Moving beyond traditional one- or possibly two-way causality involving foreign direct investment (FDI), a systematic approach is implemented for delineating both short- and long-run flows of causality involving FDI and a comprehensive set of FDI's possible determinants. Granger causality procedures incorporating error correction terms are implemented, using provincial panel data from China. In both the short and long run, growth in GDP directly influences FDI, while growth in local infrastructure and local investment provide indirect but not direct influence.
Foreign direct investment (FDI) and its relationship with economic growth have led to numerous empirical studies (see DeMello, 1997 and DeMello, 1999). Traditionally, in the literature, one or possibly two-way causality involving FDI and one factor such as GDP, local investment and infrastructure or quality of labor have been examined. Our intention is to widen the discussion by providing a systematic approach for considering all possible flows of causality involving FDI and the other variables of influence. We outline and implement a sequential econometric methodology that firstly determines whether the relevant variables of potential interest are stationary or integrated, that secondly identifies the exogeneity and endogeneity of these variables, that thirdly determines whether relationship among variables are long or short run in nature, and that finally establishes the relevant causality relationships. We do this in the context of regional panel data from China over the period, 1995–2010. China has the highest FDI among developing countries, but within China a great deal of diversity exists across regions. While not limited to the coastal provinces of China, traditionally most of the FDI inflows have been in the coastal region. Furthermore, with a limited length of reliable annual data in China, a regional panel approach provides statistical efficiency benefits.
While originally the focus in the literature had been on the one-way causal link from FDI to growth, following criticisms such as Kholdy (1995) and others, newer studies emerged allowing for the possibility of two-way causality among FDI and growth. That is, FDI can Granger cause GDP growth and GDP growth could also affect the inflow of FDI. Nevertheless, it soon becomes apparent that possible linkages inevitably involve other determinants. For example, the contribution of FDI to growth is influenced by such factors as domestic investment, technology and skills of the labor force (e.g., Apergis et al., 2006, Gholami et al., 2006, Kartircioglu and Naraliyeva, 2006 and Sun et al., 2002).
Indeed, how these other determinants influence FDI and the growth in GDP undoubtedly can shape the prescriptive policies that lead to targeted growth and investment. In searching for the relevant interaction and causality among FDI, growth in GDP and other determining factors, this study proposes a comprehensive search strategy for uncovering the most suitable dynamic specification. While section 2 of this paper will provide a number of plausible explanations of the relationship between growth in FDI and possible determinants, ultimately we have to go to the data to statistically identify the appropriate endogenous variables, control variables and specification that best describes these relationships. Section 3 documents the definitions of the data variables and the data sources. Section 4 provides a preliminary exploration of what variables are exogenous and what are endogenous, the nature of the stationarity of the variables, and whether the variables are cointegrated. Taking into account the findings of Section 4, Section 5 presents the causality exploration strategy and Section 6 presents the empirical results of the Granger causality tests and policy implications.
This paper has extended the discussion of causality involving FDI by providing a systematic approach for examining all possible flows of causality involving FDI, both in the short run and in the long run. This is the first systematic attempt to examine causality in both the short and long run for a comprehensive set of variables affecting and being affected by FDI. For FDI and its relationship with many of the key economic variables that we examine, this is an important issue. We dichotomize the short and long run impacts for such variables which have significant implications on policy related matters. Both in the short and long run we find FDI has a significantly positive effect on GDP, which in turn has a significantly positive impact on local investment and telecommunications. Consistent with the market size hypothesis, in the short run, FDI is positively affected by GDP. However, in the long run, FDI is not directly or indirectly influenced by local investment or telecommunications, but has a direct significantly positive impact on GDP. Another contribution of this study has been to provide a systematic way of dealing with causality in the context of panel data analysis. As noted in the paper, there are difficulties and challenges in dealing with causality tests in the context of panel data.
