This paper, which I obviously think is worth noting, is forthcoming in AEJ Macroeconomics:
A Pitfall with DSGE-Based, Estimated, Government Spending Multipliers, by Patrick Fève, Julien Matheron, Jean-Guillaume Sahuc, December 5, 2012: 1 Introduction Standard practice in estimation of dynamic stochastic general equilibrium (DSGE) models, e.g. the well-known work by Smets and Wouters (2007), is to assume that government consumption expenditures are described by an exogenous stochastic process and are separable in the households’ period utility function. This standard practice has been adopted in the most recent analyses of fiscal policy (e.g. Christiano, Eichenbaum and Rebelo, 2011, Coenen et al., 2012, Cogan et al., 2010, Drautzburg and Uhlig, 2011, Eggertsson, 2011, Erceg and Lindé, 2010, Fernández-Villaverde, 2010, Uhlig, 2010).
In this paper, we argue that both short-run and long-run government spending multipliers (GSM) obtained in this literature may be downward biased. This is so because the standard approach does not typically allow for the possibility that private consumption and government spending are Edgeworth complements in the utility function and that government spending has an endogenous countercyclical component (automatic stabilizer)... Since, as we show, the GSM increases with the degree of Edgeworth complementarity,... the standard empirical approach may ... result in a downward-biased estimate of the GSM.
In our benchmark empirical specification with Edgeworth complementarity and a countercyclical component of policy, the estimated long-run multiplier amounts to 1.31. Using the same model..., when both Edgeworth complementarity and the countercyclical component of policy are omitted,... the estimated multiplier is approximately equal to 0.5. Such a difference is clearly not neutral if the model is used to assess recovery plans of the same size as those recently enacted in the US. To illustrate this more concretely, we feed the American Recovery and Reinvestment Act (ARRA) fiscal stimulus package into our model. We obtain that omitting the endogenous policy rule at the estimation stage would lead an analyst to underestimate the short-run GSM by slightly more than 0.25 points. Clearly, these are not negligible figures. ...
1 We say that private consumption and government spending are Edgeworth complements/substitutes when an increase in government spending raises/diminishes the marginal utility of private consumption. Such a specification has now become standard, following the seminal work by Aschauer (1985), Bailey (1971), Barro (1981), Braun (1994), Christiano and Eichenbaum (1992), Finn (1998), McGrattan (1994).
Let me also add these qualifications from the conclusion:
In our framework, we have deliberately abstracted from relevant details... However, the recent literature insists on other modeling issues that might potentially affect our results. We mention two of them. First, as put forth by Leeper, Plante and Traum (2010), a more general specification of government spending rule, lump-sum transfers, and distortionary taxation is needed to properly fit US data. This richer specification includes in addition to the automatic stabilizer component, a response to government debt and co-movement between tax rates. An important quantitative issue may be to assess which type of stabilization (automatic stabilization and/or debt stabilization) interacts with the estimated degree of Edgeworth complementarity. Second, Fiorito and Kollintzas (2004) have suggested that the degree of complementarity/substitutability between government and private consumptions is not homogeneous over types of public expenditures. This suggests to disaggregate government spending and inspect how feedback rules affect the estimated degree of Edgeworth complementarity in this more general setup. These issues will constitute the object of future researches.