# Quant Investing: Application of Investor Sentiment Constructed by Overnight Return

Investor sentiment is an important topic that has been researched in these two decades There are diverse top-down or bottom-up approaches and use market or individual-level variables to measure investor sentiment. In the report, we focus on the sentiment constructed by the overnight return of industry ETFs provided by Lee, Y.-H. (2022) and make some extensions on their result. The most important in paper is that investor sentiment constructed from overnight returns of industry ETFs provides useful information for predicting VIX and Stock index futures returns at the monthly level.

Figure 1 shows the correlations for the sentiments of this month and the next month’s return of the VIX index, E-Mini S&P 500 future, and Mini Dow Jones future, which is very similar to the result in Lee’s paper. Although RCM has a higher value of correlation between volatility and market return, we use CM to extend the result. Since RCM is calculated by quantile regression using small size data, it may not stable as CM using regular regression. Consider for the convenience, we can replace S&P 500 future with SPY and VIX future with VIXM to trade in the strategy since they have the same correlation level with the sentiment.

We can construct the timing strategies for SPY (Monthly) and VIX (Daily) separately easily and make some optimization for VIX strategy. Figure 2 shows that the performance of both. We also can combine the VIX stats as a hedging strategy with the SPY stats (SENCM). Figure 3 shows that the performance of these mix strategies Figure 4 shows the statistics of each strategy. We can observe that both mix strategies’ Sharpe (1.48 and 1.28 v.s. 0.77)and return are higher than SPY and have smaller max drawdown (−12.92% and −22.24% v.s. −33.72%). The result shows that it can improve the performance of SPY very much and substantially reduce the time that money is in the market.

Investor sentiment is somewhat predictable of future market return and volatility. We also can build some naive strategies which have great performance. But we still need to be cautious about the low correlation, the period of causal relationship. In summary, even if the performance is good, we still need to be more careful when using this strategy again.

Yu-Chin Lin

*Quantitative Research Intern, Quantitative Finance*, **Gamma Paradigm**