When do passive bond managers use sampling versus full replication?

Question

AcadiFi · Accepted Answer

Full replication (owning every bond in the index at index weights) is feasible for small, liquid equity indices like the S&P 500's 500 names. For bond indices with 12,000+ constituents (Bloomberg Global Aggregate), full replication is impractical: many bonds are illiquid, expensive to source, and round-lot constraints make small positions uneconomic.

Representative sampling selects a subset of bonds that together mirror the index's risk exposures: duration, convexity, key-rate durations along the curve, credit quality distribution, sector allocation (Treasury/agency/corporate/MBS), and currency exposure. Optimization minimizes tracking error subject to holding a limited number of bonds (often 200-500 for a broad aggregate) and respecting trade-size constraints.

Stratified sampling divides the index into cells (e.g., AA financials 5-10 year maturity, BBB industrials 10-30 year), then holds representative bonds in each cell at cell's index weight. This ensures sector and duration fidelity but may miss idiosyncratic spread moves.

For Lanesbury Aggregate fund managing $8 billion tracking Bloomberg US Aggregate: might hold 600 bonds representing 7% of the 13,000+ constituent index. Tracking error typically 10-25 bp annualized, driven by credit-specific spread moves and timing of reinvestment flows. The fund uses Treasury futures overlay to neutralize residual duration mismatch.

Full replication is viable for smaller indices: 1-3 year Treasury indices with ~50 bonds, ETF replication funds, or bespoke separately managed accounts. Growing use of TBA (to-be-announced) MBS and credit index futures reduces the friction further.

When do passive bond managers use sampling versus full replication?

Master Level III with our CFA Course

Related Questions

Practice Questions