Exam 2012 Introductory Econometrics Answers
Exam 2012 Introductory Econometrics Answers
Exam 2012 Introductory Econometrics Answers
In what follows, the brief (and sufficient for 100% mark) solutions are given in blue and
some comments given in green. Also, the essential parts are highlighted in red.
Researcher A decided to use the ordinary least squares (OLS) regression method
to estimate this model.
List the assumptions that Researcher A should admit to rely on his/her model?
(4 marks)
Note, it’s also ok (100% credit) if the above is stated informally, in words, e.g.,:
Page 1 of 6
2. After using some statistical software, Researcher A obtained the following results:
[ ]
̂ (estimated variance of )
̂ ̂ (estimated variance of )
̅ (sample mean of )
(minimal or smallest in the sample)
(maximal or largest in the sample)
The estimates suggest that, on average, $1 increase in price will cause an increase in
sales by $ per month.
.
√ √
The regression results must be interpreted with caution, in a local sense: for small
changes of explanatory variables, and within the range of observed data that was used in
the estimation), and under assumptions of the model.
b) Indicate and briefly explain how well the estimated model explains the data
(compute relevant measures from the provided data, if needed).
(4 marks)
∑ ̅ ̂
∑ ̂ ̂ ̂
So,
This means that only about 35% of the total variation in is explained by the variation in
in the model of researcher A. (A relatively poor fit.)
Page 2 of 6
3. Forecasting (prediction).
a) Using results from Researcher A, forecast (predict) sales for a store that will set the
price at $1000 and briefly interpret it.
(4 marks)
i.e., the predicted sales, (on average), of a store charging price=$ for the product is
$ per month.
(i) ̂
where
̂
(ii) √̂ ̅ ̂
(it is from Formula sheet, so if write immediately with numbers as below—also ok!)
√ ( )
This forecast can hardly be trusted, because it attempts to forecast for the price beyond
the observed range of the price in the data.
(Notes: Students might also mention other valid (although not as critical) reasons to doubt
this forecast, such as (i) low , (ii) potentially omitted explanatory variables, (iii) relatively
low sample size, (iv) too restrictive assumptions (SR1-SR6), so they deserve partial credit
in the answer.)
Page 3 of 6
Researcher B also decided to use the OLS method to estimate the model. After
using some statistical software and the same data as that used by Researcher A,
Researcher B obtained the following results:
̂ (estimated variance of )
̂ ̂ (estimated variance of )
Formally, the influence is the same for all types of the stores and is given by:
̂
(i)
(ii) i.e., the influence of price on monthly sales is diminishing or even an inverted U-shape
(i.e., at some point, increase in price could lead to decrease in sales).
Page 4 of 6
b) If an average store (i.e., with average level of price and advertising) were to
increase price for the product (YPad) by 1 dollar, other things being the same—
what would be the impact on sales? (e.g., positive or negative? Justify your
answer.)
(4 marks)
̂
̂
the average store has ̅ , so , i.e., this is where ̂
reaches maximum w.r.t. and so increase in price for such a store will cause decrease
in sales.
For markers, note, alternatively, one could arrive to the same conclusion without
derivatives, by noting:
̂ ̂
(i) On average and ceteris paribus, for stores of type (i.e., located where there are
no major competitors) an additional $1000 of advertising induces an increase of about
$2000 in sales per month.
(ii) On average and ceteris paribus, for stores of type an additional $1000 of
advertising brings increase of about $4000 in sales per month.
For markers: note that, alternatively, a student might answer this question as:
̂
(i) for
and
̂
(ii) for
Or as
̂
,
Page 5 of 6
(i) Via intercept dummy: On average and ceteris paribus, stores near competitors have
monthly sales that are about $300,000 smaller than in the other stores. It is significant (at
1%)
(ii) Via slope dummy: On average and ceteris paribus, for stores near competitors (i.e.,
) an additional $1000 of advertising brings increase of $2000 of sales per month on
the top of the $2000 effect for all stores. It is significant (at 1%)
For markers: note that, alternatively, a student might answer this question as:
Multiple choice questions – clearly indicate your answer in the writing booklet.
(please select only one answer (out of a, b, c, d, e) per question)
a. 0.978
b. the same as for the model of Researcher A but adjusted- is smaller.
c. smaller than that for the model of Researcher A but adjusted- is larger.
d. larger than that for the model of Researcher A.
e. none of the above
(2 marks)
END OF PAPER
Page 6 of 6