Fixed Income Attribution

Brinson attribution splits a stock portfolio's return into sector allocation and stock selection. Apply the same idea to a bond portfolio and it mostly falls apart, because bond returns aren't primarily driven by which "sector" you're in, they're driven by what happens to interest rates and credit spreads, and a bond manager's real decisions are about duration positioning and curve shape, not about which industry to overweight. Fixed income attribution replaces Brinson's two-way split with a decomposition built around the things that actually move bond prices.

What moves a bond, and who gets credit for it

A bond's price return over a period comes from three broad sources: the passage of time (carry, or rolling down the curve), the level and shape of the risk-free yield curve moving, and the bond's credit spread moving. Fixed income attribution isolates the manager's contribution to each:

$$R_{portfolio} - R_{benchmark} \approx \Delta_{\text{curve positioning}} + \Delta_{\text{spread positioning}} + \Delta_{\text{selection}}$$

In words: the total outperformance splits into how well the manager's duration and curve exposure was matched to actual rate moves, how well the manager's credit spread exposure was matched to actual spread moves, and how much extra return came from picking better bonds within each rate/spread bucket rather than from the bucket itself. The curve effect typically uses key-rate durations rather than a single duration number, because a manager might be correctly positioned for a parallel shift but wrong-footed by a steepening or flattening the single-duration number can't see.

Worked example

A portfolio is overweight duration by 1 year relative to its benchmark going into a quarter where the 10-year yield falls by 50 basis points (bond prices rise as yields fall). The portfolio also holds slightly lower-quality credit than the benchmark, and spreads widen by 20 basis points that quarter (widening spreads hurt lower-quality bonds more).

1. Curve effect. Extra duration of 1 year times a 50 basis point rate decline: approximately $1 \times 0.50\% = 0.50\%$ of extra return from being long duration into a rally.
2. Spread effect. If the portfolio's spread duration exposure is 0.3 years more than the benchmark's, and spreads widen 20 basis points (a loss for spread-long positions): approximately $-0.3 \times 0.20\% = -0.06\%$, a drag from the extra credit risk.
3. Net positioning effect. $0.50\% - 0.06\% = 0.44\%$ — the manager's curve call more than paid for the credit-risk drag this particular quarter, even though the credit call itself lost money.

Selection effect is computed separately, comparing the portfolio's actual bond-level returns within each bucket against the benchmark's bonds in that same bucket, isolating pure credit-picking skill from the rate and spread bets already accounted for.

What this means in practice

Fixed income attribution is standard at any bond shop reporting to institutional clients, because a client wants to know whether their manager's outperformance came from a repeatable skill, curve and spread positioning, security selection, versus a one-off correct macro call that could reverse next quarter. The framework also flags risk: a manager who has been "outperforming" purely through a large, static duration overweight isn't demonstrating selection skill, they're running a leveraged rates bet that will lose exactly as much when rates move the other way.