A system reduction method to efficiently solve DSGE models

The paper presents a system reduction method (SRM) to improve the computational time to solve a large class of dynamic stochastic general equilibrium (DSGE) models with the methods of Anderson and Moore (1985), Klein (2000), Sims (2002) or Uhlig (1995). I measure the efficiency gains with seven models ranging from 47 to 333 equations. The time reduction for the Anderson–Moore algorithm aim ranges from 10% to 71%; Klein's function solab reduces its time between 51% and 79%; the time reduction for Sims' function gensys increases from 25% to 59%; Uhlig's function solve reduces its time between 31% and 87%. The time reduction can be crucial for Bayesian estimation of medium to large scale models.