“It’s an important question. The administration’s forecast is within the range of CBO’s post-stimulus forecast, it’s within the range of the full range of private forecasts that are out there.”
Here's the image of what he means:
Source: 2010 FY Federal Budget, CBO Macro Effects of ARRA (March 2, 20009)
The Low/High CBO lines are estimates produced by the Congressional Budget Office that include the impact of the stimulus. The budget line is the forecast utilized by the administration in it's budget projections, which also includes estimates of the stimulus' effect, and the Blue Chip is a trend reported in the budget taken from information by traders.
I'm not going to respond directly to the private forecasts although the blue chip consensus provided in the budget itself is well below the Budget projections. Yet the CBO projections do little better. To establish the CBO low and high lines above, the CBO estimated low and high estimates of how much growth the stimulus bill (ARRA) will add to growth on top of its 2009 baseline (reported in the budget).
As we can see the CBO estimates are, well, high. The high estimate forecasts growth of 5.4 percent in 2011. According to data from the Bureau of Economic Analysis the last time that real GDP grew that much was 1984.
Even the budget baseline and the CBO low forecasts three years of growth greater than 4 percent from 2011 to 2013. The US had a streak like that recently, from 1997 to 1999. Of course that was following years of sustained growth in the early 1990's and during the dot com bubble not in a recovery.
Even beyond the large numbers, Geithner's comments ring a bit hollow. Since CBO provides no information on how likely it's high or low estimates are evaluating them is more difficult. If, as is likely the CBO high represents a possible scenario well above the median outcome, say the 75the percentile , then calling on this hopeful range as good proxy for likely outcomes may be a poor choice.
CBO could have used stochastic simulation to solve this problem. Stochatic simulations, run multiple iterations of large models to produce an array of results and then attempting to assign probabilities to each outcome, or more precisely to assign outcomes to a probability distribution function. The Social Security Administering (SSA) has included such models in its Trustees Report for some time (see Figure II.D7 in the link). Although SSA still uses low/high estimates as well. The benefit of low/high, and the reason that CBO likely generated them, is that they are simple to read and much faster to compute.