Your Research Question

1. A claim about something

   • Capital punishment is the most efficient deterrent for violent crimes.
   • Women are paid, on average, 33% less than men performing the same work.

2. As specific as possible, given the length constraints

   • Do candidates that spend more money than their opponents tend to win Congressional races?

3. Testable, with data that can provide some evidence one way or another

   • One study will never be “the” definitive proof of something, only suggestive evidence

Structure of an Empirical Paper

1. Introduction

2. Literature Review

3. Theory/Model

4. Data Description

5. Empirical Model

6. Results/Implications

7. Bibliography

Introduction

• Get to your research question ASAP! Make it the first sentence even.

• Hook your reader

  • Who cares? Why is this important? Why is this relevant? How does this affect people?
    • Statistics and background information can often help

Introduction II

• State your research question clearly and quickly

• Do NOT write a “blog post” about how you became interested in the question, or all the work (and dead-ends) that led you on the journey to reaching your final answer

  • Nobody cares about the labor pains, they just want to see the baby!

• Provide an outline of the rest of the paper:

  • Why your question matters
  • How you answer the question in this paper
  • What your identification strategy is and what models you use
  • What data you use
  • What your most important results are

Introduction III

Introduction IV

• Most people do not write enough in their introductions

• Consider the incentives of a (skimming) reader pressed for time

  • If someone only skims your intro, what do you want them to know??

• My rough suggestion: make your introduction about 15-20% of your paper:

Literature Review

• Literature Review can be summarized into the introduction or given its' own section (debatable)

• No work is totally original. It's okay!

  • What have other relevant researchers written and discovered about your topic?
  • What data and models did they use? What did they find?
  • How does your paper connect and stand apart from what's been done?
  • Does your paper use different data? A different model? Different controls?

Theory

• This is an economics course, so you must describe some economic theory behind the question you are asking and answering

• Most scholarly papers have a formal economic model, which then generates predictions that they test for with data

• You do not need a theoretical model, but you do need to discuss economic principles or concepts that are relevant

  • Often there may be multiple theories that might conflict, or our expectations might not be clear (these are the best papers!)
  • There may be a significant tradeoff between competing goals, values, or expectations

Theory II

Data I

• Describe your data sources
  • Who collected or compiled the data and how?
    • e.g. government agencies, businesses, nonprofits, social surveys, etc.
  • If you collected your own data (unlikely), what was your procedure?

Data II

• Describe the data itself
  • What are your variables? What—specifically, and in English—does each measure?
  • How many observations do you have?
  • If you transformed your variables—how and why?
  • e.g. recoded into categories or dummies
  • e.g. took logs or rescaled units

Data III

• Show your data! Show us basic summary statistics and any patterns
  • Use your judgment: .hi-purple[we don't want or need to see everything]
  • What do you think is interesting or important?
  • Plots > Tables > Words > Nothing

• Good ideas to always have:
  1. A table(s) of all variables used and their description
  2. A table(s) of summary statistics of variables
  3. A table of correlations of key variables (optional)
  4. Plots of (only) the most important variables & relationships (histograms, boxplots, scatterplots, etc)

Data: Variables

Data: Correlations

Data: Summary Statistics of Quantitative Variables

Data: Counts of Categorical Variables I

Data: Histogram I

Data: Histogram II

Data: Scatterplot I

Data: Scatterplot II

Data: Scatterplot III

Data: Scatterplot IV

Data: Scatterplot V

Data: Scatterplot VI

Empirical Model I

• Describe your empirical model and your identification strategy

  • for most of you, just OLS and trying to include as many controls to remove omitted variable bias

• Why did you pick certain variables?

• How do you battle endogeneity?

• Hypothesize your expected size and magnitude of key variables

  • Give some economic intution behind what we would expect!

Empirical Model II

Empirical Model II

Empirical Model III

• Length is the most important variable we care about

• Length probably endogenous, correlated with those other Grade-determining factors:

  • Why I included these controls!

• Likely expect Length to be positive and small

Empirical Model III

• You are probably interested specifically in the relationship only for econometrics papers, so we can focus Course specifically to a binary variable Metrics to see how the results differ between non-econometrics courses

• Alternatively, we can restrict our sample to only past econometrics classes

Empirical Model IV

• Describe the limitations of your model
  • Every paper, even Nobel prize-winning ones, have limitations and problems!
  • Limited and/or poor quality data
  • Endogeneity, simultaneous causation, omitted variable bias

Empirical Model IV

• Are your results robust across different model specifications?

  • Do the size(s) of the marginal effect(s) you care about change or reverse direction? Become/lose significance?

• At minimum, you must run several models, including a multivariate regression

  • Run several variations of your model with and without controls (e.g. just Y and X, Y and X1 and X2, etc.)
  • Check for nonlinearities: polynomials, logs, etc.

Results I

• Print a table(s) of your regression(s) results

  • R packages can help: huxtable, stargazer, modelsummary

• Interpret your data

  • What does a marginal (1 unit) change in X mean for Y, a 1% change, etc?
  • Is each coefficient statistically significant (at 10%, 5%, or 1% levels)?

Results I

Results: Interpretation!

• Are your estimates economically significant?

• How big is "big"?

“No economist has achieved scientific success as a result of a statistically significant coefficient. Massed observations, clever common sense, elegant theorems, new policies, sagacious economic reasoning, historical perspective, relevant accounting, these have all led to scientific success. Statistical significance has not.” — McCloskey & Ziliak (1996: 112)

Results: Interpretation!

Results: Interpretation!

Results: Implications

• Describe several implications of your paper
  • Policy implications
  • Proposals for new research
  • Effects on current understanding
  • What else should we try to found out to answer the question better?

Don't Get Discouraged

Don't Get Discouraged

Deadlines and Reminders (From the Assignment Page)

Grading of Final Paper (From the Assignment Page)

Submitting your Final Paper

Example 1

“Exploring the Effects of Children and Marriage on Men’s and Women’s Incomes”

• Cross-sectional data for individual i

Example 2

“Does Spending More on the Offensive Line & the Defensive Line Affect NFL Team Wins?”

• Panel data for team t in year y

Example 3

“Buy You a Vote”

• Panel data for individual i at time t, with individual fixed effects (αi) and year fixed effects (τt)

Example 4

“A Cross-Sectional Study on the Effect of State Minimum Wage on Youth Unemployment at the State Level”

• Cross-sectional data for U.S. State i

Example 5

“Is Twitter Strong Enough to Measure NBA Player Performance?”

• Cross-sectional data for player i

#