This study investigates the impact of both social and news sentiments indices on the dynamic stock–bond correlation across wavelet-based time-scales over the period 1998–2016. Our results show that the news sentiments namely unemployment, tsunami and sanctions exhibit significant effects during expansion at the shortest time-scale of [2–4] days. These predictors remain significant with reverse signs during recession on the long investment horizon. Yet, the predictability of social media sentiments differs from that of news sentiments with the pattern of reversal in sign also presents for some proxies including windstorm and investment flows. Statistically, our further analysis confirmed the predictability of the sentiments out-of-sample. Excluding the news and social media sentiment effects has also resulted in minimizing the value-at-risk of the (40/60) stock/bond portfolios the most at the investment horizon of [32–64] days during recessions. Our results remain the same after performing some robustness checks.

这项研究调查了1998年至2016年期间，社会情绪和新闻情绪指数对基于小波的时间尺度上的动态股票-债券相关性的影响。我们的结果表明，新闻情绪，即失业，海啸和制裁在扩张的最短时间内（2-4天）表现出显着影响。在长期投资前景不景气的情况下，这些预测指标仍然具有显着的反向迹象。但是，社交媒体情绪的可预测性与新闻情绪的可预测性有所不同，符号反转的模式还针对某些代理（包括暴风雨和投资流）呈现。从统计学上讲，我们的进一步分析证实了样本外情绪的可预测性。排除新闻和社交媒体的情绪影响，还导致在经济衰退期间[32-64]天的投资期限内，最大程度地降低了（40/60）股票/债券投资组合的风险价值。进行一些鲁棒性检查后，我们的结果保持不变。