The parallel trends assumption is the identifying condition in a [[Design Pattern III - Difference-in-Differences (DID)|difference-in-differences]] design: absent treatment, the treated and control groups would have changed by the same amount on average over the study window: $\mathbb{E}\big[Y_{\text{post}}(0) - Y_{\text{pre}}(0) \mid G = 1\big] = \mathbb{E}\big[Y_{\text{post}}(0) - Y_{\text{pre}}(0) \mid G = 0\big]$ It concerns *trends*, not baseline *levels*, and is fundamentally untestable. Similar pre-trends in the pre-treatment period are often taken as evidence that parallel trends may hold. > [!info]- Last updated: June 11, 2026