I dont know the answer but this is what AI said.
What you are describing is usually called a “stacked event study” or “stacked difference-in-differences,” not simply a trimmed staggered dataset.
The basic idea is:
1. For each treatment event, create an event-specific subpanel around the treatment date.
2. Define event time relative to that event, for example, -12 months to +12 months.
3. Include treated units and appropriate control units for that event window.
4. Stack all the event-specific subpanels into one dataset.
5. Estimate the model with event-time indicators, treatment/cohort or stack fixed effects, and appropriate clustering.
So your Korea-Brazil 2001 window and US-Canada 2006 window example is conceptually close to a stacked event-study design.
The important caution is that the control group matters. You usually do not just trim the global dataset around each event and combine everything mechanically. For each RTA event, you need to decide which dyads are valid controls in that event window: never-treated dyads, not-yet-treated dyads, or dyads not affected by any overlapping RTA or related trade-policy shock. If the same country pair appears in multiple stacks, you also need to handle duplicated observations correctly.
Papers/methods to look at:
* Cengiz, Dube, Lindner, and Zipperer, “The Effect of Minimum Wages on Low-Wage Jobs,” QJE 2019. This is a well-known applied example using a stacked/localized event-study structure.
* Deshpande and Li, “Who Is Screened Out? Application Costs and the Targeting of Disability Programs,” AEJ: Economic Policy 2019. Another applied example often cited in connection with stacked DID/event-study designs.
* Wing, Freedman, and Hollingsworth, “Stacked Difference-in-Differences.” This is more methodological and directly addresses what stacked DID estimates and how to weight stacked samples.
* Sun and Abraham, “Estimating Dynamic Treatment Effects in Event Studies with Heterogeneous Treatment Effects.” This is not exactly the same data compilation method, but it is central for staggered event-study problems.
* Callaway and Sant’Anna, “Difference-in-Differences with Multiple Time Periods.” Also not necessarily a stacked-sample paper, but it is one of the standard references for staggered treatment timing.
For trade specifically, look at:
* Nagengast and Yotov, “Staggered Difference-in-Differences in Gravity Settings: Revisiting the Effects of Trade Agreements.” That paper is directly about RTAs/trade agreements and staggered DID in a structural gravity setting.
My guess is that, for an RTA project, reviewers may prefer that you frame this as a stacked event-study or stacked DID design rather than saying you are trimming and combining panels. The trimming is just the data-construction step. The econometric design is the important part.
One more point: if this is trade-flow data, you probably need to think carefully about gravity-model fixed effects. In many trade applications, country-pair fixed effects plus exporter-time and importer-time fixed effects are important because trade flows are affected by multilateral resistance, global shocks, exporter conditions, and importer conditions. A simple stacked DID without those controls may be criticized.
So the short answer is: yes, this approach exists. Search for “stacked DID,” “stacked event study,” “cohort-specific event study,” and “staggered DID gravity trade agreements.” But make sure your stacked windows define valid controls and that your inference accounts for duplicated observations and treatment timing.
