Doing Research in Social Science

1

• Doing Research in Social Science
• A tutorial
• Yexiao Xu
• School of Management
• The University of Texas at Dallas
• 2009 JiaoTong University

2
Outline

• The process of doing research
• The six steps
• Research methodologies
• Different estimation techniques
• Issues in social science research
• Data mining and data snooping
• Getting you feet wet!
• Good training and learning by doing
• Concluding comments

3
The Process of Doing Research

• Choosing the right topic
• What do you like to know?
• Formulating the problem
• What is your hypotheses?
• Collecting data
• Available data versus hand collected data
• Implementation
• Estimation methodology and mathematical technique
• Draw conclusion
• Analyzing the results
• Presenting your results
• The effective writing

4
Choosing the Right Topic

• The most important step
• It is not something that a textbook teaches you
• How to Choose a Topic?
• Read extensively
• Think logically
• Combing real life experience
• Listening to others
• Be skeptical
• Read an article at least three times
• It is an evolving process
• Your research question can change along the way

5
Choose Topics by Asking Questions

• Are there problems with existing studies?
• Sample size, sample selection, robustness
• Can we extend the current studies?
• Fama and French 3 factors to Carhart 4 factors
• Extend the MM theory to real world scenarios
• Labor income in asset pricing model
• Understanding puzzles
• IPO underpricing, returns from characteristic
  sorted portfolios, equity premium puzzle, earning
  announcement drift, closed-end fund

6
Choose Topics by Asking Questions (Contd)

• Why different studies finds different result?
• What might be the competing theory
• Are certain phenomena against your intuition?
• What are the consequences of new results?
• New theory
• Are there natural experiments?
• Endogeneity issues
• Can we apply new methodologies?
• Better and more robust estimates

7
Formulating the Problem

• Understanding current theories
• Application of current theories
• Carefully formulate your hypothesis
• Dont be too ambitious, but be thorough
• What can be used as your control variables?
• Be clear about the alternatives
• What is the proper statistical models?
• From mathematical models to statistical models
• A wrong but popular approach
• Starting from data

8
Data Collection

• Readily available data
• Understanding the limitation of your data
• Checking for potential problems
• Statistical distribution of the data
• Missing data issue
• Survivalship bias
• Filtering
• Hand collected data
• Thinking carefully first
• What potential variables do you need?
• Getting around with your limitation

9
Implementation

• Doing a simple analysis first
• Understanding your data
• Establishing a correlation
• One way or two way table analysis
• Difference of difference
• A regression analysis
• The fundamental variables versus control
  variables
• More sophisticated approaches
• MLE and GMM
• Monte Carlo simulation
• Bayesian approach
• Event studies

10
Draw Conclusions

• Analyzing your results
• Statistical inference
• Robust standard deviation
• Association versus causality
• Regression analysis can only establish
  association
• Granger causality
• The anticipation effect
• Event study
• Robust check
• Could your results be consistent with other
  theories

11
Presenting Your Results

• Introduction
• Your contributions
• The relevance of literature
• Everybody can understand! It is the first
  impression
• Presenting your theory
• Be super careful in exposition
• Data and methodology
• State clearly of your data source
• How did you process your data?
• Empirical analysis
• Guide your readers
• Do not expecting your readers to analysis your
  tables or graphs
• Both tables and graphs should be self-contained
• Concluding comments
• What have you tackled?
• What have you not been able to address?

12
Research Methodology

• Regression analysis
• Assumptions and limitations
• Efficiency versus robustness
• MLE versus GMM
• Different view about parameters
• Frequencist versus Bayesian
• The controlled experiment
• Monte Carlo simulation
• Estimating unobservable states
• Regime switch, state variable
• Resampling
• Bootstrapping versus Jackknife
• Cross-validation

13
Issues in Social Science Research

• The non-experimental nature of social science
  study
• Results could be consistent with other theories
• We are doing a detective job!
• Potential endogeneity problem
• An explanatory variable is influenced by the
  dependent variable
• Spurious regression
• Sharing a trend
• Outliers
• Influential data points
• Results are sensitive to subsample periods

14
Issues in Social Science Research (Contd)

• Sample selection bias
• Select a particular sample that is more likely to
  exhibit the characteristic you desire
• The data mining problem
• Seeking pattern from historical data
• Finding a model that produces what you want
• Data snooping
• Using knowledge about the data to guide
  subsequent research (hypotheses) on the same data
• Try different (aspect) data that is consistent
  with a model

15
Getting you feet wet!

• Learning by doing
• Experience accumulates
• The five steps
• Finding an interesting (classical) paper
• Understanding the paper
• Replicating the basic results
• Extending the paper
• Writing a good report
• Talk to different professors

16
Concluding Comments

• Learn your basic theory well
• Start to ask why
• Take a skeptical view
• Have a simple model in mind

17
Concluding Comments

• Thank You!