The 2009 stimulus promised unemployment under 8 percent, then came 10

In January 2009, a forecast went out with a chart attached. It said that if Congress passed a large spending and tax package, unemployment would stay at or below roughly 8 percent. If Congress did nothing, it would go higher. It was a clear, falsifiable prediction, made by serious economists, and that is exactly why it deserves to be tested. A policy that makes a prediction invites a grade.

The grade came in October 2009, when the unemployment rate reached 10 percent. This post is a rebuttal, and it follows the rules of the form. I will state the opposing argument at its strongest, in words its defenders would accept. Then I will answer it in numbered points, using the record.

What was passed

The American Recovery and Reinvestment Act became law on February 17, 2009. It was a combination of tax cuts, aid to state governments, spending on infrastructure and energy, expanded unemployment and nutrition benefits and other programs. The headline figure at signing was about 787 billion dollars, and the Congressional Budget Office later estimated the total cost at roughly 830 billion dollars as the details were updated. (Check the CBO's own reports for the current accounting; the figure changes with revisions.)

It passed in a hurry, in the middle of the worst financial crisis in decades, with no Republican votes in the House and only three in the Senate. The argument for speed was that the economy was falling and delay would deepen the fall.

The multiplier argument at its strongest

Here is the case for the stimulus, as its best defenders would make it.

One. An economy can fall into a slump in which total spending is too low to employ everyone. In 2008 and 2009, households were losing wealth as house prices and stocks fell, banks were cutting credit and businesses were cancelling orders. Every one of those decisions was rational for the individual, and together they made the downturn worse. This is the paradox that John Maynard Keynes described in the 1930s.

Two. Interest rates were already close to zero. The Federal Reserve had cut the federal funds rate to nearly nothing by the end of 2008, which meant that the usual tool of lowering rates had run out of room. When monetary policy is at its limit, fiscal policy has to carry more of the load.

Three. In such a slump, a dollar of government spending can produce more than a dollar of output. The government hires a contractor, the contractor pays workers, the workers buy groceries and the grocer hires help. This chain of respending is the multiplier. When there are idle workers and idle factories, the extra demand can be met without raising prices, so the multiplier can be larger than one.

Four. The government can borrow cheaply, and unemployed resources cost something anyway. A worker who is not working produces nothing, pays no taxes and may draw benefits. Putting him to work on a useful project is close to free in terms of what the economy gives up.

Five. The counterfactual cannot be seen, and critics are unfair to demand that it be. The unemployment rate would have gone higher without the stimulus, and the CBO and other analysts estimated that the Act raised output and employment compared with what would have happened without it. The forecast was wrong not because the policy failed but because the economy turned out to be in worse shape than the numbers available in January suggested. Data revisions later showed that the fall in output in the fourth quarter of 2008 was far deeper than first reported.

That is a serious argument, and the last point deserves respect. The people who made the forecast were working with data that were later revised for the worse. I accept that. Now let me answer.

The forecast and the record

Start with the forecast itself. In January 2009, Christina Romer and Jared Bernstein, economists advising the incoming administration, published a report on the job impact of the Recovery and Reinvestment plan in early January 2009. It projected that with a stimulus package of roughly the size that was later enacted, the unemployment rate would stay at or below roughly 8 percent, compared with a rise above 9 percent without it. The report was open about the uncertainty in its estimates, and it said so. But the headline was plain: no higher than about 8 percent with the plan.

The record, from the Bureau of Labor Statistics, is that the unemployment rate rose to 10.0 percent in October 2009, eight months after the law was signed. It remained above 8 percent for most of the following three years, and did not fall below that mark until well into 2012. The economy's recovery was among the slowest after a deep recession in the modern record.

Now, the honest defender says that the baseline was worse than anyone knew. Fine. But a policy is not graded only against the numbers in a January forecast. It is graded against whether it delivered the recovery that was promised and paid for. And the promise was specific.

The answer, in numbered points

1. The "worse baseline" excuse cuts both ways. If the economy was weaker than anyone knew, then the stimulus was a bet made in ignorance, and the people who wrote the forecast should not have presented it with such precision. A forecast that says "8 percent with the plan, 9 percent without" implies a knowledge of the economy that nobody had. That is the Hayekian point about the limits of expert knowledge. The officials who designed the package did not know how deep the hole was, and they did not know how much the spending would help. They knew a theory. (See Solyndra to Amtrak, why government picks losers when it picks winners for the broader problem with officials claiming to know what only markets reveal.)

1. The multiplier is not a constant of nature. The argument for the stimulus rests on a particular number, and the number is contested. The Congressional Budget Office, in its analyses of the Act, did not give a single figure. It gave a range, with the low estimates for the spending multipliers well below one and the high estimates well above one. Read the CBO's reports for the exact table. What matters is that the profession was divided, and that the lower end of the range means that a dollar of spending produces less than a dollar of output. When the range is that wide, a confident forecast of unemployment is not science. It is a choice of which estimate to believe.

