Elon Musk's tweets have a causal impact on the price of Dogecoin
You can't help but think of Elon Musk as you believe of Dogecoin, the cryptocurrency founded on a meme. That guy adores the doge, and the price rises every time he mentions it on Twitter.
Elon Musk is famous for swinging stocks. Alex Hayes used S&P500 to quantify the causal impact of the tweet below on Tesla's stock price in a brilliant blog post last year. He uses Brodersen et al.'s excellent CausalImpact R bundle (2015).
Switch to Dogecoin's latest craze. The following figure shows Dogecoin and Bitcoin prices for a selected time frame.
That week, Dogecoin erupted, mostly as Redditors were rallying around GameStop on the moon. About 18 million Bitcoins are currently in circulation, and there are no more than 21 million Bitcoins accessible. There are around 127 billion Dogecoins in circulation, and there is no upper limit on what the number will be, as opposed to Bitcoin.
We standardize them (regarding themselves) in the figure below to help equate the two-time sequences.
Initially, we see Bitcoin as more unpredictable, but both cryptocurrencies will rise from about 28 January at 12:00 pm. The black vertical line tells when Elon Musk tweeted. What did he share?
What is this tweet's causal effect? Although the S&P500 is very different from the Cryptocurrencies, the counterfactual Dogecoin cost is predicted by Bitcoin. I use the databases as mentioned earlier, beginning on 28 January from 12:00, since Bitcoin is not especially practical with Dogecoin. Likewise, I look after the tweet for a subset of the results. This is because of the unpredictable cryptocurrencies, and the causal impact of Elon Musk's tweet may be easily removed.
We see that the model reasonably well before Elon tweet forecasts the price of Dogecoin. The counterfeit Dogecoin price (which means Elon did not tweet the Dogecoin price) is expected to remain somewhat stable, while the actual cost is rising. But it doesn't grow right away but with a pause — maybe because he tweeted midnight? Dogecoin shows an overall 33% rise in any case (95% accurate range ranges from 23% to 42%), but note the expected dependence of this number on the post-tweet period that we take into consideration.
The figure above shows in particular that the Dogecoin price falls after the initial rise. However, overall, tweet Elon seems to have a significant causal impact on Dogecoin's price.
It is worth remembering that the study assumes that Elon Musk's tweet was only relevant for Dogecoin. However, Redditors rallied behind the blockchain and perhaps confused the causal influence of the message at the same time.
Sending Dogecoin to the Moon:
A week after the first frenzy, Musk fired several Dogecoin tweets. Let's have the details zoomed in.
The vertical black line shows the time of Elon's first tweets, followed by several other tweets. So how do we glean ideas from it?
Following this avalanche of tweets, Dogecoin grew significantly. But again, Elon's tweets did not mean that the price was could. I use the same analysis as before to evaluate whether these tweets had causal effects. As Musk has tweeted several times, I use the first tweet as my benchmark. Like above, this time, from 3 February midnight, I only select a part of the results.
Tweeting about Dogecoin on 4 February
The average estimation of causal effects is a 23% rise in price, with a reliable 95% 19 to 28 percent interval. (But again, remember that the post-tweet degree is the vulnerable period we consider). Between the first tweet and the price, the raise is a little late, and Redditors rallying around Dogecoin is not so concerned as before. But the opposite forecasts are less compelling than before and show the fragile relationship between Dogecoin and the pre-tweet Bitcoin. Of course, the process is unsure (for details, see Brodersen et al., 2015).
There are still theories on causal deduction. We assumed in this connection that the price of Doges at the time of the tweetings of Elon Musk may not have been affected by any significant incident and that the third variable did not cause Musk and Doges to tweet. In the second study, these conclusions are more realistic.
We have believed that the price of Bitcoin is relatively consistent with Dogecoin prices and that the relationship continues after the tweets. One might verify how proper Bitcoin is as control by analyzing different sub-sets and comparing the expected dogecoin price to the accurate dogecoin price. However, because I want to waste too much time dreaming about Dogecoin on a Sunday afternoon, I leave it to others.
One could find it more controllable by mixing many cryptocurrencies rather than only depending on Bitcoin — or drop the whole control game and hit the Dog in a Gaussian manner in a disrupted time series (e.g., Leeftink & Hinne, 2020). More philosophically, the study supposes that contrasting arguments are meaningful, non-disputable (e.g., Dawid 2000; Peters, Janzing, & Schölkopf 2017, p. 106).
Further, the analyses assume that prices in Bitcoin do not depend on tweets from Musk. If they were affected by them — claim they trigger Bitcoin price rises — so they will have a declining causal impact on Dogecoin. It seems probable that Musk's tweets would affect Bitcoín if they were to influence Dogecoin (e.g., merely by focusing on cryptocurrencies), and if you were interested in a non-partisan – or rather, less biassed – calculation, you could think harder.