The finding arrives as the global trading system becomes increasingly dependent on overlapping networks of regional agreements. By 2023, more than 356 regional trade agreements were in force, while the World Trade Organization had received more than 585 notifications related to such arrangements. These agreements include free-trade areas, customs unions and other preferential frameworks that reduce tariffs or establish common rules among participating economies. Their effects are not uniform. The terms of each agreement, the sectors it covers, the products exchanged and the economic characteristics of its members can all shape the outcome. Some agreements may rapidly increase trade in manufactured goods, while others may affect agriculture, services or transportation only after years of adjustment.

The central statistical challenge is timing. Regional trade agreements rarely begin affecting commerce on a single, clean date. Negotiations may influence firms before a treaty formally enters into force, while tariff reductions and regulatory changes can be phased in over many years. Countries may also join different agreements at different moments, and some agreements eventually expire or are replaced. This staggered pattern makes it difficult to compare countries fairly. A simple before-and-after calculation can confuse the effect of an agreement with unrelated changes in economic growth, exchange rates, wars, financial crises or shifts in global demand. The economists therefore applied recent advances in difference-in-differences estimation, a family of methods used to compare changes over time between units exposed to a policy and units that are not.

Traditional difference-in-differences models often rely on two-way fixed effects. In a typical application, country-pair observations are adjusted for persistent differences between pairs and for shocks affecting all countries in a given year. The approach can be useful, but it becomes problematic when treatment effects vary across countries or over time. Under staggered adoption, a country pair treated early may inadvertently serve as a comparison group for a pair treated later, even though the earlier agreement is already influencing trade. If the effects grow or shrink over time, the resulting estimate can blend together distinct phases of the policy response and, in some circumstances, give misleading weights to different treatment cohorts.

Abreha and Robertson’s design explicitly accommodates these complications. Their study uses bilateral trade data covering more than 200 countries from 1962 through 2020, drawn from the Centre d’Études Prospectives et d’Informations Internationales, or CEPII. The long time span allows the researchers to observe trade relationships well before agreements are introduced and for many years afterward. The wide country coverage also extends the analysis beyond the narrower samples common in earlier work. Rather than treating an agreement’s influence as identical immediately after implementation and a decade later, the model permits dynamic effects that can vary with time since adoption. That distinction is crucial because companies may need years to redirect supply chains, establish distribution networks, meet new standards or develop relationships with customers abroad.

The resulting pattern is one of gradual expansion rather than an instant trade surge. The estimated effect grows during the years following implementation and reaches approximately 25 to 30 percent after ten years. This estimate is larger than the effect obtained from standard two-way fixed-effects models, indicating that conventional specifications may understate the longer-term consequences of trade agreements when they fail to represent staggered adoption and changing treatment effects. The result does not mean that every agreement produces the same increase, nor that the agreement alone explains every change in commerce. Instead, it represents an average estimated response across a broad and historically diverse set of bilateral relationships, under the assumptions built into the researchers’ econometric framework.

One especially important feature of the analysis is its treatment of anticipation. Trade policy can influence behavior before the legal implementation date. Firms may begin investing, signing contracts or shifting sourcing plans after an agreement is announced but before tariffs or other provisions formally change. Importers and exporters may also adjust inventories in expectation of future rules. If researchers mark the policy’s beginning only at the enforcement date, some of the genuine response may already have occurred and could be incorrectly assigned to the pre-treatment period. Abreha and Robertson find that accounting for anticipation changes the estimated timing, magnitude and statistical significance of the dynamic trade response. The result reinforces the idea that economic policies have calendars more complicated than a single legal start date.

The paper also expands the scope of the evidence beyond manufacturing, incorporates agreements that can expire and uses a lagged dependent-variable specification as an alternative way to address the persistence of trade relationships. Bilateral commerce is highly durable: countries that trade extensively in one period are likely to continue doing so in the next because firms have already built logistics networks, established suppliers and learned how to navigate one another’s markets. A lagged dependent variable captures part of that persistence, although it also introduces additional identifying assumptions that must be considered when interpreting the estimates. By examining non-absorbing treatments—agreements that do not remain permanently active—the study reflects the fact that international policy arrangements can end, change status or be superseded.

The award committee selected the article from eligible papers published in the journal during 2025, including papers rated in the top quarter or better during the review process. The committee was chaired by N. Gregory Mankiw of Harvard University and included members of the Atlantic Economic Journal’s editorial board and sponsors of the International Atlantic Economic Society Endowment Fund. The recognition highlights not only the headline estimate but also the methodological problem behind it: trade agreements are complex interventions whose effects may vary by country, sector and time. The authors acknowledge important boundaries to their conclusions. Their study does not examine cases in which agreements create entirely new trading relationships, and it does not separately account for individual provisions such as intellectual-property protections or regulatory rules. Those details could explain why some agreements generate stronger effects than others. Even so, the analysis offers a more technically realistic picture of how trade policy reshapes the global economy—slowly, unevenly and often beginning before the paperwork says it has started.