The effect of corruption on foreign direct investments in South America : a panel data analysis

(1)

1

The effect of corruption on foreign direct investments in

South America: A panel data analysis

Bachelor thesis July 2016

Name:

Adriaan van Hall

Studentnumber: 10248080

Specialization:

Economics and Finance

Supervisor

Oana Furtuna

Abstract

This paper looks at the highly debated effect corruption has on foreign direct investments (FDI). Using data from 10 countries in South America, a panel data analysis is performed, controlling for unobserved country specific factors, to analyze whether the general theory of corruption – which predicts that corruption is negatively related to FDI – holds true. Transparency International’s Corruption Perceptions Index (CPI) is used as a proxy for corruption, with a high CPI score indicating low levels of corruption. The results show that CPI is positively related to FDI, but when controlling for a country’s democracy level and openness, this effect becomes insignificant. However, after adding a control variable which measures whether corruption has a different effect on more corrupt countries, and adding time-specific dummy variables, the results show that the CPI coefficient stays positive at 1% significance, providing evidence that a 100-point increase in CPI leads to a 22% increase in FDI inflows.

(2)

2

Hierbij verklaar ik, Adriaan van Hall, dat ik deze scriptie zelf geschreven heb en dat ik de volledige verantwoordelijkheid op me neem voor de inhoud ervan.

Ik bevestig dat de tekst en het werk dat in deze scriptie gepresenteerd wordt origineel is en dat ik geen gebruik heb gemaakt van andere bronnen dan die welke in de tekst en in de referenties worden genoemd.

De Faculteit Economie en Bedrijfskunde is alleen verantwoordelijk voor de begeleiding tot het inleveren van de scriptie, niet voor de inhoud.

(3)

3

1. Introduction

As Brazil entered the twenty first century, it had a stable economy and a tested democratic regime (Langevin & Stackhouse, 2015). The government of President Fernando Henrique Cardoso had tamed inflation in the 1990’s and after the global crisis in 2008, Brazil was one of the first countries to recover with 7,5% economic growth in 2010 (Renwick, 2015). Brazil was considered one of the biggest emerging economies in the world, and because of its desirable characteristics, such as an enormous potential consumer market and abundance of natural resources (Vijayakumar, Sridharan & Rao, 2010), it was an appealing investment opportunity for foreign investors (Khan, Barua & Bhuiya, 2015).

However, Brazil’s economy, which had already been sluggish for the past few years due to a decline of demand for Brazilian exports in Europe and Asia (Khan, Barua & Bhuiya, 2015), started to take another downturn and is now dealing with its worst recession in recent history (Kiernan & Jelmayer, 2016) as GDP growth rate has been negative since the second quarter of 2014, and inflation and unemployment rates are rising rapidly (Jain, 2015).

Incidentally, this economic downturn coincided with the uncovering of one of the biggest corruption scandals Brazil has ever seen in March 2014, when Brazilian state-controlled petroleum company Petrobras was put under investigation in a large operation which has already seen many government officials and company executives been charged with federal corruption crimes. Furthermore, current President Dilma Rousseff is facing impeachment due to illegal manipulation of fiscal accounts (Arruda de Almeida & Zagaris, 2015). The uncovering of the corruption scandal has shown that the level of corruption seems to have been underestimated, as indicated by a 50-point drop in

Transparency International’s 2015 Corruption Perceptions Index (CPI) compared to 2014 due to the scandal, corresponding with the biggest drop for a country in the index during the year

(Transparency International, n.d.-b).

Brazil is not the only country with corruption problems in South America. Even though the third wave of democratization in South America brought widespread structural reforms, corruption seems to be a term that repeatedly comes up (Biddle, 2006). Galán Pachón (2009) shows that from 2001 until 2009, the average level of corruption has gradually decreased as indicated by the CPI score of South American countries. However, in 2009 no country scored higher than 7.0 out of 10.0, and some countries dropped below 2.0. A quick look at table 1, in which the CPI scores of the autonomous economies in the South America are shown suggests that corruption is still a widespread problem in South America (Galán Pachón, 2009). The effects of corruption on the attractiveness of a country’s economy, while intuitively is negative, is still widely debated.

As Foreign Direct Investment (FDI) is a conventional method to assess the attractiveness of a country’s economy for potential investors, and previous research has found that there is a potential

Table 1: CPI scores in South America

Year Highest Lowest Median Average

2003 7,40 1,60 3,10 3,52

2006 7,30 2,30 2,95 3,54

2009 6,70 1,90 3,30 3,55

2012 7,20 1,90 3,55 3,93

2015 7,40 1,70 3,50 3,85

(5)

5

relation between a country’s CPI and FDI (Tokunova, 2014; Habib & Zurawicki, 2002; Al Sadig, 2009), the increase in perceived corruption in a certain country should yield changes in the attractiveness of South American economies for foreign investors, and therefore in potential inflows of FDI. Whether the increase of corruption perception has a positive or negative effect on FDI is debated in numerous studies. Li (2005) for instance found that FDI inflows is higher in what he calls ‘relation-based’ countries, in which there are relatively higher levels of corruption, as opposed to ‘rule-‘relation-based’ countries. Habib & Zurawicki (2002) on the other hand concluded that foreign investors generally avoid corruption because it is considered wrong and harmful to the economy as it creates operational inefficiencies. Habib and Zurawicki’s view corresponds with the general theory on corruption, that countries with lower levels of corruption attract more FDI (Quazi, 2014).

While the general theory of corruption predicts that corruption is negatively related to foreign direct investments (Quazi, 2014), this relation is still debated and there has been given ample evidence that certain countries do not react the same to differences in the level of corruption. This thesis aims to isolate the effects of a change in corruption perception on FDI for South American countries, to predict the potential effect of changes in CPI on FDI, for example for Brazil after the most recent corruption scandal.

Realizing that certain countries react differently to corruption, and motivated by the persistent issues of corruption in South America, this thesis aims to examine the effect of corruption on FDI inflows by running a panel data regression on data from the ten largest South American economies: Brazil, Argentina, Colombia, Venezuela, Peru, Chile, Ecuador, Bolivia, Uruguay and Paraguay. A panel data regression is chosen to control for unobserved country-specific effects that may vary across countries and may be correlated with corruption (Al Sadig, 2009). The research question is

formulated as follows: What is the effect of corruption on net inflows of foreign direct investment in South America?

To answer the research question, a similar panel data analysis will be employed as Al Sadig (2009) uses, but whereas he focuses on the entire world, this thesis will use only data for South American countries. To control for different factors, four different regressions will be run, each adding new control variables. The first model will employ the full sample of countries and economic variables, omitting the effect of openness of countries and level of democracy. The second model adds index values for democracy and openness of the countries. The third model adds an interaction variable to assess differences in the effect of corruption for more corrupt countries and the last model adds time-specific variables to control for unobserved effects that vary across time periods

The second part of this thesis will consist of a literature review on the definition of corruption and the effect it has on foreign investments. Furthermore, an analysis will be included discussing the other determinants of FDI and why they should or should not be included as control variables. In part three, an expectation and hypothesis will be formed for the results of this research based on empirical and theoretical research. Part four will provide an explanation on the data and

methodology that is used in this research. Part five will provide and analyze the results of the panel data regression and part six offers a conclusion.

