Cluster analysis on determinants of foreign direct investment and determination of homegeneous country groups for 1980-2005 period

In this study, it is aimed to determine the factors affecting foreign directinvestment (FDI) globally. Most of the variables studied here, are the subjects ofinternational researches on FDI determinants. Four periods are used to study thevariables. The general period of the study is from 1980 to 2005. Then three subperiodsare used: first period from 1980 to 1989 years second period from 1990to1999 years and third period from 2000 to 2005 years. Cluster Analysis is the maintechnique to examine the variables. Countries are clustered according to selectedvariables. While using cluster analysis, statistical meaning of the cluster division isexamined by using discriminant analysis.The cluster including the USA and developed European countries aredetermined according to any variable used in this study, and this cluster is called astop level foreign investment structure. This cluster stayed as the top level one in allperiods without any passing to other clusters. In addition, the number of countries inthis cluster has been increased during 2000 and 2005 period. Turkey, Poland,Portugal, Slovakia, Mexico, South Korea, Czech Republic and Hungary are thecountries came into top level cluster in this period. The same countries were in thecluster just under the top level in 1990 ? 1999 period. When the cluster members arechecked in 1980 ? 2005 and 1980 ? 1989 periods, top level cluster only covers theUSA and developed European countries.When the countries grouped in low and lower-middle level of foreigninvestment structure, it is observed that variables used here do not neatly determinethe clusters at these levels. Lower-middle group consists of less developed andAfrican countries. In this group for the period from 2000 to 2005 it is possible to seecountries like Belarus, Bosnia, Tajikistan, Ukraine and Kazakhstan. However, thiscluster is very unstable from one period to another. Low level cluster shows similarbehavior to lower-middle one. Low level cluster is a little bit healthier than lowermiddleone according to stability, but still passing through clusters from one period toanother is possible.For the whole periods studied here per capita GNP, developmentclassification and being an emerging market or not, OECD Membership, number oflandlines and mobile phones for each 1000 people are the most effective variables indetermining the clusters.Keywords: Foreign Direct Investment, Investment Development Path, ClusterAnalysis, Multivariate Analysis of Variance, Discriminant Analysis