Shortening the Sovereign Balance Sheet: Treasury Bill Stock Share, Primary-Auction Absorption, and Carrying-Cost Pass-Through Across Quantitative Tightening Regimes

Sovereign debt-management models posit an operational trade-off between shortening debt maturity to capture money-like convenience demand in primary auctions and exposing the fiscal balance sheet to faster interest-rate pass-through when policy rates rise. We evaluate both margins across Federal Reserve balance-sheet regimes from January 2015 through October 2026 by linking 4,136 primary-market auctions of U.S. Treasury bills and nominal coupon securities (1,774 unique securities across 142 year-months and 76 fine remaining-maturity tenors) with 142 monthly marketable debt and book interest-rate statements. Conditional on auction month and fine maturity tenor, competitive tenders expand less than proportionally with offering size: during the second quantitative tightening (June 2022–1 December 2025), the log competitive-tender elasticity for Treasury bills is 0.8792 (two-way clustered SE = 0.0314; wild cluster bootstrap p = 0.050 against unit proportionality) and declines to 0.7178 (SE = 0.0386) during post-runoff reserve management from December 2025 onward (regime difference = -0.1614, bootstrap p = 0.005), steepening the bill bid-to-cover slope from -0.4392 (bootstrap p = 0.041) to -0.9237 (bootstrap p = 0.002). Larger offerings during the second quantitative tightening coincide with higher indirect-bidder award shares (+7.0407 percentage points per log offering for bills, bootstrap p = 0.031; +7.5916 percentage points for coupons, bootstrap p = 0.017) and lower primary-dealer residual shares, although wild cluster bootstrap p-values soften above 0.05 under 15-benchmark-family clustering (0.090 and 0.094), without cash management bills (0.113), or outside extraordinary-measures periods (0.085 and 0.633). On the carrying-cost margin, published marketable average interest rates exclude inflation-protected and floating-rate principal and diverge from par-weighted fixed-rate reconstructions by 0.1936 basis points on average (0.8877 basis points maximum) due to net-carrying-value bill discounting. In dynamic monthly models (N = 138), an isolated 100-basis-point federal funds rate hike yields a direct four-lag coefficient sum of 15.10 basis points (HAC SE = 3.11) and, after autoregressive propagation (rho = 0.3507, p = 0.0047), a cumulative four-month impulse response of 22.84 basis points (delta-method HAC SE = 2.66, p < 0.0001; long-run multiplier 23.25 basis points), aligning closely with forward-controlled local projections (24.38 basis points at horizon h = 3). By contrast, the four-month differential for a 10-percentage-point higher bill stock share is statistically inconclusive (+6.72 basis points, HAC SE = 7.11, p = 0.3459; +1.21 basis points, p = 0.8967 under predetermined matched-lag shares), while symmetric shift-share decompositions confirm that 98.3% of direct four-lag policy pass-through operates through within-class rate resets.