A Generalized Empirical Model of Corruption, Foreign Direct Investment, and Growth

Michael S. Delgado Department of Economics Binghamton University Subal C. Kumbhakar Department of Economics Binghamton University

Nadine McCloud Department of Economics University of the West Indies at Mona March 27, 2011

Abstract We propose a generalized empirical model for estimating the eect of foreign direct investment on GDP growth rates, as well as for determining the eect of corruption on the growth rate, and on the relationship between foreign direct investment and growth. Our model allows for parameter heterogeneity between all conditioning variables (including foreign direct investment) and growth, as well as in the eects of corruption on growth. We estimate the regression using a recently developed nonparametric method of moments estimator that allows us to concurrently use instrumental variables to mitigate any endogeneity bias that may be present in the relationship between foreign direct investment and growth, and model parameter heterogeneity. We nd that there is substantial heterogeneity in the relationship between foreign direct investment and growth, and that foreign direct investment has a positive and signicant eect on growth for many of the countries in our sample. Corruption is shown to signicantly diminish the eectiveness of foreign direct investment at improving growth rates, but overall has an insignicant net eect on growth. Keywords: Foreign direct investment; corruption; parameter heterogeneity; economic growth; nonparametric method of moments; instrumental variables.

Michael S. Delgado, Department of Economics, State University of New York at Binghamton, PO Box 6000, Binghamton, NY 13902-6000. Email: mdelgad1@binghamton.edu Corresponding author: Subal C. Kumbhakar, Department of Economics, State University of New York at Binghamton, PO Box 6000, Binghamton, NY 13902-6000. Phone: 607-777-4762. Fax: 607-777-2681. Email: kkar@binghamton.edu Nadine McCloud, Department of Economics, University of the West Indies at Mona, Kingston 7, Jamaica. Email: nadine.mccloud02@uwimona.edu.jm

Introduction

Foreign direct investment (FDI) is generally thought to be an important factor of growth and development in developing countries. It is through the investments of large multinational corporations that developing countries have access to advanced technologies, management practices, and research and development that are crucial for growth, but are otherwise unavailable in the developing world (e.g., Borensztein et al. 1998 and Carkovic and Levine 2005). Unfortunately, while there has been a broad consensus as to the theoretical importance of FDI for growth and development in developing countries, there has yet to be a consensus among empirical researchers as to the signicance of FDI at increasing growth rates. Blomstrom (1986), Borensztein et al. (1998) and Alfaro et al. (2004) all nd evidence that FDI positively contributes to economic growth, whereas Haddad and Harrison (1993), Aitken and Harrison (1999), and Carkovic and Levine (2005) nd no evidence in support of growth-enhancing eects of FDI. These conicting empirical results among studies on the FDI-growth relationship may be due to the failure to appropriately incorporate parameter heterogeneities, which can lead to a misspecied model and inaccurate estimation of the relationships of interest. Durlauf (2001), for example, advocates modeling all the parameters in growth regressions as functions of developmental variables, rather than as constants. Constant parameter growth models may suer from misspecication since they ignore crucial heterogeneities induced by the developmental variables that are fundamental to the growth process. Moreover, constant-parameter models will most likely be sensitive to dierent specications of functional forms, or samples of observations. Growth models that allow for constant parameters provide a description of the average relation, at best. In the presence of substantial heterogeneity in the growth process, constant-parameter models are unlikely to accurately estimate the relationship between FDI and growth. One important developmental element that is likely correlated with the absorptive capabilities of host countries and ultimately inuences the eectiveness of FDI at improving growth rates is institutional quality. While there are dierent measures of institutional quality that may result in heterogeneity in the eect of FDI on growth across developing countries, corruption may be of extra importance because of its eect on many avenues that all ultimately inuence absorptive capabilities and growth rates. Mauro (1998), Gupta et al. (2002) and Tanzi et al. (2002) all document a negative relationship between corruption and human capital. Countries that are more corrupt tend to invest less in human capital, which ultimately decreases the ability of the country to absorb new technologies from developed nations (Borensztein et al. 1998). Bribery, for example, which is associated with higher levels of corruption, may lead to an imbalance in the relative payos between productive and unproductive sectors in the economy (Baumol 1990 and Murphy et al. 1991). Workers are less likely to move to domestic from foreign rms (i.e., the multinational corporations) where their payos are relatively higher; the result is less diusion of technology from the domestic rms to foreign rms, and a weakening of the eect of FDI on growth. Building on previous studies that have identied heterogeneity within the relationship between FDI and GDP growth (e.g., Borensztein et al. 1998 and Alfaro et al. 2004), as well as studies that have shown an important interaction between institutional factors (e.g., corruption) and the eectiveness of FDI at improving growth rates (e.g., McCloud and Kumbhakar 2

2011), we present a generalized empirical growth model with which to re-analyze the relationship between corruption, FDI, and GDP growth (as well as the relationship between growth rates and other conditioning variables). Our generalization is based on a standard growth regression that assumes homogeneous parameters. We generalize the standard model to allow for a heterogeneous relationship between GDP growth and all conditioning variables, by making the coecients unknown smooth-functions of an index of corruption, and country- and time-specic indicators. Thus, our model allows us to obtain estimates that are specic to each country in each year, and estimates that depend on the level of corruption in each country and in each year. While we present an alternative approach that complements previous studies that have allowed for heterogeneity within the corruption-FDI-growth relationship, our approach also allows us to analyze the eect of corruption on GDP growth rates through its eect on all conditioning variables (e.g., trade openness, or ination), and not solely through its inuence on the FDI-growth relationship. To estimate our generalized regression model, we use a recently developed nonparametric version of a standard method of moments estimator (Cai and Li 2008) that assumes the primary conditioning variables (e.g., FDI) enter linearly into the regression model, but allows the intercept and slope coecients to vary nonparametrically (i.e., either linearly or nonlinearly) with respect to certain environmental factors (e.g., corruption). Hence, this model is a version of the varying coecient models of Hastie and Tibshirani (1993), or more recently Li et al. (2002). One advantage of this nonparametric generalized methods of moments (NPGMM) estimator over other smooth coecient models, e.g., Durlauf et al. (2001) and Li et al. (2002), is that it allows all of the conditioning variables to be endogenous. This, in part, addresses one concern raised by Durlauf (2001), who argues that in a growth specication, all conditioning variables can be taken to be endogenous; that is, all variables typically included in a growth specication are determined, in part, by omitted factors that also determine growth rates. In particular, Borensztein et al. (1998) provide a discussion of the potential endogeneity of FDI in a growth regression. Hence, it is important to consider an instrumental-variables approach to estimating the relationship between FDI (or any other conditioning variables of interest) and growth in order to obtain consistent estimates.1 Although the generalized model diers from standard homogeneous models by incorporating parameter heterogeneity in the coecients, the generalized model maintains the traditional functional form assumptions embedded in the standard models. The advantage of maintaining such assumptions (e.g., additive separability and linearity of the regressors), is that the standard model exists as a special case of the generalized model. We can econometrically test whether the data support the assumption of parameter homogeneity inherent in the standard model. Another advantage of maintaining such functional form assumptions is that we can avoid dimensionality issues that often arise in fully-specied nonparametric models. In the present context, dimensionality issues can only arise when estimating the coecient functions in the smooth coecient model; because the number of continuous environmental factors is likely to
In the empirical growth literature, Liu and Stengos (1999) and Durlauf et al. (2001) also use semiparametric models to examine parameter heterogeneity. Their works dier from that of the present paper in many ways including the use of cross-sectional and not panel data, exclusion of FDI from the set of independent variables, sample selection (inclusion of OECD and non-OECD countries), assumed sources(s) of parameter heterogeneity, and analysis of endogeneity.
1

