Aggregate Output Measurements: A Common Trend Approach

The authors analyze a model for N different measurements of persistent latent time series when measurement errors are mean-reverting, which implies common trend among measurements. study the consequences overdifferencing, finding potentially large biases in maximum likelihood estimators (MLE) dynamics parameters and reductions precision smoothed estimates variable, especially multiperiod objects such as quinquennial growth rates. also develop an R2 measure observability that determines severity misspecification. Finally, apply their framework to US quarterly data on GDE GDI, obtaining improved aggregate output measure.