Performance Eval & Attribution Flashcards
Simple Portfolio Return Decomp
P = M + S + A M = market return S = investment style (B-M, bench rel to mkt) A = active investment decisions
Seven Properties of a Valid Benchmark
- Specified in advance
- Appropriate
- Measurable
- Unambiguous
- Reflective of the manager’s current investment opinions
- Accountable (ownership by the manager)
- Investable
Seven Benchmark Types
“1. Absolute (return benchmark)
- Manager universes - usually median manager (only Measurable, BAD benchmark)
- Broad market indices
- Style Indices
- Factor-based model
- Returns-based
- Custom security-based (most appropriate)”
Six Tests of Benchmark Quality
“1. Minimal systematic bias
- Low tracking error
- Similar risk characteristics
- High coverage ratio
- Low benchmark turnover
- Positive active positions (enough liquidity in members to take active positions)”
Hedge Fund Benchmarks
”. Long-only benchmark is inapprpriate
. Measure sum of returns on individual positions (value-added return)
. Sharpe ratio (but stdev in denom not likely appropriate)
. Create separate long/short benchmarks”
Six Levels of Macro Attribution Analysis
“1. Net contributions
- Risk-free asset
- Asset Categories (pure index fund/index)
- Benchmarks (investment style - val/growth)
- Investment Managers (Active decision)
- Allocation Effects (““Plug value””)”
Micro Attribution Analysis
“R_v = (w_pj-w_bj)(R_bj - R_b) ““pure sector allocation”” +
w_bj(R_pj - R_bj) ““within sector selection”” +
(w_pj - w_bj)(R_pj - R_bj) ““alloc/selec interaction”””
Fixed-Income Performace Attribution
“1. External interest rate effects (passive)
a. expected
b. unexpected
2. Interest rate management effect (active)
3. Other management effects (active)
a. sector or quality
b. bond selection
c. transaction costs
4. Trading activity (active - plug value)”
Risk-Adjusted Return Measures
“1. Alpha = R_p - (R_f + Beta_p(R_mkt - R_f))
- Treynor Meas. = (R_p - R_f)/Beta_p (systematic risk)
- Sharpe Meas. = (R_p - R_f)/sigma_p (total risk)
- M^2 = R_f + ((R_p - R_f)/sigma_p)sigma_mkt
- IR = (R_p - R_b)/sigma_te = alpha/tracking error”
Quality Control Hypothesis Test
“H_0: Manager fails to add value
H_1: Manager adds value
Type 1 error: reject the null when it’s true
Type 2 error: fail to reject when it’s false
Type 1 (you fail to get rid of non-performing managers)
Type 2 (you get rid of outperforming managers)”
Global Portfolio Decomposition
“R_p = CG + CF + Curr
CG: capital gains in local curr.
CF: cash flow yield in local curr.
Curr: return from curr movements = e(1+CG+CF)
e= chg local ccy relative to domestic ccy “
Global Portfolio Attribution
“Expands upon decomposition when given benchmark
R_p = benchmark domestic return + market alloc + security selection + yield contrib + currency alloc =
sum( w_jb R_jbd ) + sum( (w_jp - w_jb) R_jbl ) + sum(w_jp(R_jpl - R_jbl)) + sum(w_jp I_jl) + sum(w_jp C_jp - w_jb C_jb)
C_jp = R_jpd - R_jpl
C_jb = R_jbd - R_jbl”
Multi-Period Active Return
"R_A2 = R_a1(1+R_b2) + R_a2(1+R_p1) R_A2: active return over two periods R_a1: act ret over period 1 R_b2: benchmark return over period 2 R_p1: portfolio return over period 1"
Order-Driven Submarkets
1) ECN - Electronic Crossing Network
2) Auction markets
3) Automated auction markets
4) Brokered market, or “upstairs market”
Dealer “adverse selection”
When a dealer has less information than a trader in the transaction of a security, and is thus at a disadvantage in the trade.
