Tolerable Level of Corruption for Foreign Direct Investment in Africa

Corruption has become endemic in many African countries and is difficult to eradicate completely; therefore, reducing corruption to a tolerable level that will not deter foreign investors must be the aspiration of all political leaders and stakeholders. This study tries to identify the level of corruption that is tolerable to foreign investors, which is referred to as the Tolerable Level of Corruption for Investment (TLCI). The study proposes that below the TLCI, corruption plays the role of “sand in the wheels of commerce” and thus has a negative impact on FDI inflows, but above the TLCI, corruption functions to “grease the wheels” and has a positive impact on FDI inflows. The study is based on secondary data collected from the World Bank World Development Indicators. Using a dynamic panel data estimation technique while controlling for other variables, the estimated TLCI in Africa is -0.27 on the control of corruption scale, which ranges from approximately -2.5 (weak) to 2.5 (strong). Therefore, all African leaders and stakeholders, especially in countries that fall below the TLCI, should intensify their efforts in the fight against corruption to reduce corruption in their respective countries to at least the TLCI to attract foreign investors.


Introduction
. Economists generally agree that FDI inflows lead to an increased rate of economic growth (Wijeweera, Villano, & Dollery, 2010). Particularly in developing countries, FDI inflows should exert positive effects on economic growth because these countries suffer from low productivity and capital stock deficiencies (Johnson, 2006).
Neoclassical theory predicts higher marginal returns to the factor that is relatively scarce. Thus, capital should flow from rich countries to Africa -where capital is relatively scarce. For example, the rates of return on FDI were 7 percent globally and higher in both developing (8 percent) and transition (13 percent argued that a strong policy and regulatory regime, appropriate institutions, good infrastructure, and political and economic stability are important for attracting FDI inflows (Mwilima, 2003). A non-policy factor that plays a role in the attraction of FDI into a country is its level of institutional quality. Most research on the effect of institutional quality on FDI inflows reveals that countries that have weak institutions, especially high corruption and unreliable legal systems, tend to receive less FDI (Gastanaga, Nugent, & Pashamiova, 1998;Wei, 2000b). African countries still suffer from varying levels of negative perceptions from the outside world despite all the pro-FDI policies implemented to encourage FDI inflows. Factors contributing to these include corruption and political instability due to internal conflict, external conflict, military involvement in politics and religious tension. Over the years, the majority of African countries scored a 3.0 or below on the corruption perception index rating produced by Transparency International. For example, 87% of the countries in Africa scored a 3.0 or below in the years 2011, 2009 and 2008; 83% in 2010; and 77% in 2012. These percentages are alarming. The findings of Treisman (2000) suggest that fighting corruption in many countries has proved so difficult because it greatly varies between countries. Because of the difficulty of eradicating corruption, reducing its prevalence to a tolerable level must be the aspiration of all political leaders and stakeholders. In 2012, the number of countries that scored a 3.0 or below fell to 77%. In the same the year, however, Africa reversed the downward trend in FDI, exhibiting a 5 percent increase in FDI inflows to $50 billion. This gives an indication that there is a level of corruption that is tolerable to investors.
Notwithstanding perceptions of corruption levels, FDI continues to flow to some countries, which also supports the idea of a level of corruption that is tolerable to investors. This level of corruption will likely not deter potential investors from investing in Africa. As corruption cannot be completely eradicated, reducing it below a threshold that can be accommodated by investors is a realistic goal for African leaders. In this study, this threshold is referred to as the Tolerance Level of Corruption for Investment (TLCI). This research not only seeks to establish that corruption generally has a negative impact on FDI inflows to Africa but also to show that there is a threshold (TLCI) below which corruption is expected to have a positive impact on FDI inflows to Africa. The TLCI will motivate leaders in Africa to try to control corruption in their countries to levels that will not deter FDI inflows because corruption is difficult if not impossible to eradicate completely. This will lead to an increase in economic growth and reduce both unemployment and poverty on the African continent. The TLCI will also serve as openness, natural resources and political stability, economic stability and growth prospects, and infrastructure facilities. This is followed by a presentation of the methodology deployed in the study. The rest of the paper presents the results, discussion, and finally, the conclusion.

