Question 1 - Approximately 600 words [7 marks]

This question does not require you to undertake any analysis yourself. Rather, you are     being asked to critique an existing paper on “The impact of COVID- 19 on racial inequality in business earnings” by Fairlie (2022). The paper is available at the following link:

https://www.nber.org/system/files/working_papers/w30532/w30532.pdf

You can download a PDF version of the paper directly, or if you are off-campus you may

need to enter your ANU e-mail address on that page and it will be emailed to you (see

https://www.nber.org/papers/w30532).

Answer the following questions in your own words.

1a) What are the research questions and what data is used to test those questions? Include a discussion of the strengths and weaknesses of the data used in this paper. [2 marks]

1b) What are the main data analysis techniques used to test those research questions (from part a)? Describe at least one of those techniques in detail. [3 marks]

1c) How could the analysis be extended using a different model or technique? [2 marks]

Question 2 [6 marks]

These questions are based on the content from our guest lectures in Weeks 9 and 11. Choose one best answer for each question.

2.1 The bias-variance trade off does not apply in machine learning problems when –

a-   The machine learning models used are complex

b-  Data has missing values

c-  Linear regressions are used

d-  None of the above

2.2 Cross-validation can be used to compare different machine learning models.

a-   True

b-  False

2.3 One advantage of using a decision tree over linear regression model is that a decision tree model will always produce more accurate predictions than a linear regression model.

a-   True

b-  False

2.4 Why was the expansion of the Medicaid program in Oregon so exciting to economists?

a-   It provided affordable health care to all vulnerable individuals in Oregon.

b-  It provided a large sample size so that economists could use quantile regressions to

address endogeneity.

c-   It enabled causal inference because it was a natural experiment based on lottery

drawings.

d-   It was a large-scale program that successfully reduced emergency department visits in

Oregon.

2.5 Difference-in-difference methods are a special case of random-effect models.

a-   True

b-  False

2.6 RCTs cannot be used to establish causal inference.

a-   True

b-  False

Question 3 – Approximately 600 words [7 marks]

This question requires you to undertake your own analysis. For one of the datasets used in the computer labs this year:

3a) Identify a dependent variable that could be estimated by a binary, ordered or multinomial logit model. Which of these models would you recommend for estimating this variable?        Justify why you would choose this model over the other. [2 marks]