the maximum imprisonment for an individual at 5 years. There is some evidence that the FCPA changed the behavior of US firms and made them more sensitive to host country corruption because it increased the cost of bribing abroad. Hines (1995) analyzes the difference in the growth of investments by American firms between 1977, the year of the implementation of the FCPA, and 1982. He finds that US firms reduced their investments in countries with high corruption after the passage of the FCPA. He interprets this finding as evidence that the FCPA served to discourage US investors from engaging in bribery abroad. US firms were now reluctant to invest in countries where they would have to bribe to operate. This idea leads to hypothesize that: Hypothesis 1a. US investors will invest less in corrupt countries than investors from other countries. However, there is also evidence that the FCPA has not been effective in altering the behavior of US firms. Wei (2000b) compares 1993 FDI stocks in 45 host economies of US investors to those of investors from other 11 developed countries. He finds that US investors do not reduce their investments in corrupt countries more than others. He discusses three plausible and not mutually exclusive explanations for this finding: corruption reflects poor governance in the country and hurts all investors regardless of whether the country of origin has laws prohibiting it, all investors are deterred by corruption regardless of whether the country of origin has laws prohibiting bribery, or US firms find ways of bribing despite the existence of the law. Smarzynska and Wei (2000) confirm this finding when they study foreign investments in transition economies in 1989-1995; US firms do not invest less than others. This leads to the competing hypothesis that: Hypothesis 1b. US investors will not invest less in corrupt countries than investors from other countries. These two hypotheses will help settle the debate regarding the effectiveness of the FCPA in making US investors more sensitive to host country corruption and leading them to reduce their investments in corrupt countries. OECD Convention on Combating Bribery of Foreign Public Officials in International Business Transactions of 1997 After 20 years during which the United States was the only country with legislation banning bribery abroad, on November 21, 1997, the 30 members of the OECD and an additional six non-members (Argentina, Brazil, Bulgaria, Chile, Estonia and Slovenia) signed the Convention on Combating Bribery of Foreign Public Officials in International Business Transactions (OECD, 1997). The Convention, which came into effect on February 15, 1999, established a general framework that criminalizes bribing foreign officials to provide the firm with an improper advantage. The Convention prohibits bribing not only government officials but also officials of public international organizations. It encourages mutual legal assistance in the investigations and allows for extraditions. Additionally, the Convention requires stricter accounting standards, external auditing, and internal controls. A companion agreement disqualifies bribes from being taxdeductible business expenses (OECD, 1996). However, the Convention does not make an offence the use of facilitation payments or grease money to expedite procedures. The Convention provides guiding principles for dealing with corruption abroad, but each signatory country has to develop or adapt its national legislation to criminalize the

5 payment of bribes abroad. The OECD does not have the ability to monitor or impose fines on firms that break the Convention. Instead, it uses a name and shame process to pressure governments to introduce legislation and implement it. The OECD, through its Working Group on Bribery, provides systematic monitoring of the implementation of the Convention s standards by each signatory country. This monitoring involves two phases. In phase 1, the national legislation is checked for conformity with the Convention. This is done by the country examined and two other countries that act as lead examiners, with all countries participating in the final evaluation. Progress in creating the appropriate legislation has varied widely across countries (OECD, 2006). Table 1 lists the countries and the year when national legislation criminalizing bribery abroad entered in force 3. Some countries, like Austria, Greece, Hungary, Iceland and the US, incorporated the requirements of the Convention into their national laws by 1998, but others took much longer. For example, Brazil, Chile, and the UK only did so in 2002, Turkey in 2003 and Estonia in 2004. Insert Table 1 about here Having laws that criminalize bribery abroad is the first step towards reducing the supply of bribes, but these laws need to be enforced. For this reason in phase 2 the effective implementation of the legislation is evaluated. The evaluation of the implementation of the legislation and prosecution of bribery involves an on-site visit and consultation not only with the national government, including the police and judiciary, but also with the private sector, including firms and unions, and civil society. The findings are then compiled in a report that generates appropriate recommendations. By 3 In the US, the FCPA was altered in 1998 by The International Anti-Bribery and Fair Competition Act of 1998 (US Congress, 1999) to bring it in line with the OECD Anti-Bribery Convention. Among the changes was the redefinition of foreign official to include officials of public international organizations, and the application of the Act to all issuers of securities in the US and not only US ones. For a history of this legislation see Larson (1997) and US Department of Justice (1998). 