Does government stimulus spending create jobs?

1. James Pethokoukis @JimPethokoukis

This from the St. Louis Fed offers another look at the theory (bold is mine):

Looking at government spending more generally suffers from the problem that the spending may be correlated with economic activity: The government may spend more during a recession (as with ARRA) or more during an expansion (when tax revenues are high). This might bias the results, which economists call “an endogeneity bias.” Military spending, on the other hand, is likely to be determined primarily by international geopolitical factors rather than the nation’s business cycle. Specifically, we used a dataset developed in pioneering work by Valerie Ramey and, on several of the papers, her co-authors Michael Owyang and Sarah Zubairy.

To control for potential “anticipation effects,” Owyang, Ramey and Zubairy used historical documents to construct a time series of military spending news shocks. This allowed them to disentangle the time of military spending from when the public learned that military spending was going to change in the future.

Those authors looked at the output response to the spending shocks. They found a small effect of military spending on output and that the size of this effect did not depend on whether the economy was slack or not.3 Specifically, a $1 increase in government spending caused a less than $1 increase in gross domestic product (GDP).
We used a similar methodology and found that military spending shocks had a small effect on civilian employment. Following a policy change that began when the unemployment rate was high, if government spending increased by 1 percent of GDP, then total employment increased by between 0 percent and 0.15 percent. Following a policy change that began when the unemployment rate was low, the effect on employment was even smaller.
In the event of another recession, policymakers have a number of stabilization tools at their disposal, including quantitative easing, negative interest rates and tax relief. The research discussed above suggests that one other device, namely countercyclical government spending, may not be very effective, even when the economy is slack.