Defenders of the NRH [natural rate hypothesis] might choose to respond to my empirical findings by arguing that the natural rate of unemployment is time varying. But, I am unaware of any theory that provides us in advance [emphasis mine] with an explanation of how the natural rate of unemployment varies over time. In the absence of such a theory, the NRH has no predictive content. A theory like this -- which cannot be falsified by any set of observations -- is religion, not science.
I want to emphasise that this is not purely an issue for the natural rate of unemployment (u*) -- it can be applied to all of the other "star variables": the "natural rate of interest" (r*) and potential output (y*) . (Potential output is used to generate a GDP-based output gap, which is the usual way of framing the hidden variable.)
From a practical perspective, this question of predictive content shadows the mainstream methodology. My impression of recent research is that complexity is being added as a means of distracting away from this basic issue.
As I have written earlier, I will be discussing Roger Farmer's theory in a series of articles, which will likely end up as a chapter in the second volume of my book on recessions. (Link to volume one.) I want to focus on what his theory says, and not his criticisms of New Keynesian theories. As such, I pulled out this particular point as a separate article as it will be of interest to my heterodox-leaning readers.