Some people think that if one takes into account intertemporal dynamics, policy must necessarily be ineffective
In defense of Heritage, reader Jeff opines:
Zero crowding out? You can’t be serious? Perhaps it is you who should re-visit new keynesian models. A quick review of, say Smets-Wouters (2007), will reveal to you significant crowding out effects. But I’ll give you the benefit of the doubt and assume you did not just completely misunderstand mainstream new keynesian models and move on.
This quote is representative of a “read the abstract” or “read the conclusion” and skip the rest approach to fighting policy debates, so I thought it useful to step back and think through why certain papers have certain results, and whether they are directly relevant to the issues at hand.
First, some historical perspective. Early on in the Fresh Water/Salt Water debates, an oft heard argument was that New Classical models demonstrated that rational expectations invalidated the proposition of efficacious discretionary monetary policy. However, economists like Stanley Fischer and John Taylor demonstrated that price (or wage) rigidity in the presence of rational expectations still provided a role for policy. In other words, price flexibility drove the policy ineffectiveness proposition. Hence, the New Classical enterprise was partly a great marketing program employing, as in a card trick, a bit of clever misdirection. A similar attempt to confuse the debate of policy efficacy is going on now (not necessarily by academics, but more typically by those trying to influence policy), and there are many people eager to fall into the trap.
The issue here, as I have repeatedly stated, is not the inclusion of intertemporal aspects, or optimizing behavior; all sorts of DSGEs can produce different results.  (previously discussed in this post) This point is illustrated in this survey of DSGEs used by central banks. Even then, these models indicate a positive impact on GDP (rather than a zero multiplier as say the folks at Heritage would contend).
Figure 4, Panel A from Coenen, et al. (2010), United States: Effect of 1 Year of Fiscal Stimulus on Real GDP (Instrument: Government Consumption), no monetary accommodation.
The models are European Economic Commission (EC QUEST), European Central Bank New Area Wide Model (ECB NAWM), Federal Reserve Board’s Sigma (Fed Sigma), IMF’s Global Integrated Monetary and Fiscal model (IMF GIMF), Federal Reserve Board’s FRB/US and Bank of Canada’s Global Economy Model (BoC GEM).
One key assumption regards the monetary policy reaction function. Smets-Wouter (2007), a Bayesian estimated model, incorporates a Taylor rule (as do many papers, especially pre-2008). Cogan, Cwik, Taylor and Wieland (2009), often cited as a case of low fiscal multipliers, assume two years of monetary accommodation. Just to remind readers, this is what the Fed funds rate looks like.
Source: Deutsche Bank, Global Economic Perspectives, December 19, 2012.
The US policy rate has been flat at zero for nearly four years, and is projected to stay flat by Deutsche Bank for at least another year.
And this is what several DSGEs indicate the response of GDP looks like for a two year accommodation (not the four year accommodation we’ve actually experienced).
Figure 4, Panel C from Coenen, et al. (2010), United States: Effect of 1 Year of Fiscal Stimulus on Real GDP (Instrument: Government Consumption), 2 years of monetary accommodation.
I think it of interest that that cumulative output typically exceeds the one year government consumption impulse, so the multiplier exceeds unity. In all cases, the multiplier exceeds zero.
That’s crowding out of output. What about crowding out of investment? Certainly, the no-accommodation scenario seems to confirm the fiscal-policy skeptics’ view:
Figure 6, Panel A from Coenen, et al. (2010), United States: Effect of 1 Year of Fiscal Stimulus on Consumption and Investment (Instrument: Government Consumption), no monetary accommodation.
But what about the more relevant case where monetary policy has been accommodative?
Figure 6, Panel C (consumption on left, investment on right) from Coenen, et al. (2010), United States: Effect of 1 Year of Fiscal Stimulus on Consumption and Investment (Instrument: Government Consumption), 2 years of monetary accommodation.
In this scenario, one sees only one model predicting a decline in investment eventually (so the net effect looks close to zero) – that’s the ECB’s New Area Wide Model. The two Fed models, FRB/US and Sigma, indicate substantial crowding in.
By the way, see this post to see how John Taylor and Robert Barro used the counterfactual approach, so criticized by groups like Heritage. And for those who want a nuanced survey of the New Keynesian/New Classical landscape, see Simon Wren-Lewis’s posts  .