Gravity models, PPML estimation and the bias of the robust standard errors

ABSTRACT

In gravity models with exporter and importer dummies, the robust standard errors of the slope parameters tend to be severely downward biased when estimated by PPML. The coverage rate of confidence intervals of the estimated slope parameters may thus be much too small in cross-sections of the size typically used in empirical research.

KEYWORDS:

• Poisson pseudo maximum likelihood estimation
• heteroskedasticity-robust inference
• gravity equation

JEL CLASSIFICATION:

• F10
• F15
• C13
• C50

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

¹ An Appendix provided as a supplement provides details on all derivations of the paper. https://eeecon.uibk.ac.at/~pfaffermayr/Pfaffermayr_PPML_Estimation_and_the_Bias_of_the_Robust_Standard_Errors_Technical_Appendix_AEL_2019.pdf .

² The star indicates that these matrices are evaluated at parameter values lying in between the estimated and true one's element by element.

³ Results for ${\alpha }_2$ are rather similar and available upon request.