RELATIONSHIP BETWEEN FOREIGN DIRECT INVESTMENT AND ECONOMIC DEVELOPMENT

CHAPTER 7 RELATIONSHIP BETWEEN FOREIGN DIRECT INVESTMENT AND ECONOMIC DEVELOPMENT 7.0. INTRODUCTION The existing approach to the MNE theory treats the decision of a firm to go international as an extension of the firm theory. The general theory of FDI as given by Buckley and Casson (1976) works on two principles:(a) firms internalize missing or imperfect external markets until the costs of further internalization outweigh the benefits and (2) firms choose locations for their constituent activities that minimize the overall costs of their operations. Dunning s eclectic paradigm (Dunning, 1977, 1981) of ownership, location and internalization (OLI) advantages focuses on the sources of competitive advantage that allow a firm to compete abroad, the locational choices that firms make, and the mode of entry into foreign markets. In further analysis Dunning (1998) considers the developments in comparative advantage theory. Our approach is to incorporate the role economic development (socio-economic variables) into MNE theory. It arises out of a critical examination of Dunning (1998) wherein the author refers to macroeconomic aspects of the changing international allocation of economic activity. The argument contained in the above work needs to be recapitulated because it is based on a complex interplay of trade theory, MNE theory and macroeconomics. In the section on macroeconomics from Dunning s paper, he argues that even the recent developments in comparative cost advantage theory do not explain the international allocation of production. He suggests that the general understanding of literature is that the changes in geography of FDI over the last two decades have been broadly in line with the capital expenditure of all firms (sic) - MNE or otherwise. In reemphasizing the research by international business scholars, he points out that 129

This could mean that ownership or multinationality of firms was not a significant variable in explaining such changes: and that trade in intermediate or final product internalized and controlled by MNEs, is no differently determined than trade between independent firms, i.e., arm s length trade (Dunning (1998), pp. 55). The conclusion is that the main impact of foreignness or multinationality of firms has not been on the level of economic activity and/or trade of the countries in which they operate but on the structure of these variables. This implies that FDI will have a differential impact on the geography of economic activity. Our approach combines this fundamental understanding that MNEs are different with the implications of the new global business environment ushered in by globalization (WTO) that promotes changing international allocation of economic activity. It is this thrust that is expected to bring about a uniform development globally. FDI along with other is likely to lead to economic development (growth of socio-economic variables). In turn, economic development is likely to influence the pattern of FDI flows internationally. This leads to a two level analysis: 1. Where patterns of FDI flows (and stock) are studied in terms of concentration of inward FDI and dominance patterns of outward FDI. 2. The resultant economic development arising out of resource allocation due to FDI 1 flows in turn influence patterns of FDI through developmental variables as determinants of FDI In the light of this our study concentrates on the changes in geography of FDI over the last two decades that has let to changing international economic activity and the impact of economic development. 7.1. DETERMINANTS OF FDI Most studies concentrate on the effect of FDI on economic growth. Our study in the core chapter identifies and measures the determinants of international FDI patterns. 1 FDI resources imply not only capital but all the other complimentary resources which enable an international relocation of production such as technology transfer, managerial talent, organizational methods, etc. 130

Principal Component Analysis When we consider developmental variables like population, GDP, and so on, there is bound to be a high degree of correlation amongst independent variables. There could be three strategies that can be used for dealing with such an econometric problem: 1. If we drop all correlated variables there is a great loss of information. 2. We could use Principal Component Analysis (PCA) to determine the principal variables. 3. We could use PCA for formation of a composite index. The method of Principal Component Analysis (PCA) has two purposes. Firstly, we use PCA for data reduction, especially where the variables are interrelated. Secondly, we use PCA for compilation of a composite index. For estimating the determinants of international FDI patterns we have used a two step procedure. Firstly, components of FDI patterns are many and are correlated. Variables like GDP, human resources, trade openness, and so on which may be correlated. Under such circumstance it is not possible to use the variables directly in a regression framework on account of multicollinearity. Secondly, when there are a large number of variables we need to collapse them into a single independent variable with the help of PCA. The variable should be such that it captures all the information contained in all the individual variables. In view of these weaknesses of an ordinary regression framework, we opt for an alternative method - Principal Component Analysis (PCA). PCA is based on a linear transformation of the regressors so that they are orthogonal to each other by design. Hence, no information contained in the points in the event space is lost. Second, the normality assumption is not essential in PCA. Third, with such a dispersed set of outcomes, PCA is ideally suited because it maximizes the variance rather than minimizing the least square distance. One aim of our empirical work is to evolve a set of composite indices so as to include them as the causal variables consisting of developmental variables such as human resource, infrastructure, labour, market, openness and resources. Hence, we need to choose the essential variables by a data reduction procedure and arrive at relative weights 131

for the purpose of consolidating these variables into a single index. We chose Principal Components Analysis (PCA), which is popular in the literature since it has a number of desirable properties. It retains the maximum information, allows the composite of variables to remain uncorrelated amongst each other. The data reduction procedure involves selection of the most relevant variables that capture the maximum information and diverse information. Both the unrotated and rotated solutions explain exactly the same amount of variation in the variables. The choice between them hinges upon the interpretative power of each solution. The component scores (both rotated and unrotated), with respect to the first component are calculated. The most popular orthogonal rotation procedure is Kaiser s Varimax rotation. We therefore retain this procedure. The following consideration should be kept in mind while applying PCA: 1. For determining the retained component we need a criterion. 2. The PCA methodology tells us the total variance explained by each retained principal component as well as the cumulative percentage of the explained variation. This is a measure of the explanatory power of the component for the information content of the procedure. 3. There were various methods of rotation but the most popular method is the Varimax with the Kaiser normalization. The purpose of the rotation is to make the interpretation of the PCA more meaningful. Method of rotation however retains the same information and explanatory power. After doing these procedures there was a choice between retained principal components in a regression framework or selecting the principal variables that are associated with each of those components. This involves the Jolliffe procedure. In the first case regression is known as principal component regression and in the second case it is known as principal variable regression. We have chosen the latter because it is difficult to interpret the principal component regression. We have chosen to retain three components so that we finally land up with three principal variables. The reason is that using the Kaiser criterion of Eigen value less than one leaves only two components while retaining all seven variables leads to multicollinearity. 132

