International Trade, Economic Freedom, and Foreign Direct Investment in OECD Countries: A Panel ARDL Approach with Mixed Integration Orders and Structural Breaks

Abstract

This study examines the long‑run equilibrium and predictive Granger causality among international trade, economic freedom (EF), and foreign direct investment (FDI) in 20 OECD member countries over 2000–2022. Using annual panel data, we address methodological challenges including cross‑sectional dependence, mixed orders of integration (FDI is I(0) while trade and EF are I(1)), structural breaks, and country heterogeneity. We employ second‑generation panel unit root tests (CIPS) that account for cross‑sectional dependence, and a Panel ARDL (Pooled Mean Group) estimator that accommodates mixed integration orders. We test the homogeneity assumption of long‑run coefficients using the Hausman test (PMG vs. MG). Granger causality is examined using a modified Dumitrescu‑Hurlin test that respects the different orders of integration: FDI in levels, trade and EF in first differences. We also include control variables (GDP per capita, inflation, political stability, capital account openness), structural break dummies (2008‑2009 crisis, 2020‑2021 pandemic), and we winsorize extreme outliers (FDI in Ireland). Country heterogeneity is assessed via slope homogeneity tests and interaction terms rather than low‑power country‑specific cointegration tests.

Our main findings are: (i) CIPS tests confirm mixed orders: FDI is I(0), trade and EF are I(1). (ii) The Panel ARDL error correction term is negative and significant, confirming a long‑run equilibrium relationship without requiring all variables to be I(1). (iii) The Hausman test does not reject homogeneity (p = 0.27), but the test has low power given N=20, T=23, and the slope homogeneity test is borderline (p = 0.104). We therefore report Mean Group (MG) estimates as the primary specification, which reveal substantial cross‑country heterogeneity (EF→TRADE coefficients range from –0.09 to 0.61). (iv) Long‑run effects (Mean Group estimates, primary specification): a 1% increase in trade openness is associated with a 0.36% increase in FDI inflows (p = 0.066); a 1‑point increase in EF is associated with a 0.31% increase in trade (p = 0.10). The direct EF‑FDI effect is weak and not statistically significant (0.13, p = 0.21). Pooled Mean Group estimates (Appendix), which assume homogeneous coefficients, yield more precise estimates (0.38 and 0.33, both p<0.01) but mask substantial cross‑country heterogeneity. (v) Structural break dummies are significant for 2008‑2009 (negative effect on trade and FDI) but the system returns to equilibrium (ECT = -0.28 for trade, -0.15 for FDI). (vi) Residual cross‑sectional dependence tests and CUSUM stability tests are satisfied.

The study contributes a methodologically rigorous analysis of the trade‑EF‑FDI nexus in OECD countries, explicitly handling mixed integration orders, structural breaks, and heterogeneity. Policy implications emphasize improving regulatory efficiency and rule of law, leveraging trade as a driver of institutional reform, coordinating trade and investment policies, and adopting crisis‑resilient strategies