The Stacked-Time Simulator in TROLL: A Robust Algorithm for Solving Forward-Looking Models

Peter Hollinger ()
Additional contact information
Peter Hollinger: Intex Solutions, Inc., Needham, MA

Computing in Economics and Finance 1996 from Society for Computational Economics

Abstract: The TROLL econetric modelling system provides two methods to solve forward-looking ``rational expectations models with model-consistent endogenous leads. The traditional Fair-Taylor algorithm solves the model through a specified time horizon by treating the leads as exogenous and then iterating over that process until the leads converge. Fair-Taylor can be considered a sort of ``Gauss-Seidel- over- time. Like Gauss-Seidel, Fair- Taylor may work very well, or it may have trouble converging. Larger shocks, more time periods, or tighter convergence criteria can greatly increase the number of iterations and computer time required to converge or can result in failure to converge. When convergence is successful, the ``solution values may have errors that are large relative to the convergence criterion. One advantage of Fair-Taylor is that it does not require much computer memory. The alternative Stacked-Time algorithm ``stacks all the time periods into one large system of equations and solves them simultaneously using Newton-Raphson. For typical macroeconometric models, Newton-Raphson is usually an efficient and very robust method. With quadratic convergence near the solution, the number of iterations is barely affected by the convergence criteria. The number of iterations also does not seem to change substantially with the number of time periods or the size of the shock, although in some cases large shocks may cause numerical problems such as ``log of a negative number. At convergence, solutions are generally accurate relative to the convergence criterion.

View list of references

Downloads: (external link)
http://www.unige.ch/ce/ce96/ps/hollinge.eps (application/postscript)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: http://EconPapers.repec.org/RePEc:sce:scecf6:_026

Access Statistics for this paper

More papers in Computing in Economics and Finance 1996 from Society for Computational Economics
Address: Department of Econometrics, University of Geneva, 102 Bd Carl-Vogt, 1211 Geneva 4, Switzerland
Contact information at EDIRC.
Series data maintained by Christopher F. Baum ().