Bullet versus barbell under each curve scenario
| Curve scenario | Bullet portfolio (concentrated mid-curve) | Barbell portfolio (split short/long) |
|---|---|---|
| Parallel shift up 50 bps | Loss ≈ 50 bps × effective duration | Same loss; effective duration is identical by construction |
| Steepening (2yr -25, 30yr +50) | Smaller offset; larger net loss | 2yr leg gains, 30yr leg loses; partial offset |
| Flattening (2yr +50, 30yr -25) | Smaller offset; larger net loss | 2yr leg loses, 30yr leg gains; partial offset |
| Butterfly (mid up, ends down) | Largest loss (concentrated at the worst point) | Both ends gain; mid is empty -- net gain |
| Inverse butterfly (mid down, ends up) | Largest gain (concentrated at the best point) | Both ends lose; mid is empty -- net loss |
How key-rate duration measures the bullet-barbell choice
Using key-rate buckets for P&L attribution
The most common application of key-rate duration is RISK ATTRIBUTION after a curve move. If 10-year rates moved 30 bps up over a quarter while 2-year rates were flat, a portfolio with high 10yr KRD will explain most of the loss; a portfolio with high 2yr KRD will be roughly flat. By decomposing P&L into key-rate buckets, a manager can verify whether the portfolio actually behaved as designed under the realized curve move -- and adjust position-sizing if a particular maturity bucket carries more risk than intended.
Compare short, mid, and long Treasury ETFs
Why correlated curve moves complicate the isolation
Key-rate duration assumes you can isolate movement at one maturity bucket while holding others constant -- but in practice, curve moves are correlated. A 'pure' 10-year rate move without any 2-year or 30-year movement is rare; most actual moves involve correlated changes across the curve. Quantitative bond managers use principal-component analysis (PCA) of historical rate moves to identify the dominant curve factors (level, slope, curvature -- the first three principal components explain 90%+ of curve variance) and align position-sizing to those factors rather than treating key-rate buckets as independent.
Key-rate duration decomposes risk by maturity bucket
Key-rate duration decomposes effective duration into maturity-bucket sensitivities (2yr, 5yr, 10yr, 30yr). Bullets concentrate exposure at a single point; barbells split between short and long. Bullets win under parallel shifts; barbells provide natural offsets under curve-shape changes. The metric is the foundation for risk attribution after non-parallel curve moves -- which is what real bond markets produce most of the time.
Sit with the ideas.
Two bond portfolios both have effective duration of 5.0 years. Portfolio A has key-rate durations of (2yr: 0.8, 5yr: 3.5, 10yr: 0.6, 30yr: 0.1). Portfolio B has key-rate durations of (2yr: 2.5, 5yr: 0.2, 10yr: 0.3, 30yr: 2.0). Which portfolio better protects against a curve STEEPENING (2yr down, 30yr up)?