Homework 3.2

Author
Affiliation

Alex Cardazzi

Old Dominion University

All materials can be found at alexcardazzi.github.io.

Completion Requirements: Complete the following questions in RStudio via the homework template. When you are ready, submit your rendered html to Canvas.

Grading Criteria: Full credit will be given to correct, well formatted, and detailed answers. Partial credit will be given if I can follow your work and/or see your thought process via code, comments, and text. Point totals are listed next to each question.

This homework focuses on a relatively hot topic in the United States: marijuana legalization. In 1996, California became the first state to legalize medical marijuana. Between then and 2010, 14 other states legalized medical marijuana. Since then, states have begun legalizing recreational marijuana, but that’s a topic for another day. One of the arguments against legalizing marijuana is that marijuana and alcohol are complements (which is testable) and traffic safety will reduce by way of increased intoxicated driving (which is also testable). In their 2013 Journal of Law and Economics paper titled Medical Marijuana Laws, Traffic Fatalities, and Alcohol Consumption, Anderson, Hansen, and Rees examine these hypotheses in a difference-in-differences framework. You should read through this paper before completing the Homework.1

Question 1

Summarize the main research question in the paper, the setting analyzed, and why DiD is appropriate for this research question. (2 Points)

Question 2

Describe the data analyzed, including general structure of the data set, sources, and the nature of the data.

  1. First, read in the fatalities data and call it fars. (1 point)

  2. Replicate the mean and standard deviation of the total fatality rate in Table 4. Note: these are weighted summary statistics. (1 point)

  3. Use Table 1 in Anderson et al. (2013) to help you create a variable to denote the treatment status for each state in each year.2 (1 point)

  4. Finally, create a variable that measures the time since the MML was passed for each state. Think carefully about how to denote states that are never treated. (1 point)

Question 3

Discuss the general assumptions underlying the causal inference method used in the empirical analysis.

  1. How do the authors address these assumptions in the paper (if at all)? (1 point)
  1. Include a discussion of threats to validity of the results in terms of causal inference. (1 point)
  1. Since this empirical setting examines a staggered rollout, what do you think about the authors empirical strategy? (1 point)

Question 4

Create a figure similar in spirit to Figures 1-3 in the paper. Instead of breaking down the fatality rate by age group as the authors do, just use the total fatality rate supplied in the data. Note: these are also weighted numbers. This is difficult, and you have to read the paper carefully. (2 points)

Question 5

Run a Goodman-Bacon decomposition.3 Interpret the output. (2 points)

Question 6

Replicate the main result reported in the paper.

  1. Recreate columns (1) and (2) of Table 6. Note that this is a simple TWFE result. Be sure to include all the bells and whistles4 sans their controls. Present the coefficient estimates in a modelsummary table. (1 point)

  2. Use the two stage difference-in-differences method in Gardner (2021)5 to re-estimate columns (1) and (2). (1 point)

  3. Estimate event study versions of columns (1) and (2). When plotting, use xlim = c(-10, 10). (1 point)

  4. Estimate event study versions of columns (1) and (2) using Gardner (2021). When plotting, use xlim = c(-10, 10). (1 point)

  5. Using Gardner (2021) and the specification in columns (1) and (2), examine the effect of MMLs on the log of per-capita VMT. (1 point)

Question 7

Discuss the results generated. What is the interpretation of the causal parameter of interest? Place the causal effect in the context of the setting.

  1. Compare the TWFE and 2SDiD results for fatalities. What do you think about the results in Anderson et al. (2013)? (1 point)
  1. Discuss the results for per-capita VMT. (1 point)

Question 8

Do you believe the results are credibly causal? Are the signs and magnitudes plausible? Why or why not? (1 Point)

Footnotes

  1. The PDF will be made available on Canvas.↩︎

  2. You might have to make some judgement calls about which year should be the first year of treatment. This is what makes research tricky!↩︎

  3. Do not print the giant output, just the small summary table version.↩︎

  4. See the table notes for the specifics.↩︎

  5. Kyle Butts’s vignette; Asjad Naqvi’s vignette↩︎