In this paper, we analyze the relative impact of attention measures either on the mean or on the variance of Bitcoin returns by fitting nonlinear econometric models to historical data: Two non-overlapping subsamples are considered from January 1, 2012, to December 31, 2017. Outcomes confirm that market attention has an impact on Bitcoin returns and volatility, when measured by applying several transformations on time series for the trading volume or the SVI Google searches index. Specifically, best candidate models are selected via the so-called Box–Jenkins methodology and by maximizing out-of-sample forecasting performance. Overall, we can conclude that trading volume-related measures affect both the mean and the volatility of the cryptocurrency returns, while Internet searches volume mainly affects the volatility. An interesting side finding is that the inclusion of attention measures in model specification makes forecast estimates more accurate.

中文翻译：

---

市场关注是否会影响比特币的回报和波动性？

在本文中，我们通过将非线性计量经济模型拟合到历史数据来分析注意措施对比特币收益的均值或方差的相对影响：考虑了两个非重叠子样本（从2012年1月1日到2017年12月31日） 。结果证实，通过对交易量或SVI Google搜索指数的时间序列应用几种转换来衡量，市场关注度对比特币回报和波动性有影响。具体而言，通过所谓的Box-Jenkins方法并通过最大化样本外预测性能来选择最佳候选模型。总体而言，我们可以得出结论，与交易量相关的指标会影响加密货币收益的均值和波动率，而互联网搜索量主要影响波动率。