Sharp vs Fuzzy RDD
You'll learn: How regression discontinuity designs differ when treatment assignment at the threshold is guaranteed versus merely more likely.
The Core Distinction
In a sharp RDD, crossing the threshold deterministically assigns treatment. Everyone above the cutoff gets treated; everyone below does not. There's no ambiguity—it's a clean on/off switch.
In a fuzzy RDD, crossing the threshold increases the probability of treatment, but doesn't guarantee it. Some people above the cutoff don't receive treatment, and some below do. The discontinuity is "fuzzy" because treatment uptake jumps but doesn't flip from 0% to 100%.
Real-World Examples
Sharp: A scholarship given to all students scoring ≥70 on an entrance exam. If you score 70, you get it. Score 69.9, you don't. No exceptions.
Fuzzy: A job training program offered to all unemployed workers over age 55. Some decline, and some younger workers sneak in through appeals. Age 55 creates a discontinuity in treatment probability, not certainty.
Why It Matters
Sharp RDD lets you estimate the treatment effect directly by comparing outcomes just above and just below the cutoff—the jump in treatment perfectly aligns with the jump in outcomes.
Fuzzy RDD requires an extra step. You first estimate the jump in treatment probability at the cutoff, then scale the outcome jump accordingly. It's essentially an instrumental variables approach, using threshold-crossing as an instrument for actual treatment receipt.
Key Takeaway:
Sharp RDD has deterministic treatment assignment at the cutoff (all-or-nothing), while fuzzy RDD has probabilistic assignment (treatment likelihood jumps but isn't absolute), requiring adjusted estimation methods.