Major Project Seminar

Assignment:

 

Write a full draft for your major project paper.  The paper should consist of seven sections:

• Introduction
• Literature Review
• Data
• Theory
• Estimation
• Results
• Conclusion

 

Detailed Requirements for All Sections

 

Introduction

The main purpose of the introduction is to tell the reader what your research is about.  The introduction should do the following:

• State your research question.
• Explain why your research question is interesting. You may choose to briefly reference relevant literature or facts or data in the course of arguing that your research question is interesting or important, but you are not required to do so.
• Provide an overview of how your project will address the research question. You should give a general statement of your current thinking on how you will approach answering the research question with data.
• Provide an overview (1-2 sentences) of the results that you got.

What is most important is that it is easy for the reader to understand right from the beginning, precisely what your proposed paper is about, that the ideas are organized into clear, well-structured paragraphs, and that you “make a case” for why your project is worth doing.

 

As a rough guide, for a paper that is fourteen pages long, the introduction should be about two pages.

 

Literature Review

The main purpose of the literature review is to establish your credibility by showing the reader that you have made an effort to learn about your subject.  Your literature review should summarize existing research that is closely related to your research question.  For each paper you mention, you should write 1-4 sentences stating what the paper does, and how it relates to your major project paper.

 

Your literature review should focus on two types of papers:

• Key papers, meaning credible and well-known papers, whether theoretical or empirical, on subjects that are closely related to your research question.
• Empirical papers on subjects that are identical to or very similar to your proposed paper.

For some topics, the best way to show the reader that you are aware of the existing literature on the subject will be a more in-depth engagement with one, two, or three key papers.  For other topics, the best way to do this will be to provide a broad overview that mentions a larger number of papers, but in less detail.  For the purposes of this course, there is no “hard and fast” rule for how many papers should be in your literature review.  However, at the very minimum, I want you to do a Google Scholar search for existing research that is closely related to your question, and to write up a brief summary of what is or isn’t out there.

 

The best literature reviews will explain what your paper’s main contributions are by comparing your paper to what has already been done in the literature.

 

As a rough guide, for a paper that is fourteen pages long, the literature review should be about two pages.

 

Data

The main purpose of the data section is to orient your reader to your data set so that they are able to engage with the analysis.  A secondary purpose is to document your data in detail so that a future researcher could, in principle, reconstruct your data set.

 

The data section should give a detailed description of the data that you will use.  The data section must, it minimum, state your data source is with a degree of exactness that would allow your reader to track down the specific database or data tables.  For example, it would not be sufficient to say “I will use World Bank Data,” you would want to say “I will use [variables] from the World Bank’s Global Consumption Database.”  Many of you will want to use data from more than one data source; for example you may want to use a specialized database on a subject directly related to your topic, plus, for example, data on demographics from the American Community Survey.  If you are using proprietary data, data that you plan on collecting yourself, or data that is otherwise not publicly available, you must explain how you obtained access to or collected the data, or how you plan to do so.

 

The data section should be written in clear, professional prose with a sensible paragraph structure.  You must use your judgment about what information to include in your data section and how to structure it.  Some of the sort of things that you might choose to mention are:

• the “level” of observation – individual, city-level, state/province, country level
• the number of observations
• whether the data is based on a survey versus administrative data
• the time period covered by the data
• what variables are available

 

Your data section must include a table or tables of descriptive statistics for important variables in your data-set.  A descriptive statistics table should include at least the following information for each variable:  number of observations, mean, standard deviation, min, max.   All tables should have a readable, professionally dignified format.  Output directly copy-and-pasted from STATA or R is not acceptable.  You can use the STATA “sum/summarize” command to generate descriptive statistics.  Alternately you can generate descriptive statistics with Excel or many other software packages.  Regardless, you must format the output into a dignified and attractive table.

 

Reproducing your entire data set is not a substitute for a descriptive statistics table, and it is almost never a good idea to reproduce your entire set as a table in your paper, as these tables tend not to be very readable or useful.

