601 INVESTMENT RISK REPORTING NEW REPORT: ACTIVE EQUITY RISK

Transcription

1 601 INVESTMENT RISK REPORTING NEW REPORT: ACTIVE EQUITY RISK Committee on Investments / Investment Advisory Committee August 17, 2004

2 RISK QUESTIONS What are the components of portfolio total risk and active risk? How do they compare to the policy benchmark? How much, at the margin, does each risk exposure contribute to portfolio risk? Is the risk contribution proportional to expectations for return? How does active performance over the last quarter relate to the risks taken? Over time, is there evidence of skill (positive risk adjusted performance)? Page 1

3 REPORTING RISK AND RETURN Goal of Treasurer s report to Committee What returns were achieved What risks were taken to earn the returns Are risks intentional and within appropriate bounds Risk includes Portfolio characteristics Performance attribution Sector / country / quality exposures Risk also includes Measures of volatility Contribution of sector / portfolio risk to total Measures of sensitivity (which factors have the most impact on total risk) Page 2

4 RISK MODELS Treasurer is now implementing holdingsbased factor models for forecasting equity and bond risk Same model as previously used for analyzing internal equity strategy Similar in concept to returns based models presented in Q report, however More granular forecasts to aid in decision making (manager allocation) More responsiveness to current market conditions Page 3

5 REPORT CONTENTS Portfolio characteristics Risk measures and risk decomposition Underlying risk factor exposures Industry factors Style factors Performance attribution based on risk factors Active US Equity includes the 25% of total US Equity which is or will be managed actively; as of , it includes a portfolio managed passively against the Russell 1000 tobacco free index and nine actively managed small capitalization US equity portfolios managed against various small cap style benchmarks Overall benchmark for Active US Equity is the Russell 3000 Tobacco Free Page 4

6 RISK MEASURES Measures of risk are estimates of volatility, and show the amount by which asset values could increase or decrease over a given time period Portfolio risk measures are based on the volatility of each security, the size of each position, and the degree to which security prices move together Units of risk are standard deviation of returns, annualized Range around expected return in which the actual one-year return should fall in roughly two of every three years Beta is a measure of the overall degree of market exposure A beta of 1.0 is neutral or very similar to the benchmark Both total and active risk may be decomposed into a common factor (systematic) portion and portion due to security selection. Page 5

7 PORTFOLIO CHARACTERISTICS PORTFOLIO CHARACTERISTICS VS BENCHMARK UCRP RUSSELL 3000 Dividend Yield Price - Earnings Ratio (historical) Price - Book Ratio Debt - Equity Ratio Fundamental Beta Market Capitalization (Cap Wtd) $68.45 bn $74.6bn Number of Holdings 1,972 3,000 The aggregate active portfolio is currently structured very similar to the benchmark Weighted security characteristics such as dividend yield, P/E, P/B have historically been successful in explaining crosssectional stock volatility, and are used in the development of factor based risk models Average market capitalization is slightly smaller than the benchmark, reflecting the active small cap strategies Page 6

8 RISK DECOMPOSITION RISK MEASURES / DECOMPOSITION (Qtr End) TOTAL RISK ACTIVE RISK Risk Contrib Risk Contrib %Std dev %Total %Std dev %Active Common Factor Style Factors Industries Factor Interaction N/A 2 N/A (2) Asset Selection Total Benchmark Beta 0.99 Shown above are measures of portfolio and benchmark risk, which are very similar The majority of total risk is attributed to industry exposures Only a small amount comes from style factors or asset selection Most of active risk though is attributed to style exposures different from the benchmark (see below) Industries and asset selection have a small contribution Active risk (tracking error) is very low (80 bp), due to the large component of passively managed funds Page 7

9 PERFORMANCE ATTRIBUTION PERFORMANCE ATTRIBUTION (one quarter) (3) Return Risk Total Portfolio Benchmark Active (0.25) 0.86 Common Factor 0.03 N/A Style Factors Industries Asset Selection (0.28) 0.23 Market Timing Trading Performance relative to a benchmark can be explained as follows: Risk exposures different from the benchmark Performance of those risk factors during the period Most of the underperformance was attributed to asset selection, not common factors Two thirds of asset selection return came from over-weighted assets, onethird from under-weighted assets Risk column in performance attribution refers to variability of each component of earnings over that quarter Page 8

