Chapter 8. Record Portfolio Theory and Capital Market Yiyang Yang Department of Applied Mathematics and Statistics State University of New York at Stony Brook March 2012 Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 1 / 35 Outline 1 Introduction 2 Measuring Performance 3 Mutual-Fund Performance 4 The Shapes of Distributions 5 Using the Past to Predict the Future Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 2 / 35 Introduction Past vs. Future portfolio theory and capital market theory: prediction empirical work: past record Motivation: portfolio theory+predictions from past record measuring tools: probabilities; relative frequencies expected return; average return variability; relative frequencies of various deviations Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 3 / 35 Introduction The Market Portfolio Measure of market portfolio: Indexes: Dow-Jones’ Index, S&P Composite Index average values for individual securites: RMt = N1 ∑N i =1 Rit , where RMt =return on the market portfolio in time period t, Rit =rate of return on security i in time period t, N =number of securities Figure: Rate of return on the market portfolio Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 4 / 35 Introduction The Effectiveness of Diversification Motivation: How many securities must be included to obtain a reasonably well-diversified portfolio? Figure: Variability and portfolio diversification 8.63 n , where ( σ )2 thus (σS )2 = T can be approximated by σp = 11.91 + n =number of securities. When n = 1, σT = 20.5, σS = 11.9, 0.34. Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 5 / 35 Itroduction Market and Industry Factors Table: Proportion of security risk attribute to market factors Period June 1927-September 1935 October 1935-Feburary 1944 March 1944-July 1952 August 1952-December 1960 Average proportion(%) 58.4 55.7 41.2 30.7 Remarks: The proportion decreases overtime. Analysis over the entire period indicates, market fluctuation accounts for 52% variance of a typical security, a group of industry accounted for another 11%. In an index model, the market can be represented by an index and additional indexes can be added to represent industry factors. Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 6 / 35 Measuring Performance Reward-to-Variability Definition Capital market line re = EM − p σM where re =price of risk reduction for efficient portfolios, EM =expected return, p =pure rate of interest, σM =standard deviation. Corollary The past performance of any portfolio can be represented as: Ap − p 0 r ( )p = v σp0 where ( vr )p =reward-to-variability ratio for portfolio, Ap =actual average return, p 0 =actual pure interest rate, σp0 =actual variability of portfolio. Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 7 / 35 Measuring Performance Reward-to-Variability Figure: reward-to-variability Remark: The slope of the line associated with the portfolio is the reward-to-variability ration. The steeper the line, the better the performance of portfolio. Used to measure the performance of portfolio. Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 8 / 35 Measuring Performance Reward-to-Volatility Definition Security market line rs = Ei − p bi where rs =price of risk reduction for securities, Ei =expected return of security i, p =pure rate of interest, bi =volatility of security i. Corollary The past performance of any portfolio can be represented as: r Ai − p 0 ( )i = b bi0 where ( br )i =reward-to-variability ratio of security i, Ai =actual average return, p 0 =actual pure interest rate, bi0 =actual volatility of security i. Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 9 / 35 Measuring Performance Reward-to-Volatility Figure: Reward-to-volatility Remark: The slope of the line is the reward-to-volatility ratio. The steeper the line, the better it is. Reward-to-volatility ratio is close related to characteristic line. Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 10 / 35 Measuring Performance Differential Return Definition Actual characteristic line represents the relationship of actual return of a security or portfolio and that of market portfolio. Thus, the line passes through the point at which both returns equal their actual average value Ai and AM ; and the slope is actual volatility bi0 . Figure: Actual characteristic line Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 11 / 35 Measuring Performance Differential Return Figure: Differential return Derive the actual characteristic line Y from Ai , AM , bi0 . Line Z represents the corresponding efficient portfolio with same volatility. 0 x = Aib−0 p is the reward-to-volatility ratio, separate it: i Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 12 / 35 Measuring Performance Differential Return r = dh + (AM − p 0 ) b If dh > 0, performance of the security or portfolio was superior to that of a market-based portfolio of comparable volatility; if dh < 0,it was worse. dv = dh · bi0 is the vertical distance from point P to the characteristic line. Definition dv is denoted as differential return. A positive differential return indicates that performance was superior to that of a market-based portfolio of comparable volatility; a negative differential return indicates that it was worse. Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 13 / 35 Measuring Performance Differential Return Figure: compare the performance of two securities Remarks: Based on differential returns, line j was superior to that of i. Based on reward-to-volatility ratio, line i was superior to that of j. The reward-to-volatility ratio can compare securities mutually. The reward-to-volatility ratio and differential return both can compare the performance of a security with that of market. Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 14 / 35 Mutual-Fund Performance Introduction Definition An open-end mutual fund is an institution designed to provide both diversification and professional management at relatively low cost. Some facts about mutual fund: The managers are ready to issue new shares or retire old shares at virtually any time. The net asset value per share=(current market value of the fund’s holding)/(number of shares) Load charge, typically 8 to 10 percent, goes to the sales organization. The managers of the fund are paid separately, usually 0.5% of the total net asset value. Most mutual funds hold over 100 different securities. Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 15 / 35 Mutual-Fund Performance Correlation with Market Figure: Proportion of variance attribute to market Remarks: The graph incudes the result of 115 mutual funds. On average, 85% of the variance of the mutual fund can be attributed to market fluctuations. In sum, most mutual funds are well-diversified. Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 16 / 35 Mutual-Fund Performance Volatility Figure: Volatility of corresponding 115 mutual funds Remarks: The average volatility is 0.84. Most mutual funds perform more conservative than one made up of the securites in S&P Index. The differences in volatility may due to the objectives of the funds. Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 17 / 35 Mutual-Fund Performance Volatility Table: Volatility by type of fund Classification Growth Funds: primary objective is long-term growth of capital Growth-income Funds: emphasis on long-term, consider income Income-growth Funds: emphasis current income, consider long-term Income Funds: primary objective is current income Blanced Funds: relative stability and continuity of income Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) Number of funds 31 Average volatility 0.970 30 0.941 15 0.856 9 0.674 30 0.645 March 2012 18 / 35 Mutual-Fund Performance Performance of Mutual Fund Figure: Reward-to-variability ratios, 1954-1963 (net returns) Remarks: 11 funds had ratios exceeding Dow-Jones’ portfolio; 23 are smaller. The results are based on net return. Apply same analysis on gross return, 19 had ratios larger than that of Dow-Jones’ portfolio; 15 had smaller ratios. Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 19 / 35 Mutual-Fund Performance Performance of Mutual Fund Figure: Average return and volatility 1955-1964 (115 mutual funds) Remarks: M indicates the performance of market portfolio, based on S&P Index. Left graph indicates net performance, more funds plot below security market line than above it. Right graph indicates gross performance, points scatter randomly around the line. Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) Plausible explanation: excessive managing expenditures. March 2012 20 / 35 Mutual-Fund Performance Performance of Mutual Fund Figure: Differential returns 1945-1964 (115 mutual funds) Remarks: Based on net values, the average of differential return was -1.1%; 76 had negative differential returns. Based on gross values, the average was -0.4%; 55 had negative values. Mutual funds do no better on average than market based portfolios of comparable volatility. Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 21 / 35 Mutual-Fund Performance Fund Managers got talent or luck? Figure: Rank of reward-to-variability (net value) Remarks: There are slight positive relationships in both graphs. Comparison suggests that difference in performance based on net returns may be due more to cost of management than effectiveness. Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 22 / 35 Mutual-Fund Performance Fund Managers got talent or luck? Well-managed funds with enjoy runs of successive superior performance. Table: Frequency of successive superior Length of run (number of successive years of superior performance) 1 2 3 4 Number of instances 574 312 161 79 Percent of instances followed by another year of superior performance (%) 50.4 52.0 53.4 55.8 Remarks: Typical fund obtained superior performance 50.2% of the time. Funds with one to four successive prior years of superior performance had slightly (less than 6%) better success. Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 23 / 35 Mutual-Fund Performance Fund Managers got talent or luck? Figure: Volatilities of 56 funds over two periods Remarks: The relationship is clearly positive, though not perfect. Regardless of the overall performance, the target level of volatility is well preserved. Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 24 / 35 Mutual-Fund Performance Summary The conclusions of this section can be summarized: Most funds diversified well. Most managers selected a general-risk class and maintained their stated positions reasonably well. On the average, funds did no better, before expenses, than market-based portfolios of comparable volatility. On the average, funds did worse, after expenses, than market-based portfolios of comparable volatility. Few, if any, funds consistently performed better than market-based portfolios of comparable volatility. Most funds appear to have spent too much searching for mispriced securities. Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 25 / 35 Mutual-Fund Performance Further Discussion Other factors: Sales charge: no-load funds outperform load funds, switch funds perform worst. Turnover: managers with low turnover outperform managers with high turnover. The ratio of expenses to assets: the less, the better. Fund size: unable to find any impact of size on performance. Large funds have more to spend for information and analysis. Large funds have more impact on market, when they engage in purchase and sales. Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 26 / 35 The Shapes of Distributions Stable Paretian (Pareto-Levy) Distributions Figure: rate of return on market portfolio from 1926 to 1965 Remark: Normal Distribution contradicts reallity. Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 27 / 35 The Shapes of Distributions Stable Paretian (Pareto-Levy) Distributions Definition Random variables {Xn } are independent copies of X , then X is said to follow Stable Distribution if: X1 + X2 + · · · + Xn = Cn X + Dn 1 where Cn = n α dictates the stability property, Dn is a real number. Four parameters characterized Stable Distribution α the characteristic exponent; a measure of the height of the extreme tail areas of the distribution α ∈ (0, 2) β an index of skewness β ∈ [−1, 1] γ a scale parameter γ ∈ (0, +∞) δ a location parameter δ ∈ (−∞, +∞) Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 28 / 35 The Shapes of Distributions Stable Paretian (Pareto-Levy) Distributions Figure: Stable Distributions and parameters Remarks: When α = 2, β = 0, it is a Gaussian Distribution with 2 γ = σ2 , δ =expected value Empirical work suggests that actual rates of return are best approximated by 1.7 ≤ α ≤ 1.9 When 1.7 ≤ α ≤ 1.9, E (X ) = δ, Var (X ) = ∞ Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 29 / 35 The Shapes of Distributions Index Model Based on Stable Distribution Definition Single-index model of return on security i can be represented as: Ri = ai + bi I + ci where ai , bi =parameters, ci =uncertain variable, I =level of index. Corollary Assume I and {ci } follow Stable Paretian Distribution with same α∗ . Then, the risk of a portfolio can be represented as: N γp = ∑ Xiα γc ∗ ∗ i + bpα γI i =1 where γp = portfolio risk, γci =risk unique to security i, bp = ∑N i =1 Xi bi , γI =index risk. Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 30 / 35 The Shapes of Distributions Index Model Based on Stable Distribution Corollary Assume portfolios are well-diversified and half the risk of a typical security is due to uncertainty of index, then: ∗ 1 ∗ γp = [n · ( )α + 1]bpα γI n where n =the number of securities in the portfolio. Figure: portfolio risk and number of security Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 31 / 35 The Shapes of Distributions Summary Practical implications: Gaussian distrubtion fails to descirble the actual rate of return. Stable Paretian Distribution with 1.7 ≤ α ≤ 1.9 approximates reallity best. Variance is an untrustworthy indicator of risk. Volatility can still be used. Index models are efficient in normative applications. Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 32 / 35 Using the Past to Predict the Future Is Future like Past? The cause of difference between future and past: Chance events Changes in management The price process might be greatly different. Diversification can stabilize volatilities of securities and portfolios. Figure: Volatility in two periods Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 33 / 35 Using the Past to Predict the Future Solve the Problem Based on above discussion, an investor could select the portfolio based on past records: maximize ∑N i = 1 X i Ei ∗ subject to ∑N i =1 Xi bi = bp and 0 5 Xi 5 1 n for security i. where Ei is average past return of security i, bi is actual past volatility of security i, bp∗ is some desired level of volatility, n is a number large enough to force adequate diversification. Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 34 / 35 References William F. Sharpe, Portfolio Theory and Capital Markets, McGraw-Hill Book Company 1790. Edward J. Elton, Martin J. Gruber, Stephen J. Brown, William N. Goetzmann, Modern Portfolio Theory and Investment Analysis, John Wiley & Sons. Inc. 2009. Svetlozar T. Rachev, Young Shin Kim, Michele Leonardo Bianchi, Frank J. Fabozzi, Financial Models with Levy Process and Volatility Clustering, June 2010 Yiyang Yang (SUNY at Stony Brook) Record (AMS 512) March 2012 35 / 35