Dynamic analysis of calendar anomalies in cryptocurrency markets: evidences of adaptive market hypothesis

ABSTRACT

This paper analyses the effects known as the day of the week and the month of the year in the cryptocurrency markets. The closing values of eleven cryptocurrencies have been considered. The study employs dummy variable regression techniques, ANOVA and Friedman tests for assessing two calendar anomalies, the day-of-week and month-of-year effects. To test these calendar effects, we have applied both full sample and rolling-regression techniques for two lengths of the rolling sample intervals. Furthermore, we have examined the existence of long memory in day-of-the- week and month-of-the-year cryptocurrency returns. The results provide evidence about the existence of day-of-the-week and month-of-the-year effects in cryptocurrency returns, in particular, on Thursdays and in November. In addition, it should be added that the general results of the current study show that the calendar effect in the cryptocurrency market is dynamic rather than static, which indicates that the calendar effect is a phenomenon that varies over time.

KEYWORDS:

• Calendar anomalies
• cryptocurrencies
• market efficiency
• dummy regression
• adaptive market hypothesis
• Hurst exponent

JEL CLASSIFICATION:

• C12
• C22
• G14
• G15

Acknowledgements

This work was supported by the Spanish Ministerio de Ciencia e Innovación under Grant PID2020-113367RB-I00; Young Talent Research Project UNED 2022 under Grant 076-044355 ENER-UK. I am grateful to the Editorial Team and two anonymous referees for useful comments and suggestions.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. https://coinmarketcap.com/.

2. Some authors have analysed the adequacy of data in the cryptocurrency market. Thus, Alexander & Dakos (2020) showed that there could be problems with the different kinds of data sources; to solve this problem, the authors recommend scholars should only consider data from exchanges, which is quite restrictive. Vidal-Tomás (2022) showed that most of the datasources provide price series that are statistically equal, provided the cryptocurrencies are large, and the exchanges included in the calculation of prices are liquid. Investing creates a weighted price series using the exchanges that they show on the website. Given that Binance (the largest and most liquid exchange in cryptos) is included in the calculation, the time series used in this study should be appropriate and the results should not be different to other papers. https://www.investing.com/crypto/.

3. Tables 8 and 9 only present the p-values. It has been decided not to present the estimation of all the coefficients to save space. In those cases in which the coefficients are statistically significant, a mention of their sign is made. Full results of estimates of all coefficients for all currencies are available to anyone interested in them upon request.

Additional information

Funding

The work was supported by the Spanish Ministry of Science and Innovation [PID2020-113367RB-I00]; Universidad Nacional de Educación a Distancia [076-044355 ENER-UK].