What determines cryptocurrencies’ price?

Uncertainty surrounding cryptocurrencies has been widely increased by news spread and high variance with bubbles followed by strong bursts particularly in late 2017 and early 2018. This uncertainty has made it particularly demanding to find definitive evidence on whether they can be considered a mean of exchange or rather a speculative investment. Furthermore, a question has risen: what determines their price and how does it evolve along time? In turn, their high volatility makes it harder to understand if their risk-adjusted return is higher than that of the global market portfolio. These questions have been answered in a Swedish research (Jenny Asplund and Felicia Ivarsson, 2018) by using three types of variable with respect to Bitcoin, Ethereum, Ripple and Litecoin: a) macro-financial variables (MSCI index of the 23 stock exchanges of developed countries, gold price, WTI oil price index, euro, South-Korean won, yean exchange rates with respect to the US dollar (as these four currencies are the most used in the four above mentioned cryptocurrencies cross-transactions); b) supply variables (market capitalization) and demand variables (exchange volume); c) popularity variables like the number of visualization of Wikipedia pages of the four analyzed cryptocurrencies and the number of times when at least one the words “Bitcoin”, “Ethereum”, “Litecoin”, “Ripple”, “Cryptocurrency” and “Cryptocurrencies” have been mentioned by an online article.
In their econometric models, the two researchers also use US sovereigns’ three-month yield as a control variable representing the risk-free rate, and a proxy of the macroeconomic situation.
Using a classic Ordinary Least Square (OLS) model[1], they find that supply and demand variables and popularity variables show significant effects on the four cryptocurrencies’ prices. Only one of them, namely online articles, has a negative sign. This is a strong evidence that – at least in the period analyzed – those articles had a negative stance. Oppositely, macro-financial factors do not seem to have any significant effect on those prices, except a negative and statistically strong relation with the US dollar/yen exchange rate.
As for the latter inquiry – i.e. their risk-adjusted return -, the so called M2 measure[2] suggests that investors are well compensated for the cryptocurrencies’ risk in a one year period with some exceptions: for example, Ripple between September 2015 and September 2016 and Ethereum between September 2017 and April 2018. If the same analysis is carried out monthly, the market risk-adjusted return of the four cryptocurrencies is more in line with the return of MSCI marker portfolio. In hindsight, this is evidence of a stronger convenience to hold cryptocurrencies with a longer time horizon, while at a monthly level, their return has been subject to sharpened volatility.
To sum up, the authors show that there is no correlation between crypto- and fiat currencies. Unfortunately, this result does not come out in favor of their role as a store of value and thus they are still unsuitable to substitute fiat money, while being a good speculative tool. Furthermore, as their price is driven by “soft” (i.e. popularity, supply and demand variables) rather than “hard” (macro-financial) factors is an evidence of low market risk but a high idyosincratic one, enabling them to be a good hedge. The risk-adjusted annual return (not the monthly return) is almost always higher that that of the MSCI market portfolio although decreasing.

[1] Firstly, the researchers evaluate panel data funding hypotheses, ruling out the presence of correlation between the error term and the independent variables and thus accepting the superiority of the random effects model with respect to the fixed effects model (where that correlation is non zero). Secondly, they prefer the OLS to the random effects, because they cannot find any strong within-variance in the times series of the four cryptocurrencies. To find out more on the econometric literature on static panel data models using the software Stata © see Torres-Reyna (2007). For theoretical, deeper analyses on the difference between fixed and random effects models see Wooldridge (2014).
[2] The M2 index measures the Modigliani’s risk-adjusted performance, given by the Sharpe ratio multiplied by the annualized standard deviation of the reference benchmark (or the market portfolio, which in this case is the MSCI world index). Having multiplied these two measures, the risk free rate (i.e. the three-month US sovereign yield) is added.