An important lesson that correlation does NOT imply causation
Here’s a headline from a Washington Post article by Paulina Villegas on June 16, 2021 (there were many similar headlines around the world as well): “Cristiano Ronaldo snubbed Coca-Cola. The company’s market value fell $4 billion.” The article referred to soccer’s (football’s) Portuguese superstar, who sat down at a press conference on June 14, 2021 during the European Championship and proceeded to remove two bottles of Coke that were prominently displayed on the table in front of him (Coca-Cola was one of the tournament’s official sponsors). He replaced them with a bottle of water, saying, “Agua. No Coca-Cola.” It was immediately big news. Ronaldo has 308 million Instagram followers and this YouTube video of the incident has been viewed almost 16 million times.
The Washington Post article sounded very conclusive: “The simple gesture had a swift and dramatic impact: The soft drinks giant’s market value fell $4 billion, highlighting the power and impact that celebrities and influencers can have on the market.” The key word in this sentence for me is “impact.” That sentence suggests a clear cause-and-effect. The cause was the action by Ronaldo, a celebrity and influencer, snubbing Coca-Cola by removing Coke bottles. The effect or impact of Ronaldo’s action was a substantial drop in the market value of Coca-Cola. Case closed! Or was it? Following an interesting argument provided in this Forbes article, I’ll deliver evidence that strongly suggests otherwise. I’ll also show a better method known as an event study, a mainstay among academics, to better investigate potential cause-and-effect.
The Missing Link: Ex-Dividend Days Explained
First, an important digression and explainer of the missing link in this story. Companies that pay regular dividends, like Coca-Cola, tend to do so on a predictable quarterly cycle, so it usually isn’t a surprise to the market as to the timing of upcoming cash dividend payments. In order to prepare for the distribution of dividends to shareholders, there is a cut-off point, known as the ex-dividend date, whereby anyone who becomes a new owner of shares on or after that date doesn’t receive the upcoming dividend, which is usually scheduled to be distributed a couple of weeks after the ex-dividend date.
As an example, suppose on May 1 a stock is selling for $10 per share with an upcoming dividend of $1 per share to be paid on May 15. The stock is priced at $10 in anticipation of the upcoming $1 dividend. On May 2 the stock trades ex-dividend, and so anyone buying shares then won’t be receiving the upcoming May 15 dividend. Absence any new information, we would expect the stock to drop by $1 (to $9) on May 2, the ex-dividend date. Empirical evidence supports this conjecture.
Coca-Cola’s (KO) recent ex-dividend date was June 14, 2021, which just happened to coincide with the day of Ronaldo’s press conference. That meant that anyone who first owned the stock on that date wasn’t eligible for the $0.42 per share dividend that was going to be paid on July 1. With 4.32 billion shares outstanding, that’s a cash payment of about $1.8 billion. So, absent any new information, we would expect KO’s market value to drop by that amount. Note that for anyone who owned the shares prior to June 14, 2021, the anticipated share price drop wouldn’t affect their overall wealth. Like many major companies, KO has a dividend reinvestment program (DRIP), and so investors can elect to have any cash dividends immediately converted into purchasing more stock.
What Really Happened on June 14, 2021
Now let’s see what was really happening with KO around the time of the infamous snub. On Friday, June 11, KO’s stock price closed at $56.16 a share, and KO’s shares had an overall market value of $242.61 billion. On Monday, June 14, when the market opened at 9:30am ET, the stock was at $55.69, down $0.47 a share or a market value loss of $2.03 billion. The price continued to drop in the next several minutes, to $55.27 at 9:43am. By that time, the overall stock value had declined by $3.85 billion. That sounds like a lot of money, but to put that in perspective, it’s just 1.5 percent of the Friday value. Furthermore, half of that amount (0.75 percent or $0.42 per share) was because of the ex-dividend date effect. Since the S&P 500 index had changed very little since Friday (just a slight decline of -0.1 percent), then we can conclude that there was some reason, probably related to the outlook for KO or its industry, that caused the unexplained drop of about 0.75 percent.
What’s important to note is that this all occurred prior to the start of Ronaldo’s press conference, during which he removed the Coke bottles, at 9:43am ET. The chart below tracks minute-by-minute June 14 prices for KO, as well as the S&P 500 (normalized to the KO price at that precise time).
