NEW MODELS FOR HIGH AND LOW FREQUENCY VOLATILITY Robert Engle NYU Salomon Center Derivatives Research Project Derivatives Research Project
FORECASTING WITH GARCH
DJ RETURNS
DOW JONES SINCE 1990 Dependent Variable: DJRET Method: ML - ARCH (Marquardt) - Normal distribution Date: 01/13/05 Time: 14:30 Sample: Included observations: 3789 Convergence achieved after 14 iterations Variance backcast: ON GARCH = C(2) + C(3)*RESID(-1)^2 + C(4)*GARCH(-1) CoefficientStd. Errorz-StatisticProb. C Variance Equation C9.89E E RESID(-1)^ GARCH(-1) R-squared Mean dependent var Adjusted R-squared S.D. dependent var S.E. of regression Akaike info criterion Sum squared resid Schwarz criterion Log likelihood Durbin-Watson stat
DEFINITIONS r t is a mean zero random variable measuring the return on a financial asset CONDITIONAL VARIANCE UNCONDITIONAL VARIANCE
GARCH(1,1) The unconditional variance is then
GARCH(1,1) If omega is slowly varying, then This is a complicated expression to interpret
SPLINE GARCH Instead, use a multiplicative form Tau is a function of time and exogenous variables
UNCONDITIONAL VOLATILTIY Taking unconditional expectations Thus we can interpret tau as the unconditional variance.
SPLINE ASSUME UNCONDITIONAL VARIANCE IS AN EXPONENTIAL QUADRATIC SPLINE OF TIME For K knots equally spaced
ESTIMATION FOR A GIVEN K, USE GAUSSIAN MLE CHOOSE K TO MINIMIZE BIC FOR K LESS THAN OR EQUAL TO 15
EXAMPLES FOR US SP500 DAILY DATA FROM 1963 THROUGH 2004 ESTIMATE WITH 1 TO 15 KNOTS OPTIMAL NUMBER IS 7
RESULTS LogL: SPGARCH Method: Maximum Likelihood (Marquardt) Date: 08/04/04 Time: 16:32 Sample: Included observations: Evaluation order: By observation Convergence achieved after 19 iterations CoefficientStd. Errorz-StatisticProb. C(4) E W(1)-1.89E E W(2)2.71E E W(3)-4.35E E W(4)3.28E E W(5)-3.98E E W(6)6.00E E W(7)-8.04E E C(5) C(1) C(2) Log likelihood Akaike info criterion Avg. log likelihood Schwarz criterion Number of Coefs.11 Hannan-Quinn criter
ESTIMATION Volatility is regressed against explanatory variables with observations for countries and years. Within a country residuals are auto- correlated due to spline smoothing. Hence use SUR. Volatility responds to global news so there is a time dummy for each year. Unbalanced panel
ONE VARIABLE REGRESSIONS
MULTIPLE REGRESSIONS
IMPLICATIONS Unconditional volatility varies over time and can be modeled Volatility mean reverts to the level of unconditional volatility Long run volatility forecasts depend upon macroeconomic and financial fundamentals
HIGH FREQUENCY VOLATILITY
WHERE CAN WE GET IMPROVED ACCURACY? USING ONLY CLOSING PRICES IGNORES THE PROCESS WITHIN THE DAY. BUT THERE ARE MANY COMPLICATIONS. HOW CAN WE USE THIS?
ONE MONTH OF DAILY RETURNS
INTRA-DAILY RETURNS
ARE THESE DAYS THE SAME?
CAN WE USE THIS INFORMATION TO MEASURE VOLATILITY BETTER? DAILY HIGH AND LOW DAILY REALIZED VOLATILITY
PARKINSON(1980) HIGH LOW ESTIMATOR IF RETURNS ARE CONTINUOUS AND NORMAL WITH CONSTANT VARIANCE,
TARCH MODEL WITH RANGE C1.07E E RESID(-1)^ RESID(-1)^2*(RESID(-1)<0) GARCH(-1) RANGE(-1)^ Adjusted R-squared S.D. dependent var S.E. of regression Akaike info criterion Sum squared resid Schwarz criterion Log likelihood Durbin-Watson stat
Robert F. Engle Giampiero M. Gallo A MULTIPLE INDICATOR MODEL FOR VOLATILITY USING INTRA-DAILY DATA Robert F. Engle Giampiero M. Gallo Forthcoming, Journal of Econometrics
Absolute returns Insert asymmetric effects (sign of returns) Insert asymmetric effects (sign of returns) Insert other lagged indicators Insert other lagged indicators
Repeat for daily range, hl t : And for realized daily volatility, dv t :
Smallest BIC-based selection
Forecasting one step-ahead one step-ahead multi-step-ahead multi-step-ahead
Term Structure of Volatility 1
IMPLICATIONS Intradaily data can be used to improve volatility forecasts Both long and short run forecasts can be implemented if all the volatility indicators are modeled Daily high/low range is a particularly valuable input These methods could be combined with the spline garch approach.