Loading [MathJax]/jax/output/HTML-CSS/jax.js
  • 1 Introduction
    • 1.1 Introduction
    • 1.2 DID in Figure (on screen)
    • 1.3 Plan of the Lecture
    • 1.4 Reference
  • 2 Framework
    • 2.1 Framework
    • 2.2 Identification
    • 2.3 Parallel Trend Assumption
  • 3 Estimation
    • 3.1 Estimation Approach
    • 3.2 Plug-in Estimator
    • 3.3 Example: Card and Kruger (1994)
    • 3.4 Regression Estimators
    • 3.5 Regression Estimators with FEs
  • 4 Parallel Trend
    • 4.1 Discussions on Parallel Trend
    • 4.2 Diagnostics for Parallel Trends: Pre-treatment trends
    • 4.3 Example (Fig 5.2 from Mastering Metrics)
    • 4.4 Unit-Specific Time Trends
    • 4.5 Other Diagnostics: Placebo test
  • 5 Research Strategy
    • 5.1 Research Strategy using DID

1 Introduction

1.1 Introduction

  • Difference-in-differences (DID)
    • Exploit the panel data structure to estimate the causal effect.
  • Consider that
    • Treatment and control group comparison: selection bias
    • Before v.s. After comparison: time trend
  • DID combines those two comparisons to draw causal conclusion.

1.2 DID in Figure (on screen)

1.3 Plan of the Lecture

  • Formal Framework
  • Implementation in a regression framework
  • Parallel Trend Assumption

1.4 Reference

2 Framework

2.1 Framework

  • Consider two periods: t=1,2. Treatment implemented at t=2.
  • Yit: observed outcome for person i in period t
  • Gi: dummy for treatment group
  • Dit: treatment status
    • Dit=1 if t=2 and Gi=1
  • potential outcomes
    • Yit(1): outcome for i when she is treated
    • Yit(0): outcome for i when she is not treated
  • With this, we can write Yit=DitYit(1)+(1Dit)Yit(0)

2.2 Identification

  • Goal: ATT at t=2 E[Yi2(1)Yi2(0)|Gi=1]=E[Yi2(1)|Gi=1]E[Yi2(0)|Gi=1]

  • What we observe

    Pre-period (t=1) Post (t=2)
    Treatment (Gi=1) E[Yi1(0)|Gi=1] E[Yi2(1)|Gi=1]
    Control (Gi=0) E[Yi1(0)|Gi=0] E[Yi2(0)|Gi=0]
  • Under what assumptions can we the ATT?

    • Simple comparison if E[Yi2(0)|Gi=1]=E[Yi2(0)|Gi=0].
    • Before-after comparison if E[Yi2(0)|Gi=1]=E[Yi1(0)|Gi=1].
    • Other (more reasonable) assumption?

2.3 Parallel Trend Assumption

  • Assumption: E[Yi2(0)Yi1(0)|Gi=0]=E[Yi2(0)Yi1(0)|Gi=1]
    • Change in the outcome without treatment is the same across two groups.
  • Then, E[Yi2(1)Yi2(0)|Gi=1]ATT=E[Yi2(1)|Gi=1]E[Yi2(0)|Gi=1]=E[Yi2(1)|Gi=1]E[Yi1(0)|Gi=1](E[Yi2(0)|Gi=1]E[Yi1(0)|Gi=1])=E[Yi2(0)Yi1(0)|Gi=0] (pararell trend)

  • Thus, ATT=E[Yi2(1)Yi1(0)|Gi=1]E[Yi2(0)Yi1(0)|Gi=0] which is why this is called “difference-in-differences”.

3 Estimation

3.1 Estimation Approach

  1. Plug-in estimator

  2. Regression estimators

3.2 Plug-in Estimator

  • Remember that the ATT is ATT=E[Yi2(1)Yi1(0)|Gi=1]E[Yi2(0)Yi1(0)|Gi=0]

  • Replace them with the sample average. ^ATT={ˉy(t=2,G=1)ˉy(t=1,G=1)}{ˉy(t=2,G=0)ˉy(t=1,G=0)} where ˉy(t,G) is the sample average for group G in period t .

  • Easy to make a 2×2 table!

3.3 Example: Card and Kruger (1994)

image

3.4 Regression Estimators

  • Run the following regression yit=α0+α1Gi+α2Tt+α3Dit+βXit+ϵit

    • Gi: dummy for treatment group

    • Tt:dummy for treatment period

    • Dit=Gi×Tt. α3 captures the ATT.

  • Regression framework can incorporate covariates Xit, which is important to control for observed confounding factors.

3.5 Regression Estimators with FEs

  • With panel data yit=αDit+βXit+ϵi+ϵt+ϵit where ϵi is individual FE and ϵt is time FE.
  • Do not forget to use the cluster-robust standard errors!
    • See Bertrand, Duflo, and Mullainathan (2004, QJE) for the standard error issues.

4 Parallel Trend

4.1 Discussions on Parallel Trend

  • Parallel trend assumption can be violated in various situations.
  • Most critical issue: Treatment may depend on time-varying factors
    • DID can only deal with time-invariant factors.
  • Self-selection: participants in worker training programs experience a decrease in earnings before they enter the program
  • Targeting: policies may be targeted at units that are currently performing best (or worst).

4.3 Example (Fig 5.2 from Mastering Metrics)

image


4.5 Other Diagnostics: Placebo test

  • Placebo test using other period as treatment period. yit=τγτGi×It,τ+μi+νt+ϵit
    • The estimates of γτ should be close to zero up to the beggining of treatment (Fig 5.2.4 of Angrist and Pischke)
  • Placebo test using different dependent variable which should not be affected by the policy.

5 Research Strategy

5.1 Research Strategy using DID

  • Ishise et al (2019)
    1. How to find a research question
    2. What outcome dataset to look for
    3. What policy to look for (except for example 1 and 2).