Trading Stocks Make Sense - Post Earnings & Long Bias

Opportunities you should not miss after earnings announcements

In this short article, we will present you the proof that trading stocks make sense after earnings announcements, and that there are opportunities you should not miss. We all know that earning’s announcement can make a big difference – huge gains or losses. It is very risky because before earnings, we never know if the surprise will be positive or negative. An earnings surprise is percentual comparison of a company’s reported quarterly profits and analysts’ expectations.

post earnings key takeaways trade with science

We decided to do research for you: we will use all earnings on the US market during the years 2018 and 2019. This study could be beneficial for investors who do not have access to before/after market hours. Since we do not want to play with random surprise we will analyze post earnings trading.

Post Earnings

Post earnings mean: if the announcement is before the market open, we will hold the position from the market open till the close of that day. If the announcement is after the market close, we will open it the next day when the market opens. We will compare the performances for holding the position for 1, 3, 5, 10, 15, and 21 trading days (a full month). We also distinguish the performance with positive or negative surprises and consider if the actual earnings were positive or negative.

We have to take into account that the stock market is naturally long biased. Because of this bias the average yearly return for the last 90 years on the US market is 9.8% (Composite Index, later S&P 500). It means when you short stock, your probability of win is generally lower than 50% (if you do it randomly). On the other hand, when you are long, you have a higher chance to end with gains (again when you trade randomly). We will also compare trading after the earnings with this bias.

Below you can see a simple table of average performance while the earnings’ surprise is in 3 categories: lower than -10%, between -10% and 10% and over 10%. We want to catch if there is significant difference between the groups. It is natural to analyze the surprise because that is the thing which moves with the price.
 

< -10%

[-10%, 10%]

> 10%

1 day return

-0.74%

0.07%

0.16%

3 days return

-0.78%

0.10%

0.28%

5 days return

-0.90%

0.07%

0.20%

10 days return

-1.23%

0.05%

0.21%

15 days return

-0.93%

0.22%

0.16%

21 days return

-1.16%

0.42%

0.36%

count

5470

11565

9440

In the table below we have % change of price after x days and in the last row the counts. From the counts we can see that there are almost double positive surprises than negative, which confirms the long bias assumption. We can clearly see that the returns for negative surprise are significantly negative (tested with different statistical tests), but the return after positive surprise is not significantly higher than after no surprise. 

Two years of observation is a statistically significant sample and we observed a growth period in 2019 as well as downtrends in 2018. We should also look for winning ratio, where we short stocks for after negative sentiment:

 

 

Winning ratio

Counts

surprise < -10%

55.02% (shorts)

5469

-10% <= surprise < 10%

51.37% (longs)

11564

surprise > 10%

50.04% (longs)

9442

From this we can clearly see that creating some strategy for shorting after a negative surprise can have an interesting edge (do not forget that we get these results in a long-biased market). But trading long after earnings does not have expected results, and the return is straightly random 50:50.

Let’s finish this study by a short analysis of the long bias of the stock market. We will only compare the results after a positive surprise.

 We have 9440 observations of positive surprises during those 2 years. So we create random samples of 9440 observations – we buy a random stock at random time at an open price and sell it after 3 or 10 days at close price. 

We will compare only 3 day and 10 day returns. This random sample we created 1000 times so we could get the distribution of randomly selected long positions. The results are in the histograms and the table

 

post-earnings mean

randomized mean

randomized median

3 day return

0.28%

0.74%

0.63%

10 day return

0.21%

1.13%

1.12%

 

post-earnings mean

randomized mean

randomized median

3 day return

0.28%

0.74%

0.63%

10 day return

0.21%

1.13%

1.12%

Without doing any statistical tests we can clearly say that post-earnings long positions have worse performance than trading just randomly. It doesn’t even beat the long-bias of the market, so you should not waste your time trading long post-earnings. 

From this analysis, it looks like the main movement is during the gap or first day. We didn’t want to do a theoretical analysis that we hold position during the announcement because we consider the surprise as an unknown variable that we can not predict.

We add histograms to see the distribution of long bias during 2018-2019 (we accept the fact that for this analysis of long-bias, we should have used more extended data, but found these two years enough representative).

In the table above you only see an average performance, for demonstration purposes we look at histograms. Histogram is constructed by 1000 random samples, each sample consists of 9440 long positions, with random start and holding period 3 and 10 days. For each sample we calculate the average return.

3 day returns

10 day returns 

We can see that there are not even the observations where the average return of 9440 longs positions is negative, that strong is long-bias (only one in 3 days’ returns). Do not forget about this fact when you plan to hold shorts for a longer time.
We analyzed trading post-earnings, which means we started a day after the earnings announcement. Next to it, we showed how the market is long-biased. This bias is so strong that buying post-earnings stocks with positive surprise does not beat the random stock buying. But on the other hand, shorting the stocks with negative surprises has a significant edge.

Comments (1)

Excellent topic, thank you!

Leave a comment