On the other hand eliminating some variables through PCA does not affect the explanatory power of the equation because the retained variables contain the information of those which are eliminated. We have used the Joliffe s procedure for selecting principal variables. We take up each rotated component and select the variable that has the highest component score. Then we move to the next component and so on. This way we get the three principal variables which represent the maximum information and eliminate the variables that are correlated to them and hence create multicollinearity. Method for Construction of the Index The method for construction of a composite index is given by Jha and Murthy (2006). Once the number of retained principal components is determined and the rotated component scores obtained, we have the choice of using the principal components as such or selecting certain sub-set of variables from the larger set of variables. Jolliffe proposes selecting one variable to represent each of the retained principal components. The variable that has the highest loading on a component is chosen to represent that component, provided that it has not already been chosen. If it has been chosen, the variable with the next largest loading is selected. The procedure starts with the largest principal component and proceeds to the smallest retained component 2. Index = 3 j wjxj Xj = retained variables Wj = component scores (weights). This procedure has been used for creating the following indices: 1. Index of Human Resources; 2. Index of Infrastructure 2 An alternative approach is to delete variables by using the discarded principal components. Starting with the smallest discarded component, the variable with the largest weight on that component is deleted. Then the variable with the largest loading on the second smallest component would be discarded. If the variable has previously been discarded, then the variable with the next highest loading would also be discarded. This procedure continues up to the largest such discarded component. 133

3. Index of Labour 4. Index of Market 5. Index of Trade Openness 6. Index of Resources 7.2. RESULTS OF PRINCIPAL COMPONENT ANALYSIS OF WORLD DEVELOPMENTAL VARIABLES 7.2.1. Human Resource Human resource represents skilled manpower, which would have an impact on the FDI patterns. For measuring the variable we have identified variables- Expenditure on Education (EDUX), Primary Education, Pupils (EDU_P) and Population (POPT) and it has been explained about how we shall be developing a composite index, that summaries the information contained in all these variables. It involves three steps: First step involve the Kaiser-Meyer-Olkin (KMO) test which tells us about the adequacy of sample and appropriateness of PCA as a methodology. In general, value of KMO test should be on higher side which represents good. In case of human resource, value is 0.45, which is just reasonable. Bartlett test measures sphericity which states about suitability of using of PCA. If it is statistically significant then it represents about suitability of using principal component analysis. In case of human resource, Bartlett test is highly statistically significant (Table 7.1a). Table 7.1: Results of Principal Component Analysis of Human Resource (a) KMO and Bartlett s Test Kaiser-Meyer-Olkin Measure of Sampling Adequacy.452 Approx. Chi-Square 2732.534 Bartlett s Test of Sphericity Df 3 Sig..000 Next step is to be found out number of principal components which are being retained. In this case, we got three variables and we have imposed the condition that all three variables have to be retained. Therefore total variance explained by these three variables is 100 percent. Table 7.1b gives the total explained variation captured by three retained components. 134

(b) Explained Component Initial Eigenvalues Extraction Sums of Squared Loadings of of 1 2.013 67.094 67.094 2.013 67.094 67.094 2.976 32.531 99.624.976 32.531 99.624 3.011.376 100.000.011.376 100.000 In the next step, we used Varimax rotation method, to arrive at rotated component score. This would enable us to have a better interpretation of components. Moreover it helps us in generating the value weights obtained from the factor loading for constructing the composite index. (c) Rotated Component Matrix Variable Component 1 2 3 EDU_P.997.030 -.074 POPT.993.085.076 EDUX.057.998.002 Rotated component scores of EDU_P, EDUX and POPT are 0.997, 0.998 and 0.076 respectively (Table 7.1c). These scores are used for construction of composite index of human resource. Composite Index of Human Resources I HR = 0.997EDU_P + 0.998EDUX + 0.076POPT 7.2.2. Infrastructure Infrastructure refers to the facilities through which others resources can be efficiently and optimally used. For measuring this variable we have identified following variables- Energy Production (ENP), Electricity Production (ELP), Air Transport (ATS), Air Transport-Passengers (ATP), Road Sector Energy Consumption (ROAD), Telephone Lines (TEL) and Telephone Lines (per 100 People) (TEL_P). It shall be used to develop composite index that summaries the information contained in all these variables. 135

KMO test tells us about the adequacy of sample and appropriateness of PCA as a methodology. In general, value of KMO test should be on higher side which represents good. The value of KMO test is 0.815, which is high and good. Bartlett test suggests infrastructure variable is highly statistically significant (Table 7.2a). Table 7.2: Results of Principal Component Analysis of Infrastructure (a) KMO and Bartlett s Test Kaiser-Meyer-Olkin Measure of Sampling Adequacy.815 Approx. Chi-Square 9192.486 Bartlett s Test of Sphericity Df 21 Sig..000 Next step is to be found out number of principal components which are being retained. In this case, we have seven variables and we have imposed the condition that three variables have to be selected. The retained variables are ROAD, TEL, and TEL_P. variance explained by these three variables is 96.56 percent. Table 7.2b gives the total explained variation captured by three retained components. (b) Explained Component Initial Eigenvalues Extraction Sums of Squared Loadings of of 1 5.290 75.572 75.572 5.290 75.572 75.572 2.993 14.187 89.759.993 14.187 89.759 3.479 6.840 96.599.479 6.840 96.599 4.151 2.164 98.763 5.065.929 99.691 6.014.199 99.891 7.008.109 100.000 Now, we used Varimax rotation method, to arrive at rotated component score. This would enable us to have a better interpretation of components. Moreover it helps us in generating the value weights obtained from the factor loading for constructing the composite index. 136

