The nexus between institutions, foreign aid and foreign direct investment

In both of the explanations, foreign aid should be asignificant factor of private capital inflows, which are generally accepted to vigorouslypromote growth, technology, and employment in

UNIVERSITY OF ECONOMICS INSTITUTE OF SOCIAL STUDIES

HO CHI MINH CITY

VIETNAM - NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

THE NEXUS BETWEEN INSTITUTIONS, FOREIGN AID,

AND FOREIGN DIRECT

INVESTMENT

BY

LE M TUE

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

HO CHI MINH CITY, JULY 2015

VIETNAM

INSTITUTE OF SOCIAL STUDIES THE HAGUE

THE NETHERLANDS

VIETNAM - NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

THE NEXUS BETWEEN INSTITUTIONS, FOREIGN AID,

AND FOREIGN DIRECT

INVESTMENT

A thesis submitted in partial fulfilment of the requirements for the degree of

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

I am grateful to Dinh Cong Khai, my academic supervisor, and Pham Khanh Nam, a member of the VNP scientific committee, for helpful and detailed comments

This paper examines the mutual relationship between foreign aid and foreign directinvestment (FDI), which might be ambiguous by reverse causality or simultaneity problems.Using the dual-approach dynamics-balanced (DADB) model, we are able to point out thatboth bilateral and multilateral aid could lead to more FDI, and the impact of the latter could

be even larger than that of the former The institutional effect of multilateral aid is proposed

to explain this phenomenon Interestingly, the role of political stability could surpass those ofdemocracy and control of corruption in having more aid disbursements

1 INTRODUCTION 1

1.1 Practical Motivation and Research Problems 1

1.2 Research Objectives 2

1.3 Structure 2

2 LITERATURE REVIEW 3

3 MODEL AND DATA 6

3.1 Dual-Approach Framework 6

3.2 Dual-Approach Dynamics-Balanced Model 8

4 RESULTS 12

4.1 Independent Marginal Effects between Institutions, Foreign Aid, and FDI 12

4.2 Reliability and Robustness Checks 14

5 CONCLUDING REMARKS 29

5.1 Empirical Findings 29

5.2 Policy Implication 29

5.3 Research Contribution, Implication, and Limitations 30

5.4 Future Research 30

REFERENCES 31

APPENDIX A 33

APPENDIX B 34

LIST OF FIGURES

Figure 1: The Nexus between Institutions, Foreign Aid, and FDI 8

LIST OF TABLES Table 1: Descriptive Statistics 10

Table 2: Independent Marginal Effects between Institutions, Foreign Aid, and FDI, 1996-2012 18

Table 3: The Impacts of Different Institutional Measures on FDI and Foreign Aid 22

Table 4: Independent Estimation of the Dynamic FDI Equation 25

Table 5: Independent Estimation of the Dynamic Aid Equations 27

LIST OF APPENDICES Table A 1: Variables and Data Sources 33

Table B 1: Regression Result of Table 2, column (1) 34

Table B 2: Regression Result of Table 2, column (2) 35

Table B 3: Regression Result of Table 2, column (3) 36

Table B 4: Coefficient of INSA index in Table 3, Panel A, column AA 37

Table B 5: Coefficient of INSA index in Table 3, Panel A, column BA 38

Table B 6: Coefficient of INSA index in Table 3, Panel A, column MA 39

Table B 7: Regression Result of Table 4, column (7) 40

Table B 8: Regression Result of Table 4, column (8) 40

Table B 9: Regression Result of Table 4, column (9) 41

Table B 10: Regression Result of Table 5, column (7) 41

Table B 11: Regression Result of Table 5, column (8) 42

Table B 12: Regression Result of Table 5, column (9) 43

Figure B 1: Scree Plot of Eigenvalues of Components for Five Variables of INSF index 44

Figure B 2: Scree Plot of Eigenvalues of Components for Three Variables of INSA index 44

