ECON2206: Introductory Econometrics
Problem Set 2
Assignment instructions
• You may submit your work either:
– One hard copy of the assignment to your tutor before the end of your Week 10
tutorial class.
– Email a copy of the assignment to your tutor by 5PM on November 22nd.
• Name, student number, course title, tutorial group number and tutor’s name should
e clearly included in the submission.
• The submission should be no longer than 3 pages. Penalties will otherwise apply.
• If you hand in a hard copy do not use plastic sheets or binders, simply staple the
pages together.
• The assignment is INDIVIDUAL work. You may discuss the assignment with you
peers, but you must submit YOUR OWN answers.
• If the answer requires some mathematical calculation show the steps don’t just
eport the final results.
• If the answer requires the use of STATA show the STATA code. You don’t need to
eport the STATA output.
• This assignment has a total of 100 points awarded.
1
Problem Set 2
A book publisher is interested in what makes a book successful. In order to answer this
question the publisher provides you with a dataset of books they have published in the
last 5 years (PS2.dta). For each book this dataset contains information on the number of
thousands of books sold (sales), the length of the book in number of pages (length), if the
ook was originally written in english (english = 1 if it was originally written in english,
english = 0 otherwise) and whether the book is fiction or non-fiction (fiction = 1 if the
ook is fiction, fiction = 0 if the book is non-fiction).
In order to analyze the data you first estimate the following population regression models:
Sales = α0 + α1length + α2fiction + α3english + U, (1)
1. (13 points) Use OLS to estimate equation (1). Report α̂0 and α̂2 Assume that the
Gauss-Markov assumptions hold and interpret these parameters. (Answer in less
than 100 words)
2. (21 points) Imagine that you don’t believe that equation (1) is the co
ect pop-
ulation regression model. Instead you think that the co
ect population regression
model should take into account that the length of a book affects sales by a different
amount depending on if the book was originally written in english and if the book is
fiction. Estimate this regression using OLS and interpret all the estimates. You can
assume that all Gauss-Markov assumptions hold when interpreting the estimates
(Answer in less than 60 words for each parameter you interpret)
3. (13 points) Using the estimates from question 2. provide a prediction of how much
an english non-fiction book of 200 pages should sell.
4. (20 points) Perform a statistical test to determine whether equation (1) suffers
from heteroskedasticity. What is the result of this test (use 5% significance level)?
5. (20 points) Assume that V ar(U |length, fiction, english) = σ2 × (length/500)
where U is the e
or term of equation (1). Use this information to estimate equation
(1) using Weighted Least Squares (in STATA if you need to perform a regression of
y on x where you don’t want to include the constant you can write the following:
eg y x, nocons).
6. (13 points) The book publisher tells you that the length of a book may have been
measured with some e
or. They assure you that while the e
ors are quite frequent
they are completely at random. Do you expect that α̂1 you estimated in question
1. to be biased? In which direction you expect the bias to go? (Answer in less than
80 words)