We examine the effects of an important technology diffusion channel—foreign direct investment (FDI)—on the growth of total factor productivity (TFP) an d the role played by natural resources in this relationship. Based on cross-sectional data from 71 developing countries, we found that the net effect of FDI on TFP growth decreases with rents provided by natural resources. This result highlights the phenomenon of the natural resource curse applied to foreign direct investment and the non-linearity of the effect of FDI on the TFP growth.
Since the 1980s, inflows of foreign direct investment (FDI) have risen rapidly [
Although the increase in FDI flows and its positive spillovers have been identified in the literature, the beneficial effect of FDI on productivity gains or economic growth is not empirically conclusive, making it an important field of research.
Many authors argue that differences in total factor productivity (TFP) are the key to understand countries’ income differences [
At the microeconomic level of the firm, the evidence of benefits from FDI remains unclear and contradictory [
The macroeconomic literature review shows that three key factors are involved in the relationship between FDI and TFP: human capital, financial development and trade regime (openness versus protectionism). Carkovic and Levine [
We postulate in this study that the ambiguity of the impact of FDI on TFP identified in previous studies is related to the fact that the presence or absence of natural resources in the host countries is not considered. Our hypothesis is that investors going to resource-rich countries are concerned about the extraction of natural resource-related rents and not about a potential increase in the productivity of local factors of production. FDI in resource-poor countries should therefore be expected to have positive impacts on TFP, unlike that in resource-rich countries.
The purpose of this study is therefore to identify the role of natural resources in determining the net effect of FDI on the growth of TFP. Is there any natural resource curse in the relationship between foreign direct investment and total factor productivity? To answer this question, we relied on cross-sectional data from 71 developing countries over the period 1995-2005 (average of the period)1.
In the rest of the paper, we will first present and describe our data; then, attention will be paid to the econometric strategy; finally, the results will be presented and discussed before concluding and drawing economic policy recommendations.
This section describes the data used in the empirical analysis, specifically, measures of TFP growth, FDI, natural resource rent, and some control variables. Our sample is made up of 71 developing countries over the period 1995-2005. This period is relevant to our study because of the rapid increase in net inflows of FDI during this period, especially towards developing countries. The sample was selected based on the availability of data and the relevance of the issue for developing countries. In addition, we have averaged the data over the period because we plan to work in cross-sections.
This is the dependent variable in our model. We obtained this by calculating the average annual growth rate of TFP over the period 1995-2005. Our measure of total factor productivity is based on the standard framework of growth accounting. Consider a standard Cobb-Douglas production function as follows:
where Y is the aggregate product (GDP); A is total factor productivity (TFP); K is the physical capital stock; L is the labor force of the economy; and α is the elasticity of the product relative to the capital. With the data on Y, K, L, and α, it becomes simple to calculate the value of TFP. The labor force (L) and GDP (Y) are extracted from the World Development Indicators (WDI) [
The growth rate is then generated, taking the first difference of lnA and then averaging over the period. We will designate the annual growth rate of TFP by TFPGR.
FDI is the main explanatory variable of our model. Data on FDI come from WDI. We use the share of net FDI flows as a percentage of GDP. Since we focus on the technology transfers that FDI can provide for host countries, we think it is more relevant to consider net values rather than gross values. More precisely, we use the average value of FDI, defined as a percentage of GDP, for our study period.
