Asymptotic Results for Dyadic Data

Xavier D'Haultfoeuille (CREST - ENSAE)

March 12, 2019, 15:30–17:00

Room MS 001

Econometrics and Empirical Economics Seminar


Dyadic data are samples indexed by two units from the same population. Examples include trade flows between countries, relationships in a network or results of sports matches. Such data exhibit particular dependence patterns, affecting the statistical properties of the estimators. We establish in this context simple and uniform law of large numbers and central limit theorems, implying the asymptotic normality of a large class of linear and nonlinear estimators. Next, we derive a new consistent variance estimator, and show that it can be computed easily with standard statistical software. We also show the general consistency of a particular bootstrap scheme. Monte Carlo simulations suggest that the two methods work well even with few observations. Finally, we show that accounting for dependence patterns in trade data has potentially large effects on standard errors. Joint with Laurent Davezies et Yannick Guyonvarch (CREST)