We revisit the evidence of the existence of a long-run link between financial intermediation and economic growth, by testing of cointegration between the growth rate of real GDP, control variables and three series reflecting financial intermediation. We consider a model with a factor structure that allows us to determine whether the finance-growth link is due to cross countries dependence and/or whether it characterises countries with strong heterogeneities. We employ techniques recently proposed in the panel data literature, such as PANIC analysis and cointegration in common factor models. Our results show differences between the developed and developing countries. We run a comparative regression analysis on the 1980-2006 period and find that financial intermediation is a positive determinant of growth in developed countries, while it acts negatively on the economic growth of developing countries.