Introduction (3-4 pages)
Although you might deviate from the following outline, it is worth using as a starting point for virtually any empirical paper:
Paragraph 1 – Motivate the question
Paragraph 2 – State the question
Paragraph 3 – Explain how you answer the question; depending on the complexity of the analysis, you may need one paragraph to sketch our your theoretical approach and an- other paragraph to describe your empirical analysis, or one paragraph to lay out identifi- cation/estimation issues and another to explain exactly what you do
Paragraph 4 – Overview your findings
Paragraph 5 – Put your paper in context with the previous literature pointing out the main contributions of your paper in the literature
Paragraph 6 – Relate your findings to important policy issues (if possible)
Literature Review (2-3 pages)
This section will help you put your paper in the context of the literature. Choose papers that are closely related to your topic to briefly present what they have done. Each reference to the literature should logically flow to the next reference. The purpose is to illustrate that even though similar topics have been examined before there are some issues that they omit or could not examine, and so you will be addressing these issues. Therefore, the last paragraph in this section should give the reader the place of your paper in the literature and your main contributions.
Theory/Empirical Specification (4-8 pages)
Sometimes you have a separate model section and econometrics section, and sometimes you com- bine the theory and econometrics into a single section.

?Even if your analysis consists of running a regression (be it a log-linear earnings model, a discrete choice health or schooling model, etc.) you should sketch out—or at least address—the underlying decision-making. You want to make clear (i) what relationship(s) you intend to identify; (ii) why those relationships are expected to exist in the data, and what they might represent; (iii) why it is (or is not) difficult to identify the relationship(s) of interest.
Your econometrics discussion should clarify how you identify the relationships discussed in the theory section.
The theory section should use economic theory combined with a discussion of the literature to clarify the question posed in the introduction. This is where you make your statement of the ques- tion more precise. After reading this section, readers should understand exactly what you want to identify in the data, and what you want to learn.
The econometrics section should make clear what you intend to estimate. Whether you analyze one outcome or multiple outcomes, whether you estimate a single-equation model or a multi-equation model, whether you estimate one specification of the model or multiple specification, everything should be described and justified here.
Once the reader gets to the “Findings” section, there should be no surprises regarding what you estimate, how you estimate it, or why you estimate it.
Data and Descriptive Statistics (4-6 pages)
In this section you explain how the econometric model you presented before can be estimated empirically. That is, from where you get the information to apply the estimation strategy and how you construct different variables. The choice of the variables should directly come from the model you described in the previous section and be motivated by the literature review section.
Subsection A – Name the data sources and give salient characteristics. As necessary, describe how you chose your data source(s).
Subsection B – Explain how you selected the sample(s) for your analysis. Clearly state your selection criteria, and give sample sizes.
Subsection C – Discuss each covariate used in your analysis. Distinguish between dependent and independent variables. This is the place to give basic summary statistics (i.e., a table of means and standard deviations for each sample analyzed).
The descriptive analysis could be an independent section or part of your data section depending on the extent of your analysis. In some papers, all you need is a table of means and standard deviations and you are ready to present estimates from a parametric model. In other papers, a descriptive analysis is an important part of the paper, insofar as you need to establish what types of variation exists in the data, or what types of patterns are seen in the data. Often, this is the nonparametric analog to what you then explore more formally via a parametric model.
Bear in mind that (a) every table included in the paper must be discussed or at least alluded to; (b) everything discussed in the text must further the argument you are trying to make.
?Findings/Results (6-8 pages)
This section serves one purpose: to answer the question. All you have to do is present the estimates, interpret the estimates, and present more estimates to convince readers that your findings are “real.”
This section should not (a) state hypotheses, or give theoretical justification for your regressions (except to reiterate and flesh out what was discussed in the theory section); or (b) explain how you specify/estimate your regression models (except to remind readers what was explained in the econometrics/empirical specification section). There should be no surprises whatsoever. This is not the place for readers to discover what the question is, what the theoretical explanation is, what outcome is of interest, what sample you use, etc.
Conclusion (2 pages)
By convention, we typically provide a brief review of the question, the methods, and the findings. More importantly, it is an opportunity to put the findings in perspective, underscore their impor- tance, and suggest how the research could be extended in the future.
References
All papers mentioned in the main text should appear in the references section and all papers in this section have to be cited within the text. In the main text reference a paper using the format (Last name, year) and not the full citation.
Tables
There is no “right” number of tables, but 19 is always too many. Most empirical papers will include between 6 and 10 tables.
Use as uniform format as possible for your tables—both in terms of the structure (horizontal lines, etc.) and the presentation of information (number of significant digits, standard errors versus t- statistics; standard errors in parentheses versus in separate column, variable names, etc.)
Make sure your tables are easily readable, and that they can be understood without reference to the text. This includes avoiding nonstandard abbreviations (i.e., names you have given to your variables while working in Stata) and using directly the output tables from State without format- ting them.
Above all, make sure the tables further the paper’s goal. If you are including a bunch of variables in your regressions that we do not care about, consider putting them in an appendix table. The featured tables should give us the answers, clearly and succinctly. The other tables should simply provide reassurance that there is not monkey business going on.