Writing an Empirical Economics Paper

This document contains instructions for writing an empirical economics paper.

The Structure of the Empirical Paper

Empirical papers in economics have a consistent look and feel. Follow the usual outline:

Title Page: Includes title, your name, date, and anyone you want to thank for help

Abstract: In 100 words or less, state the main contribution or finding.

I.   Introduction

A. Statement of the topic and question to be analyzed

B. Rationale for choice of the topic (or why you find this interesting)

C. Explanation of the organization of the remainder of the paper

II. Literature Review

Choose some form (e.g., chronological or thematic) to organize the literature review. Mere listing and summary of several sources is not acceptable.  A good literature review interweaves the various articles in a seamless way.

III. Theoretical Analysis

Present a brief version of a model or highlight the theoretical source of the hypothesis to be tested.

IV. Empirical Analysis (the main and longest part of the empirical paper)

A.  The Data

i.      Provide sources on all variables

ii.      Provide summary statistics on all variables in a well-organized table

B.   Presentation and Interpretation of Results

V. Conclusion

A.  Restate the topic or question that was analyzed

B.   Provide your answer or conclusion, and compare to previous results in the literature

C.   Point out the best areas for further research

VI. References

Key Style Issues

.    Use the outline labeling scheme and section headings (e.g., IV.A.ii Summary Statistics) to organize your paper.

.    Citation style.: When referring to someone’s work, simply list the author’s last name and publication year (e.g., Jones [2004]). The full citation is in the references section.

.    Display regression results in the standard table format (see below for more detail).

Literature Review

A literature review is a summary of what other people have thought about your question or questions closely related to your topic. More specifically, it should explain how others have dealt with the issues you will be addressing in your paper.  The literature review usually serves two equally important purposes. First, it will explain how others have tackled your question. Second, it will provide you with some theory (economic or otherwise) which you can use in trying to answer the question or test someone else's answer.

The quality of your review depends on the quality of the papers you include, how on point they are, and your ability in distilling and presenting the findings in the literature.

You may certainly read nonprofessional sources like Newsweek or google your research question in order to stimulate the development of a policy topic, but these sources are not suitable for upper-  level undergraduate research.  Do not rely on mass media sources for your literature review.

For your literature review, you need work published in professional, journals. JSTOR, www.jstor.org, is an archive that contains the full text of a select group of journals in economics and other disciplines up through about four years ago (this varies from journal to journal). It is a good place to start, but you will want to go beyond JSTOR.

The references of the papers you find can lead you to other interesting papers and make your  literature search easier.  Once you find a single paper that addresses your research question, its bibliography is a gold mine of other papers that asked that question, or related questions.

Citation is important. After paraphrasing findings or explicitly quoting text, give credit by simply listing the last name of the author (use “et al.” when there are more than two authors) and year of  publication.  Do not include the entire reference in the text of your paper or in a footnote. Here is an example:  “Smith [2003] finds that more schooling lowers the probability of smoking.”

In the references, a full citation of the Smith [2003] article is presented. A standard referencing format you can use is the Chicago style. http://www.wisc.edu/writing/Handbook/DocChicago.html

Be warned of the dangers of plagiarism. It is very easy to plagiarize someone's work unintentionally; but this fact does not make plagiarism any less serious of an offense. Make certain that you either directly quote and attribute the quote, or paraphrase the source (no more than three consecutive words alike). Remember this: In general, direct quotation should be used sparingly in an economics research paper. Repeated use of direct quotation gives the impression of laziness and is often disruptive of your own style and method of organization.

A good strategy is to make sure that you paraphrase the work when you are actually taking the notes from the source, in case you forget to do so later on. Remember that the whole point of a literature review is to present others' work—your contribution will come a bit later. It is perfectly acceptable to say something like, "In his recent book on medical malpractice, Smith [2003] contends that ..."

Theoretical Section

"Why is a theory section needed in an empirical paper?" Because a complete answer to your question must rely on theory and data. You will need some theory to guide you in deciding which variables are relevant for your question. Common sense alone is not a sufficient reason for including or excluding certain variables in your analysis. Theory can also help in choosing the functional form and whether or not autocorrelation or heteroskedasticity are part of the data generation process.

In some cases, the theory section is quite clear. For example, earnings function papers have a solid theoretical foundation that underlies the use of the semi-log functional form. If your paper utilizes a measure of earnings as the dependent variable, you can present a theoretical argument for using the  semi-log form. as well as, for comparison, a regression that uses wage as the dependent variable.

However, it is also possible that there is no well developed theory for your question. In this case, it is common to see the literature review and theoretical sections combined. Your functional form and explanatory variables are chosen based on the work of others.

The theoretical section is a difficult piece of the empirical paper because some questions have precious little theory behind them.  Even those questions that do have a solid theoretical foundation are often difficult to explain.  When deciding what to say in terms of the theory section, remember that you are writing an empirical paper so the main function of the theory is to justify your empirical work. In other words, use the theoretical section to explain why you chose the particular explanatory variables you selected and the functional forms you used.

Empirical Results

This is the most important part of your paper. It is always divided into two main subsections: the data and the results.

The Data

Do not forget to provide the sources of your data and to help the reader by making a table that offers summary statistics on each variable. You should define each variable carefully and, if necessary, point out how the empirical measure deviates from its theoretical counterpart. Typical summary statistics that are offered include: max, min, average, and SD values for each variable. It is not unusual to offer histograms and other information for variables with skewed distributions. Excel is a fabulous tool here, and it is easy to get carried away. Remember, your goal should be clarity!

