Statistical analysis of Bitcoin price manipulation by Tether issuance in late 2017
In this article, we analyze the correlation between the printing of new tokens of a USD-backed dollar cryptocurrency called Tether and the upraise of Bitcoin prices in late 2017, constituting probably one of the latest large-scale market manipulations and bubbles whose consequences only recently we started to see. The ideas and analysis methods of this article are not new, they are all taken from or inspired by the recently published paper “Is Bitcoin Really Un-Tethered?” by John M. Griffin et al. and the so-called Tether Report. However, the results shown here are probably the currently most updated on the internet and, hopefully, a bit easier to understand than the cited publications. The Jupyter Notebook used for this post can be found in the end of the article.
Tether (or USDT)
Tether is a clearly controversial cryptocurrency (or token, to be more correct) born in late 2014 and whose more than shady creators claim to be backed 1:1 by USD dollars held in reserve by Tether Limited. For this reason, Tether is known to be a “stablecoin”, because its value fluctuates very little around the US dollar value. Interestingly, the Paradise Papers leaks in November 2017 connected Tether Limited to two of the Bitfinex officials, unveiling an array of dark bonds and mutually beneficial relationships between Bitfinex and Tether. Despite all the rumors and evidence that Tether Limited may actually not have enough dollars to back the amount of Tether issued, together with a subpoena issued by the U.S. Commodity Futures Trading Commission, the cryptocurrency has steadily kept its USD-pegged value and is trusted by traders all over the world that use it everyday as hedge to keep their value during cryptocurrency bear markets.
From a technical point of view, Tether does not own its on blockchain. Instead, it runs on top of the Bitcoin blockchain using something called the Omni Layer Protocol. In practice, just as Bitcoin, all the Tether transaction chain or ledger is public and can be accessed through https://www.omniexplorer.info. In fact, in an attempt to provide transparency, Tether Limited made available the list of the top richest Tether accounts in which appear several accounts with balances and annotations about the ownership of the accounts (most of them cryptocurrency exchanges). Tether Ltd. also made available the current reserve of USD and the Tether printer address which has the following id:
On the omniexplorer page of the Tether printer account, one can see all the events in which Tether has been issued or printed. In the following plot I show the Tether printing events (gray vertical bars) corresponding from March 2017 to July 2018, together with the cumulative printed Tether in red, the Bitcoin price evolution and the Bitfinex/Tether subpoena issuance the 6th of December of 2017. Bitcoin prices have been obtained as 1-h candlesticks from the Coindesk Bitcoin Price Index.
An interesting observation about the plot is that most of the Tether printing events, as well as the majority of the volume printed seem to happen from November 2017 to January 2018, precisely during the rise and burst of the Bitcoin bubble.
Statistical analysis of Tether manipulation
The hypothesis explored in depth in John M. Griffin’s publication as well as in the Tether Report is that Tether was printed and transferred to several exchanges as a response to Bitcoin dips or transient bearish markets in order to buy Bitcoin with it and provide a false sensation of bullish market that, in turn, would lead other investors to buy more Bitcoin as expected in a bullish-looking market. Therefore, the idea is that the Bitcoin bubble, that reached almost 20k in value, was a consequence of a synchronized orchestra of Tether printing and Bitcoin buying events in response to transient Bitcoin price dips.
If this hypothesis was true, one would expect that:
- Few hours before every Tether print, on average, prices of Bitcoin would go lower
- Few hours after every Tether print, on average, prices of Bitcoin would rise
A way we can solve this riddle statistically is by comparing the distribution of:
- (A) The difference on prices (price returns) between, for instance, 10 hours before each Tether printing event and the price during the printing event
- (B) All historical price differences (price returns) observed in 10h intervals.
Then, we can use a statistical test called Kolmogorov-Smirnov (KS) that tells us whether is statistically possible that the two distributions are different and, in fact, one is higher or lower than the other. This statistical test returns us something called the p-value, which tells us what is the probability that we sample the two set of samples A and B from the same distribution. Usually, we say two distributions are different when the p-value is lower than 0.05, meaning that there’s a 5% probability (very low) that we randomly get the two set of samples A and B from the same distribution. This suggests that the two distributions are statistically different with a 95% (1-pvalue) confidence.
Distribution of price differences before Tether printing
We performed two-sample one-tailed KS tests to see if the distributions of price difference between x hours before Tether printing and the price at Tether printing was lower from the distribution of all x hours price differences. We did this for x ranging from -48 hours to -1 hour. Note that, in this case, the KS test is one-tailed because we are trying to assess whether one distribution (e.g. price difference 10h before Tether print) is lower than the other (e.g. all 10h interval price differences) instead of just assessing if both distributions are different (this would be a two-tailed KS test). These are some of the most significant results:
One can see that in the cases in which the p-value is significant (p-value < 0.05) we can clearly see that the distribution is “shifted” towards the left or present several spikes on the left. The following plot shows all the p-values for x ranging from -48 hours to -1 hour.
