What is empirical duration and how does it differ from analytical duration?

Question

AcadiFi · Accepted Answer

Empirical duration is estimated by regressing observed bond price changes on observed yield changes over a historical window. It captures real-world price-yield dynamics including liquidity, credit, and flow effects, not just the analytical cash-flow response.

Formula: Empirical duration = -β in the regression ΔP/P = α + β × Δy + ε

Procedure:

1. Collect daily (or weekly) bond total return and Treasury yield changes over a 1-3 year window
2. Run OLS of return on yield change
3. The slope β, negated, is the empirical duration

Example — Alderton Credit Corp 7-year bond analyzed over 504 trading days:
- Analytical modified duration (from cash flow formula): 5.8
- Empirical duration (from regression of returns on 7y Treasury yield changes): 4.2

Why the gap?
- The bond is a BBB credit; in risk-off episodes credit spreads tighten as Treasury yields fall (correlated), offsetting duration risk
- Bond liquidity is thin; stale pricing dampens observed sensitivity
- Investor flows sometimes move independently of rates

For high-yield bonds the gap is even wider. Empirical duration for a B-rated bond might be 2.1 against an analytical of 4.9, because credit spread widening in sell-offs offsets much of the rate move (spreads are negatively correlated with Treasury yields in stress).

For AAA-rated sovereigns, the two measures converge. Analytical and empirical duration for the on-the-run 10-year Treasury should be nearly identical because there's no credit overlay to disturb the relationship.

When to use empirical duration:
- Credit-sensitive instruments (corporates, EM, HY)
- Structured products with non-standard cash flows
- Portfolio-level risk measurement where liquidity matters
- Verification of model-based risk numbers

When NOT to use:
- Fresh issues with limited price history
- Regimes where relationships are shifting (regression estimates backward-looking)
- When analytical duration is required by regulation (many capital models)

Best practice: report both, explain discrepancies in commentary, and use empirical for actual hedging P&L attribution.

What is empirical duration and how does it differ from analytical duration?

Master Level II with our CFA Course

Related Questions

Practice Questions