Our paper explores the dynamic relationship between the call rate and the term structure of interest rates. Under inflation targeting regime, the monetary authority intends to change the shape of the yield curve by adjustingshort-term interests rate such as the call rate in Korea, the movement of which, in turn, is anticipated to influence the yields of bonds with varying maturities. Employing wavelet transformation known for its flexibility and effectiveness in dealing with time series data, we investigate the relationship between the movement of call rate and that of the yield curve to fill the gap untouched by the current literature.
We choose an orthogonalized wavelet filter, Daubechies wavelets, and apply it to the time series of various interest rates with different maturities to obtain five wavelet details and a wavelet approximation, whose sum, by definition, equals to the corresponding original time series. Jointly with these properties of an orthogonal wavelet method, the no-arbitrage conditions in the bond market are extended to hold for each wavelet detail and wavelet approximation. Based on the extended bond market equilibrium conditions, we construct VAR/VECM models with the decomposed time series by frequencies.
A couple of important empirical findings merit our attention. First, the shorter the time scale of a wavelet detail and time to maturity, the stronger the tie between the movement of the call rate and the consequent movement of the yield curve. Second, we could discern the liquidity effect from the Fisher effect as each effect manifests itself on different time scales of wavelets. Third, the call rate adjustment induces the contemporaneous shift of the yield curve, which in turn supports the presence of the liquidity effect. These results could not be obtained by the traditional analysis utilizing unfiltered raw time series. The merit of wavelet decomposition is self-evident because it closely resembles de-noising processes. Noisy signals do not contain useful information and should be filtered out so that we benefit from more efficient analysis with time series sharpened by a de-noising procedure.