(c) Rotated Component Matrix Variable Component 1 2 3 ROAD.896.407.096 ATP.885.428.133 ATS.869.348.216 ELP.716.683.094 TEL.350.910.090 ENP.532.805 -.030 TEL_P.148.039.987 Rotated component scores of ROAD, TEL and TEL_P are 0.896, 0.910 and 0.987 respectively (Table 7.2c). These scores are used for construction of composite index of Infrastructure. Composite Index of Infrastructure I INFRA = 0.896ROAD + 0.910TEL + 0.987TEL_P 7.2.3. Labour Labour represents raw human work force. Cheap raw labour may influence cost side. For measuring this variable we have identified following variables- Employment, 15-24 age (EMPTEEN), Employment (EMP), GDP Per Person (GDPPC), Labour Participation Rate (LRATE), Labour Force, (LFT) and Population Working Ages (POPWA). It shall be used to develop composite index that provides the information contained in all these variables. KMO test tells us about the adequacy of sample and appropriateness of PCA as a methodology. In general, value of KMO test should be on higher side which represents good. The value of KMO test is 0.666, which is good. Bartlett test suggests labour variable is highly statistically significant (Table 7.3a). 137

Table 7.3: Results of Principal Component Analysis of Labour (a) KMO and Bartlett s Test Kaiser-Meyer-Olkin Measure of Sampling Adequacy.666 Approx. Chi-Square 2909.533 Bartlett s Test of Sphericity Df 15 Sig..000 Next step is to be found out number of principal components which are being retained. In this case, we have six variables and we have imposed the condition that three variables have to be selected. The retained variables are LRATE, LFT and POPWA. variance explained by these three variables is 89.01 percent (Table 7.3b). (b) Explained Component Initial Eigenvalues Extraction Sums of Squared Loadings of of 1 2.983 49.715 49.715 2.983 49.715 49.715 2 1.566 26.093 75.808 1.566 26.093 75.808 3.793 13.211 89.018.793 13.211 89.018 4.363 6.051 95.069 5.243 4.054 99.123 6.053.877 100.000 Next, we used Varimax rotation method, to arrive at rotated component score. This would enable us to have a better interpretation of components. Moreover it helps us in generating the value weights obtained from the factor loading for constructing the composite index. (c) Rotated Component Matrix Variable Component 1 2 3 LRATE.932.183.102 EMP.929.243.104 EMPTEEN.908 -.113.097 LFT.345.860.080 GDPPC.268 -.673.562 POPWA.076.002.969 138

Rotated component scores of LRATE, LFT and POPWA are 0.932, 0.860 and 0.969 respectively. These scores are used for construction of composite index of labour. Composites Index of Labour I LAB = 0.932LRATE + 0.860LFT + 0.969POPWA 7.2.4. Market Market is a place where production is used for consumption. For measuring this variable we have identified following variables- Market Capitalization of Listed Companies (MKTCAP), Listed Domestic Companies (COS), Population Density (POPDEN), Population in Largest City (POPL), Manufacturing-Value Added (MFWA), Industry-Value Added (INVA) and Services-Value Added (SVA). It shall be used to develop composite index that summaries the information contained in all these variables. KMO test tells us about the adequacy of sample and appropriateness of PCA as a methodology. In general, value of KMO test should be on higher side which represents good. The value of KMO test is 0.824, which is high and good. Bartlett test suggests market variable is highly statistically significant (Table 7.4a). Table 7.4: Results of Principal Component Analysis of Market (a) KMO and Bartlett s Test Kaiser-Meyer-Olkin Measure of Sampling Adequacy.824 Approx. Chi-Square 7823.383 Bartlett s Test of Sphericity Df 21 Sig..000 Next step is to be found out number of principal components which are being retained. In this case, we have seven variables and we have imposed the condition that three variables have to be selected. The retained variables are MKTCAP, POPL and POPDEN. variance explained by these variables is 91.37 percent (Table 7.4b). 139

(b) Explained Component Initial Eigenvalues Extraction Sums of Squared Loadings of of 1 4.734 67.628 67.628 4.734 67.628 67.628 2 1.072 15.318 82.947 1.072 15.318 82.947 3.590 8.424 91.371.590 8.424 91.371 4.432 6.171 97.542 5.135 1.928 99.470 6.034.487 99.957 7.003.043 100.000 We applied Varimax rotation method, to arrive at rotated component score. This would enable us to have a better interpretation of components. Moreover it helps us in generating the value weights obtained from the factor loading for constructing the composite index. (c) Rotated Component Matrix Variable Component 1 2 3 MKTCAP.962.093 -.043 SVA.961.200 -.012 INVA.907.354.052 MFVA.897.355.067 COS.695.363.141 POPL.328.929.106 POPDEN.023.092.993 Rotated component scores of MKTCAP, POPL and POPDEN are 0.962, 0.929 and 0.993 respectively. These scores are used for construction of composite index of Market (Table 7.4c). Composite Index of Market I MKT = 0.962MKTCAP + 0.929POPL + 0.993POPDEN 140

