Regression Discontinuity in Time (RDiT)

Regression discontinuity in time (RDiT) is a quasi-experimental design that uses time as the running variable and estimates an intervention's effect as the jump in the outcome at the date the intervention began.

시간 회귀불연속설계(RDiT)는 시간을 배정변수로 사용하여 개입이 시작된 시점에서 결과변수의 단절적 변화(점프)로 개입효과를 추정하는 준실험 설계이다.

How do I estimate the effect of a sudden policy change using a single time series?

단일 시계열에서 갑작스러운 정책 변화의 효과를 어떻게 추정하나요?

Regression discontinuity in time fits flexible trends in the outcome on each side of the implementation date and attributes a discontinuity at that date to the intervention. It fits high-frequency data, such as daily pollution or traffic counts, when a policy started abruptly and no untreated comparison group exists.

When to use Regression Discontinuity in Time (RDiT)

• A policy or event started abruptly on a known date
• High-frequency outcome data such as daily or hourly measurements
• No untreated comparison group is available

When to use something else

• A comparable untreated group exists: use difference in differences
• The effect is expected to build gradually rather than jump: use interrupted time series with slope changes

Assumptions and requirements

• No other change occurred at the same date
• Smooth trends in the outcome around the cutoff absent the intervention
• Bandwidth and polynomial order chosen carefully, with sensitivity checks
• Serial correlation addressed in standard errors

What the analysis reports

• Estimated discontinuity at the cutoff with robust standard errors
• Plot of the outcome over time with fitted lines on each side
• Sensitivity to bandwidth and polynomial order
• Placebo tests at false cutoff dates

How to report it

Regression discontinuity in time estimates showed a [value] change in [outcome] at the [policy] start date (SE = [value], p = [value]), robust to alternative bandwidths.

Related methods

• Regression Discontinuity Design
• Interrupted Time Series Analysis
• Difference in Differences (DiD)

Run Regression Discontinuity in Time (RDiT) on Quorant