This subsection is the place to offer interesting information about the data. You should also point  out the limitations, if any, of your data.  You will want to describe your procedure in obtaining the data, making sure to point out key decisions in how you drew your sample. For example, in describing the wage variable, you might explain that you decided to remove all observations with negative values.  You will want to clearly state the time period (survey month and year, if CPS data) of your data set.

Do not go into excruciatingly painful detail on every step of your data collection.

Presentation and Interpretation of Results

This subsection is the heart of an empirical paper. Having set out the question, reviewed the previous literature, explored the theoretical perspective, and collected data, you are finally ready to do some econometrics.

Use subheadings to lead the reader through the different levels of your analysis. You might start with a table that compares averages for two groups, then move to a regression analysis, considering a variety of specifications and different sets of explanatory variables. You may also want to have a subheading for advanced analyses, such as robust standard errors.

You do not need to report every regression you run.

If you ran multiple models, use a table to report your results. The table is used to easily display the results from various models and invites comparison of coefficients.  Below is a template you can use to organize your results:

	Model 1: Dependent Variable	Model 2: Dependent Variable
Intercept	Est. Coefficient (est. SE)	Est. Coefficient (est. SE)
X1	Est. Coefficient (est. SE)	Est. Coefficient (est. SE)
X2		Est. Coefficient (est. SE)
X3		Est. Coefficient (est. SE)
X4		Est. Coefficient (est. SE)
X42
n
R2

The table shows how the first regression has no control variables. It is a simple, bivariate regression of X1 on the dependent variable. Model 2 adds three explanatory variables (presumably selected on the basis of some theoretical reasoning).

Notice how the table invites comparison of the models. In the discussion of the results, you would explain the results from each model and offer your opinion on the best answer to your research question.

The table can be augmented with asterisks (for statistically significant coefficient estimates) or other information (e.g., DW statistics for autocorrelation).  You can add notes at the bottom as needed.

To be sure, there is no consensus on the matter of significant figures in the economics profession.  One thing that is quite clear is that reporting ten or fifteen decimal places is silly and embarrassing. Avoid this. Some rounding must be applied to computer output.  While applying “pleasing to the eye,” the common practice in the social sciences, is better than nothing, you can do better by considering the likely size of the error in the results.

Writing up your results

Here is a list of good practices that should help you in writing up your results:

●    Motivate the error term: Begin the results subsection with a paragraph or two that explains the functional form. you are using and why there is a random disturbance term. What is the source of chance error in your data generation process?

●    Produce a table to present the results of various regressions. Use this table to discuss and compare the various models.

●    Interpret your results: Explicitly mention the estimated coefficient of the crucial explanatory variable in your analysis and comment on what it means.

●    Provide units: When discussing numbers, such as estimated coefficients or predicted Y values, remember to present the units of the numbers.

●    If you are testing a hypothesis, present the null hypothesis, compute the test statistic, and report the P-value.  State whether you reject or do not reject the null.

●    If you are estimating a parameter, report the estimated SE and a 95% confidence interval.

●    Discuss whether or not a coefficient is practically important—i.e., comment on the magnitude of the coefficient.

●    If needed, compute and interpret elasticities.

●    Compare your results to others in the literature. Do they support or contradict the relevant economic theory?

At the end of the empirical section of your paper, you should be able to draw a conclusion, even if it is a negative one. For example, you may find that there is no relationship between divorce and schooling in your data; this is still worth reporting.

Remember this: No study is absolutely perfect, but if you have done a thorough job in your empirical section, you should be able to reach some answer to your research question. This conclusion will then be inserted into your introductory paragraph in a slightly different form.

Things to Avoid

Here are common mistakes and poor practices to avoid like the plague:

●   Writing out a long equation with a beta in front of each variable. This is silly because the regression results table will list the variables in each model.

●   Even worse is: y =  β0   + β1x1  + β2 x2   + … + βkxk  + ε .  This has zero information content.

●   Splitting tables over more than one page.  Make every effort (adjusting font size and column width) to keep an entire table on the same page.

●   Repeating the same information, e.g., n =  12’662, for each variable in a summary statistics table.  This is visible noise pollution. Simply put n =  12’662 at the bottom of the table.

●   Referring to a paper by an author’s first name or title such as Mr. or Dr.  Use last name [year].

Conclusion

In the conclusion, your job is to give the paper's greatest hits. That is, you should restate your research question, give the high points of the literature survey and theory, remind your reader of the data and the methods you used, and restate your conclusion.

You may then go on to talk about the limitations of your analysis, any data you wish you could have garnered but couldn't, and what you would have liked to have done with your analysis but couldn't given the time limitations. This needn't be a long section—do not apologize for your work, but do suggest avenues for further research.

After writing the conclusion, you should then go to the beginning of the paper and write the Introduction. It should be a snap.

"And then I turn it in?" No, not quite yet. The last thing you should do is PROOFREAD your

paper. Even after spell checking the paper with your word processor, you should take the time to read it one last time before turning it in. Fix typographical errors, improve wording, and make sure the numbers make sense.

Your paper will be evaluated with the rubric below. The rubric relies on the material presented above. Follow these instructions and you will write a solid paper.