The maximum errors in Phillips match the blue bars

In Menzie's test of the Phillips.

Phillips does not work when we are rescaling.  The curve itself has a zero, a cycle, and it can be tuned well enough when velocity equation is accurate.  But that blue bar is not adiabatic, ergosis fails. So the estimator is momentarily like a tail whipping about the requant.

Our theory says the central banker wants a more liquid line of symmetry.  We want the longer sequence, we need to introduce the error term associated with the repeated blue bars, a longer sequence, a higher order queue,  And Menzie says something similar, and the curve sort of works on average.