UNIVERSITY OF ESSEX EC963 PUBLIC POLICY EVALUATION SPRING ASSIGNMENT NOTE: The deadline for handing in this assignment is April 24th 2024, 12:00 noon (online via FASer). It is acceptable to copy/paste output from R. However, clarity of presentation will be taken into account and to get full marks you are required to comment appropriately on your output. You MUST provide an R script that executes all relevant analyses. This script must be clear. You may include comments within your script if relevant. Please submit your answers in .pdf format including an appendix with your script. The R script is worth 15 marks. ALL QUESTIONS MUST BE ADDRESSED Disclaimer: The dataset provided comes from the official website of the French Ministry of Health. 1 On December 15th 2021, the French government decided to lift the lockdown that was introduced on October 30th to reduce the prevalence of the covid-19 and replaced it with a curfew. Starting on December 15th 2021, it was decided that a curfew would be imposed, prohibiting individuals to go out without ’reasonable excuses' from 8 pm to 6 am. Despite this curfew, the incidence rates of the covid-19 continued to rise and became exceed- ingly high in some regions of France. The government decided to introduce a "reinforced curfew" that would last from 6 pm to 6 am in the following manner: • On January 2nd, the "reinforced curfew" was introduced in some regions while the "standard curfew" kept being applied in other regions. On January 10th and January 12th, the "reinforced curfew" was extended to another set of regions of France. On January 16th, the "reinforced curfew" was extended to the entire territory. Your task: Your task is to evaluate the impact of the "reinforced curfew", relatively to the "standard curfew", on the incidence rates of covid-19. Two data sets are provided for this assignment. data_covid_daily.dta provides information at the daily level. data_covid_weekly.dta provides information at the weekly level. Information on the variables contained in the two data sets is provided at the end of the assignment (p.6). Important: when assessing the impact of the curfew, remember that there is a lag between the date of contamination and the date of a positive test. You can consider that this lag is on average 5 days. Terminology and variables of interest: • The "standard curfew" designates the curfew that lasts from 8 pm to 6 am and the "reinforced curfew" designates the curfew that lasts from 6 pm to 6 am. You need to construct the related variables based on the identifiers of the Départements (regional subdivision of France) contained in the data set. To do that, use the information provided on page 5. • The incidence rate designates the number of positive covid cases divided by the pop- ulation size. You need to compute it based on the information contained in the data set. IMPORTANT: you need to use R for questions 1, 2 and 5. Provide the R script at the end of the document containing your answers. You can copy and paste, or use screen shots as long as it is readable. This R script is worth [15 MARKS] 1. [5 MARKS] You hear in the news that some individuals are assessing the impact of the "reinforced curfew" by comparing the incidence rates between places with the standard curfew and places with the reinforced curfew on January 15th. 2 Implement this approach using the data data_covid_daily.dta by comparing the average incidence rates between places with the "standard curfew" and places with the "reinforced curfew" on January 15th. You can use a linear regression or a t-test to compute the difference in means between the 4 groups (standard curfew, reinforced curfew on January 2nd, reinforced curfew on January 10th and reinforced curfew on January 12th). What is the conclusion of this approach regarding the impact of the reinforced curfew? Do you think that this approach is right? Why? 2. [10 MARKS] You hear in the news that some individuals are assessing the impact of the "reinforced curfew" by comparing the evolution of the incidence rates after the curfew to those before the curfew in places where the "reinforced curfew" was introduced. Implement this approach using data_covid_daily.dta by comparing the av- erage incidence rate between January 15th and December 15th among the group of Départements where the "reinforced curfew" was introduced on January 2nd. You can use a linear regression or a t-test to compute the difference in means. What is the conclusion of this approach regarding the impact of the reinforced curfew? Do you think that this approach is right? Why? 3. [5 MARKS] Since you have taken a public policy evaluation course, you think that the setting would be appropriate for a difference-in-differences. Recall what this method is and what the identifying assumption is. 4. [10 MARKS] You decide to use a difference-in-differences strategy. You want to compare the evolution of the incidence rate between two groups: Départements where the reinforced curfew was introduced on January 2nd and Départements where the reinforced curfew was introduced on January 16th. What are the treatment and control groups? Is the treatment binary? Write down the econometric specification you will estimate and the main hypotheses on the coefficients of interest. 5. Perform the following tasks: • [5 MARKS] Using the data data_covid_daily.dta, plot the evolution of the incidence rate by day from December 1st to February 12th for the two groups of interest (as described in question 4). • [5 MARKS] Using the data data_covid_weekly.dta, plot the evolution of the incidence rate by week for the two groups of interest (as described in question 4). ⚫ [20 MARKS] Using the data data_covid_weekly.dta, estimate the difference- in-differences model at the weekly level comparing the two groups of interest (as described in question 4). Does the identifying assumption hold? What is the conclusion of this approach regarding the impact of the reinforced curfew? Compare your findings with the results of questions 1 and 2. 3 6. [10 MARKS] Given that the "reinforced curfew" is applied at the Département level (a regional subdivision of France), one of your classmates tells you that it could be possible to evaluate the impact of the policy with a regression discontinuity design. The idea would be to use the distance to the border between two Départements as running variable, then compare the contamination rates of two villages close to the border. Recall what a regression discontinuity design is and its two identifying assumptions. Do you think that these two assumptions are likely to hold in this case? Why? 7. [5 marks] Using the insight from the answer to question 6, what elements could threat the internal validity of the difference-in-differences strategy? Provide an an- swer to this question using the notion of selection bias. 8. [10 MARKS] After answering all the previous questions, you realize that it is difficult to precisely estimate the impact of the "reinforced curfew". You decide to write to the French government describing the ideal setting in which you could test the impact of the "reinforced curfew". Describe this ideal setting, name the method you would use, the identifying assump- tion and how you would estimate the impact of the "reinforced curfew". REMINDER: do not forget to include the R script at the end of the document con- taining your answers. This script is worth [15 MARKS]. + Details on the Curfew The "standard curfew" lasted from 8pm to 6 am. The "standard curfew" was introduced on December 15th. The "reinforced curfew" lasted from 6pm to 6 am. The introduction of the reinforced curfew was staggered. It started on January 2nd in some regions, was expanded on January 10th and January 12th to another set of regions and finally applied to entire territory starting on January 16th. The "reinforced curfew" was introduced at the level of Départements. It corresponds to a regional subdivision of France. France is divided in 101 Départements.. In the data sets, the information is provided for 96 Départements. The number in parentheses below designate the identifier of the Département targeted by the reinforced curfew. You need to use these numbers to construct the variables of interest. • Départements where the reinforced curfew was introduced on January 2nd: Hautes- Alpes (05), Alpes-Maritimes (06), Ardennes (08), Doubs (25), Jura, (39), Marne (51), Haute-Marne (52), Meurthe-et-Moselle (54), Meuse (55), Moselle (57), Nièvre (58), Haute-Saône (70), Saône-et-Loire (71), Vosges (88) and Territoire de Belfort (90). • Départements where the reinforced curfew was introduced on January 10th: Allier (03), Alpes de Haute-Provence (04), Bouches-du-Rhône (13), Cher (18), Côte d'Or (21), Bas-Rhin (67), Haut-Rhin (68), Vaucluse (84). • Départements where the reinforced curfew was introduced on January 12th: Drôme (26) and Var (83). ● On January 16th, the reinforced curfew was applied to the rest of the Départements. 5