Simulating computable overlapping generations models with TROLL

Abstract

In this paper, we examine the performance of the TROLL stacked-time algorithm in the simulation of large scale overlapping generations (OLG) models. At each period of time, the number of equations is proportional to the individual length of lifetime. The model size and the data requirements may thus be very large. Given the repetitive structure of the equations, we show how TROLL specific macrocommands can be used to explode a generic version (in which the lifetime is parameterized) into a complete model. A similar technique applies to explode the initial dataset on the whole simulation horizon. The stability properties of the model are derived and the stacked-time algorithm performances are checked for a large scale model with endogenous labour supply and uncertain lifetime. It turns out that TROLL performances are very attractive even for a system of 300000 simultaneous equations. © 2004 Kluwer Academic Publishers.