Estimating gravity equations: to log or not to log?

Arbeitspapier

This study compares two alternative approaches to estimate parameters in gravity equations. We compare the traditional OLS approach applied to the log-linear form of the gravity model with the Poisson Quasi Maximum Likelihood (PQML) estimation procedure applied to the non-linear multiplicative specification of the gravity model. We use the trade flows for all products, for all manufacturing products as well as for manufacturing products broken down by three-digit ISIC Rev.2 categories. We base our conclusions on the generalised gravity model of Bergstrand (1989) that allows us to investigate differences in factor-proportions and home-market effects at the industry level. In addition, we compare the effects of other explanatory variables such as exporter and importer total income, distance, preferential trade agreements, common border, historical ties, and common language on the volume of trade. Our study provides comprehensive evidence on likely qualitative and/or quantitative differences in the values of estimated coefficients as a result of application of an alternative estimation method. Our main conclusion is that both estimation results as well as results of the regression misspecification tests provide supporting evidence for the PQML estimation approach over the OLS estimation method.

Object Details

• Language: Englisch

References and Relationships

• Bibliographic citation: Series: DIW Discussion Papers ; No. 739

Classification and Topics

• Classification: Wirtschaft ; Models of Trade with Imperfect Competition and Scale Economies; Fragmentation
• Subject: generalised gravity equation ; Poisson regression ; Gravitationsmodell ; Internationale Wirtschaftsbeziehungen ; Schätztheorie ; Statistische Verteilung ; Maximum-Likelihood-Methode ; Schätzung ; OECD-Staaten

Contributors, Places and Time

• Event: Geistige Schöpfung
• (who): Siliverstovs, Boriss ; Schumacher, Dieter
• Event: Veröffentlichung
• (who): Deutsches Institut für Wirtschaftsforschung (DIW)
• (where): Berlin
• (when): 2007

Further information

• Handle: http://hdl.handle.net/10419/27263
• Last update: 10.03.2025, 11:41 AM CET