2. Literature review

In this part, we will emphasize the definition which will be used for corruption. Furthermore, theoretical and empirical research will be analyzed to find the relation between corruption and

(6)

6

foreign direct investment (FDI). To conclude, an analysis is offered on other potential determinants of FDI.

2.1 Definition of corruption

Throughout the years, corruption has been given many definitions. Hindess (2004), however claims that there is no authoritative consensus on the definition of corruption as it is inevitably linked to subjective perceptions of what is in the interest of the public. Von Alemann (2004) agrees with this and concludes that political scientists have been unable to find a single definition that is acceptable to the public. To illustrate the difficulties of defining corruption, Von Alemann (2004) analyzes five different definitions of corruption in his research, but disqualifies all of them as he argues none of them include every aspect of corruption.

Despite the difficulties of defining corruption while incorporating all its facets, there have been many efforts to define corruption in a generalized way. The Oxford Learner’s dictionary uses two different relevant definitions of corruption; 1: dishonest or illegal behavior, especially of people in authority, and 2: the act or effect of making somebody change from moral to immoral standards of behavior (The Oxford Learner’s dictionary, , while a more formal definition is given by The Foreign Corrupt Practices Act (FCPA): ‘illegal payments to foreign official representatives in order to obtain permission to create or retain business’ (Tokunova, 2014). Another definition, and the one that is followed in this thesis, is given by Langbein & Sanabria (2013), who state that corruption is generally defined as ‘the use of public office for private gains’. The same definition is used by Transparency International, and as this thesis uses their Corruption Perceptions Index (CPI) as a measure of corruption, it is justifiable to use their definition so that the data matches the definition used (Transparency International, n.d.-c).

TI adds that corruption can be classified as grand, petty and political, and that these classifications depend on the volume of money lost and the sector where it occurs (Transparency International, n.d.-c). Acts of corruption committed at a high level of government that distort policies or the central functioning of the state are classified as grand, whereas petty corruption refers to everyday abuse of power by low- and mid-level public officials during interactions with citizens when trying to access basic goods or services. Political corruption consists of manipulating policies, institutions and rules of procedure in the allocation of resources and financing by political decision makers, who abuse their position (Transparency International, n.d.-c)

2.2 Effect of corruption on FDI

As described in section 2.1, there are different definitions and dimensions of corruption, and all seem to point toward unlawful and socially unacceptable behavior. In this section the effect of corruption perception is analyzed based on empirical and theoretical research.

The prevailing theory on foreign investments is that countries with lower levels of corruption tend to attract more FDI (Li, 2005), and that corruption breeds inefficiencies and distortions, which harm the economy (Quazi, 2014). This is in line with the theory of the grabbing hand, which emphasizes the negative effects corruption can have by raising transaction costs for foreign investors (Quazi, 2014). Higher transaction costs in turn raise the overall cost of investment and reduce profitability.

Investors will therefore, when having the alternative of two investment opportunities that offer the same return on investment, opt for the investment in the less corrupt country. Wei (1997), and Mathur and Singh (2013) came to the same conclusion as they found that corruption perception does play a big role in investors’ decision of where to invest, and that in general, countries that rank poorly on the Corruption Perceptions Index receive low FDI flows compared to those ranking above

(7)

7

them. Wei (1997) argues that foreign investors are not attracted to countries with high levels of corruption because of the uncertainty of business conditions. Langbein and Sanabria (2013) add that corruption is widely regarded as a wicked social problem, as it robs countries of economic growth in favor of costly transfers of money, often from the poor to the relatively more powerful gatekeepers or providers of public services. Mauro (1995) agrees with this, suggesting that corruption worsens poverty and impacts income distribution.

For now, we assume that in general, corruption has a negative effect on FDI. There is however ample evidence that this does not hold true for all countries, and that there could be many reasons why FDI and corruption could be positively related for certain sets of countries. This follows the theory of corruption as helping hand, rather than grabbing hand, ‘greasing’ the economy and leading to a more efficient economy. Some arguments in favor of corruption state that corruption can improve efficiency and result in a Pareto optimal outcome, and that limited corruption can boost innovation and weaken monopoly, which promotes economic growth (Quazi, 2014).

Bardhan (1997) offers two arguments in favor of the helping hand theory. Firstly, economists have shown that when a country has pre-existing governance distortions, additional distortions and black market activity can actually improve welfare, even if there are extra costs to these activities

(Bardhan, 1997). This argument claims that when a government has failed to provide a fair economic environment, the course made possible by corruption can function as grease, and improve economic efficiency. The second argument he provides is that corruption can be seen as “speed money”. The act of corruption can reduce delay in highly bureaucratic decision-making and can form a Nash equilibrium in a non-cooperative game in which waiting costs are minimized, reducing inefficiency in public administrations (Bardhan, 1997). Bardhan (1997) does however argue that the presence of corruption could also backfire and potentially slow down, rather than speed up economic decision, as to attract more bribes.

Furthermore, Tokunova (2014) argues that corruption can lead to less risky investments, as investors can gain access to inside information through bribes, which can function as a hedge. This argument is also made by Li (2005), who found evidence that relatively more corrupt countries, who he

classifies as ‘relation-based’ countries, do in fact attract higher levels of FDI compared to ‘rule-based’ countries, in which there is less corruption. He considers that the prevailing theory - which views a healthy government environment as conductive to economic activities such as investment, whereas in the absence of such governance, investments cannot be protected - as incomplete since there is no complete vacuum of protection mechanisms between these two governance environments. Li (2005) concluded that in a country with an unhealthy environment, while firms cannot rely on laws and public trust, great emphasis is put on personal loyalty, and as ‘relation-based’ countries tend to favor the most powerful groups and individuals, personal relationships with these groups or

individuals can drastically reduce risk.

Alemu (2012) and Mauro (1995) disagree with the argument that corruption can increase FDI inflows. While Alemu (2012) does acknowledge that relatively corrupt countries can have high levels of FDI, he argues that these countries can further improve their inflows of FDI if they managed to reduce the present level of corruption. Mauro (1995) adds that while cumbersome and dishonest bureaucracies can indeed slow down economic growth, he presents evidence that indicate

corruption lowers private investment and decreases economic growth, even in subsamples of highly bureaucratic countries.

An interesting argument from Alemu (2012), is that once corruption becomes entrenched, its negative effects multiply and the public will begin to regard it as the norm. This argument follows

(8)

8

the theory of corruption equilibrium, which implies that corruption is a repeated game and state-dependent, so the play at a certain time, is dependent on the previous play, implying that the amount of corruption in a country, when no exogenous shock is presented, is likely to remain stable (Langbein & Sanabria, 2013). From this theory, it follows that a change in a country’s CPI score is most likely caused by an exogenous event. The fact that the uncovering of the Petrobras scandal in Brazil caused a 50-point decrease in CPI attests to this.

