Summary: | 碩士 === 淡江大學 === 財務金融學系碩士班 === 94 === We apply Ederington and Guan (2005), to examine the forecasting ability of sixth time-series volatility models, including historical variance, EWMA, GARCH(1,1)-N, GARCH(1,1)-G, and restricted least squares (RLS) and GEN. We seek to determine why one model or group of models forecasts another focusing on three issues: 1、the proper weighting of older versus recent observations, 2、the relevance of the parameter estimation procedure, and 3、we use four criterions to measure forecast ability. Our evidence indicates
1、The GARCH(1,1)model puts too much weight on the most recent observations and not enough on older observations.
2、Different parameter estimation procedures result in quite different parameter estimates for the same model.
3、The GEN model is the best volatility forecasting model, RLS model is the second, and GARCH(1,1)-G model is always superior to GARCH(1,1)-N model.
|