If a meteorologist was asked what the day's high temperature had been, would it be acceptable to simply repeat his/her earlier forecast? Of course not. The forecast was merely a prediction, which should now be replaced with what actually happened.

Yet that is the approach the Congressional Budget Office (CBO) used when declaring that the stimulus had saved 1.5 million jobs. Rather than actually examine the performance of the post-stimulus economy, it essentially re-released its old forecast that the stimulus would likely create jobs.

CBO Confirms Its Methodology

In a recent speech to the National Association of Business Economics, CBO Director Doug Elmendorf confirmed this by stating:

[W]e don't think one can learn much from watching the evolution of particular components of GDP [gross domestic product] over the last few quarters about the effects of the stimulus … so we fall back on repeating the sort of analysis we did before. And we tried to be very explicit about it that it is essentially repeating the same exercise we did rather than an independent check on it.[1]

When asked if this means that any actual underperformance of the stimulus would fail to show up in the CBO's stimulus jobs count, Elmendorf replied "That's right." This means the 1.5 million jobs saved estimate was pre-determined.

Of course, the stimulus was originally promised to create (not just save) more than 3 million jobs.[2] Instead, the economy has since lost more than 3 million additional net jobs. The abject failure of the stimulus policies recommended by Keynesian economic models should induce some fundamental re-analysis of these models' assumptions. Instead, the CBO is re-releasing the same jobs analysis--with the same economic assumptions--that they had used a year ago.

The "Begging the Question" Fallacy

The CBO's conclusion that the stimulus created jobs is based on an economic model that began with the premise that all stimulus bills create jobs. In other words, the conclusion is already assumed as a premise. Logicians call this the fallacy of begging the question. Mathematicians call it assuming what you are trying to prove.

More specifically, the CBO's model started by automatically assuming that government spending increases GDP by pre-set multipliers, such as:

• Every $1 of government spending that directly purchases goods and services ultimately raises the GDP by $1.75;
• Every $1 of government spending sent to state and local governments for infrastructure ultimately raises GDP by $1.75;
• Every $1 of government spending sent to state and local governments for non-infrastructure spending ultimately raises GDP by $1.25; and
• Every $1 of government spending sent to an individual as a transfer payment ultimately raises GDP by $1.45.[3]

(Note that all CBO figures in this paper represent the midpoint between their high and low estimates.)

Then the CBO plugged the stimulus provisions into the multipliers above, came up with a total increase in GDP of 2.6 percent, and then converted that additional GDP into 1.5 million jobs.

The problem here is obvious. Once the CBO decided to assume that every dollar of government spending increased GDP by the multipliers above, its conclusion that the stimulus saved jobs was pre-ordained. The economy could have lost 30 million jobs, and the model would have said that the economy would otherwise have lost 31.5 million jobs without the stimulus. An asteroid could have hit the United States, wiping out everyone outside of Washington, D.C., and (as long as Washington still spent the stimulus money) the CBO's economic model would have produced the same stimulus jobs data. There is no adjustment made to reflect what actually happened in the economy after the stimulus was enacted.

Computer Model Trumps Economic Reality

The inability to precisely count stimulus jobs is not the CBO's fault. It is extremely difficult to determine how many jobs were created or saved by the stimulus bill because there is no counterfactual showing how the economy would have performed without the stimulus. In addition, detailed economic data can take years to acquire. However, the proper response would be for the CBO to declare they lack the resources to measure the performance of the stimulus--or at the very least incorporate the economy's continued underperformance into their models--rather than try to present their earlier predictions as some sort of actual post-stimulus job count.

The CBO will likely release updated stimulus jobs estimates in the future. Why wait? If the CBO is merely repeating their prediction of the stimulus's effects--rather than actually examining the economy--they can release their end-of-2010 update now. It is all just a matter of multiplying scheduled stimulus outlays times the pre-set multiplier. There is no need to wait and see how the economy and stimulus performs.

Test the Multipliers

The debate over the efficacy of Keynesian stimulus is essentially a debate over the correct multipliers. Some believe the multipliers are high[4]; others believe they are as low as zero[5] (or even negative). Testing the stimulus requires testing the multipliers. Yet by simply assuming large multipliers, the CBO effectively pre-ordained its conclusion that the stimulus worked regardless of what actually happened in the economy.

Elmendorf has confirmed that the CBO's stimulus analysis consists of little more than re-releasing its pre-stimulus projections. Policymakers and analysts should not mistake this analysis for an actual examination of the stimulus's impact.

Brian Riedl is Grover M. Hermann Fellow in Federal Budgetary Affairs in the Thomas A. Roe Institute for Economic Policy Studies at The Heritage Foundation.