Reader Steven Kopits writes “potential GDP model is also a binding constraint model”, so GDP “…is subject to some sort of natural speed limit which cannot be exceeded”. This assertion is so amazingly absolutist in nature, and represents such a misunderstanding of how macroeconomists typically think of potential, that I am moved to observe that if this were so, output would never exceed potential GDP in our frameworks. Now, let's consider the relevant depiction implied by the CBO estimates (using a production function approach [1]).

Figure 1: Log output gap (blue, left scale) and log GDP (red, right scale) and potential GDP (gray scale), all in bn. Ch.2009$, SAAR. NBER defined recession dates shaded gray. Source: BEA, 2014Q3 final release, CBO budget and Economic Outlook (February 2014), NBER, and author's calculations.

Note that output exceeds potential on numerous occasions according to CBO estimates, and hit 3.5% (log terms) as recently as in 2000. That's because in this framework (the neoclassical synthesis, as forwarded in for instance Samuelson's textbook, or e.g., here), factors of production can be utilized at greater than “normal” rates.

Now there are interpetations of potential GDP that fit Kopits' description; from this post:

In Summers and Delong (1988), the authors provide an interpretation of potential as a level of output that can't be exceeded (to me this is reminiscent of the Friedman “plucking model” (re-iterated in a 1993 publication, see also [1], [2])). Figure 5 depicts per capita potential under this interpretation.

Figure 5 from Summers and Delong (1988).

The formula used is given by recursive application of their equation 17:

Where y* is potential GDP, and k=3 to 5. It's of interest to consider what an updated version of the De Long-Summers procedure implies for the output gap. I apply the procedure to log GDP instead of per working age worker, and set k=3.