Capital Market Research Forum 4/2555

Transcription

1 Capital Market Research Forum 4/2555 Hedging Effectiveness of SET50 Index Futures: Empirical Studies and Policy Implications Thaisiri Watewai, Ph.D. Chulalongkorn Business School Chulalongkorn University 23 March 2012

2 Hedging Effectiveness of SET50 Index Futures: Empirical Studies and Policy Implications Thaisiri Watewai, Ph.D. Chulalongkorn Business School Chulalongkorn University 1

3 Motivations SET50 Index Futures o Launched on April 28, 2006 o Adjust portfolio exposure to the index o Effectiveness of managing the exposure depends on many factors: Correlation between the return of the futures and that of the index Liquidity cost Transaction cost (brokerage commission fees and taxes) 2

4 Motivations (cont.) Alternatives o ThaiDEX SET50 Exchange Traded Fund (TDEX) Require full capital investment Short selling can be costly o SET50 Index Options Highly illiquid 3

5 Motivations (cont.) SET50 index futures o Pros Requires only margin deposits Cost of shorting the futures is small Liquid o Cons Relatively large contract size Predetermined expiry date High transaction costs (?) 4

6 Main Findings Relatively small liquidity cost Relatively large transaction cost Significantly improve cost-adjusted Sharpe ratio Lower global minimum variance Improvement depends on ability to forecast market trend 5

7 Outline Literature Review Cost-Adjusted Mean-Variance Model Liquidity Cost Estimation Factor Model Hedging Effectiveness and Cost Contributions Extensions Policy Implications and Conclusions 6

8 Literature Review Hedging Effectiveness: Objectives o Based on a given and fixed portfolio o Determine the optimal hedge ratio Minimum Variance (Ederington; Johnson; Myers and Thompson) Mean-Variance (Cecchetti, Cumby and Figlewski; Howard and D Antonio; Hsin, Kuo and Lee) Expected Utility Maximization (Benninga, Eldor and Zilcha) Mean Extended-Gini Minimization (Cheung, Kwan and Yip) Generalized Semivariance Minimization (De Jong, De Roon and Veld) 7

9 Literature Review (cont.) Hedging Effectiveness: Econometrics o How to accurately estimate the hedge ratio OLS (Junkus and Lee) GARCH (Baillie and Myers) Random coefficient (Grammatikos and Saunders) Cointegration (Ghosh) Mean Extended-Gini (Kolb and Okunev) Generalized Semivariance (Lien and Tse) 8

10 Literature Review (cont.) Hedging Effectiveness: Trading costs o Always ignore associated trading costs o Few exceptions Lence (1995, 1996) Brokerage fee and initial margin deposit Economic value of complicated estimation techniques for minimum variance hedge ratio is negligible Maybe optimal not to hedge at all Does not consider liquidity costs Price impact: Chan and Lakonishok; Keim and Madhavan; Sharpe et al. 9

11 Literature Review (cont.) Contributions o Interaction between the use of futures and the portfolio choice of stocks o Include liquidity cost in addition to transaction cost Asymmetric liquidity cost curve 10

12 Cost-Adjusted Mean-Variance Model Objective o Maximize cost-and-risk adjusted mean return (Mean - L Cost - T Cost) - Variance Universe o Stocks in SET50,TDEX, SET50 index futures Budget: 11

13 Cost-Adjusted Mean-Variance Model (cont.) Decision variables o Weight in stock : o Risk-free weight : o Futures weight : = o Vector of weights : 12

14 Cost-Adjusted Mean-Variance Model (cont.) Mean - Variance o Mean : where : risk-free rate : vector of mean returns Portfolio Return Returns of stocks, futures, cash Weights of stocks, futures, cash 13

15 Cost-Adjusted Mean-Variance Model (cont.) Mean Variance o Variance : where : covariance matrix of returns Portfolio Variance Weights of stocks, futures, cash Covariance of stocks, futures, cash Weights of stocks, futures, cash 14

16 Cost-Adjusted Mean-Variance Model (cont.) Liquidity Cost o Cost Asymmetric Liquidity Cost Curve Total Cost Trading Value 15

17 Cost-Adjusted Mean-Variance Model (cont.) Liquidity Cost o Cost Asymmetric Liquidity Cost Curve where : traded value : liquidity cost parameter for buy : liquidity cost parameter for sell 16

18 Cost-Adjusted Mean-Variance Model (cont.) Liquidity Cost o Re-balance from to Buy Sell o Percentage of Liquidity Cost : 17

19 Cost-Adjusted Mean-Variance Model (cont.) Transaction Costs o Variable cost + Fixed cost o Stocks and TDEX : = 0.25% + 7% VAT = % o Futures : = % VAT = baht 18

20 Cost-Adjusted Mean-Variance Model (cont.) Transaction Costs o Re-balance from Stocks: per traded value to Futures: per contract o Percentage of Transaction Cost : 19

21 Cost-Adjusted Mean-Variance Model (cont.) Transaction Costs o Pre-determined Expiry Date of Futures Buy futures T-cost Expiry date L-cost T-cost 12 February 31 March 20

22 Cost-Adjusted Mean-Variance Model (cont.) Transaction Costs o Pre-determined Expiry Date of Futures o Percentage of Transaction Cost : 21

