Estimation of a Small Macro-model under the Assumption of Contaminated Distributions.

Robust methods have seldom been implemented with real economic data. This paper tries to fill the gap by estimating a small macro-model by the bounded-influence instrumental variables (BIV) estimator. The BIV estimates yield a standard error of the regression that is smaller than its least squares counterpart and a reduction in errors forecast. The BIV results signal instability in the last period of the sample. This instability is particularly evident in the financial sector of the economy, where BIV uncovers a case of a poorly modeled structural break. Thus, the paper also provides an example of the value of BIV as a diagnostic tool. Copyright 1991 by Blackwell Publishing Ltd