7.2.5. Trade Openness Trade openness refers to openness of domestic country for international trade activities. It facilitates free movement of goods and services amongst countries. For measuring the variable we have identified variables- reserves (TRES), Trade Openness (TOPEN) and Official exchange rate (EXCG) and it has been explained about how we shall be developing a composite index, that summaries the information contained in all these variables. KMO test which tells us about the adequacy of sample and appropriateness of PCA as a methodology. In general, value of KMO test should be on higher side which represents good. The value of KMO test is 0.485, which is reasonable. Bartlett test suggests trade openness variable is significant at 10 percent level of significance (Table 7.5a). Table 7.5: Results of Principal Component Analysis of Trade Openness (a) KMO and Bartlett s Test Kaiser-Meyer-Olkin Measure of Sampling Adequacy.485 Approx. Chi-Square 6.958 Bartlett s Test of Sphericity Df 3 Sig..073 Next step is to be found out number of principal components which are being retained. In this case, we got three variables and we have imposed the condition that all three variables have to be retained. Therefore total variance explained by these three variables is 100 percent. Table 7.5b gives the total explained variation captured by three retained components. (b) Explained Component Initial Eigenvalues Extraction Sums of Squared Loadings of of 1 1.086 36.191 36.191 1.086 36.191 36.191 2 1.020 34.002 70.193 1.020 34.002 70.193 3.894 29.807 100.000.894 29.807 100.000 141

In the next step, we used Varimax rotation method, to arrive at rotated component score. This would enable us to have a better interpretation of components. Moreover it helps us in generating the value weights obtained from the factor loading for constructing the composite index. (c) Rotated Component Matrix Variable Component 1 2 3 EXCG 1.000 -.014 -.020 TRES -.014.999 -.042 TOPEN -.020 -.043.999 Rotated component scores of EXCG, TRES and TRADE are 1.00, 0.999 and 0.999 respectively (Table 7.5c). These scores are used for construction of composite index of trade openness. Composite Index of Trade Openness I TOPN = 1.00EXCG + 0.999TRES + 0.999TOPEN 7.2.6. Resource Resource includes Gross Fixed Capital Formation (GFCF), Gross Domestic Products (GDP), GDP Per Capita (GDPPC), Gross Domestic Savings (GDS) and Natural Resources (TNRES). These resources are used to developed composite index, which gives information contained in these variables. KMO test which tells us about the adequacy of sample and appropriateness of PCA as a methodology. In general, value of KMO test should be on higher side which represents good. The value of KMO test is 0.560, which is good. Bartlett test suggests resource variable is highly significant. Table 7.6: Results of Principal Component Analysis of Resource (a) KMO and Bartlett s Test Kaiser-Meyer-Olkin Measure of Sampling Adequacy.560 Approx. Chi-Square 4820.623 Bartlett s Test of Sphericity Df 10 Sig..000 142

Next step is to be found out number of principal components which are being retained. In this case, we have five variables and we have imposed the condition that three variables have to be selected. The retained variables are GFCF, GDPPC and TNRES. variance explained by these variables is 98.02 percent (Table 7.6b). (b) Explained Component Initial Eigenvalues Extraction Sums of Squared Loadings of of 1 3.078 61.552 61.552 3.078 61.552 61.552 2 1.016 20.328 81.879 1.016 20.328 81.879 3.807 16.143 98.022.807 16.143 98.022 4.094 1.877 99.899 5.005.101 100.000 In the next step, we used Varimax rotation method, to arrive at rotated component score. This would enable us to have a better interpretation of components. Moreover it helps us in generating the value weights obtained from the factor loading for constructing the composite index. (c) Rotated Component Matrix Variable Component 1 2 3 GFCF.989.118 -.068 GDS.968.125 -.066 GDP.960.152 -.062 GDPPC.172.982 -.079 TNRES -.080 -.076.994 Rotated component scores of GFCF, GDPPC and TNRES are 0.989, 0.982 and 0.994 respectively (Table 7.6c). These scores loading are used for construction of composite index of resource. Composite Index of Resources I RES = 0.989GFCF + 0.982GDPPC + 0.994TNRES 143

7.3. RESULTS OF PRINCIPAL COMPONENT ANALYSIS OF DEVELOPED COUNTRIES DEVELOPMENTAL VARIABLES 7.3.1. Human Resource Human resource represents skilled manpower, which would have an impact on the FDI patterns. For measuring the variable we have identified variables- Expenditure on Education (EDUX), Primary Education, Pupils (EDU_P) and Population (POPT) and it has been explained about how we shall be developing a composite index, that summaries the information contained in all these variables. KMO test which tells us about the adequacy of sample and appropriateness of PCA as a methodology. In general, value of KMO test should be on higher side which represents good. The KMO test value is 0.766, which is high. Bartlett test is highly statistically significant. variables within human resource are only three. Hence, these three variables become principal variable. Therefore total variance explained by these variables is 100 percent. Human resource includes Expenditure on Education (EDUX), Primary Education, Pupils (EDU_P) and Population (POPT) as principal variables. Table 7.7: Results of Principal Component Analysis of Human Resource (a) KMO and Bartlett s Test Kaiser-Meyer-Olkin Measure of Sampling Adequacy.766 Approx. Chi-Square 2308.477 Bartlett s Test of Sphericity Df 3 Sig..000 Next step is to be found out number of principal components which are being retained. In this case, we got three variables and we have imposed the condition that all three variables have to be retained. Therefore total variance explained by these three variables is 100 percent. Table 7.7b gives the total explained variation captured by three retained components. 144

(b) Explained Component Initial Eigenvalues Extraction Sums of Squared Loadings of of 1 2.924 97.479 97.479 2.924 97.479 97.479 2.058 1.933 99.412.058 1.933 99.412 3.018.588 100.000.018.588 100.000 In the next step, we used Varimax rotation method, to arrive at rotated component score. This would enable us to have a better interpretation of components. Moreover it helps us in generating the value weights obtained from the factor loading for constructing the composite index. (c) Rotated Component Matrix Variable Component 1 2 3 EDU_P.797.582.163 POPT.733.588.342 EDUX.566.797.211 Rotated component scores of EDU_P, EDUX and POPT are 0.797, 0.797 and 0.342 respectively (Table 7.7c). These scores loading are used for construction of composite index of human resource. Composite Index of Human Resource I HR = 0.797EDU_P + 0.797EDUX + 0.342POPT 7.3.2. Infrastructure Infrastructure refers to the facilities through which others resources can be efficiently and optimally used. For measuring this variable we have identified following variables- Energy Production (ENP), Electricity Production (ELP), Air Transport (ATS), Air Transport-Passengers (ATP), Road Sector Energy Consumption (ROAD), Telephone Lines (TEL) and Telephone Lines (per 100 People) (TEL_P). It shall be used to develop composite index that summaries the information contained in all these variables. 145

