Social security, self-control problems and unknown preference parameters

We develop a general equilibrium model with overlapping generations to show that Social Security may increase welfare in dynamically efficient economies where agents are affected by self-control problems à la Gul and Pesendorfer (2001, Econometrica 69, 1403). In calibrating the model to the US economy, we make no assumption on agents’ preference parameters and set them to match target levels of capital-output and consumption-output ratios. Our simulationsinform that Social Security improves welfare with degrees of temptation not below 11%.We also find support for a program with a tax rate similar to the one in the real US economy.