According to Quazi (2014), empirical evidence on the effect of corruption on FDI is not conclusive and other economic variables, particularly domestic institutions seem to be more significant

determinants of FDI than corruption. To facilitate comprehension of the empirical evidence it should be noted that CPI scores and levels of corruption are inversely related; a high score on CPI

corresponds to low levels of corruption. Therefore, a positive relation between CPI and FDI inflows indicates that corruption is negatively correlated with FDI. While evidence on the effect of

corruption is inconclusive, several studies found a positive relation between CPI and FDI. Udenze (2014) found a statistically significant positive relation between CPI and FDI, and concluded that a one-point increase on the Corruption Perceptions Index leads to a 0,88% decrease in net FDI inflows as a percentage of GDP for low- and middle-income countries. Tokunova (2014) found inconclusive results as developed countries showed a significant positive relation between CPI and FDI, while the relation between these variables for the BRIC countries yielded insignificant results. Li (2005) found a negative relation between CPI and FDI, concluding that ‘relation-based’ countries, which are

relatively more corrupt, receive more FDI inflows as a percentage of GDP. He however focuses on FDI for a single time period for each country. Therefore, high levels of FDI in a country do not necessarily imply that these levels could not be increased further by reducing corruption. Al Sadig (2009) also found inconclusive results in his panel data regression, namely that the coefficient of corruption levels was negative but insignificant when testing for all countries, but highly significant when excluding the high-income OECD countries from the sample. He however found that when controlling for the quality of institutions in the host country, the negative effect of corruption disappears. He concludes however that this does not prove that the level of corruption is unimportant but emphasizes the importance of the quality of institutions. Finally, Alemu’s (2012) research found that for 16 Asian countries, a 1% increase in corruption level triggers an approximate 9,1% decrease in FDI inflows.

2.3 Other determinants of FDI

It goes without saying that there are other determinants of FDI besides corruption. Yasmin, Hussain & Chaudhary (2003) conclude in their analysis of factors affecting FDI inflows in developing

countries, that relevant determinants for FDI are among others; urbanization, GDP per capita, unemployment, inflation, current account, external debt, domestic investment, trade openness and wages.

Furthermore, Mathur and Singh (2013) found that more democratic countries receive more FDI than less democratic countries, and Tokunova (2014) adds that economic growth has a positive effect on FDI inflows and provides arguments for including unemployment and population growth as

determinants of FDI. Udenze (2014) provides theoretical arguments for openness of a country, Inflation and GDP per capita as determinants of FDI, and finds in his empirical research that the openness, and GDP growth variables are significant determinants of FDI, and to conclude, Al Sadig (2009) ran a panel data regression which found significant values for some previously mentioned determinants of FDI and some that have not been discussed. Among others, he found significant effects of GDP per capita, openness of the host country, inflation, secondary school enrollment,

(9)

9

population growth, urban population growth, and previous levels of FDI per GDP. Furthermore, he mentions that the level of democracy and rule of law are significant determinants of FDI.

Some of the aforementioned determinants of FDI will be used as control variables in this research. Excluded variables will range from variables that have no empirical or theoretical support as determinant for FDI, or variables which have insufficient data to be included. Section four will provide an overview and explanation of the included and potentially relevant excluded determinants of FDI.

3. Hypothesis

After reviewing the literature, a hypothesis regarding the effect of corruption on FDI is formulated in this section based on the research question: What is the effect of corruption on net inflows of foreign direct investment in South America?

It seems that the general view on corruption - that it has a negative effect on the economy and foreign investments - holds true. This however does not seem to be the case for every country, as Tokunova (2014) found insignificant results regarding the effect of the level of corruption, based on the Corruption Perceptions Index (CPI), on foreign direct investments (FDI) for the BRIC countries, and Udenze (2014) found insignificant results for sub-Saharan countries. Furthermore, Al Sadig (2009) only found a significant negative effect of corruption on FDI when he excluded the high-income OECD countries, and these results turned insignificant again when controlling for domestic institutions of the host country. This matches the findings of Quazi (2014), who found that FDI is more likely linked to other determinants than corruption.

To answer the research question, a panel data regression will be run for the ten biggest economies in South America using data for the past twenty years, starting with the year the CPI measure was introduced, and analyzing four different models.

The first model we analyze will include all the economic variables but will exclude data for democracy and country openness as these are more likely correlated with the Corruption

Perceptions Index (CPI). The expectation is that CPI and foreign direct investment (FDI) are positively correlated, meaning that a higher score on CPI leads to higher levels of FDI.

In the second model, values for a democracy index are added as well as a measure of country openness. According to Al Sadig (2009) and Quazi (2014), this should yield less significant results for CPI as a determinant of FDI, as these variables are correlated with CPI, but have higher predictive power than CPI.

The third model adds a variable which measures whether the effect of CPI is different for more corrupt countries compared to the less countries. The expectation is that these countries do differ. Empirical research so far has mostly found that more corrupt countries show mostly insignificant correlation between CPI and FDI, while results for developed countries are significant (Tokunova, 2014). It does seem reasonable to assume that exogenous corruption effects such as the Petrobras scandal in Brazil, discussed in section one, would have a less significant impact in countries where investors know that corruption is widespread.

The final model adds time-specific variables which accounts for unobserved effects in certain years. The expectation is that, when controlling for time-specific effects, a larger portion of the variance in

(10)

10

the panel data model will be explained, so that the effect of other determinants of FDI can be more accurately predicted.

Ultimately, it is expected the level of CPI is positively correlated with net inflows of FDI, proving both, that a higher score on CPI in a certain country will attract more FDI, and showing that countries with higher scores on CPI attract larger amounts of FDI stock than countries with low CPI scores.

4. Methodology and Data Selection

To isolate the effects of corruption on foreign direct investment (FDI), a panel data regression will be used. This approach has been chosen, because as opposed to cross-sectional analysis, a panel data analysis has certain advantages, particularly useful for the data analyzed in this thesis. One of these advantages, according to Hsiao (2007), is that there is a more accurate inference of model

parameters: “Panel data usually contains more degrees of freedom and more sample variability than cross-sectional data” (Hsiao, 2007, p. 3). Furthermore, according to Wooldridge (2001), a panel data analysis allows us to look at dynamic relationships and most importantly allows us to control for unobserved cross sectional heterogeneity. As the data used in this analysis is rather limited, we have between fifteen and twenty observations per country, the first argument in favor of panel data regression is relevant. Secondly, by controlling for cross section heterogeneity, we will be able to capture time invariant unobserved effects that vary across countries. According to Al Sadig (2009), when failing to hold these variables constant, the estimated effects may be biased in either direction.

Unless otherwise stated, data for the economic variables are retrieved from the World Bank database. For the years since the introduction of Transparency International’s Corruption

Perceptions Index (CPI), most of the data was available for the countries included in the database. However, as limited data was found for Guyana and Suriname, these countries have been excluded from the database. Furthermore, as French Guiana and the Falkland Islands are not fully

autonomous countries, these have also been excluded from the dataset. In the end, the ten largest economies in South America are used in the sample: Brazil, Argentina, Colombia, Venezuela, Peru, Chile, Ecuador, Bolivia, Uruguay and Paraguay.

4.1 Included variables

The dependent variable in this analysis is the logarithm of the level of net inflows of FDI measured in current US dollars, divided by the country’s population. Previous research has mostly measured FDI as a percentage of GDP to control for country size, but intuitively this leads to problems. After all, it is unlikely that FDI and GDP share all of their determinants, which makes it a necessity to control for exogenous changes in GDP, as changes in GDP would affect levels of FDI. Population growth is a much more stable variable, making sure that changes in the dependent variable are only caused by changes in FDI. One limitation of using the logarithm of the net inflows of FDI per capita, is that negative values, which are possible, will automatically be dropped. Overall, the database which was used had 4 observations of negative FDI inflows which will be replaced by missing values.

