Bitcoin Is a Classic Bubble InvestmentIn large data sets, correlations are easy to find. Useful relationships are more elusive
Cryptocurrencies like bitcoin are digital currencies that operate independently of central banking systems. They appeal to people who don’t trust central banks. They appeal to people who think computers are taking over the world. They appeal to people who want to get rich by making offbeat investments.
I’ve been investing and teaching investing for decades, so my main interest in cryptocurrencies is in whether they are sound investments. I’m a value investor—I even wrote a book about that—and the short answer is no, cryptocurrencies are no more a sound investment than were tulip bulbs during the infamous Dutch tulip-bulb bubble (1637), which ruined investors who bought overpriced bulbs. As an investment, cryptocurrencies are nothing more than digital tulip bulbs.
Investors who buy bonds get paid interest. Investors who buy stocks get paid dividends. Investors who buy apartment buildings get paid rent. People who buy cryptocurrencies get nothing more than the hope that they can sell their cryptocurrency to a Greater Fool for a higher price than they foolishly paid.
This reality hasn’t dissuaded delusional people from buying cryptocurrencies any more than it dissuaded earlier fools from speculating in tulip bulbs in the 1600s, South Seas stock in the 1700s, and dot-com stocks in the 1990s.
Beyond rumors, gossip, and wishful thinking, how can anyone predict the price of cryptocurrencies, which are even more detached from reality than were tulip bulbs and dot-com stocks? At least tulip bulbs can be planted and later they generate more bulbs. At least some dot-com companies were profitable. With cryptocurrencies, there is no there there. People would buy bonds, stocks, and apartment buildings even if a law was passed that forbid them to sell them—because of the coupons, dividends, and rental income. No one would buy cryptocurrencies if they could never sell them.
Although the value of bitcoin is not logically related to anything tangible, there will inevitably be fleeting coincidental correlations with other variables. In 2018, Aleh Tsyvinski (a Yale professor of economics) and Yukun Liu (then a graduate student, now a professor himself) reported that the price of bitcoin, the most popular cryptocurrency, could be predicted from, of all things, how often the word bitcoin is mentioned in Google searches. Specifically, they concluded that an increased number of bitcoin searches typically preceded an increase in bitcoin prices. A decline in bitcoin searches preceded a decline in prices. They attributed this correlation to changes in “investor attention,” although they also said that, “We don’t give explanations, we just document this behavior.”
For the months they studied, January 2011 through May 2018, the correlation between monthly bitcoin searches and the market price of bitcoin on the first day of the next month was a remarkable 0.78. Figure 1 below shows this close correlation, right up until the spring of 2018 when bitcoin prices did not fall as much as predicted by the decline in bitcoin searches.
If this pattern continued, it would be a novel way to get rich: Using digital data to predict the price of a digital currency. In months when bitcoin searches are high, buy on the last day of the month; in months when bitcoin searches are low, sell on the last day of the month.
Figure 1 Predicting Bitcoin Prices from Bitcoin Searches:
Consider, however, the fact that many (if not most) people searching for information about bitcoin may not be speculators, but are simply curious about this new thing called bitcoin. In real estate, people who go to open houses with no intention of buying are called lookie-loos. There are surely millions of bitcoin lookie-loos, which makes Google searches a poor proxy for investor enthusiasm.
After the lookie-loos stop looking, will prices fall? Not if they weren’t potential buyers. Did the rise and subsequent fall in bitcoin search activity shown in Figure 1 reflect investors chasing and then fleeing bitcoin, or looky-loos who were curious about bitcoin and then satisfied their curiosity?
Figure 2 shows what happened subsequent to the Tsyvinski/Liu analysis. Bitcoin search activity stabilized at a low level while bitcoin prices fell, rebounded, and then fell again, causing search activity to be a terrible predictor of bitcoin prices.
The original correlation had scant basis in theory. There was no compelling reason for a lasting correlation between bitcoin searches and bitcoin prices and the temporary correlation was useless for predicting future bitcoin prices. Buying bitcoin based on bitcoin searches was even nuttier than buying bitcoin in the first place.
Figure 2 Mis-Predicting Bitcoin Prices from Bitcoin Searches:
It is easy to find coincidental patterns in large data sets, and Tsyvinski/Liu discovered this one by considering literally hundreds of possible correlations between bitcoin prices and other variables, including stock prices in the beer, book, and bowling center industries. (Not to mention greeting cards, musical instruments, and shoe stores.)
I looked at a few correlations too. Figure 3 shows a comparable lookie-loo relationship between bitcoin prices and searches for Jumanji, which relates to the movie Jumanji: Welcome to the Jungle that was released in December 2017, near the peak of the bitcoin bubble. For the period January 2011 through May 2018, the correlation between Jumanji searches each month and the market price of bitcoin on the first day of the next month is 0.73, which is comparable to the 0.78 correlation between bitcoin searches and bitcoin prices. Those who believe that correlation supersedes causation would not be troubled by the fact that Jumanji has nothing to do with bitcoin. Correlation is enough.
Figure 3 Jumanji and Bitcoin:
Figure 4 shows that, like the correlation between bitcoin prices and bitcoin searches, the correlation between bitcoin prices and Jumanji searches was short-lived. (The surge in Jumanji searches in late 2019, right when bitcoin prices were falling, was due to the release of the sequel, Jumanji: The Next Level.)
Figure 4 Déjà vu, Déjà vu:
In both of these cases, bitcoin and Jumanji searches, there is a close correlation for a while, suggesting that statistical trends in a Google search term can profitably predict bitcoin prices. Then the models subsequently whiff on bitcoin prices.
In large data sets, correlations are easy to find. Useful relationships are more elusive.
If you enjoyed this piece by Gary Smith on bitcoin investment, you may also want to look at some of his other thoughts on what statistics do (and don’t) really tell us about investment:
Investors, AI isn’t your big fix In investing and elsewhere, an AI label is often more effective for marketing than for performance
A BABY, A GEEK, and a COW all walk into a bar looking for some BEER and VINO… …
Some other thoughts at Mind Matters News on bitcoin investment:
Is lack of trust Bitcoin’s biggest security threat? It’s almost a parable: Everyone can see, no one can access, the millions trapped in the ether by a password known only to a dead man (Jonathan Bartlett)
How Bitcoin works: The social value of trust (Jonathan Bartlett)
Chair of Forbes media says money is about trust Facebook Likes cryptocurrency but the Faceless bureaucrats are not so impressed. Stay turned.