Stocks and bonds are the two most important asset classes traded on financial markets, which play a crucial role in asset allocation and portfolio management. It is now the stylized fact that the stock-bond relations have changed dramatically over the last two decades, shifting from sizably positive to predominantly negative in the late 1990s (Baele, Bekaert, and Inghelbrecht, 2010; Bansal, Connolly, and Stivers, 2014, Chiang, Li, and Yang, 2015; Lee, Marsh, Maxim, and Pfleiderer, 2006). Efforts also have been made to explore various economic forces driving the time-variation in stock-bond relations, including macroeconomic state variables (Baele, et al., 2010; Yang, Zhou, and Wang) and financial market uncertainty (Connolly, Stivers, and Sun, 2005; Chiang, Li, and Yang, 2015). Although the aforementioned studies have provided insight on the time-variations in stock-bond relations, few have ever been on investigating the dynamics of their relations across frequencies. This paper expands extant literature to explore their dynamics simultaneously across various frequencies over time and further examine whether the relations can be linked to fundamental economic factors and stock market uncertainty. Understanding the dynamics of stock-bond relations across various frequencies over time and their determinants is important for deciding the asset allocation and making macroeconomic policy. For example, investors with various investing horizons concern the characteristics of asset returns in the corresponding investing frames respectively. Thus, exploring the time-frequency relations of stock and bond returns is crucial to asset allocation as the market participants pursue different investing horizons. Moreover, given stocks and government bonds account for a dominant share in all traded financial assets, the main economic forces driving their relations also become of interest for regulatory and monetary authorities. The determinants of their relations provide useful information for policy maker on understanding the status of financial markets and the expectations of investors. In order to capture the time-varying structures of stock-bond relations, most existing studies have investigated by the rolling-window correlation (Gulko, 2002; Ilmanen, 2003; Andersson et al., 2008) or worked with the family of GARCH models (DeGoeij and Marquering, 2004; Connolly et al., 2005; Chiang et al., 2015). In addition, copula method has been applied to characterize extreme dependence of stock and bond returns recently (Chui and Yang, 2012; Jammazi, Tiwari, Ferrer, and Moya, 2015; Sun, Rachev, Stoyanov, and Fabozzi, 2008; Wu and Liang, 2011). The benefit of copula method is that it can provide the whole dependence between asset returns beyond only linear association and result holds regardless of the distribution of returns. However, as mentioned above, investors with various investing horizons concern the corresponding dependences of investing frames respectively. Aforementioned approaches failed to capture the time-varying dependences across different frequencies.