This is a very important explanatory variable for our study. The variable “RENT” is taken from the World Bank Adjusted Saving Project database. This database is used in the empirical works on natural resource exploitation [
The computation of the values of this variable is done in several stages. In a first step, we obtain the unit rent by the difference between the price on the world market and the unit cost of extraction. For negative values of the unit rent, it is assumed that the result is due to incomplete data on extraction costs [
Econometric growth regressions, in cross-sections, usually include the initial level of GDP as an explanatory variable to control the convergence effects. Although there is no clear theoretical basis for TFP convergence across countries, recent studies have suggested convergence towards a common technological frontier. Ayhan Kose, Prasad [
This indicator considers loans granted to firms and households by banking and non-banking institutions (loans from the Central Bank to commercial banks and credits granted to the government and public enterprises are excluded), in relation to GDP. We chose it as an indicator of the size of financial development. It has been extracted from the WDI. In the literature on the determinants of TFP and economic growth, financial development plays an important role. King and Levine [
This is the consumer price index also extracted from the WDI. We consider it as an indicator of the macroeconomic environment, which is one of the main determinants of total factor productivity growth. Macroeconomic stability tends to stimulate long-term productivity growth, reduce interest rates and encourage entrepreneurs to spread their projects over a longer horizon. Aghion, Angeletos [
Trade liberalization also plays an important role in the literature on the determinants of TFP growth (see for example [
POPGR is the annual growth rate of the population calculated over our study period. Data on population growth rate are extracted from the WDI. Population growth is one of the main variables found to be robust in growth regressions in the economic literature. Ayhan Kose, Prasad [
After the presentation of the main variables used in our study, we give in this section an overview of their descriptive statistics (
Our study attempts to show the net effect of FDI on the TFP growth, considering the rent provided by natural resources and controlling for several variables. In the logic of Mankiw, Romer [
Variables | Mean | Standard dev. | Min | Max |
---|---|---|---|---|
TFPGR | 0.97 | 1.91 | −4.67 | 4.79 |
FDI | 3.37 | 3.42 | 0.10 | 20.04 |
RENT | 3.95 | 6.15 | 0.00 | 26.05 |
INFLATION | 16.07 | 23.81 | 2.04 | 138.33 |
TRADE | 76.35 | 38.20 | 21.79 | 207.21 |
CREDIT | 28.60 | 30.42 | 3.53 | 172.68 |
TFPGR | FDI | RENT | TRADE | INFLATION | CREDIT | |
---|---|---|---|---|---|---|
TFPGR | 1.000 | |||||
FDI | 0.088 | 1.000 | ||||
RENT | −0.215 | 0.185 | 1.000 | |||
TRADE | 0.313 | 0.346 | −0.088 | 1.000 | ||
INFLATION | 0.017 | −0.033 | 0.148 | 0.048 | 1.000 | |
CREDIT | 0.285 | −0.087 | −0.090 | 0.327 | −0.201 | 1.000 |
the following equation:
We will focus on the estimated coefficients of the Equation (1) and more particularly on the coefficients in front of the FDI and RENT variables (β2 and β3, respectively). We check the consistency of the signs of the coefficients of our control variables with what we expected and presented in the presentation of the variables. Second, and this is also the key point of our study, we interact the variable FDI with the variable RENT and use it as an explanatory variable to capture the role of natural resources in determining the effect of FDI on TFP growth. The two variables FDI and RENT are still controlled in the equation, which is specified as follows:
In the specification of Equation (2), we will focus on the coefficients in front of the variables FDI and FDI*RENT.
An important issue in the relationship between FDI and TFP growth is the endogeneity of the FDI variable. Indeed, while FDI can broadly be a source of technology transfer and economic growth, it is also plausible that FDI itself will be determined to a large extent by the TFP growth. More specifically, a country with higher productivity growth may attract more FDI than a country with relatively smaller TFP growth. It is expected that a country with high-growth TFP will be more efficient in terms of adoption of new technologies and innovations, and this can be a fundamental determinant of foreign investors’ decisions regarding location. This is a specific case of a simultaneity bias that can be a source of endogeneity of the FDI variable. In addition, there may be an omission bias caused by the omission of a relevant variable correlated with FDI, as well as by errors in FDI measurement.
To solve this endogeneity problem, we will use an instrumental variable estimation method of two-stage least squares (IV). The instrumental variables must meet two conditions to be valid: they must be effectively correlated to the endogenous variable, and they must have no direct influence on the dependent variable (TFP growth). We will test the validity and non-weakness of our instruments and the endogeneity of the FDI variable in the following section. Here, we are just posing the problem.