Literature review
Both policy and non-policy factors have been identified as drivers of FDI inflows in the literature (Fedderke & Romm, 2006). Policy factors include openness, product market regulations, labor market arrangements, corporate tax rates, trade barriers, and infrastructure. Non-policy factors include corruption, market size of the host country, distance/transport costs, factor endowments, and political and economic stability (Mateev, 2009). The framework on Multinational National Enterprises (MNEs) posits that firms invest abroad to look for three types of advantages: Ownership (O), Location (L), and Internalization (I); hence, it is called the OLI framework. A firm can use its specific advantages in the foreign country to earn a higher marginal profit or decrease its marginal cost compared to its competitors (Dunning, 1973;1988). The institutions in the domestic country have the potential to attract MNEs depending on whether, given the existing institutions, the foreign firm can capitalize on its location advantage. Internalization theorists opine that FDI occurs when the benefits of internalization outweigh its cost (Fina & Rugman, 1996). Firms therefore exploit their ownership and location advantages to minimize their transaction costs.
One area of institutions of the domestic country that has generated much interest in recent times is corruption. Public corruption according to Svensson (2005) is the misuse of public office for private gain, which includes the sale of government property, kickbacks in public procurement, bribery and embezzlement of government funds.
Differences among countries with respect to the extent of corruption may depend on the degree to which officials compete against each other to sell mutually substitutable benefits to private agents (Shleifer & Vishny, 1993). Svensson (2005) found the highest levels of corruption to be associated with developing or transition countries. Corruption has become endemic in many African countries and is difficult to eradicate completely. Countries with strong institutions are expected to reduce or maintain corruption at a tolerable level to attract investors. The "grease the wheels" hypothesis is more prominent in the early economics literature, with much emphasis on the effects of corruption on efficiency (e.g., Leff, 1964;Leys, 1965;Huntington, 1968). The "grease the wheels" hypothesis suggests that an inefficient bureaucracy creates a major impediment to economic activity, so some ''grease" money may be needed to circumvent this impediment. Beck and Maher (1986) and Lien (1986) suggested that corruption can increase efficiency because inefficient regulations constitute an obstacle to investment which can be removed by bribing bureaucrats. Some studies (Egger & Winner, 2005) found positive shortand long-run impacts of corruption on FDI, which supports the position of Leff (1964). There is some current empirical evidence in support of the "grease the wheels" hypothesis (Vial & Hanoteau, 2010). Corruption may be beneficial in a second-best world by alleviating the distortions caused by ill-functioning institutions. However, some economists are of the view that corruption would tend to reduce economic growth (Shleifer & Vishny, 1993). This negative impact of corruption is viewed as "sand in the wheels of commerce" (Cuervo-Cazurra, 2008). Malfunctioning government institutions have been contended to constitute severe obstacles to investment, entrepreneurship, and innovation by many economists (Mauro, 1995 Earlier studies on corruption and firm efficiency found corruption to negatively affect the efficiency of firms (Dal Bo & Rossi, 2007;Picci, 2005;Yan & Oum, 2011).
At low levels, corruption is seen as "greasing the wheels", and at high levels, it is seen as "sand in the wheels of commerce". This study argues that in countries in which bureaucratic regulations are cumbersome, corruption might be a means to achieve certain benefits by foreign investors; thus, they are motivated to invest in those countries. In such countries, corruption is expected to have a positive impact on private investment. However, when corruption goes beyond the paying of bribes to levels of malfunctioning government institutions, corruption is expected to have a negative impact on private investment. In such countries, corruption may deter foreign investment. The corruption variable is captured as the perception of corruption in the public sector of the host country and is expected to have both negative and positive effects on the inflows of FDI into a country depending on the levels of institutional quality and corruption.