The main explanatory variable in this study is Transparency International’s Corruption Perceptions Index (CPI). This is a measure, based on informed views of analysts, businesspeople and experts in countries around the world, which estimates the level of corruption in a certain country

(Transparency International, n.d.-b). This data is retrieved from the Transparency International website and is used as a proxy for corruption as there is no true measure of corruption. Von

(11)

11

Alemann (2004) states that while it is practically impossible to find an objective reliable measure for corruption, CPI provides an astonishingly reliable estimation of corruption. Furthermore, corruption perception is what investor’s base their decisions on, so it is likely that if there is a relation between corruption and FDI, this relation will be observed by using CPI. CPI scores rank between 0-10 with a higher score representing lower levels of corruption. This means that levels of corruption and CPI scores are inversely correlated which should be kept in mind when analyzing the data.

From section 2.3, it can be concluded that there are many other determinants of FDI. These should be added as control variables, but to find and include all of them would not be feasible as firstly, different studies find different determinants of FDI, and secondly, secondary determinants of FDI seem to only affect certain countries. For this reason, only the determinants that are supported by literature, have significant empirical results as determinants of FDI, and for those which data is available, will be included in the analysis. Following the methodology of Al Sadig (2009), all the explanatory variables have been lagged one year in order to avoid simultaneity and show that decisions to invest abroad take time.

GDP per capita (GDPPC) and GDP growth (GDPG) are included as control variables, as according to Tokunova (2014) they are important macroeconomic variables which are used to indicate economic growth and standards of living in a country. GDPPC is included in logarithmic form. Al Sadig (2009) included both these variables in his analysis to control for the host country’s market size and market potential. Intuitively these are important variables, as investors are more likely to invest in a country with strong economic growth as there should be more investment opportunities. Furthermore, in a country where standards of living are higher, more capital should be available for investment. Empirical results on GDPPC are not conclusive. Udenze (2014) found an insignificant negative effect of GDPPC on FDI inflows, but Al Sadig (2009) found a 1% significant positive effect for all the models he analyzed. GDP growth was found to have a positive effect, which was 1% significant, in the analysis of both Udenze (2014) and Al Sadig (2009).

The level of inflation (INF) is also used as a control variable as according to Demirhan and Masca (2008), inflation has a significant negative effect on levels of FDI, Yasmin et. al (2003) came to the same conclusion and found that inflation is a significant determinant of GDP, at least for lower- and upper middle income countries. The economic channel for the negative relation between inflation and FDI inflows seems reasonable, as an increase in inflation leads to a real drop in return on investment, making investment opportunities in high inflation countries less attractive. In this analysis, the level of inflation also controls for economic stability.

Two other control variables included in the analyses are population growth (POPG) and urbanization (URBG), measured by urban population growth. Yasmin et. al (2003) concluded in their analysis that urbanization and population growth are relevant determinants of FDI, and Tokunova (2014) found significant results for population growth as a positive determinant of FDI in parts of her analysis. Surprisingly, Al Sadig (2009) found a negative effect of population growth on FDI in all of his models and found both positive and negative effects of urbanization in his models. This necessitates further research.

Another variable, which is added to control for the fact that investors might be attracted to a host country that already has large amounts of existing FDI stock, is existing FDI stock as a fraction of GDP (FDIGDP). The reason this should have a positive effect on net FDI inflows, is that a high level of existing FDI stock may be viewed as a signal for a good investment environment (Al-Sadiq, 2009).

(12)

12

The degree of a country’s openness, measured by the sum of imports and exports divided by GDP, is also added to control for a country’s foreign competitiveness and government intervention. It should be easier to invest in countries that have high levels of imports and exports. The level of openness is therefore expected to have a positive effect on net FDI inflows. As anticipated, Udenze (2014) and Al Sadig (2009) found a significant positive effect of openness on the level of FDI inflows.

The last variable to be added to the dataset, is an index for the level of democracy. Freedom House’s democracy index, which started providing indices in 1998 and scores countries between 1 (highest) and 7 (lowest), will be used to measure democracy levels. Prior to 1998, these values will be given as missing. Mathur and Singh (2013) found that more democratic countries receive higher levels of FDI, as democratic countries are more likely to move to free trade as opposed to implementing

protectionist measures, potentially blocking foreign investments (Milner & Kubota, 2005). The level of democracy, which is inversely related to the democracy index, is expected to have a positive effect on FDI inflows, meaning that the coefficient for democracy is expected to be negative. Some variables have not been included in the panel data analysis even though there is literature which suggests it could be a determinant of FDI. The level of external debt and current account are such variables. The reason for their exclusion is that there are no empirical results which indicate that these are significant determinants of FDI, and since data is quite limited, it would not be wise to fill the panel data with variables that are unlikely to show significant results. Wages and

unemployment are also omitted from the model as the effects of these variables are already incorporated in the model through other variables. The inclusion of GDP per capita should be a better measure than wages as it incorporates standards of living. Unemployment can be classified as a variable controlling for economic stability, making it an unnecessary variable as inflation already controls for this effect. Besides already having variables included in the model which control for similar effects, there is also little evidence to suggest either wages or unemployment are significant determinants of FDI. Some other variables, such as school enrollment and rule of law are also excluded from the database. This however, is due to data unavailability, as these variables have proven to be significant determinants of FDI inflows (Al Sadig, 2009). By running a panel data analysis, the expectation is that the model will control for these country specific omitted variables.

4.2 The model

The model used in the first regression excludes the democracy index and level of country openness, and is structured as follows:

log (FDI/POP)i,t-1 = β0 + βCPI*CPIi,t-1 + βGDPPC*log(GDPPC)i,t-1 + βGDPG*GDPGi,t-1

+ βFDIGDP*FDIGDPi,t-1 + βPOPG*POPGi,t-1+ βURBG*URBGi,t-1 + βINF*INFi,t-1 + ηi + εi,t

In this model, i and t are the country and time subscript respectively. The β’s are coefficients which will be estimated by the panel data regression. η is the variable added to account for country specific effects, and ε is the error term. The second model adds values for country openness and the

democracy index. These are denoted by their respective coefficients:

+ βFDIGDP*FDIGDPi,t-1 + βPOPG*POPGi,t-1+ βURBG*URBGi,t-1 + βINF*INFi,t-1

(13)

13

The third model adds the interaction variable: Corrupt*CPI (CPIC), to measure whether the effect of corruption is different in more corrupt countries. The variable Corrupt is a dummy variable which has value 1 for countries with an average CPI score, over the last 20 years, between 0 and 5:

+ βOPEN*OPENi,t-1 + βDEM*DEMi,t-1+ βCPIC*CPICi,t-1 + ηi + εi,t

The final model adds time-specific dummy variables, denoted by γ, which account for unobserved differences in overall levels of FDI per year. The expectation is that the significant determinants of FDI will provide a more accurate prediction of its effect on FDI inflows. The structure of the model is identical to that of the third model.