We use three variables as instruments of the FDI variable: the country’s isolation (LANDLOCK), the lagged FDI (FDILAG) and the real exchange rate (EXR). Our variable LANDLOCK is a dummy variable that takes the value 1 if the country is landlocked and 0 if not. We think that the fact that a country is landlocked or not can affect the investment choice of foreign entrepreneurs especially because of the difficulties that can appear in the routing of imported products in the case of a landlocked country. Regarding the instrumental variable FDILAG, which represents the previous FDI (for the period 1985-1990), we drew on the economic literature. Wheeler and Mody [
Our third instrument―the real exchange rate (EXR)―is an important determinant of FDI among many others. Real exchange rates, by modifying the relative costs (or relative wealth), can have an impact on the investment decisions of multinational firms. For example, Froot and Stein [
Finally, it should be noted that since the FDI variable is suspected of endogeneity, the introduction of any interactive variable involving FDI makes this new variable endogenous. This is the case of FDI*RENT in our Equation (2), which is our equation of interest. For this reason, we will also instrument this interactive variable, which to some extent represents the FDI in the natural resource sector generating rents in the corresponding countries. This interactive variable will be instrumented by FDILAG*RENT and LANDLOCK*RENT.
The normality test applied to our equation of interest gives us a statistic of 3.12 and a probability of committing a first-order error of 0.212. In other words, if we reject the null hypothesis of normality, there is a 21 percent chance of making a wrong decision. Since this probability is higher than the usual tolerance thresholds (1%, 5% or 10%), the hypothesis of normality of residuals is not rejected.
The White test applied to the equation of interest gives us a White statistic of 42.26 with a p-value of 0.50. The hypothesis of homoscedasticity is therefore not rejected at the usual tolerance thresholds, as the probability of committing an error by rejecting the null hypothesis is 50 percent. Correction of heteroscedasticity is therefore unnecessary3.
In the section on endogeneity issues, we underlined the likely endogeneity of the FDI and FDI*RENT variables. Thus, it appears necessary to apply the orthogonality test to confirm or refute our suspicion. However, one must make sure of the non-weakness of the instruments used. Using weak instrument variables to realize the orthogonality test makes this test less powerful, as the orthogonality hypothesis is often accepted wrongly. The test of non-weakness of instruments consists in checking whether the instruments are sufficiently correlated with the endogenous variables or whether the explanatory power of the instrumentation equations is quite important. As already highlighted above, the variables LANDLOCK, FDILAG and EXR are used as instruments of the FDI variable, and for the interactive variable FDI*RENT, the instrumental variables are LANDLOCK*RENT and FDILAG*RENT. When there are two or more endogenous variables, the instruments of each endogenous variable also instrument the other endogenous variables. We thus have two instrumentation equations, one for FDI and the other for FDI*RENT, in which the endogenous variables are regressed on the exogenous explanatory variables and on their five instruments, without omitting the constant. To ensure the non-weakness of the instruments, we test the joint significance of these instruments in the two equations with an F-test and calculate the partial R2 (representing the explanatory power of the instruments).
We approximate the partial R2 by the difference between the R2 of the instrumentation equation and the R2 of the same equation omitting the instrumental variables. Thus, for the FDI instrumentation equation, the joint significance test of the instruments gives us a Fisher statistic of 11.10 (p-value of 4.28 × 10−7), and the partial R2 is 0.38. For FDI*RENT, we get a Fisher statistic of 67.19 (p-value of 2.98 × 10−10) and a partial R2 of 0.55. These indicators show that the explanatory power of the instruments is important; they cannot be considered as weak4.
The aim is to test whether our suspected endogenous variables are endogenous or not, particularly the FDI and FDI*RENT. The orthogonality test used is that of Durbin, Wu and Hausman in the version of Nakamura and Nakamura [
The principle of the test is to remove the residuals of our two instrumentation equations estimated by the OLS and to introduce them as explanatory variables in our equation of interest. We obtain the test equation, and the objective will be to check whether the two residual variables are significant or not. In the latter case, we do not reject the null hypothesis of orthogonality of the variables suspected of endogeneity.
