Вы находитесь на странице: 1из 7

TO MODEL STOCK PRICE

VOLATILITY USING GARCH


MFEM Assignment
Abstract
Modelling time series data of stock returns using MSARIMA and EGARCH

Objective
To forecast stock prices of All Cargo company using ARIMA GARCH modelling
Data
Daily data has been used from the period 2007 to 2012
Step 1
Check for stationarity of data.
We checked for stationarity using ADF test and found the data to be non-stationary in level and
stationary in first difference. However, on checking the correlogram in first difference we reached
white noise process, hence, we cannot use de trend option. We de seasonalized the data using
below command.
Series dall=dlog(allcargo,0,5)

Re ran the ADF test and found the series to be stationary.

For estimation we use regression analysis and give the below command
ls dlog(allcargo,0,5) ar(1) ma(5) sar(10) sma(10)

We check the correlogram of residuals squared and find that the probability values are<0.05
pointing to heteroskedasticity of variance.

To confirm we run arch test and find that variance is heteroskedastic

We estimate using GARCH EGARCH model

We find that the constant, size, asymmetry and persistence terms are significant as the null
hypothesis of no persistence and no asymmetry is rejected (p value<0.05). The coefficient of the
asymmetry term is -0.11 indicating presence of negative shock that increase volatility more than
positive shocks, coefficient of persistence is 0.87 indicating there is persistence of shocks in volatility.
Next we again check for heteroskedasticity to see if arch effect is left, we find that now there is no
arch effect.
We run the forecast static forecast and get MAPE of 1.77%

Вам также может понравиться