There are many studies of the stimulus, but finally there is one which goes behind the numbers to see what really happened. And it’s not an entirely pretty story. My colleagues Garett Jones and Daniel Rothschild conducted extensive field research (interviewing 85 organizations receiving stimulus funds, in five regions), asking simple questions such as whether the hired project workers already had had jobs. There are lots of relevant details in the paper but here is one punchline:
…hiring people from unemployment was more the exception than the rule in our interviews.
In a related paper by the same authors (read them both), here is more:
Hiring isn’t the same as net job creation. In our survey, just 42.1 percent of the workers hired at ARRA-receiving organizations after January 31, 2009, were unemployed at the time they were hired (Appendix C). More were hired directly from other organizations (47.3 percent of post-ARRA workers), while a handful came from school (6.5%) or from outside the labor force (4.1%)(Figure 2).
One major problem with ARRA was not the crowding out of financial capital but rather the crowding out of labor. In the first paper there is also a discussion of how the stimulus job numbers were generated, how unreliable they are, and how stimulus recipients sometimes had an incentive to claim job creation where none was present. Many of the created jobs involved hiring people back from retirement. You can tell a story about how hiring the already employed opened up other jobs for the unemployed, but it’s just that — a story. I don’t think it is what happened in most cases, rather firms ended up getting by with fewer workers.
There’s also evidence of government funds chasing after the same set of skilled and already busy firms. For at least a third of the surveyed firms receiving stimulus funds, their experience failed to fit important aspects of the Keynesian model.
This paper goes a long way toward explaining why fiscal stimulus usually doesn’t have such a great “bang for the buck.” It raises the question of whether as “twice as big” stimulus really would have been enough. Must it now be four times as big? The paper also sets a new standard for disaggregated data on this macro question, the data are in a zip file here.