Applying this test, we get a Fisher statistic of 13.80 for a p-value of 1.84 × 10−5. We therefore reject the null hypothesis of orthogonality of the FDI and FDI*RENT variables and confirm the endogeneity of FDI and FDI*RENT. The two-stage least squares method will be applied to estimate our Equation (2)5.
This is a test to ensure the quality of instrumental variables [
Let us now give an overview on the control variables of Equation (1). We can already note that the coefficient in front of TFP init (in logarithm) is negative and highly significant, which confirms the hypothesis of convergence. Indeed, it is likely that the higher the initial level of productivity, the lower the TFP growth will be. The natural resource rent variable has a significantly negative effect on the TFP growth. This result seems to validate the phenomenon of the natural resource curse. The financial development (CREDIT) also has a significant and positive effect on TFP growth, as expected. Inflation rate has a significantly
TFPGR | TFPGR | FDI | FDI*RENT | TFPGR | |
---|---|---|---|---|---|
Ln(TFPinit) | −0.009819*** | −0.00951*** | −0.93253 | 17.97145* | −0.005235 |
(−2.997) | (−2.867) | (−0.698) | (1.765) | (−1.141) | |
FDI | 0.000321 | 0.001543** | 0.006557* | ||
(0.435) | (2.342) | (1.841) | |||
FDI*RENT | −0.00011*** | −0.000301* | |||
(−3.165) | (−1.690) | ||||
RENT | −0.000442* | −0.000042 | −0.09574 | 1.122294** | −0.000254 |
(−1.864) | (−0.185) | (−1.598) | (2.037) | (−0.556) | |
Ln(TRADE) | 0.001015 | −0.000114 | 2.48538*** | 17.32759** | −0.012889* |
(0.281) | (−0.030) | (2.988) | (2.221) | (−1.814) | |
Ln(CREDIT) | 0.005374*** | 0.005450** | −1.16961** | −8.383227 | 0.011721*** |
(2.686) | (2.653) | (−2.267) | (−1.526) | (2.854) | |
Ln(INFLATION) | −0.003983** | −0.003288* | −0.497567 | 0.983595 | −0.002649 |
(−2.090) | (−1.869) | (−1.374) | (0.365) | (−0.955) | |
POPGR | −0.768257*** | −0.74263*** | −130.644** | −399.5595 | −0.257186 |
(−4.243) | (−4.090) | (−2.255) | (−1.353) | (−0.718) | |
LANDLOCK | −0.665290 | −32.146*** | |||
(−0.559) | (−3.178) | ||||
LANDLOCK*RENT | 0.81017*** | 16.65108*** | |||
(6.534) | (6.978) | ||||
FDILAG | 0.09532 | −2.533505 | |||
(0.244) | (−0.830) | ||||
FDILAG*RENT | 0.14305* | 1.064964 | |||
(1.992) | (1.279) | ||||
EXR | 0.04233 | −0.580877* | |||
(0.841) | (−1.804) | ||||
CONSTANT | 0.060972** | 0.058217** | 2.89184 | −110.7236* | 0.045327 |
(2.334) | (2.141) | (0.439) | (−1.688) | (1.357) | |
R2 | 0.36 | 0.39 | 0.64 | 0.91 | 0.25 |
R2−adjusted | 0.29 | 0.32 | 0.56 | 0.89 | 0.13 |
Observations | 71 | 71 | 59 | 59 | 59 |
***, **,* significant at 1, 5, and 10 percent respectively; student t values in parentheses.
negative effect on the TFP growth, as macroeconomic instability may discourage investment and innovation. Population growth has a significant negative impact on productivity growth. Finally, although the coefficient of trade openness is positive as expected, it is not significant.
After this overview of our control variables, we turn to the results of the OLS estimation of Equation (2), in which we introduced an interactive variable FDI*RENT. The results are reported in the second column of
We previously suspected (and confirmed our suspicions using the orthogonality test) the FDI and FDI*RENT variables of endogeneity. It is therefore necessary to estimate our Equation (2) using the two-stage least squares method, as the OLS method may substantively bias our results. The results are reported in the last column of
It would be interesting to have an estimate of the important role that natural resource rent can play in determining the net effect of FDI on TFP growth. For this purpose, we calculate the effect of a one standard deviation increase of the RENT variable on the TFP growth for a country receiving the average level of FDI in our sample. The effect is measured by β”3*mean(IDE)*σ (RENT). It appears that a one standard deviation increase in natural resource rent reduces the FDI effect on TFP growth by 0.000062 percent.
