Non-standard errors in the cryptocurrency world
Date Issued
2024
Abstract
Motivated by recent findings from the equity market, we investigate non-standard errors in cryptocurrency research. We examine ten prevalent decisions related to data sources, sample preparation, and portfolio construction, generating 20,736 research designs for 43 sorting variables. Our findings reveal remarkable variation in portfolio performance tied to seemingly trivial choices. The non-standard errors in cryptocurrency studies not only surpass those in the stock market but also clearly exceed standard errors—though varying considerably across coin characteristics. Notwithstanding the above, the most prominent cryptocurrency factors, such as size and momentum, remain consistently robust across numerous specifications. Lastly, we find that reducing the influence of the smallest coins effectively decreases the non-standard errors.
Subjects
Cryptocurrency markets
;
Non-standard errors
;
Portfolio sorts
;
Data mining
;
p-hacking
;
Risk factors
;
Anomalies
Publisher
Elsevier
Institution
Type
journal article
Journal/Edited collection
Issue
92
Start page
103106
Language
German