What works in fiscal policy.

In the New York Times Greg Mankiw reviews empirical evidence that suggests tax cuts work as fiscal policy, while increases in government spending do not.

The results are striking. Successful stimulus relies almost entirely on cuts in business and income taxes. Failed stimulus relies mostly on increases in government spending.
All these findings suggest that conventional models leave something out. A clue as to what that might be can be found in a 2002 study by Olivier Blanchard and Roberto Perotti. (Mr. Perotti is a professor at Boccini University in Milano, Italy; Mr. Blanchard is now chief economist at the International Monetary Fund.) They report that “both increases in taxes and increases in government spending have a strong negative effect on private investment spending. This effect is difficult to reconcile with Keynesian theory.”

You might almost think that the supply side people were on to something.