+ βOPEN*OPENi,t-1 + βDEM*DEMi,t-1+ βCPIC*CPICi,t-1 + ηi + γt + εi,t

5. Empirical data and results

This section presents the results of the model estimations and an analysis on the results. The subject of interest in this panel data regression was the effect of corruption, measured by the Corruption Perceptions Index (CPI), on net inflows of foreign direct investment (FDI). The results of the panel data analysis are found on the next page.

As mentioned, the analysis consists of four different models, the first one consisting of most of the economic variables, and including country-specific variables. This model has an adjusted R-squared value of 0,714, meaning that over 71% of the variance in net inflows of FDI can be explained by the model.

The second model added a democracy rating and level of openness, to confirm whether these are more important determinants of FDI, as suggested by Quazi (2014), and Al Sadig (2009). The results show that the significance of the coefficient for CPI has indeed dropped and while the coefficients for the democracy rating and level of country openness are not significant, the level of adjusted R-squared increased to 0,734, indicating that the model has improved.

Model three adds a variable to find whether more corrupt countries react differently to changes in corruption perception. As discussed in the literature review in section two, the expectation is that corruption has a smaller effect on FDI in these countries. This expectation is intuitive as when corruption is already widespread in a country, investors should be less surprised by an increase in corruption perception. The results show that, as expected, the effect of CPI on FDI is negated for more corrupt countries, and actually becomes negative. This means that for more corrupt countries, a decrease in levels of corruption leads to a decrease in net inflows of FDI. Furthermore, the

coefficient for CPI is now significant at the 10% level, and the adjusted R-squared value increased to 0,738.

The final model adds specific variables for each observed year. The expectation is that time-specific events such as the recent global crisis will explain some of the variance making the model more accurate, leading to a higher significance level of the relevant included variables. The adjusted

(14)

14

R-squared value for the final model has increased to 0,756, meaning that over 75% of the variance in inflows of FDI can now be explained by the model. Furthermore, as expected, the significance of some coefficients has changed as the time-specific variables now explain some of the variance in the model.

5.1 Empirical results: The effect of CPI on FDI

The objective of this thesis was to research the effect of corruption on net inflows of foreign direct investment for countries South America. To that end, data was analyzed using four different models to control for certain factors. Considering the fact that countries in South America are unlikely to be

-0,836 -1,046 -0,457 0,925 (0,618) (0,738) (0,805) (2,077) 0,101** 0,052 0,175*** 0,220* (0,040) (0,056) (0,089) (0,096) 0,710* 0,833* 0,799* 0,361 (0,132) (0,157) (0,157) (0,530) 0,009 0,011 0,008 0,008 (0,006) (0,007) (0,007) (0,009) 0,060* 0,045* 0,041* 0,031*** (0,011) (0,014) (0,014) (0,017) -0,808* -0,965** -1,135* -1,178* (0,298) (0,373) (0,382) (0,385) 0,458** 0,608** 0,668* 0,605** (0,180) (0,245) (0,245) (0,280) 0,003 0,005 0,003 0,004 (0,002) (0,004) (0,004) (0,005) 0,001 0,001 0,001 (0,003) (0,003) (0,005) -0,036 -0,033 -0,101 (0,064) (0,064) (0,074) -0,206*** -0,014 (0,117) (0,013)

Country-specific effects yes yes yes yes

Time-specific effects no no no yes

R-squared 0,740 0,766 0,772 0,816

Adj. R-squared 0,714 0,734 0,738 0,756

F-value 28,782 23,480 22,763 13,671

p-value 0,000 0,000 0,000 0,000

Table 2: Results of panel data regression

The effect of corruption on foreign direct investment

Note: All the independent variables are lagged one period. Standard errors are in parenthesis. *, ** and *** indicate

statistical significance at 1, 5 and 10 percent, respectively

Model 1 Model 2 Model 3 Model 4

GDPG Log(GDPPC) CPI Constant OPEN CPIC DEMO INF URBG POPG FDIGDP

(15)

15

identical regarding FDI inflows, and the absence of certain relevant variables due to data

unavailability (see section 4.1), a panel data regression was used as it can control for unobserved country-specific and time-specific variables.

In all the tested models, the coefficient for CPI was positive, as was expected, and although it was insignificant in model two, its coefficient was at the 5, 10 and 1 percent significance level

respectively for models one, three and four. These findings suggest that the level of CPI in a host country does have an effect on its net inflows of foreign direct investment, meaning that holding all other independent variables constant, countries with a higher level of CPI do in fact have higher levels of FDI inflows. From the first model, it can be concluded that a 100-point increase in the level of CPI leads to an increase of 10,1% in net inflows of FDI. After controlling for democracy and openness of the host country however, the coefficient drops to 0,052 and is no longer significant. This corresponds to Quazi’s (2014) findings that other variables besides corruption are more significant determinants of FDI.

Interestingly, when controlling for a different effect of CPI on FDI in more corrupt countries (average CPI score between 0 and 5), not only is the interaction variable CIPC negative and significant at the 10 percent level, after controlling for this effect, the CPI coefficient also becomes significant at the 10 percent level. From the results, it follows that for this model, countries that are relatively more corrupt, actually show a negative relation between CPI and FDI, and FDI in these countries drops by 3,1% for every 100-point increase in CPI. Countries that are relatively less corrupt show a larger effect of changes in CPI than in models one and two as they receive 17,5% more FDI when CPI increases by 100 points. This effect gets stronger when controlling for time-specific effects in the final model. Model 4 shows that when controlling for time-specific effects, a 100-point increase in CPI leads to a 22% increase in FDI. Furthermore, the effect of CPI on FDI for more corrupt countries, while still negative, has been reduced to near 0 and is no longer significant. This suggests that the negative relation between CPI and FDI for more corrupt countries which was found in model 3 was most likely due to time-specific events. Considering the coefficient of CPI in the fourth model is significant at the 1 percent level, we can conclude that in general, CPI is positively related to net inflows of FDI, and when there are control variables for time-specific events, the relation between CPI and FDI is not significantly different in more corrupt countries.

5.2 Empirical results: The effect of the other explanatory variables on FDI

Most of the control variables have the expected effect on foreign investments. GDP per capita (GDPPC) has a positive sign and is significant at the 1 percent level for the first three models. In the first model, a 1% increase in GDPPC leads to a 0,71% increase in FDI net inflows. After controlling for openness of the host country and democracy level the effect of GDPPC increases to a 0,83% increase per 1% increase in GDPPC and in the third model, after controlling for the level of corruption in more corrupt countries, a similar increase in GDPPC leads to an 8% increase in FDI inflows. In the final model, when controlling for time-specific events, GDPPC becomes insignificant while still having a positive coefficient. This follows the expectation that richer countries tend to attract more FDI inflows as was hypothesized in section 4.

The coefficient for GDP growth (GDPG), which functions as a measure of market potential, has a positive sign and comparable values, between 0,8% and 1,1% increase in FDI per 1% increase in GDPG, for all models. This is consistent with the expectation that investors tend to invest in countries with higher levels of economic growth. However, the coefficient while positive, is insignificant for all analyzed models.

(16)

16

It was expected that prior levels of FDI stock (FDIGDP) as a fraction of GDP would have a positive effect on new foreign direct investment. The results from the panel data regression are consistent with this expectation as the coefficient for FDIGDP is positive and significant at the 1 percent level for the first three models and significant at the 10 percent level for the last model.

