Dynamic Risk Profile of the U.S. Term Structure

Abstract:     
Unlike theoretical interest rate time series data, empirical interest time series data indicate obvious deviations from the conventional assumptions of theoretical term structure models. A careful examination of interest rate time series data from different U.S. Treasury maturities by wavelet multiresolution analysis (MRA) suggests that the first differences of the term structure of interest rate series are periodic or, at least, cyclic, non-stationary and dependent, i.e., they exhibit long memory. In addition, each time series data from a particular maturity has its own unique Hurst exponent and accordingly supports the Market Segmentation theory. The findings also imply that simple affine term structure models are insufficient to describe the complete dynamics of the various term structure diffusion processes and call for more intensive research that might provide better, most likely fractal or nonlinear, term structure models for each maturity in a system-like fashion. If this is correct, empirical term structure models may describe even chaotic, i.e., diffusion processes with non-unique dynamic equilibria, although it will require research of high-frequency (intra-day) data to establish that fact.

Keywords: Wavelet, interest rates, Hurst exponent, term structure, yield curve

JEL Classification: E43

Accepted Paper Series