VPIN能否預測比特幣崩盤？

Could VPIN Predict Bitcoin's Crash?

本文欲探討Easley，Lopez de Prado and O’Hara（2012b）提出的「市場知情交易」（Volume-Synchronized Probability of Informed Trading, VPIN）是否與比特幣市場的日內價格崩盤有關？不論使用Chen，Hong and Stein（2001）的價格崩盤風險指標（Negative Coefficient of Skewness［NCSKEW］或Down-to-Up Volatility［DUVOL］）及Pöppe，Moos and Schiereck（2016）計算VPIN的Tick Rule或Bulk Volume Classification Rule，分別於Coinbase與Bitstamp兩家交易所在2015年1月1日至2018年12月1日期間的比特幣日內交易資料，實證結果皆顯示VPIN對比特幣價格崩盤風險有顯著的預測能力。另外，我們也發現回報率對崩盤風險有著顯著預測能力，惟若僅看當期流通成交量增加，則價格崩盤風險未必將會增加，雖然價格崩盤時往往伴隨回報率波動性放大，但是崩盤風險與波動性並非互為替代變數。

This study focuses on whether volume-synchronized probability of informed trading (VPIN) proposed by Easley, Lopez de Prado and O'Hara (2012b) has the ability to predict the short-term crash in the bitcoin market. VPIN is shown to have significant predictive power in both Bitcoin exchanges, Coinbase and Bitstamp, for crash risks during the period between January 1, 2015 and December 1, 2018 with both crash proxies, the negative coefficient of skewness (NCSKEW) and the down-to-up volatility (DUVOL), proposed by Chen, Hong and Stein (2001) and both trading direction classification algorithms, Tick Rule and Bulk Volume Classification Rule, examined by Pöppe, Moos and Schiereck (2016). Returns and their lags are also found to have significant predictive power for crash risks. However, an increase in trading volume does not necessarily imply that the risk of crash increases. The predictive power of turnover and return volatility for crash risks may depend on the features of bitcoin exchanges. In addition, we find that return volatility is not as proxy for crash risk although market crash often comes with increasing volatility.