This paper focuses on testing long run macroeconomic relations for interest rates, equity, prices and exchange rates suggested by arbitrage in financial and goods markets. It uses the global vector autoregressive (GVAR) model to test for long run restrictions in each country/region conditioning on the rest of the world. Bootstrapping is used to compute both the empirical distribution of the impulse responses and the log-likelihood ratio statistic for over-identifying restrictions. The paper also examines the speed with which adjustments to the long run relations take place via the persistence profiles. We find strong evidence in favour of the UIP and to a lesser extent the Fisher equation across a number of countries, but our results for the PPP are much weaker. Also the transmission of shocks and subsequent adjustments in financial markets are much faster than those in goods markets.