Your performance on this course will be evaluated via a 3000 word research paper. This paper will form 100% of your grade.
The research paper should follow the basic elements of a novel research project, in that it should address a specific research question, briefly identify the theoretical contribution, provide testable hypotheses, and implement a suitable design based on one of the methods that we study in the course.
The goal here is to put one of the methodological approaches we study on this course in to practice. Your paper must implement at least one of the following: experiments, matching, regression, regression discontinuity, difference-in-differences/fixed effects, synthetic control, or instrumental variables. Which approach you choose should be guided by a) your research question, and b) the type of data that is available to you. For instance, it is unrealistic to think that you will be able to design, pay for, run and analyse a field experiment over the next ten weeks.
Your paper can address any issue of your choice from political science or related disciplines, provided you are asking a clear causal question.
- Introduction/Research question statement
This section of the paper should be used to introduce the main topic that you will be addressing, and the central research question that you will be aiming to answer. It will be helpful to be very precise here as to the causal relationship of interest (e.g. “In this research paper I assess the causal impact of D on Y”). You should also very briefly locate your study in the existing literature.
- Description of methodology
In this section of the paper you should explain the methodological approach you are taking, what assumptions it relies on, how it helps to overcome difficulties in making causal inferences in your setting, and so on.
- Description of data
You need to clearly and concisely describe the data you will use to implement your design. You should say where the data comes from. This section is also the place to to discuss the scope of the study: are you focusing on a particular country/set of countries? What is the scope of the analysis? You may also wish to include some descriptive statistics of the key variables in your data.
Results should be presented in well-formatted tables and figures. Do not include any raw R output (marks will be deducted). Remember that in addition to presenting the main substantive results of the analysis, you may wish to include empirical evidence that pertains to the identification assumptions that lie behind the design, though some of these might be better placed in the appendix to your paper.
What have we learned from your paper? What are the limitations of the analysis?
- Your paper should not exceed 3000 words. This includes footnotes, but does not include the bibliography, tables, the title page (including abstract), or any appendices.
Rcode used in the empirical parts of your assignment should be included in the appendix. This does not count towards your word count. Note that your
Rscript file should be neatly presented and easy to follow, and not include everything you have tried out, but only what is included in your final paper. Essentially, it needs to be possible for us to fully understand the analyses you have done and possibly recreate all of the main tables/plots in your paper. You should not include any
Rcode in the main body of your paper.
- Plagiarism will be taken very seriously. See this page for details on UCL’s plagiarism policy.
The research paper will be due on April 24th, 2023 at 2pm. Papers will be submitted online via Turnitin.
We encourage you to book in to see one of us during office hours to discuss your project before the end of term.
The following are some example papers that have been produced on this course over the past few years.
For your project, you will need to assemble a dataset to which you can apply one of the methods that we cover during the term. We say assemble because we do not expect you to spend many days and weeks painstakingly collecting a huge amount of new data. Instead, you should think whether you can use a dataset that already exists, or combine existing datasets together in order to be able to apply your chosen method.
When assembling your dataset, you should ask yourself some key questions:
What is the unit of analysis that I am going to be working with?
- The units are the rows of your data, they are the objects that you are studying. For example, you might be studying individuals, or countries, or companies, or sub-national regions within a certain country, etc. Until you know what the units are, it is very hard to move forward and collect information about those units.
What is the outcome variable of my analysis?
- The outcome is the variable that you want to explain, or the variable that you expect to be causally related to your treatment.
- For example, if your research question is “What is the effect of unemployment on support for right-wing parties in UK constituencies?”, then your outcome variable will be the level of support for right-wing parties in UK constituencies (where constituencies are the unit of analysis)
What is the treatment variable in my analysis?
- The treatment variable is the causal factor of interest. It is the variable that you believe to be causally related to your outcome.
- For example, in the example above, the treatment variable would be the level of unemployment in each UK constituency
Do I need data on these variables over time?
- Many of the methods we study on this course require data for more than one time period. For instance, the difference-in-differences design requires over-time data, and so does the synthetic control method. By contrast, selection-on-observables strategies like matching and regression can be applied only using a single cross-section of data. Whether you need to collect data from multiple time points will depend on which of the methods you intend to apply.
A good new resource for finding datasets is the Google Dataset Search website. Here you can enter a search term for the type of data you are looking for and Google will return to you some possibilities. This may be a good way to start looking for data.
Below are links to some commonly used datasets in political science. Please note that this is not an exhaustive list and that you may need to look elsewhere to find relevant data for your project. Nevertheless, it may be useful for you to use these as resources to help with your projects.