One can also determine the level of RENT from which the net effect of FDI on TFP growth becomes negative, as this effect decreases with the natural resource rent. We derive our dependent variable (the TFP growth) relative to the FDI variable, and we equalize it to zero. We find a value of 21.78. This result just means that in our sample, if natural resources in a country generate an income above 21.78 percent of GDP, then the FDI has a negative net impact on TFP growth.
The effect of the FDI on TFP growth, which decreases with natural resource rents, can be explained by the combination of two factors. On the one hand, natural resources are themselves an important determinant of FDI, which means that FDI that goes to resource-rich countries is more oriented towards the natural resource sector. On the other hand, FDI in the sector of natural resources does not generate the expected benefits in terms of technology diffusion that can contribute to the growth of TFP. In fact, local companies can only benefit from the technology of multinational firms established in the country if they are either complementary or in competition. In the natural resources sector in many developing countries, this is not the case, and technology and skills transfers are often not effective.
One of the fundamental reasons for the observed growth in foreign direct investment flows and the interest that countries, especially developing ones, have in receiving this FDI is the idea that FDI can contribute immensely to the development efforts of the host country. In fact, the FDI can be a source of technology transfer, strengthening the workforce through managerial skills and other externalities that benefit the host economies by increasing the total factor productivity. With the rise in these FDI flows, especially in developing countries, the natural question that may come to the mind of policymakers is whether an economy grows faster through FDI. The empirical literature, at both macroeconomic and microeconomic scales, has not found convincing evidence in favor of FDI.
The benefits of foreign investment may depend on local conditions in the host country. Based on a sample of 71 developing countries, we have attempted to show that the availability of natural resources in the countries and more precisely the rents that these natural resources provide can be an important determinant of the net effect of FDI on the growth of TFP. We empirically found that the effect of FDI on the TFP growth decreases with the natural resource rent. Thus, the gains of FDI in terms of productivity growth are lower in more natural resource-rich countries, highlighting the phenomenon of the natural resource curse.
This result is highly important, especially for sub-Saharan African countries where extractive industries receive the bulk of FDI. To turn this curse into a “blessing”, it would be wise to examine the establishment of policies and institutions regulating the participation of multinationals in the extractive industries in a development-friendly way and to put in place some measures to encourage the industrialization and diversification of the developing economies based on the extraction of natural resources.