The effect of population growth and urbanization on FDI was expected to be positive for both variables. Corresponding to the expectations, the coefficient for urbanization is indeed positive and significant at the 5 percent level for models one, two and four, and significant at the 1 percent level for model three, which suggests that countries with higher levels of urbanization receive

approximately 45,8%-66,8% more FDI inflows per capita for every 1% increase in population growth in urban areas. The coefficient for population growth in general is negative, and while this was not expected, it corresponds to the results of Al Sadig (2009) who ran a similar panel data regression for the entire world. These results, which are significant at the 1 percent level for models one, three and four, and at the 5 percent level for model two, suggest that population growth is actually a negative determinant of FDI inflows.

Inflation was expected to be a negative determinant of FDI inflows, as real rate of return diminishes when inflation increases, as well as the fact that high levels of inflation suggest economic instability, but was found to be positive and insignificant. Udenze (2014) found similar insignificant results, which suggests that this area needs further research. Al Sadig (2009) did find a negative effect of inflation on FDI at the 1 percent level of significance.

Both the democracy index as well as the level of openness show the expected sign as it was hypothesized that a lower score on the democracy index, indicating a high level of democracy, corresponds to higher levels of FDI inflows. This can be explained by the fact that democratic

countries are more likely to move to free trade as opposed to implementing protectionist measures, potentially blocking foreign investments (Milner & Kubota, 2004). The level of openness was

expected to have a positive effect on FDI, and while the coefficient is positive, its value is weak at 0,001 and insignificant for all models in which it is included.

6. Conclusion

Motivated by the recent Petrobras corruption scandal in Brazil, this thesis looks at the effect corruption has on a country’s economy, and in particular on foreign direct investments (FDI). As the implications of corruption are still inconclusive, and the fact that corruption has been widespread in South American countries throughout the years, as indicated by their scores on the Corruption Perceptions Index (CPI), it is relevant to find the effect corruption has on net inflows of FDI to assess what corruption means for the attractiveness of a country’s economy leading to the following research question: What is the effect of corruption on net inflows of foreign direct investment in South America?

This thesis has sought to find an answer to this question by analyzing data for the ten largest economies in South America for the last twenty years, employing four different panel data sets and using a wide set of control variables which were all lagged one year in order to avoid simultaneity and show that decisions to invest abroad take time. The expectation was that the level of corruption, inversely measured by the CPI index, is negatively correlated with net FDI inflows, that is, there is a positive relation between CPI and FDI, but that there are other factors, particularly domestic institutions, which are correlated with CPI that are better determinants of FDI inflows.

(17)

17

The results show that the coefficient of CPI was, as expected, positive for the first model at a

significance level of 5 percent, but turned insignificant went controlling for level of country openness and level of democracy. However, in model three, when adding a variable which measures the difference of the effect of CPI on FDI inflows for more corrupt countries (average CPI score between 0 and 5), its coefficient turns significant at the 10 percent level. Finally, when controlling for time-specific effects, the coefficient of CPI becomes significant at the 1 percent level and suggests that for every 100-point increase in CPI, FDI inflows increase by 22%.

The results for the control variables were mostly as expected. GDP per capita proved to have a positive effect and was significant at the 1 percent level for each of the first three models. After controlling for time specific events, the effect of GDP per capita remains positive but becomes insignificant. GDP growth was also positively correlated with FDI inflows, but its coefficient was insignificant for all models. Previous FDI stock measured as a fraction of GDP also proved to be positively correlated with FDI inflows and was significant at the 1 percent level for the first three models, and at the 10 percent level for the last model.

The level of urbanization, measured by the population growth in urban areas, showed a significant positive effect for each of the four models, suggesting that countries receive approximately 45,8%-66,8% more FDI inflows per capita for every 1% increase in urban population growth. Unexpectedly, however, population growth in general showed a significant negative effect on FDI inflows. While this does not match the theory, Al Sadig (2009) found similar results. This suggests that population growth is actually a negative determinant of FDI inflows. Inflation was expected to have a negative effect on FDI inflows, but while being insignificant, turned out to have a positive coefficient. Furthermore, the democracy index and level of openness, while having the expected effect on FDI, both proved to be insignificant determinants of FDI inflows.

While this thesis shows that CPI has a significant positive effect on net inflows of FDI per capita, there are some limitations to this research. First of all, data availability was an issue, as some potentially important variables such as rule of law and school enrollment could not be added to the model. Using a panel data regression however should solve this, as it controls for unobserved country specific effects. Furthermore, as Transparency International only started their Corruption Perceptions Index in 1995, there were at most twenty possible observations for each country. This leads to a relatively small dataset. Another limitation is that CPI might not be an ideal measure of corruption as it is based on surveys and questionnaires and could therefore be subjective. Finally, a possible limitation is that of reversed causality. It can be argued that it is not CPI which is a

determinant of FDI inflows, but the other way around. Future research can potentially focus on the aspect of reversed causality, as well as provide measures for variables that were excluded due to lack of data availability in this research. Furthermore, future research could find a more reliable measure for corruption to replace CPI, and include a larger dataset to provide more reliable results.

(18)

18

7. Bibliography

Al Sadig, A. (2009, June 14). The Effects of Corruption on FDI Inflows. Cato Journal, 29(2), 267-294. Retrieved from http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2264898

Alemu, A. (2012). The Effect of Corruption on FDI Inflow: Empirical Evidence from Panel Data of East Asian Countries. Seoul Journal of Economics, 25, 387-412.

Arruda de Almeida, M., & Zagaris, B. (2015). Political Capture in the Petrobras Corruption Scandal: The Sad Tale of an Oil Giant. Association of Certified Financial Crime Specialists, 39(2), 87-99. Retrieved from Association of Certified Financial Crime Specialists:

http://www.acfcs.org/wp-content/uploads/2015/06/Arruda-Zagaris-2015.pdf Bardhan, P. (1997, September). Corruption and Development: A Review of Issues. Journal of

Economic Literature, 35(3), 1320-1346. Retrieved from

http://siteresources.worldbank.org/INTEASTASIAPACIFIC/Resources/226262-1253782457445/6449316-1261623644460/corruption-and-development.pdf

Biddle, L. (2006). Corruption in Latin America: Political, Economic, Structural and Institutional Causes. Master's Thesis, University of North Carolina, Department of Political Science, Chapel Hill. Retrieved from https://cdr.lib.unc.edu/indexablecontent/uuid:d28e0e16-d999-4759-9508-efae5c0c4c1c

Demirhan, E., & Masca, M. (2008). Determinants of Foreign Direct Investment Flows to Developing Countries: A Cross-Sectional Analysis. Prague Economic Papers, 17(4), 356-369.

Freedom House. (n.d.). Freedom in the World. Retrieved from Freedom House: https://freedomhouse.org/report/freedom-world/freedom-world-2016

Galán Pachón, J. (2009). Corruption and Oversight in Latin America. Retrieved from Global Organization of Parliamentarians Against Corruption:

http://gopacnetwork.org/Docs/SenatorGalansPresentation_CorruptionandOversightinLatina America_EN.pdf

Habib, M., & Zurawicki, L. (2002). Corruption and Foreign Direct Investment. Journal of International Business Studies, 33(2), 291-307.

