Guidelines for paper replication
Choosing a paper to replicate
- The paper must be published in a peer-reviewed journal.
- The paper must have empirical econometric or statistical content.
- You should understand the main points, and the paper should interest you.
- The data should be available and the methods accessible.
- The data must be cross sectional and not time series or panel data.
Please submit “12 Steps to Understanding a Quantitative Research Report”(attached) for the paper that you are replicating; and a copy of the tables that you plan to replicate.
General guidelines for paper replication
The document that you author should have four or five main sections, depending on what you find:
- Summary
- In this first section, summarize the paper. What was the objective of the paper? What did the authors find? If you found anything different, preview it here. Explain the research problem or question that the original paper addresses.
- Empirical strategy and data
- Explain the paper in plain English? How did the authors answer their research question? What data did they use? What variation did they exploit to answer their question? Any weaknesses in their approach that you discovered as you replicated their results?
- What is the dependent variable?
- What is the key independent variable?
- What is the proposed model?
- What are control variables that the original paper addresses?
- How big is the estimated effect?
- What is the unit of analysis?
- What are other relevant dimensions of the data? (cross section etc)
- How did you gather the data?
- Replicated results
- Every table in the paper should be replicated (unless otherwise cleared with me)
- The tables should appear as they do in the original paper
- Explain the results as you write
- Differences between your results and the authors’ results should be highlighted in the text of this section
- Compare the sign, size, and significance of the results to those in the original paper.
- Differences
- Comment on differences in your results (if any)
- Extensions
- Explain your extension of the original paper. You might consider overlooked variables, shortcomings in the estimation procedure, outliers and leverage points, or limited external validity of the model.
