Testable implications of affine term structure models

Affine term structure models have been used to address a wide range of questions in macroeconomics and finance. This paper investigates a number of their testable implications which have not previously been explored. We show that the assumption that certain specified yields are priced without error is testable, and find that the implied measurement or specification error exhibits serial correlation in all of the possible formulations investigated here. We further find that the predictions of these models for the average levels of different interest rates are inconsistent with the observed data, and propose a more general specification that is not rejected by the data.