Reason magazine’s Peter Suderman has a lengthy but eye-opening examination of President Obama’s 2009 American Recovery and Reinvestment Act — aka the stimulus bill — and why even after spending $833 million through it the economy continues to suck.

The article is a top-to-bottom dissection that exposes the many layers of folly involved. Several passages stand out but this one in particular is worth noting because it points out the central flaw in reports that argue the stimulus was a success: There is literally no way to measure those claims.

“According to the non-partisan Congressional Budget Office,” says, the Obama administration’s stimulus website, “the Recovery Act supported as many as 3.5 million jobs across the country.” As the stimulus ran its course over roughly three years, the capital’s top newspapers kept printing similar, supportive-sounding figures from the budget office. “CBO Says Stimulus May Have Added 3.3 Million Jobs,” a Washington Post headline trumpeted in 2010. “CBO: Stimulus Added Up to 3.3 million Jobs,” declared a Politico headline in 2011. Senate Democrats touted the estimates as proof of ARRA’s success. So did the vice president.

When the stimulus passed, the White House promised more than just results; it promised accountability. “If the verdict on this effort is that we’ve wasted the money, we built things that were unnecessary, or we’ve done things that are legal but make no sense, then, folks, don’t look for any help from the federal government for a long while,” Biden told local government officials at a conference on stimulus spending in March 2009. The administration set up to help collect data, track stimulus spending, and see how many jobs it created.

ARRA also tasked the CBO with issuing quarterly reports estimating both the size of the boost the stimulus had given to the economy and the number of jobs it had created or saved. When the CBO put together its estimates, it ran into the same trouble that every economist attempting to measure the impact of government spending eventually faces: the lack of a counterfactual. There was no way around it. In a November 2009 report, the agency flatly declared that “it is impossible to determine how many of the reported jobs would have existed in the absence of the stimulus package.”

Impossible, yes, but the CBO was still required by statute to produce a progress report every quarter. So rather than attempt to measure the law’s output — the actual number of real-world jobs created or saved — the CBO measured inputs: how much money was spent and on what. Once the CBO knew how much money had been spent, it combined that number with an estimate of the multiplier. Because it relied on a range on multipliers that went relatively high — up to 2.5 for federal purchases — the results were quarterly reiterations of estimates the agency made before the law was even passed. (The White House Council of Economic Advisers also made estimates using a similar technique.)

The CBO estimated that the stimulus created or saved up to 3.6 million jobs. But CBO Director Douglas Elmendorf has also noted that if the real-world results were different — if the law created 5 million jobs, or if it created none at all — the agency wouldn’t know. At a March 2010 presentation, Elmendorf characterized the CBO’s follow-up reports as “repeating the same exercises we did rather than an independent check.” At the same event, Elmendorf was asked, “If the stimulus bill did not do what it was originally forecast to do, then that would not have been detected by the subsequent analysis?” His response: “That’s right. That’s right.” (Emphasis added.)