be relatively small, or at least smaller than the entire conditioning set in a fully nonparametric regression, the curse of dimensionality can often be avoided. Our results conrm that there exists substantial heterogeneity in the relationship between FDI and growth, and we nd strong evidence that FDI has a positive and signicant inuence on growth rates for about 80 percent of the developing countries in our sample. In addition to providing observation-specic estimates of the coecients (e.g., the FDI coecient), our model also provides estimates of the marginal eect of corruption on each of the coecients. Our estimates show that corruption signicantly reduces the eectiveness of FDI on growth, which supports previous studies that suggest that corruption inuences the absorptive capabilities of developing countries. However, when considering the total eect of corruption on growth rates (i.e., the sum of the indirect eects of corruption on all of the coecients in the model), we nd that corruption does not signicantly inuence growth rates. Through our heterogeneous parameter estimates, we analyze separate groups of countries that have substantially dierent coecients and isolate characteristics common within such groups. This type of analysis is useful for international investment policies: knowledge of whether there is a positive and signicant relationship between FDI and growth rates for any particular country, or how this relationship varies with respect to corruption is crucial for designing policies aimed at improving growth rates. While we do not nd evidence of regional or geographical groups, we nd that with respect to heterogeneity within the FDI-growth relation, many countries with the highest returns to FDI also have the lowest returns to corruption. Conversely, countries with an insignicant or relatively low correlation between FDI and growth have the highest estimated returns to corruption. Hence, our results suggest that developing countries with relatively low correlations between FDI and growth may benet substantially from a reduction in corruption. Our empirical results are robust to using dierent instruments for FDI, allowing for all conditioning variables to be endogenous, using dierent measures of corruption, controlling for other measures of institutional quality that may be correlated with the dependent and independent variables. Moreover, a specication test suggests our semiparametric model that allows for endogeneity is more consistent with the data than the standard-homogeneous model. The structure of the rest of the paper is as follows. Section 2 presents and discusses our generalized empirical growth model and its nonparametric method of moments estimator. Section 3 discusses the data. Section 4 provides the main empirical results and discusses their implied policy prescriptions. Section 5 investigates the robustness of our main results. Section 6 concludes. The excluded empirical results can be furnished on request.

2
2.1

Empirical Methodology
Growth Models

We consider a standard growth model with the growth rate of real GDP per capita as the dependent variable and a set of control variables. Letting git denote the real GDP per capita growth rate in country i at time t, we write the standard model as:

 git = 0 + Yit 1 + Xit 2 + it ,

i = 1, . . . , n

t = 1, . . . , T,

(1)

in which Yit is our measure of FDI, Xit is a vector of control variables, (0 , 1 , 2 ) is a

vector of parameters to be estimated, and it is a zero-mean random error. The advantage of (1) is that, under certain regularity assumptions, it is easy to consistently estimate the eect of FDI on growth using a least squares criterion. One primary drawback of this model, however, is that it fails to incorporate parameter heterogeneity that more likely exists in a cross-country panel of observations. In particular, model (1) does not allow the eect of FDI on growth, 1 , to vary with respect to the index of corruption. Allowing the parameter estimates to vary with respect to corruption is a pragmatic way to identify the indirect eect of corruption on growth. An alternative way to incorporate the level of corruption into the growth regression would be to add the corruption index as another conditioning (i.e., X) variable, but this approach does not identify indirect channels through which corruption inuences growth. Corrupt governments (or ocials) are more likely to embezzle funds and redirect public expenditures towards personal and private ventures, rather than direct them towards more publicly benecial avenues. The eect is that FDI and other correlates of economic growth may be directly inuenced by the level of corruption. Through these channels, corruption may indirectly inuence GDP growth rates. However, it is important to incorporate the direct eects of corruption in the regression model to obtain an accurate picture of the eect of corruption on growth rates and on the relationship between the conditioning variables (e.g., FDI) and growth, and results that are comparable to those of existing growth studies. Since our interest is on the estimation of the eect of FDI on growth, and how this eect varies with respect to the level of corruption, we generalize the model to incorporate the eect of corruption on the -parameters. Specically, we generalize (1) by allowing the -parameters in the model to vary with respect to a particular set of environmental variables, Zit , which contains the index of corruption. Hence, we write our generalized model as:
git = 0 (Zit ) + Yit 1 (Zit ) + Xit 2 (Zit ) + it ,

i = 1, . . . , n t = 1, . . . , T.

(2)

An advantage of using the generalized model in (2) is that it provides observation-specic estimates of the coecients of the model thereby allowing us to analyze the heterogeneity in the eect of FDI (and other control variables) on growth rates. As previously argued, an accurate modeling of parameter heterogeneity is crucial for designing cross-country policies to increase growth rates in developing countries; if the eect of FDI on growth rates varies substantially across dierent countries (or dierent groups or types of countries), investment policies governing FDI should be tailored to each specic country (or group of countries). The policy prescriptions for boosting growth through FDI, which are implied by homogeneous parameter estimated, may be too passive or active for some developing countries. If we assume that corruption is orthogonal to the error term, then it is straightforward to extract the direct and indirect eects of corruption on growth. The direct eect of corruption on growth comes through the eect of corruption on the intercept function, 0 (); we can obtain

an estimate of this direct eect through the partial derivative of the intercept function with respect to corruption at a particular point, z: 0 /z. The indirect eects come through the eect of corruption on the other coecient functions in the model, j=0 (); we obtain these eects through the partial derivatives of each of the slope coecient functions with respect to corruption, j=0 /z. The total eect of corruption on GDP growth rates is the sum of the direct and indirect eects of corruption on growth. Hence, taking a partial derivative of the growth rate in (2) with respect to corruption at a particular point yields the total eect of corruption on growth: git 0 1 2 = + Yit + Xit . z z z z (3)

2.2

Estimation

To exploit the generality of our model in (2), we assume that the coecient functions are unknown smooth functions of Z. For ease of exposition, we rewrite (2) more compactly as:
git = Xit (Zit ) + it ,

i = 1, . . . , n t = 1, . . . , T,

(4)

in which Xit is a vector of dimension k with the rst column containing a one and the remaining columns containing the (k 1) regressors (including FDI); () is a vector of smooth coecient functions of unknown form; Zit is a vector of dimension p containing environmental factors that are assumed to be the sources of parameter heterogeneity. If we assume also that all regressors in X are exogenous then (4) is a standard semiparametric smooth coecient model that can be consistently estimated using the nonparametric kernel estimator proposed by Li et al. (2002). This exogeneity assumption seems strong in the present growth application, hence we allow the variables in X to be endogenous. This key endogeneity assumption distinguishes our model in (4) from other semiparametric smooth coecient models. In the special case where all regressors in X are exogenous, our model is equivalent to a standard semiparametric smooth coecient model.
If any element in X is endogenous, then E[git |Xit , Zit ] = Xit (Zit ) and estimation using

typical semiparametric estimators (e.g., Li et al. 2002) will provide inconsistent estimates of the unknown coecient functions. Several nonparametric estimators have been proposed to deal with the problem of endogeneity in smooth coecient models, for example, Das (2005), Cai et al. (2006), and Cai and Li (2008). Both the estimators in Das (2005) and Cai et al. (2006) are two-step estimators that require nonparametric estimation of the endogenous variables on the instruments and exogenous variables in the rst step followed by semiparametric regression of the dependent variable on the rst stage estimates of the endogenous variables. Cai and Li (2008), however, propose a one-step estimator of (4) when X is allowed to be endogenous. We apply this one-step Cai and Li (2008) estimator to reap the gains in eciency that the one-step estimator likely has over the two-step estimators. We note that the Cai and Li (2008) framework allows for all X variables to be endogenous, and explicitly assumes that the environmental variables in Z are exogenous. To circumvent the endogeneity problem and obtain consistent estimates of the coecient

functions, Cai and Li (2008) propose the following conditional moment restriction:
E[Q(it )it |it ] = E[Q(it ){git Xit (Zit )}|it ] = 0,