GDP growth rate and GDP per capita
The economics literature indicates that FDI has led to economic development of the host country because FDI inflows facilitate the acquisition of valuable tangible and intangible assets, such as enhanced technology, managerial skills, expertise, innovation capability, capital formation and related physical assets (Liu, Shu, & Sinclair, 2009;Vu, Gangnes, & Noy, 2008;Wang, 2009). Elsewhere, market size has been predicted to be a positive and significant determinant of FDI flows (Garibaldi et al., 2002;Nunes, Oscategui, & Peschiera, 2006;Sahoo, 2006

Trade openness
Trade openness refers to the sum of exports and imports of goods and services into a country and gives an indication of how liberalized a country is in terms of trade. The impact of trade openness on economic growth can be positive and significant mainly due to the accumulation of physical capital and technological transfer as a result of FDI inflows. Therefore, trade openness is an important vehicle for technological spillovers. According to Eicher (1999), Lee (1993) and Young (1991), openness to trade also stimulates domestic investment by encouraging competition in domestic and international markets and generating higher returns on investment through economies of scale. Trade openness is generally a positive and significant determinant of FDI inflows (Asiedu 2002;Sahoo 2006). Trade openness is captured as trade as a share of GDP and it is expected to facilitate the flow of FDI into the host country.

Natural resources and Political stability
FDI attraction to Africa can also be influenced by the availability of natural resources on the continent. Jadhav (2012) suggest that resource-seeking FDI is motivated by the availability of natural resources in host countries. This resource seeking remains a relevant source of FDI for various developing countries. Studies have shown that natural resources play vital roles in overall attraction of FDI to Africa (Asiedu, 2002;Dupasquier & Osakwe, 2006). In Africa, countries that have natural resources were more attractive than those without such resources (Asiedu, 2005). According to North and Weingast (1989) and Li (2009)

Economic stability and Growth prospects
Economic stability has been found to be a positive indicator of FDI inflows (Mateev, 2009). A country that has stable macroeconomic conditions with high and sustained growth rates is expected to have more FDI inflows than a more volatile economy (Ranjan & Agrawal, 2011). Proxies for the macroeconomic stability of a country include GDP growth rates, industrial production index values, interest rates, and inflation rates (Dasgupta & Ratha, 2000). High inflation rates are associated with economic disarray and lower purchasing power, so inflation risk becomes an important factor in long-run investment plans. Inflation has been found to have a negative relation with FDI inflows though its magnitude is much smaller (Ranjan & Agrawal, 2011).
However, research on the influence of exchange rates on FDI inflows has shown varied results. While Kyereboah-Coleman and Agyire-Tettey (2008) posit that the volatility of the real exchange rate has a negative influence on FDI inflows, Jeon and Rhee (2008) show that FDI inflows have significant association with both the real exchange rate and expected exchange rate changes. Nonetheless, Brahmasrene and Jiranyakul (2001) and Dewenter (1995) find no statistically significant relationship between the level of the exchange rate and FDI inflows (Anyanwu, 2012).
When a country's currency depreciates, foreign investors take advantage of the ability to purchase assets at a reduced cost. Investment in countries whose currencies face high depreciation is relatively less expensive.
Therefore, it is expected that a high inflation rate in the host country attracts less FDI, while higher pressure to depreciate the exchange rate of the host country attracts more FDI.

Infrastructure facilities
The importance of infrastructure development to attracting FDI inflows cannot be ignored. Studies by Musila and Sigue (2006) and Dupasquier and Osakwe (2006) show that FDI in Africa is dependent on the development of infrastructure. Similar results were obtained by Kersan-Skabic and Orlic (2007)

Data
With the exception of the control of corruption index, the variables used in this study are based on secondary data collected from the World Development Indicators This index is chosen not only because of its authenticity but also because of its free availability on the internet. The control of corruption index is one of the six dimensions of governance included in the Worldwide Governance Indicators.

