Investor sentiment, defined as the optimism or pessimism about stocks in general (Baker & Wurgler, 2006), can be transmitted to financial markets through investors’ transactions and choices. Specifically, behavioral biases, such as loss aversion, pessimism, and herding, can have a considerable influence on the market during a crisis. This may have a significant impact on the long-term correlation between U.S. stock and bond markets, which can be explained by the flight- to-quality theory (Caballero & Krishnamurthy, 2008). Asgharian, Christiansen, and Hou (2016) focus their analysis of the impact of macro-finance factors on correlation, and also analyze the impact of individual indexes of investor sentiment, including PMI (the log- difference of the producer confidence index), CC (the log-difference of the consumer confidence index) and VXO (the log-difference of the volatility index), on the correlation of stock and bond markets. In contrast to Asgharian et al. (2016), we focus on the impact of the composite index of investor sentiment on the long-term stock- bond correlation and further analyze the time-varying relationship reacts to financial crises. Specifically, we extend the DCC-MIDAS model proposed by Colacito, Engle, and Ghysels (2011) to allow long-term correlation driven by investor sentiment and consider structural breaks to adjust the correlation during different periods. Through the Bai and Perron (2003) test, we detect the multiple structural breakpoints during the time-varying correlation. Considering structural changes in July 1997 and April 2007, we divide the sample period into three sub-sample periods to test whether or not a financial crises impacts the long-term stock-bond correlations. The model can combine daily stock and bond returns with a monthly composite index of investor sentiment and take financial market turmoil into account. The results show that the composite index of investor sentiment has a significantly positive influence on the long- term stock-bond correlation, and the shock of crises significantly decrease the average correlation but the effect of sentiment does not change significantly. Furthermore, our out-of-sample analysis presents significant improvement for the performance of portfolio allo- cation after involving the effect of investor sentiment on the long-term stock-bond correlation.