1 INTRODUCTION

1.1 Practical Motivation and Research Problems

From the behavior aspect, when a country receives more aid from a donor, it will

acknowledge the generosity of that donor, easily cooperate with that sovereign partner, andcreate a concessionary legal environment for the enterprises of that donor (Kimura & Todo,2010; Rodrik, 1995, p 25) From the effectiveness aspect, when a country receives more aid

from all donors, it is able to improve the social and economic infrastructures, thus the human

capital as well as the total factor productivity increase accordingly (Harms & Lutz, 2006).Hence, the recipient country would be able to attract more FDI from any other countries due

to its increasing competitiveness In both of the explanations, foreign aid should be asignificant factor of private capital inflows, which are generally accepted to vigorouslypromote growth, technology, and employment in the host country Nevertheless, researchstudies have not found a robust relationship between foreign aid and FDI (Alesina & Dollar,2000; Harms & Lutz, 2006)

Harms and Lutz (2006) suggest that one should consider the role of political andinstitutional characteristics when quantifying this relationship Indeed, institutional quality ofthe host country itself is an important direct magnet of private capital inflows Abundantempirical studies have pointed out the negative causality of a bad institution to the inflows ofFDI Ironically, several countries which are perceived as having high corruption and lowpolitical, institutional profiles still have large inflows of FDI (Habib & Leon, 2002)

In the aspect of modeling, the influence of foreign aid on FDI is difficult to estimate due

to the problems of simultaneity and reverse causality By using lagged variables asinstruments, 2SLS and GMM methods, to some extent, could alleviate such endogeneity.However, the treatment is purely technical and does not reflect the nature of the problems.Asiedu, Jin, and Nandwa (2009) propose a simultaneous equations model that could solvethese problems In this approach, foreign aid and FDI are determined at the same time, andeach of them is the determinant of the other While the dual approach is undoubtedly a superbidea, the applied model and the results of this research nonetheless contain some flaws andcontradictions First, there is no institutional determinant in the aid equation Second, in theaid equation, the positive coefficient of FDI could be interpreted that while foreign aid

1.2 Research Objectives

This paper aims to modify the simultaneous equations model of Asiedu et al (2009) andvisualize the intricate relationship between foreign aid and FDI, which might comprisesimultaneity and reverse causality With regard to purposes, while Asiedu et al (2009) focus

on the alleviating role of foreign aid on the adverse effect of expropriation risk on FDI, weconcentrate on the effect of foreign aid on final FDI With regard to samples, we use bothlow-income and middle-income subsamples to support our analysis, whereas it is only thelow-income countries in Asiedu et al (2009) In comparison with Harms and Lutz (2006), weapply a different model with different proxies of variables and a more recent period to assessthe effect of foreign aid on FDI

After setting up the framework and specifying the according model, we use the data toillustrate the mutual relationship between foreign aid and FDI Using this result, we expect tofigure out whether multilateral aid could actually lead to more FDI As for researchersconcerned with the determinants of FDI and foreign aid, this paper provides more empiricalevidence on the role of institutions In particular, we reexamine whether better institutionscould attract more FDI as postulated in theory and found in many research studies By theway, we also appraise the importance of different institutional measures on FDI In the otherside, do democracy (freedom), control of corruption, and political stability help a countryreceive more foreign aid?

1.3 Structure

In Chapter 2, this paper briefly reviews a trade theory which is widely used to explain theinvestment decision of foreign investors and some empirical results based on this theory Wemainly concentrate on the papers that have institutions and foreign aid as the determinants ofprivate capital inflows In Chapter 3, we explain the dual-approach framework and theregression model The variables and data sources are also described in this chapter Theempirical findings and associated explanations are located in Chapter 4 Chapter 5recapitulates the results for making policy and research

2 LITERATURE REVIEW

We first review the OLI theory on the investment decision of foreign investors; then we

come into the papers mentioning institutions and foreign aid as separate explanatory

variables of FDI; next, we have a look on the research which embeds the political and

institutional factors into the influence of foreign aid on FDI Lastly, we summarize the

institutional measures which might affect foreign aid

Dunning (1988, 1998, 2001) built up the OLI paradigm as a general framework to explainthe activities of foreign investors The ownership advantages are classified as the Ocomponent and emphasize the comparative advantages of firms which can expand theirbusiness abroad Analyzing the O component provides us with information about the nature

of products and the ability of firms The location advantages define the L component and arerelated to the human and natural resources, the favorable conditions for production, business,research activities, and the market size in the host country The internalization advantagesbelong to the I component and focus on the aspect of how to lower transaction costs, as firmsdecide whether importing intermediate products from markets or internalizing foreignsuppliers into their production chain The I component could be taken into analysis by firms

at the time of choosing the destination

Political and institutional factors of the host country are considered as the locationadvantages in the OLI framework The influence mechanism of these factors on foreigninvestors are mentioned in the papers such as Habib and Leon (2002) and Dunning andLundan (2008) On the empirical side, Habib and Leon (2002) find out a negative relationshipbetween corruption levels in the host countries and their inflows of FDI According to Habiband Leon, foreign investors might see corruption as violating social and professional ethicsand increasing unnecessary costs.1 Moreover, paying bribes is strictly prohibited in the homecountries of some foreign investors such as the United States (Hines, 1995)

