QBUS6830 Financial Time Series and Forecasting
Financial Time Series and Forecasting
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: THEend8_
Business School
QBUS6830
Financial Time Series and Forecasting
Group Assignment Part II
Group Assignment Part II will contribute 25% towards your final grade and is to be
completed in your existing group for Part I. The due date is Friday 12th November, by
11:59pm via online submission in Canvas.
Obtaining Data and Excel Questions
For this assignment you will need to use the same assets you used in Part I. The Jupyter
notebook file part_2_data_prep.ipynb will assist you to extract 3 of these assets and price
data on those assets for the dates required.
Submission Requirements
The assignment submission consists of 3 parts; a Jupyter notebook file with numerical and
written answers and all Python code used, plus group Meeting Minutes and a Peer
Assessment form.
In the Jupyter notebook file, you should provide your answers to all questions below, by
including Python output and written answers (you can type in any input cell by using the
“markdown” option in Jupyter). Only one student should submit this file, per group.
A template for the group Meeting Minutes will be placed on Canvas and should have entries
for at least 3 group meetings. Only one student should submit per group.
Finally, the Peer Assessment Form, requires each group member to assess the
contributions of their fellow group members and will be used to adjust marks in the case
where student contribution differs significantly across group members. Each member of the
group must submit their own Peer Assessment form.
Submission links will become available in Canvas shortly before the due date.
QBUS6830, Financial Time Series and Forecasting
Volatility Modelling and Risk forecasting
In this assignment you are required to build and assess a range of models for forecasting
market risk for asset 1 in your data set.
You will employ a range of volatility models and will examine the accuracy of 1-period ahead
volatility forecasts. Note that once you have chosen models based on your in-sample period,
you should use those same models for all of your forecast periods: e.g. if you use information
criteria to determine model 2 is an ARCH(3) model, based on your in-sample data, then you
should use the ARCH(3) model for all of your model 2 forecasts. The volatility proxies your
group must use are proxies 2 and 4. Use a fixed-size rolling window, fixed horizon, moving
origin approach for generating these forecasts, updating your model estimates daily.
Next you will generate 1-period ahead Value at Risk and Expected Shortfall forecasts, at the
2.5% and 1% levels, for all volatility forecast models, as well as a symmetric CAViaR model.
Your data set should range Jan 4, 2010 to October 5, 2021. The last 500 days of your data
should form your forecast period, whilst all the data before that forms your in-sample period.
Part I (55 marks)
(a) (10 marks) You must choose 3 parametric models to fit to your in-sample data for
asset 1, at least one of these must be an asymmetric volatility model, at least one
other must be symmetric. At least one model must use lagged asset 2 as a regressor
in the mean equation. At least one model must use lagged asset 3 in the same way.
Motivate your model specification choices in detail for one model (only), e.g. order
selection, etc. Briefly mention your choices for the other two.
(b) For (only) one of the asymmetric models:
i. (10 marks) Provide your estimation results for asset 1 and write down the
estimated model equations. Discuss the statistical significance and interpret
the estimated parameters. Discuss the asymmetry found by this model.
ii. (10 marks) Perform a thorough diagnostic analysis to assess the fit of the
model.
iii. (7 marks) Discuss any component(s) you might add to the model that might
potentially capture any model mis-specifications found; motivate your choices.
Propose a model based on your findings (it can be the same, if appropriate, or
different). This is the asymmetric model you will now use below.
Below, you will use all models above to produce volatility, VaR and ES forecasts (at both
2.5% and 1% levels); as well as some historical simulation methods specified below.
You are also required to include and consider three forecast combination techniques (for
each of volatility, VaR and ES separately): (i) an equally weighted forecast combination of
volatility forecasts and of VaR and of ES forecasts, using only the parametric volatility
models (and CaViaR for VaR); (ii) an equally weighted forecast combination of all volatility
forecasts and of all VaR and of all ES forecasts for all models/methods; and (iii) a forecast
combination for volatility that weights each model/method’s forecast in proportion to its in-
sample accuracy (by RMSE); a combination for VaR and ES that weights each
model/method’s forecast in proportion to its quantile (VaR) or joint (ES) loss in the in-sample
period.
QBUS6830, Financial Time Series and Forecasting
(c) (5 marks) Form all your parametric 1-step-ahead volatility model forecasts for asset
1 for the entire forecast period of 500 days. Also do this using adhoc historical
simulation measures, using 25 days and 100 days. Present and discuss all the
volatility forecasts and accuracy measures.
(d) (6 marks) Form all your 1-step-ahead tail risk model forecasts for asset 1 for the entire
forecast period of 500 days. Also use adhoc methods using 100 days and 200 days.
Present and discuss the 1- period ahead VaR forecasts for all volatility models and
methods, plus the CaViaR model. Also present and discuss tests and accuracy
measures for all these VaR forecasts.
(e) (7 marks) Present and discuss the 1-period ahead ES forecasts for all volatility
models in (d). Also use adhoc methods using 100 days and 200 days. Present and
discuss accuracy measures and tests for ES as well as joint accuracy measures for
all the VaR and ES forecasts.