Did the ARRA Hire the Unemployed or Just Shift Jobs Around the Employed?

The Meracatus Center recently published a study by Garrett Jones and Daniel Rothschild titled, “Did Stimulus Dollars Hire the Unemployed?” It uses surveys and seeks to point out flaws in the achievements of the ARRA, or Obama’s stimulus program. Matt Yglesias, Kevin Drum, and Jon Chait have already commented on it, pointing out that the results of the study seem to look like a major win for the stimulus.

Specifically, the major debate is over this data point: stimulus-funded projects hired 42.1 percent from the pool of unemployed people. The authors of the paper argue that this is a problem, as it is similar to the rates of hiring from the unemployed that have been observed from the 1990s-2000s. Let’s take a look at their argument (my bold, numbering):

Worker responses

Did stimulus-funded projects hire the unemployed or the already employed? Our surveys indicate a near-tie on this question. Of the 277 respondents hired after January 31, 2009, 42.1 percent had been unemployed immediately beforehand and 47.3 percent had come directly from another job. Of the rest, 4.1 percent had been out of the labor force, and 6.5 percent had been in school. Thus, the weight of the evidence suggests that ARRA did an enormous amount of “job shifting” rather than “job creating.” There is evidence of the latter, but, under Keynesian reasoning, every worker hired away from another job reflects some weakening of the stimulus.

We saw this “worker poaching” tendency in our interviews as well. [1] This is similar to the amount of job shifting that goes on in relatively normal economic times. Eva Nagypal (2008, p.1) notes that “employer-to-employer transitions…ma[de] up 49 percent of all separations from employers” in the decade prior to her study. Robert Hall (2005) finds a similar number, roughly 40 percent. [2] Since on average separations equal hires (the minor factor of net job growth aside), there is little difference between the recent U.S. average and our sample average. In other words, we find little evidence that stimulus spending was particularly effective at moving the unemployed into work. During the worst recession in generations, the ARRA receiving organizations in our sample hired away employed workers at roughly the same rate as in normal economic times.

Firstly, from [1] it is clear that they compare the ARRA job hires to the job hiring environment from studies that look at the pre-recession environment, roughly 1994-2007. They justify this by [2], which says that separations roughly equal hires during the relative time periods, so there should be “little difference” between the two samples.

That second point is clearly false. A quick random search found this graph from this site using JOLTS:

During 2008-2009, or one of the time periods in question, separations were significantly higher than hires, which should require us to re-estimate what a “normal” hiring distribution looks like. And that baseline should look like the rest of the economy in the depressed state during 2008-2009.

The flows between unemployment and employment and within the two themselves are distorted relative to any reasonable baseline, and as such a direct comparison isn’t trivial. I’m not going to re-estimate a baseline here, but I have two clues that indicate what should be expected. First, transitions from Unemployed to Employed have plummeted:

This recession is characterized by a major increase in the difficulty the unemployed have in finding a job. If the stimulus is associated with a pre-recession rate of hiring from the unemployed, we should consider that a major win.

What about the rate at which those already employed find a job? This is harder to track, but this graph from The Labor Market and the Great Recession (April 2010) by Elsby, Hobijn, and Sahin sheds some light:

This is complicated, so let me take a second to explain. The separation rate is the rate at which people quit their jobs or suffer a layoff. The inflow rate is the rate at which they enter unemployment (flow into the unemployment category from the employed category). The layoff separation and inflow rate are almost identical in the recession: quits decline, and the rate at which quits flow into unemployment decline even faster.

This last part makes sense — you aren’t going to quit your job in a labor market with 10 percent unemployment and a unemployment to vacancy ratio of 6-to-1 unless you are absolutely sure you have the next job lined up. As such, my initial reaction to this graph is that the amount of hiring from Employed -> Employed should go up — the gap between the separation and inflow line have increased slightly, it seems.

These are looking at different spaces within the labor market, so comparisons between these two data points and the Jones/Rothschild study aren’t simple. However, there has been a major change in the movement between categories here and my initial reaction is that keeping a consistent ratio of hiring the unemployed at a time characterized by a plummeting Unemployed -> Employed rate should be a major win, no? That’s before a probable increase in employed people separating only when they have new jobs and the new quirk of a massive increase in the ‘Not in the Labor Force’ population (which hasn’t been conclusively studied in the past). At the very least, a relatively full employment market, like that of the 1990s-2000s, can’t serve as a benchmark.

That’s my reaction; what do readers with more familiarity with the labor market readers see in this study?