If the government spends a dollar to "stimulate" the economy, that dollar is coming from somewhere. The government is thus destimulating the economy at the point from where it is obtaining the money for its "stimulus."
Since the money obtained is being taken away from where it otherwise would be put into personal exchange, we know it has a negative impact at that point.
On the other hand, the money that is then cycled through a group of politicians may end up on any type of boondoggle. And even if the politicians were sincere in their attempt to properly stimulate the economy, there is still no way that they can say, in the way you can in a personal exchange, that all sides benefited, since the group the money came from (ultimately taxpayers or those squeezed by inflation) have no direct say in how the money is spent. Thus, the greatly hailed Keynesian multiplier effect could very well be negative.
In a column at FT, economist Tim Harford almost gets to this point in his analysis on the topic, but not quite. Nevertheless, his insights and perspective on the subject are worth reading, especially his observation on the zero to negative characteristics of the multiplier. Read the column here.