Stabsseminar ved Richard Harris

Dr. Richard Harris fra University of Exeter holder stabsseminar ved Institutt for foretaksøkonomi på fredag.

06.12.2000 - Elin F. Styve

Tema er Finite Sample Biases in Tests of the Expectations Hypothesis of the Term Structure of Interest Rates

Tid og sted:
Fredag 8. desember kl. 1215-1330 i Karl Borchs aud.

Abstract
Bekaert, Hodrick and Marshall (1997, BHM) demonstrate that if agents assume that the short yield follows a highly persistent but stationary first order autoregressive (AR) process, empirical tests of the expectations hypothesis are substantially biased in favour of accepting the null hypothesis even in relatively large samples, thus strengthening the reported rejection of the expectations hypothesis. BHM derive analytical results for the bias, and support these with evidence from simulation experiments. However, noting that the simple AR process is unrealistically simple, BHM extend their simulation experiments using a vector autoregression (VAR) to generate forecasts for the short yield.
In this paper, we demonstrate that in going from the AR model to a more general model such as the VAR, in which agents are assumed to have information about next period's short yield beyond that contained in the current short yield, a further source of bias is introduced. This bias serves to offset the bias identified by BHM, making the rejection of the expectations hypothesis less likely than their results would imply. Moreover, we show that if agents believe that the short rate follows a unit root, the bias identified by BHM disappears, leaving just a single bias that serves to weaken the rejection of the expectations hypothesis, not strengthen it as BHM conclude.

KEYWORDS: Expectations Hypothesis of the Term Structure of Interest
Rates; Finite Sample Bias; Vector Autoregression; Monte Carlo Simulation.

Kontaktperoner er Steinar Ekern (9279) og Hans K. Hvide (9283)
Se også www.nhh.no/for/seminars/fall2000/.