Different studies provide surprisingly a large variety of controversial conclusions aboutthe forecasting power of an indicator, even when it is supposed to forecast the sametime series. In this study we aim to provide a thorough overview of linear forecastingtechniques and draw conclusions useful for the identification of the predictive relationshipbetween leading indicators and time series. In a case study for Germany we forecastfour possible representations of industrial production. Further on we consider alarge variety of time-varying specifications: ex post vs. ex ante, rolling vs. recursive andmodel specifications such as restricted vs. unrestricted, AIC vs. BIC vs. OSC, direct vs.indirect. In a horse race with nine leading indicators plus benchmark we demonstrate thevariance of assessment across target variables and forecasting settings (50 per horizon).We show that it is nearly always possible to find situations in which one indicatorproved to have better predicting power compared to another.
Forecasting competition leading indicators model selection