The Impact of the TPP on Trade Between Member Countries: A Text-As-Data Approach | Asian Development Bank

At the forefront of growth and development in Asia and the Pacific

The Impact of the TPP on Trade Between Member Countries: A Text-As-Data Approach

Publication | June 2017
The Impact of the TPP on Trade Between Member Countries: A Text-As-Data Approach

With World Trade Organization negotiations in deadlock, countries are increasingly turning to preferential trade agreements to integrate their economies into the global trading system.

We propose a new method to predict the impact of preferential trade agreements (PTAs) on trade and welfare, taking the Trans-Pacific Partnership (TPP) agreement as a case study. Relying on a novel dataset of treaty texts covering all trade agreements notified to the World Trade Organization, we first construct an indicator comparing existing PTAs to the TPP in terms of textual similarity. In a second step, we include this indicator into a gravity model of international trade in order to estimate the impact of different dimensions of PTA design on trade flows between TPP member countries. We derive predictions for two scenarios, the TPP with and without the United States, and compare the results to approaches using PTA incidence or depth indicators. At the aggregate level, our approach yields a slightly higher effect: In the scenario with the United States, total trade between TPP partners is predicted to increase by 9.4% ($162 billion in absolute terms), as opposed to the 2.6–6.4% ($45–$110 billion) obtained with conventional methods. Without the United States, the absolute increase would be much lower ($67 billion), but the percentage increase of trade among the other 11 members higher (16%). A closer look at the exports of individual Asian TPP members reveals that yet more fine-grained variables are necessary to obtain reliable predictions at a more disaggregate level. Text-as-data methods offer the possibility to generate such variables through, for example, chapter- and article-level similarity measures.

WORKING PAPER NO: 745