DeLong's Stimulus Accounting: A Deconstruction
Ever since the Obama administration released its report detailing the number of jobs "saved or created" from the $787 billion stimulus package, cynics have been having a field day playing with the numbers. Investigations have revealed that many of the respondents' claims of job creation are bogus, and simple division shows that some states received more than $500,000 in stimulus awards for each job "saved or created."
Not surprisingly, UC Berkeley Keynesian economist Brad DeLong defended the stimulus plan and actually one-upped the Obama official who was defending the numbers. Yet as we'll see, I think DeLong himself pulls figures out of the air to justify the program. The whole affair serves to remind the public that conventional macroeconomics is an art, not a science — and a heavily politicized art to boot.
DeLong's Dubious Display
Before delving into DeLong's numbers, we have to first set up the context. DeLong quotes from ABC News' White House correspondent Jake Tapper who wrote,
Posting its results late this afternoon at Recovery.gov, the White House claimed 640,329 jobs have been created or saved because of the $159 billion in stimulus funds allocated as of Sept. 30.…
The White House argues that the actual job number is actually larger than 640,000 — closer to 1 million jobs when one factors in stimulus jobs added in October and, more importantly, jobs created indirectly, such as "the waitress who's still on the job," Vice President Biden said today.
So let's see. Assuming their number is right — 160 billion divided by 1 million. Does that mean the stimulus costs taxpayers $160,000 per job? Jared Bernstein, chief economist and senior economic advisor to the vice president, called that "calculator abuse." He said the cost per job was actually $92,000 — but acknowledged that estimate is for the whole stimulus package as of the end of 2010.
Then, DeLong comments on Tapper's cynicism and Jared Bernstein's attempts to defend the stimulus:
Jared fell into a trap. The right way to do the accounting is, for each extra year of employment we get:
Direct federal cost
Extra federal and state revenue
Extra costs of debt financing
Value of goods and services produced
Discount because we are buying different goods and services than we would ideally, or buying them at a different time
Value of having a job to the person who gets one — these aren't people who are indifferent between going to work and getting their head together, after all
Net Benefit to Economy
Let's go through each of DeLong's line items. I do believe that I have an objection to every single point in his analysis.
Direct Federal Cost of $92,000
As Tapper indicates at the end of the first block quotation, Jared Bernstein has the decency to admit that this figure of $92,000 does not come from an empirical analysis of the actual number of jobs "saved or created" compared to how much stimulus money has been expended. Rather, it refers to the administration's economic forecast through the end of 2010! In other words, DeLong is taking the administration's economic modeling as his starting point, and ends up concluding that the stimulus is apparently working. No kidding!
Remember that Obama's economic team predicted back in January that if the stimulus package were approved, then unemployment would never break 8 percent throughout the course of the recession. (See the graph at the bottom of page 4 in their original forecast.) That makes me skeptical of the team's assurances that the plan will create (or save) so many jobs that the $787 billion expenditure will translate into only $92,000 per job.
Of course, DeLong and other Keynesians could say that the January forecast was wrong, because economists didn't realize how bad the recession was at the time. Since it's obviously turned out worse than people thought at the time, it means more stimulus is needed to boost aggregate demand back up and close the "output gap." So what's the problem?
The problem is that many economists dispute the entire Keynesian model. By starting off his table of calculations with the theoretical $92,000 number — which itself comes from a Keynesian model touting the benefits of deficit spending — DeLong isn't even trying to grapple with critics who reject the Keynesian approach. He is simply assuming his conclusion without apparently realizing it.
Benefit of $27,000 in Extra Federal and State Revenue
The idea here is that the direct payouts from the federal government are partially offset, because people holding jobs pay taxes. I suppose that I don't have a problem with DeLong using the point for federal taxes, but I'm not sure I endorse his inclusion of state revenues.
