Cryptocurrency return reversals

ABSTRACT

Analysing a set of 200 cryptocurrencies over the period from 2015 to 2019, we document a significant return reversal effect that holds at the daily, weekly, and monthly rebalancing frequencies and is robust to controls for differences in size, turnover, and illiquidity. Moreover, the reversal effect persists during both halves of our sample period and following periods of both high and low market implied volatility. Consistent with the effect being driven by a combination of market inefficiency and compensation for liquidity provision, we find reversals are most pronounced among smaller capitalization and less liquid cryptocurrencies.

KEYWORDS:

• Cryptocurrency
• short-term reversal
• market efficiency
• market liquidity
• return predictability

JEL CLASSIFICATION:

• G11
• G14

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ Details available at: https://coinmarketcap.com/methodology/#market-data

² To limit possible microstructure effects we remove observations with monthly turnover of less than 1% or corresponding rates for our daily and weekly tests. The results are robust to imposing no turnover requirement or higher requirements.

³ We exclude the cryptocurrency Vertacoin (VTA) from our analysis, as it represents an extreme outlier with a monthly time series standard deviation of at least 20 times more than any other coin; however, our results are robust to its inclusion.

⁴ To ensure our results are not driven by outliers, we recompute average long-short portfolio returns when excluding observations in excess of 2.24 standard deviations from the mean as recommended by Aguinis, Gottfredson, and Joo (2013). The loser-minus-winner return spread remains highly significant at 6.78% (t = 37.70), 8.69% (t = 9.60), and 15.72% (t = 3.54) for daily, weekly, and monthly frequencies, respectively.

⁵ The monthly loser-minus-winner return spread is also economically large among small, liquid cryptocurrencies at 12.15% (t= 1.45), but the lack of statistical significance may partially reflect the reduced power of subsample tests.

⁶ In unreported tests, we also partition the sample using the Global and U.S. Economic Policy Uncertainty (EPU) Indexes and find similar results with a large and significant reversal effect following periods of both high and low uncertainty.