45th Congress of the European Regional Science Association: "Land Use and Water Management in a Sustainable Network Society", 23-27 August 2005, Amsterdam, The Netherlands
Trade costs remain an important barrier to international trade in today's globalizing economy. Despite the popular discussion on the "death of distance”, distance is still an important source of trade costs and continues to have an irrevocable impact on the patterns of international trade. The literature identifies various factors that can explain the importance of geographical proximity for bilateral trade. First, transport costs and costs of timeliness increase with distance. Moreover, psychic distance increases as well. Because of cultural unfamiliarity and information costs, traders have less knowledge of distant markets. Empirical estimates of the distance effect in trade abound. The evidence indicates that distance still matters for trade. However, differences in estimated effects across the literature make generalizations about the distance effect and its development over time more difficult. This paper performs a meta-analysis of existing empirical studies of bilateral trade, in order to contribute to our understanding of distance decay in trade. Meta-analysis is a statistical analysis of a set of existing empirical results in a specific research area, in order to integrate the findings. It constitutes a quantitative survey of the literature that explicitly addresses the causes of cross-study variation in empirical outcomes. To perform the meta-analysis, a sample of gravity studies was constructed that is as representative as possible. For this purpose, a literature search has been conducted on the Internet, using the Econlit database. Using the search string "trade and/or distance, and gravity, in all fields”, a list of 214 applicable studies has been identified. From this list, 30 studies were randomly selected into the meta-analysis sample. The paper focuses on two key issues. First, it investigates cross-estimate variation in the distance effect according to differences in, e.g., time period concerned, data type used, or empirical specification and estimation method used. Then, we analyse whether the impact of distance has declined over time.