Does a temporary increase in government expenditure increase output in the standard new-Keynesian DSGE model? Is it possible to draw any intuitive parallels between the impact of government expenditure in these model and in the IS-LM model?y
The "standard" New Keynesian model could be many things, but suppose that we're dealing with the basic log-linearized 3-equation model (intertemporal Euler equation, New Keynesian Phillips curve, and monetary policy rule) exhibited, for instance, in Gali's textbook.
In most variants of this simple model, a temporary increase in government expenditure will almost always increase output, but the intuition is different from the Old Keynesian IS/LM model. The extent of the increase depends on the monetary policy rule.
If the monetary policy rule targets inflation of exactly zero (assuming trend inflation is also zero), then it turns out that there is no difference between the effects of government expenditure in the NK model and the effects in a purely real model; by successfully targeting zero inflation, we effectively negate the effect of the NK model's nominal frictions. Since the underlying real model produces an increase in output when government spending increases (ignoring effects of possible distortionary taxation used to fund this spending), we get this in the NK model with inflation targeting as well, but it's not very exciting. It's a purely classical story: all else equal, consumers work harder when their consumption is lower, and thus consumption does not drop one-for-one with increases in government spending.
If the monetary policy rule happens to target a constant real interest rate, then in the basic NK model the "multiplier" is exactly 1: the constant path for the real interest rate leaves consumption unchanged, and the increase in output exactly equals the increase in spending. If the monetary policy rule targets a constant nominal interest rate during the stimulus, then there is an additional effect because future stimulus spending produces expected inflation that lowers the real interest rate and boosts consumption. This is the source of "multipliers" >1 in the basic NK framework. See Woodford's 2011 AEJ macro paper for more details on this.
None of this comports with the traditional Keynesian 1/(1-MPC) multiplier. This is for a couple reasons, but most important is the fact that the basic NK model features fully Ricardian consumers who intertemporally smooth consumption and anticipate future taxes levied to pay for current spending. Gali, Lopez-Salido and Valles's 2007 JEEA comes much closer to the traditional story by adding ad-hoc "non-Ricardian" households.