Semester

Fall

Date of Graduation

2002

Document Type

Dissertation

Degree Type

PhD

College

Chambers College of Business and Economics

Department

Economics

Committee Chair

Ronald Balvers.

Abstract

This thesis illustrates the effects of outliers on forecasting and on simulated general equilibrium models. The first essay demonstrates that using robust-estimation techniques may greatly improve long-run autoregressive forecasts if the model to be estimated is generated from a heavy-tailed distribution. The second essay continues with a similar theme. It shows that in the context of random contamination models, automatic autoregressive selection criteria may be severely inaccurate. If the estimated model is misspecified, forecasting accuracy may be considerably reduced. Once again, robust-estimation techniques limit the influence of the aberrant data---in this case from the contaminating distribution. The third essay demonstrates that calibrations, and thus simulation results, for International Real Business Cycle models can change if aberrant data is accounted for by using robust-estimation techniques. In the particular case examined in this essay, there is more evidence for higher levels of persistence in technology than indicated by OLS.

Share

COinS