Evidence for state transition and altered serial codependence in US$ interest rates

This paper studies the codependence among, and drawdown and drawup properties of, US$ interest rates. The problem is attacked from the angle of regime switching. Different regimes are identified using the Hidden Markov Models (HMMs). The statistical properties in each state are examined separately and reconciled to form a coherent picture. We found that high fractions of reversals exist in the normal state and that consecutive bursts exist in the excited state. In large drawdowns and drawups (draws), long draws tend to be 'democratic', short draws tend to be 'oligarchic' and medium-size draws stay in either 'democratic' or 'oligarchic' mode, while conditionally independent draws are rarely found. We also investigated the distributions of draws. We found that HMMs recover the draw properties well and that the overall distribution of draws is an informationally-rich indicator about the correlation regime(s) in the various Markov states.