An Empirical Analysis of Affine Term Structure Models Using the Generalized Method of Moments

In this paper we formulate two tractable two-factor affine term structure models, imposing weak assumptions on the distributions of the measurement errors involved in the different yields. Exploiting the implied moment conditions, the models are estimated by the generalized method of moments using weekly term structure data for Germany, Japan, the UK and the USA. Despite our relatively weak assumptions, for each of these countries the overidentifying restrictions tests indicate that the estimated two-facotr models should be rejected. Apparently, the fact that many affine term structure models are rejected empirically, is unlikely to be due to the assumptions about the joint distribution of the measurement errors, but more likely to the lack of flexibility to explain certain aspects of the term structure.