Akademik Veri Yönetim Sistemi
7th International Conference on Energy Economics and Energy Policy (ICEEEP 2023), Barcelona, İspanya, 28 - 30 Nisan 2023, ss.1-8, (Tam Metin Bildiri)
This paper determines the existence of “herd behavior” and the factors of
herding in cryptocurrency markets. We examine the causability between herding
and sentiment signals such as (FoMO, hopeful, negative, positive, and uncertain)
as well as between herding and volatility and valume as a measure of overconfidence
for Bitcoin. Our dataset covers January 1, 2019 – September 10, 2022,
consisting of COVID 19 pandemic. First, we use intraday aggregate trade data
and construct daily herding intensity statistics for negative, positive, and
zero trades in the sense of Patterson and Sharma (2006). Then, we compute
realized volatility series exploiting the Parkinson’s (1980) range-based
measure. Following Balcilar et al. (2017), the volatility series are
log-detrended. The sentiment signals are smoothed using the exponential
smoothing (ETS) models in the statsmodule Python module. Lastly, we
estimate Fourier-type Granger causality tests developed by Nazlioglu et al.
(2019). Our results show the bi-directional causality between sentiment signals
from Twitter and the herding intensity statistics at the conventional
significance levels. The sentiment signals from Reddit have a limited impact on
the herding statistics, but most of the signals significantly cause the
volatility measures. Bitcointalk sentiment signals do not cause any herding,
volatility, and volume measures. Our results provide important implications for
investors and portfolio managers interested in cryptocurrency investments