How to Write an Empirical Paper
Steps
1. Choose a topic. 2. Find a (good) question. 3. Figure out an identification strategy.
- Find your data and methodology
4. Econometric estimation. 5. Write down your results.
2
1. Choose a Topic
Choose the area and a specific topic
Read the newspaper, look at the course syllabus Browse journals and textbooks (EconLit)
Discuss with classmates, professors, friends, ...
You have to like it! Start small! But keep it interesting!
1. Choose a Topic
Literature review
Browse journals and textbooks (EconLit) Think of related areas
Build a bibliography of research papers on the topic that seem interesting to you (about 10-15)
Read previous research for background.
4
2. Find a (good) question
Something that hasnt been answered yet!
The hard way: Come up with a new question
Unlikely.
AGAIN: START SMALL!
The easy way: Add something new to something old
5
2. Find a (good) question
The easy way: Replicate and extend the results of an existing paper.
come up with a new explanation for an empirical fact apply a different methodology/data/model/experiment change one of the assumptions use a new instrument
The question can be out there, but you may come up with a new way of approaching it. You may find a gap in the literature.
6
3. Identification Strategy
How are you planning to answer your question?
a. Methodology.
b. Data
3. Identification Strategy
a. Methodology
Reduced form regression analysis Instrumental variables Structural model Matching, Propensity score estimation Non-parametric bounds Experiment (lab or natural) etc
3. Identification Strategy
a. Methodology It will depend on the specific question, the data available, and your own strengths!
Be very aware of (and honest about) the assumptions that are driving identification.
Functional form, independence,
3. Identification Strategy
b. Data Household surveys.
US: CPS, NLSY, PSID (Census, SIPP, NELS, etc). Europe: ECHP, national employment and consumption surveys, BHP, GSOEP. Canada: Survey of Household Spending, Labor Force Survey, Census, etc.
Firm surveys (CompuStat, CRSP)
Aggregate data.
Penn World Tables, World Bank, OECD, etc.
10
3. Identification Strategy
Before starting with the estimation be aware of the type of data:
Cross-section Census Time series Panel
the type of variables:
Numerical (Continuous, Discrete) Categorical (Ordered, Unordered)
11
3. Identification Strategy
GET TO KNOW YOUR DATA Read manual and codebook. Become familiar with the structure of the data set
identifiers, household vs. individual variables
Identify the variables of interest What are important controls Run a lot of descriptive statistics. Get to know your variables (which currency, missing values, etc.) and detect mistakes
12
4. Econometric estimation
Main specification
What are your assumptions and are the reasonable? What are your sources of identification? Identify possible problems, caveats, alternative explanations
Run a lot of robustness checks
change the sample size experiment with control variables relax or strengthen assumptions fake counterfactuals
13
Present your preliminary results and get feedback
5. Write the Paper
Typical structure of a paper:
1. Introduction. 2. Background, previous literature.
3. 4. 5. Model Data and methodology. Results and robustness checks.
6. Conclusion.
14
5. Write the Paper
3. Model
-
State equations to be estimated
derived from a structural model or reduced form (clarification of the dependent and the independent variable) stating the equation allows you to discuss assumptions and potential problems (endogeneity, reverse causality, etc.)
Xi = age, age squared, marital status, number of children, etc.
discuss estimation method appropriate to solve problems (panel data, IV-rergession, DiD, rergession continuity, etc. discuss assumptions underlying your estimation method (e.g. linearity of the model, E[|X]=0, semi-elasticity, etc.) if not all assumptions are likely to be fullfilled at least discuss the arising bias
15
5. Write the Paper
4. Data -
describe your data source and the sample selection
e.g. Swiss Labor Force Statistics 2006: 4125 women age 15-98, 1347 not working, 70 no wage observation 2708 observations)
describe your sample: descriptive statistics of main variables - evtl. by comparions groups, subsets, etc.
Variable wage university age married child Obs 2708 2708 2708 2708 2708 Mean 3.269529 .0705318 40.7031 .5029542 .6458641 Std. Dev. .4671046 .2560884 10.98031 .5000836 .9753867 Min .2850343 0 16 0 0 Max 5.60919 1 81 1 4
discuss how variables are constructed, measurement problems, missing control variables, etc.
16
5. Write the Paper
5. Results and robustness checks. - recall estimation method - regression results (e.g. coefficient, significance, etc.) - alternative specifications:
17
5. Write the Paper
6. Conclusion
Summarize shortly the paper Discuss your results with potential interpretation Policy conclusion
18
IMPORTANT
set personal deadlines
write down the question
write down a proposal (question, contribution to the literature, identification strategy)
19
Bibliography
Wooldridge (Introductory Econometrics). Chapter 19. Carrying out an Empirical Project
Thomson (A Guide for the Young Economist) Chapter 1. Writing Papers
20
