Using rule-based updating procedures to improve the performance of composite indicators

Ideally, the set of variables underlying composite indicators is checked and updated when needed on a regular basis. In practise, the timing and procedures of these updates are usually chosen ad hoc. We suggest a rule-based indicator selection updating procedure, performed at regular intervals, which reduces the arbitrariness of this process. We apply this procedure to one of the most prominent targeted composite leading indicator for Switzerland, which is based on bivariate associations of potential variables with a reference series reflecting the Swiss growth rate cycle. We show that in a simulated real-time analysis the targeted indicator selection procedure outperforms the widely used approach to combine as many potential variables as possible. Furthermore, the regular updating procedure preserves the leading properties of the composite indicator with respect to the reference time series, as compared to the same composite indicator without such updates.