Wavelet analysis of stock returns and interest rate changes: Evidence from Turkey

This thesis undertakes an attempt to re-examine the interdependence between 604 weekly observations of stock and bond returns in Turkey. The wavelet analysis provides a deeper understanding about the relationship considering the heterogeneous agents trading at different investment horizons. First, test findings reveal cointegration and causal relationships running from bond yields to stock prices for several indices. In line with these findings, time-domain tests suggest that bond yields Granger-cause stock returns, while the reverse does not hold for any indices in the short run. After implementing causality tests to the decomposed series, however, the paper shows that causal relationship is mostly concentrated on the higher frequencies, i.e. mid- and long-term horizons at the both sides. This finding implies that stock returns and changes in bond yields can be used as predictive power on each other. These results are also corroborated by the frequency causality test. Moreover, the asymmetric causality test reveals significant relationships between different return components. The positive component (shock) of bond returns, for example, leads the negative components of "RXU100" and "RXBANK", while, on the other hand, there are causal linkages from the negative components of "RXU100" and "RXBANK" to both positive and negative components of bond returns. Conversely, the wavelet-based outcomes indicate significantly negative relationships at varying significance and magnitudes between variables up to the fourth scale. Moreover, almost all stock indices are more volatile than bond market. "The higher scales, the lower volatility" finding suggests that short-term investors should respond to every variation in asset returns.