Summary
In a previous report, we aggregated REACTION signals for all stocks to create a daily sentiment score for earnings activity on a given day that predicted down moves in the stock market (SPY) on the next trading day. In this report we conduct the same study but only aggregate REACTION scores for the most liquid 500 stocks to create our daily sentiment score since the largest, most liquid stocks contribute most to the performance of SPY. Between 2017 and 2021 H1, this sentiment score produced 217 indicators predicting down day in SPY out of 1,383 total trading days. The average return of SPY on days with an indicator was -0.10% as compared to an overall average return of 0.05% for all trading days indicating the sentiment score is predictive of down moves in the market. When we use the indicator as an overlay to SPY to provide downside protection and enhance returns, we find that the indicator boosts annualized returns of SPY between 2017 and 2022 H1 from 11.8% to 16.7%, decreases maximum drawdown from -33.7% to -26.9%, and increases Sharpe ratio from 0.67 to 0.98.
Read as PDF
Introduction
The Leapday REACTION signals predict short-term moves in individual stocks after earnings announcements. In a previous report, we aggregated these signals for all stocks to predict down days in the stock market (SPY). Given the significant amount of investor attention to earnings events, the thesis of our previous paper is that negative earnings events affect overall market sentiment on days in which there are lots of earnings activity (i.e., peak earnings season) and/or on days in which the largest companies report earnings and can thereby predict down moves in the stock market when aggregated.
This paper conducts the same study but focuses only on the most liquid 500 stocks when aggregating REACTION scores to predict down moves in SPY. The most liquid and largest stocks tend to drive the market given that SPY is a cap-weighted index. Therefore we wanted to test whether focusing on signals for stocks that are likely part of the market index provided a stronger market prediction.
We proceed by creating an indicator to predict down moves in SPY using Leapday’s REACTION signals for the most liquid 500 stocks, producing summary statistics for the performance of this indicator, using the indicator as an overlay on SPY to provide downside protection and enhance returns, and discussing the findings.
Creating The Downside Earnings Sentiment Indicator
This paper examines if the REACTION scores for the most liquid 500 stocks can be aggregated to predict down moves in SPY. Our focus will be aggregating the signals from the first day of trading after an announcement (t0) to predict moves in SPY for the next trading day (between the t0 close and t1 close). To create our “Downside Earnings Sentiment Indicator”, we take the following steps which we also used in the first part of the study, except in this study we filter for only the 500 most liquid stocks, cap average traded value at $500M, and only include days when earnings activity for the most liquid 500 stocks exists:
We use both the t0 open + 60 min and the t0 close REACTION signals to create the Downside Earnings Sentiment Signal as both entry points make predictions that cover the desired period of this study (between the t0 close and t1 close) and past research has indicated that there is low overlap in trades between the entry points.
We filter REACTION signals for the most liquid 500 stocks as that is the main distinction between this study and our previous report.
We filter REACTION signals for approximately the top one-third of signals as our studies have shown that approximately the top one-third of signals have the highest predictive power.
For each of the 2 REACTION entry points (t0 open + 60 min and t0 close), we multiply an individual company’s REACTION score by that company’s trailing 20-day average traded value and create a daily sum for all companies that report earnings on that day (“Earnings Sentiment”). By multiplying the trailing average traded value by the signal, we give greater weight to larger companies, which have more influence on the market’s performance. To prevent extremely large and active companies from dominating that day’s Earnings Sentiment, we cap trailing 20 day average traded value at $500M. (Our previous paper that used all stocks to create Earnings Sentiment capped trailing 20 day average traded value at $250M. Since this paper only uses this 500 most liquid stocks, we increase this upper threshold to $500M.)
For each of the 2 REACTION entry points, we calculate a percentile rank for the Earnings Sentiment based on the trailing year and only include days in which there was earnings activity for the most liquid 500 stocks in the percentile. We then take an average of these 2 percentile ranks (“Earnings Sentiment Percentile”).
We finally create our Downside Earnings Sentiment Indicator by using the bottom one-quarter of the Earnings Sentiment Percentile as a trigger. These days should be the weakest in terms of market sentiment because they have the highest weighting of REACTION signals.
Summary Statistics
Next, we examine the performance of SPY on days in which the Downside Earnings Sentiment Indicator is generated as designed above using only the most liquid 500 stocks. As the REACTION signals have out-of-sample history since January 2016 and the Downside Earnings Sentiment Indicator uses a 1-year rolling period to rank the Earnings Sentiment, the indicator’s history begins in January 2017.
We begin by examining the average return of SPY on days when there is a Downside Earnings Sentiment Indicator and compare that to the overall average daily return of SPY.
As Table 1 indicates, the average return of SPY is significantly less on days in which there is a Downside Earnings Sentiment Indicator versus the average SPY day indicating that the indicator has ability to predict downside moves and can potentially be used as an overlay on SPY to provide downside protection and enhance returns.
To better examine the robustness of this signal, we next look at average returns of SPY on days when there is a Downside Earnings Sentiment Indicator by year and compare this to the average daily SPY return by year.
