THE UNIVERSITY OF SYDNEY

FACULTY OF ARTS AND SOCIAL SCIENCES

SCHOOL OF ECONOMICS

ECMT6003: Applied Business Forecasting

Take-home assignment

Due on 18 November 2019, by 9 AM

Instructions:

• This assignment consists of two questions. Each question is worth 15 marks, so the entire assign-

ment is worth 30 marks. The marks obtained in this assignment will be added to those obtained in

the mid-semester examination (maximum 20 marks) and those obtained in the final examination

(maximum 50 marks) to determine the final mark for this unit.

• This paper consists of one front page and one page with questions. Hence, the total number of

pages is two.

• This assignment must be completed in groups of two or three students.

• Please hand in your answers no later than 9 AM on Monday 18 November 2019, using the tool on

the Canvas page for this unit. Your submission should be in the form of a zip file containing your

report, your data for Question 2, and your computer code. One submission per group is sufficient;

just make sure it contains the SIDs of all students in the group.

• The maximum number of words is three thousand, not counting any tables, graph descriptions, or

computer code. Please note that the word “maximum” implies that 1200 words is fine too.

Good luck!

Question 1. Refer to the data set that I have posted on Canvas, containing quarterly data on several

Australian macroeconomic and financial variables, as well as the prices of some commodities that are

relevant to the Australian economy and the exchange rates between the Australian dollar and the curren-

cies of its major trading partners. All series start in 1999 and end in the most recent quarter for which

data are available at the time of writing. This is all raw data; no detrending or deseasonalisation or similar

trickery has been performed.

The goal in this half of the assignment is to come up with forecasts for the unemployment rate in

Australia. More specifically, I am interested in your forecasts (point predictions and 95% intervals) for

the unemployment rate (a) at the end of the first quarter for which no unemployment data are known yet,

Q3 of 2019; (b) at the end of this year, Q4 of 2019; and (c) at the end of next year, Q4 of 2020.

Clearly, this is a very open-ended question, with no obvious right or wrong answers – except for

part (a), I suppose. You may wish to include all, some, or none of the other variables as predictors, and

many different types of models are possible. For this reason, much of my interest is in which forecasting

techniques you choose to try, and why you end up working with the technique that you pick. Thus, make

sure to motivate your decisions carefully!

Feel free to use whatever software you are most familiar with – Excel, Stata, EViews, Matlab, R,

Python; anything goes, as long as it is capable of estimating the models that you need. Please include

your code when you hand in your answers, as described on the front page of this assignment.

Question 2. This is the BYO version of Question 1. Obtain a data set that is related to economics or

business, contains at least thirty observations, and leaves something to forecast. Then, forecast it! If you

have an interesting forecasting problem at your workplace (and you’re allowed to share the data), it’s

fine to use that for this assignment. If in doubt, contact me on peter.exterkate@sydney.edu.au, but almost

anything will be accepted. As in Question 1, I am mostly interested in your choice of forecasting model,

and in the (statistical or economical) justifications you provide for this choice.

In addition to your code, please also include the data in your submission.