10 STYLE EXPOSURES AND RISK Currency (0.000) Earnings Var. (0.000) Earnings Yield (-0.005) Growth (0.005) Leverage (0.000) Momentum (-0.001) Size(NEU) (-0.005) Size (0.004) Size (NL) (0.007) Trading (0.014) Value (0.000) Volatility (0.050) MCTR in parentheses Yield (-0.004) Portfolio Benchmark Active RISK MEASURES / DECOMPOSITION (Qtr End) (2) TOTAL RISK Risk Contrib Risk Contrib %Std dev %Total %Std dev %Active Common Factor Style Factors Industries Factor Interaction N/A 2 N/A (2) Asset Selection Total Benchmark Beta 0.99 ACTIVE RISK The graph shows portfolio and benchmark exposure to each style factor, measured in standard deviation Note on magnitude: as a rule of thumb, style exposures < 0.20 are not considered significant The active bars show the difference between portfolio and benchmark exposure The aggregate of all exposures, weighted by the volatility and covariances among the risk factors, yields the total volatility attributed to style Only the Size and Size NL (non linearity) factors are significant, and are indicative of the tilt toward smaller cap stocks together they account for a large portion of the active risk MCTR, or Marginal Contribution to Total Risk, is a risk sensitivity measure (see below for explanation) Page 9

11 SECTOR EXPOSURES AND RISK Basic Mat'ls (0.15) Cnsmr Cyc. (0.15) Cnsmr Non-cyc. (0.12) Cnsmr Serv. (0.14) Comm Serv. (0.14) Energy (0.12) Financial (0.15) Health Care (0.12) Industrials (0.15) Technology (0.18) Telecom (0.14) Transport (0.14) Utility (0.11) Cash (0.00) Portfolio Benchmark Active RISK MEASURES / DECOMPOSITION (Qtr End) (2) TOTAL RISK Risk Contrib Risk Contrib %Std dev %Total %Std dev %Active Common Factor Style Factors Industries Factor Interaction N/A 2 N/A (2) Asset Selection Total Benchmark Beta 0.99 ACTIVE RISK The graph shows portfolio and benchmark exposure to each industry group factor, measured in percent (adds to 100) The active bars show the difference between portfolio and benchmark exposure The aggregate of all exposures, weighted by the volatility and covariances among the risk factors, yields the total volatility attributed to industries There are no significant active industry exposures, indicating a well diversified portfolio; this explains the small contribution to active industry risk MCTR, or Marginal Contribution to Total Risk, is a risk sensitivity measure (see below for explanation) Page 10

12 Marginal Contribution to Total Risk (MCTR) This is the sensitivity to total portfolio risk of a small change in the factor exposure It is measured in basis points, or 1/100 of a percent of standard deviation E.g., a 1% increase* in exposure to the Financials (industry) risk factor will cause portfolio risk to increase by 0.148% or 15 bp E.g., a.01 standard deviation increase* in exposure to the Momentum (style) risk factor will cause portfolio risk to decrease by % or.12 bp *Calculation assumes that an increase (decrease) in asset weights resulting in increase (decrease) in factor exposure is offset by a decrease (increase) in cash of the same amount Also may be calculated for active risk to aid portfolio manager in determining impact of active (benchmark relative) decisions Page 11

13 CONCLUSIONS Risk exposures are similar to the benchmark, so active risk is low Most of the active portfolio is still passively managed, until the transition to active large cap managers later in the year The Risk Management group has established a quantitative review and oversight process of active managers to complement the qualitative oversight process of the Externally Managed Investments group A risk model is a key element in this quantitative analysis and control process Page 12

14 NOTES The benchmark for aggregate of all managers is Russell 3000 Tobacco Free Risk factor exposures and risk estimates are based on actual holdings as of quarter end using proprietary data and analytics from Barra, Inc. Performance covers the period to (single quarter, cumulative, not annualized) Performance attribution is based on month end holdings using a buy-hold methodology; thus calculated returns will differ slightly from official performance results Style factor exposures are expressed as % Standard Deviation (annualized) Industry exposures are expressed in terms of percent weights Page 13

15 APPENDIX: MODEL RISK FACTORS US equity risk factors measure systematic (common) risks shared by groups of securities; divided into two types Style factors 13 factors, measuring risks such as Large vs. small capitalization Value vs. growth styles Volatility, momentum, trading activity Leverage, exposure to foreign currency Industry factors 54 Industries, grouped into 13 sectors as shown on graph A security can have exposure to more than one industry Based on segment sales, assets, and earnings Page 14

16 APPENDIX: RISK MEASUREMENT First use of a risk model is to measure risk exposures and contributions Then compare those measures to: A benchmark, some pre-assigned limits, or qualitative understanding of the portfolio s strategy This is an important component of the total investment oversight process Includes both qualitative and quantitative analysis The value of a quantitative risk model: Contribution of a single position or sector to total portfolio risk depends on more than just its size Depends on the volatility of that position Depends on its covariance with the rest of the portfolio Page 15

17 APPENDIX: RISK BUDGETING Second use of a risk model is to aid in risk budgeting: Allocation of risk in proportion to expectations for return Manager will take positions different from consensus only if she has expectation to earn excess returns Size of the active position should depend on Expected out-performance Degree of confidence in one s beliefs Risk it contributes to the rest of the portfolio. Risk budgeting is the process of constructing a portfolio so that each active weight should be sized so that, at the margin, its contribution to expected return is proportional to its contribution to risk Application: allocation to managers within asset class Page 16