From 9:43am through to the remainder of the trading day, KO’s stock price actually rose, both in absolute terms as well as relative to the overall market. So how about this for a revised story headline: “Cristiano Ronaldo snubbed Coca-Cola. The company’s market value then rose by $1.2 billion.” If that were the new headline, would we then try to infer cause-and-effect and conclude that Ronaldo was a has-been and no longer an influencer
A Better Investigative Approach: Event Studies
So far, I’ve shown that there was no immediate impact on KO’s stock price by Ronaldo’s actions on June 14, 2021. Notice that I also examined what was happening with the S&P 500 index. The idea to compare a stock’s price movement relative to the market in order to isolate firm-specific effects came from a seminal 1969 research paper. The University of Chicago authors, including Nobel Laureate Gene Fama, developed a technique known as an event study to investigate and try to quantify potential cause-and-effect of events on market values.
The methodology starts with examining stock price movements prior to a known event (like the Ronaldo snub) relative to overall market movements, and then tracks what happens around and after the event. Each day, the overall market return (like the S&P 500) is subtracted from the stock’s return in order to estimate an abnormal return (AR) that suggests what the impact of the event on the stock might be, absent any movements related to the overall stock market (for example, an announced interest rate change by the Fed). By tracking beyond the one event day, we can look for any cumulative effects (that is, by summing the ARs to get cumulative ARs or CARs). This way we can track events that might have had a more gradual impact on a stock’s price. For example, in the case of the snub, it might take a while for consumers to reassess whether they want to support the company or perhaps stop buying the product, and so we might see an impact over several weeks.
Starting in May 2021, the month before the snub, we can track CARs through June 11, 2021 (the trading day before the snub) through to the end of July. The chart below shows the results.
Between May and June 11, KO’s stock performed about 2 percent better than the market. We’d have to search to see what news might have been driving that. After June 11, we actually do see negative abnormal returns subsequent to the event (note that the prices used in this graph adjust for the effects of the dividend payment). We can’t say with certainty that the Ronaldo snub caused the negative abnormal returns, but there is an association. By early July, cumulative abnormal returns are around -3.2 percent. But then over the next few weeks the abnormal returns were positive and the CARs turn positive and so any negative impact around the snub event has disappeared. Since the beginning of May 2021, KO’s stock has actually outperformed the overall market. What can we conclude? Subsequent to the Ronaldo snub, there is a negative association with KO’s returns relative to the market that appears to last several weeks but is only short-term in nature, and there is no long-term negative association with the event.
Correlation versus Causation
Notice that I was very careful in concluding about associations rather than causation. Let’s go back to the Washington Post headline: “Cristiano Ronaldo snubbed Coca-Cola. The company’s market value fell $4 billion.” Taken literally and separately, each sentence is a true statement. There’s no question that Ronaldo’s removal of the Coke bottles and his statement “No Coca-Cola” was a deliberate snub. And based on the change in KO’s closing stock price between Friday, June 11, 2021 and Monday, June 14, 2021, KO’s market value dropped by more than $4 billion.
Is there a correlation between the snub and the stock price decline? No question, as they both occurred on the same day. But the evidence above is strong that within that particular day, the snub did not cause the drop in the stock price. What we do know — and is well established — is that the stock going ex-dividend was the primary cause of the price decline at the opening of trading. This Associated Press article on June 21, 2021 was much more accurate than the Washington Post article when it stated: “A drop in Coca-Cola’s share price this week was attributed by some to Ronaldo’s snub, but without any evidence that the two things were connected.”
Don’t Believe Everything You Read
What to take away from all of this? Here’s my favorite nerdy cartoon that distinguishes between correlation and causation. (It’s one of those cartoons that you might need to give some time to sink in.)
It’s important to note that, even applying event study methodology, proving causation related to stock price movements is very difficult, as stock prices are driven by all kinds of information, some which may impact the overall market, like economic news, and some that is specific to a particular firm. At best, we can look for evidence of correlation coupled with a plausible explanation that is isolated from other possible explanations. In the case of the Ronaldo snub, the Washington Post and many other news outlets overlooked important information — the missing link — related to the ex-dividend date.
In writing her story, I don’t believe Paulina Villegas had the objective of creating fake news. I suspect she didn’t realize that June 14, 2021 was KO’s ex-dividend date and then proceeded to attribute correlation to causation. The lesson here isn’t to not trust mainstream news outlets but rather, more broadly, to gather as much relevant information as you can before jumping to any conclusions, or making any decisions.
Oh, and by the way, according to the Guardian, the bottle of water that Ronaldo replaced the Coke bottles with was also a Coca-Cola product.