KMO test tells us about the adequacy of sample and appropriateness of PCA as a methodology. In general, value of KMO test should be on higher side which represents good. The value of KMO test is 0.876, which is high and good. Bartlett test suggests infrastructure variable is highly statistically significant (Table 7.8a). Table 7.8: Results of Principal Component Analysis of Infrastructure (a) KMO and Bartlett s Test Kaiser-Meyer-Olkin Measure of Sampling Adequacy.876 Approx. Chi-Square 6588.390 Bartlett s Test of Sphericity Df 21 Sig..000 Next step is to be found out number of principal components which are being retained. In this case, we have seven variables and we have imposed the condition that three variables have to be selected. The retained variables are ROAD, TEL_P and ATS. variance explained by these three variables is 98.35 percent. Table 7.8b gives the total explained variation captured by three retained components. (b) Explained Component Initial Eigenvalues Extraction Sums of Squared Loadings of of 1 5.808 82.970 82.970 5.808 82.970 82.970 2.924 13.202 96.172.924 13.202 96.172 3.153 2.180 98.352.153 2.180 98.352 4.077 1.101 99.453 5.025.352 99.805 6.010.140 99.944 7.004.056 100.000 Now, we used Varimax rotation method, to arrive at rotated component score. This would enable us to have a better interpretation of components. Moreover it helps us in generating the value weights obtained from the factor loading for constructing the composite index. 146

(c) Rotated Component Matrix Variable Component 1 2 3 ROAD.989.103.077 ELP.983.123.096 ATP.979.116.116 ENP.974.133 -.117 TEL.950.153.159 ATS.890.137.417 TEL_P.125.992.025 Rotated component scores of ROAD, TEL_P and ATS are 0.989, 0.992 and 0.417 respectively (Table 7.8c). These scores loading are used for construction of composite index of Infrastructure. Composite Index of Infrastructure I INFRA = 0.989ROAD + 0.992TEL_P + 0.417ATS 7.3.3. Labour Labour represents raw human work force. Cheap raw labour may influence cost side. For measuring this variable we have identified following variables- Employment, 15-24 age (EMPTEEN), Employment (EMP), GDP Per Person (GDPPC), Labour Participation Rate (LRATE), Labour Force, (LFT) and Population Working Ages (POPWA). It shall be used to develop composite index that provides the information contained in all these variables. KMO test tells us about the adequacy of sample and appropriateness of PCA as a methodology. In general, value of KMO test should be on higher side which represents good. The value of KMO test is 0.667, which is good. Bartlett test suggests labour variable is highly statistically significant (Table 7.9a). 147

Table 7.9: Results of Principal Component Analysis of Labour (a) KMO and Bartlett s Test Kaiser-Meyer-Olkin Measure of Sampling Adequacy.667 Approx. Chi-Square 1532.005 Bartlett s Test of Sphericity Df 15 Sig..000 Next step is to be found out number of principal components which are being retained. In this case, we have six variables and we have imposed the condition that three variables have to be selected. The retained variables are EMP, LFT and POPWA. variance explained by these three variables is 85.41 percent (Table 7.9b). (b) Explained Component Initial Eigenvalues Extraction Sums of Squared Loadings of of 1 2.910 48.498 48.498 2.910 48.498 48.498 2 1.278 21.299 69.797 1.278 21.299 69.797 3.937 15.609 85.406.937 15.609 85.406 4.579 9.646 95.052 5.256 4.273 99.325 6.040.675 100.000 Next, we used Varimax rotation method, to arrive at rotated component score. This would enable us to have a better interpretation of components. Moreover it helps us in generating the value weights obtained from the factor loading for constructing the composite index. (c) Rotated Component Matrix Variable Component 1 2 3 EMP.961.129.004 LRATE.946.130 -.080 EMPTEEN.896.028 -.100 LFT.056.943.030 GDPPC.256.562 -.538 POPWA -.008.016.941 148

Rotated component scores of EMP, LFT and POPWA are 0.961, 0.943 and 0.941 respectively. These scores are used for construction of composite index of labour. Composite Index of Labour I LAB = 0.961EMP + 0.943LFT + 0.941POPWA 7.3.4. Market Market is a place where production is used for consumption. For measuring this variable we have identified following variables- Market Capitalization of Listed Companies (MKTCAP), Listed Domestic Companies (COS), Population Density (POPDEN), Population in Largest City (POPL), Manufacturing-Value Added (MFWA), Industry-Value Added (INVA) and Services-Value Added (SVA). It shall be used to develop composite index that summaries the information contained in all these variables. KMO test tells us about the adequacy of sample and appropriateness of PCA as a methodology. In general, value of KMO test should be on higher side which represents good. The value of KMO test is 0.824, which is high and good. Bartlett test suggests market variable is highly statistically significant (Table 7.10a). Table 7.10: Results of Principal Component Analysis of Market (a) KMO and Bartlett s Test Kaiser-Meyer-Olkin Measure of Sampling Adequacy.824 Approx. Chi-Square 5204.363 Bartlett s Test of Sphericity Df 21 Sig..000 Next step is to be found out number of principal components which are being retained. In this case, we have seven variables and we have imposed the condition that three variables have to be selected. The retained variables are MKTCAP, POPL and POPDEN. variance explained by these variables is 95.18 percent (Table 7.10b). 149

