The Problem with Trade Measurement in International Relations

Trade statistics are widely used in studies and policymaking focused on economic interdependence. Yet, researchers in International Relations (IR) have largely disregarded half the data available to study trade. Bilateral trade flows are usually recorded twice: by the sending economy as an export and by the receiving one as an import. These two values should match, but discrepancies between them tend to be large and pervasive. Most studies ignore this issue, which we label the “mirror problem” for short, by using only one entry. However, it is not self-evident which one is consistently most accurate. Hence, IR's reliance on error-prone trade statistics may be distorting its study of economic interdependence. This article explores this problem in three steps: first, we quantify the mirror problem in trade data. Second, we investigate the origins of the mirror problem, using statistical analyses, archival records, and interviews with statistical experts. Third, we illustrate the implications of the mirror problem through replications covering diverse topics in IR. We find that accounting for the mirror problem can variably strengthen, undermine, or overturn conclusions of such analyses. The findings underscore the severity of measurement problems in IR and suggest particular ways to address those problems