VOLATILITY FORECASTING

1
VOLATILITY FORECASTING

• Steven Poher
• Ramzi Rached
• Ricardo Uribe
• Dongting Zheng

Global Investment Management
2
AGENDA

• Objective
• Background Information
• Forecasting Models
• Data set
• Methodology
• Results
• Conclusion

3
OBJECTIVE

• Objective
• To establish a variance forecasting model
• Why?
• Important for risk managers (VaR)
• Used to price options
• Volatility Return investment decision

4
BACKGROUND INFORMATION

• Realized / observed volatility is measured by
  squared returns
• Volatility displays a positive correlation with
  its own past
• Simple Model
• PB Equal weights on the past m observations

5
FORECASTING MODELS

• More flexible model ? Simple GARCH or GARCH (1,1)
  
• Extended to Local and Global Instruments
• Models to be tested for this project
• GARCH (1,1)
• GARCH (1,1) Local
• GARCH (1,1) Global
• GARCH (1,1) Local Global

6
DATA SET

• Source DataStream
• Period 3/27/1998 - 3/28/2008 (10 years)
• Granularity 1 day

Country Index
NIKKEI 225
CAC 40
DAX 30
FTSE 100
SP 500
7
DATA SET

• Local Instruments
• Change in Exchange Rates
• EUR / USD / JPY / GBP
• Change in short-term interest rates
• T-Bill (US) / BTAN (FR)
• Global Instruments
• Change in Short-term Eurodollar rate
• Change in the Term Structure spread

8
METHODOLOGY

• Using EXCEL, test our 4 models for each of our 5
  markets
• Use Maximum Likelihood Estimation (MLE) to
  estimate ? / ? / ? / ? EXCEL Solver
• Test the models using a regression of Squared
  Returns vs. Forecasted Variance
• Discuss the statistical significance of the
  regression / Select the best model for a given
  country

9
METHODOLOGY - EXAMPLE
10
RESULTS

• Best models for each country

Country Model R2
GARCH Change in Term Structure Spread (G) 1.21
GARCH Change in / (L) 14.12
GARCH Change in / (L) Change in Term Structure Spread (G) 15.56
GARCH (1,1) Change in / (L) Change in ST Eurodollar (G) 14.23
GARCH 10.58
11
RESULTS

• Best model for German Market
• R2 of 15.56
• Final equation
• Simple GARCH
• Change in / Exchange (L)
• Change in Term Structure Spread (G)

12
RESULTS
13
CONCLUSION

• No universal model
• Different countries different models
• Good proxy for DE / Bad for JP
• GARCH could also be extended
• Leverage effects
• Day-of-week effects
• Jumps
• Economic intuition reality check

14
QUESTIONS?

• THANK YOU