2007, 30 of the 36 countries had been evaluated for implementation of the legislation; OECD (2007) summarizes the evaluations. The multi-country nature of the Convention has been hailed as supporting the effective limitation in the supply of bribes abroad. First, by imposing the same legal requirements on companies from the leading sources of FDI, the Convention levels the playing field among competitors and reduces the incentives to bypass the laws against bribery abroad. The firm and its main competitors are bound by similar legal requirements. Second, the self and mutual monitoring of the Convention and its application generates pressure on the home governments to enforce the legislation and effectively increase the cost of bribing foreign officials. Moreover, the firm is subject to additional scrutiny not only by the government but also by competitors and non-governmental organizations that increases the probability of detection of bribery, and hence the actual cost of bribing. These arguments support the hypothesis that: Hypothesis 2a. Investors from countries signatories of the OECD Anti-Bribery Convention will invest less in corrupt countries than investors from other countries. However, critics of the Convention argue that even a multi-country approach would not be effective. The non-governmental anticorruption organization Transparency International praises the efforts in combating corruption that followed the Convention, but remains critical of its implementation (Transparency International, 2001). Hamra (2000) finds skepticism among a sample of business leaders regarding the effectiveness of the Convention. Heinmann (2004) discusses the limitations of the Convention. First, it is difficult to prepare foreign bribery cases because it requires investigations abroad and prosecutors may not have the expertise or resources to do so. Moreover, although there is mutual legal assistance among countries participating in the Convention, this is not the case with other countries. Second, in some countries responsibility for prosecuting cases is at the state or provincial level and these may not have the interest or resources to prosecute bribery abroad. Third, companies that loose to competitors who

6 bribe foreign officials may be reluctant to bring complaints for lack of hard evidence, the threat of defamation suits, and the fear of loosing future contracts. Fourth, there is little public awareness in the home country and abroad that bribery abroad has been criminalized. In addition to these limitations, firms may still bribe but have to be more careful about it, such as invoicing a bribe as consulting fees. As a result, investors will not be discouraged from investing in corrupt countries where they will be asked for bribes, but merely be more cautious about how they conduct business there. These ideas support the competing hypothesis that: Hypothesis 2b. Investors from countries signatories of the OECD Anti-Bribery Convention will not invest less in corrupt countries than investors from other countries. These two competing hypotheses and the two on the effectiveness of the FCPA discussed before implicitly test four alternative arguments regarding the effectiveness of laws against bribery abroad in changing the behavior of foreign investors in corrupt countries. The first argument is that the laws are always effective instruments for tackling corruption abroad. Evidence of this argument will be finding that both the Convention and the FCPA alter the behavior of foreign investors in corrupt countries. The second and opposite argument is that laws against bribery abroad are never effective instruments for addressing corruption abroad. This argument will be supported when neither the Convention nor the FCPA alter the behavior of investors in corrupt countries. A third argument is that the laws are effective in tackling corruption, but legislators and firms need time to comply. Evidence of this argument will be finding that the FCPA has affected the behavior of foreign investors in corrupt countries while the Convention has not. The fourth argument is that laws against bribery abroad are effective but only when they are implemented in multiple countries. Evidence of this will be finding that the Convention has affected the behavior of foreign investors while the FCPA has not affected the behavior of US investors until the Convention was in place. I will touch upon which of these ideas is supported in the discussion of the results. RESEARCH DESIGN I test the hypotheses using data on bilateral FDI inflows to 103 host economies 4. I have a cross-sectional panel of 7 years. I include data on FDI between 1996 and 2002, with 1999, the year the OECD Anti-Bribery Convention was ratified as the middle point. The bulk of FDI data comes from the United Nations Conference on Trade and Development (UNCTAD) country profiles (UNCTAD, 2005). This database provides the widest coverage of bilateral FDI inflows in terms of the number of countries available. I complemented this database with information on FDI from the OECD (2005), which has been a usual source of data in other studies. Data on the countries size (GDP, population) came from the World Bank s World Development Indicators (2005). Information on geographic characteristics and languages were derived from the Central Intelligence Agency s World Factbook (2005). Data on colonial histories were derived from the Factbook and from Encyclopedia Britannica (2005). 