Author Details

Daniel Barrera

Daniel Barrera
Senior Associate, MSCI Research

Simon Minovitsky

Simon Minovitsky
Vice President, Equity Research Team

Social Sharing

Extended Viewer

Could Factors Have Explained Cryptocurrency Risk?

  • Given the proliferation and popularity of cryptocurrencies, we test whether factors that are important in the cross section of equity returns were also important for cryptocurrencies.
  • We created a multifactor model with seven factors and found that basic price- and market-based factors such as size, volatility, beta and momentum were important in explaining systematic risk for cryptocurrencies.
  • The low-liquidity and size factors reported the highest cumulative returns over our study period and also during the last year.

With thousands of cryptocurrencies in existence and their rise in popularity as investment vehicles, the need for standardized investment tools applied to the crypto market has increased as well. Indeed, academics have already started to apply standard asset-pricing tools, such as factor models.1 These academic studies found that some factors that are important in explaining equity returns, such as market beta, size and momentum (specifically, reversal or short momentum), were also important in explaining the cross section of cryptocurrency returns.

To build on these studies, we created a multifactor model that contains seven technical factors based on price, trading volume, age and market capitalization.2 While the age factor is not commonly used in the equity market, all the others are traditional equity factors that have a long history of inclusion in equity models.

 

Factors in the Multifactor Cryptocurrency Market Model

FactorDescription
Market Intercept of the cross-sectional regression. Exposure=1 for every asset.
AgeLog of days since cryptocurrency inception
SizeLog of market capitalization
LiquidityLog of average traded volume over the last month
MomentumRelative strength (cumulative return over the last 90 days) and historical alpha
BetaHistorical beta from the capital asset-pricing model (CAPM), applying the exponentially weighted moving average (EWMA) within a six-month window and three-month half-life
Residual VolatilityCumulative return range and daily standard deviation of returns

Alpha, beta and sigma are outputs of a previous time-series CAPM-style regression against a market-capitalization-weighted index (arbitrarily reducing weights for Bitcoin and Ethereum) comprising the top 100 cryptocurrencies.

Given the high level of concentration in the crypto market — Bitcoin comprises 44% of the total market capitalization while Ethereum takes a share of 19% — we had to determine a crypto-specific weighting scheme for the regression. We investigated a variety of approaches and narrowed it down to two: equal weights and the traditional square root of market capitalization modified by arbitrarily underweighting Bitcoin and Ethereum. Equal weights, however, gave disproportionate importance to smaller cryptos with extreme returns and underweighted Bitcoin and Ethereum too much, given their importance in the opportunity universe. In the end, we went with the modified square-root approach. This option ensured that the estimated factors were not dominated by the top two coins, but instead represented a more diverse estimation universe.

 

What Did We Find?

Over our sample period of July 3, 2017, to March 24, 2021, our seven-factor model yielded an average R2 of 0.45, indicating these seven factors explained 45% of the variation in the cross section of cryptocurrency returns. While this number is well within the historical range found in our equity-factor models, it should be interpreted with caution given the short history of the available data and our limiting the study set to 100 cryptocurrencies. In contrast, our equity models contain thousands of stocks as part of their estimation universe and more than 20 years of historical data.

In our short data history for cryptocurrencies, the characteristics that showed a distinguishable premium were size and low liquidity. Size and low liquidity also provided the most-positive information ratios (risk-adjusted return), while low momentum and low beta followed closely. Factor returns can vary over time and tend to move in cycles, but it’s too soon to tell if these factor trends will continue or move in cycles as well.

 

Cumulative Returns for the Seven-Factor Model

Model estimated from July 3, 2017, to March 24, 2021. Source for prices, volume and market capitalization used as input data to the model: Coinmarketcap.com

 

Sample Statistics of Tested Factors

 t-statt-stat>2 Avg Return (%) Volatility(%) IRCorr Mkt 
Market5.9472.1589.6082.191.091.00&
Beta2.04 39.60 -14.91 39.30 -0.38 0.38
Age1.5527.55-9.7696.21 -0.10 -0.08
Residual Volatility1.4626.898.1537.220.220.19
Liquidity1.4324.32-56.0541.12-1.36-0.13
Size1.4124.4722.3529.390.760.17
Momentum1.3723.81-18.2434.73-0.53-0.14

Model estimated from July 3, 2017, to March 24, 2021. Source for prices, volume and market capitalization used as input data to the model: Coinmarketcap.com

In terms of explanatory power, we found that beta and age were the strongest factors in our backtest as measured by t-statistics. Notably, the age factor also presented the largest volatility out of all seven factors, which made it the leading candidate to explain risk of the represented coins.

 

Individual Coins’ Exposures to Specific Factors

As of March 24, 2021, the final day of our study period, we observed that Bitcoin’s exposures to size, liquidity, age and momentum were high compared to other cryptocurrencies, as the standardized values were above zero, while its exposures to residual volatility and beta were relatively low. In contrast, Dogecoin’s exposures to all seven factors increased steeply in 2021 as its popularity spiked, reflecting an increase in total risk. The historical exposures of Bitcoin and Dogecoin are shown in the exhibit below.

Knowing the exposures of an individual currency or portfolio along with the factor returns allows investors to understand which risks had the most influence on the returns of individual assets in any past period.

 

Factor Exposures for Bitcoin and Dogecoin

Bitcoin

Model estimated from July 3, 2017, to March 24, 2021. Source for prices, volume and market capitalization used as input data to the model: Coinmarketcap.com

Dogecoin

Model estimated from July 3, 2017, to March 24, 2021. Source for prices, volume and market capitalization used as input data to the model: Coinmarketcap.com

 

Factors Have Potential to Explain Cryptocurrency Risk

We found that traditional asset-pricing tools and factors showed promise in analyzing the cross section of cryptocurrency returns. Moreover, we believe these factors can provide an interesting first approach for estimating cryptocurrency risk and the correlations among them. In our next post in this series, we will explore the efficient frontier of cryptocurrency portfolios using our prototype model.

The authors thank Ian D’Souza for his contributions to this post.

 

 

1See Yukun Liu, Yukon, Tsyvinski, Aleh, and Wu, Xi. 2019. “Common Risk Factors in Cryptocurrency.” National Bureau of Economic Research.
Shen, Dehua, Urquhart, Andrew, and Wang, Pengfei. 2020. “A three-factor pricing model for cryptocurrencies.” University of Reading.

2It is difficult to define “fundamental” factors such as value or quality in the cryptocurrency market.

 

 

Further Reading

Bitcoin: Good as Gold?

Foundations of Factor Investing

Regulation