Unformatted Attachment Preview
Econ 3390.01 Applied Health Economics
Empirical Paper Assignment
Results & Discussion
This phase of the project entails the completion and submission of two components, including:
1. Results & Discussion [Documents]
2. Data analysis command file (analysis.do) [Command Files]
where [brackets] denote where this file(s) should be saved in the “Replication Documentation”
1. Results & Discussion. This document includes what will become the Results and Discussion
sections of your final paper.
a. The Results section should follow directly from your Data & Methods section where you
developed your model specification. However, your regression specification can/may have
changed since you submitted your Data & Methods section and received feedback. This is
acceptable and, in fact, encouraged. Just be sure to update the Data & Methods in the final
version of your paper. Here, you provide a table of your regression results (potentially
including multiple columns) – or several tables if necessary – and a brief description of
those results. This section should be concise (1-3 pages, including tables), but will vary
depending on the number of specifications included. Some things to consider in presenting
and describing your results are:
Use outreg2 to generate your tables, or otherwise format them to disciplinary
conventions – do not copy/paste tables from Stata.
Do not report your estimates in terms of β’s or by using obscure abbreviations used
in your codebook (such as lrgdppc). For the purposes of the results table, write out
the variable name (such as logged real GDP per capita) that is consistent with the
name you used in the Data & Methods section.
Do not include too many significant digits. The goal is readability, not false
Provide enough information so that the reader can understand your results without
looking back at the text excessively. The notes on a table can be used for these
purposes. For example, are you reporting standard errors or t-statistics in
parentheses under the coefficient (standard errors are more common and generally
preferred) and what do the “*” next to the coefficients denote (which level of
If you estimate several different specifications of your model (using the same
dependent variable), include these as several different columns of the same table to
facilitate comparison and improve readability. Table 1 below (from Nikolov, 2013)
is a good example of this.
Page 1 of 4
In describing your results – be brief. Your reader has already seen the table. For
example, in a paper focused on the effect of education on wages, following Table
1, the text might say:
Table 1 shows that including a measure of ability in the wage
equation dramatically lowers the predicted effect of education on
earnings. Column 1 does not include an ability measure and
indicates that a year of education raises wages by 9.1 percent.
Column 2 adds the ability measure and the education effect drops to
3.1 percent. Columns 3 and 4 show that this general pattern is
repeated even when state level dummy variables are included. The
estimates in Table 1 are therefore consistent with the hypothesis that
the OLS estimates suffer from an upward ability bias (Nikolov,
2013, p. 11).
Notice that this description of the results begins and ends with statements that put
the results in the context of the paper’s thesis. This is a good technique.
Always interpret coefficients using the correct units.
If you are testing hypotheses other than standard significance tests (such as a test
against a one-sided alternative, comparing two coefficients, or an exclusion
restriction test), be sure to present the null and alternative hypothesis, compute the
test statistic, report the p-value, and state your conclusion about the null hypothesis.
You should not do this for every coefficient you estimate, your table will already
make it clear to the reader which coefficients are statistically significant.
Pay attention to practical/economic significance (effect size), not just statistical
significance. If you are using a large data set even very small effects can be
statistically significant. Provide the reader with some context to put the effect in
If any of your results are strange or unexpected be sure to note this, you can and
should return to it in your discussion section.
Are your results consistent with or counter to the existing empirical research? Note
this so that you can return to it in your discussion section.
Page 2 of 4
The purpose of crafting these short descriptive paragraphs is to guide your reader through
the results to focus their attention on the most important aspects of the table(s). Focus on
what is important and be as clear as possible.
By the end of your results section, you should be able to draw a conclusion, even if it is not
the one you expected. An unexpected result is a result nonetheless. Insignificant results can
also be important and have significant policy implications. For example, perhaps we
discover the 3-pt field goal percentage does not have an impact on a NBA team’s winning
percentage, holding 2-pt field goal percentage (among other variables) constant. This might
lead coaches to focus less of 3-pt field goals during practice and in recruiting players. As
another example, suppose we find that after controlling for education and ability, work
experience has no impact on worker productivity. This may lead employers to focus less
on experience in employee recruitment. Remember, no study is a perfect study, but if you
have done thorough empirical work, you should be able to reach some answer to your
b. The Discussion section includes a discussion of the interpretation and implications of your
empirical results, a discussion of potential explanations for the results based on your
conceptual framework and/or previous literature, and a discussion of how your results are
consistent with, an extension of, or at odds with the existing literature. This section should
be concise (1-3 pages) and provide the reader with all of the relevant context needed to
understand and interpret your results. Some things to consider in discussing your results
If you have done a good job writing your results section, the discussion section
should be relatively straightforward.
Many good research questions in economics have policy implications. This is your
chance to mention these (you may have already alluded to these in your introduction
and/or literature review). Be careful not make value judgements – let your analysis
speak for itself.
If your results are strange or unexpected, suggest reasons why. This could be due
to limitations of your methodology or data or perhaps inconsistencies in the
Discuss these potential limitations of your research, such as: small sample size, less
than ideal unit of analysis (state or national vs. individual), simplicity of the
functional form, data (un)availability, selection bias, time period, omitted variables,
If there are several different possible explanations or interpretations of your results,
which of these do you think are more likely? Are your interpretations based on a
theoretical or conceptual understanding of the issue or based in the previous
literature, or both?
I am being hypocritical here. Academic writing should generally not use “etc.”
Page 3 of 4
2. Analysis.do command file. In one command file (Analysis.do), write code that produces all
empirical results presented in the Results section, using your analysis.dta data file that you
built during data construction. Save this file in the Command Files folder. This may include
mean estimations and additional graphics, but for most of you this includes only the code to
estimate your regression(s) and to test any additional hypotheses. Remember, for full credit all
command files should be fully commented and easily readable. In the case of your analysis.do
command file, you should provide comments noting which commands produce which results
tables in the final paper.
Page 4 of 4