To see the problem, consider this analogy. I live in Tennessee. Suppose the federal government wrote a check for $50 million to the Tennessee state government, say, to build a giant museum honoring Al Gore. (It's expensive because it's powered exclusively by hamster treadmills.) Using DeLong's logic, I think he would agree that every US taxpayer living in the 49 other states could be upset at their share of the $50 million expenditure. However, nobody in Tennessee would have a right to complain, because, after all, that money wasn't really a $50 million federal expenditure — it went right into the pockets of Tennessee politicians, and so showers benefits on all Tennessee residents.
Now, I suppose DeLong could answer by saying that extra revenue for state governments — due to employed workers who owe their jobs to the stimulus spending — obviates the need for other tax hikes. Yet, this assumes that state spending would be the same, regardless of stimulus awards. This clearly can't be right, because one major argument for the stimulus was that it would reduce the need for states to slash their spending.
To say that federal spending isn't "really" as expensive as it first appears, because some of the money flows into state treasuries, seems just as wrong to me as arguing that federal spending isn't "really" that expensive, because some of the money flows into the cash registers of gas stations. I see no reason to view state legislatures as somehow representing the economic interests of all Americans, in contrast to any other arbitrarily chosen group. If anything, I view money flowing into the hands of state politicians as an undesirable event by itself.
Zero Extra Costs of Debt Financing
Here, DeLong refers to the extremely low interest rates the government must pay to borrow money. Even putting aside the objection that technically the yields on short-term Treasuries are positive (i.e., not literally zero), DeLong is implicitly assuming that the entire $787 billion (and more, if we followed DeLong's advice to expand the stimulus) will be paid back before interest rates rise. I would be willing to bet $787 billion that this won't happen.
Benefit of $110,000 for the Value of Goods and Services Produced
I truly have no idea where DeLong got this number from, so it's hard for me to comment on it. (I emailed him but he didn't respond.) Just keep in mind that DeLong's final tally involves this particular number, which, I presume, he pulled out of the air. He certainly doesn't justify it in his article.
Cost of $18,000 Because Government Spending Is Wasteful
For this point, let me quote DeLong's fuller explanation:
[B]ecause we are pulling forward spending from the future into the present — spending the $92,000 now rather than in the future — we are buying stuff too soon, and because the government is all thumbs we are to some degree buying less valuable stuff than we would ordinarily be buying. Figure a 20% discount — that's an $18,000 cost.
I personally would use a higher discount than 20 percent when comparing the value I would get from spending my own money, versus the value I get from having politicians seize it from me and spend it "on my behalf." When I think about all the things the federal government spends money on — be it the war on drugs, the occupation of Iraq and Afghanistan, funding the IRS and SEC, etc. — I would actually rather my tax dollars be burned, than flow into the projects and thus boost the "services" I get.
Fair enough, DeLong has in mind roads, bridges, and so forth, and he is at least acknowledging that there is a discount because of government incompetence and the rushing of stimulus spending. Yet even so, why does DeLong apply his arbitrary 20 percent discount rate to the figure of $92,000? Shouldn't it be applied to the $110,000 figure for the value of goods and services produced? Unfortunately, I can't answer the question for sure, since we have no idea where that $110,000 figure came from. But it seems as if DeLong used $18,000 when he should have used $22,000 (i.e., 20 percent of $110,000).
Value of $50,000 for Having a Job
Again, DeLong apparently plucked this figure out of the clear blue sky. He could just as well have used $40,000 or $60,000, with the accompanying movements in the final tally.
Putting aside the arbitrariness of the number, is it even correct to consider this aspect in the first place? Let's quote DeLong on this point to make sure we understand his view:
Why shouldn't we be doing more deficit spending all the time? Usually because of [the last point]: when the economy is in its normal state, the marginal worker is somebody who doesn't value having a job all that much — the [last] number is usually on the order of $10,000 rather than $50,000, and so isn't worth the $18,000 cost of having the government actually do the buying. (emphasis added)
It's not perfectly clear what DeLong has in mind, but I believe he is underscoring the fact that having any particular job is not nearly as important during normal times as during a recession. In other words, DeLong isn't saying that an unemployed person during normal times has no burning desire to get a job; he's just arguing that no particular job opening is all that valuable.
