M1-088M1: Mathematics & Statistics FundamentalsCausal InferenceMedium
⚡Flashcard Drill➔Mastery:

Explain Regression Discontinuity Design (RDD) and Its Core Continuity Assumption

🎯 Active Recall Self-Check (Draft First, Verify After)
Login-free practice tracker: recall core answer, audit pitfalls, and log review status.
Step 1 · Draft your concise answer & assumptions
Step 2 · Check Pitfalls & Follow-up Inquiries
Next: M1-089 →
📐 Mathematical Definition
τ=lim⁡x↓cE[Y∣X=x]−lim⁡x↑cE[Y∣X=x]\tau=\lim_{x\downarrow c}\mathbb E[Y\mid X=x]-\lim_{x\uparrow c}\mathbb E[Y\mid X=x]
⚡ One-Sentence Core Answer
Regression Discontinuity Design (RDD) exploits exogenous threshold cutoffs on continuous running variables (e.g. Credit score ≥700\ge 700 gets loan), comparing units immediately above and below the threshold as a quasi-randomized trial.

📌 Key Takeaways

  • •
    Sharp RDD: Treatment assignment is a deterministic step function of running variable XX: Ti=I(Xi≥c)T_i = \mathbb{I}(X_i \ge c).
  • •
    Fuzzy RDD: Cutoff increases the probability of treatment discontinuously: P(Ti=1∣Xi)P(T_i=1\mid X_i) jumps at cc (analyzed via Instrumental Variables).
  • •
    Core Continuity Assumption: Potential outcomes E[Y(0)∣X]E[Y(0)\mid X] and E[Y(1)∣X]E[Y(1)\mid X] must be smooth and continuous at cutoff cc; units cannot precisely manipulate their running variable around the threshold.

📐 Mathematical Derivations

Identification theorem: The treatment effect at cutoff cc is identified as the discontinuous jump in conditional expectation: τRDD(c)=lim⁡x↓cE[Y∣X=x]−lim⁡x↑cE[Y∣X=x]\tau_{\text{RDD}}(c) = \lim_{x \downarrow c} E[Y\mid X=x] - \lim_{x \uparrow c} E[Y\mid X=x]. Expanding potential outcomes: lim⁡x↓cE[Y∣X=x]=lim⁡x↓cE[Y(1)∣X=x]=E[Y(1)∣X=c]\lim_{x \downarrow c} E[Y\mid X=x] = \lim_{x \downarrow c} E[Y(1)\mid X=x] = E[Y(1)\mid X=c] (by continuity). Similarly, lim⁡x↑cE[Y∣X=x]=lim⁡x↑cE[Y(0)∣X=x]=E[Y(0)∣X=c]\lim_{x \uparrow c} E[Y\mid X=x] = \lim_{x \uparrow c} E[Y(0)\mid X=x] = E[Y(0)\mid X=c]. Subtracting gives: τRDD(c)=E[Y(1)−Y(0)∣X=c]\tau_{\text{RDD}}(c) = E[Y(1) - Y(0)\mid X=c], which is the exact causal Local Average Treatment Effect at the threshold. Local linear regression fits separate slopes on each side of cc within bandwidth hh: min⁡α,β,τ,γ∑i:∣Xi−c∣≤h(Yi−α−β(Xi−c)−τTi−γ(Xi−c)Ti)2K(Xi−ch)\min_{\alpha, \beta, \tau, \gamma} \sum_{i: |X_i - c| \le h} \left(Y_i - \alpha - \beta(X_i - c) - \tau T_i - \gamma(X_i - c)T_i\right)^2 K\left(\frac{X_i - c}{h}\right).

🏭 Production Trade-offs

Validating RDD assumptions: (1) McCrary Density Test: Checks whether the density of the running variable f(X)f(X) exhibits a sudden spike/cliff at cutoff cc. A spike indicates that users strategically manipulated their score (e.g. Submitting fake receipts to cross a bonus threshold), violating local randomization. (2) Covariate Continuity: Pre-treatment covariates (age, gender, prior activity) must be smooth and continuous across the cutoff.
⚠️ Common Interview Pitfalls
  • ✕
    Extrapolating RDD treatment effects far away from the cutoff (RDD estimates are strictly local to X=cX=c).
  • ✕
    Using high-degree global polynomials (e.g. 5th-order polynomial regression), which Gelman & Imbens (2019) proved produces noisy, biased boundary artifacts; local linear or local quadratic regression with triangular kernels must be used.
🎯 Interviewer Follow-up Inquiries
  • ?
    Why did Gelman & Imbens prove that high-order global polynomials should never be used in RDD?
  • ?
    How does the McCrary Density Test detect manipulation of the running variable?
⚡
Test muscle memory on this topic via 3-Layer Flashcards
Active Recall: Mentally formulate Prompt ➔ Math Derivation ➔ Benchmark Implementation in 3 minutes.
🎯 Drill This Card Now➔
📚

Associated Knowledge Base Guides & Mindmaps

← PreviousM1-087: Explain Methodologies for Measuring Long-Term Treatment Effects Beyond Short-Term Metrics📋 Back to BankNext →M1-089: Explain the Synthetic Control Method and Its Application in Geo-Testing and Policy Evaluation