Course contentsShow
← Data Science
Lesson 1453 of 2,14536. Causal Inference BasicsFree lesson

The Parallel Trends Assumption

Understand why parallel trends is crucial and how to assess it using pre-treatment data visualization.

The Parallel Trends Assumption

What you'll learn: You'll understand why parallel trends is the critical assumption underlying Difference-in-Differences analysis and how to visually assess it before treatment occurs.

Why Parallel Trends Matters

In Difference-in-Differences (DiD), you compare how a treated group changes over time versus how a control group changes. But this only works if both groups would have followed the same trend in the absence of treatment.

Think of it like two runners on parallel tracks. If one runner suddenly speeds up after drinking a new energy drink, you can attribute the speed boost to the drink—but only if both runners were maintaining the same pace before. If one was already accelerating naturally, you can't tell whether the drink worked or if that runner would have sped up anyway.

The Core Assumption

Parallel trends means that, without the intervention, the treated and control groups would have experienced the same change over time. Not the same levels—they can start at different places—but the same trajectory.

If this assumption fails, your DiD estimate is biased. You might attribute a pre-existing divergence (or convergence) to your treatment when it was already happening.

Assessing with Pre-Treatment Data

The key test: plot both groups' outcomes over multiple pre-treatment periods. The lines should move in parallel. Look for:

  • Similar slopes before treatment
  • No diverging or converging patterns
  • Consistent gaps between groups

If trends diverge before treatment, parallel trends is violated, and DiD is inappropriate without adjustment.

Key Takeaway: Parallel trends requires that treated and control groups would have followed the same trajectory absent treatment; always visualize pre-treatment periods to check this assumption before trusting your DiD estimate.