4 Not all the economies listed are independent countries (e.g. Hong Kong). I nevertheless treat them as independent sources or destinations of FDI. Additionally, data for Belgium and Luxembourg appears together in the UNCTAD reports. I use them as one single source or destination. The host economies for which I have FDI data are: Algeria, Angola, Argentina, Armenia, Australia, Austria, Azerbaijan, Bahamas, Barbados, Belgium/Luxembourg, Belize, Benin, Bermuda, Bolivia, Brazil, Brunei, Bulgaria, Burkina Faso, Burundi, Cambodia, Cameroon, Canada, Cape Verde, Central African Republic, Chad, Chile, Colombia, Comoros, Costa Rica, Cuba, Czech Republic, Denmark, Djibouti, Dominican Republic, Ecuador, El Salvador, Eritrea, Estonia, Ethiopia, Finland, France, Gambia, Germany, Greece, Guatemala, Guyana, Haiti, Honduras, Hungary, Iceland, Ireland, Italy, Jamaica, Japan, Kazakhstan, Korea, Kyrgyzstan, Latvia, Lithuania, Macau, Macedonia, Malawi, Mali, Mauritius, Mexico, Moldova, Mongolia, Morocco, Mozambique, Myanmar, Netherlands, New Zealand, Nicaragua, Norway, Panama, Paraguay, Peru, Poland, Portugal, Russia, Rwanda, Saint Kitts Nevis, Saint Lucia, Sierra Leone, Slovakia, Slovenia, Somalia, Spain, Suriname, Sweden, Switzerland, Tanzania, Trinidad Tobago, Tunisia, Turkey, Uganda, UK, Uruguay, USA, Uzbekistan, Venezuela, Zambia, and Zimbabwe.

7 Table 2 summarizes the variables, measures, and sources of data. The dependent variable is the natural logarithm of the bilateral FDI inflows from a home country to a host country, measured in US$ using the average foreign exchange rate for the year. Insert Table 2 here Corruption is illegal and, as such, difficult to measure with precision. Like previous studies, I have to rely on subjective measures of corruption (see Wei (2000b) for a discussion of the alternative measures of corruption). I use two indicators of corruption. First, I use the indicator of corruption of Transparency International (2005), which is available on an annual basis since 1995. The original indicator measures corruption within an interval of 10 (low corruption) to 0 (high corruption). To avoid awkwardness in the interpretation of the coefficients, I rescale the indicator so that a higher number indicates higher corruption by subtracting the original index from 10. Second, I use the measure of control of corruption developed in the World Bank by Kaufmann, Kraay and Mastruzzi (2003). This measure is only available every other year from 1996 onwards. I complete the missing years with the average of the indicator in the year before and after each missing year. The indicator has a spread of -2.5 (high corruption) to 2.5 (low corruption). I rescaled this it to facilitate the interpretation of the results by deducting it from 2.5 so that a higher number represents higher corruption. I am interested in understanding whether investors from home countries with laws against bribery abroad are more sensitive to the level of corruption in the host country than other investors and, as a result, further reduce their FDI in corrupt countries. Therefore, to capture this additional sensitivity to host country corruption and test the hypotheses I use an interaction between an indicator that the home country has laws against bribery abroad and an indicator of the level of corruption in the host country. I then analyze the impact of this interaction on FDI inflows. Wei (2000b: 8) provides a detailed explanation of this procedure, which he used to test whether US investors were more sensitive to host country corruption than other investors. To test hypotheses 1a and 1b, I multiply a dummy indicator that the country of origin of the investor is the US, which had the FCPA in place since 1977, by the level of corruption in the host country. Hypothesis 1a is supported if the coefficient of this product is negative and statistically significant. Hypothesis 1b is supported if the coefficient of such product is not statistically significant. I further analyze whether the sensitivity of US investors to corruption changed after the adoption of the Convention in 1998 by creating two time-variant variables. One indicates that the investor s country of origin is the US before 1998, the year when the FCPA was modified to incorporate the OECD Anti-Bribery Convention. The other indicates that the investor s country of origin is the US after 1998. I multiply each by the level of corruption in the host country to analyze whether the sensitivity of US investors to corruption changed after the implementation of the Convention. To test hypotheses 2a and 2b I created a time-variant dummy indicator that takes a value of 1 when the home country of the investor has domestic laws against bribery abroad in force and multiplied this by the level of corruption in the host country. Table 1 lists the years when the countries had laws against bribery abroad in force. For US investors, the time-variant dummy indicator takes a value of 1 for the whole period because the FCPA has been in force since 1977. Hypothesis 2a is supported if the coefficient of the product is negative and statistically significant. Hypothesis 2b is supported if the coefficient of the product is not statistically significant. All models include an indicator of the level of corruption in the host country to control for the impact of corruption on FDI regardless of the country of origin of the investor. The models also include an indicator of the home country to control for other characteristics of the home country that may influence FDI. I control for other variables that may explain FDI inflows using a gravity model. The gravity model has been applied to the study of the determinants of FDI flows (e.g. Bevan & Estrin, 2004), and in particular to the impact of corruption on FDI (e.g. Wei, 2000b). The theoretical basis of the gravity models to the

