Each month the U.S. Census Bureau receives from the Customs and Border Protection Service documents describing all U.S. exports over $2,500. Most exporters now file the required documents via the Automated Export System. Because of filing options, for the majority of exporters, we have no information on "low-valued" exports (LVEs), those valued at or below the exemption level of $2,500; for the remainder, we have records for all LVEs. As exporters are not required to report their LVEs, the Census Bureau estimates the low-valued component for any country by multiplying its total of exports valued over $2,500 by a pre-determined factor. This factor, specific to the country of export, was determined in the late 1980's based on data collected up to that point in time. However, while export patterns--commodities, modes of transport, quantities, and values--have changed, the factors remain the same. Some experts believe that the current system may underestimate LVEs.

In this paper, we look back at research into several approaches to estimate LVEs. The goal was to develop robust procedures that accurately represent what is currently being exported, and can adapt over time with minimal review to reflect changes in export patterns. After reviewing background on the collection of export data and the estimation of LVEs, we briefly describe several methods for estimating LVEs. In all of these strategies, we start by combining exports into groups determined by common features, such as the export's commodity or its mode of transport. We review methods based on (1) data from the small portion of exporters who report their LVEs, (2) export data to Canada and from courier shippers, (3) models that try to predict how an exporter will file, and (4) models that hypothesize that some shippers split their exports into smaller pieces to avoid filing export documents. The research reveals promising aspects of the approaches as well as serious flaws.