Except for 2017, the Downside Earnings Sentiment Indicator occurred on days when SPY had weaker performance than the average day of that year. This result has been particularly evident in 2022 H1 indicating a strong downside indicator in recent market conditions.
Creating An Overlay On SPY To Provide Downside Protection
Below we use the Downside Earnings Sentiment Indicator as an overlay on a long SPY portfolio to provide downside protection. As the Downside Earnings Sentiment Indicator is available before the market close (using the aggregate of REACTION signals from t0 open + 60min and t0 close), we will exit a long SPY position at the close on any day with an indicator and re-enter a long SPY position at the close on the following day without an indicator. In other words, portfolio performance will be 0.0% on days with an indicator.
Below we show the performance since 2017 of a long SPY portfolio that exits the long position on the 217 days in which there is a Downside Earnings Sentiment Indicator. We compare this portfolio to SPY performance over the same period.
Tables 3 and 4 provide metrics associated with the chart above.
Table 3 above indicates the Downside Earnings Sentiment Indicator works well as an overlay that provides downside protection on a long SPY portfolio. Annualized returns since 2017 are significantly boosted and returns are improved in every year except for 2017. Table 4 additionally shows that a portfolio using the Downside Earnings Sentiment Indicator as an overlay has significantly lower drawdown and stronger risk-adjusted metrics.
Next, we take a deeper dive into the indicator performance by examining additional metrics including hit rate and average winner divided by average loser. We study all days in which there is a Downside Earnings Sentiment Indicator and consider a winning day one in which the SPY was down and thus the signal worked as intended and consider a losing day as one in which the SPY was up.
With a hit rate of 43.3% and a Avg. W / Avg. L of 1.65, these results indicate that the indicator boosts portfolio performance not because it is more often right than wrong but because it helps in protecting against large losing days.
Earnings Sentiment Indicator Comparison
Finally, we compare the performance from this report of using only the most liquid 500 stocks to create our Downside Earnings Sentiment Indicator to the performance from our previous report which uses all stocks to create the indicator. Tables 6 and 7 below show the performance since 2017 of a long SPY portfolio that exits the long position when there is a Liquid 500 Downside Earnings Sentiment Indicator and when there is an All Stocks Downside Earnings Sentiment Indicator.
The All Stocks Downside Earnings Sentiment Indicator outperforms the Liquid 500 Indicator in the higher volatility years of 2018, 2020, and 2022. This is perhaps due to the All Stocks Indicator being triggered on more days than the Liquid 500 indicator. The Liquid 500 Indicator, however, still outperforms SPY significantly in high volatility years and outperforms the All Stocks indicator in other years, in particular in 2021.
Discussion
We have shown that by aggregating company-specific REACTION scores for the most liquid 500 stocks into a daily sentiment score, we can create an indicator that predicts large down moves in SPY. This indicator can be used by investors to obtain downside protection and significantly enhance returns by using the indicator as an overlay for a long equity portfolio. We believe these results exist because a large number of negative REACTION signals and/or the largest companies having negative REACTION signals affects market sentiment and is a destabilizing factor driving big downward moves.
The indicator is robust and helped provide downside protection in every year since 2018. It has been particularly helpful in providing downside protection in 2022 amidst a macro-driven market in which inflation and rising interest rates have created greater uncertainty and where earnings announcements of the largest and most actively traded companies provide foresight into the health of the economy.
Investors can use a Downside Earnings Sentiment Indicator in a variety of ways. Most obviously, they can use the indicator as an overlay to reduce drawdown and enhance returns in a long equity portfolio. A sophisticated investor will likely implement an options strategy in lieu of the methodologies described above. In addition, more active investors can use the indicator as a risk management tool in existing long equity strategies.
Conclusion
In this report, we extend previous research which used Leapday’s REACTION signals as an overlay on SPY to provide downside protection and enhance returns. As a follow up to the previous study we focused specifically on the most liquid 500 stocks on a given day to create an indicator that predicts down moves in the stock market (SPY) on the next trading day. Between 2016 and 2022 H1, this sentiment score produced 217 downside indicators predicting down days in SPY. When we use this indicator as an overlay to SPY by exiting a long SPY position on days with a down indicator, we find that the indicator boosts annualized returns of SPY between 2016 - 2022H1 from 11.8% to 16.7%, decreases maximum drawdown from -33.7% to -26.9%, and increases Sharpe ratio from 0.67 to 0.98. This overlay can be used to reduce drawdown and enhance returns in any long equity portfolio.
Disclaimer
This material is solely for informational purposes and is not an offer or solicitation for the purchase or sale of any security, nor is it to be construed as legal or tax advice. References to securities and strategies are for illustrative purposes only and do not constitute buy or sell recommendations. The information in this report should not be used as the basis for any investment decisions. We make no representation or warranty as to the accuracy or completeness of the information contained in this report, including third-party data sources. The views expressed are as of the publication date and subject to change at any time. Hypothetical performance has many significant limitations and no representation is being made that such performance is achievable in the future. Past performance is no guarantee of future performance.
Comments