8 study of FDI is the proximity-concentration hypothesis (e.g. Horstmann & Markusen, 1992; Brainard, 1997). The barebones gravity model explains FDI flows based on indicators of the host country s size (GDP and population) and distance (geographic) between home and host countries (Linneman, 1966). To these, I add indicators of the common political and cultural backgrounds, which are common in studies of bilateral trade and bilateral FDI. Theoretically, they build on the idea of distance between countries affecting MNE behavior (e.g. Ghemawat, 2001; Johanson & Vahlne, 1977). Therefore, I use the following set of controls, which are common in gravity models (e.g. Frankel & Rose, 2002; Wei, 2000b) First, I control for the size of the country using indicators of gross domestic product and of population. Larger countries are more likely to attract FDI because MNEs can achieve economies of scale in the country. Second, I control for the geographic distance between the countries using an indicator of the great circle distance, which measures distance on the surface of the earth using longitude and latitude coordinates. Distance indicates the existence of transportation costs that would discourage trade and favor FDI. This distance measure is traditionally complemented with indicators that the country is landlocked or an island or has a common border with the host country, as these characteristics affect the difficulty of transporting products, and therefore the likelihood of undertaking FDI. Cultural similarities are captured with an indicator of the existence of a common language between home and host country. This facilitates the transfer of information across borders and reduces psychic distance. Commonalities in administration are measured with indicators of the existence of a colonial relationship and of a common colonizer. Colonial powers traditionally imposed their administrative traditions in the colonies. Similarities in administrative practices facilitate FDI, as investors benefit from a reduced psychic distance between home and host countries. I use a double-log model with quasifixed-effects and one-year lag to analyze the data, as done by Wei (2000b). In the double-log model I apply natural logarithms to the dependent variable (FDI) and the independent variables (GDP, population, distance) to help make the error term close to homoskedastic (Wei, 2000b: 4). I run analyses with only those bilateral FDI flows that have positive values because only the logarithms of positive numbers are defined. The independent variables are measured one year earlier than the dependent variable to account for the time lag that occurs between the decision to invest and the actual FDI. I use a quasi-fixed-effects specification whereby I control for the home country using a dummy indicator for each country. These home country dummies are designed to capture characteristics of the home country that may affect its FDI abroad, such as size and level of development. I use several methods to analyze the data. The analysis of the effectiveness of laws against bribery abroad in altering the behavior of investors in corrupt countries belongs to a class of treatment estimations known as differencesin-differences, where one compares the differences in outcomes before and after the treatment in a group to another group that has not been subject to the treatment. These difference-in-differences estimations are increasingly popular with applied researchers, but they suffer from problems in the standard errors, such as serial correlation 5. Bertrand, Duflo and Mullainathan (2004) discuss the problems and propose several solutions to them 6. I follow their recommendations and use alternative methods to correct for some of the limitations in the estimation. First, I start with a random effects panel regression as done in most analyses of differences-in-differences and 5 I have a cross-sectional panel of seven years of investment relationships between home country i and host country j. The problems of serial correlation will appear at the home-host country dyad, especially in the interaction between the indicator that the home country has laws against bribery abroad and the indicator of host country corruption. This interaction will start taking values above zero once the laws are in place, and will continue with similar values in years after the laws are passed because the level of corruption in the host country changes little from year to year. Therefore, in the models I will control for problems in the error term in the home-host dyad. 6 I thank an anonymous referee for suggesting this paper.

