2.1 Review of Empirical Studies
A common way to assess the effects of trade preferences has been to estimate gravity models, incorporating dummy variables for various preference schemes. An early such study is Sapir (1981) which uses yearly cross-sectional regressions for 1967–1978 to estimate the effect of the EU’s GSP regime. He finds a significant and positive effect for 1973 and 1974. Using the estimated coefficients to calculate gross trade creation (GTC), the estimations suggest that the GSP created 91-93% extra trade.2
In another early gravity study, Oguledo and MacPhee (1994) find positive and statistically significant effects for GSP, Mediterranean and Lomé preferences for the year 1976. The Lomé effect is larger than the Mediterranean effect, which in turn exceeds that of the GSP. The corresponding gross trade creation is very large indeed, with the value for the Lomé preferences actually approaching 2000%. In a similar study, Nilsson (2002) estimates the gravity model on three-year-averages for 1973–1992. Nilsson finds a significant and positive effect for most though not all years for GSP and Lomé, and that the effect of the latter is larger. The Mediterranean preferences are mostly insignificant. Again, the estimated gross trade creation is quite sizable, with figures for the Lomé preferences at most exceeding 400%.
Sapir (1981), Oguledo and MacPhee (1994) and Nilsson (2002) made important contributions by being able to show that – contrary to the bleak view taken by many commentators regarding the effectiveness of preferences – when controlling for other important factors that may work against developing countries’ export prospects, preferences actually have positive effects. On the other hand, the cross-sectional methods used in these papers made it impossible to control for unobserved heterogeneity between countries, and the implied omitted variable bias may be an explanation for the remarkably large effects that were found. A more recent paper which takes this into account is Péridy (2005), estimating the effect of Mediterranean preferences for 1975–2001 in a sample of OECD and some developing countries. Péridy uses various panel data methods that can control for time- and country unobserved heterogeneity, and finds a Mediterranean dummy that is highly significant in all cases, and with similar magnitudes in all specifications. Using the coefficient from the fixed effects specification to calculate gross trade creation, this would be about 38%, which intuitively seems like a more reasonable magnitude of the effect than what is found in previous studies.
While Péridy (2005) focuses on the Mediterranean preferences, Persson and Wilhelmsson (2007) use data for a very long time period, 1960-2002, to estimate the effects of all types of non-reciprocal trade preferences that have been used by the EU. Estimating a gravity model with fixed effects to capture country-pair and time specific unobserved heterogeneity, and bilateral time trends to capture changes over time in the heterogeneity across country-pairs, significant and positive effects are found for most, though not all versions of trade preferences. The magnitude of the estimated gross trade creation is again much more modest than in the earlier studies, with for example the GSP regime being estimated to have increased trade by about 4%, and the Lomé convention by about 30%.3