Hindess, B. (2004). Corruption and Democracy in Australia. Retrieved June 6, 2016, from National Criminal Justice Reference Service:

http://unpan1.un.org/intradoc/groups/public/documents/APCITY/UNPAN019091.pdf Hsiao, C. (2007, May). Panel data analysis - advantages and challenges. TEST, 16(1), 1-22. Jain, S. (2015, September 16). Brazil's Inflation Continues to Soar While Growth Is Stagnant.

Retrieved June 5, 2016, from Market Realist: http://marketrealist.com/2015/09/brazils-inflation-continues-soar-growth-stagnant/

Khan, T., Barua, U., & Bhuiya, I. (2015, July). BRICS Economy: An Appealing Investment Opportunity on the International Stage. Journal for Worldwide Holistic Sustainable Development, 1(3), 15-42. Retrieved from http://wwhsdc.org/wp-content/uploads/2014/12/15-42-Khan-Barua-Bhuiya..pdf

(19)

19

Kiernan, P., & Jelmayer, R. (2016, June 1). Brazil's Recession Deepens. The Wall Street Journal, p. 1. Retrieved June 7, 2016, from http://www.wsj.com/articles/brazils-recession-deepens-1464786002

Langbein, L., & Sanabria, P. (2013). The Shape of Corruption: Colombia as a Case Study. The Journal of Development Studies, 49(11), 1500-1515. Retrieved from http://www-tandfonline-com.proxy.uba.uva.nl:2048/doi/pdf/10.1080/00220388.2013.800858

Langevin, M., & Stackhouse, T. (2015, August). Brazil Initiative. Retrieved from Elliot School of International Affairs:

https://brazil.elliott.gwu.edu/sites/brazil.elliott.gwu.edu/files/downloads/Brazil%20and%20

Development-%20Growth,%20Equity%20and%20Sustainability%20into%20the%2021st%20Century.pdf Li, S. (2005, July-August). Why a poor governance environment does not deter foreign direct

investment: The case of China and its implications for investment protection. Business Horizons, 48(4), 297-302. Retrieved from

http://www.sciencedirect.com.proxy.uba.uva.nl:2048/science/article/pii/S00076813050004 43

Mathur, A., & Singh, K. (2013). Foreign direct investment, corruption and democracy. Applied Economics, 45(8), 991-1002.

Mauro, P. (1995, August). Corruption and Growth. The Quarterly Journal of Economics, 110(3), 681-712. Retrieved from http://homepage.ntu.edu.tw/~kslin/macro2009/Mauro%201995.pdf Milner, H., & Kubota, K. (2005, January). Why the Move to Free Trade? Democracy and Trade Policy

in the Developing Countries. International Organization, 59(1), 107-143. Retrieved from https://www.princeton.edu/~hmilner/forthcoming%20papers/LDCdem_IO.pdf

Quazi, R. (2014). Corruption and Foreign Direct Investment in East Asia and South Asia: An

Econometric Study. International Journal of Economics and Financial Issues, 4(2), 231-242. Retrieved from http://www.econjournals.com/index.php/ijefi/article/view/695/pdf Renwick, D. (2015, April 3). After Brazil's Boom, Bust? Retrieved May 13, 2016, from Council on

Foreign Relations: http://www.cfr.org/brazil/after-brazils-boom-bust/p36393 The World Bank. (n.d.). World Bank Database. Retrieved from The World Bank:

http://data.worldbank.org/

Tokunova, S. (2014). A Comparative Study on the Effects of Corruption on FDI. Master Thesis, Erasmus Universiteit Rotterdam, School of Economics, Rotterdam. Retrieved from https://thesis.eur.nl/pub/15948/Tokunova.pdf

Transparency Interantional. (n.d.-a). Corruption Perceptions Index. Retrieved from Transparency International: http://www.transparency.org/research/cpi

Transparency International. (n.d.-b). Corruption Perceptions Index 2015. Retrieved from Transparency International: http://www.transparency.org/cpi2015

Transparency International. (n.d.-c). How do you define corruption? Retrieved from Transparency International: http://www.transparency.org/what-is-corruption/#define

(20)

20

Udenze, O. (2014). The Effect of Corruption on Foreign Direct Investments in Developing Countries. The Park Place Economist, 22(1), 85-95. Retrieved from

http://digitalcommons.iwu.edu/cgi/viewcontent.cgi?article=1416&context=parkplace Vijayakumar, N., Sridharan, P., & Rao, K. (2010). Determinants of FDI in BRICS Countries: A panel

analysis. Int. Journal of Business Science and Applied Management, 5(3), 1-13. Retrieved from http://www.business-and-management.org/download.php?file=2010/5_3--1--13-Vijayakumar,Sridharan,Rao.pdf

Von Alemann, U. (2004, August). The unknown depths of political theory: The case for a

multidimensional concept of corruption. Crime, Law & Social Change, 42(1), 24-34. Retrieved from http://link.springer.com/article/10.1023/B:CRIS.0000041035.21045.1d

Wei, S.-J. (1997, November). Why is corruption so much more taxing than tax? Arbitrariness Kills. Retrieved from National Bureau of Economic Research:

http://www.nber.org/papers/w6255.pdf

Wooldridge, J. (2001). Econometric Analysis of Cross Section and Panel Data. London: The MIT Press. Yasmin, B., Hussain, A., & Chaudhary, M. A. (2003). Analysis of Factors Affecting Foreign Direct

Investments in Developing Countries. Pakistan Economic and Social Review, 41(1&2), 59-75. Retrieved from

http://pu.edu.pk/images/journal/pesr/PDF-FILES/5%20YASMIN%20Analysis%20of%20Factors%20Affecting%20FDI%20in%20Developing -V41-2003.pdf

(21)

21

8. Appendix

8.1 Table 1 – country CPI scores

8.2 Regression results

8.2.1 Regression results model 1:

+ βFDIGDP*FDIGDPi,t-1 + βPOPG*POPGi,t-1+ βURBG*URBGi,t-1 + βINF*INFi,t-1 + ηi + εi,t

Model Summary Model R R Square Adjusted R Square Std. Error of the Estimate Durbin-Watson 1 .860 .740 .714 .2745473 2.178 ANOVA

Model Sum of Squares df Mean Square F Sig.

1 Regression 34.711 16 2.169 28.782 .000 Residual 12.211 162 .075 Total 46.922 178 Coefficientsa Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) -.836 .618 -1.353 .178 COL .077 .094 .047 .812 .418 ARG .056 .102 .034 .548 .584 PER .005 .100 .003 .050 .960 VEN .179 .163 .102 1.095 .275 CHI -.073 .164 -.045 -.444 .658 ECU -.072 .125 -.041 -.576 .565

Year Brazil Colombia Argentina Peru VenezuelaChile Ecuador Bolivia Paraguay Uruguay Suriname Guyana

2003 3,90 3,70 2,50 3,70 2,40 7,40 2,20 2,30 1,60 5,50

2006 3,30 3,90 2,90 3,30 2,30 7,30 2,30 2,70 2,60 6,40 3,00 2,50

2009 3,70 3,70 2,90 3,70 1,90 6,70 2,20 2,70 2,10 6,70 3,70 2,60

2012 4,30 3,60 3,50 3,80 1,90 7,20 3,20 3,40 2,50 7,20 3,70 2,80

2015 3,80 3,70 3,20 3,60 1,70 7,00 3,20 3,40 2,70 7,40 3,60 2,90

Year Highest Lowest Median Average

2003 7,40 1,60 3,10 3,52

2006 7,30 2,30 2,95 3,54

2009 6,70 1,90 3,30 3,55

2012 7,20 1,90 3,55 3,93

(22)