23 Cost-Adjusted Mean-Variance Model (cont.) Formulation o Objective: o Constraints: o No cash-borrowing o No short-selling o Limit stock concentration at 20% o Limit position by trading value at 50% o Maximum number of futures contracts at 20,000 contracts o Margin deposit at 50,000 baht per futures contract 22

24 Liquidity Cost Estimation Data o Intraday bid-ask prices of each stocks, TDEX and futures (SET, TFEX, Thomson Reuters) o Sample 20 points for every 5 minutes from three best bid and ask prices Method o Approximate the piecewise linear liquidity cost curve by two quadratic functions 23

25 Liquidity Cost Estimation (cont.) Example: 24

26 Liquidity Cost Estimation (cont.) Results : 25

27 Liquidity Cost Estimation (cont.) Results : By security Ticker Estimate (x 10-8 ) Rank PTT PTTEP BANPU TOP Futures LH TDEX TRUE MAKRO

28 Liquidity Cost Estimation (cont.) Forecasting o Average of last 10 trading days as forecast of next day o Forecasting performance: 27

29 Factor Model Multifactor model (Chincarini and Kim) Factor choices o Market : 1 o Value : PE ratio o Size : log(market Capt) o Momentum : past 12-month performance o Recommendation : recommendation score 28

30 Factor Model (cont.) Data o Thomson Reuters Datastream : Daily PEs, market capitalizations, total index returns, and recommendation scores (IBES) of stocks in the SET50 index from the database. o Bloomberg : Daily total index return of TDEX and futures o Thai BMA : Daily yield of one-month treasury bill 29

31 Factor Model (cont.) Descriptive Statistics o The market factor : most volatile standard deviation and excess kurtosis. 30

32 Factor Model (cont.) Forecasting o Mean : where : factor mean Expected Stock Return Factor Beta Expected Factor Return o One-year period with exponential weights 31

33 Factor Model (cont.) Forecasting o Variance : where : factor covariance : residual covariance Stock Variance Factor Beta Factor Covariance Factor Beta o One-year period with exponential weights 32

34 Hedging Effectiveness and Cost Contributions Setup o Time period: January 2008 December 2009 o Frequency: daily trading o Initial budget : 1,000 million baht Analysis o Ex-ante : expected returns before re-balancing o Ex-post : realized returns o Both are cost-and-risk adjusted 33

35 Hedging Effectiveness and Cost Contributions (cont.) Scenarios o Futures not allowed : MV o Futures allowed: Liquidity + Transaction costs : MV Liquidity cost : MV Transaction cost : MV No cost : MV o Always include liquidity and transaction costs of stocks and TDEX 34

36 Hedging Effectiveness and Cost Contributions (cont.) Performance Analysis o Cost-adjusted mean-variance frontier where o Liquidity cost o Transaction cost : portfolio s Sharpe ratio Return Risk 35

37 Hedging Effectiveness and Cost Contributions (cont.) Results: Ex-ante MV frontier 36

38 Hedging Effectiveness and Cost Contributions (cont.) Results: Ex-post MV frontier 37

39 Hedging Effectiveness and Cost Contributions (cont.) Results: Ex-post liquidity costs 38

40 Hedging Effectiveness and Cost Contributions (cont.) Results: Ex-post transaction costs 39

41 Hedging Effectiveness and Cost Contributions (cont.) Results: Ex-post cumulative returns and weights 40

42 Extensions Two extensions o Naïve forecasting model One-year equally weighted sample means for Five-year equally weighted sample covariances for o Minimum stock holdings (LTF) Minimum of 65% in stocks 41

43 Extensions (cont.) Naïve forecasting model : Ex-ante MV frontier 42

44 Extensions (cont.) Naïve forecasting model : Ex-post MV frontier 43

45 Extensions (cont.) Minimum stock holdings o MV frontier where : expected value at min global variance : std deviation at min global variance : min global variance Sharpe ratio o LTF : 65% minimum stock holdings 44

46 Extensions (cont.) Minimum stock holdings : Ex-ante MV frontier 45

47 Extensions (cont.) Minimum stock holdings : Ex-post MV frontier 46

48 Extensions (cont.) Minimum stock holdings : Ex-post liquidity costs 47

49 Extensions (cont.) Minimum stock holdings : Ex-post transaction costs 48

50 Extensions (cont.) Minimum stock holdings : Ex-post weights 49

51 Policy Implications and Conclusions Significant improvements on both ex-ante and ex-post Sharpe ratio (given ability to forecast market trend) Factor model Naïve model Min stock holdings* MV MV FLT Increase Ex-ante Ex-post Ex-ante Ex-post Ex-ante Ex-post * Minimum global variance Sharpe ratio 50

52 Policy Implications and Conclusions (cont.) Lower global minimum variance on both exante and ex-post bases Min stock holdings MV MV FLT Decrease Ex-ante Ex-post Relatively small liquidity cost Relatively large transaction cost 51

53 Policy Implications and Conclusions (cont.) Current market structure and liquidity for the SET50 index futures well facilitate investors with large portfolio values (1,000 million baht) Realized benefits depend also on the ability to forecast the market trends, and constraints faced by fund managers 52

54 Policy Implications and Conclusions (cont.) Fund managers must understand the role of the futures in improving the risk-adjusted performance Do not be misled by the fact that using a short position of the futures to hedge the market risk may reduce the realized return during the market upturn 53