Econometric studies using the gravity model

The gravity model provides a useful framework for assessing the impact of policy variables on the behavior of bilateral flows between countries. Its name is derived from its passing similarity to Newtonian physics, in that flows between two countries increase in proportion to their economic mass (as measured by GDP) and are constrained by the friction between them (due to trade and other costs, which is proxied by distance). It is also common to use so-called dummy variables to capture geographical effects (such as whether the two countries share a border, or if a country has access to the sea), cultural and historical similarities (such as if two countries share a language or were linked by past colonial ties), and regional integration (such as belonging to a free trade agreement or sharing a common currency).

A disadvantage of using dummy variables is that they may capture the impact of a range of other effects that occurred during the same time period as the RTA (Regional Trade Agreements). For example, most applications do not distinguish the extent of multilateral trade liberalization. Ideally, specific trade policy variables would be included in the estimating equations, such the level of multilateral and preferential tariffs. However, the complexity of preferential trade arrangements precludes such an approach. A notable exception is the study done by Estevadeordal and Robertson (2004), who included a measure of preferential tariffs in their analysis of the impacts of RTAs on regional trade in Latin America.

Although widely used because of its empirical success, the gravity model had lacked rigorous theoretical underpinnings and was long criticized for being an ad hoc model. Recent theoretically grounded gravity equations are derived from models with strong constraints on preferences and technology, which undermines a straightforward interpretation of some of the estimated coefficients.

Another weakness of many applications of the gravity model is the proxying of trade costs by distance, and the implicit assumption that cargoes traveling 1,000 miles in Africa face exactly the same trade costs as similar cargoes traveling 1000 miles in, say, Europe.