1. Even a multiplier above one has to be weighed against what the money costs. The funds were borrowed. Someone must repay, and the debt has an interest cost. A conservative asks not only "did it create jobs?" but "at what price per job?" Even if the CBO's estimates of jobs supported are taken at face value, the cost per job was very high. Compare that with what the same resources might have done in the private economy, and you see the opportunity cost. For the arithmetic of interest on the debt, see do the arithmetic on $40 trillion and the interest bill it carries.

1. Much of the money went to the wrong place at the wrong time. The plan was sold as a quick injection of spending on projects that were "shovel-ready," that is, ready to start immediately and put people to work. In practice, large infrastructure projects need planning, environmental review, bids, contracts and permits. Critics noted afterward that shovel-ready had turned out to be less ready than hoped. A bridge cannot be built next month, and a recession is over, or not, by then. The share of the money that went to aid for state governments and to benefit extensions did flow quickly, but those programs mostly propped up existing spending rather than creating new projects.

1. Temporary spending does not change what people expect. Milton Friedman's permanent income hypothesis, published in 1957, held that households base spending on their long-run expected income, not on windfalls. The 2009 package included a payroll tax credit of up to 400 dollars per worker, spread through paychecks, which many households did not notice and some saved. A subsidy that is small, temporary and hidden in withholding tables has little power to change behavior. (For the Chicago school's skepticism about stimulus, see Austrian, Chicago and Keynesian economics compared on one table.)

1. The slow recovery points to structural problems that spending could not fix. The housing wreckage, the damaged balance sheets of households and banks, the uncertainty over taxes and regulation and the mismatch between workers' skills and open jobs are not demand shortfalls that a road project solves. A government that treats every downturn as a demand problem keeps prescribing the same medicine for different illnesses.

What a fair reading concludes

Let me be careful about what I am claiming and what I am not. I am not claiming that the Act did nothing. Almost any large injection of money does something. Some people kept their jobs because their state government avoided cutbacks. Some people were helped by extended benefits. The CBO's analysis concluded that the Act raised employment relative to what it would otherwise have been, and I do not ask you to ignore that.

My claim is narrower and, I think, stronger. The package was sold with a prediction, and the prediction failed. It was sold on the speed and the readiness of its projects, and the projects were slower than advertised. It was sold on a multiplier whose range was wide, and the policy was justified as though the high end were certain. And its cost, in debt, was large. A policy that costs this much and delivers this little certainty should not be the default response to the next downturn. The recovery from 2009 was slow, and the burden of proof should be on the people who say that more of the same would have been faster.

What the Founders would ask

Let me put a constitutional question to it as well. The Constitution grants Congress the power to tax and spend for "the common Defence and general Welfare of the United States" (Article I, Section 8, Clause 1). It does not say "for the management of the business cycle." The founding generation did expect the government to borrow in emergencies, Hamilton's program of 1790 made that plain, but they thought of debt as a tool for extraordinary needs, to be paid down in better times, not a standing instrument of economic management. See Hamilton's 1790 Report on Public Credit and what honest borrowing requires for how Hamilton thought about it.

The modern view, associated with Keynes, treats borrowing as a countercyclical tool: borrow in bad times and repay in good. In practice, Washington has done the first half with enthusiasm and the second half almost never. A tool that is used in only one direction is not countercyclical. It is simply spending. If a Keynesian wants to defend a stimulus, the honest test is whether he will also support the surplus that his own theory requires in a boom. Few ever have. (For the history of rules meant to enforce that discipline, see From Gramm-Rudman to PAYGO, a timeline of fiscal rules that bent.)

What to do instead

If not a stimulus, then what? Conservatives have a list, and it is worth stating, because "do nothing" is not a serious answer to millions of people out of work.

• Fix the monetary side. If the money supply is shrinking, the central bank should expand it, and it has the tools. This is the Friedman and Schwartz lesson.
• Cut barriers to hiring and investment. Lower the cost of employing someone, simplify the tax code and stop adding new uncertainty about rules, so businesses that were waiting can proceed.
• Let the housing and banking correction happen cleanly. Resolve failed institutions in an orderly way, as I describe in Bailouts and moral hazard, from Chrysler in 1979 to Silicon Valley Bank, and don't drag out the repair.
• Make unemployment insurance a bridge, not a trap. Support workers during a transition without creating cliffs that discourage work.
• Keep the debt on a sustainable path, so that the country has room to act in a real emergency.

These policies do not promise a number on a chart. They don't need to. They are the policies of a government that is honest about what it can know.

The case in brief

1. The Recovery Act of February 17, 2009, was justified by the multiplier argument and by a January 2009 forecast that unemployment would stay at or below roughly 8 percent with the plan.
2. Unemployment reached 10 percent in October 2009 and stayed above 8 percent for most of the next three years, so the forecast failed.
3. The defense, that the baseline was worse than known, concedes that the forecast claimed a certainty nobody had, and the CBO's multiplier range shows the profession was divided.
4. Shovel-ready projects were slow, temporary spending changed little behavior, and the cost was borrowed money with interest to pay.
5. A better response would fix the money supply, remove barriers to hiring, resolve failed firms cleanly and keep the debt sustainable.

What you can do: read the Romer and Bernstein report and the Bureau of Labor Statistics unemployment series side by side, then ask your member of Congress what result would convince them that a stimulus had failed.