ECON 474 Final Project Guidelines FA21
Preliminaries
-
Due 05/14/2022, 11:59 pm (submission on Moodle).
-
Form groups and choose one paper to replicate before April 29th! Put the name of your groupmates and the selected paper in the spreadsheet to secure your first choice. No more than two groups per paper allowed.
-
Here you find a template for the Final Project.
General requirements
You must read and summarize the paper you choose. There is no page limit, but extra points will be given for concise writing with correctly displayed plots and tables. Each summary should be divided as follows:
- Question
What is the causal link the paper is trying to reveal? What would be the ideal experiment to uncover that relationship? Still on the “ideal experiment,” why would it be so difficult to implement?
- Methodology
What is the identification strategy? Which assumptions are made? What are potential threats to the identification? Also, briefly explain the dataset.
- Results
Replicate the main figures and tables (check the specific requirements for each paper below) and comment on the results.
- Conclusions and Limitations
What conclusions does the author reach? How robust are the results? What about external/internal validity?
Specific requirements
To find the published paper, go to https://www.library.illinois.edu/ and search for the article’s title. Follow the specific requirements in the section Results of your report, depending on the paper you choose.
Regression Discontinuity Design (+ up to 2% EC)
Replicate the paper Andrew C. Eggers and Jens Hainmueller “MPs for sale? Returns to office in postwar British politics.” American Political Science Review, 2009.
Part of their dataset can be found here. The table below describes the data:
| Variable | Definition |
|---|---|
| surname | surname of the candidate |
| firstname | first name of the candidate |
| party | party of the candidate (labour or tory) |
| ln.gross | log gross wealth at the time of death |
| ln.net | log net wealth at the time of death (main outcome) |
| yob | year of birth of the candidate |
| yod | year of death of the candidate |
| margin.pre | margin of the candidate’s party in the previous election |
| region | electoral region |
| margin | margin of victory/vote share (running variable) |
Reproduce Table 4 (columns 1 and 3, page 524) and Figure 4 (page 525) for both Tory and Labour candidates. Also, run placebo studies for the variables age (yod-yob) and margin.pre.
Instead of using local linear regression, use the linear and quadratic specifications we discussed in class. Limit the margin of victory to [-0.075, 0.075] and rerun the linear and quadratic specifications as a robustness check.
Fixed Effects (+ up to 1.5% EC)
Replicate the paper Marcus Casey, Jeffrey Schiman, and Maciej Wachala “Local Violence, Academic Performance, and School Accountability” AEA Papers and Proceedings, 2018.
The dataset can be found here. The table below describes the data:
| Variable | Definition |
|---|---|
| AYP_ind | Indicator if school is meeting Adequate Yearly Progress |
| Math_MEO | Percentage of students in the performance category of “meeting” or “exceeding” on the ISAT math portion |
| Math_WO | Percentage of students in the performance category of “ warning” on the ISAT math portion |
| Property_Crime_After_dist_d | Count of property crimes within 0.1, 0.2, 0.3, and 0.5 mile of a school that happened within two weeks after the testing window |
| Property_Crime_Before_dist_d | Count of property crimes within 0.1, 0.2, 0.3, and 0.5 miles of a school that happened within two weeks before the testing window |
| Property_Crime_dist_d | Count of property crimes within 0.1, 0.2, 0.3, and 0.5 mile of a school that happened during the testing window |
| Violent_Crime_After_dist_d | Count of violent crimes within 0.1, 0.2, 0.3, and 0.5 mile of a school that happened within two weeks after the testing window |
| Violent_Crime_Before_dist_d | Count of violent crimes within 0.1, 0.2, 0.3, and 0.5 mile of a school that happened within two weeks before the testing window |
| Violent_Crime_dist_d | Count of violent crimes within 0.1, 0.2, 0.3, and 0.5 mile of a school that happened during the testing window |
| violent_pp_avg_R | Average per capita crime rate for the tract that a school is located in. |
| avg_hh_size | Average number of individuals in a household |
| med_age_pop | Median age within the population |
| med_inc_hh | Median household income |
| prop_below_pov | Proportion of families below poverty level |
| prop_hisp | Proportion of individuals that identify as Hispanic |
| prop_white | Proportion of individuals that identify as white |
| unemp | Unemployment Rate |
| year | Year |
| schoolid | School identifying number |
| commid | Community area identifying number |
Reproduce tables 1 and 2 running fixed effects regressions with fixest and clustering standard errors at schoolid. Interpret the results.
Synthetic Control Method (+ up to 1% EC)
Replicate the paper Alberto Abadie, Alexis Diamond, and Jens Hainmueller “Synthetic control methods for comparative case studies: Estimating the effect of California’s tobacco control program.” Journal of the American Statistical Association, 2010.
The dataset can be found here. The table below describes the data:
| Variable | Definition |
|---|---|
| state | State name |
| id | State number |
| year | Year |
| lnincome | Per-capita state personal income (logged) |
| beer | Per-capita beer consumption |
| age15to24 | State population and percent of state population aged 15–24 |
| retprice | Average retail price per pack of cigarettes (in cents) |
| cigsale | Per-capita cigarette consumption (in packs) |
Reproduce Table 1 (page 499), Table 2 (page 500), Figure 2 (page 500), and Figure 3 (page 501). Also, using the package SCtools, conduct placebo studies reassigning the treatment to all comparison units. Plot the Post period RMSPE/Pre period RMSPE ratio and interpret the results. Finally, show the “exact p-value” of this exercise.
Difference-in-Differences (+ up to 0.5% EC)
Replicate the paper Rafael Di Tella and Ernesto Schargrodsky “Do Police Reduce Crime? Estimates Using the Allocation of Police Forces After a Terrorist Attack.” The American Economic Review, 2004.
The dataset can be found here. The table below describes the data:
| Variable | Definition |
|---|---|
| observ | Block id |
| barrio | Neighborhood |
| calle | Street |
| altura | Street Numbering |
| institu1 | 1 if there is a Jewish institution in the block, 0 otherwise |
| institu3 | 1 if there is a Jewish institution one block away, 0 otherwise |
| distanci | Distance to closest Jewish institution (in blocks) |
| edpub | 1 if there is a public building/embassy, 0 otherwise |
| estserv | 1 if there is a gas station, 0 otherwise |
| banco | 1 if there is a bank, 0 otherwise |
| totrob | Car Theft |
| mes | Month |
Replicate figure 2. However, instead of using the weekly evoulution of car thefts, group the data by month and treatment status and plot the evolution of average car theft in the treated (institu1==1) and the control (institu1==0) city blocks from April to December. What can you say about the common trends assumption in this setting? Also, reproduce column (A) of table 3. Considering a roughly approximated increase in police presence of 223%, what is the crime-police elasticity in their setting?