Busse and Hefeker (2007) examine the impacts of government stability, law and order,absence of internal and external conflicts, lack of ethnic tensions, control of corruption,

democracy, and bureaucratic performance on FDI inflows to developing countries in the

period 1984-2003 The paper does show positive relationships between such measures andprivate capital inflows With the same period of research, Bénassy-Quéré, Coupet, and Mayer

(2007) use the gravity model to test the influences of various institutional data on FDI stocks.

In general, a country owning higher institutional quality would have a larger aggregate stock

of FDI, and higher institutional distance between the home country and the host countrylowers the FDI stock of the former in the latter In other words, bad institutional measureswould hinder FDI to a country For example, one of the results in Asiedu et al (2009) is thatexpropriation risk restrains the investment decision of foreign enterprises The samples of thisresearch are low-income and Sub-Saharan African countries

On the other hand, the number of research studies that find out bad institutions as anincentive of FDI is quite limited Egger and Winner (2005) show the empirical evidence onthe positive relationship between corruption and inward FDI The research uses a sample ofboth developed and developing countries in the period 1995-99 to strengthen the position ofLeff (1964): bribing could reduce uncertainty among low-informational countries andsafeguard foreign investors under major economic and political changes

In contrast to institutions, the role of foreign aid per se on FDI is ambiguous whenanalyzing the data As hypothesized by some authors, foreign aid does not only have thepositive effect on FDI, but also the adverse effect The two simultaneous contradictory effectscould be infrastructure effect and rent-seeking effect (Harms & Lutz, 2006), or complementand substitution (Selaya & Sunesen, 2012) As a result, many research studies could not find

a significant influence of aggregate aid on FDI (Bird & Rowlands, 1997; Harms & Lutz,2006; Kimura & Todo, 2010) Selaya and Sunesen (2012) is a sparse research which finds outthe dominance of the positive effect However, the role of bilateral aid in attracting FDI ismore robust and could be found in Rodrik (1995) and Kimura and Todo (2010) Formultilateral aid, there has been no empirical evidence that it could instantly enhance FDI(Rodrik, 1995) In the worst case, negative coefficients of aggregate aid, bilateral aid, andmultilateral aid in the FDI regressions are exposed in the research of Asiedu et al (2009) forlow-income and Sub-Saharan African countries

Aiming to address the vague impact of foreign aid on FDI, Harms and Lutz (2006) allowthe marginal effect of foreign aid in the FDI equation to be dependent upon institutionalvariables, use different estimation techniques, time periods and country groups, and evendisaggregate the types of foreign aid and private foreign investment Unfortunately, in linewith previous papers, the authors still conclude that: on average, there is no effect of foreignaid on FDI Yet, the authors discover a remarkable result: only in the countries with highregulatory hindrance, i.e low institutional quality, foreign aid has a positive effect on FDI Asnoted by the authors, this phenomenon emerges because the role of foreign aid as

reducing expropriation risk, thus increasing foreign investment, becomes conspicuous inthose countries Similarly, the positive coefficient of the interaction term betweenexpropriation risk and foreign aid in Asiedu et al (2009) means that expropriation risk ispositively added to the effect of foreign aid on FDI Given the negative coefficient of foreignaid at the beginning, the increase of expropriation risk level, to some extent, could makeforeign aid moving in the same direction with FDI, although expropriation is perceived as abad practice.

