#3 - Long
Instructions: While cooperating on the assignment is encouraged, plagiarism is not. I
will only accept hand written answers to the interpretation part. Show your work as no
marks will be allocated for the final answer alone.
Use Stata or R to do these. If you choose to use another software, please get my
approval at least a week before the assignment is due.
Using the dataset that you prepared while doing Assignment #3 - Short (or the data
file that was uploaded to D2L).
(a) Run a pooled OLS regression of the money spent on repairs/maintenance on the
unemployment rate, the interest rate, housing starts and a trend. Comment on your
results.
(b) Should you include province-specific dummies in (a)? Why or why not? Run such
a model and use Ontario as your benchmark (please also control for trends).
(c) Should you include year specific dummies in (a)? Why or why not?
(d) Run the model in (a) with year specific fixed effects instead of trends (omit the first
year for which you have data for all variables as your baseline) and controls for provinces.
Which model do you prefer and why?
(e) Estimate a Fixed Effect model of the money spent on repairs/maintenance on the
unemployment rate, the interest rate, housing starts and a trend. Compare your model to
the results you have in (b). To make the models comparable, you may need to estimate the
model in (b) without a constant but allowing for a binary variable controlling for Ontario.
(f) Estimate a first differenced model of the money spent on repairs/maintenance on
the unemployment rate, the interest rate, housing starts. Should you include a trend?
Why or why not? Comment on your results.
(g) Estimate a Random Effect model of the money spent on repairs/maintenance on
the unemployment rate, the interest rate, housing starts and a trend. Comment on your
results.
(h) Which model do you think is better between (e) and (f)? Why?
(i) Which model do you think is better between (e) and (g)? Why?
(j) A big part of being an economist is being able to tell when a model makes sense and
when it doesn’t. I have asked you to run a number of models in this assignment; which
ones do you think worked better than others and why? In light of your results to questions
(a) to (k) what is the optimal model in your opinion? Your answer can be a combination
of some of the things you did above. If it is, please run your “optimal” model. Justify
each of your decisions.
(k) What other regressors do you think would have been helpful to include here? Do
you have other suggestions on how to make the model better?