In the Bitcoin community, Bitcoin price analysis is a growing trend—although it is often the target of many jokes and sarcastic comments. This sarcasm and cynicism towards Bitcoin price analysis does not lack justification, however; most analysts make dreadfully inaccurate predictions and fail to use data that is actually useful, a practice that most assuredly comes from employing the improper methodology.
But we will not get into the problems with the Bitcoin price analysis community in this article, since we have discussed them at length elsewhere. Instead, this article will be the first installment in a two-part series on how far the mainstream is lagging behind the Bitcoin community in terms of their understanding of Bitcoin. This article specifically focuses on a recent research paper coming from a Swiss university that studies fluctuations in the Bitcoin price as they relate to changes in social awareness of the digital curency. Throughout the course of this article, we will discover that the Swiss researchers ultimately added nothing of value to discussions on Bitcoin and also employed extremely flawed economic theories and methodologies.
Swiss Study of Bitcoin Price Fluctuations and their Connection to “Social Signals:” Why this Study is Valueless and Where it Went Wrong
The inspiration of this article comes from a recent article from CoinDesk, which reported on a Bitcoin study that was recently published by ETH Zürich—a university in Switzerland that is a leader in technology education. This study gathered empirical data from mid 2010 to November of 2013, at the start of the Bitcoin price peak, just a few months before Mt. Gox crashed, and employed several different statistical tools to analyze this data.The research team took four datasets, treated them as variables, and superimposed them upon each other in order to find correlations and interdependences between the four variables as they grew and shifted throughout the time period considered in the study. The first of the four variables were downloads of the Bitcoin client and the Bitcoin blockchain. The research team analyzed the activity on the blockchain and tracked the downloads of the Bitcoin client to generate an approximate number of new Bitcoin users.
Secondly, the team compiled data on the exchange rates between Bitcoin and three different, fiat currencies and three different exchanges, respective of the currencies. They used the Mt. Gox Bitcoin price, denoted in United States dollars (USD); they tracked Chinese activity through BTC-China, denoted in Chinese renminbi (CNY); lastly, they tracked the activity of BTC-de—a European Bitcoin exchange—which is denoted in the Euro (EUR). Both Mt. Gox and BTC-de allowed for trade in EUR, so the research team tracked the EUR movement in both exchanges to study European Bitcoin activity.
The third variable used by the team was Internet search information. They compiled data on the amount of searches done on Google for the term “bitcoin.” As an alternative, they used searches on Wikipedia, the world’s largest online encyclopedia.
The fourth, and final, variable used in the study was word-of-mouth (WOM) information sharing. To track the growth in WOM Bitcoin information sharing, they measured the number of Bitcoin-related tweets per 1 million posts on the team’s Twitter feed. As an alternative, the study used the number of “shares” for the posts on the oldest, from what the researchers could gather, Bitcoin-related Facebook page: http://www.facebook.com/bitcoins.
Then the researchers employed their chosen methodology to analyze the data. We will not go into that methodology here, for it is little more than myriad statistical models and empirical analysis that holds no weight in the world of economics. In fact, as we will be arguing, their lengthy, empirical research project did not elicit data that the Bitcoin community had not already figured out on their own. In a later part of this article, though, we will briefly cover the flaws of empirical economics and the reader will be referred to a few, previous articles, written by this author, that have provided a more in-depth analysis of the fallacy of empirical economics.
However, the researches did use one analytical method, that this author thinks is very interesting, which we will outline here. The research team attempted to identify a “fundamental” value for a single bitcoin, so that they could have a baseline measurement of Bitcoin value that they could compare to the actual Bitcoin price as the market responded to “social signals.” They conceded that it was difficult to pin down the precise “fundamental” value of a bitcoin; but, they also argued that this value had to be at least the production cost of one bitcoin, so they used that cost as the benchmark for “fundamental” value. This particular method rests upon an age-old, economic fallacy: the cost-of-production theory of value . We will discuss this fallacy later on in this article, when we identify the problems with the research project as a whole.
A direct quote from the research paper regarding the fundamental Bitcoin value:
It is difficult to calculate an estimate of the fundamental, or intrinsic, value of one bitcoin, which is different to its ‘fair’ value . However, we argue that the fundamental value of one bitcoin equals at least the cost involved in its production (through mining), and therefore that we can use this cost as a lower bound estimate of the fundamental value. This definition has the advantage of being independent from any subjective assessment of future returns.
