Causal relationships between cryptocurrencies: the effects of sampling interval and sample size

Abstract

Abstract For this paper, the relationship between seventeen popular cryptocurrencies was analyzed by multivariate Granger causality tests and simple linear regression, using data spanning the period 1 September 2020 to 8 December 2021. The novelty of this work is that it studies the effects of sampling interval and sample size in cryptocurrency markets, which can yield significantly different results. Minute-by-minute, hourly and daily data were collected to examine the Granger causality relationship between cryptocurrencies. It was found that all the currencies demonstrated a significant causality relationship when high frequency (such as minute-by-minute) data was used, in contrast to hourly and daily data. The bigger the sample size, the higher the probability of rejecting the null hypothesis. Hence, the null hypothesis for the Granger causality test can be rejected for minute-by-minute time series data because of too large a sample size. Granger causality test results for hourly and daily data indicated that Bitcoin, Ethereum Classic, and Neo were leading indicators among the cryptocurrencies included in the research. In addition, according to simple linear regression analysis, the short term marginal effect of Bitcoin plays an important role by creating significant impacts on other cryptocurrencies.