Writing Papers: A Checklist
Michael Kremer
The following guidelines are not ironclad; there will sometimes be reasons to break them. However, you should check over them in preparing your paper, and make sure you have a good reason if you are not following them.
Empirical Work
1. Describe the data set. Where is it from? How many observations?
2. Include a table showing the means of the variables, and the standard deviations.
3. As a rule, it is usually better to show standard errors, rather than t-statistics.
4. Look at a journal to see how tables are laid out. Avoid horizontal lines in tables. 5. Include basic information like sample size and R* in all tables.
6. To the extent possible, make tables self contained with clearly-explained titles and
extensive notes.
7. Explain clearly where the identification comes from. As a disciplinary device, try to
include an explanation that would be intelligible to a non-economist.
8. Did you drop any data? What rules did you use in deciding what data to drop?
9. In general, you should explain your empirical procedures in enough detail that someone could replicate your work.
10. If you use instruments, explain why you think they might be valid.
Modeling
Think about the purpose of your model. If you are trying to prove the existence of an effect, it is generally good to start with the simplest, most stripped-down model that will deliver your result. Later, you can do extensions to examine how general the result is. There is no need to strive for realism in early drafts.
On the other hand, if you are building a theoretical model for empirical implementation, it may be more important to have realistic assumptions, and not to abstract too much from other factors. Similarly, if you are trying to prove generality, you may want to make minimal assumptions about functional form.
For an “existence of an effect” type of paper, it is useful to think about the following questions:
1. Would it be easier to derive the results in continuous or discrete time?
2.Can I demonstrate the effect in two or three time periods, or do I need to do an infinite- horizon model?
3.Are the functional forms for the utility and production functions the ones that make the derivations as simple as possible?
4.If the model involves agents of different types, how many different types are actually needed to generate the result?
2
In general, it is a good idea to explicitly write out the utility function, the production function, and if applicable, the timeline of moves, the solution concept (competitive equilibrium, Markov Perfect Equilibrium etc.), and the assumption about who has what information when.
Exposition
1. The introduction and literature review should generally take no more than five double- spaced pages. After that, you should get into the model.
2. Always include an abstract.
3. In the abstract and introduction, don’t just say what issues you have examined, but describe your results.
4.Papers should very rarely be more than forty-five pages long. Twenty to thirty is usually appropriate.
3