📊

Ch 2 - V4 (Ceteris Paribus)

May 9, 2025

Lecture Notes: Randomized Controlled Trials and Regression Analysis

Introduction to Randomized Controlled Trials (RCTs)

  • Definition: RCTs create two comparable groups - a treatment group and a control group.
  • Purpose: The control group provides a counterfactual, helping to determine causal relationships by understanding what would have happened if the treatment was not applied.

Importance of Counterfactuals

  • Causal Analysis: A correct counterfactual allows us to conclude causation.
  • Alternative Approach:
    • When RCTs are not feasible, identify a correct counterfactual for those receiving treatment.
    • Example: To assess the impact of college on wages, compare individuals who attended college with those who went straight to the workforce.

Concept of Ceteris Paribus

  • Definition: Latin phrase meaning "all else equal."
  • Application: In causal analysis, to determine the impact of college, imagine splitting a person's life into two paths—one attending college, the other not.

Challenges in Causal Analysis

  • Non-Comparable Groups: Differences in IQ, family income, and work experience can skew results.
  • Experience Factor: Skipping college provides work experience, influencing income.

Regression Analysis

  • Purpose: Correct biases by estimating relationships between dependent and independent variables.
  • How It Works: Controls for specified variables, estimating independent effects.
    • Example: Regression compares individuals with similar IQ, family income, and work experience to isolate the effect of college on income.
    • Outcome: A more accurate estimation shows college increases income by 28%, not 37%.

Issues with Regression: Omitted Variable Bias

  • Definition: Leaving out variables that affect both the treatment and outcome leads to biased results.
  • Examples of Omitted Variables:
    • Student aptitude, motivation, and inclination.
    • Different college opportunities, such as legacy admissions.

Case Study on Omitted Variable Bias

  • 1999 Study on Nightlights and Myopia:
    • Initial Claim: Nightlights cause myopia in infants.
    • Overlooked Factor: Genetics - parents with myopia were more likely to use nightlights for their convenience.
    • Conclusion: The correlation was due to a third factor (genetics), not the nightlight itself.

Conclusion

  • Regression is a powerful tool for causal analysis, but must account for all relevant variables.
  • Continuous effort is required to identify and control for omitted variables to avoid biased conclusions.