Regression discontinuity design

Definition

Regression discontinuity design (RDD) is a quasi-experimental research design used to identify causal effects by exploiting a predetermined cutoff point in a continuous variable that assigns treatment or intervention. This design helps to assess the impact of the treatment by comparing outcomes for observations just above and just below the cutoff, thus mimicking random assignment and addressing potential confounding variables.

5 Must Know Facts For Your Next Test

1. RDD relies on the assumption that units just above and below the cutoff are similar in all respects except for the treatment, allowing for valid causal inference.
2. It is particularly useful when random assignment is not feasible, making it a valuable tool in observational studies where ethical or practical considerations prevent randomized control trials.
3. RDD can be implemented in both sharp and fuzzy forms, with sharp RDD having a strict cutoff and fuzzy RDD allowing for some noncompliance around the cutoff.
4. Estimation of treatment effects in RDD often involves non-parametric methods, using local linear regression techniques to improve accuracy near the cutoff.
5. RDD can be visually represented through graphs that plot the outcome against the running variable, showing a clear jump at the cutoff point, which indicates the effect of the treatment.

Review Questions

• How does regression discontinuity design help in establishing causal relationships in observational studies?
  • Regression discontinuity design helps establish causal relationships by comparing outcomes of units that are very similar but differ in treatment based on a specific cutoff. This allows researchers to mimic random assignment, as those just above and below the cutoff are likely to have similar characteristics. By focusing on this narrow band around the cutoff, RDD can effectively control for confounding variables that could bias results.
• What are the key differences between sharp and fuzzy regression discontinuity designs, and how do they affect the interpretation of treatment effects?
  • Sharp regression discontinuity design has a strict adherence to the cutoff where individuals either receive or do not receive treatment based solely on their position relative to it. In contrast, fuzzy regression discontinuity allows for some individuals around the cutoff to not comply with treatment assignment, meaning not everyone above the cutoff receives treatment. This distinction affects interpretation; sharp RDD provides clearer causal inference while fuzzy RDD requires careful consideration of noncompliance when estimating treatment effects.
• Evaluate how regression discontinuity design can be applied in real-world scenarios and its limitations compared to other causal inference methods.
  • Regression discontinuity design can be applied in various real-world scenarios, such as assessing educational interventions based on test scores or policy changes tied to age thresholds. However, its limitations include the requirement of a well-defined cutoff and assumptions that units near the cutoff are comparable. Compared to randomized control trials, RDD may provide less robust causal estimates if there are concerns about manipulation of the running variable or if there are unobserved differences between groups that could bias results.

Related terms

Cutoff Point: The specific value on a continuous variable that determines whether an individual receives the treatment or not in regression discontinuity design.

Local Average Treatment Effect (LATE): The average effect of a treatment on individuals who are induced to receive the treatment due to being near the cutoff in a regression discontinuity design.

Instrumental Variable: A variable that is used in regression analysis to account for unobserved confounding, helping to isolate the causal effect of a treatment on an outcome.