International business cycle co-movement: the role of FDI

Abstract

This article investigates the relationship between FDI and business cycle synchronization in the period 1982 to 2011 for eight industrialized countries. We find that more synchronized business cycles are associated with stronger FDI relations in the period 1995 to 2011, but not before 1995. More intensive FDI links are also associated with a greater vulnerability to lagged output spillovers from abroad. Our findings suggest that FDI has become a separate channel through which economies may affect each other and that FDI stocks are now an essential aspect of economic interdependence.

Keywords:

• business cycle synchronization
• international linkages
• trade
• FDI
• vertical integration

JEL Classification:

• F21
• F44

Views expressed here are those of the authors and do not necessarily reflect official positions of De Nederlandsche Bank.

Notes

¹ See Appendix A.1 for details on all of the data series employed in this article. FDI data for a shorter period of time are available for Switzerland (1993–2011) and Japan (1996–2011). We include these countries in the analysis at a later stage (see Table 2). The complete data set is available for download from our personal webpage on http://www.dnb.nl.

² In theory, the stock of inward FDI of Canada originating from Australia according to Canada’s statistical records should be the same as the stock of outward FDI of Australia invested in Canada according to Australia’s records. However, in practice, these numbers sometimes differ substantially.

³ Many of the instruments have been used before in the literature. Our inclusion of labour market regulations is motivated by Nicoletti et al. (2003) and Dewit et al. (2009), EMU-membership by Petroulas (2007) and Schiavo (2007) and FDI regulations by Nicoletti et al. (2003). See Appendix A.1 for the complete list of instruments.

⁴ As said before, the country-specific dummies will take account of systematic methodological differences between national FDI statistics.

⁵ We applied the following procedure. We created an auxiliary data set of 120 observations in which pairs consisting of reporting countries appear once and pairs involving a nonreporting country appear twice. We then estimated by TSLS. The effective doubling of the sample will not affect the point estimates of the coefficients, but will understate the SEs. The relevant correction factor is (n = 60 and k = 10). For the White statistic, the correction factor is 0.5.

⁶ In Tables 1–3, we also report a White test for heteroscedasticity which usually indicates that heteroscedasticity is not a significant issue. However, to be on the safer side, we consistently use heteroscedasticity consistent SEs throughout the article.

⁷ Judged by the Schwartz information criterion (SIC), we found virtually no country pairs for which m was larger than 2. Furthermore, to avoid a discontinuous data series ρ_L is based on the point estimate of the λ’s irrespective of statistical significance.