Cointegration testing in dependent panels with breaks

In this paper we test for the existence of a long-run relationship between investment and savings (the Feldstein-Horioka puzzle) in a panel of 18 OECD countries, 1970-2007, allowing for heterogenous breaks in the coefficients. For this purpose we develop a bootstrap panel cointegration test with breaks robust to cross-section dependence. The test suggests that, even allowing for breaks in the countries where capital control regulations changed in the sample period, there is no evidence of an investment-savings long-run relationship for the panel as a whole.