22 BOL -.032 .123 -.019 -.263 .793 PAR -.236 .123 -.124 -1.916 .057 URU -.418 .183 -.239 -2.283 .024 CPI .101 .040 .326 2.507 .013 LogGDPPC .710 .132 .428 5.355 .000 GDPG .009 .006 .068 1.617 .108 FDIGDP .060 .011 .298 5.421 .000 POPG -.808 .298 -.716 -2.712 .007 URBG .458 .180 .528 2.544 .012 INF .003 .002 .079 1.537 .126 Residuals Statistics

Minimum Maximum Mean Std. Deviation N Predicted Value .889717 3.185029 2.085248 .4415975 179

Residual -1.2344667 .5864968 .0000000 .2619176 179

Std. Predicted Value -2.707 2.490 .000 1.000 179

Std. Residual -4.496 2.136 .000 .954 179

8.2.2 Regression results model 2:

+ βOPEN*OPENi,t-1 + βDEM*DEMi,t-1 + ηi + εi,t

1 Regression 33.689 18 1.872 23.478 .000 Residual 10.284 129 .080 Total 43.973 147 Coefficientsa Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) -1.045 .738 -1.417 .159 COL .062 .135 .036 .458 .648

(23)

23 ARG -.041 .127 -.023 -.321 .749 PER .000 .137 .000 -.003 .998 VEN .021 .253 .011 .083 .934 CHI .040 .247 .023 .160 .873 ECU -.213 .191 -.118 -1.116 .267 BOL -.165 .204 -.089 -.810 .420 PAR -.385 .288 -.200 -1.339 .183 URU -.363 .272 -.201 -1.336 .184 CPI .052 .056 .161 .927 .356 LogGDPPC .834 .157 .469 5.304 .000 GDPG .011 .007 .078 1.582 .116 FDIGDP .045 .014 .207 3.179 .002 POPG -.965 .373 -.782 -2.588 .011 URBG .608 .245 .615 2.488 .014 INF .005 .004 .071 1.163 .247 DEMO -.036 .064 -.065 -.564 .574 OPEN .001 .003 .042 .357 .722 Residuals Statistics

Residual -1.1486051 .6303648 .0000000 .2644928 148

Std. Residual -4.068 2.233 .000 .937 148

8.2.3 Regression results model 3:

+ βOPEN*OPENi,t-1 + βDEM*DEMi,t-1+ βCPIC*CPICi,t-1 + ηi + εi,t

(24)

24 Residual 10.284 129 .080 Total 43.973 147 Coefficients Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) -1.045 .738 -1.417 .159 COL .062 .135 .036 .458 .648 ARG -.041 .127 -.023 -.321 .749 PER .000 .137 .000 -.003 .998 VEN .021 .253 .011 .083 .934 CHI .040 .247 .023 .160 .873 ECU -.213 .191 -.118 -1.116 .267 BOL -.165 .204 -.089 -.810 .420 PAR -.385 .288 -.200 -1.339 .183 URU -.363 .272 -.201 -1.336 .184 CPI .052 .056 .161 .927 .356 logGDPPC .834 .157 .469 5.304 .000 GDPG .011 .007 .078 1.582 .116 FDIGDP .045 .014 .207 3.179 .002 POPG -.965 .373 -.782 -2.588 .011 URBG .608 .245 .615 2.488 .014 INF .005 .004 .071 1.163 .247 DEMO -.036 .064 -.065 -.564 .574 OPEN .001 .003 .042 .357 .722 Residuals Statistics

Residual -1.1486051 .6303648 .0000000 .2644928 148

Std. Residual -4.068 2.233 .000 .937 148

8.2.4 Regression results model 4:

(25)

25 Model Summary Model R R Square Adjusted R Square Std. Error of the Estimate Durbin-Watson 1 .903 .816 .756 .2700420 2.393 ANOVA

1 Regression 35.878 36 .997 13.667 .000b Residual 8.094 111 .073 Total 43.973 147 Coefficients Model Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta 1 (Constant) .929 2.078 .447 .656 COL .087 .148 .051 .587 .558 ARG -.022 .180 -.012 -.124 .902 PER -.051 .166 -.028 -.305 .761 VEN .283 .287 .142 .983 .328 CHI -1.005 .823 -.573 -1.222 .224 ECU -.160 .227 -.089 -.706 .482 BOL -.297 .349 -.159 -.851 .397 PAR -.457 .344 -.237 -1.328 .187 URU -1.522 .854 -.843 -1.783 .077 CPI .220 .096 .678 2.284 .024 logGDPPC .361 .530 .203 .681 .497 GDPgrowth .008 .009 .061 .962 .338 FDIperGDPt-1 .031 .017 .145 1.867 .065 POPgrowth -1.179 .385 -.956 -3.062 .003 URBANpopGrowth .606 .280 .613 2.161 .033 INF .004 .005 .061 .815 .417 DEMO -.101 .074 -.180 -1.361 .176 OPEN .001 .005 .041 .229 .819 CPIxCorrupt -.137 .125 -.351 -1.097 .275 T98 .160 .330 .024 .484 .629 T99 .393 .201 .142 1.957 .053 T00 .232 .216 .096 1.075 .285 T01 .137 .307 .021 .447 .655 T02 -.072 .159 -.030 -.453 .651

(26)

26 T03 .013 .154 .006 .084 .933 T04 .121 .140 .056 .863 .390 T06 .064 .133 .029 .479 .633 T07 .211 .152 .097 1.385 .169 T08 .297 .180 .137 1.645 .103 T09 -.001 .213 .000 -.003 .998 T10 .296 .242 .136 1.224 .224 T11 .354 .236 .163 1.498 .137 T12 .401 .269 .185 1.492 .139 T13 .182 .297 .084 .612 .542 T14 .125 .307 .055 .406 .686 T15 .170 .315 .071 .540 .590 Residuals Statistics

Residual -.9944679 .4828741 .0000000 .2346573 148

The effect of corruption on foreign direct investments in South America : a panel data analysis

The effect of corruption on foreign direct investments in

South America: A panel data analysis

Name:

Adriaan van Hall

Studentnumber: 10248080

Specialization:

Economics and Finance

Supervisor

Oana Furtuna

Table of contents

1.

Introduction

2.

Literature review

2.1

Definition of corruption

2.2

Effect of corruption on FDI

2.3

Other determinants of FDI

3.

Hypothesis

4.

Methodology and Data Selection

4.1

Included variables

4.2

The model

5.

Empirical data and results

5.1

Empirical results: The effect of CPI on FDI

Table 2: Results of panel data regression

The effect of corruption on foreign direct investment

5.2

Empirical results: The effect of the other explanatory variables on FDI

6.

Conclusion

7.

Bibliography

8.

Appendix

8.1

Table 1 – country CPI scores

8.2

Regression results

8.2.1 Regression results model 1:

8.2.2 Regression results model 2:

8.2.3 Regression results model 3:

8.2.4 Regression results model 4: