Evaluation of Previous Post’s GDP Forecast – Off By 1/5 of 1%

In a previous posts the 4th quarter GDP was forecasted using an Autoregressive Intergrated Moving Average model. Using the Box-Jenkins methodology including model identification, seasonal adjustments, estimation and diagnostic testing a simple model was calculated via regression for the purposes of forecasting.

Previous post with GDP forecast:


The first 4th quarter’s GDP numbers were recently published by the Federal Reserve bank and the model did a fairly good job at forecasting this single quarter.  The model suggested that 4th quarter GDP would be 14,900 million dollars and the actual number was 14,870 million dollars.  This difference represents a forecast error of about of less than 1/5 of 1% which is a pretty close the expected number.

Federal Reserve of Actual GDP Numbers:


Once the first quarter is over and the GDP numbers are published I will follow up on those forecast to again analyze how accurate the model is.  The most difficult part about forecasting is recognizing the sharp turns that are coming, given the recently sluggish growth with little external shocks the economy has been experiencing the accuracy isn’t surprising.  It will be interesting to see how the model stands up over time and whether or not a multivariate time series approach might increase it’s accuracy.