Data Analysis
In order to meet the objectives of the study, a dynamic panel data estimation technique is used.
Several studies have found lagged FDI to be correlated with current FDI (Asiedu, 2013) System GMM estimate also has an advantage over the Difference GMM with respect to variables that exhibit "random walk" or are close to random-walk variables (Baum, 2006;Bond, 2002;Roodman, 2006;2007). According to Efendic, Pugh and Adnett (2009), because model specifications including macroeconomic variables are known in economics to be characterized by random walk statistical generating mechanisms, the System GMM approach seems more suitable. Empirical research with dynamic models shows that the System GMM is a good estimator, or at least better than the Difference GMM, which is severely downward biased (Hoeffler, 2002;Nkurunziza and Bates, 2003;Presbitero, 2005). Moreover, Roodman (2006) suggests that it is better to avoid Difference GMM estimation, which has a weakness of magnifying gaps if one works with an unbalanced panel .
The general model is of the form presented in equa- where it i it u v ε = + , for i = 1,…, N and t = 2,…, T, with 1 α < . The disturbance term it ε has two orthogonal components, which are the fixed effects i u and the id- The framework for evaluating the associations among FDI, corruption, and other determinants of FDI is presented in equation (2).
where it y is a measure of FDI in country i at time period t , , 1 i t y − is a measure of FDI in country i at time period 1 t − , it x is an index of the control of corruption in country i at time t , 2 it x is the squared index of control of corruption in country i at time period t , it z are control variables in country i at time period t , 1 β , 2 β , 3 1 , β α and ω are parameters to be estimated, and finally, it ε denotes the disturbance term. StataCorp 2013 is the statistical software used in the data analysis.    No external instruments are used. In this panel, there are 50 countries (N) that are analyzed over a period of 17 years (T), which means there are more countries (N) than years (T). It has been argued by many authors that dynamic panel models are especially designed for situations wherein T is smaller than N to control for dynamic panel bias (Baltagi, 2008;Baum, 2006;Bond, 2002;Roodman, 2006;2007;Sarafidis, Yamagata, & Robertson, 2009). Nerlove (2002) argues that economic behavior is inherently dynamic, and thus, most econometrically interesting relationships are explicitly or implicitly to remove serial correlation in the disturbance term (Beck & Katz, 1996). Dynamic panel models are useful when the dependent variable depends on its own past realizations. In addition, models including lagged dependent variables can also control, to a large extent, for many omitted variables.

Dynamic Panel Model
Once a lagged dependent variable is included as part of the panel model specification, there is a violation of strict/strong exogeneity because the lagged dependent variable, which is one of the regressors, is correlated with past values of the error term. The correlation of the idiosyncratic error term it v with the lagged dependent variable , 1 i t y − at time t + 2 is the source of the strict exogeneity. There is also a violation of the weaker condition of no contemporaneous correlation of the regressors with the composite error term ( When it y is correlated with the fixed effects in the error term, it gives rise to "dynamic panel bias" (Nickell, 1981). The endogeneity problem renders the estimators inconsistent and inferences from the estimated model less accurate.

The GMM Estimator
The panel dynamic model takes the following form (equation (4)) where y exhibit state dependence: as standard instruments and the lags of the endogenous variables to generate the GMM-type instruments as described in Arellano and Bond (1991) and includes lagged differences of the endogenous variables as instruments for the level equation.

Specification Testing in Dynamic Panel Models
Specification testing in dynamic panel models is con-

Testing for Residual Serial Correlation
The degree of serial correlation of the residual term in i.e., 0 . One should therefore test for second order serial correlation because its presence indicates a specification error. The idiosyncratic disturbance term it v is related to , 1 i t v − ∆ mathematically via the shared , 1 i t v − term, so a negative first order serial correlation is expected in differences meaning that its evidence is of no importance. Therefore, to check for first order serial correlation in levels, it is important to check for second order correlation in differences as well, as this will detect correlation be- Therefore, serial correlation of order l in levels is checked by looking for correlation of order 1 l + in difference (Roodman, 2009). These tests lose power when the number of instruments i is large relative to the cross section sample size n . The rule of thumb is to keep the number of instruments less than or equal to the number of groups so that when the ratio r of the sample size to the number of instruments is less than one, 1 n ratio i = < , the assumptions underlying the two procedures may be violated.