(5)

in which it = (Wit , Zit ) , Wit is a vector of instrumental variables such that E[it |Wit ] = 0,

and Q(it ) is some vector function such that the conditional moment restriction in equation (5) is satised. While in principle any vector for Q(it ) that satises the conditional moment restriction in (5) can be used, Cai and Li (2008) suggest using Q(it ) =
Wit Wit (Zit z)/h

, where

h is a smoothing parameter, and is the Kronecker product operator, to make use of the instrumental variables in Wit . Cai and Li (2008) suggest estimating the coecients, (Zit ), with nonparametric kernel methods which, combined with the conditional moment restriction in (5), yields a nonparametric equivalent of a GMM estimator (or NPGMM). We assume the coecients, (Zit ), are twice continuously dierentiable, so that we can apply local-linear least-squares to estimate the unknown functions. A rst order Taylor expansion
around a given point z yields an approximation to the function j (Zit ) given by j (z)+j (Zit z),

in which j is a gradient vector of the partial eects j (z)/z. Thus the local-linear procedure provides a vector of estimated coecient functions, j (Zit ), along with their rst order gradient vectors, j (z)/z. Letting Uit =
Xit Xit (Zit z) and = (j (z), j ) be the vector of coecients

and their rst order partial derivatives, the conditional moment restriction in equation (5) gives rise to the following locally weighted orthogonality condition:
n T Q(it )(git Uit )Kh (Zit z) = 0 i=1 t=1

(6)

in which Kh (Zit z) is a generalized product kernel of dimension p that admits a mix of continuous and discrete environmental factors contained in Z (see Racine and Li 2004), and h denotes a vector of smoothing parameters. Cai and Li (2008) show that a consistent estimate of can be obtained from:
= (Sn Sn )1 (Sn Tn ),

(7)

in which 1 Sn = n and 1 Tn = n

T Q(it )Uit Kh (Zit z)

(8)

i=1 t=1 n T

Q(it )Kh (Zit z)git .

i=1 t=1

(9)

To avoid any pitfalls associated with an ad hoc choice of smoothing parameters, we use least-squares cross-validation to select the parameters in h. The least-squares cross-validation criterion function is given by
nT

min
h i=1

(gi gi )2

(10)

in which gi is the leave-one-out estimate of the conditional mean, X (Z). All standard errors

are estimated using a wild-bootstrap.

3
3.1

Data
Overview

The data set comes from McCloud and Kumbhakar (2011). It consists of a balanced panel of 60 non-OECD countries spanning the period 1985-2002 giving a total of 1080 observations. Our primary interest is in estimation of the eect of FDI on GDP growth and how this eect varies with respect to the level of corruption in each country in each year. Our secondary interest is in identifying the overall eect of corruption on GDP growth, both directly through its inuence on the intercept term and indirectly through its role in the eects of other control variables on growth. Our measure of GDP growth is the per capita GDP growth rate that comes from the Penn World Table (version 6.2).

3.2

FDI

Our measure of FDI is the percentage of FDI inow relative to GDP in constant 2002 dollars, which comes from the United Nations Conference on Trade and Development online statistical database. It is generally believed that FDI may be correlated with any factors that inuence growth rates but are omitted from the regression model; that is, FDI may be endogenous in a growth specication such as (1). The empirical FDI literature has been unable to identify an ideal instrumental variable to completely control for any endogeneity bias. Several studies have proposed several dierent instrumental variables that have been shown to mitigate, at least part of, the endogeneity of FDI. Borensztein et al. (1998), for example, suggest using lagged values of FDI or measures of institutional quality. Carkovic and Levine (2005) suggest using lagged FDI as well as lagged dierences of FDI as instrumental variables. We nd that in our data set, lagged values of FDI work reasonably well and appear to mitigate, at least part of, the endogeneity of FDI; see section 4.1. Measures of institutional quality (i.e., ethnolinguistic fractionalization and latitude from La Porta et al. (2009), and the log of the life expectancy and log of the fertility rate from the 2005 World Development Indicators) and lagged dierences of FDI appear irrelevant based on their low explanatory power in the rst-stage parametric regressions. Moreover, semiparametric regressions using these latter instrumental variables did not yield meaningful estimates or appear to mitigate any endogeneity bias. In addition to the above instrumental variables proposed in previous studies, we propose using total world FDI ows and total FDI ows to developing countries as alternative instrumental variables. Currently, we are unaware of other studies that use these total FDI ows as instrumental variables for individual country FDI ows. Our rationale for using these variables is that measures of total FDI ows will cause uctuations in individual country FDI ows, but are uncorrelated with growth rates of individual countries. Parametric rst-stage regressions suggest that total world FDI ows and total FDI ows to developing nations may be reasonable alternative instrumental variables for FDI. The sample correlations between FDI (the endogenous variable of interest) and each of the instrumental variables are 0.74, 0.32, and 0.13, for lagged FDI, world FDI ows, and developing world FDI ows, respectively. We use lagged FDI 8

as our preferred instrumental variable since it provides the strongest rst-stage correlations with FDI (as well as strongest sample correlation), but we consider both measures of total FDI ows as alternative instrumental variables in our sensitivity analysis.

3.3

Corruption

We combine two indices of corruption that are widely used in the existing empirical literature. One index of corruption is from Knack and Philip (1998), which is for the period 1984 to 1996. This index ranges from 0 to 6 and lower scores indicate lower levels of corruption in that high government ocials are likely to demand special payments and illegal payments are generally expected throughout lower levels of government in the form of bribes connected with import and export licenses, exchange controls, tax assessment, police protection, or loans. The other corruption index is from Transparency International (TI) for the period 1997 to 2002.2 The TIs corruption index measures the overall extent of corruption and therefore does not distinguish between administrative and political corruption, nor between petty and grand corruption. It ranges from 0 to 10 with higher values indicating lower levels of corruption. For ease of exposition, we rescale the TI index so that lower values represent lower levels of corruption. An important assumption in this analysis is that our proxies for corruption are time invariant. Our rationale is based on the fact that the extent to which corruption is entrenched in many non-OECD countries makes it dicult for these countries to lower their corruption levels in the absence of proper legal recourse through institutional reform. Consequently, we transform the Knack and Philip index to be within the range of 0 to 10 and then construct an aggregated corruption index by using the time average of the Knack and Philip and TI indices as the measure of corruption for the entire time span. We note that the Spearmans rank correlation coecient for the average PRS and TI indices is 0.6151 with a p-value of 0 for the null hypothesis of independence. Hence, combining the dierent measures of corruption from these two sources should not bias the qualitative implications about the eect of corruption on FDI-growth relation.