 

If there are special features of your data that are particularly interesting or important, you may choose to include additional tables or charts that illustrate these features of the data.  You may choose to include additional tables, charts, or graphs to illustrate interesting or important features of the data, but you have to use your judgment about what to include.  Do not include tables or graphs as “filler”.

 

By the time you are writing your draft, you should have collected and worked with your data.  I expect your data section or tables to be “complete”, in the sense that I expect you to report descriptive statistics for all of the key variables that will be required to execute your analysis plan.

 

As a rough guide, for a paper that is fourteen pages long, the data section should be about two pages long.

 

Theory

The main purpose of the theory section is to establish your credibility by showing that you have thought carefully about how the world works.

 

The theory section should be a clear description of how you think the phenomenon you are studying works, with a special emphasis on cause and effect.  For example, if you are studying the effect of railroad construction on average land prices, then your theory section should be a clear description of how you think average land prices are determined:  what factors determine them, how do those factors interact with each other, and how does your variable of interest, in this example railroad construction, fit into the picture.

 

For the purposes of the major project paper, your theory section should not include your regression equations or discussion of econometric techniques like difference-in-differences.  Those things should be in the “Estimation” section of the paper.

 

Your theory section does not need to include a mathematical model or any equations.  It is normal in economics to describe relationships using mathematical equations and algebra, and you are encouraged to use equations in your model section if you are comfortable doing so.  However it is not required, and if you believe that you can more clearly explain your subject using words without any equations you are welcome to do that.  You are also welcome to use tables, flow-charts, diagrams, or any other presentation devices that you think are helpful.  It is usually best to “keep it simple”.

 

Your theory section does not need to refer to a specific named economic model that you learned in a previous class or discovered when researching your topic.  Depending on your topic, there may be named model that is particularly relevant and it may be most natural to draw on that model to frame your discussion.  For example, in a paper that is about trade patterns it will probably be natural to refer to the Heckscher-Ohlin model, or at least to the theory of comparative advantage.  In my example about railroad construction and land prices it would be natural to refer to the model of supply and demand in describing what determines average land prices.  It is important to emphasize, however, that what matters is not whether you “name-check” economic theory, it is that your model section gives a clear and sensible description of “how things work” and shows high quality critical reasoning about your subject.  Irrelevant expositions of economic theory will not earn you points, and may lose you points if they are lengthy, erroneous, or unclear.

 

Your theory section should include an explicit and careful discussion of any potential sources of endogeneity in the phenomenon you are studying. We are not looking for a technical discussion of what endogeneity means in econometrics.  We are looking for critical thinking regarding the ways in which cause and effect might be complicated in your research subject.  If you are having trouble working out where to start thinking about endogeneity here are two questions to ask:  First, are there any ways in which your outcome variable might actually cause your right hand side variable of interest?  In the example, are there any ways in which land value might determine railroad construction?  Second, are there any ways in which your outcome variable and your right hand side variable may be jointly determined by something else?  For example, if you are studying the effect of railroad construction on average land value, you might consider whether different levels of railroad construction in different places are the result of different local rates of economic development, which might also affect land prices.

 

As a rough guide, for a paper that is fourteen pages long, the theory section should be about two pages long.

 

Estimation (or Methods)

The main purpose of the estimation section is to precisely document exactly what data analysis you have done and why.

 

The estimation section should describe how you analyze the data in order to investigate your research question.  If you have ever taken a lab science class, it may be helpful to think of the “Estimation” section as corresponding to the “Methods” section of a lab report.  That is, the estimation section should specify exactly what data analysis you executed.

 

The estimation section should be written in clear, professional prose with a sensible paragraph structure.  You must use your judgment about what information to include in your estimation section and how to structure it.  However at the very least the estimation section should include

• A regression equation for the primary regression model that you will use to answer your research question.
• An explanation of how this data analysis addresses the research question. For example if there is a key regression coefficient that captures the relationship that you are trying to measure, say which coefficient it is, and explain how it captures what you’re trying to measure.