In regard to institutional determinants of foreign aid, Alesina and Dollar (2000), Dollarand Levin (2006) have consensus on the positive impact of democracy Corruption level,however, is not found to ultimately affect the decision of a majority of donors (Alesina &Weder, 2002)

3 MODEL AND DATA

3.1 Dual-Approach Framework

Figure 1 graphically illustrates the influence of institutions on foreign aid and FDI, andthe interaction between these two foreign inflows The former induces simultaneity and thelatter causes reverse causality Solid arrows are the foci of this paper In the bottom of thefigure, institutions are the major factors of both FDI and foreign aid The OLI theory suggeststhat a better institution could have more FDI inflows Similarly, foreign aid might be

disbursed more to a country with better governance Hence, α2 and β2 are expected to bepositive

The amount of foreign aid received by a recipient could affect aggregate private capital

inflows with an average magnitude α1 In case of multilateral aid, it aims to enhance the

social and economic infrastructures and is typically thought as creating the “infrastructure

effect.” As a matter of fact, this effect does not cover all the cases and is likely to occur in thelong run In the short run, suppose that country A and country B nearly have the sameconditions as well as the institutional measures When country A receives more multilateralaid than country B in a period, it does not necessarily mean that country A could improve itsinfrastructure instantly and have better infrastructure than country B, and thus could attractmore foreign investors This argument especially makes sense insofar as multilateral aid, ingeneral, is not targeted to support a specific industry Rather, it is the information andconditional policy functions of multilateral aid that could invisibly protect foreign investors

in destination countries.2 With this reasoning, this paper proposes the “institutional effect” ofmultilateral aid Two features of the institutional effect are that it could safeguard any foreigninvestor, but it could not change the institutional nature in the short run

In case of bilateral aid, besides productive sectors, a number of its projects are also forinfrastructural development, and it should also have the same infrastructure effect as that ofmultilateral aid In addition, bilateral aid has its own special effect which is known to greasebilateral private capital inflows and is referred to as the “vanguard effect.” It is noticed thatthe vanguard effect of bilateral aid is quite similar to the institutional effect of multilateral

aid, but only applied for bilateral investors The coefficient α1 is conjectured to be positive inboth cases of foreign aid The impact level is expected to vary systematically among

2 See Rodrik (1995) for detailed discussion on the roles of multilateral aid.

governments having different performances Therefore, the marginal effect of foreign aid

could be expressed as a linear function of institutions, i.e α1 + δ institution The idea of

putting political and institutional variables into the effect of foreign aid on FDI was firstemployed by Harms and Lutz (2006)

The mechanism through which FDI affects foreign aid could be direct or indirect The

direct mechanism is suitable to explain the inflows of bilateral aid occurring in the short run

It means that, when having more FDI in a host country, the government of home country hasmore motivation to disburse aid to that location Of course, it is not mandatory for thegovernment of home country to do so because any country would welcome FDI However, aseconomic activities somewhat reflect the intimacy in diplomatic relationship between twocountries, the use of FDI as a determinant of bilateral aid is still acceptable On the otherhand, multilateral donors certainly are not concerned about FDI in recipient countries To

sum up, in the short run, the coefficient β1 of FDI inflows could be positive in the bilateralaid equation and neutral in the multilateral aid equation

In the long run, it is the indirect mechanism that FDI could lead to the change in bothbilateral and multilateral aid FDI might enhance GDP growth and GDP per capita, and, inturn, these outcomes decide the amount of foreign aid.3 In other words, if the inflows of FDIare actually transformed into the wealth of nation, that host country, eventually, would not

rely on concessionary loans or grants of donors anymore With this argument, the β1

coefficients of FDI inflows in the bilateral and multilateral aid equations should turn negative

in the long run

Also from FDI to foreign aid, the marginal effect of the former on the latter could beadjusted by the institutional quality Different measures of institutions, such as democracyand corruption, are found to affect GDP growth and GDP per capita (Acemoglu, Naidu,Restrepo, & Robinson, 2014; Barro, 1996; Mauro, 1995) By adding institutional factors into

the coefficient of FDI inflows, i.e β1 + γ institution, we could, in principle, evaluate how an

economy absorbs FDI and makes itself less relied on the support of foreign donors

Figure 1: The Nexus between Institutions, Foreign Aid, and FDI.

3.2 Dual-Approach Dynamics-Balanced Model

From the dual-approach framework above, we have the general dual-approach model:

fdi = f1 (α1 aid ,α 2 insf ,λ1others), aid = f 2 ( β1 fdi , β 2 insa , λ2others).

