Having trouble with the arsenal of democracy multipliers

I thought the issue was how much does the total economy grow when central government expands spending.  So I look at the [ Blue Line] change in total GDP over the change in central government spending.  During the pre-war period we see the classic result,  just prior to the spending, the ratio is high.  But as soon as the central government spending ramps up, the ratio drops rapidly.  I keep seeing this in the data, and economists actually report this, both Romer and Taylor see this drop off though they disagree on how fast the multiplier drops off.   The drop off is natural, if it never dropped off then it would not be a stimulus, but a validation of socialism.

As reported the drop off in GDP starts in late 1941 when we hit constraints, but just prior to that we have GDP growing in the face of central government spending. Spending starts in 1940, and immediately we see the ratio drop, but it takes a year for GDP to drop.  Remember, the national banking system takes a half a year to equilibriate, and the other sectors take longer.  So, yes, we can con folks into a short term GDP burst, but is is merely a con based on our need to have adaption time.  The rapid drop off in the ratio confirms Taylor and Company more than Romer and Company.