(b) Explained Component Initial Eigenvalues Extraction Sums of Squared Loadings of of 1 5.134 73.340 73.340 5.134 73.340 73.340 2 1.132 16.173 89.513 1.132 16.173 89.513 3.396 5.663 95.177.396 5.663 95.177 4.241 3.444 98.620 5.077 1.103 99.724 6.017.237 99.961 7.003.039 100.000 We applied Varimax rotation method, to arrive at rotated component score. This would enable us to have a better interpretation of components. Moreover it helps us in generating the value weights obtained from the factor loading for constructing the composite index. (c) Rotated Component Matrix Variable Component 1 2 3 MKTCAP.964.169 -.038 SVA.941.305.016 INVA.857.489.090 MFVA.847.493.108 COS.731.495 -.191 POPL.415.868.213 POPDEN -.014.111.985 Rotated component scores of MKTCAP, POPL and POPDEN are 0.964, 0.868 and 0.985 respectively (Table 7.10c). These scores are used for construction of composite index of Market. Composite Index of Market I MKT = 0.964MKTCAP + 0.868POPL + 0.985POPDEN 150

7.3.5. Trade Openness Trade openness refers to openness of domestic country for international trade activities. It facilitates free movement of goods and services amongst countries. For measuring the variable we have identified variables- reserves (TRES), Trade Openness (TOPEN) and Official exchange rate (EXCG) and it has been explained about how we shall be developing a composite index, that summaries the information contained in all these variables. KMO test which tells us about the adequacy of sample and appropriateness of PCA as a methodology. In general, value of KMO test should be on higher side which represents good. The value of KMO test is 0.458, which is reasonable. Bartlett test suggests trade openness variable is highly significant (Table 7.11a). Table 7.11: Results of Principal Component Analysis of Trade Openness (a) KMO and Bartlett s Test Kaiser-Meyer-Olkin Measure of Sampling Adequacy.458 Approx. Chi-Square 17.619 Bartlett s Test of Sphericity Df 3 Sig..001 Next step is to be found out number of principal components which are being retained. In this case, we got three variables and we have imposed the condition that all three variables have to be retained. Therefore total variance explained by these three variables is 100 percent. Table 7.5b gives the total explained variation captured by three retained components. (b) Explained Component Initial Eigenvalues Extraction Sums of Squared Loadings of of 1 1.172 39.074 39.074 1.172 39.074 39.074 2 1.052 35.073 74.147 1.052 35.073 74.147 3.776 25.853 100.000.776 25.853 100.000 151

In the next step, we used Varimax rotation method, to arrive at rotated component score. This would enable us to have a better interpretation of components. Moreover it helps us in generating the value weights obtained from the factor loading for constructing the composite index. (c) Rotated Component Matrix Variable Component 1 2 3 EXCG.998 -.029 -.062 TRES -.030.996 -.078 TOPEN -.062 -.079.995 Rotated component scores of EXCG, TRES and TRADE are 0.998, 0.996 and 0.995 respectively (Table 7.11c). These scores are used for construction of composite index of trade openness. Composite Index of Trade Openness I TOPN = 0.998EXCG + 0.996TRES + 0.995TOPEN 7.3.6. Resource Resource includes Gross Fixed Capital Formation (GFCF), Gross Domestic Products (GDP), GDP Per Capita (GDPPC), Gross Domestic Savings (GDS) and Natural Resources (TNRES). These resources are used to developed composite index, which gives information contained in these variables. KMO test which tells us about the adequacy of sample and appropriateness of PCA as a methodology. In general, value of KMO test should be on higher side which represents good. The value of KMO test is 0.610, which is good. Bartlett test suggests resource variable is highly significant (Table 7.12a). Table 7.12: Results of Principal Component Analysis of Resource (Developed Countries) (a) KMO and Bartlett s Test Kaiser-Meyer-Olkin Measure of Sampling Adequacy.610 Approx. Chi-Square 2903.366 Bartlett s Test of Sphericity Df 10 Sig..000 152

Next step is to be found out number of principal components which are being retained. In this case, we have five variables and we have imposed the condition that three variables have to be selected. The retained variables are GFCF, GDPPC and TNRES. variance explained by these variables is 98.86 percent (Table 7.12b). (b) Explained Component Initial Eigenvalues Extraction Sums of Squared Loadings of of 1 3.023 60.461 60.461 3.023 60.461 60.461 2 1.320 26.399 86.861 1.320 26.399 86.861 3.600 12.004 98.864.600 12.004 98.864 4.051 1.028 99.893 5.005.107 100.000 In the next step, we used Varimax rotation method, to arrive at rotated component score. This would enable us to have a better interpretation of components. Moreover it helps us in generating the value weights obtained from the factor loading for constructing the composite index. (c) Rotated Component Matrix Variable Component 1 2 3 GFCF.993.091 -.051 GDP.983.082 -.038 GDS.977.124 -.060 GDPPC.138.974.177 TNRES -.073.172.982 Rotated component scores of GFCF, GDPPC and TNRES are 0.993, 0.974 and 0.982 respectively (Table 7.12c). These scores are used for construction of composite index of resource. Composite Index of Resources I RES = 0.993GFCF + 0.974GDPPC + 0.982TNRES 153

