I imagine this is how it worked for court alchemists. The CBO says that it has been ordered to give a prediction of the effect the ARRA and so it has.
The Economists suggests this reverse crystal ball gazing is valid because its model driven. Alchemy was model driven: if there is a Philosopher’s Stone, then you can turn copper to gold with it. Hey, alchemists models were even internally consistent; Philosopher’s Stone had the properties needed to make the transfiguration possible. The problem is they weren’t empirically relevant. Despite the evidence from the Harry Potter franchise, it turns out there is no Philosopher’s Stone.
Why is this alchemy? The CBO report talks about uncertainty of so-called “multipliers” as if it was trying put a confidence interval around a fixed, but unknown value. “The ranges between high and low multipliers are designed to encompass most economists’ views about the direct and indirect effects of different policies.” ((BTW, why is “most economists’ views” considered a valid way to estimate parameter uncertainty. Even assuming multipliers are structural and thus throwing 30 years of research in macroeconomics out the window, estimates of multipliers are uncertain because we don’t have sufficient data to estimate them NOT because 3/4 of economists get their estimates of multipliers from Mankiw’s textbook.)) But we know multipliers are not structural, that for different policy and economic environments the multipliers will be different. If we had Sumnerian monetary policy, fiscal multipliers would be zero. If we were in a liquidity trap or if the Fed gave up on monetary policy, multipliers would quite high. By assuming constant multipliers, the CBO is manufacturing the result.
How good are the CBO at forecasting (or hindcasting in this case)? Not good at all. Since 1976 their forecasts have been WORSE than random walk forecasts.
So the King orders copper to be turned to gold and the alchemists comply. The King’s happy. His subjects are happy. Their reality is unperturbed by actual reality.