The authors would like to thank the anonymous referees and the editor of the journal for useful comments and suggestions on the original version of this article.
Zidouemba, P.R. and Elitcha, K. (2018) Foreign Direct Investment and Total Factor Productivity: Is There Any Resource Curse? Modern Economy, 9, 463-483. https://doi.org/10.4236/me.2018.93031
Countries | Countries | ||
---|---|---|---|
1 | Albania | 37 | Macedonia, FYR |
2 | Argentina | 38 | Madagascar |
3 | Azerbaijan | 39 | Malawi |
4 | Belarus | 40 | Malaysia |
5 | Benin | 41 | Mauritius |
6 | Botswana | 42 | Mexico |
7 | Brazil | 43 | Moldova |
8 | Burkina Faso | 44 | Morocco |
9 | Cambodia | 45 | Mozambique |
10 | Cameroon | 46 | Nicaragua |
11 | Chad | 47 | Pakistan |
12 | Chile | 48 | Peru |
13 | China | 49 | Philippines |
14 | Costa Rica | 50 | Poland |
15 | Cote d’Ivoire | 51 | Romania |
16 | Croatia | 52 | Russian Federation |
17 | Ecuador | 53 | Rwanda |
18 | Egypt | 54 | Senegal |
19 | El Salvador | 55 | South Africa |
20 | Ethiopia | 56 | Sri Lanka |
21 | Gambia, The | 57 | Sudan |
22 | Georgia | 58 | Swaziland |
23 | Ghana | 59 | Syrian Arab Rep. |
24 | Guatemala | 60 | Tanzania |
25 | Guyana | 61 | Thailand |
26 | Honduras | 62 | Togo |
27 | Hungary | 63 | Tunisia |
28 | India | 64 | Turkey |
29 | Indonesia | 65 | Uganda |
30 | Iran | 66 | Ukraine |
31 | Jordan | 67 | Uruguay |
32 | Kazakhstan | 68 | Venezuela, RB |
33 | Kenya | 69 | Vietnam |
34 | Kyrgyz | 70 | Zambia |
35 | Lesotho | 71 | Zimbabwe |
36 | Lithuania |
(Estimated residuals)2 | t−student | |
---|---|---|
Constant | −0.012064 | −1.086949 |
Ln(TFPinit) | −0.000582 | −0.248578 |
(Ln(TFPinit))2 | 0.000169 | 0.928578 |
(Ln(TFPinit))*FDI | −0.000100 | −0.675570 |
(Ln(TFPinit))*FDI*RENT | 3.73E−05 | 0.676337 |
(Ln(TFPinit))*RENT | −0.000111 | −0.652552 |
(Ln(TFPinit))*(Ln(TRADE)) | 7.31E−05 | 0.281620 |
(Ln(TFPinit))*(Ln(CREDIT)) | −0.000159 | −0.763931 |
(Ln(TFPinit))*(Ln(INFLATION)) | −0.000340 | −1.694893 |
(Ln(TFPinit))*POPGR | 0.003563 | 0.250563 |
FDI | −0.000131 | −0.152009 |
FDI2 | 3.00E−05 | 1.608165 |
FDI*FDI*RENT | −1.29E−05 | −1.495710 |
FDI*RENT | −0.000265 | −0.716871 |
FDI*(Ln(TRADE)) | −8.53E−05 | −0.456798 |
FDI*(Ln(CREDIT)) | 0.000243 | 1.896597 |
FDI*(Ln(INFLATION)) | 6.41E−06 | 0.078808 |
FDI*POPGR | −0.001066 | −0.212562 |
(FDI*RENT) 2 | 3.16E−07 | 1.002939 |
(FDI)*RENT | −1.16E−06 | −0.388743 |
(FDI*RENT)*(Ln(TRADE)) | 6.26E−05 | 1.053034 |
(FDI*RENT)*(Ln(CREDIT)) | −4.56E−05 | −1.077390 |
(FDI*RENT)*(Ln(INFLATION)) | 2.31E−06 | 0.100833 |
(FDI*RENT)*POPGR | 0.002013 | 1.589030 |
RENT | 0.000758 | 0.699599 |
RENT2 | 7.36E−06 | 0.963122 |
RENT*(Ln(TRADE)) | −0.000148 | −0.930505 |
RENT*(Ln(CREDIT)) | 0.000137 | 1.281784 |
RENT*(Ln(INFLATION)) | −5.12E−05 | −0.712068 |
RENT*POPGR | 0.004718 | −1.052031 |
Ln(TRADE) | 0.003543 | 1.135144 |