The Findings of the Zürich Bitcoin Study
After laying out the chosen methodologies and datasets to be used while conducting this Bitcoin price study, the authors of the paper then spent several pages going over the results of their study. Rather than going over these results in detail, for they are riddled with unnecessary technical jargon that does not actually provide any valuable data to Bitcoin economics, we will give a brief summary of the findings and provide a few direct quotes from the text.So, to briefly summarize the results of the study, the researcher’s essential findings were that there existed a positive correlation between increased Google searches, increased WOM information sharing, and an increasing Bitcoin price. They double checked this correlation by running their statistical models for both sets of Internet search and WOM sharing variables—with Wikipedia serving as the alternative to Google and Facebook serving as the alternative to Twitter. Even when looking at Wikipedia searches and Facebook shares, increased activity on both platforms positively correlated with increases in the Bitcoin price.
In addition to their “discovery” of this connection between social signals and Bitcoin price, they found that this correlation actually creates two positive feedback loops. The first feedback loop is that increased Internet searches and social media shares was followed by an increase in the price, which was then followed by more searches and shares, etc. The second loop is one that involves growth in new Bitcoin users. They found that increased searching and sharing led to an increase in adoption, which created an increase in the Bitcoin price, leading to more searching and sharing, etc.
These feedback loops did not go on indefinitely, however. The study also found a correlation between peak Internet searching and WOM sharing and a falling Bitcoin price. Several times during the time period studied by these researchers, searching and sharing would reach a peak, after which the Bitcoin price would “crash.” After discovering this correlation, the study came to the conclusion that these social signals are definitely linked to, and play a significant role in, the “bubble cycle” of Bitcoin; increased social awareness leads to an explosion in the Bitcoin price, which is followed by a sharp drop in the price, along with a decrease in searching and WOM sharing.
Here are some direct quotes from the paper on the findings of the study:
We disentangle the feedback cycles in our system by means of a VAR , which captures time-dependent multidimensional linear relationships between the four variables of the analysis, with a lag of 1 day. . . The VAR reveals the following feedback cycles:
– ‘social’ cycle: search volume increases with price ( f P,S 1/4 0.386), word of mouth increases with search volume ( f S,W 1/4 0.243), and price increases with word of mouth ( f W,P 1/4 0.1). Simultaneously accounting for all dependencies between the four variables emphasizes the influence of word of mouth on price, revealing a stronger relation than cannot be observed with pairwise correlation analysis (more details in the electronic supplementary material, §S3). The three-way loop between S t , W t and P t represents the feedback cycle between social dynamics and price in the Bitcoin economy.
– ‘user adoption’ cycle: search volume increases with price ( f P,S 1/4 0.386), the number of new users increases with search interest ( f S,U 1/4 0.158) and price increases with increases in user adoption ( f U,P 1/4 0.137). This second three-way loop between S t , U t and P t models how the exchange rate of Bitcoin to other currencies depends on the number of users in the Bitcoin economy
In addition to these two cycles, we find a negative relation from search to price ( f S,P 1/4 20.233). This is illustrated by a clear dyadic relation between the two variables’ extremes: three of the four largest daily price drops were preceded by the first, fourth and eighth largest increases in Google search volume the day before.
The cycles presented above provide an explanation for the generation of bubbles in the Bitcoin economy. Recent findings indicate that the driving forces behind Bitcoin prices changed since its invention , motivating our decomposition of the study period into characteristic time windows, each of which corresponds to a distinct bubble. We do this by estimating a lower bound for the fundamental value of Bitcoin: we approximate the energy cost of producing one bitcoin, which is derived directly from the Bitcoin difficulty  (see Material and methods). Throughout our study period, the price stayed almost always above the fundamental value (figure 3a: the trajectory of the weekly weighted mean price is almost exclusively on the left of the price/fundamental equality line). The trade of bitcoins at a much higher price indicates the possible presence of a bubble , and the events at which the market price starts diverging from the fundamental value mark the beginning of bubbles.
Lastly, a quote on how the researchers believe that this type of quantitative analysis can be useful in predicting the future Bitcoin price and future Bitcoin bubbles:
The statistical technique we used in this paper thus proves to be a robust way of identifying the coupled dynamics of the socio-economic variables we study. It also produces accurate estimates of the future levels of any variable (including price and word of mouth) based on the past history of the system.
Continued on Page 2