3.4

Additional Control Variables

We use the following list of covariates to control primarily for any omitted variables bias between GDP growth and FDI, but also to serve as possible channels through which corruption may eect growth. The variables include initial GDP per capita dened as GDP per capita in the previous year; openness, dened as the ratio of exports plus imports as a percentage of GDP; government consumption, dened as the ratio of general government consumption as a percentage of GDP; domestic investment as a percentage of GDP; the US treasury bill rate; and the ination rate. All variables come from the Penn World Table (version 6.2) except for the US treasury bill rate which comes from the IMF International Financial Statistics Database. We include the US treasury bill rate to control for changes in the growth rate that are caused by macroeconomic conditions that are exogenous to each individual country.
The TI index was rst launched in 1995 with only a small number of countries. Using earlier years of this index would have reduced our eective sample size.
2

In addition to corruption, the vector of environmental variables, Z, also contains an unordered categorical country indicator and an ordered categorical indicator for year to control for country and year xed eects, respectively. Alternative specications include the fertility rate (total births per woman) and an index of democracy as additional environmental variables. Our index of democracy comes from the Polity IV database and ranges from -10 to +10 with +10 representing complete democracy and -10 complete autocracy. With the exception of the variables already measured in percentage terms or growth rates, all continuous variables are measured in logs.

4
4.1

Results
Ordinary Least Squares

We rst estimate the standard homogeneous model in (1) using ordinary least squares. Since the coecients do not vary in this model, we include corruption and country and time dummy variables as standard conditioning (i.e., X) variables. The purpose of estimating the homogeneous model is to provide estimates that are directly comparable to other studies that do not use semiparametric estimators, and to anchor our semiparametric results to the standard case. Table 1 contains the results from the dierent model specications. The rst three columns in Table 1 show estimates from three standard models: the rst column reports the results from a parsimonious model in which the only regressors are FDI and xed eect dummy variables; the second column adds corruption; and the third column adds the rest of the conditioning set. We nd that FDI has a positive and signicant eect on growth rates in columns 1 and 2; that is, including corruption in the regression does not erode the eect of FDI on growth. In particular, an increase of 10 percent in the FDI inows to GDP is associated with an increase of 3 percent in economic growth rate. Interestingly, the coecient on corruption is positive and signicant. A positive coecient on corruption implies that holding everything else constant, increasing the level of corruption in a country will increase its rate of real GDP per capita growth. We nd this result to be counter-intuitive; our prior expectation is that corruption has a negative (or perhaps insignicant) eect on GDP growth rates. Moreover, the R2 does not improve substantially after including corruption into the regression, which suggests that corruption may not contain much predictive power. Including the rest of the conditioning set (column 3) does not change the eect of corruption on growth, but it does erode the eect of FDI on growth by approximately 67 percent. This large reduction in the estimated FDI coecient suggests that the FDI-growth eects in the columns 1 and 2 may be driven by omitted variables bias. Indeed, there may be other factors that are subsumed in the errors and are correlated with both FDI and economic growth. To explore this possibility, we use instrumental variables methods. Columns 4 through 6 report estimates from two-stage least squares regressions that use the one-period lagged value of FDI to control for any possible endogeneity of FDI in the standard models. Specically, we include all exogenous variables including corruption in the rst stage regression, as well as country and time dummy variables to control for country and time xed eects. We nd that for each rst-stage specication, the FDI instrument is positive and sta10

tistically signicant, and the F-statistic for the null hypothesis of no regression exceeds 10, suggesting strength of the instrumental variable (Staiger and Stock, 1997). We nd that in the second-stage regressions, including the fully specied model in column 6, the instrumented FDI variable has a positive and signicant eect on growth rates. In addition, the magnitude of the FDI coecient is substantially larger than in the simple ordinary least squares models. This suggests that there is a downward bias in the ordinary least squares estimates of FDI, most likely caused by endogeneity; the lagged FDI instrument is able to correct for (at least part of) the downward bias on the FDI coecient. Corruption remains positive and signicant, and many other conditioning variables maintain their sign and signicance observed in their ordinary least squares counterparts. The explanatory power in the two-stage least squares models is comparable to that from the ordinary least squares models. The aforementioned ordinary least squares and two-stage least squares homogeneous estimates yield three important observations: One, on average, FDI has a positive and signicant eect on GDP growth rates in non-OECD countries, but the eect is subject to a downward endogeneity bias that will potentially mask the signicance of FDI. Two, the use of lagged FDI as an instrument for FDI is able to mitigate this downward bias and provides more precise estimates of the eect of FDI on growth. This positive mean eect of FDI on growth, as implied by the homogeneous models, should not be taken to imply that in all countries, FDI has a positive and signicant eect on growth rates. Three, corruption appear to have strongly positive and signicant, albeit counter-intuitive, eect on growth rates. This latter counterintutive result may be a manifestation of model misspecication, at least with regards to the way in which corruption is included in the model. A maintained hypothesis is this paper is that corruption itself is an environmental variable and thus it should be included as such in the regression, and not included as a typical conditioning (i.e., X) regressor. In light of these observations, we move on to the results from our generalized semiparametric specications that concurrently allow for (a) corruption to enter into the model indirectly through its inuence on the relationship between the conditioning variables and GDP growth rates, (b) parameter heterogeneity and (c) endogenous regressors.

4.2

Semiparametric Smooth Coecient Models

Columns 7 through 10 in Table 1 report the mean coecients and standard errors for four semiparametric smooth coecient specications. The rst two specications (columns 7 and 8) do not control for endogeneity of FDI; hence, these models are standard semiparametric smooth coecient models (SPSCM). Columns 9 and 10 control for endogeneity of FDI using lagged FDI and the Cai and Li (2008) NPGMM estimator. Columns 7 and 9 include only FDI as a conditioning variable, whereas columns 8 and 10 include all other conditioning variables. Each of the four models include corruption and indicators for country and year as environmental variables. 4.2.1 Mean Parameter Estimates

We nd that the mean coecient on FDI is positive and statistically signicant in all semiparametric specications. In the semiparametric models that do not allow for instrumental variables 11

(columns 7 and 8), we observe FDI coecients that are similar in magnitude to their ordinary least squares estimates in columns 1 and 2. After controlling for endogeneity, we observe that the mean FDI coecients in the semiparametric models (columns 9 and 10) are close in magnitude to their counterparts from the two-stage least squares models. In essence, the FDI coecients are substantially larger in magnitude in the semiparametric models that control for endogeneity, suggesting that the endogeneity of FDI biased its coecients downward bias. Thus, we again nd evidence that supports the validity of using lagged FDI values to correct for any endogeneity bias associated with FDI. Turning to the additional conditioning variables in the fully-specied models (columns 8 and 10), we nd that regardless of whether we instrument for FDI, the mean estimates of initial income, openness and the ination rate are negative and highly statistically signicant, whereas those associated with domestic investment are positive and highly statistically signicant. The mean estimates of all other conditioning variables are insignicant. Turning to the total eect of corruption on growth, g/Z, we nd in each of the four semiparametric models that corruption has an insignicant eect on GDP growth rates. This result is more in line with our prior expectations, and is in contrast to the results from the ordinary least squares and two-stage least squares estimates that found corruption to have a positive and signicant eect on growth rates. Table 2 shows the means and standard errors of the direct eect of corruption on the coecients in the semiparametric models. At the mean, we nd that corruption has a negative and signicant eect on the coecient on FDI (columns 2 to 4). Thus, coupling this result with the positive and signicant FDI coecients in Table 1 (columns 8 to 10) implies that an increase in corruption will decrease the eectiveness of FDI on growth, and through this channel, holding everything else constant, decrease GDP growth. From Table 2, we also nd that corruption has a positive and signicant eect on the coecient on openness (column 2) and on ination (column 4). Since both the openness and ination coecients are negative and signicant (see columns 8 and 10 of Table 1), these positive partial derivatives imply that the eect of openness and ination on growth is dampened by an increase in corruption. In addition, we nd that in each of the four semiparametric models, the R2 is substantially higher than in the corresponding homogeneous regressions. These higher R2 values suggest that there are sizeable parameter heterogeneities across countries and modeling these parameter heterogeneities substantially improves the t of the model to the data. We now turn to the distribution of the coecients and their partial eects with respect to corruption to further assess the degree of parameter heterogeneity that exists in the estimates, and to understand the policy implications of heterogeneity in the relationship between FDI and growth. 4.2.2 Heterogeneous Parameter Estimates