7.4. RESULTS OF PRINCIPAL COMPONENT ANALYSIS OF DEVELOPING COUNTRIES DEVELOPMENTAL VARIABLES 7.4.1. Human Resource Human resource represents skilled manpower, which would have an impact on the FDI patterns. For measuring the variable we have identified variables- Expenditure on Education (EDUX), Primary Education, Pupils (EDU_P) and Population (POPT) and it has been explained about how we shall be developing a composite index, that summaries the information contained in all these variables. KMO test which tells us about the adequacy of sample and appropriateness of PCA as a methodology. In general, value of KMO test should be on higher side which represents good. The KMO test value is 0.552, which is good. Bartlett test is highly statistically significant (Table 7.13a). Table 7.13: Results of Principal Component Analysis of Human Resource (a) KMO and Bartlett s Test Kaiser-Meyer-Olkin Measure of Sampling Adequacy.552 Approx. Chi-Square 1279.916 Bartlett s Test of Sphericity Df 3 Sig..000 Next step is to be found out number of principal components which are being retained. In this case, we got three variables and we have imposed the condition that all three variables have to be retained. Therefore total variance explained by these three variables is 100 percent. Table 7.13b gives the total explained variation captured by three retained components. (b) Explained Component Initial Eigenvalues Extraction Sums of Squared Loadings of of 1 2.294 76.467 76.467 2.294 76.467 76.467 2.695 23.163 99.630.695 23.163 99.630 3.011.370 100.000.011.370 100.000 154

In the next step, we used Varimax rotation method, to arrive at rotated component score. This would enable us to have a better interpretation of components. Moreover it helps us in generating the value weights obtained from the factor loading for constructing the composite index. (c) Rotated Component Matrix Variable Component 1 2 3 EDU_P.975.210 -.071 POPT.965.250.078 EDUX.228.974.002 Rotated component scores of EDU_P, EDUX and POPT are 0.975, 0.974 and 0.078 respectively (Table 7.13c). These scores are used for construction of composite index of human resource. Composite Index of Human Resource I HR = 0.975EDU_P + 0.974EDUX + 0.078POPT 7.4.2. Infrastructure Infrastructure refers to the facilities through which others resources can be efficiently and optimally used. For measuring this variable we have identified following variables- Energy Production (ENP), Electricity Production (ELP), Air Transport (ATS), Air Transport-Passengers (ATP), Road Sector Energy Consumption (ROAD), Telephone Lines (TEL) and Telephone Lines (per 100 People) (TEL_P). It shall be used to develop composite index that summaries the information contained in all these variables. KMO test tells us about the adequacy of sample and appropriateness of PCA as a methodology. In general, value of KMO test should be on higher side which represents good. The value of KMO test is 0.778, which is high and good. Bartlett test suggests infrastructure variable is highly statistically significant (Table 7.14a). variables included are seven. Out of these variables, principal component analysis selects three variables as principal infrastructure variables. These principal variables are 155

Electricity Production (ELP), Telephone Lines per 100 persons (TEL_P) and Road Sector Energy Consumption (ROAD). variance explained by these variables is 96.6 percent. Table 7.14: Results of Principal Component Analysis of Infrastructure (a) KMO and Bartlett s Test Kaiser-Meyer-Olkin Measure of Sampling Adequacy.788 Approx. Chi-Square 3502.683 Bartlett s Test of Sphericity Df 21 Sig..000 Next step is to be found out number of principal components which are being retained. In this case, we have seven variables and we have imposed the condition that three variables have to be selected. The retained variables are ELP, TEL_P and ROAD. variance explained by these three variables is 95.15 percent. Table 7.14 gives the total explained variation captured by three retained components. (b) Explained Component Initial Eigenvalues Extraction Sums of Squared Loadings of of 1 5.061 72.293 72.293 5.061 72.293 72.293 2 1.323 18.898 91.191 1.323 18.898 91.191 3.277 3.955 95.146.277 3.955 95.146 4.188 2.685 97.831 5.104 1.481 99.312 6.037.523 99.835 7.012.165 100.000 Now, we used Varimax rotation method, to arrive at rotated component score. This would enable us to have a better interpretation of components. Moreover it helps us in generating the value weights obtained from the factor loading for constructing the composite index. 156

(c) Rotated Component Matrix Variable Component 1 2 3 ELP.965.156.182 ENP.947 -.031.197 TEL.934.213.088 ATP.889.325.264 ROAD.717.213.652 TEL_P -.036.961.142 ATS.539.788 -.028 Rotated component scores of ELP, TEL_P and ROAD are 0.965, 0.961 and 0.652 respectively (Table 7.14c). These scores are used for construction of composite index of Infrastructure. Composite Index of Infrastructure I INFRA = 0.965ELP + 0.961TEL_P + 0.652ROAD 7.4.3. Labour Labour represents raw human work force. Cheap raw labour may influence cost side. For measuring this variable we have identified following variables- Employment, 15-24 age (EMPTEEN), Employment (EMP), GDP Per Person (GDPPC), Labour Participation Rate (LRATE), Labour Force, (LFT) and Population Working Ages (POPWA). It shall be used to develop composite index that provides the information contained in all these variables. KMO test tells us about the adequacy of sample and appropriateness of PCA as a methodology. In general, value of KMO test should be on higher side which represents good. The value of KMO test is 0.671, which is good. Bartlett test suggests labour variable is highly statistically significant (Table 7.15a). 157

Table 7.15: Results of Principal Component Analysis of Labour (a) KMO and Bartlett s Test Kaiser-Meyer-Olkin Measure of Sampling Adequacy.671 Approx. Chi-Square 1575.516 Bartlett s Test of Sphericity Df 15 Sig..000 Next step is to be found out number of principal components which are being retained. In this case, we have six variables and we have imposed the condition that three variables have to be selected. The retained variables are EMP, LFT and POPWA. variance explained by these three variables is 92.42 percent (Table 7.15b). (b) Explained Component Initial Eigenvalues Extraction Sums of Squared Loadings of of 1 3.333 55.543 55.543 3.333 55.543 55.543 2 1.469 24.485 80.028 1.469 24.485 80.028 3.743 12.388 92.415.743 12.388 92.415 4.230 3.835 96.250 5.179 2.989 99.239 6.046.761 100.000 Next, we used Varimax rotation method, to arrive at rotated component score. This would enable us to have a better interpretation of components. Moreover it helps us in generating the value weights obtained from the factor loading for constructing the composite index. (c) Rotated Component Matrix Variable Component 1 2 3 EMP.960.077.198 LRATE.940.089.155 EMPTEEN.919.201 -.070 GDPPC -.031 -.902.337 LFT.535.694.349 POPWA.115 -.139.960 158

