Predicting market direction during peak earnings season
In this report we investigate whether aggregating Leapday’s REACTION signals on busy earnings days can be used to predict market direction. We construct a market direction signal that achieves an average return of 0.41% over a 3 day holding period. In comparison, going long SPY results in an average 3 day return of 0.19% during 2016-20 and an average 3 day return of -0.18% during the days that overlap with the market direction signal. This signal is especially useful in providing protection against large down moves in a long equity portfolio.
Read as PDF
Leapday’s REACTION signals predict short-term moves for stocks after earning events. These signals are delivered as scores ranging from -100 to +100 indicating the direction and strength of the signal. The signals are available for 3 entry points after the event date (which is defined as the first day the market is open subsequent to the event’s public dissemination): (1) 60 minutes after the open on the event day; (2) just before the market close on the event day; (3) before the market open on the next trading day after the event day. For each delivery time, signals are available for multiple holding periods ranging from 2 to 15 trading days. Each combination of delivery time and holding period is a unique signal generated from an independent model that has been optimized for that specific entry and holding period.
Below are the 11 unique combinations of REACTION signals including their entry and holding period:
In our previous research note “Trading Events: Predicting Reactions After Earnings Announcements“ (available upon request by emailing firstname.lastname@example.org), we demonstrate the predictive power of the REACTION signals to forecast individual stock returns. In this report, we examine how an aggregate score created from the individual predictions can predict moves in the broader stock market on busy earnings days.
Creating a Market Direction Signal
This study is motivated by our anecdotal observation that when a large number of positive or negative REACTION signals are generated on the same day the market tends to move in that direction. Our goal is to analyze whether a skew in aggregate scores on days with higher earnings activity is indicative of positive or negative market sentiment, which can thereby predict market direction. To examine this, we create a market direction signal by aggregating the REACTION signals and establishing a trade signal based on the skew in the aggregate score on busy earnings days. We test the performance of this signal using SPDR S&P 500 ETF (SPY).
We target an entry point at the close after the event date and use a 3 day holding period. As shown previously, there are 11 unique REACTION signals. For this study, we focus on the t0 open + 60min 4d signal and the t0 close 3d signal. Both of these signals would be available before the close after the event date, and both correspond to the 3 day hold we set up above and end at the t3 close. We decide to average the scores from both of these signals to create a more robust signal. Because the signals are already normalized from -100 to 100, averaging the 2 signals will also produce values from -100 to +100 and we expect a mean around 0 for the entire sample. Below is a distribution of the daily aggregated REACTION scores which have been averaged for the 2 signals selected above.
The next step in creating our market direction signal is an important one. This step involves filtering tradeable days down to days in which there is significant earnings activity. It would not make much sense to hypothesize that earnings will impact the broader market on days with light earnings activity. If we look at Leapday’s proprietary event dataset, we notice that the median day has 18 earnings events. We decide to focus purely on the top quartile of days in which there are 65 events or more.
Finally, since our focus is to predict market direction, we need to determine a threshold to apply to the mean daily scores to create buy and sell signals. We use a threshold of +/- 8 to create signals. This threshold represents approximately the strongest one-third of scores on the tradeable days established in the previous step.
To summarize, we create a market direction signal to trade the market close on busy event days and hold this trade for 3 trading days by: (1) aggregating the REACTION scores of the t0 open + 60 min 4d signal and t0 close 3d signal for each day; (2) averaging the aggregate t0 open + 60 min 4d and t0 close 3d signals for each day to create an average daily score; (3) filtering for days in which event activity is in the top quartile of event days; (4) applying a trade threshold of +/- 8 to the average daily score to trade only the strongest one-third of signals on tradeable days.
Below we present metrics from trading the market direction signal created above by trading SPY on the market close on busy event days and holding that trade for 3 trading days. This study uses a sample period from January 1, 2016 to December 31, 2020. There are 1,259 total trading days during the period. Filtering for the top quartile of event days in which there are 65+ events per day, our universe shrinks to 302 tradable days. From these 302 days, we trade on approximately one-third of days in which we have the strongest signals. The result is 99 trades.
The signals do well with an average return of 0.41% across both long and short signals. Next, we compare this result to the return of an average 3 day long and 3 day short hold in the SPY between 2016 and 2020.
As expected, the long trade has positive performance, given that this period was a prolonged bull market. However, this underperforms the signal performance. Next, we compare the market direction signal to the performance of long SPY and short SPY on only the days for which we have a market direction signal.
On days when the signal was triggered the market had a negative return on average. However, because our signal was able to identify positive and negative returns correctly with a hit rate above 50%, we were easily able to beat the baseline return of going either only long or only short on these days.
We have shown that by creating a market direction score based on aggregating company-specific REACTION scores, an investor can predict the outcome of short-term price returns for the broader stock market. We believe that these results exist because earnings activity on busy earnings days contributes significantly to market sentiment.
Investors can use this market direction signal in a variety of ways. First, an opportunistic investor can trade these signals. Second, an investor can use these signals to manage risk around diversified portfolio positions. Particularly, the short signals perform extremely well and can be a valuable overlay in an enhanced return strategy. Finally, an investor looking to gain maximum value can use our underlying REACTION signals for individual earnings events to trade those events based on the strongest signals, as inputs into their existing models to enhance performance, or to manage risk around positions held through earnings events.
In this report, we present an approach to create market direction signals using Leapday’s REACTION signals for individual stocks. This signal can be used on busy earnings days to profit from short-term momentum in the market that is driven by earnings activity. By using these signals, an investor can achieve an average return of 0.41% over 3 days compared to an average 3 day return of 0.19% for long SPY between 2016-20 and an average 3 day return of -0.18% for long SPY during the days that overlap with the market direction signal. Short trades are particularly strong during the prolonged bull market in the sample period with an average return of 0.61% over 3 days, a hit rate of 53.2%, and winners that are 102% larger than losers. This result indicates that the market direction signals are particularly useful in providing protection against downside moves.
In a future report, we will take a look at using these short signals to provide drawdown protection and enhanced returns in a long equity portfolio.
Update March 23, 2021 - to read the report about using the short signals to provide downside protection and enhanced returns, click here.
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.