US composite economic indicator with nonlinear dynamics and the data subject to structural breaks

Composite economic indicator is a very useful tool designed to trace and predict the business cycle conditions. This paper studies possible extensions of this approach intended to cope with the potential data problems caused by various structural breaks affecting both level and volatility of the component series. The structural shifts are introduced in the composite economic indicator model via deterministic dummies capturing breaks in the observed variables' intercepts and in the residual variances of the specific factors. As an illustration the Post-Second World War US monthly macroeconomic series are utilized for which different specifications of the single-factor linear and regime-switching model are evaluated.