This paper sets out to explore whether convergence and herding phenomena exist for digital currencies. Daily data cover a large spectrum of cryptocurrencies in separate bull and bear periods. Empirical estimations for detecting club convergence and clustering are performed by the methodology proposed by Phillips and Shu (2007, 2009). Econometric outcomes reveal preliminary evidence of powerful herding behaviour. The lowest of large-cap digital currencies attract mainly cryptocurrencies which are about in the middle of the large-cap and medium-cap categories whereas the highest-cap cryptocurrencies are parts of convergence clubs with mostly large-cap or purely medium-cap digital currencies during bear markets. Notably, segmentation is higher during bear markets as clusters are formed around more numerous cryptocurrencies than during bull markets. Convergence is stronger during flourishing periods. Secondary herding is also realized among pairs of clubs. Our findings enable investors to better diversify their portfolios and ameliorate their risk-return trade-off during extreme events.

本文着手探讨数字货币是否存在趋同和羊群现象。每日数据在独立的牛市和熊市期间涵盖了广泛的加密货币。菲利普斯和舒（2007，2009）提出的方法进行了检测俱乐部趋同和聚类的经验估计。计量经济学结果揭示了强大的放牧行为的初步证据。大盘数字货币中最低的货币主要吸引大约在大盘和中盘类别中间的加密货币，而大盘数字货币则是会聚俱乐部的一部分，其中大多数都是大盘或纯中盘数字货币在熊市期间。值得注意的是，在熊市期间的细分程度更高，因为集群围绕着比牛市期间更多的加密货币形成。在繁荣时期，收敛性更强。在成对的俱乐部之间也实现了次级放牧。我们的发现使投资者能够在极端事件中更好地分散投资组合并改善风险收益权衡。