This model augments the simultaneous equations model of Asiedu et al (2009) by adding the

political and institutional determinant, insa, in the aid equations The detailed model for

+ λ23debtgdp i,t−2 + λ24lnpop it + u 2it ,

where infdigdp it is the net inflows of foreign direct investment, and aidgdp it is the net

inflows of official development assistance (ODA) to country i in year t Both figures are in

percentage of GDP The first lag of aid inflows in the aid equations, aidgdpi ,t −1 , is a majordifference in comparison with the previous research of Asiedu et al (2009) It could be addedbecause foreign aid is disbursed gradually with the completeness of projects Together withthe first lag of FDI inflows, infdigdpi , t −1 , this addition makes the system of two equations

balanced, thus the model is referred to as dual-approach dynamics-balanced model.4

Most of the macro and demographic determinants are inherited from Asiedu et al (2009),but some adjustments are made In the FDI equation, the proxy of economic infrastructure is

the total density of communication utilities (utilcom), which is comprised of telephone lines,

4 Some papers using the first lag of FDI as an explanatory variable are Asiedu et al (2009), Busse and Hefeker (2007), Jensen (2003), and Gastanaga, Nugent, and Pashamova (1998).

mobile cellular, and internet subscribers However, the increase of current FDI inflows mightprovoke the usage of communication system Thus, we lag this variable to avoid theinteraction with the dependent variable and its first lag in the right-hand side of the equation.

The trade openness (tradeopen) as measured by the total exports and imports over GDP is

also lagged two periods because there is a direct contribution of FDI to this variable in thesame year As foreign investors expand their business activities to other countries, either inthe form of greenfield or merger and acquisition, they are more likely to export machines,equipment, materials, and experts to, and import final products from the host country Alsothe consumption of expatriates raises the in-border-exports component of the destinationcountry

As depicted in Figure 1, FDI inflows could affect the GDP growth of a country within that year, so gdpgrow is also lagged two times to eliminate the reverse contribution of FDI

inflows into this explanatory variable Finally, the institutions that might affect FDI inflows

(insf ) are regulatory quality, control of corruption, government efficiency, rule of law, and

political stability The impact of these institutional measures will be examined separately inthe regression

In the aid equations, to avoid the component relationship of foreign aid and FDI in the

current GDP per capita, we lag the variable lngdppc Apart from the projected negative

relationship between the income level of a country and the amount of aid that it receives, wealso expect that the decreasing rate does not hold constant, in particular diminishes Hence,

the quadratic component of lngdppc is added The second macro variable which is debt over GDP ( debtgdp) is lagged due to the inclusion of foreign aid to the current sovereign debt.

Moreover, the debt in the past is more appropriate to be a determinant of aid disbursements inthe future

The political and institutional variables of foreign aid (insa) are supposed to be

democracy, control of corruption, and government efficiency As a demographic factor,

population (lnpop) is used because it could create the small-country effect on the inflows of

foreign aid With the same institutions and other things equal, a country with lowerpopulation is more homogenous and more likely to receive the support of foreign community

In comparison with the trade openness used in the FDI equation, a smaller populationrepresents a larger social openness at the country level.5

3.3 Sample and Data

Table 1: Descriptive Statistics

5 Dudley and Montmarquette (1976) discuss in detail the relationship between small country and

foreign aid.

6 The 2013 income classifications of the World Bank are low income (no more than 1,035 US$),

middle income (1,036-12,615 US$), and high income (no less than 12,616 US$).

the sources of these subjective variables Table 1 summarizes the descriptive statistics More details of the variables and the data sources are shown in Table A1 (Appendix A).

4 RESULTS

4.1 Independent Marginal Effects between Institutions, Foreign Aid, and FDI

Table 2 shows the regression results based on the DADB model Panel A contains thebase regression, which follows exactly the specification of the DADB model, for each type offoreign aid—aggregate (AA), bilateral (BA), and multilateral (MA) The three variables ofinterest are FDI inflows, foreign aid, and institutions In the FDI equation, five availablemeasures of institutions—which are rule of law, regulatory quality, control of corruption,government efficiency, and political stability—are examined one by one Each variable doeshave a positive impact on FDI inflows The coefficients as well as their significance levelswithin the full sample are reported in Panel A of Table 3 Except for the governmentperformance, all other institutional and political measures are valuable to attract more FDI.The significance levels are at least 5-percent Surpassing other measures in the category, thequality of regulation in the private sector has the largest impact on the investment decisions