Rotated component scores of EMP, LFT and POPWA are 0.960, 0.694 and 0.960 respectively (Table 7.15c). These scores are used for construction of composite index of labour. Composite Index of Labour I LAB = 0.960EMP + 0.694LFT + 0.960POPWA 7.4.4. Market Market is a place where production is used for consumption. For measuring this variable we have identified following variables- Market Capitalization of Listed Companies (MKTCAP), Listed Domestic Companies (COS), Population Density (POPDEN), Population in Largest City (POPL), Manufacturing-Value Added (MFWA), Industry-Value Added (INVA) and Services-Value Added (SVA). It shall be used to develop composite index that summaries the information contained in all these variables. KMO test tells us about the adequacy of sample and appropriateness of PCA as a methodology. In general, value of KMO test should be on higher side which represents good. The value of KMO test is 0.547, which is good. Bartlett test suggests market variable is highly statistically significant (Table 7.16a). Table 7.16: Results of Principal Component Analysis of Market (a) KMO and Bartlett s Test Kaiser-Meyer-Olkin Measure of Sampling Adequacy.547 Approx. Chi-Square 606.027 Bartlett s Test of Sphericity Df 10 Sig..000 Next step is to be found out number of principal components which are being retained. In this case, we have seven variables and we have imposed the condition that three variables have to be selected. The retained variables are MKTCAP, POPDEN and POPL. variance explained by these variables is 88.57 percent (Table 7.16b). 159

(b) Explained Component Initial Eigenvalues Extraction Sums of Squared Loadings of of 1 2.342 46.838 46.838 2.342 46.838 46.838 2 1.315 26.295 73.133 1.315 26.295 73.133 3.772 15.440 88.573.772 15.440 88.573 4.426 8.514 97.087 5.146 2.913 100.000 We applied Varimax rotation method, to arrive at rotated component score. This would enable us to have a better interpretation of components. Moreover it helps us in generating the value weights obtained from the factor loading for constructing the composite index. (c) Rotated Component Matrix Variable Component 1 2 3 MKTCAP.951.091.103 MFVA.937.058.202 POPDEN.065.925 -.067 COS.088.764.433 POPL.216.102.940 Rotated component scores of MKTCAP, POPDEN and POPL are 0.951, 0.925 and 0.940 respectively (Table 7.16c). These scores loading are used for construction of composite index of Market. Composite Index of Market I MKT = 0.951MKTCAP + 0.925POPDEN + 0.940POPL 7.4.5. Trade Openness Trade openness refers to openness of domestic country for international trade activities. It facilitates free movement of goods and services amongst countries. For measuring the 160

variable we have identified variables- reserves (TRES), Trade Openness (TOPEN) and Official exchange rate (EXCG) and it has been explained about how we shall be developing a composite index, that summaries the information contained in all these variables. KMO test which tells us about the adequacy of sample and appropriateness of PCA as a methodology. In general, value of KMO test should be on higher side which represents good. The value of KMO test is 0.493, which is reasonable. Bartlett test suggests trade openness variable is not significant (Table 7.17c). Table 7.17: Results of Principal Component Analysis of Trade Openness (a) KMO and Bartlett s Test Kaiser-Meyer-Olkin Measure of Sampling Adequacy.493 Approx. Chi-Square 3.044 Bartlett s Test of Sphericity Df 3 Sig..385 Next step is to be found out number of principal components which are being retained. In this case, we got three variables and we have imposed the condition that all three variables have to be retained. Therefore total variance explained by these three variables is 100 percent. Table 7.17b gives the total explained variation captured by three retained components. (b) Explained Component Initial Eigenvalues Extraction Sums of Squared Loadings of of 1 1.093 36.428 36.428 1.093 36.428 36.428 2 1.008 33.613 70.042 1.008 33.613 70.042 3.899 29.958 100.000.899 29.958 100.000 In the next step, we used Varimax rotation method, to arrive at rotated component score. This would enable us to have a better interpretation of components. Moreover it helps us in generating the value weights obtained from the factor loading for constructing the composite index. 161

(c) Rotated Component Matrix Variable Component 1 2 3 TRES 1.000 -.007 -.016 TOPEN -.007.999 -.046 EXCG -.016 -.046.999 Rotated component scores of TRES, TRADE and EXCG are 1.00, 0.997 and 0.999 respectively (Table 7.17c). These scores are used for construction of composite index of trade openness. Composite Index of Trade Openness I TOPN = 1.00TRES + 0.999TOPEN + 0.999EXCG 7.4.6. Resource Resource includes Gross Fixed Capital Formation (GFCF), Gross Domestic Products (GDP), GDP Per Capita (GDPPC), Gross Domestic Savings (GDS) and Natural Resources (TNRES). These resources are used to developed composite index, which gives information contained in these variables. KMO test which tells us about the adequacy of sample and appropriateness of PCA as a methodology. In general, value of KMO test should be on higher side which represents good. The value of KMO test is 0.666, which is good. Bartlett test suggests resource variable is highly significant (Table 7.18a). Table 7.18: Results of Principal Component Analysis of Resource (a) KMO and Bartlett s Test Kaiser-Meyer-Olkin Measure of Sampling Adequacy.666 Approx. Chi-Square 2513.542 Bartlett s Test of Sphericity Df 10 Sig..000 Next step is to be found out number of principal components which are being retained. In this case, we have five variables and we have imposed the condition that three variables have to be selected. The retained variables are GFCF, GDPPC and TNRES. variance explained by these variables is 99.09 percent (Table 7.18b). 162