Phillips Curve Flattening Mechanically Weakens NAIRU Identification

This paper shows that, in the standard random-walk state-space model, the likelihood information about an interior NAIRU state relative to the smoothness precision supplied by the state equation is proportional to the squared Phillips curve slope. Canadian quarterly data from 1977 to 2026 show the mechanism is quantitatively severe: across the February 1991 inflation-targeting break this ratio falls by a factor of twenty-three, driven almost entirely by the slope. The credible interval for the unemployment gap then contains zero in 97 per cent of quarters, against 73 per cent before, and without a sign restriction the posterior probability of a negative slope falls from 0.99 to 0.78.