Reading 35: Portfolio Performance Evaluation Flashcards
Analysing Portfolio Returns
Regression analysis (risk factors like systematic risk, size, value) is used to determine how portfolio returns compare with index returns in returns-based attribution. Holdings-based attribution considers the holdings over time to evaluate the decisions that contributed to returns, while transactions-based attribution uses both holdings and transactions to explain performance. Returns based is least accurate as it isn’t based on underlying securities, transaction based is most accurate but also very complex.
Capture Ratio
A capture ratio greater than 1 indicates positive asymmetry of returns, or a convex return profile. Capturing more of the upside and less of the downside.
Portfolio Evaluation
Performance measurement: overall return and risk evaluation, to a benchmark (relative) or to a target (absolute).
Performance attribution: key drives that generate performance, used as quality control. Must account for 100% of portfolio’s risk and return.
Performance appraisal: were returns driven by luck in overall market or investment decisions.
Micro Vs Macro Attribution
Micro attribution analyses to portfolio at the portfolio managers level, understanding the drivers of the portfolio’s return. Macro attribution analyses investments at the fund sponsor’s level, including decisions to deviate from strategic asset allocation.
Arithmetic Vs Geometric Attribution
Arithmetic subtracts benchmark return from portfolio return for single period, they do not compound to equal active returns over multiple periods without a smoothing adjustment. Therefore geometric attribution is used which defined return as a ratio (most common for investment professionals).
The Branson-Hood Beebower Method (BHB)
Allocation effect, measures value added from overweighting or underweighting the sector relative to sector benchmark performance (BH method uses relative to sector benchmark excess return over the overall benchmark). Selection effect measure security selection value add through return on portfolio vs return on benchmark. Interaction effect measures allocation and selection effects working together. Note that the Brinson Fachler (BH) method is exactly the same except for allocation effect mentioned above (more common to use BH).
Fixed Income Attribution Analyses
The three methods include exposure decomposition - duration based; yield curve decomposition - duration based; yield curve decomposition - full repricing based.
Yield Curve Decomposition - Duration Based
Major active bets a fixed income manager could take to generate active return include: duration, curve shape, sector selection, bond selection. Exposure decomposition portions both portfolio and the benchmark into duration buckets, then into sectors. This method allows simple presentation of the output of attribution with low data requirements. Used for marketing and client reporting as it is clear and understandable.
Yield Curve Decomposition - Duration Based
Decomposes the active return into the following sources of yield to maturity: yield or income, roll, parallel shift in curve (duration and convexity), shape of curve, spread, residual. This method requires more data and is used by analysts and portfolio managers rather than for marketing.
Yield Curve Decomposition - Full Repricing
Most comprehensive form of fixed income attribution. Breaks down active return of the manager using individual spot rates for cash flows occurring at different maturities. Can accomodate the broadest range of instruments and yield curve changes. It is the most complex and less likely to be easily understood.
Risk Attribution
For bottom up relative attribution analyses use securities marginal contribution to tracking risk (tracking error), for absolute bottom up use securities marginal contribution to total risk. For both top down and factor based absolute, use factors marginal contribution to total risk and specific risk. For relative top down attribute tracking error to relative allocation and selection. For relative factor based analyse factors marginal contribution to tracking error and active specific risk.
Liability Based Benchmarks
Likely used by fund sponsors and portfolio managers when a firm has specific liabilities to pay in the future (defined benefit plan). Nominal bonds, inflation adjusted bonds and high quality stocks are frequently used assets.
Asset Based Benchmarks
There are seven primary types: Absolute aims to exceed a minimum return benchmark (not investable); Broad Market Indexes, are investable, measurable, specified in advance but style may deviate; Style Indexes are investable and widely understood but investment style definitions can vary and some also may have large weightings in certain securities and sectors; Factor Model Based may not have available data, may be ambiguous, and factor exposure isn’t always relevant to managers; Returns Based use portfolio returns over time and corresponding returns on style indexes for same period, doesn’t reflect what managers owns, sufficient historic returns are needed, won’t work if style changes.
Asset Based Benchmarks 2
The final two are: Manager Universes that use broad universe of managers or funds and the median manger falling in the middle when ranked from highest to lowest, it is measurable but prone to survivorship bias, reliance on representation of universe, cannot be identified or specified in advance (not investable so not an appropriate benchmark). Custom Security Based, designed to reflect managers security allocations and investment process, meets all validity criteria, allows continual monitoring, can be expensive, lack of transparency is possible.
Valid Benchmark
A valid benchmark includes the following: specified in advance, appropriate, measurable, unambiguous, reflective of managers current investment opinions, accountable (accepting key differences between portfolio and benchmark), investable (can replicate the benchmark and forge active management).
