Uncovering Aggregation Bias in Trade Policy Analysis

Professor Marcelo Olarreaga, along with Jean‑Marc Solleder, and Fulvio Silvy have recently published an article in the Journal of International Economics that revisits how governments set tariffs and reassesses the balance between public interest and lobbying.

The authors show that earlier research relied on highly aggregated data, whereas tariff decisions are made at a much more detailed product level. This mismatch created aggregation bias, leading to an overestimation of the importance given to social welfare.

Using machine‑learning techniques to reconstruct detailed production data for 142 countries, the study finds that the actual weight governments place on social welfare is 77% lower than previously estimated.

These results indicate that lobbying plays a far more significant role in trade protection than earlier work suggested, highlighting the importance of using finely disaggregated data to understand trade‑policy decisions.

ABSTRACT

Estimates of Grossman and Helpman (1994) Protection For Sale (PFS) model yield unrealistically high estimates of the weight governments put on social welfare relative to lobbying contributions. Estimates of the former are often close to 1. We argue this is due to the level of aggregation at which the model is estimated. While protection is determined at the tariff line level, production data are only available at the industry level. Using a new production dataset at the tariff level, our estimates show that the average weight on social welfare in a sample of 142 countries declines by 77 percent.

The study is available here: Protection for sale without aggregation bias

> Click here to view the GSEM faculty’s publications in top-tier journals.