The Estimation of Tolerable Level of Corruption for Investment
Relationships between two economic variables are predicted to be non-monotonic in various economic theories. A popular empirical test of such theories, accord-ing to Plassmann and Khanna (2003), is to estimate an equation using a polynomial of the variable that is

Test of the U-Shaped Relationship
The control of corruption variable scale ranges from approximately -2.5 (weak) to 2.5 (strong), which means that the higher a country is on the scale, the better governance performance against corruption and, thus, the smaller the level of corruption. Therefore,

Descriptive statistics
The descriptive statistics of the variables deployed in the study are presented in Table 2. All the variables have values ranging from 727 to 877, as the highest observations. The period under study is from 1996 to 2012. Variables obtained from Worldwide Governance Indicators (control of corruption and political stability) have data missing for three years (1997, 1999, & 2001), and this accounts for those variables having

Empirical Results of the Dynamic Panel Model
The results of the dynamic panel model estimated including endogenous and exogenous variables in addition to the lagged dependent variable are presented in Table 3. The FDI net inflow per GDP is used as the dependent variable in the estimation of the FDI model.
The control of corruption variable and its squared values as well as other control variables are used as independent variables. The two-step estimator is deployed in the estimation, with control of corruption and trade openness variables treated as endogenous and all other independent variables treated strictly as exogenous.
No external instruments are used.

Model Specification Diagnostics Test
The validity of the estimated results in System GMM depends on the statistical diagnostics tests. The results indicate that the specification pass the Han-  (2) is not rejected, and therefore, the Arellano-Bond test for serial correlation supports the validity of the model specification (Basu, 2008). As the number of instruments (47) is less than the number of groups (50), the assumptions underlying the two procedures are not violated. The 47 instruments came from the restriction to use two lags for levels and two for differences in the data (i.e., the restriction is set to (2 2) in xtabond2).
The two-step estimates that report the Hansen Jstatistic test yield theoretically robust results (Roodman, 2006). The Hansen J-statistic tests the null hypothesis of correct model specification and valid over-identifying restrictions, i.e., the validity of instruments. The rejection of the null hypothesis means that either or both assumptions are violated. Baum (2006) argues that the Hansen J-test is the most commonly giving an indication that the model has valid instrumentation. The difference-in-Sargan/Hansen test, also known as the C-test (Baum 2006;Roodman, 2006), is used to test the validity of subsets of instruments (i.e., levels, differenced, and standard IV instruments). It estimates the System GMM with and without a subset of suspect instruments enabling investigation of the validity (i.e., exogeneity) of any subset of instruments as well as their contribution to "the increase in J-test" (Roodman, 2007). The null hypothesis of the model diagnosis test, which states that the specified variables are proper instruments, i.e., the set of examined instruments is exogenous with p-value 0.471 for GMM differenced instruments and 0.137 for system instruments cannot be rejected. This shows that the exogeneity of any GMM instruments used, i.e., levels and differenced instruments, are valid instruments. Similarly, the null hypothesis of the model diagnosis test states that the specified variables are proper standard "IV" instrument subsets cannot be rejected. Efendic et al. (2009)

