Extracting GDP signals from the monthly indicator of economic activity

Abstract

Real-time data are analysed for information on the Chilean monthly economic activity indicator IMACEC and what it indicates of the final GDP, defined as the growth rate that has been subject to at least two annual revisions. Data are presented and revisions analysed briefly. Mincer-Zarnowitz tests suggest that forecast rationality is rejected with respect to the three-month IMACEC growth rate as a nowcast of the first released quarterly GDP, as well as the first published GDP as a nowcast of the final GDP. An out-of-sample nowcasting analysis was conducted using only data which were available in real-time. The results show that small models nowcast better than less parsimonious ones. The evidence from the empirical study suggests no improvement in the nowcasting performance when historical data are supplemented with the first monthly IMACEC of the quarter. On the other hand, when two monthly observations IMACEC are available, the root mean squared nowcast error (RMSNE) decreases by 24%, and a further decline of 33% is obtained when the third monthly observation of the quarter is published. Both of these advances are statistically significant. No further improvement is obtained with the publication of the first release of the quarterly GDP. JEL classifications: C89, E17 Keywords: Real-time data, data revisions, nowcasting