Final Prep: Topics

EC 421: Introduction to Econometrics

Author

Edward Rubin

README: The final exam will cover the following topics.

I will also assume you understand the material from the midterm.

As with the midterm, you do not need to memorize proofs. Understand the steps/reasoning. Intuition is paramount. I might ask how you get from one step to the next. I won’t ask you to write down a full proof.

Slide Set 7: Time Series

  • Notation
  • Assumptions
  • Static models
  • Dynamic models
    • With lagged explanatory variables
    • Autoregressive, distributed lag (ADL) models
    • ADL(p,q)
    • Long-run vs. short-run effects
  • Contemporaenous exogeneity

Slide Set 8: Autocorrelation

  • Definition of autocorrelation
  • Negative and positive autocorrelation
  • Equation/formula for AR(p) processes
  • OLS implications for static models and dynamic models with lagged explanatory variables
  • OLS implications/bias for dynamic models with lagged dependent variables
  • Breusch-Godfrey test
    • Hypotheses
    • Interpretation of results
    • Conclusion
  • Misspecification
  • FGLS: General idea

Slide Set 9: Non-Stationary Time Series

  • Definition/requirements of nonstationarity
    • Mean
    • Variance
    • Covariance
  • Random walks
  • Spurious correlations
  • Differencing

Slide Set 10: Causality

  • Prediction vs. causal inference/estimation
  • Correlation vs. causation
  • Reverse causality
  • Experiments/RCTs
  • The “ideal experiment” (ideal dataset)
  • Treatment effects:
    • Individual effects
    • Average treatment effects
    • Constant treatment effects
  • Selection bias

Slide Set 11: Instrumental Variables

  • Exogeneity and endogeneity
  • Requirements for a valid instrument
    • Relevant
    • Exogenous
  • Probability limit
  • First stage and reduced form
  • Two-stage least squares
    • Motivation
    • First stage
    • Second stage
  • Interpretation

Slide Set 12: Panels

  • Why panel data can help
    • Within-unit vs. between-unit variation
    • Time-invariant omitted variables
  • Pooled OLS vs. fixed effects (FE)
  • Unit and time fixed effects
  • Within transformation intuition
  • FE interpretation
    • Identification from within variation
    • Limits of FE (time-varying confounders, reverse causality, measurement error, weak within variation)
  • Difference-in-differences (DiD)
    • Core idea: difference between two differences
    • DiD regression setup and interaction term interpretation
    • Two-way fixed effects (TWFE) framing
  • Parallel trends assumption
    • Meaning
    • Common threats/violations
  • When DiD is most useful