The Universality Premium in Treasury Yield Curve Risk

Abstract

This study develops a scope–efficiency–reliability framework for Treasury yield curve risk and quantifies the universality premium associated with protecting arbitrary linear key-rate exposures. Four principal prediction-set designs are compared over 6,502 walk-forward forecast dates from 2000 through 2025 using six US Treasury maturities. At the 99% nominal level, the Gaussian full-curve benchmark misses on 2.169% of dates; rolling and adaptive conformal full-curve sets reduce this to 1.000%. Broader coverage has a measurable set-size cost: the adaptive full-curve design requires a 57.1% larger mean standardised radius than calibration to a fixed family of five portfolios. The paired bootstrap 95% interval for the radius difference is [1.784, 1.944] standardised units. Full-curve conformal misses nevertheless remain temporally clustered, whereas the adaptive fixed-family design records 0.984% simultaneous misses and is not rejected by the full-sample conditional-coverage test. Adaptive updating increases radius relative to rolling calibration without resolving full-curve clustering in the base specification. The contribution is an explicit, empirically measured distinction between coverage scope, prediction-set size, and temporal reliability, giving practitioners a concrete basis for selecting and validating portfolio-aware risk bounds.