 

Some of the sort of things that you might choose to mention in this section are:

• The basic “methodology” that you plan to use. For example, if your study is going to use a difference-in-differences methodology, say that.  If your study is going to use an OLS regression on a cross-section of countries, say that.
• Some specific regressions that you plan to run in order to investigate your research questions.
• If your research question involves some difficulty in identifying cause-and-effect, any plans you have for making sure that your regressions measure the effect you want them to measure, or for exploring alternative possible cause-and-effect mechanisms.
• If you plan on using a sophisticated or high-tech type of estimation technique, how you will learn or have learned to use that technique correctly (e.g. “I will use a propensity-score-matching model, as learned in my advanced policy evaluation course at the Humphrey School in fall 2017).

 

As a rough guide, for a paper that is fourteen pages long, the estimation section should be about two pages long.

 

Results

The results section should present and interpret the results of your data analysis.  The results section is the most important section of your paper.  You will need to exercise a lot of judgment about what to include in your results section and how to structure it.  The main purpose of the results section is to explain and discuss what you discovered in your data analysis, and your interpretations and conclusions about the data analysis are an important part of this section.

 

Do not include results from regression techniques or tests that you do not understand, just because the option to run a particular analysis exists in the software command you are using.  You run an unnecessary risk by putting in fancy stuff unless you understand it and have a good case for why it is appropriate.

 

The results section should be written in clear, professional prose with a sensible paragraph structure.  It should include regression output tables.  It can also include other tables or charts that illustrate in detail any features of your results that are particularly interesting or important.  All tables and charts should have a readable, professionally dignified format.

 

At least one results table should be in a standard results table format with the following characteristics:

• Different versions of the regression should be arranged as different columns a single results table.
• The RHS variables used in your models should be rows in your results table.
• Row headings should be descriptive variable names, so that the reader can tell at a glance what the meaning of the RHS variable names is rather than having to decode them by reference to other parts of your paper.
• For each regression, you should report at least N, R², each regression coefficient, and the standard error associated with each regression coefficient.
• If you are comfortable translating the P-value (“P > |t|” in STATA) into asterisks next to the coefficients representing statistical significance at the 5% and 10% levels, you may do this. Otherwise, please also report the P-values themselves in your results table.
• Your table should have blank spaces in the rows corresponding to variables that are not used in each particular version of the regression model – it should thus be possible to tell exactly what RHS variables are used in each of the different versions of the model.
• If you vary the LHS variable, it should be clear from the column headings what the LHS variable is in each regression.

 

To the greatest extent possible, the results section should be self-contained – a paper within a paper.  It should be possible for an educated reader to skip to the results section and understand what you’ve done and what you’ve found.  Thus, for example, regression output tables should be introduced with enough information that the reader can work out what type of regression you ran and what variables you used.  The written prose should draw the reader’s attention to the most important and interesting features of the results.  The written prose should also relate the results to your research question.  It should state what conclusions you draw about the answer to your research question, and justify those conclusions based on the results of the data analysis.

As a rough guide, for a paper that is fourteen pages long, the results section should be about 3-4 pages.

 

Conclusions

The results section should contain a “Conclusions” subsection, that summarizes your research question, what you have done to answer the research question with evidence, and states your main conclusions based on your analysis. You should spend at least a few sentences at the end of the conclusion section discussing what further areas of exploration there could be based on your analysis.

As a rough guide, for a paper that is fourteen pages long, the conclusion section should be about three-quarters of a page long, and no longer than a page.

 

Formatting and length guidelines:

• Tables and figures should be included in the main body of the paper
• Use times New Roman font, size 12
• Use double-spacing except inside tables
• All page margins should be 1 inch
• Maximum length is 15 pages
• No minimum length
• Use Chicago “author-date” style for references, including both in-text references and the reference list at the end
• Charts and tables in the text do count towards the page limit
• Reference list at the end of the paper does not count towards the page limit

 

While there is no minimum length, it would be difficult to do the required tasks in fewer then ten pages, and if your paper is very short, or relies on extensive charts and tables to make up the length, you could consider this a signal that you need to do more work on some or all of the sections.