EC639/EC410C: Replication Project
This will be the major evaluation item in EC410C/EC639. The project will consist of two graded components: a paper (75% of the grade) and a presentation (25% of the grade). The project will be completed in teams of two. If you want to work alone, you can ask me and in I can consider it, although I would strongly encourage students to work together.
Copyright By PowCoder代写 加微信 powcoder
Each team must select a published economics paper to do their project on. The list of acceptable papers is already available in the course syllabus, available on MyLS. These papers have already been checked by me, and have all data and code available somewhere online. There are more papers than there are teams (assuming all teams of two), so there should be enough choice for all teams. Papers will be selected in the following way:
1. After teams are selected (by yourselves), I will generate a random ranked ordering of all the teams.
2. The ranking of teams will determine the order that teams select their paper for their project.
3. The selection will be done in-lecture (after our econometrics review session) to ensure transparency and fairness. To facilitate the selection process, please bring a list of several papers that you would like to do as a team, keeping in mind the order that you are selecting in. Each team will have a maximum of 2 minutes to select their paper after the previous team is done.
The focus of the project will be on the replication, which I will describe below immediately. I will also break down expectations of the paper and presentation below. I will also separately upload the grading schemes (breakdowns) for the paper and presentation to MyLS.
Replication
Replication is the process of trying to reproduce another paper’s empirical results. The most basic form of replication, which you will be doing, is to use the same data and code as the paper, to see if the results can be reproduced. Even in this simplest form, lots of things can potentially make a replication unsuccessful:
· The data provided are incorrect or incomplete
· The code provided are wrong or incomplete
· The results generated do not match up with the results published in the paper
As mentioned in the first day of class, each team will select their paper, based on the random ordering of teams that I will generate after students submit their teams to me. Papers will then be selected by each team based on this order, likely at some point in class that I will announce later. The papers to be selected come from the list of replication papers I have included in the syllabus.
You will determine what the main table of results is, in your paper. If there is more than one main table of results, you will choose the one that seems best to you. You must inform me of which table you have chosen as your main table by 5pm on March 17 (email or office hours, only one person per team necessary). Your first task in the replication project will be to use your data and code to reproduce the table chosen as best as you can; this is a so-called baseline replication.
In addition to a baseline replication that tests whether the data and code can actually reproduce the table’s results, replications will also test the “robustness” of a set of results by estimating different regression specifications (i.e. does taking logs matter?) or making sensible sample restrictions or expansions of samples (i.e. do the results rely on the inclusion or exclusion of particular observations or sets of observations, like certain industries, provinces, types of workers, etc.?). Other things people have done is include extra control variables, which can require getting extra data from elsewhere and merging it into the original data; adding more data is obviously more ambitious and few (if any) students are likely to attempt this but its success will be rewarded accordingly in the replication component of your project grade if the exercise is meaningful and a good extension of the original table. You can also consider alternative forms of the main outcome or treatment variables. This is not a complete and exhaustive list of things you can try, but suggestions to get you started.
For your projects: you will need to:
1) first conduct a baseline replication exercise, attempting to reproduce your chosen table’s estimates or values using the provided data and code.
2) You must then also extend the analysis by testing the robustness of the table’s results in some way by doing additional analyses, as listed in the previous paragraph or in some other way that you can think of.
3) Finally, you must also suggest one final robustness/replication check that extends the analysis from the baseline table’s methods and data that is NOT what you did in 2). Be ambitious in this suggestion, since you are not constrained by whether you yourself can do this exercise. Rather, think of something that would be useful in testing the robustness or replicability of the table chosen.
I should note that it is not important to me if your baseline replication is successful or not; in fact, unsuccessful replications can sometimes be even more informative and helpful to further research!
Finally, for your replication, you should describe your replication exercises in both the paper and presentation, but importantly document the entire replication analysis in a log file which you should include in the paper, in the Appendix.
The paper should be at most 9 pages, double spaced (not including references, tables, figures, and Appendices). This is a limit, not a page suggestion; if you feel your paper could be shorter but still accomplishes all the needed tasks, feel free to submit a paper under that length. The paper should be uploaded into the dropbox on MyLS as a Word document or PDF. Your paper must consist of the following components:
1. A summary of the paper you have chosen. You must state the research question, the contribution of the paper (relating it to the relevant literature), and its main findings. You should also make clear what the policy implications (or implication to economists) are, given the findings of the paper.
2. A description of the methods used. This includes a description of the main data used (particularly in the replication results you will do). You should clearly describe the empirical methodology the paper uses, with a particular focus on the methods used in the results you are replicating.
3. A description and discussion of the replication you have conducted. Present the table of results that you have replicated, as well as the original table of results you are replicating (if the results differ). For your extra analysis beyond just reproducing the results (i.e. you tried other specifications, or added more controls), include them in the same table (or another table) and explain what you have done, how you did it, why you did it, and what you concluded from these additional exercises. Based on your replication, conclude whether or not the original paper’s findings are correct. Explain also another replication/robustness exercise that you think would be useful here, how you would conduct it (if you had the data and time), and why this is a useful additional exercise.
4. A critical (i.e. your thoughts) analysis of the paper, highlighting what you consider to be the strengths and weaknesses of the paper you have chosen. For example, is the research design convincing in arguing for a causal relationship?
5. An Appendix, with the log file of you running the replication analysis, clearly showing the regressions that produce the table of the main results (successful or not).
Presentation
Presentations will occur in the last two weeks of the course, tentatively (depending on progress on the material and number of students presenting, etc). The presentations will be 15 minutes each, per team. Questions can and will be asked by me, or by other students, in the class. Teams’ responses to questions will form part of my grade to each team. Each team member must present a non-trivial amount of the presentation; you can’t just read the title page, for example. The presentation should mirror the format of the paper, including the following things:
1. A brief summary of the paper: its question, its contribution, and what they find.
2. A discussion of its methods and data, with a focus on the data and methods that will be used in the results that you are replicating.
3. A discussion of the replication. Present the results in a table, and discuss whether you were successful or not. What have you learned about this paper, and generally?
4. Finally, a brief conclusion, summarizing your thoughts on the paper, and what you thought was good and bad about it (strengths and weaknesses of the paper).
I will ask groups to upload the final versions of their slides to a dropbox on MyLS before their presentation times, so I have them and to commit you to finishing the slides ahead of time.
Why are we doing this?
A fair question! I believe that empirical research is a lot harder than many students realize. In econometrics courses, we often provide students with a lot of the code, or the assignments involve very specific tasks. Research is different: you might need lots of different tools and data combined to solve some problem or accomplish some task that might seem simple at first. This project will be a good first attempt for students to attempt to do some research using code and data that SHOULD already work. This will get you more practice in actually using real data and code to do empirical analysis.
In studying how some of these well-published studies actually did their empirical analysis, this will teach students a lot about how researchers do their work. As a student, I had taken a lot of econometrics, but had not actually done very much actual work at a high level, and certainly none that had contributed to our knowledge of economics. My first replication project taught me so much about how research is actually conducted, from the code used, best practices for documenting work in do-files, combining multiple data sets, etc. These lessons proved very helpful for me once I started doing research and data analysis, and hopefully you will find this similarly helpful.
程序代写 CS代考 加微信: powcoder QQ: 1823890830 Email: powcoder@163.com