of multinational enterprises (MNEs) A one-point higher in this measure of sound policyresults in 0.66 percentage-of-GDP higher of FDI inflows after controlling for the persistence

of these inflows and other conditions

Since all of the institutional measures have a positive impact on the dependent variable,

we then use the principal component analysis (PCA) to construct a composite institutionalindex which represents the role of government policies in promoting FDI This index isdenoted by INSF and will be used in later FDI-determined regressions Apart from reducingthe workload, another reason behind this composition is that these measures often movetogether A sound government normally has high scores on rule of law, regulatory quality,control of corruption, and government efficiency, and there would be little space for politicalinstability, residential violence, and external conflicts

In column (2) of Table 2, bilateral aid, as expected, has a positive coefficient in the FDIequation This result is similar to Rodrik (1995, p 44) It could be interpreted that when theinflows of bilateral aid is higher by 1 percentage of GDP, the inflows of FDI is approximatelyhigher by 0.18 percentage of GDP, ceteris paribus Interestingly, in column (3), the effect ofmultilateral aid is twice larger than that of bilateral aid with a coefficient of 0.39 Theexplanation for the more profound impact of multilateral aid is that while bilateral aid onlyimproves the inflows of FDI from the countries of bilateral donors, multilateral aid, on theother hand, attracts foreign investors from all countries In short, the institutional effect

outperforms the “extended” vanguard effect To check the robustness of the results by usingsmaller samples, the impacts of institutions and foreign aid on FDI within the low-incomeand middle-income subsamples are reported in the first half of Panel B and C of Table 3 Asobserved, the effect of foreign aid on FDI within the low-income countries is more thandouble within the middle-income countries, but there is no signal of the effect of institutions

on FDI within the former subsample whatsoever There are two possible reasons for thesedifferences First, there might be a virtual institutional threshold under the perception offoreign investors When institutional measures of host countries are below a “thresholdvalue,” then any difference between them would not be considered As a matter of fact, thedescriptive statistics in Table 1 show that, on average, the quality of institutions of low-income countries is lower than that of middle-income counterparts in any aspect Instead, theextended vanguard effect of bilateral aid and the institutional effect of multilateral aid have amore important role in the low-income countries, and it forms the second reason In the need

of low-interest development finance, less developed countries have to comply withregulations and requirements of international organizations Incidentally, these obligationssomewhat affect policies toward foreign private sectors, and in reverse, foreign investorscould have more confidence in these countries This phenomenon is especially obvious withinthe low-income countries, of which the average of foreign aid is much higher than that ofmiddle-income countries

Other control variables in the FDI equation are statistically significant and have theexpected signs When GDP growth or trade openness increases by 1 percentage of GDP, thenet inflows of FDI after two years could be higher by 0.06 or 0.01 percentage of GDP,respectively When the total number of telephone lines, mobile subscribers, and internet users

is higher by 10 units per 100 people, private capital inflows in the next two years could belarger by 0.1 percentage of GDP

In the aid equations, the insignificant coefficients of FDI inflows in the second part ofcolumns (1), (2), and (3) of Table 2 illustrate the non-impact of FDI on foreign aid It meansthat a country with more FDI does not necessarily have more or less of bilateral ormultilateral aid In other words, although FDI could motivate more aid from some specificdonors, the totals of bilateral aid inflows between recipient countries are not ultimatelydifferent due to their total received FDI This result is in line with the statement of Alesinaand Dollar (2000): there is no relationship between FDI and foreign aid

The democracy index, which includes political rights and civil liberties, has an importantrole in the distribution of foreign aid, both bilateral and multilateral The negative coefficientsmean that when a country is appraised to have more freedom, it would have more foreign aidthan others Similarly, we also point out that control of corruption and political stability couldinfluence the aid disbursements from both bilateral and multilateral donors at the aggregatelevel The coefficients and the significance levels of these explanatory variables within thefull sample are attached in Panel A of Table 3 The empirical result that corruption reducesthe amount of foreign aid coming to a country contrasts with the finding of Alesina andWeder (2002) Interestingly, the coefficients of political stability are much larger than those

of control of corruption in the same context Like INSF index for FDI, we also construct acomposite institutional index, which is INSA, to record which of governance indicators couldhelp a country in receiving more foreign aid The impacts of FDI and institutions on foreignaid within the low-income and middle-income subsamples are consolidated in the second half

