Beyond Conventional Sentiment Indicators: Cryptocurrency’s Hidden Potential in VIX Forecasting

These data and codes replicate all the empirical results presented in the paper "Beyond Conventional Sentiment Indicators: Cryptocurrency’s Hidden Potential in VIX Forecasting" by Ming Gu, Juan Lin, and Siyi Liu.

1. Input Data Files (/Data)
This file contains the main dataset and sub-dataset (/Data/Controls) used for the empirical analyses in the paper.
(1) match_BTC.dta: Daily decomposed return of Bitcoin from January 2018 to December 2025, aligning with the schedule of US market, based on 5-min OHLCV data for matching that can be obtained using paid API from Coindesk.
(2) match_ETH.dta: Daily decomposed return of Ethereum from January 2018 to December 2025, aligning with the schedule of US market, based on 5-min OHLCV data for matching that can be obtained using paid API from Coindesk.
(3) Bitcoin_active_addresses.xlsx: Daily unique active addresses of Bitcoin from Glassnode.
(4) Ethereum_active_addresses.xlsx: Daily unique active addresses of Ethereum from Glassnode.
(5) gs_btc: Google search volumes of the term “Bitcoin”.
(6) gs_eth: Google search volumes of the term “Ethereum”.
(7) VIX_History.xlsx: Daily prices of the CBOE VIX Index.
(8) VXD_History.xlsx: Daily prices of the CBOE DJIA Volatility Index.
(9) VXN_History.xlsx: Daily prices of the CBOE NASDAQ Volatility Index.
(10) VXX_HistoricalData.xlsx: Daily prices of the iPath Series B S&P 500 VIX Short-Term Futures ETN.
(11) VIXY_HistoricalData.xlsx: Daily prices of the ProShares VIX Short-Term Futures ETF.
(12) fomc.xlsx: The dummy of pre-schedule FOMC announcement day.
(13) SPX_FUT.xlsx: Daily prices of S&P 500 (SPX) futures from Wind.

2. Code Files (/Codes)
(1) Stata codes (/Codes/Stata_codes): Obtain_Crypto_Sum.do and OtherVolatility_Merge.do provide the processed data for analysis. Summary_Statistics.do describes the decomposed cryptocurrency return and VIX changes. Univariate_Sort.do check the overnight cryptocurrency return as a sentiment proxy. Regression_Insample.do gives the in-sample forecasting for the intraday change of VIX. Regression_OutofSample_Bitcoin.do and Regression_OutofSample_Ethereum.do give the out-of-sample forecasting for VIX involving Bitcoin and Ethereum, respectively. OutofSample_Pandemic.do and OutofSample_NonPandemic.do assess the out-of-sample forecasting within and without the impact of COVID-19. Strategy_Performance.do and Strategy_Performance_VIXY.do assess the economic benefits of strategy based on out-of-sample forecasting. Cumulative_Gain.do draws the cumulative gain of the best strategies selected. ExpandingWindow_OutofSample.do give the out-of-sample forecasting with an expanding window.
(2) R codes (/Codes/R_codes): mcstest.R applies the MCS test for out-of-sample forecasting.

3. Results Saving file (/Save): Save the essential results of every step of empirical analysis, the sub-file (/Save/Predict) saves the forecasting losses for MCS test.

本数据集与代码可复现Ming Gu、Juan Lin与Siyi Liu所发表论文《超越传统情绪指标：加密货币在VIX预测中的隐藏潜力》中的全部实证结果。

1. 输入数据文件（/Data）
本文件夹包含论文实证分析所用的主数据集与子数据集（/Data/Controls）。
(1) match_BTC.dta：2018年1月至2025年12月的比特币每日分解收益率，匹配美国市场交易时段，数据基于5分钟OHLCV（开盘价、最高价、最低价、收盘价、成交量）行情，可通过Coindesk付费API获取。
(2) match_ETH.dta：2018年1月至2025年12月的以太坊每日分解收益率，匹配美国市场交易时段，数据基于5分钟OHLCV行情，可通过Coindesk付费API获取。
(3) Bitcoin_active_addresses.xlsx：来自Glassnode的比特币每日唯一活跃地址数。
(4) Ethereum_active_addresses.xlsx：来自Glassnode的以太坊每日唯一活跃地址数。
(5) gs_btc：关键词"Bitcoin"的谷歌搜索量。
(6) gs_eth：关键词"Ethereum"的谷歌搜索量。
(7) VIX_History.xlsx：芝加哥期权交易所（CBOE）波动率指数（VIX）的每日收盘价。
(8) VXD_History.xlsx：芝加哥期权交易所（CBOE）道琼斯工业平均指数波动率指数（VXD）的每日收盘价。
(9) VXN_History.xlsx：芝加哥期权交易所（CBOE）纳斯达克波动率指数（VXN）的每日收盘价。
(10) VXX_HistoricalData.xlsx：iPath系列B S&P 500 VIX短期期货交易所交易票据（ETN）的每日收盘价。
(11) VIXY_HistoricalData.xlsx：ProShares VIX短期期货交易所交易基金（ETF）的每日收盘价。
(12) fomc.xlsx：美联储联邦公开市场委员会（FOMC）预定公告日的虚拟变量。
(13) SPX_FUT.xlsx：来自Wind资讯的标普500（SPX）期货每日收盘价。

2. 代码文件（/Codes）
(1) Stata代码（/Codes/Stata_codes）：Obtain_Crypto_Sum.do与OtherVolatility_Merge.do用于生成分析所需的处理后数据。Summary_Statistics.do用于描述加密货币分解收益率与VIX变动情况。Univariate_Sort.do用于检验隔夜加密货币收益率作为情绪代理变量的有效性。Regression_Insample.do用于实现VIX日内变动的样本内预测。Regression_OutofSample_Bitcoin.do与Regression_OutofSample_Ethereum.do分别实现纳入比特币与以太坊的VIX样本外预测。OutofSample_Pandemic.do与OutofSample_NonPandemic.do用于评估新冠疫情影响期间与非疫情期间的样本外预测效果。Strategy_Performance.do与Strategy_Performance_VIXY.do用于评估基于样本外预测构建的交易策略的经济效益。Cumulative_Gain.do用于绘制所选最优策略的累计收益曲线。ExpandingWindow_OutofSample.do用于实现扩展窗口下的样本外预测。
(2) R代码（/Codes/R_codes）：mcstest.R用于对样本外预测结果应用模型置信集（MCS）检验。

3. 结果保存文件（/Save）：用于存储各实证分析步骤的核心结果，其子文件夹（/Save/Predict）用于存储模型置信集（MCS）检验所需的预测损失值。