(Ln(TRADE))2 | −0.000173 | −0.701349 |
(Ln(TRADE))*(Ln(CREDIT)) | −0.000390 | −1.575658 |
(Ln(TRADE))*(Ln(INFLATION)) | −0.000321 | −1.139595 |
(Ln(TRADE))*POPGR | −0.016021 | −0.877402 |
Ln(CREDIT) | 0.001174 | 0.873517 |
(Ln(CREDIT))2 | 9.66E−05 | 0.790305 |
---|---|---|
(Ln(CREDIT))*(Ln(INFLATION)) | −1.43E−05 | −0.101693 |
(Ln(CREDIT))*POPGR | −0.011264 | −0.966158 |
Ln(INFLATION) | 0.003604 | 1.845971 |
(Ln(INFLATION))2 | −4.50E−05 | −0.662162 |
(Ln(INFLATION))*POPGR | −0.008223 | −0.719776 |
POPGR | 0.110877 | 0.906588 |
POPGR2 | −0.198122 | −0.356670 |
R2 | 0.60 | |
P-value (F-statistic) | 0.60 | |
P-value (N*R2) | 0.50 | |
Observations | 71 |
FDI | FDI | FDI*RENT | FDI*RENT | |
---|---|---|---|---|
Ln(TFPinit) | −0.93253 (−0.698) | −0.089802 (−0.171) | 17.97145* (1.765) | 1.683306 (0.276) |
RENT | −0.09574 (−1.598) | 0.113740 (1.126) | 1.122294** (2.037) | 4.622865** (2.103) |
Ln(TRADE) | 2.48538*** (2.988) | 2.908001*** (2.907) | 17.32759** (2.221) | 20.80291* (1.747) |
Ln(CREDIT) | −1.16961** (−2.267) | −1.336879** (−2.247) | −8.383227 (−1.526) | −13.35797 (−1.361) |
Ln(INFLATION) | −0.497567 (−1.374) | −0.565103 (−1.402) | 0.983595 (0.365) | 0.037923 (0.007) |
POPGR | −130.644** (−2.255) | −31.21034 (−1.365) | −399.5595 (−1.353) | −107.4232 (−0.389) |
LANDLOCK | −0.665290 (−0.559) | −32.146*** (−3.178) | ||
LANDLOCK*RENT | 0.81017*** (6.534) | 16.65108*** (6.978) | ||
FDILAG | 0.09532 (0.244) | −2.533505 (−0.830) | ||
FDILAG*RENT | 0.14305* (1.992) | 1.064964 (1.279) | ||
EXR | 0.04233 (0.841) | −0.580877* (−1.804) | ||
CONSTANT | 2.89184 (0.439) | −3.067298 (−0.807) | −110.7236* (−1.688) | −55.79534 (−0.992) |
R2 | 0.65 | 0.26 | 0.91 | 0.36 |
R2-adjusted | 0.56 | 0.19 | 0.89 | 0.30 |
Observations | 59 | 71 | 59 | 71 |
R2 partial | 0.39 | 0.55 |
***, **,*: significant at 1, 5, and 10 percent respectively; in parentheses the values of the student t.
TFPGR | |
---|---|
Ln(TFPinit) | −0.005235 (−1.377) |
FDI | 0.006557*** (3.058) |
FDI*RENT | −0.000301*** (−2.783) |
RENT | −0.000254 (−0.931) |
Ln(TRADE) | −0.012889*** (−2.969) |
Ln(CREDIT) | 0.011721*** (4.785) |
Ln(INFLATION) | −0.002649 (−0.993) |
POPGR | −0.257186 (−0.932) |
RES(FDI)estimated | −0.004908* (−1.749) |
RES (FDI*RENT)estimated | 8.99E−05 (0.449) |
CONSTANT | 0.045327* (1.812) |
R2 | 0.55 |
Observations | 59 |
***, **, *: significant at 1, 5, and 10 percent respectively; in parentheses the values of the student t.
Residuals estimated by IV | |
---|---|
Ln(TFPinit) | 0.003204 (0.628) |
LANDLOCK | 0.004468 (0.457) |
LANDLOCK*RENT | −0.000522 (−0.623) |
FDILAG | 0.000164 (0.061) |
FDILAG*RENT | 0.000109 (−0.000) |
EXR | −0.000230 (−0.788) |
RENT | 0.000144 (0.399) |
Ln(TRADE) | −0.001876 (−0.354) |
Ln(CREDIT) | −0.000424 (−0.131) |
Ln(INFLATION) | −0.000148 (−0.057) |
POPGR | −0.047808 (−0.177) |
R2 | 0.018 |
Observations | 59 |
The statistic of Sargan S = N*R2 = 0.018*59 = 1.062 and χ2(3) = 7.8147 for an error of the first kind of 5 per cent, hence the non-rejection of the null hypothesis of the validity of the instruments.