of Panel B and C of Table 3

Other control variables in the aid equations have statistically significant impact on thedistribution of foreign aid The squared component of the income level implies not only anegative marginal effect but also a diminishing rate of this effect Additionally, the incomelevel is much more sensitive and significant in the multilateral aid equation than in thebilateral aid version This figure proves the target of reducing poverty of multilateralorganizations Finally, a country with more debt or lower population is more open and likely

to look for the help of foreign counterparts and international organizations

4.2 Reliability and Robustness Checks

In this part, we check the reliability and the robustness of the specified model Thereliability check is done by not lagging the control variables of the two equations Besides thesmaller-sample approach mentioned above, the robustness check is undertaken by three moreexperiments First, a control variable is removed from the specification Second experiment isthe insertion of some irrelevant variables Third, different methods of regression—which arePOLS, difference GMM (Arellano & Bond, 1991), and system GMM (Blundell & Bond,

1998)—are applied separately to each of the equations.

Panel C [columns (7)-(9)] of Table 2 reports the results when we do not lag the controlvariables There are several reasons that the lag version is more reliable than the non-lagversion of the specification First, the effect of GDP growth in the non-lag specification isattenuated to 0.05, in comparison with 0.06 in the lag specification This attenuation is caused

by the feedback of the dependent variable The procedure of reverse causality is explained asfollowing GDP growth, which represents the production and consumption capabilities of aneconomy, is one of the most important criteria when MNEs choose a country to invest Thehigher GDP growth of a country, the more opportunities are that the investment could beabsorbed and turned into profits In turn, when foreign investors decide to expand theirbusiness to a higher-growth country, they also contribute to the production of goods andservices within that country As a result, GDP growth is spirally affected by FDI inflows.

To mathematically illustrate the attenuated coefficient of a positively feedback-affectedexplanatory variable x, let denote a self-change of this assumed exogenous variable as ∆x > 0,

and the following marginal effect on the dependent variable y is β real Then, we have

β real = ∆y / ∆x Next, the change ∆y of y enlarges the self-change of x an additional amount

∆x feed ≥ 0 Hence, the total change of x is equal to ∆xnominal = ∆x + ∆x feed The coefficient of x

which we obtain in the multivariate regression of y is

In this case, we have β nominal = 0.05, β real = 0.06, and h = 0.05 / 0.06 ≈ 0.83 It means that,

when taking into account the reverse causality of FDI inflows, the marginal effect of GDPgrowth is downward-biased and equal to 83 percent of its real magnitude

Second, in contrast to GDP growth, the coefficient of trade openness in the non-adjustedspecification is much larger and more significant than that in the adjusted specification This

phenomenon occurs when x or y is the component of each other Suppose that the effect of x

on y is β , thus the effect relationship is y = β x In addition, due to the componential nature between y and x, the component relationship is y = bx The multivariate regression of y on x

negative, in the regression In our case, we have β = 0.01 and β + b = 0.03 Similarly, a small

component relationship could also be found between the amount of debt and the inflows of

15

aid in the aid equations of the non-adjusted specification Foreign aid is a component ofsovereign debt within the same year.

Third, a model or a specification is considered more reliable when its regressioncoefficients are stable despite changes in the proxies of other variables On this criterion, thetwo-lag specification is a little better than the non-lag counterpart The coefficient of lag ofFDI variable in the former is stable at the value 0.56, in spite of the foreign aid proxies,whereas that in the latter fluctuates between 0.54 and 0.55 Fourth, we do not empiricallyprefer the non-lag specification because it is less robust than the two-lag version As seen in

Panel D (Table 2), the coefficient of INSF index becomes insignificant after utilcom is

removed from the model

Next, we turn our attention to the robustness checks First, we remove some control

variables from the model, which are utilcom in the FDI equation and lnpop in the aid equations The results are reported in Panel B, G, and H of Table 2 In the case of utilcom, all

of the control variables which might share its role in the FDI equation—GDP growth, tradeopenness, and the lag of FDI inflows—increase significantly in their impacts In contrast, thecoefficients of our variables of interest—INSF index and foreign aid—are downward biased.The coefficient of INSF index becomes unstable Nevertheless, their impacts on FDI inflowsremain significant, and the impact of multilateral aid still doubles that of bilateral aid There

is not much change in the three aid equations with this omission of utilcom In the case of

lnpop, the absence of this demographic variable just induces the role of control of corruption

and political stability in having more foreign aid

Second, we add some irrelevant or weakly relevant variables into the base specification ofthe DADB model In particular, the FDI equation is added with the money stock M2

