Risks and returns of cryptocurrencies

Yukun Liu, Aleh Tsyvinski 06 September 2018



There has been a great deal of attention on cryptocurrencies both in the academic world and among practitioners. So far, academic attention has mostly focused on developing theoretical models (e.g. Cong and He 2018, Cong et al. 2018, Sockin and Xiong 2018, Schilling and Uhlig 2018, Abadi and Brunnermeier 2018). 

In a recent paper, we use textbook empirical asset pricing methods to compare cryptocurrency returns to those of other asset classes (Liu and Tsyvinski 2018). There are three broad questions that we are interested in that may contribute to the current popular debate and describe the stylised empirical facts that may be useful for developing theoretical models: 

  • Do cryptocurrencies behave similarly to other major asset classes? More specifically, do the same factors that determine returns of other asset classes also determine the returns of cryptocurrencies? 
  • What factors may predict the behaviour of cryptocurrency returns? 
  • The returns of which industries are exposed to cryptocurrency returns?

We find that cryptocurrency returns do not co-move with traditional asset classes. However, some cryptocurrency-specific factors –momentum and investor attention –strongly predict cryptocurrency performance. Lastly, we create an index of exposures to cryptocurrencies of 354 industries in the US and 137 industries in China.

Basic properties

We focus on three popular cryptocurrencies: Bitcoin, Ripple, and Ethereum. For the periods we investigate, these three cryptocurrencies exhibit high average return and high volatility. For example, the mean and standard deviation of the historical monthly Bitcoin returns are 21.60% and 69.46%, respectively, both of which are an order of magnitude higher than those of traditional assets. However, although cryptocurrencies’ Sharpe ratios are higher than those of traditional asset classes, they are not drastically so. Looking only at Sharpe ratios, cryptocurrencies appear similar to traditional asset classes.

Stocks, currencies, and precious metals commodities

We compare cryptocurrencies to traditional asset classes such as stocks, currencies, and precious metal commodities. We find that cryptocurrencies have no exposure to stock market risk factors. For example, they do not co-move with the Fama-French five factors (Fama and French 2016). Using a set of 155 documented factors in the finance literature (Feng et al. 2017, Chen and Velikv 2017), we find that only four out of the 155 factors are significant, but these four factors do not form any discernible patterns.

Two popular narratives about the functions of cryptocurrencies are that they may serve as a medium of exchange (as traditional currencies do), or as a storage of value (as precious metal commodities do). We investigate whether the markets perceive cryptocurrencies similarly to currencies or precious metal commodities and find that, in contrast to these two popular narratives, cryptocurrency returns have little exposure to currency returns or precious metal commodity returns.

Cryptocurrency-specific factors

While we find that cryptocurrencies do not behave like traditional assets, their returns can still be predicted by several cryptocurrency-specific factors. The two strongest cryptocurrency return predictors that emerged are time-series momentum and investor attention. 

Momentum is the phenomenon that when an asset increases in value, it then tends to rise even higher. This is a feature of many known asset classes, and we find that it is strongly present in cryptocurrencies. For example, for Bitcoin weekly returns, a one-standard-deviation increase in the current week’s return is associated with a 3.16% increase in returns over a one-week horizon.

We also find that when crypto-specific investor attention is abnormally high, cryptocurrency returns tend to increase. We use Google searches and Twitter postings to proxy for investor attention and find that for Bitcoin, for example, a one standard deviation increase in the current week’s Google searches leads to a 1.84% increase in return over the next week.

At the same time, many other potential cryptocurrency-specific factors that are commonly noted in the popular discourse do not affect cryptocurrency prices. These are the costs of mining, the cryptocurrency-specific proxy for the price-to-‘dividend’, and the cryptocurrency-specific realised volatility ratio.

Industry exposures

Media and investors have speculated that cryptocurrencies could potentially affect all kinds of industries. We take 354 industries in the US and 137 industries in China and investigate whether the returns of the stocks in these industries co-move with cryptocurrencies. We find that the healthcare and consumer goods industries significantly and positively co-move with Bitcoin returns, while the finance, retail, and wholesale industries do not. At the same time, there is evidence that Ethereum returns significantly and negatively co-move with the financial industry.


Abadi, J and M Brunnermeier (2018), “Blockchain economics,” Princeton University, mimeo.

Chen, A and M Velikov (2017), “Accounting for the anomaly zoo: A trading cost perspective,” Working Paper.

Cong, L W and Z He (2018), “Blockchain disruption and smart contracts,”NBER Working Paper 24399.

Cong, L W, Z He and J Li (2018), “Decentralized mining in centralized pools,”Working paper.

Fama, E F and K R French (2016), “Dissecting anomalies with a five-factor model,” Review of Financial Studies 29(1): 69–103.

Feng, G, S Giglio and D Xiu (2017), “Taming the factor zoo,” Working paper.

Liu, Y and A Tsyvinski (2018), “Risks and returns of cryptocurrency,” NBER Working Paper 24877.

Schilling, L and H Uhlig (2018), “Some simple bitcoin economics,”NBER Working Paper 24483.

Sockin, M and W Xiong (2018), “A model of cryptocurrencies,” Working paper.  



Topics:  Financial markets

Tags:  cryptocurrency, cryptocurrencies, Bitcoin, ripple, Ethereum, asset pricing models

PhD candidate in Economics, Yale University

Arthur M. Okun Professor of Economics, Yale University


CEPR Policy Research