My 72 year model of the US economy (1921-41, 1946-96) has a 99.9%
correlation with the actual results using only five factors, military
spending, deficit, long cycle, trade and oil. When I separately model
the GNP in the WWII years, that also works very well. This model is
based on predicting manufacturing productivity, something not measured
during the WWII years due to the national emergency. You can see that
the red "model" and black "actual" lines match up almost exactly in each
subcycle, know as the Juglar cycle of 8-10 years. More information is in the book Peace Economics.