In your answers to the questions below, you should present your Eviews equation estimation output as it would be in published academic papers. (Examine several such papers, the approaches to presentation are fairly standard.) Raw Eviews output should be included only in an Appendix.
The data required for the coursework is contained in the excel file `Coursework_2.xls’ in the coursework section on Moodle. The file contains daily return and 5-min realised variance data for FTSE100 from January 2000 to December 2017 that we will use to estimate and forecast the volatility of FTSE100.
Consider the observations for the FTSE 100 stock index return series for the period ranging from 2000 to 2015. Build an appropriate ARMA model and test for ARCH effects in the series of daily returns.
Using the observations for years 2000 to 2015, estimate the following models from the GARCH family, selecting the appropriate lags:
- GARCH with normal innovations
- GJR with normal innovations
- a specification of your choice
Always using as the conditional mean equation the ARMA specification that you identified in Question 1 above.
For each of these models, forecast the daily volatility (square root of variance) for the last two years in your sample. Plot your forecasts against the actual realizations of the variable (realised volatility proxy), as well as the forecast errors (i.e. actual – forecast). Interpret the results.
Compute the MSE, MAE and MAPE for all 3 models which you estimated. According to each of these criteria, which model forecasts best and which model forecasts worst?
