Portfolio Construction
What you must know
Benchmark selection criteria (specified in advance, appropriate to the mandate, measurable, unambiguous, reflective of current investment opinions, accountable), passive replication methods (full replication, stratified sampling, optimisation), active approaches (fundamental, quantitative, factor-based, core-satellite), multifactor models and the distinction between factor beta and true alpha, tracking error and the information ratio, constraints and position sizing, risk budgeting and risk parity, and fixed-income construction (duration, curve, spread, liquidity).
Traps
- Tracking error measures relative (active) risk. A portfolio can have negligible tracking error and enormous absolute risk, or the reverse.
- The information ratio is active return divided by active risk; comparing IRs across managers using different benchmarks is meaningless.
- Regression alpha may simply be compensation for unintended factor exposures. Unsystematic risk is diversifiable and is not rewarded.
- The benchmark must be investable and chosen before the fact; selecting one after the result violates the specified-in-advance criterion.
- Full replication minimises tracking error but is costly in a broad index; sampling deliberately accepts tracking error to cut cost.
- Position sizing limits and liquidity constraints belong in the optimisation, not bolted on afterwards.
Worked example: information ratio screen
Two managers run against the same benchmark. Manager A: active return 3.0%, tracking error 8.0%, so IR = 0.375. Manager B: active return 1.5%, tracking error 3.0%, so IR = 0.500. B is the better risk-adjusted active bet; A's larger alpha is purchased with more than proportionate active risk. Rule of thumb: compare IRs only over identical periods against identical benchmarks.