9 specify an autocorrelation structure AR(1) to control for serial correlation 7. In a crosssectional panel, serial correlation may lead to the overestimation of t-statistics and significance levels, especially in the coefficient of the treatment. However, despite this, of the 65 papers with potential serial correlation problems surveyed by Bertrand, Duflo and Mullainathan (2004), only 5 have some kind of serial correlation correction. Second, I run feasible generalized least squares with corrections for heteroskedasticity among panels and for panelspecific autocorrelation; the random effects panel regression with AR(1) correction that I did first assumes homoskedastaicity and a common autocorrelation across panels, which are assumptions that in many cases do not hold in cross-sectional panels. Third, I allow for an unrestricted covariance structure over time and run a panel regression where I cluster the error terms by home-host dyad with White-Huber corrections. Fourth, I ignore the time series information and the problems in serial correlation that come with it. The easiest way to do this is to average data in two periods, before and after the laws are created, and study the influence of the laws on FDI inflows on a twoperiod panel. However, such approach is only valid when the laws are created at the same time in all countries. When the laws are created at different times in different countries, which is the case here, before and after vary across countries and are not defined for countries that did not pass the laws. In this case I have to modify this approach and run two models. In the first one I regress FDI flows against all the controls. I then take the residuals from this analysis for those countries that have passed laws against bribery abroad and separate these residuals in two groups: residuals from years before the laws were created, and residuals from years after the laws were created. To estimate the impact of the laws against bribery abroad I then run a simple regression on this two-period 7 I use a random effect panel model to be able to compute the effects of the controls, many of which are time invariant, and to correct for time variant shocks that affect investors from a home country to a host country. This is the approach taken by Bertrand, Duflo and Mullainathan (2004). panel of residuals. A more detailed explanation of the advantages and disadvantages of each method appears in Bertrand, Duflo and Mullainathan (2004). I use the following general specification of the empirical models to test the hypotheses: FDI ijt = γ 1 Home country has laws against bribery abroad (US FCPA or OECD Anti-Bribery Convention) it-1 * Host corruption jt-1 + X ijt-1 β + ε ijt where γ 1 is the coefficient of interest, X ijt-1 is a vector of the control variables, β is a vector of the coefficients of the control variables, and ε ijt is the error term. RESULTS Table 3 presents the summary statistics and correlation matrix. Although some of the variables show high correlation coefficients, this is to be expected because the data is a panel. However, the analyses are not subject to multicollinearity. The variance inflation matrix analysis indicators are well below the rule-ofthumb levels that would indicate the presence of multicollinearity except for the indicator of GDP and population when they are in the same equation. I run additional models excluding one or the other and find that the results do not change in sign or significance, revealing limited problems with multicollinearity (Greene, 2005). Insert Table 3 here Although in the theoretical section I discussed the laws against bribery abroad in temporal order, starting with the FCPA and continuing with the OECD Anti-Bribery Convention, I will discuss the results of the analyses in the reverse order, tackling the unknown effectiveness of the OECD Anti- Bribery Convention first and then discussing the effectiveness of the FCPA, which has been studied before and is subject to conflicting results, second. Table 4 presents the results of analyzing the effectiveness of the OECD Anti-Bribery Convention in leading investors to reduce FDI in corrupt countries. These results help test the competing hypotheses 2a and 2b. Models 4a through 4d use the measure of corruption from Transparency International, while models 4e

10 through 4i use the one from World Bank. The results support Hypothesis 2a and not 2b. Specifically, the coefficient of the interaction between the indicator that the home country implemented the Convention into its national laws and the measure of host country corruption is negative and statistically significant in the random effects panel regression with correction for serial correlation (Models 4a and 4e), the feasible generalized least squares with correction for heteroskedasticity and panel-specific serial correlation (Models 4b and 4f), and the panel regression analysis with clustered and White- Huber error (Models 4c and 4g), while the coefficient of the indicator of when the home country implemented the Convention into its national laws is negative and statistically significant in the two-period panel regression of the residuals from the regression of FDI inflows on all controls (Models 4d and 4h). In other words, investors from countries that implemented the OECD Anti-Bribery Convention have become more sensitive to host country corruption and reduced their investments in corrupt countries. This heightened sensitivity to corruption is in addition to the general negative influence of host country corruption on FDI that all investors experience, which has been found in other studies (e.g. Cuervo-Cazurra, 2006; Habib & Zurawicki, 2002; Lambsdorff, 2003; Wei, 2000b). This is the first empirical test of the effectiveness of the OECD Anti-Bribery Convention. Despite the criticisms raised against the Convention, it appears that it has been effective. Insert Table 4 here Table 5 provides the results of analyzing the effectiveness of the FCPA in inducing US investors to limit their FDI in corrupt countries. These results serve to test the competing hypotheses 1a and 1b. Models 5a, 5b and 5c use the measure of corruption from Transparency International and models 5d, 5e and 5f use the measure of corruption from World Bank. The results support Hypothesis 1a and not 1b. The coefficient of the interaction between the indicator that the host country is the US and the measure of host country corruption is negative and statistically significant in the random effects panel regression with correction for serial correlation (Models 5a and 5d), the feasible generalized least squares with correction for heteroskedasticity and panel-specific serial correlation (Models 5b and 5e), and the panel regression analysis with clustered and White- Huber error (Models 5c and 5f) 8. These results imply that the FCPA is effective in inducing US investors to reduce their presence in corrupt countries. This finding is contrary to those of Wei (2000b) and Smarzynska and Wei (2000b), who found that US investors were not more sensitive to host country corruption than other investors. One explanation for the difference in findings is the period of analysis of the studies. Both Wei (2000b), who studied FDI stocks in 1993, and Smarzynska and Wei (2000), who studied investments in 1989-1995, analyzed US investment before the OECD Anti-Bribery Convention was in place. Thus, US investors had the incentive to bypass the FCPA to remain competitive abroad. In this study, once the OECD Anti-Bribery Convention made the main competitors be bound by similar legislation, the incentive is to bypass the FCPA diminishes. However, to make sure that it is the period of study that explains the difference in findings between their studies and this one, we need to run additional analyses. Insert Table 5 here Therefore, I further investigate the idea that the OECD Anti-Bribery Convention made US investors sensitive to host country corruption and, as a result, they reduced investments in corrupt countries. To do this, I separate investments by US firms in two periods, before and after 1998, the year when the FCPA was modified to incorporate the requirements of the OECD Anti-Bribery Convention, and analyze whether the behavior of US investors differs between periods. Table 6 presents the results of 8 I cannot run the model where I ignore the time series information because the FCPA was in place throughout the period of analysis. Therefore, there is no before and after period where I can compare the effect of the FCPA on the residuals from the regression of the controls on FDI inflows. However, I do run this analysis when analyzing the behavior of US investors once the OECD Anti- Bribery Convention was in place.

11 these analyses. The results show that before the OECD Anti-Bribery Convention, US investors were not more sensitive to host country corruption than other investors, as found in other studies. However, after the OECD Anti-Bribery Convention, US investors became more sensitive to host country corruption than other investors and further reduced their FDI in corrupt countries. Specifically, the coefficient of the interaction between the indicator of US investments before 1998 and the measure of host country corruption is not statistically significant in any of the models. However, the coefficient of the interaction of the indicator of US investments after 1998 and the measure of host country corruption is negative and statistically significant in the random effects panel regression with correction for serial correlation (Models 6a and 6d), the feasible generalized least squares with correction for heteroskedasticity and panel-specific serial correlation (Models 6b and 6e), and the panel regression analysis with clustered and White- Huber error (Models 6c and 6f), while in the model where we ignore the time series information (Models 6d and 6g) the coefficient of US investments after 1998 on the residuals from the influence of the controls on FDI inflows is negative and statistically significant 9. Therefore, the results support the idea that US investors, which were bound by the FCPA, reduced investments in corrupt countries, but only after the OECD Anti-Bribery Convention was in place. Insert Table 6 here In sum, the results support three important new findings. First, they support the idea that laws against bribery abroad are 9 In this analysis I cannot compare US investments before and after 1998 in the same equation because they are mutually exclusive categories that become collinear in the analysis. As I explained before, the residual analysis only takes data from the country with the treatment. Hence, we can only run one indicator at a time. If instead of running the model with US investments after 1998 on the residuals we run the model with US investments before 1998 on the residuals, the coefficient is positive and statistically significant. US investors were not further deterred by host country corruption before the OECD Anti-Bribery Convention. effective in making investors become more sensitive to host country corruption and, as a result, further reduce their investments in corrupt countries. However, this occurs only when there is a coordinated effort at reducing the supply of bribes and the laws against bribery abroad are