A Question for Zandi and Blinder

Back in the days of the ARRA debate, a great deal was made of a study by Mark Zandi and Alan Blinder that described the effects of the stimulus using a complex economic model.   I don't want to renew the whole stimulus debate but a post on a different topic caused me to dust off the Zandi/Blinder paper and look at some of the forecasts for GDP that were in the model.

Here are five interesting numbers for you.  The first four are forecast GDP in 2012 (in $2005 dollars) according to the model.

No policy response (no fiscal or monetary policy) - $12620
Monetary stimulus (no fiscal policy) - $14216
Fiscal stimulus (no monetary policy) - $13774
Both fiscal and monetary stimulus - $14552

Actual GDP in 2012 - $13593

In other words, the GDP in 2012 was nearly 7% smaller than was predicted by the Zandi/Blinder model in 2010 and indeed was nearly 5% smaller than predicted by the model in a world where the stimulus never happened.  These are large differences to be sure and in a very short period of time.

My question for the authors is "why?"

A couple of answers occur to me. 

1. Fiscal and monetary policy were far more restrictive than envisioned by the model for 2011 and 2012.  While possible, this seems highly unlikely because smaller deficits were forecast for 2011 and 2012 in 2010 than actually occurred.  To the degree the authors were using public data, they would have assumed more austerity on the government side than actually existed.
2. The model's long term effects are either flawed or undermined by exogenous factors.  Again, while this is possible, it seems unlikely, particularly given the magnitude of the observed differences.
3. The model isn't very good and its conclusions are suspect.  In the absence of an alternative, this seems the most likely answer.

I'm quite sure I'll not get a reply from the authors but it just goes to show that periodically checking on the predictions of models is a good way to think about validating them, albeit it's often after the smoke of the debate has cleared.