(m2gdp), and the aid equations are added with the inflation rate (infc) Young- and

old-dependency ratios of the working-age population are also included in these placebo tests.Panel E and F show the results While the domestic financial development, the inflation, andthe younger dependents are utterly irrelevant, the older dependents reveal some concealed orindirect relevance The old-dependency ratio, to some extent, reflects the human capital, theliving conditions, and the ability to pay debt in the future On one hand, the human capitaland the living conditions are the attractive factors of foreign investment On the other hand,the low ability to work and pay debt of the old population might concern bilateral donors Inregression, the coefficients of old-dependency ratio in the FDI and bilateral aid equations are

In column (14), even though the coefficient of freedom in the bilateral aid equationreduces in the size, from -0.15 to -0.08, compared to the base regression in column (2), itssignificance level also decreases Thus, the role of freedom could be plausible in bilateral aiddisbursements Such situations also occur with INSF index and communication utilities in theFDI equation [Panel F, column (16)].7 For other cases in Panel E and F, the coefficients andtheir significance levels are almost unchanged.

7 The coefficients (t-statistics) of communication utilities are 009 (2.67) in column (1) and 005 (1.38)

in column (16).

Table 2: Independent Marginal Effects between Institutions, Foreign Aid, and

Note: 3SLS estimation, t-statistics in parentheses, *** p<0.01, ** p<0.05, * p<0.1, AA is

aggregate aid, BA is bilateral aid, MA is multilateral aid (percentage of GDP).

Table 2: Independent Marginal Effects between Institutions, Foreign Aid, and FDI,

Note: 3SLS estimation, t-statistics in parentheses, *** p<0.01, ** p<0.05, * p<0.1, AA

is aggregate aid, BA is bilateral aid, MA is multilateral aid (percentage of GDP).

Table 2: Independent Marginal Effects between Institutions, Foreign Aid, and

Observations 1,186 1,186 1,186 1,289 1,289 1,289

Note: 3SLS estimation, t-statistics in parentheses, *** p<0.01, ** p<0.05, * p<0.1, AA is aggregate

aid, BA is bilateral aid, MA is multilateral aid (percentage of GDP).

Table 2: Independent Marginal Effects between Institutions, Foreign Aid, and

Note: 3SLS estimation, t-statistics in parentheses, *** p<0.01, ** p<0.05, * p<0.1, AA is aggregate

aid, BA is bilateral aid, MA is multilateral aid (percentage of GDP).

Table 3: The Impacts of Different Institutional Measures on FDI and Foreign Aid.

equation The coefficients of foreign aid and net FDI inflows are estimated when using INSF index in the FDI equation and INSA index in the aid

As the last robustness check, we run regressions for the dynamic FDI and foreign aidequations separately using different methods Dynamic panel-data models implied whenusing GMM methods take into account unobserved individual-level effects In particular, wehave the model of FDI:

In this case, {x1it , x2it } are vectors of strictly exogenous variables, {w 1it , w2it } are vectors

of endogenous variables and predetermined variables, {v 1i ,v 2i } are unobserved country-level

effects, and {e 1it ,e 2it } are idiosyncratic errors Natural resources, specific alliances with somecountries, and colonial relationship could be entailed in the fixed effects of the FDI equationand the bilateral aid equation

Table 4 and 5 uncover the details The results generated by POLS (Panel As) and systemGMM (Panel Cs) are quite similar to each other and to the previous base regressions, whilethose generated by difference GMM (Panel Bs) are extremely odd, especially thepredetermined components In any case, either difference GMM for the FDI equation (Panel

B, Table 4) or difference GMM for the aid equations (Panel B, Table 5), some tests ofoveridentifying restrictions or exogeneity of instrument subsets are strongly rejected.8Consequently, these difference-GMM estimates are invalid

We apply an alternative transformation which is forward orthogonal deviations (FOD)(Arellano & Bover, 1995) to see whether the results of difference GMM could be better This

8 The results of these tests are not reported in the tables herein and could be sent upon requests.