In a soon to be published paper, several economists at the International Monetary Fund report estimates of government spending multipliers which are much smaller than those previously reported by the U.S. Administration. In order to obtain the estimates the IMF economists use a very large complex model called the Global Integrated Monetary and Fiscal (GIMF) Model developed by Douglas Laxton and his colleagues at the IMF . The paper is quite technical, but the bottom line summary is that a one percent increase in government purchases (as a share of GDP) increases GDP by a maximum of 0.7 percent and then fades out rapidly. This means that government spending crowds out other components of GDP (investment, consumption, net exports) immediately and by a large amount. 
The IMF estimate is much less than the multiplier reported in a paper released last year by Christina Romer of the President’s Council of Economic Advisers and Jared Bernstein of the Vice President’s Office. The attached graph shows how huge the difference is. It shows the impact on GDP of a one percentage point permanent increase in government purchases as a share of GDP reported in the IMF paper (labeled GIMF) and in the Administration paper (labeled Romer-Bernstein).
John Cogan, Volker Wieland, Tobias Cwik and I raised questions about Romer-Bernstein paper soon after it was released last year because the estimates seemed to be much different from comparable estimates based on more modern new Keynesian models. We classified the Romer-Bernstein estimates as old Keynesian. Since then many technical papers have been written on this subject, of which a recent paper by Michael Woodford is the most comprehensive in my view. The IMF model is of the new Keynesian variety and adds more evidence of the huge policy differences between new Keynesian and old Keynesian models.