To present the distribution of coecients in a concise manner, we use 45-degree gradient plots to simultaneously show the magnitude, sign, standard errors, statistical signicance and density of the coecients. To construct these plots, we rst plot the observation-specic coecients on the 45 degree line. The location of any particular coecient to the horizontal axis determines the sign and magnitude of the coecient, whereas the density can be seen by the proximity

12

of surrounding observations to any particular observation. Areas with a high concentration of coecients are areas of higher density. We then calculate observation-specic condence bounds by adding (and subtracting) twice the observation-specic standard error from each coecient. We then overlay the condence bounds above (and below) the scatterplot of coecients. This allows us to assess whether each observation is statistically signicant; if the horizontal line at zero runs between the coecient and its upper or lower condence bound, the observation has a statistically insignicant coecient. If the horizontal line at zero does not intersect the condence bound for a particular observation, that observation is statistically signicant. Figure 1 displays the 45-degree gradient plots for the distribution of observation-specic FDI coecients and standard errors for each of the four semiparametric models that are described in Table 1. In each model, many of the observations are positive. Specically, for the standard semiparametric models that do not use instrumental variables (SPSCM models), 56 percent and 65 percent of the FDI coecients are positive and signicant, respectively SPSCM1 and SPSCM2. In the semiparametric models that use lagged FDI as an instrumental variable (NPGMM models), the FDI coecients are positive and signicant for 59 percent and 76 percent of the observations, respectively NPGMM1 and NPGMM2. While the plots show a substantial amount of heterogeneity in the parameter estimates, it is clear that, on average, FDI has a positive and signicant eect on GDP growth rates. Moreover, after instrumenting for any endogeneity bias on the coecients, we nd that the fraction of coecients that are positive and signicant increases. These results provide evidence that the lagged value of FDI is also able to mitigate (at least part of) the downward bias on the FDI coecients in the semiparametric models. In reference to the additional conditioning variables in the fully-specied models that have statistically signicant mean estimates, we nd that 59 percent of the initial income coecients are negative and signicant in the model that does not instrument for FDI, whereas 72 percent of the coecients are negative and signicant in the instrumental variables model. We nd that 46 percent of the coecients on openness are negative and signicant in the non-instrumental variables model, and 75 percent of the coecients are negative and signicant in the instrumental variables model. In the non-instrumental variables model we nd that 94 percent of the coecients on domestic investment are positive and signicant, and 94 percent of the coecients on ination are negative and signicant. In the instrumental variables model, we nd that 95 percent of the coecients on domestic investment are positive and signicant, and 91 percent of the coecients on ination are negative and signicant. These results suggest that endogeneity of FDI induces biases in the estimates associated with the other regression coecients, and the size and direction of these biases appear to dier across countries and regressors. Figure 2 contains 45 degree plots for the partial eects of the FDI coecients with respect to corruption. For the standard semiparametric models, 42 percent and 74 percent of the FDI coecients show a signicantly negative partial eect with respect to corruption, respectively SPSCM1 and SPSCM2. In the semiparametric instrumental variables models, 45 percent and 84 percent of the FDI coecients show a negative and signicant partial eect with respect to corruption, respectively NPGMM1 and NPGMM2. These results, especially in the fully specied models, lend additional support to the view that corruption decreases the eectiveness of FDI on GDP growth.

13

The eect of corruption on the additional coecients in the fully-specied models is insignificant, except for the coecients associated with openness and ination. In the non-instrumental variables model, we nd that corruption signicantly reduces the eectiveness of openness on growth rates for 58 percent of the observations. In the instrumental variables model, the eect of corruption on openness ceases to be statistically signicant; however, we nd that 56 percent of the observations have a positive and signicant partial eect of the ination coecient with respect to corruption. In general, these results suggest that corruption strongly aects the FDI coecient, and has a neutral eect on the coecients of (many of) the other conditioning variables in the model. For the fully-specied models, 45-degree gradient plots for the additional conditioning variables are available upon request from the authors. Figure 3 shows the 45 degree plots for the total eect of corruption on growth rates. In the rst two semiparametric models (that do not instrument for endogeneity), we nd 64 percent and 70 percent of the partial eects to be insignicant, respectively SPSCM1 and SPSCM2. In the semiparametric instrumental variables models, we nd 66 percent and 72 percent of the partial eects to be insignicant, respectively NPGMM1 and NPGMM2. Hence, we nd strong evidence that although corruption decreases the eectiveness of FDI on growth rates, corruption has an overall insignicant eect on growth. With regards to the estimated bandwidths, we nd in each of our four semiparametric models that the bandwidth on corruption exceeds several standard deviations of the data. In the local-linear least-squares context, a relatively large bandwidth on corruption implies that corruption enters linearly into the parameters, j (). Local-linear least-squares is nothing more than weighted least-squares through the kernel function providing a local weight for each observation; a large bandwidth means that the local neighborhood includes all observations, and hence a globally linear estimate with respect to corruption.3 We leave further analysis of any potential linearity between corruption, FDI, and GDP growth for future research. Our empirical results so far are in favor of parameter heterogeneity. To formally test whether the semiparametric models yielding heterogeneous parameter estimates are indeed preferred to the homogeneous parameter models, we use the model specication test of Cai et al. (2000). The Cai et al. (2000) test allows us to determine whether the data support the null hypothesis of the simple two-stage least-squares model. We test the null hypothesis against two alternative hypotheses: the two primary semiparametric smooth coecient models that include the entire conditioning set of regressors (one using instrumental variables and the other without). We are able to reject the null hypothesis of correct specication for the constant parameter model in both tests with a p-value of 0.0000. Hence, the data support our generalized semiparametric specication that admits parameter heterogeneity as a function of corruption over the homogeneous parameter specication.
3 This does not imply any particular parametric functional form for the coecients, j (), since a large bandwidth on corruption does not necessarily point towards any specications regarding interactions between corruption and other environmental variables, or towards correct parametric specication for other environmental variables.

