Econ 204 Econometrics

Quote

Correlation doesn’t imply Causation

• Econometrics
• Bivariate Ordinary Least Squares Regression
  • Confidence Intervals and Hypothesis Testing in OLS Linear Regression
  • Problem: Autocorrelation? → Lagged Variables
  • Problem: Heteroskedatisticy? → robust standard errors

Different Models §

• Multivariate Ordinary Least Squares Regression
  • Problem: Omitted Variables → Include the variables
  • Problem: Post-Treatment Variable
    • Mediator Bias → Your choice on what you want
    • Confounder Bias → Your choice on what you want
• Nonlinear Models
  • → Choose the best one

Techniques §

• Dummy Variables
  • Multiplicative Dummies
  • Categories
  • Difference of Means test
• Fixed Effects Model for Panel Data
  • Problem: Endogeneity due to time-dependent or panel (=group)-dependent factors
  • LSDV approach
  • De-meaned approach
  • Difference In Difference
• Two-Stage Lease Squares Regression and Instrumental Variables
  • Simultaneous Equation Model

Designing and Analyzing Experiments §

• Controlled Experiments
  • Natural Experiments
• Regression Discontinuity

Endogeneity fight Summary:

1. Soak-up
   • Other causality? → Multivariate Ordinary Least Squares Regression
   • Mediator/Confounder? → Post-Treatment Variable
   • Time/Group-dependent? → Fixed Effects Model for Panel Data
2. Isolate/Create the exogenous variation
   • Isolation? → Two-Stage Lease Squares Regression and IV
   • Randomization? → Controlled Experiments
3. Regression Discontinuity