Do dispersion-based tests capture herding? Evidence from cryptocurrency markets

Ünlü U., Bayram F.

FINANCE RESEARCH LETTERS, cilt.107, ss.1-20, 2026 (SSCI, Scopus)

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 107
  • Basım Tarihi: 2026
  • Doi Numarası: 10.1016/j.frl.2026.110397
  • Dergi Adı: FINANCE RESEARCH LETTERS
  • Derginin Tarandığı İndeksler: Scopus, Social Sciences Citation Index (SSCI), ABI/INFORM
  • Sayfa Sayıları: ss.1-20
  • Akdeniz Üniversitesi Adresli: Evet

Özet

This study re-examines herding behavior in cryptocurrency markets using two dispersion-based frameworks: the threshold approach of Christie and Huang (1995) and the nonlinear CSAD model of Chang et al. (2000). Using daily returns of major cryptocurrencies over the 2021–2023 boom–bust period, both models are estimated for the full sample and for rising and declining market conditions.
Across both methodologies, we do not find robust evidence of herding as captured by standard dispersion-based measures. Instead, dispersion increases with the magnitude of market returns, indicating that periods of market stress are associated with greater cross-sectional heterogeneity rather than convergence in returns. Robustness tests that control for persistence using a lagged CSAD term confirm that dispersion is persistent, while the nonlinear effect remains positive across specifications, with stronger statistical support during downturns.
Overall, the findings suggest that standard dispersion-based herding measures in cryptocurrency markets may be more informative about heterogeneous responses to common shocks than about coordinated trading behavior at the aggregate level during the sample period.

This study re-examines herding behavior in cryptocurrency markets using two dispersion-based frameworks: the threshold approach of Christie and Huang (1995) and the nonlinear CSAD model of Chang et al. (2000). Using daily returns of major cryptocurrencies over the 2021–2023 boom–bust period, both models are estimated for the full sample and for rising and declining market conditions.
Across both methodologies, we do not find robust evidence of herding as captured by standard dispersion-based measures. Instead, dispersion increases with the magnitude of market returns, indicating that periods of market stress are associated with greater cross-sectional heterogeneity rather than convergence in returns. Robustness tests that control for persistence using a lagged CSAD term confirm that dispersion is persistent, while the nonlinear effect remains positive across specifications, with stronger statistical support during downturns.
Overall, the findings suggest that standard dispersion-based herding measures in cryptocurrency markets may be more informative about heterogeneous responses to common shocks than about coordinated trading behavior at the aggregate level during the sample period.