Friday, October 9, 2009

My GDP forecast...

... for last year, as computed by the model, called VAR(30), that I am presenting as a Master's Dissertation. Excerpt from the results section, followed by the forecast the model would have made back in December of 2007:

The annual pseudo out-of-sample forecasts (performed as of December of each year) were compared to the (December) Wall Street Journal Economic Forecasting Surveys since 2004. VAR(30) would have been the only one to predict a 2008 recession in December, 2007. Over the short, four year period (2005 through 2008) available for a benchmark, VAR(30) outperforms on average all of the private-sector forecasters polled by the Wall Street Journal. This figure is however skewed to the upside by the 2008 extreme outperformance: excluding the 2008 recession year, the rank falls to the 16th place, out of 42 (only forecasters who provided GDP growth estimates for all of the four years were included in this ranking, and thus there can be survivorship bias in the list). This illustrates the importance of Renshaw’s “right ballpark estimate”. Specifically, VAR(30) ranks 15th out of 59 for the 2005 GDP forecast, 41st out of 59 for 2006, 20th out of 63 for 2007, and 1st out of 54 for 2008.
Also presented are fan charts of each yearly pseudo out-of-sample forecasts made each December. Fan charts represent the 90%, 75% and 50% confidence bands around the central GDP forecast.


(click to enlarge)

No comments:

Post a Comment