Interpretation and discussion of results
The results of the estimated System GMM are presented in Table 3. Depending on the sign (+/-) of the estimates, a one-unit increase of the independent variable will lead to either an increase or decrease of the dependent variable with a magnitude determined by the corresponding coefficients. All variables with positive estimates have positive impact on the dependent variable, and those with negative estimates have negative impact on the dependent variable. The results show that control of corruption is negative and highly significant while the square of control of corruption is positive and highly significant. The control of corruption scale ranges from approximately -2.5 (weak) to 2.5 (strong) which mean that the higher the score of the country, the less corrupt it is. Thus, at low scores, corruption has a negative impact on FDI inflows, and at high scores, corruption has a positive impact on FDI inflows. This gives an indication that below certain level of corruption, a country is able to attract foreign direct investment, and beyond that level, potential investors are no longer motivated to invest in that country. Potential foreign investors in Africa are very sensitive to the perception of corruption in the host country. This confirms the evidence from earlier studies that corruption deters foreign direct investments (Aizenman & Spiegel, 2002;Barassi & Zhou, 2012;Cuervo-Cazurra, 2006Habib & Zurawicki, 2002;Hakkala et al., 2008;Javorcik & Wei, 2009;Voyer & Beamish, 2004;Wei, 2000a).
With exception of inflation, as expected, and GDP per capita, all the other control variables are positive.
In purpose. The findings also shows that trade openness is a positive and significant determinant of FDI inflow.
The results show that a 1-unit increase in the percentage of trade openness to the GDP of a country leads to 4.13% increase in the percentage of FDI inflow to GDP of that country supporting the assertion that trade liberalization leads to increased FDI inflow (Anyanwu, 2012;Asiedu, 2002;Ranjan & Agrawal, 2011;Sahoo, 2006). The results also show that a 1-unit increase in inflation leads to -1.59% decrease in the percentage of FDI inflow to GDP. The higher the volatility of the inflation rate, the more unstable is the macroeconomic environment of the host country and lower is the FDI inflow to that country. This results is consistent with Ranjan and Agrawal (2011) who found inflation to have a negative relation with FDI inflow though its magnitude is very less. Similarly, a 1-unit increase in previous year's GDP growth rate leads to 1.54% increase in the percentage of FDI inflow to GDP. This shows that because GDP growth rate represent a country's economic track record it is an indicator of profitable investment opportunities to the outside world. This finding is consistent with earlier assertion that market size is a positive and significant determinant of FDI flows (Garibaldi et al., 2002;Nunes et al., 2006;Sahoo, 2006). Contrary to expectations, GDP per capita have a negative and significant association with FDI inflows, but this finding is consistent with earlier findings (Dauti, 2008

The estimated Tolerable Level of Corruption for Investment
The results in Table 4 show that at certain level of corruption of the host country, investors are motivated to invest in that country, but below that level, investors decline to invest in that country. Estimating the level of corruption that is likely to attract potential investor to Africa is very important not only to African leaders but to all (new and old) potential investors in Africa. This level of corruption is the TLCI of a country, which will determine whether FDI is likely to flow to a country. The coefficient 2 β of the control of corruption variable tells both the direction and steepness of the curvature. As 2 β is a positive value, it indicates that the curvature is upwards but less steep.   Table 3 show both the control of corruption and the square of control of corruption are significant. However, these criteria, though sensible, are neither sufficient nor necessary and are too weak, as argued by Lind and Mehlum (2007). Lind and Mehlum (2007)

Conclusion
Many empirical studies have examined the influence of corruption on economic growth at the country level, but only a few have looked at the effects of the level of corruption on FDI inflows. The quality of institutions or level of corruption in the domestic country has the potential to attract foreign direct investment depending on whether the foreign firm can exploit its location advantage (Abotsi, 2015) within the existing institutions. As corruption cannot be completely eradicated, reducing it to a threshold that can be accommodated by investors must be the goal that African leaders endeavor to achieve. This threshold is referred to as the ing of government institutions. Therefore, all African leaders and stakeholders, especially in countries that fall below the TLCI, should intensify their efforts in the fight against corruption to reduce the level of corruption in their respective countries to at least the TLCI to attract FDI to enhance their development. This TLCI will also guide potential investors in selecting which African countries to invest in.
This study's limitations result from the nature and availability of the data deployed in the study. The frequency of the data is annual, and it spans from 1996 to 2012 for 50 countries in Africa with data missing for three years (1997, 1999, & 2001). More robust results would have been obtained if these data were available and included in the analysis. Another limitation to this study is the assumption that foreign investors choose a country based solely on the level of corruption of the host country because there are other country business risks and individual-specific shocks that investors take into consideration before an investment decision is made. It is recommended that in measuring corruption, researchers should endeavor to disaggregate corruption into its various components, such as bribes, kickbacks, and malfunctioning state institutions, because this will not only help stakeholders make informed decision in anticorruption policy formulation but also help them to know where to direct these policies.