14

4.2.3

Policy Implications

One of the advantages of incorporating parameter heterogeneity into the regression model is that it allows us to identify the country specic returns to both FDI and a marginal change in corruption. We can analyze the placement of countries in the distribution of FDI coecients to ascertain which countries have the highest returns to FDI. In addition, we can identify which group of countries may benet the most from a reduction in corruption; countries that have the highest partial eect of the FDI coecient with respect to corruption may benet substantially from FDI if their level of corruption were to decrease. Moreover, the present analysis may help directly with international FDI policies. One stipulation in an international FDI agreement may be a mandatory reduction in the level of corruption in the developing host country. At the very least, this analysis can assist policymakers in determining in which countries FDI will most eectively improve growth rates, and in which countries FDI may have a neutral eect. Focusing now on the fully-specied semiparametric instrumental variables model (NPGMM), we analyze which countries appear to have the highest and lowest returns to FDI. Table 3 shows the lists of countries divided based on their relative returns to FDI. For those observations with positive and signicant FDI coecients, we divide the countries into separate lists for each of the four quantiles based on the magnitude of the coecient. That is, the countries with FDI coecients that are in the highest 75 percentile are grouped together in the 4th quantile group; we do the same for each quantile.4 Since all of the countries in these quantile lists realize positive returns to FDI, we also include a list of countries that have insignicant FDI coecients.5 Because we have a panel data set, some countries have FDI coecients that appear in each category for at least one year. To provide a bit more clarity as to which countries receive high or low returns, we group the countries based on their modal classication: if a country appears most frequently in the column for insignicant returns to FDI, we classify it as insignicant. It is important to highlight which countries have consistently insignicant returns to FDI; policymakers may want to reconsider investment policies or investment stipulations aimed at these particular countries.6 We can see in Table 3 that there does not appear to be any geographical similarities between the groupings of countries. Each group of countries contains countries from each continent or geographical region. This result implies that there exists heterogeneity in FDI returns even within geographical regions. Hence, geographical-oriented investment policies may be inappropriate for enhancing growth-eects of FDI, the best FDI policies may most likely be country-specic. With this in mind, the country lists in Table 3 provide preliminary estimates of the potential returns each country may realize from further FDI. Table 4 shows a similar breakdown of countries based on the partial derivative of the FDI coecient with respect to corruption.7 Here, we classify the modal observation into quantiles of
The smallest positive and signicant coecient is 0.004, and the largest positive and signicant coecient is 3.15. The quartile values are 0.37, 0.61, and 0.78, respectively, for the 25th , 50th , and 75th percentiles. All estimated values are in percentage terms. 5 Only 2 percent of the coecients were negative and signicant; none of the negative and signicant coecients represented the modal classication, so we do not include a category for negative returns. 6 It is important to acknowledge that although these classications are based on the modal observations, many of the countries appear in the same category for many of the years in the panel. Hence the modal classication provides an accurate depiction of the distribution of the countries across the classication groups. 7 Specically, the largest negative and signicant partial eect is -4.16, and the smallest negative and signicant
4

15

negative and signicant coecients and a separate category for insignicant returns. While we again fail to nd any apparent geographical groupings emerging from the lists, comparing groups of countries between Tables 3 and 4 yields interesting results that complement our previous ndings. Of the countries with insignicant returns to FDI, we nd approximately 67 percent fall into the group of countries that record the highest marginal returns to corruption; and of this same group of FDI coecients, approximately 83 percent fall into the groups that record the highest and second highest returns to corruption. A similar analysis of countries with the lowest positive returns to FDI shows that approximately 63 percent fall into the group with the highest returns to a decrease in corruption, and approximately 88 percent fall into the groups with the highest and second highest returns to a decrease in corruption. These comparisons strongly suggest that many countries with insignicant or low returns to FDI may benet substantially from a reduction in their levels of corruption. A similar comparison is done for countries with the highest returns to FDI. We nd that approximately 28 percent of the countries with the highest returns to FDI have insignicant returns to a decrease in corruption, and approximately 61 percent have insignicant or low returns to a decrease in corruption. This suggests that the countries with the highest returns to FDI do not stand to gain as much from a marginal reduction in corruption. Overall, our estimates suggest that corruption does indeed weaken the relationship between FDI and growth.

Sensitivity Analysis

We now turn to additional model specications that can be used to examine the robustness of our core results. Our primary concern is that we have inadequately controlled for any potential endogeneity that may exist between GDP growth rates, FDI, corruption, or any other variables in our conditioning set. Our secondary concern is that we have failed to incorporate related environmental variables into our assessment of the relationship between corruption and the FDI-growth relationship. All results from this section are available on request. We rst address our concerns regarding any additional biases arising from endogeneity. To examine whether any possible endogeneity bias associated with the X variables (excluding FDI) is driving our core results, we re-estimate our fully-specied NPGMM model using one-period lags of all X variables, with the exception of the U.S. treasury bill rate.8 Indeed, one of the strengths of the NPGMM model proposed by Cai and Li (2008) is its ability to allow for the potential endogeneity of all X variables. Lagging all regressors reduces our sample size to 1020. Although the estimated model has fewer number of statistically signicant coecients for the other X variables, as well as a slightly lower R2 value, our core results remain unchanged. We nd a positive and signicant relationship between FDI and GDP growth rates, and a negative and signicant relationship between corruption and the FDI-growth relationship for a large subset of observations in our sample. To address any endogeneity brought on by serial correlation within the regressors for any
partial eect is -0.50. The quartile values are 1.08, 1.28, and 1.45, respectively, for the 25th , 50th , and 75th percentiles. All estimated values are in percentage terms. 8 The U.S. treasury bill rate is explicitly assumed to capture exogenous uctuations in macroeconomic conditions.

16

of the X regressors (including FDI), we re-estimate our fully-specied NPGMM model using two non-overlapping, aggregated panel data sets. The rst aggregated panel uses observations aggregated over 4-year intervals, and the second aggregates observations over 3-year intervals.9 In each case, we aggregate the data by taking the average of the annual observations over each time-interval, and use the level of initial GDP at the beginning of each time interval to be our measure of initial GDP. There are several advantages of using time-aggregated panels. One, time-aggregation can substantially reduce serial correlation within regressors that may cause correlation between the regressors and the error term. Two, time-aggregated panels can capture more general movements in macroeconomic conditions by ltering out business cycle uctuations and minimizing attenuation bias from measurement error, which are usually prevalent in annual data. The aggregation into a shorter panel reduces our number of observations to 300 and 360, respectively, for the 4-year and 3-year models. We caution, however, that with such few observations we may introduce dimensionality issues in terms of estimating the unknown coecient functions. In both models we nd a positive and signicant relationship between FDI and GDP growth rates, as well as a negative eect of corruption on the FDI-growth relationship. We note, however, that the number of statistically signicant coecients for other conditioning variables has been reduced, and for some observations, the sign of the coecient has been switched. This suggests that dimensionality issues may be present when estimating the coecients. Hence, the sample size may be insucient for accurate estimation of the model. Nevertheless, these results are generally consistent with our core results, suggesting that any serial correlation that may be present within the regressors is not causing us to obtain inconsistent estimates in our semiparametric models. Although our instrumental variable for FDI, lagged FDI, provides favorable results, this internal instrument may not be fully valid. It is well-known that factors that induce correlation between an endogenous regressor and the error term can also induce correlation between an internal instrument and the error term. For this reason, instruments that are external to the data can be more credibly valid than their internal counterparts. To assess whether our core results are inuenced by invalid instruments, we estimate our fully-specied NPGMM model using both total world FDI ows and total FDI ows to developing countries as instruments for FDI. As previously mentioned, total world FDI ows and total FDI ows to developing countries are observable variables that should cause uctuations in FDI, which are unrelated to conditions within an individual host country. Using these external instrumental variables, we obtain similar results as those derived from our preferred specication. Hence, our core results do not appear to be driven by invalid instruments. So far, we have only considered one measure of corruption, and have ignored any potential endogeneity and other biases associated with this corruption variable. In particular, although the Spearmans correlation test suggests that the TI and Knack and Philip indices are strongly correlated, merging dierent corruption indices may result in a conceptual ambiguity since each index aims at capturing dierent aspects of corruption. In addition, the scale of measurement diers across indices and also, aggregation causes measurement errors in these various indices to be dependent, hence increasing the variance in the measurement error. The extent of this
9

For the 4-year panel, the last interval contains only 2 years.

