While 2020 has seen plenty of unexpected market-moving events, we’re in the midst of one that at least has reliable calendar dates: the holiday shopping season. With COVID-19 having both impacting consumer spending expectations to the downside and accelerated the adoption of e-commerce platforms as people avoid in-person shopping, attention on listed companies operating in the space is high as their performance during the period will shed light on the state of the economy. 

The holiday retailing frenzy has its origins in Black Friday but has extended to include the entire Thanksgiving weekend. Major e-commerce sites have added their own proprietary “days” to the mix in the period leading into Thanksgiving, namely Amazon’s (AMZN) Prime Day and Alibaba’s (BABA) Singles’ Day. (Note: Prime Day was in October this year after having been in July in prior years; Singles’ Day is November 11.)

For educational purposes we kept it fairly simple, looking at the performance of the SPDR S&P Retail ETF (XRT), AMZN, and BABA in the ten days surrounding the “event” in question going back five years. However, there are thousands of ways to research this phenomenon.  

For AMZN, the event was Prime Day:  

For BABA it was Singles’ Day: 

And for XRT it was Cyber Monday: 

From looking at these graphs, it is not immediately clear whether or not these sales events are catalysts for an increase in stock price. Thus, we performed a few tests to determine if there are statistically significant, abnormal returns near the event date.

To do this, we:

• Gathered five years of data, computed the daily returns, and treated this as our population. From that data, we computed the population mean and standard deviation.
• Computed the returns over the five trading days before and after the events, and tested for significance.
• Calculated the single return value between the trading day before and after the events and tested for significance. 

Some caveats to this method:

• We assumed normality in stock returns, which is generally not true.
• These events only occur once per year, so we didn’t have a lot of data.

We tested at the .05 significance level.

For AMZN on Prime Day, we had a p-value of .500 for abnormal returns for both the 10-day returns as well as the single-day returns.

For BABA on Singles Day, we had a p-value of .500 and .498 for abnormal returns for the 10-returns and the single-day returns.

For XRT on Cyber Monday, we had a p-value of .500 and .499 for abnormal returns for the 10-day returns as well as the single-day returns.

From our statistical tests and by looking at our distributions, it is clear that these events do not produce abnormal returns. This is what we would expect — if there was a glaring inefficiency, the opportunity would easily be arbitraged away.

Although this basic exercise may not have produced an immediately actionable insight, it demonstrates how you can harness the data on our platform easily and efficiently to test nascent ideas.

The research notebook used to generate the plots and perform the statistical tests can be found here.

For more examples of how to use QuantConnect’s platform in action, see our Strategy Library or our Idea Streams video series.