VaR and Changing Volatility Jorion, Chapter 8 VaR and the Unreal World The Pitfalls of VaR estimates
Summary Picture of changing volatility Moving averages and rolling VaR’s Riskmetrics and weighted variances GARCH modeling of volatility Correlations and portfolios
Picture of Changing Volatility dowvolplt.m
Summary Picture of changing volatility Moving averages and rolling VaR’s Riskmetrics and weighted variances GARCH modeling of volatility Correlations and portfolios
Moving Average of Volatility Rolling moving average of returns squared madowvar.m
Moving Average of Volatility Brooks/Persand and Hoppe papers –Tradeoff between small and large samples –Conditional volatility versus large sample size –Small often looks better –Trickier with weightings Interesting question –Evaluation? (graphical)
Summary Picture of changing volatility Moving averages and rolling VaR’s Riskmetrics and weighted variances GARCH modeling of volatility Correlations and portfolios
RiskMetrics VaR h(t) = variance at time t Smooth weighting of past volatility
Riskmetrics VaR rmdowvar.m hrmdowvar.m
Summary Picture of changing volatility Moving averages and rolling VaR’s Riskmetrics and weighted variances GARCH modeling of volatility Correlations and portfolios
GARCH Modeling GARCH(1,1): –Complete model for changing variance
GARCH Modeling Forecasting Variance h(t)
How Does this Differ from Riskmetrics? For 1 horizon, not much Multi-horizon is different h(t+m) is needed
GARCH Variance T periods in the future
GARCH Variance Forecast Days Ahead Variance Shock Unconditional Variance
RiskMetrics VaR Forecasts h(t) = variance at time t
Summary Picture of changing volatility Moving averages and rolling VaR’s Riskmetrics and weighted variances GARCH modeling of volatility Correlations and portfolios
Correlations Moving averages Riskmetrics (examples) GARCH
Riskmetrics Correlation Example (rmcorr.m)
Crashes and Correlations Large down moves connected to increases in correlations Implications for risk management and portfolio construction Reliability in the data?
Final Suggestions on Volatility Options data and implied volatility High frequency data High/low range data