Econ 201 by David Kim
Measuring & Forecasting Measuring and forecasting latent volatility is important in regards to: – Asset allocation – Option pricing – Risk management
Brownlees and Gallo (2009) Looks at different volatility measures improve out-of-sample forecasting ability of standard methods Looks into issue of forcasting Value-at-Risk (VaR) by looking at various volatility measures – Realized volatility – Bipower volatility – Two-scales realized volatility – Realized kernel – Daily range
VaR Modeling Assumes – h t – conditional variance of daily return – n t – i.i.d. unit variance from an appropriate cumulative distribution F – One-day-ahead VaR is defined as maximum one- day-ahead loss
Volatility Measures Realized volatility Bipower realized volatility
Volatility Measures (cont’d) Two-scales realized volatility – Let – Define: – This estimator combines information from both slow and fast time scales
Volatility Measures (cont’d) Realized kernel – Y h (p t ) = – k( ) = appropriate weight function as the sample frequency increases, realized kernel can get the fastest convergence rate ·
Volatility Measures (cont’d) Daily Range – p high,t – largest log-price – p low,t – lowest log-price – Affected by a much lower measurement error It is as precise as realized volatility if using a sample of low frequency data and certain conditions
Companies HJ Heinz Company (HNZ) and Kraft Foods Inc. (KFT) – Consumer goods sector within the food industry – Both diversified companies
HNZ: Price
HNZ: Returns
HNZ: Relative Contribution of Jumps
HNZ: RV Volatility Signature
HNZ: BV Volatility Signature
KFT: Price
KFT: Returns
KFT: Relative Contribution of Jumps
KFT: RV Volatility Signature
KFT: BV Volatility Signature
HNZ: RV
HNZ: BV
KFT: RV
KFT: BV
Look further into all volatility measures If an appropriate area for research, include more stocks Other potential areas of interest: – How presence of jumps has information relevant to forecasting volatility HAR modelling frame