matlab代写-F 609/629
AcF 609/629 - Financial Econometrics
Professor Ingmar Nolte
Dr Sandra Nolte
Coursework Assignment
16th February 2021 12:00pm to 16th March 2021 12:00pm UK time
This assignment contains one empirical question worth 100 marks and consti-
tutes 20% for students taking AcF 609 and 35% for students taking AcF 629.
The remaining 5% will be made up by the careers element.
Your answers/reports should be easy to read, have a logical structure and
include informative tables and figures. The clarity of presentation will be
marked, as well as the contents of your answers/reports.
All the results MUST be produced by MATLAB. Please include relevant MAT-
LAB codes/programmes as appendices in your coursework. The programmes
should have a clear structure and are appropriately commented.
Your coursework must be submitted on the 16th March 2021 12:00pm UK
time via Moodle. Late submissions will lead to a penalty deduction in line
with the PG programme guidelines.
Electronic copies of reports must be submitted via Moodle.
Good Luck!
The data can be found in the Data folder under the Coursework Assignment
section on Moodle. Each group of students will choose a high-frequency stock
trade-and-quote dataset for ONE stock only from IBM, INTC and HPQ.
Part 1: ARMA-GARCH Model
a) For your chosen stock, compute the tick-by-tick log-return and the open-
to-close log return series for each trading day in the dataset.
b) Provide a descriptive analysis for both return series computed above.
Discuss your findings and relate them to the stylized facts of financial
c) For the daily open-to-close log-return time series computed in a), find
the best model in the general ARMA-GARCH modelling framework.
You should provide detailed model selection procedures and justify your
choice. Use a sensible set of competing ARMA-GARCH specifications.
d) Perform diagnostic checks on your best ARMA-GARCH model. Discuss
the goodness-of-fit of your model.
Part 2: Realized Volatility and Forecasting
e) Construct a volatility signature plot with daily annualized realized volatil-
ity estimators based on the high-frequency data for the chosen stock.
f) Select an optimal sampling frequency and construct the daily annualized
realized volatility estimator based on the optimal sampling frequency.
Justify your choice.
g) Split the dataset into an in-sample period (03/01/2005 to 31/12/2012)
and an out-of-sample period (02/01/2013 to 31/12/2014). Compute
rolling-window, one-step-ahead (one-day-ahead) volatility forecasts with
the ARMA-GARCH model and a HAR-RV specification. To reduce com-
putational burden, estimate the ARMA-GARCH and HAR-RV model
parameters only once on the basis of the initial in-sample period.
h) For the out-of-sample period use the ex-post daily annualized RV as a
proxy of the forecasting target (true volatility) and compare the fore-
casting performances of the ARMA-GARCH and the HAR-RV models.
You should compute standard forecasting evaluation statistics, conduct
appropriate tests and provide conclusions.