Output gap filters disagree because they measure different gaps
Different output gap filters can measure distinct economic concepts rather than imperfectly estimate the same quantity. Danmarks Nationalbank Working Paper No. 225, published October 22, 2026, proposes a filter that separates two theory-based gaps.
Two gaps behind one estimate
Jeppe Druedahl, Martin Andreas Kildemark and Christoffer Jessen Weissert distinguish efficient output, absent nominal and real frictions, from flexible-price output, absent nominal frictions only.
Productivity and labor-supply shocks contaminate output and employment signals; cost-push shocks separate the two gaps.
Inflation net of expectations helps identify slack but tracks the flexible-price gap when cost-push shocks occur.
Their two-gap filter adds a cost-push state to the Okun–Kuttner specification, retaining output, labor, inflation and expectations.
Tests use 100 simulations of 250 quarters across seven benchmark environments.
The filter improves gap recovery where conventional filters fail, with gains also found in richer models.
US slack tells two stories
US estimates for Q1 1984–Q2 2025 put the efficient gap near zero after the pandemic but the flexible-price gap well above zero.
After the financial crisis, efficient slack persists until around 2015.
Against the two-gap estimates, Okun–Kuttner's efficient-gap concordance is 1.000; Hodrick–Prescott's is 0.664. Kuttner correlates 0.971 with the flexible-price gap, but scale differences reduce concordance to 0.427. Congressional Budget Office concordance is 0.736 with the efficient gap and 0.392 with the flexible-price gap.
The cost-push state tracks an external energy-price cycle.
One number hides the trade-off
The authors warn that averaging one-gap estimates can create an uninterpretable measure rather than resolve disagreement.
Same-sign gaps permit a common stabilizing response; opposite signs expose a policy trade-off.
Precision remains constrained by labor-supply shocks and uncertainty over the Phillips-curve slope.