Testing the interpretation of indices in a macroeconomic index model

This paper discusses and contrasts different approaches to the construction of dynamic models containing unobservables. It is argued that some analysis is easy to carry out using Dynamic Factor Analysis (DFA), a frequency domain technique, while other analysis is easier to carry out in the context of a Dynamic Multiple Indicator-Multiple Cause Model (DYMIMIC), a time domain technique. An example is presented in which six macroeconomic variables are found to be connected by two common factors using DFA. The interpretation of these factors as 'anticipated' and 'unanticipated' aggregate demand is then tested using the DYMIMIC model. This interpretation is strongly rejected.