This indicates that prices went down more than average before Tether printing, in consistence with our hypothesis that Tether was printed in response to Bitcoin dips.
Distribution of price differences after Tether printing
According to the initial hypothesis, once the Tether had been printed and distributed to several exchanges, Tether Limited would use the USDT tokens to buy Bitcoin back. This way, one would expect that the distribution of price differences between +x hours and the time of Tether printing should be displaced towards the right, i.e. should be bigger than the distribution of all x hour price differences. The range of x we used is from +1 hour to +48 hours. These are some representative distributions:
And this is the complete plot of all the p-values for the specified range of x:
We can see that at least 6 points out of 48 show statistical differences. This indicates us that there is evidence that after Tether was printed, Bitcoin prices raised more than average.
Correlation vs Causation
An interesting dilemma presented at this point is: while there seems to be a correlation between Tether printing after prices going down and prices going up after Tether printing, how can we truly know that Tether printing actually causes these changes? The answer is hard to assess in this case. There has been plenty of studies where has been shown that correlation does not necessarily imply causation, some of them funny such as the correlation found between number of people who drowned by falling into a pool and films Nicolas Cage appeared in. While the debate may be interesting, the topic is out of the scope of this article (leave your thoughts in the comments!).
Consequences of Bitcoin manipulation via Tether
If our considerations and hypothesis are correct, Tether Limited would have manipulated all the cryptocurrency markets (remember that all cryptocurrencies show great correlation with Bitcoin) and would have greatly contributed to the Bitcoin and cryptocurrency bubble. The question now is: how much? While it is really hard to assess with exactitude, I have ventured to calculate some simulations of how much the markets would have been affected by it.
To do so, I have considered that if Tether Limited would have not manipulated the market, the market would have not experienced all the price raises that we detected before as “statistically significant”. Therefore, I have removed all the events of the statistically different distributions, consisting of a total of 341 candlesticks (approx. 14 days). Bare in mind that these distributions do not only include events in which the price raised but also candles in which price went down. To be consistent and comparable with the real Bitcoin price, I removed the price change experienced during the “deleted” candlesticks from all the subsequent candlesticks. Finally, I have interpolated the price of the deleted candlestick with the values of the neighbor candles.
The price obtained by these simulations, using alpha=0.05 and alpha=0.1 (alpha is the significance level), as well as the original Bitcoin price can be seen in the following plot.
In an attempt to quantify the changes observed in the simulations, I have calculated the price differences for the all-time-high price of bitcoin (17th of December 2017) as well as differences between real price and simulation as of today.
What we can see is that Tether manipulation is the responsible of a 10-70% increase in Bitcoin prices. These numbers are consistent with the ones reported by John M. Griffin et al., who reports as 50% the contribution of Tether manipulation to the price increase.
Conclusions and personal opinion
In this article we have explored the effect of Tether printing in the manipulation of Bitcoin price. Some of the take-home messages are:
- The statistics support the theory that Tether Limited and Bitfinex corporations used Tether to buy Bitcoin in key occasions during the meteoric rise of Bitcoin prices of 2017 and beginnings of 2018.
- These corporations would print Tether right after transient Bitcoin price dips and would distribute it to an organized network of accounts in different cryptocurrency exchanges (explored in detail in John M. Griffin paper).
- Tether Ltd. would then buy Bitcoin with the freshly minted Tether and would promote the creation of a fraudulent bullish market, which would attract more investors to buy Bitcoin, contributing this way to increase the bubble (momentum effect).
- As part of the strategy, and supported by the paper mentioned in this article, Tether Ltd. would send the freshly bought Bitcoin to their accounts in Bitfinex.
- In my personal opinion, Bitfinex then would either (a) slowly sell the Bitcoin in their own exchange and other exchanges or (b) would sell the Bitcoin to private investors over-the-counter (OTC). This last kind of operation (OTC) would be highly profitable for Bitfinex because it would allow to monetize the whole operation (in dollars) while preventing the side effect of generating bearish markets in the centralized exchanges by selling their Bitcoin.
The information provided is for educational purposes only. By no means it represents any financial advise and the information must be taken “as is” without guarantee of any kind. I do not hold any relationship with any of the parts mentioned in this article.
Dear programmer, you can now find the Python implementation used to write this post in this Jupyter Notebook: https://github.com/GerardBCN/TetherManipulationStudy/blob/master/manipulation_explorer.ipynb. I’m open for suggestions and debugging!