17

attenuation bias will further exacerbate if our assumption of a time-invariant corruption index is inappropriate for the data. To address these concerns, we consider (a) the time average of the Knack and Philip index for the period 1984 to 1996 and (b) the Knack and Philip index for 1984 in lieu of our aggregated corruption index. Given that the initial year of our time span is 1984, this latter measure of corruption seems more credibly exogenous, at least in theory, than the other measures. We nd that our core results hold intact against these alternative measures of corruption. Specically, we nd that these other measures support our earlier results that corruption weakens the relationship between FDI and GDP growth. These results suggest that any potential endogeneity or inappropriateness of the time-invariant assumption of our aggregated corruption measure is not driving our core results. Turning now to our secondary concern, we add two dierent variables to our set of environmental factors, namely the fertility rate (total births per woman) and an index of democracy. Democratic economies can better guarantee and protect property and contract rights than autocratic economies, and North (1990) argues that secure property rights are crucial for economic growth. Furthermore, Barro (1996) nds a heterogeneous democracy-growth relation in which more democracy is growth-enhancing at low levels of democracy but is growth-deterring at high levels of democracy. High fertility rates can reduce investments in health and human capital, which in the long-run can result in reduced physical work capacity in the labour force. Thus, democracy and fertility rate may aect the ecacy of FDI to promote growth, and may also be directly related to economic growth. We therefore include fertility rate and an index of democracy to control for unobserved environmental factors that may cause heterogeneity in the FDI-growth relationship but are omitted from our initial list of environmental variables. We nd that including either variable in Z does not alter the conclusions drawn from our core results.

Conclusion

In this paper we present a generalized empirical growth model that concurrently allows (a) for heterogeneity within the FDI-growth relationship, as well as between the growth rate and all other conditioning variables, (b) each of the coecients in the model (e.g., the coecient governing the FDI-growth relationship) to be a function of corruption within each country, and (c) each of the conditioning variables to be endogenous. The advantage of such a generalization is that it allows us to analyze heterogeneity within the corruption-FDI-growth nexus and address many of the concerns discussed by Durlauf (2001), in a unied framework. We nd that there exists substantial heterogeneity within the FDI-growth relationship, and that FDI has a positive and signicant eect on growth rates for many of the countries in our sample. In addition, we nd that corruption signicantly reduces the eect of FDI on growth rates for a large subsample of our data. We nd, however, that corruption has an insignicant net eect on growth rates. The nding that corruption weakens the relationship between FDI and growth suggests that international investment policies aimed at improving growth rates through FDI should carefully consider corruption levels in the developing host country prior to implementing FDI policies. Our results suggest that many countries with relatively low returns to FDI may substantially benet, in terms of the returns to FDI on growth rates, from a reduction in corruption. 18

References
[1] Aitken, B. and A. Harrison, 1999. Do Domestic Firms Benet from Foreign Direct Investment? Evidence from Venezuela American Economic Review, 89, 605-618. [2] Alfaro, L., A. Chanda, S. Kalemli-Ozcan and S. Sayek, 2004. FDI and Economic Growth: The Role of Local Financial Markets, Journal of International Economics, 64, 89-112. [3] Barro, R.J. 1996. Democracy and Growth, Journal of Economic Growth, 1, 1-27. [4] Baumol, W. J., 1990. Entrepreneurship: Productive, Unproductive, and Destructive, Journal of Political Economy, 98, 893-921. [5] Blomstrom, M., 1986. Foreign Investment and Productive Eciency: The Case of Mexico, Journal of Industrial Economics, 35, 97-110. [6] Borensztein, E., J. De Gregorio, and J. W. Lee, 1998. How Does Foreign Direct Investment Aect Economic Growth? Journal of International Economics, 45, 115-135. [7] Cai, Z., M. Das, H. Xiong, and X. Wu, 2006. Functional coecient instrumental variables models, Journal of Econometrics, 133, 207-241. [8] Cai, Z., J. Fan, and Q. Yao, 2000. Functional-Coecient Regression Models for Nonlinear Time Series, Journal of the American Statistical Association, 95, 941-956. [9] Cai, Z. and Q. Li, 2008. Nonparametric Estimation of Varying Coecient Dynamic Panel Data Models, Econometric Theory, 24, 1321-1342. [10] Carkovic, M. and R. Levine, 2005. Does Foreign Direct Investment Accelerate Economics Growth? Does Foreign Direct Investment Promote Development, eds. T. H. Moran, E. M. Graham and M. Blomstrom, Washington: Center for Global Development and Institute for International Economics, 195-220. [11] Das, M., 2005. Instrumental variables estimators of nonparametric models with discrete endogenous regressors, Journal of Econometrics 124, 335-361. [12] Durlauf, S. N., 2001. Manifesto for a Growth Econometrics, Journal of Econometrics, 100, 65-69. [13] Durlauf, S. N., A. Kourtellos and A. Minkin 2001. The Local Solow Growth Model, European Economic Review, 45, 928-960. [14] Gupta, S., H. R. Davoodi and E. R. Tiongson, 2002. Corruption and the Provision of Health Care and Education Services, Governance, Corruption, Economic Performance, eds. G. T. Abel and S. Gupta, Washington: International Monetary Fund, 245-279. [15] Haddad, M. and A. E. Harrison, 1993. Are There Positive Spillovers from Direct Foreign Investment?: Evidence from Panel Data for Morocco, Journal of Development Economics, 42, 51-72.

19

[16] Hastie, T. and R. Tibshirani, 1993. Varying-Coecient Models, Journal of the Royal Statistical Society, Series B, 55, 757-796. [17] Knack, S. and K. Philip, 1998. IRIS-3: File of International Country Risk Guide (ICRG) Data, 3rd Edition. New York: The Political Risk Service Group, Inc. [18] La Porta, R., F. Lopez-de-Silanes, A. Shliefer, and R. Vishny, 1999. The Quality of Government, Journal of Law, Economics, and Organization, 15, 222-279. [19] Li, Q., C. J. Huang, D. Li, and T. Fu, 2002. Semiparametric Smooth Coecient Models, Journal of Business and Economic Statistics, 20 (3), 412-422. [20] Liu, Z. and T. Stengos, 1999. Non-Linearities in Cross-Country Growth Regressions: A Semiparametric Approach, Journal of Applied Econometrics, 14, 527-538. [21] Mauro, P., 1998. Corruption and the Composition of Government Expenditure, Journal of Public Economics, 69, 263-279. [22] McCloud, N. and S. C. Kumbhakar, 2011. Institutions, Foreign Direct Investment, and Growth: A Hierarchical Bayesian Approach, Journal of the Royal Statistical Society: Series A, Forthcoming. [23] Murphy, K.M., A. Shleifer, and R.W. Vishny, 1991. The Allocation of Talent: Implications for Growth, The Quarterly Journal of Economics, 106, 503-530. [24] North, D. (1990) Institutions, Institutional Change and Economic Performance. New York: Cambridge University Press. [25] Racine, J. S. and Q. Li, 2004. Nonparametric Estimation of Regression Functions With Both Categorical and Continuous Data, Journal of Econometrics, 119, 99-130. [26] Staiger, D. and J. Stock, 1997. Instrumental Variables Regression with Weak Instruments, Econometrica, 65, 557-586. [27] Tanzi, V. and H. R. Davoodi, 2002. Corruption, Public Investment, and Growth, Governance, Corruption, Economic Performance, eds. G. T. Abel and S. Gupta, Washington: International Monetary Fund, 280-299. [28] Transparency International, Corruption Perception Index. URL: http :

//www.transparency.org/policyr esearch/surveysi ndices/global/cpi.

