A Comparative Study Of CARR Model And GARCH Model On Volatility Prediction

Volatility forecast has become a major topic in domestic and international capital markets and econometric research since the 1980 s.The subject of this paper is the comparative study of volatility prediction between CARR model and GARCH model.Based on this paper,the related theoretical concepts of volatility prediction are first studied,and then study the relevant models of volatility prediction.Next,the statistical characteristics of different markets are analyzed.Finally,empirical research on related models is carried out.Inconsistent,this paper has different choices in the distribution of model perturbation terms.GARCH model and CARR model will discuss the applicability of normal distribution,t distribution and generalized exponential distribution(GED),so as to select the optimal model.In order to ensure the applicability of each model in different markets,the SSE will be selected.The KLCI,Hang Seng Index,S&P 500 Index and Nikkei 225 Index demonstrate the applicability of the two models in different markets.In the choice of time period and volatility frequency,this paper selects the longest interval data according to the data availability(that is,the beginning of data recording on wind)to ensure the applicability of GARCH model and CARR model in different economic stages,in fluctuation In the choice of rate frequency,this paper selects the daily market data.This is because scholars have shown that the statistical time series of financial time series are most obvious in the daily market sequence and the weekly market sequence.This paper intends to select the daily market sequence and start the day to the end of 2018.The transaction data was model-fitted as data within the sample,and the transaction data from the beginning of 2019 to the end of June 2019 was used as model data evaluation.According to the model fitting effect evaluation index,the LLF,AIC and BIC indicators indicate that the GARCH model with the t-distribution of the disturbance term is better than the GARCH model with the normal distribution of the disturbance term;the CARR model with the GED distribution of the disturbance term is better than the disturbance term.Distribution of CARR models.According to the evaluation results of the model evaluation indicators,namely MSE,MAE and MAPE,the research results show that the CARR model has lower error results in the volatility prediction than the GARCH model,that is,the CARR model uses more information than the GARCH model,so the model prediction effect is better.