We discuss the implications of nonlinear dynamic phenomena for policy definition in a growing economy. De- parting from the hypothesis that local analysis of economic systems focus the definition of policy on short run outcomes, we propose that to have a clearer perspective on long run outcomes, we have to focus the analysis on the study of local bifurcations and global dynamics instead. This approach, in our opinion, provides a better insight on complex macroeconomic phenomena and allows for a better definition of policy rules in a long run horizon. To demonstrate this hypothesis, we set up a representative agent economy based on neoclassical assumptions, where agents face convex risk premium and investment adjustment costs. This economy describes an endogenous opti- mal growth model, that has solutions given by a nonlinear three-dimensional dynamical system. To evaluate this system, we resort to qualitative analysis methods and show the existence of fold (saddle-node), hopf and fold-hopf bifurcations, in a multiple equilibria environment. Numerical results suggest the absence of local stable solutions for a wide range of parameter values. We then focus our analysis on the complex organization of the economy phase-space and evaluate several conjectures related to the existence of complex nonlinear phenomena in the vicinity of fold-hopf bifurcation points. We relate these conjectures with the hypothesis of endogenous structural change and discuss the implications of complex global dynamic phenomena for long run policy definition.