Randomized Controlled Trials Overview

Randomized Controlled Trials (RCTs)

• RCTs are the gold standard for establishing causality.
• Random assignment ensures treatment and control groups are similar.
• Any difference in outcomes is attributed to the treatment.
• RCTs are essential for understanding how policies and interventions affect outcomes.
• They distinguish between correlation and causation.
• Correlation does not imply causation.
• Causation establishes a direct link between a cause and its effect.
• Average Causal Effect (ACE) measures the average impact of an intervention across a population.
• ACE accounts for variations in individual responses to the intervention.
• Selection bias is a key challenge in causal inference.
• Selection bias occurs when groups differ systematically due to factors unrelated to the treatment.
• Randomized Controlled Trials (RCTs) effectively eliminate selection bias.
• RCTs randomly assign participants to treatment or control groups.
• This ensures that the only difference between the groups is the treatment.
• RCTs must be well-designed to avoid biases and threats to internal and external validity.

Fixed Effects (FE)

• Controls for unchanging differences within entities (e.g., schools, people).
• Eliminates bias from time-invariant factors like school culture or teacher quality.
• FE regression isolates the effect of a treatment over time for specific groups.
• Powerful in scenarios with unobservable or unmeasurable factors that vary across entities but remain constant over time.
• Comparing outcomes before and after treatment between similar groups, helps in isolating the effect of the treatment on an outcome.

Instrumental Variables (IV)

• Fixes endogeneity when the treatment isn't randomly assigned.
• Uses an external “instrument” (e.g., random selection for laptops) to assess the treatment's impact, not confounding factors.
• IV method isolates the treatment effect, even if treatment assignment is not random.

Difference-in-Differences (DID)

• DID compares changes over time across a treatment and control group to determine the effect of intervention.
• The method identifies differences across the two groups after implementing a treatment that wasn't present in the previous time period.
• DID relies on the assumption that both groups would follow parallel trends without the intervention.
• DID regression isolates treatment effect by capturing changes in the experimental group compared to the control group over time.

Regression Discontinuity (RD)

• RD estimates causal effects by comparing people just above and below a threshold (e.g., GPA).
• Assumes that people just above and below the threshold are comparable except for the treatment.
• Helpful when there is a 'cutoff' for an intervention (e.g., students with 3.0 GPA receive laptops.)
• Can use either simple or multiple regression.