M1-059M1: Mathematics & Statistics FundamentalsCausal InferenceMedium
⚡Flashcard Drill➔Mastery:

Explain Difference-in-Differences (DID) and Its Foundational Parallel Trends 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: M1-060 →
📐 Mathematical Definition
τ^DID=(YˉT,post−YˉT,pre)−(YˉC,post−YˉC,pre)\hat\tau_{DID}=(\bar Y_{T,post}-\bar Y_{T,pre})-(\bar Y_{C,post}-\bar Y_{C,pre})
⚡ One-Sentence Core Answer
Difference-in-Differences estimates causal effects by comparing the pre-post changes in a treatment group against changes in an untreated control group over the same time horizon: ΔDID=(YˉT,2−YˉT,1)−(YˉC,2−YˉC,1)\Delta_{\text{DID}} = (\bar{Y}_{T, 2} - \bar{Y}_{T, 1}) - (\bar{Y}_{C, 2} - \bar{Y}_{C, 1}).

📌 Key Takeaways

  • •
    Core Mechanism: First difference subtracts baseline time-invariant user heterogeneity; second difference subtracts macroeconomic time trends.
  • •
    Parallel Trends Assumption: In the absence of treatment, the average outcome of treatment and control groups would have followed parallel trajectories over time.
  • •
    Regression Formulation: Yit=β0+β1Treatedi+β2Postt+δ(Treatedi×Postt)+ϵitY_{it} = \beta_0 + \beta_1 \text{Treated}_i + \beta_2 \text{Post}_t + \delta (\text{Treated}_i \times \text{Post}_t) + \epsilon_{it}, where δ\delta is the causal DID effect.

📐 Mathematical Derivations

Expectations table across groups and time periods: (1) Control Pre: E[YC,1]=β0E[Y_{C, 1}] = \beta_0. (2) Control Post: E[YC,2]=β0+β2E[Y_{C, 2}] = \beta_0 + \beta_2. Change in control: ΔC=β2\Delta_C = \beta_2 (time trend). (3) Treatment Pre: E[YT,1]=β0+β1E[Y_{T, 1}] = \beta_0 + \beta_1. (4) Treatment Post: E[YT,2]=β0+β1+β2+δE[Y_{T, 2}] = \beta_0 + \beta_1 + \beta_2 + \delta. Change in treatment: ΔT=β2+δ\Delta_T = \beta_2 + \delta. Taking the difference of differences: ΔDID=ΔT−ΔC=(β2+δ)−β2=δ\Delta_{\text{DID}} = \Delta_T - \Delta_C = (\beta_2 + \delta) - \beta_2 = \delta. The estimate isolates the true causal impact δ\delta while controlling for baseline group differences β1\beta_1 and macro shocks β2\beta_2.

🏭 Production Trade-offs

DID is widely used for policy evaluation, natural experiments, and geo-level marketing campaigns (e.g. Launching TV ads in California while using Texas as control). The Parallel Trends assumption cannot be directly proven for the post-treatment period; scientists validate it empirically using Event Study plots (lead-lag regressions showing estimated coefficients δt=0\delta_t = 0 for all pre-treatment periods t<0t < 0).
⚠️ Common Interview Pitfalls
  • ✕
    Applying standard DID when parallel trends fail pre-treatment (pre-existing divergence violates the identifying assumption).
  • ✕
    Staggered adoption bias: When different units receive treatment at different times, standard two-way fixed effects regressions produce biased estimates (addressed by Callaway-Sant'Anna estimators).
🎯 Interviewer Follow-up Inquiries
  • ?
    How do event study regressions formally test the pre-treatment parallel trends assumption?
  • ?
    Why does standard two-way fixed effects (TWFE) fail in staggered rollout designs with heterogeneous treatment effects over time?
⚡
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

← PreviousM1-058: Explain Propensity Score Matching (PSM) and Inverse Probability Weighting (IPW) for Observational Causal Inference📋 Back to BankNext →M1-060: Explain Instrumental Variables (IV), Two-Stage Least Squares (2SLS), and the Local Average Treatment Effect (LATE)