CMStatistics 2018: Start Registration
View Submission - CFE
A0233
Title: U.S. aggregate output measurement: A common trend approach Authors:  Tincho Almuzara - CEMFI (Spain)
Gabriele Fiorentini - University of Florence (Italy) [presenting]
Enrique Sentana - CEMFI (Spain)
Abstract: Signal-extraction techniques are applied to US GDP and GDI to produce an improved aggregate output measure, emphasising the presence of a common trend in levels whose absence would imply an empirically implausible diverging statistical discrepancy. We also study the consequences of ignoring this common trend, which can introduce important biases in maximum likelihood estimators and considerably reduce the Kalman smoother precision. Our theoretical and Monte Carlo results characterise the severity of misspecification as inversely proportional to an $R^{2}$ measure of common trend observability. This $R^{2}$ is high in the data (1952Q1-2015Q4). Therefore, we conclude misspecification is small but not negligible.