A singular value decomposition approach for testing the efficiency of Bitcoin and Ethereum markets

This letter revisits the informationally efficiency of the two major cryptocurrencies Bitcoin (2013- 2021) and Ethereum (2016-2021). The analysis is based on the computation of the singular value decomposition (SVD) entropy of a matrix formed by lagged vectors of price returns. The computed entropy is compared with a reference obtained from uncorrelated time series to decide whether the rows of the lagged matrix are uncorrelated. The procedure was implemented over a sliding window to assess the time variations of the entropy. The results show that the markets are informationally efficient most of time over different scales, except for some short periods that are linked to the 2016-2017 price-boom period and the 2020 Covid-19 economic downturn. (C) 2021 Elsevier B.V. All rights reserved.