In any event, DeLong's last sentence struck me as odd. For even if we substitute $10,000 rather than $50,000, his net-benefits figure still works out to $37,000 per job created by government deficit spending. DeLong also mentions that during normal economic times, interest rates are positive and thus government deficit spending is more costly than now.
But wait a second. Remember that DeLong is talking about the cost vs. benefit of the government providing an extra year of employment for one person. He is arguing that in normal times, the smaller "benefit of having a job" figure (down to $10,000 from its current value of $50,000) and higher costs on servicing the federal debt, make the total negative, rather than its current net value of $77,000.
For that to be true, it means that in normal times (i.e., not during recessions), the interest cost for the federal government to borrow an extra $92,000 works out to $37,000. As I mentioned above when discussing DeLong's claim that federal borrowing currently comes at no additional cost, it's not clear how the timing should be reckoned here, since the money will clearly not be paid back within a year.
Whatever the framework, I'll just point out that DeLong has painted himself into a corner. He tried to deal with the obvious retort — "If deficit spending is so great right now, why not always?" — but his own numbers can't force a negative outcome. In the text above, he compared $10,000 with −$18,000, but those aren't the relevant figures. The relevant figure is the bottom number of $77,000, and his two adjustments — where he bumps down the worker's valuation of having a job, and bumps up the interest cost on federal debt — don't whittle his $77,000 figure down to zero during normal times.
In short, despite his protests to the contrary, DeLong's analysis "proves" that federal deficit spending is always a good idea because it confers net benefits on society. Since that can't possibly be right, it's clear that DeLong's entire approach is flawed. We should take his $77,000 figure therefore with a pound of salt.
Two Competing Macroeconomic Visions
Let's pause just to remind ourselves of DeLong's apparent overall view. He thinks that in the depths of a recession such as the present one, the government can spend $92,000 to bring forth $110,000 worth of goods and services, from a worker who would have been willing to perform the job for a pay of only $42,000. Given these views, it's no wonder that DeLong thinks massive stimulus spending is a great idea.
Of course, a proponent of free markets would wonder how this situation could be possible. If there are unemployed workers willing to do a given task for $42,000, and consumers would value the resulting output at $110,000, then why in the world isn't there at least one entrepreneur filling the gap?
At this point, the two visions — the Keynesian and the Austrian — butt heads. The Keynesian can list all sorts of reasons that the market economy, left to its own devices, would not exploit the enormous profit opportunity suggested by the above numbers. For example, wages are "sticky." Even though plenty of unemployed workers would love to be hired at $42,000, for some reason, wages are stuck at a much higher level, causing institutional unemployment.
This article is already quite lengthy, so I won't delve into the various Keynesian arguments and the Austrian responses. I merely note that when it comes to redeploying resources after an economic disruption — so that workers can move to where they produce the most value for consumers — I trust the voluntary exchanges of the market, more than the coercive orders of politicians.
Brad DeLong runs a quick calculation and decides that each stimulus job actually showers $77,000 in net benefits on the economy. Yet, we have seen that literally every step in his reasoning is suspect. Rather than plunging the taxpayers even deeper into debt, the government would do well to cut its own spending and return resources to private hands. Only then can true economic recovery begin.
 Actually, I think what happened is that DeLong was right to cite "crowding out" but was sloppy in where he tried to plug that effect into his list of numbers. (If DeLong had been more specific about where the $92,000 figure came from, he may have seen the mistake.) Specifically, if the economy were at full employment, then a government expenditure of $787 billion would not have created any net jobs; people would have filled new stimulus jobs only by quitting their old jobs. Thus DeLong's $92,000 figure of the "direct federal cost" would have been much higher, since the given amount of government spending would be divided by a very small number of new net jobs. This large (and negative) figure would then be able to dwarf any of the other advantages, showing that in times of low unemployment, government deficits don't shower net benefits on society.