20

Table 1: Summary of results from the parametric and semiparametric regression models. Variable (1) Intercept FDI Corruption Initial Income Openness Government Consumption 21 Domestic Investment US T-Bill Ination Rate -0.0846 0.0156 0.3013 0.0827 OLS Models (2) -0.1148 0.0186 0.3031 0.0824 0.0210 0.0071 (3) 0.6014 0.1022 0.1024 0.0840 0.0200 0.0069 -0.1039 0.0126 -0.0153 0.0122 0.0809 0.0504 0.3401 0.0525 -0.5278 0.1925 -0.0004 0.0003 0.2128 (4) -0.0855 0.0156 0.6152 0.1811 2SLS Models (5) -0.1161 0.0187 0.6215 0.1804 0.0212 0.0071 (6) 0.5814 0.1024 0.4916 0.2006 0.0216 0.0069 -0.0993 0.0128 -0.0230 0.0127 0.0576 0.0515 0.2766 0.0602 -0.3858 0.2032 -0.0002 0.0003 0.2164 SPSCM (7) 0.0066 0.0041 0.3829 0.1406 0.0048 0.0062 (8) 0.0617 0.0257 0.2723 0.0824 0.0052 0.0087 -0.0089 0.0040 -0.0123 0.0052 0.0117 0.0329 0.1801 0.0388 0.0186 0.1544 -0.0013 0.0004 0.5416 NPGMM (9) 0.0041 0.0045 0.5207 0.1306 0.0058 0.0058 (10) 0.0805 0.0341 0.5187 0.1360 0.0005 0.0181 -0.0114 0.0036 -0.0205 0.0066 0.0164 0.0236 0.1738 0.0392 0.0447 0.1238 -0.0011 0.0003 0.3706

R2

0.1241

0.1307

0.1226

0.1293

0.5929

0.4673

1. Dependent variable in each regression is the growth rate of GDP per capita. 2. All OLS and 2SLS models include dummy variables to control for country and time xed eects; all semiparametric models allow coecients to vary with respect to corruption and country and time xed eects. 3. Semiparametric models report mean coecients and standard errors. 4. OLS and SPSCM models do not control for endogeneity of FDI. 5. First stage 2SLS models regress FDI on lagged FDI, exogenous variables, and country and year dummy variables. 6. The eects of corruption on GDP growth reported in both the SPSCM and NPGMM models is the total eect of corruption as dened in section 2.1.

Table 2: Summary of the eect of corruption on the coecients. Variable Intercept FDI Initial Income Openness Government Consumption Domestic Investment US T-Bill Ination Rate SPSCM (1) 0.0098 0.0077 -0.3132 0.2270 (2) 0.0706 0.0618 -0.3992 0.1142 -0.0078 0.0058 0.0186 0.0075 -0.0281 0.0402 -0.0591 0.0879 0.1892 0.2243 -0.0002 0.0003 NPGMM (3) 0.0128 0.0069 -0.4091 0.1332 (4) 0.0702 0.0722 -1.1656 0.4245 -0.0059 0.0086 0.0163 0.0160 -0.0582 0.0715 0.0129 0.0965 0.0506 0.3599 0.0006 0.0003

1. Table reports mean partial eects and standard errors of corruption on each of the conditioning coecients.

22

Table 3: Countries grouped according to FDI returns. Insignicant 1st Bangladesh Bolivia Ethiopia Haiti Hungary Indonesia Jamaica Nigeria Panama Paraguay Singapore Zambia Egypt El Salvador Ghana Guatemala Honduras Pakistan Philippines Romania* 2nd Botswana Chile India Korea Peru* South Africa Sri Lanka Tunisia Uruguay Vietnam Positive and Signicant 3rd Angola Cameroon Colombia Dominican Republic Ecuador Kenya Mexico Morocco Peru* Senegal Thailand Trinidad and Tobago Uganda Venezuela 4th Albania Argentina Brazil Bulgaria Burkina Faso China Costa Rica Jordan Madagascar Malawi Malaysia Mongolia Mozambique Nicaragua Poland Romania* Tanzania Zimbabwe

1. Countries are grouped by modal appearance across categories. 2. 1st , 2nd , 3rd , and 4th denote the relative quantiles for the positive and signicant coecients. 3. Countries with * are listed more than once.

23

Table 4: Countries grouped according to corruption returns. Insignicant 1st Angola Botswana Chile* Costa Rica Hungary Jordan Nicaragua Poland* Singapore South Africa Tanzania Zambia Burkina Faso Chile* China El Salvador Kenya* Korea Madagascar Malaysia Mongolia Mozambique Philippines Senegal Tunisia Uruguay 2nd Bolivia* Brazil Dominican Republic* Egypt* Kenya* Mexico Morocco Paraguay* Peru Poland* Sri Lanka Thailand Trinidad and Tobago Vietnam Negative and Signicant 3rd Argentina Bangladesh* Bolivia* Colombia Dominican Republic* Ecuador Ghana* Honduras Kenya* Malawi Uganda 4th Albania Bangladesh* Bulgaria Cameroon Egypt* Ethiopia Ghana* Guatemala Haiti India Indonesia Jamaica Nigeria Pakistan Panama Paraguay* Romania Venezuela Zimbabwe

1. Countries are grouped by modal appearance across categories. 2. 1st , 2nd , 3rd , and 4th denote the relative quantiles for each of the negative and signicant coecients. 3. Countries with * are listed more than once.

24

SPSCM 1

SPSCM 2

^ 1

^ 1 2 0 2 1 0 1 2 ^ 1 3 4 5

0.5

0.0

0.5

1.0

1.5

2.0

2.5

0.0

0.5 ^ 1

1.0

1.5

NPGMM 1

NPGMM 2

^ 1

^ 1 2 1 0 ^ 1 1 2 3 4 2 0

1 ^ 1

Figure 1: 45 degree gradient plot of the eect of FDI on growth for each of the four semiparametric models in Table 1.

25

SPSCM 1
1

SPSCM 2

^ (1) (Z1)

^ (1) (Z1) 8 6 4 2 0 2

10

1.5

1.0

0.5 ^ (1) (Z1)

0.0

0.5

^ (1) (Z1)

NPGMM 1
4

NPGMM 2

^ (1) (Z1)

^ (1) (Z1) 5 4 3 2 1 0 1 2

2 ^ (1) (Z1)

^ (1) (Z1)

Figure 2: 45 degree gradient plot of the eect of corruption on the FDI coecients for each of the four semiparametric models in Table 1.

26

SPSCM 1
0.4 0.4

SPSCM 2

(g) (Z1)

0.2

(g) (Z1) 0.1 0.0 (g) (Z1) 0.1 0.2

0.0

0.2

0.4

0.2

0.0

0.2

0.1

0.0 (g) (Z1)

0.1

0.2

NPGMM 1
0.3 0.4

NPGMM 2

0.2

(g) (Z1)

0.1

(g) (Z1) 0.2 0.1 0.0 (g) (Z1) 0.1 0.2

0.0

0.1

0.2

0.8

0.6

0.4

0.2

0.0

0.2

0.2

0.1

0.0 (g) (Z1)

0.1

0.2

0.3

Figure 3: 45 degree gradient plot of the eect of corruption on growth for each of the four semiparametric models in Table 1.

27