implemented in multiple countries. This multicountry effort diminishes the prisoners dilemma problem faced by investors from a country that is subject to legal constraints when investors from other countries do not face such constrains. Second, the results are the first empirical test of the effectiveness of the OECD Anti-Bribery Convention in inducing investors from home countries that have laws against bribery abroad to reduce investments in corrupt countries. Despite criticism, the OECD Anti-Bribery Convention appears to be effective. Third, the results solve some of the conflicting views on the effectiveness of the FCPA. The FCPA is effective on making US investors more sensitive to host country corruption and reduce investments in corrupt countries, but only when other countries adopted similar legislation. These results should be interpreted with caution, however. They only indicate that investors from home countries with laws against bribery abroad reduced their presence in corrupt countries after the OECD Anti-Bribery Convention was in place. Whether the OECD Anti-Bribery Convention or US FCPA have led to higher numbers of identification or prosecutions of bribery abroad was not tested here; future research can analyze this. CONCLUSIONS Laws against bribery abroad established in countries with good institutions can help deal with bribery in corrupt countries by reducing the supply of bribes. However, I argued that the laws need to be present and coordinated across multiple countries to provide a level playing field and reduce the incentive to bypass them. Despite doubts about the effectiveness of such laws, the empirical analyses showed that investors from countries that implemented the OECD Anti-Bribery Convention invested less in corrupt countries. Moreover, US investors, which are bound by the FCPA, have also invested less in corrupt countries, but only after the Convention was in place. These results serve as evidence that laws against bribery abroad

12 need to be coordinated across multiple countries to be effective. The findings provide the basis for future developments in the area of firms and institutions. They highlight the need to incorporate in future analysis of MNEs not only the institutional characteristics of the host country where the MNE invests but also the institutional characteristics of the home country where the MNE comes from. Additionally, the study indicates that future studies on corruption need to separate among foreign investors by their country of origin because they are likely to react to host country corruption differently. In terms of policy recommendations, the study illustrates how the creation and implementation of laws against bribery abroad can be effective in influencing the behavior of investors. To be effective, however, the laws need to be coordinated across countries. This highlights the benefits of cross-country collaboration and the value added provided by international institutions. All in all, the paper illustrates how we have started turning the page on viewing corruption abroad as an acceptable or necessary way of doing business and instead fight it. Coordinated efforts to reduce corruption can bear fruit. REFERENCES Ades, A., & Di Tella, R. 1997. National champions and corruption: some unpleasant interventionist arithmetic. Economic Journal, 107 (443): 1023-1043. Ades, A., & Di Tella, R. 1999. Rents, competition, and corruption. American Economic Review, 89 (4): 982-993. Bertrand, M., Duflo, E., & Mullainathan, S. 2004. How much should we trust differences-in-differences estimates? Quarterly Journal of Economics, 119 (1): 249-275. Bevan, A. A., & Estrin, S. 2004. The determinants of foreign direct investment into European transition economies. Journal of Comparative Economics, 32 (4): 775-87. Bevan, A.A., Estrin, S., & Meyer, K. 2004. Foreign investment location and institutional development in transition economies. International Business Review, 13(1): 43-64. Boddewyn, J. J. 1988. Political aspects of MNE theory. Journal of International Business Studies, 19 (3): 119-144. Brainard, L. 1997. An empirical assessment of the proximity-concentration trade-off between multinational sales and trade. American Economic Review, 87 (4): 520 544. Central Intelligence Agency. 2005. World Factbook http://www.cia.gov/cia/publications/fact book/ Accessed January 10, 2005. Cuervo-Cazurra, A. 2006. Who cares about corruption? Journal of International Business Studies, 37 (6): 803-822. Delios, A., & Henisz, W. J. 2003a. Political hazards, experience and sequential entry strategies: the international expansion of Japanese firms, 1980-1998. Strategic Management Journal, 24 (11): 1153-64. Delios, A., & Henisz, W. J. 2003b. Policy uncertainty and the sequence of entry by Japanese firms, 1980-1998. Journal of International Business Studies, 34 (3): 227-241. Djankov, S., La Porta, R., Lopez-de-Silanes, F., & Shleifer, A. 2002. The regulation of entry. Quarterly Journal of Economics, 117 (1): 1-37. Elliot, K. 1997. Corruption as an international policy problem: Overview and recommendations: In Elliot, K. (Ed.) Corruption and the Global Economy: 175-233 Washington, D.C.: Institute for International Economics. Encyclopedia Britannica. 2005. http://www.britannica.com/ Accessed January 8, 2005. Frankel, J., & Rose, A. 2002. An estimate of the effect of common currencies on trade and income. Quarterly Journal of Economics, 117 (2): 437-466. Fisman, R., & Gatti, R. 2002Decentralization and corruption: evidence across countries. Journal of Public Economics, 83 (3): 324-344. Ghemawat, P. 2001. Distance still matters. The hard reality of global expansion.