Wednesday, September 16, 2009
This morning's Wall Street Journal carries Carl Bialik's column on how to count stimulus jobs, and does a fine job taking you through the math. He begins:
So somewhere in the next 30 days we will have the state of Minnesota (through the MMB department). According to the federal Recovery.gov site, Minnesota has so far expended $1.47 billion of the $4.72 billion approved for the state. Now once upon a time I estimated that the Administration's calculation was that if they spent $93,333 they had "created or saved" one job. So the current estimate based on that $1.47b is that currently stimulus has created or saved 15,766 jobs (rounding up). Steward notes that the original proposal from the Administration indicated the state would earn 66,000 jobs, but to be fair that was over two years. Patience, Tom!
As the $787 billion federal stimulus package was being deliberated by Congress in February, the White House estimated that the act would increase employment by 3.5 million jobs, including 24,000 combined in New Hampshire and Wyoming.
So far, though, those states say the stimulus has added fewer than 1,000 jobs.
Less than a month from now, when every state receiving stimulus funds will be required to make such a report, the numbers will fall far short of White House projections -- whether it's the original 3.5 million job projection or the latest estimate, issued by the White House last week, that one million jobs have been created thus far by the stimulus act.
But Oberstar is also playing funny with the data in claiming that 2900 transportation jobs in Minnesota have already been created. Using my same divisor would make the $99 million spent by the Dept. of Transportation (most of which goes to highway funds) give us 1,061 jobs. Maybe I am not calculating that correctly, so I read on in Bialik's article.
Here's that CEA report. The report is much more careful than Oberstar in noting the tenuous nature of these estimates. However, Bialik explains the method they use is different because the White House numbers will include the multiplier effect, which data from Recovery.gov or the state sites will not.
The most visible figures available to evaluate the job market are unemployment rates, which don't speak well for the stimulus package. The national rate of joblessness last month was 9.7%, up from 8.5% in March, the month after the stimulus act was passed. A week after that number was released, the White House's Council of Economic Advisers reported that the stimulus had increased employment to a level by "slightly more than 1 million jobs higher than it otherwise would have been."
That awkward wording says a lot: It reflects the tough job facing any economist who tries to estimate job creation. In every method used, economists are forced to imagine an alternate reality -- one built on assumptions that are easily challenged. For example, to compare present unemployment rates to past rates may be straightforward but it fails to account for other economic forces that were going to affect unemployment with or without the stimulus.
The White House method assumes that things were getting worse and that the stimulus is the sole factor responsible for stopping the bleeding. So economists imagined an alternative reality whereby the present would have been much worse -- to the tune of one million more lost jobs.
I would disagree with Bialik's characterization of the multiplier as "standard". First, the 1.57 multiplier is a long-run effect; as I've already noted, the additional jobs one would expect from, say, the lunch truck operator who hires another driver because manufacturing plants are adding new shifts, are a long-run thing. The one-quarter multiplier is 1.05 by the estimates of the Administration, and the long run takes a year. They are taking credit for jobs they expect will happen in the future. Even 1.05 might be too high, as Casey Mulligan argues, but again he's looking at an impact multiplier versus a long-run multiplier. It could be 0.2 for one quarter and 1.57 over six quarters.
To arrive at that larger projection, the White House council applied standard multipliers -- reflecting that some spending has ripple effects through the economy, such as hired people increasing their consumption -- to components of the stimulus. For instance, increasing government spending by 1% of GDP would increase the GDP by 1.57%, which is the average of the Federal Reserve's estimate and that of a private firm.
Then the council converted the projected GDP growth into job figures, assuming that a 1% increase in GDP would reduce unemployment by 0.75% -- an assumption based on historical economic figures.
That assumption is based on a decades-old principle developed by Arthur Okun, former chair of the Council of Economic Advisers under Lyndon Johnson. Though Mr. Okun's name wasn't cited in the council's projection, his famous law tying GDP to employment was clearly in effect. The economist proposed that changes in the two were directly proportional -- nearly 50 years ago.
The council had doubts about its numbers for the tax-cut portion of the stimulus, writing in its report, "We confess to considerable uncertainty about our choice of multipliers for this element of the package."
Let me also point to a Time article this week regarding the CEA's use of Okun's Law. The relationship between output and employment has gotten tenuous. Bialik picks up on this as well, and ends up finding the state-by-state figures at least cleaner, if not completely covering all the good things the stimulus might have done.
One last point to make here. As Menzie Chinn noted last month, fiscal policy multipliers are much larger when the monetary authority is accommodating, and you can't be more accommodating than when you are at the zero interest bound. And so it comes as no surprise that Chairman Bernanke is also singing the praises of recovery without notice of what is happening in the local economy (my comments to the local newspaper in that link.)