Trade Potentials in Gravity Panel Data Models

Abstract

The paper shows how - using as an example the trade flows between eleven European countries and 31 OECD `reporting' countries - the result of a gravity model, in terms of potential trade, changes substantially when country heterogeneity and dynamics are taken into account. Comparing the in-sample trade potential index derived from various estimators yields three different results: (a) the average trade potential index poorly represents the distribution of yearly trade potentials; (b) the index converges towards the demarcation value corresponding to the equality between observed and predicted trade flows when country heterogeneity and dynamics are taken into account; (c) the sign of its yearly average is not the right statistic with which to determine the (in)existence of unrealized trade potentials. Finally, the index derived from a dynamic specification with multilateral fixed-effects is better able to reflect the role played by the time-variant country-specific unobservable element associated with the possible presence of positive or negative trade potentials.

Trade performances and technology in the enlarged European Union

“…Fromapurelyeconometricpointofview,wehaveadoptedatheoreticallybasedgravitymodelà la Anderson and van Wincoop (2003) by including countries' fixed effects, in a slightly different way from suggestions by Baldwin and Taglioni (2006) and Baier and Bergstrand (2007), because thenumberofobservationsfortheCEECsampleprovidesinsufficientdegreesoffreedomforthe estimationof2NT(N=countries,T=years)dummiesforunidirectionaltradeinaSystemGMM. Hence, we have adopted the approach suggested by De Benedictis et al (2005) by including exportingandimportingcountries'fixedeffects(α i andδ j respectively),andacountry-pairtimevarianttrendvariablecalculatedastheinteractionbetweentemporaltrendsandfixedeffectsfor countrypairs(trend ijt ). WehavealsoadaptedtheHMRtwo-stageprocedureinapanelsetting,byalsoincludingatimevariant control variable for firms' heterogeneity, using a first-stage probit selection equation (fhet ijt )andthestandardinverseMillsratio(mills ijt ).Inordertoincludespecifictransactioncosts related to firms' heterogeneity that are not included among the regressors of the second-stage estimation, we have used a standard dummy variable for the existence of a common language, which strongly affects the formation of trading relationships.…”

“…CEEC countries joining the EU should have benefited from the European tradeintegrationprocess;thusthevariableassumesvalue0uptothemomentwhenthecountry entered the EU, and value 1 thereafter. In particular, the variable ENL ijt embodies the so-called 'announcement effect' of the entrance of the eight new member countries into the EU (De Benedictis et al, 2005;Paas, 2001, Sapir, 2001, corresponding to the date of the European Council meeting of Laeken in December 2001. Hence, the dummy assumes the value of 1 as of 2002forallEUcountry-pairsinvolvedintheenlargementprocess.…”

Essays on Transport Economics

“…The GMM model is one of the methodologies most used to estimate dynamic gravity models (Jung, 2009). However, as argued in Blundell & Bond (1998), when the data are highly persistent -as in the case of bilateral trade flowsthis procedure can be improved through the estimation by System GMM (De Benedictis & Vicarelli, 2005). It has been shown that when the lagged values of the series approach a unit root, these instruments contain little information about the endogenous variables in differences.…”

Preference erosion and the developing countries exports to the EU: a dynamic panel gravity approach

“…In the case of the gravity model, first-differencing the equation removes the fixed effect but also the time invariant regressors from the specification and, when the regressors are of interest, the resulting loss of information may be a serious drawback (De Benedictis, Vicarelli 2005). Moreover, with highly persistent data and short panel (along the time dimension), as in the case of bilateral exports flows and of our dataset specifically, the GMM estimator may suffer marked small sample bias due to weak instruments (Blundell, Bond, 1998).…”