voigtstefan/sp500

--- license: mit tags: - finance pretty_name: SPY ETF High-frequency data size_categories: - 10M<n<100M --- # Dataset Card for SPY ETF High-frequency data This dataset was created by aggregating message-level observations for the ETF SPY which tracks the S&P 500 index. We retrieved and evaluate the entire available order book message history for the first 50 levels from July 1st 2007 to October 31st 2025. We restrict our analysis to order book activity during regular trading hours. ## Dataset Details ### Dataset Description We aggregate the order book messages into the following variables, measured at 5 second intervals: 1. Initiator net volume (in million USD) which is the net of buyer and seller initiated shares transacted during the last 5-minute interval. We sign transactions as +1 if executed against a sell-side limit order and -1 if executed against a buy-side limit order. For execution against a hidden limit order, we impose the sign +1 if the transaction executes at a price that exceeds the last observed midquote and -1 if the transaction price is below the last observed midquote. 1. Trading volume (in million USD), which is the cumulative trading volume during each 5-second interval. We compute trading volume as the number of traded shares times the transaction price. 1. Bid-ask spread (in basis points), computed as the time-weighted average difference between the best prices quoted at the sell and buy side of the order book during each 5-second interval. We compute the bid-ask spread relative to the current midquote for every order book snapshot. 1. Depth measured as the number of posted shares in visible limit orders 5 basis points from the current best price on both sides of the order book. We take the time-weighted average depth during each interval to aggregate depth from message level into 5-second intervals. - **Curated by:** Stefan Voigt ([e-mail]) - **License:** MIT ### Dataset Sources More information on the dataset can be found in the paper "Market responses to a VIX shock" by Albert J. Menkveld, Nikolaus Hautsch, and Stefan Voigt. - **Repository:** [https://github.com/voigtstefan/market-responses-to-a-vix-shock](https://github.com/voigtstefan/market-responses-to-a-vix-shock) ## Citation Please cite the paper "Market responses to a VIX shock" by Albert J. Menkveld, Nikolaus Hautsch, and Stefan Voigt when using our dataset. **BibTeX:** @unpublished{menkveld2026vix, author = {Menkveld, Albert J. and Hautsch, Nikolaus and Voigt, Stefan}, title = {Market Responses to a {VIX} Shock}, note = {Available at SSRN: \url{https://ssrn.com/abstract=6298780}}, url = {https://ssrn.com/abstract=6298780}, year = {2026}, month = feb, day = {24}, }--- 许可证: MIT协议 标签: - 金融 展示名称: SPY ETF高频数据集 规模类别: - 1000万 < 数据量 < 1亿 --- # SPY ETF高频数据集数据集卡片 本数据集通过聚合追踪标普500指数（S&P 500 Index）的交易所交易基金（Exchange Traded Fund, ETF）SPY的逐笔消息观测数据构建而成。我们获取并评估了2007年7月1日至2025年10月31日期间全部可获取的前50档委托簿消息历史，且仅对常规交易时段内的委托簿交易活动开展分析。 ## 数据集详情 ### 数据集描述 我们将逐笔委托簿消息聚合为以下以5秒为间隔的统计变量： 1. 发起方净成交量（单位：百万美元）：指最近5分钟区间内，买方发起与卖方发起的成交股份净额。我们将交易标记为+1，若成交对手方为卖盘限价订单；标记为-1，若成交对手方为买盘限价订单。若成交对手方为隐形限价订单，则当成交价格高于最新观测到的最优买卖中间价时标记为+1，低于该中间价时标记为-1。 2. 交易成交量（单位：百万美元）：指每个5秒区间内的累计成交总量。我们通过成交股份数乘以成交价格计算该指标。 3. 买卖价差（单位：基点）：指每个5秒区间内，委托簿买卖双方最优报价差值的时间加权平均值。我们以每一次委托簿快照的当前最优买卖中间价为基准计算买卖价差。 4. 委托簿深度：指委托簿双方当前最优价格上下5基点范围内的可见限价订单挂单股份总数。我们通过计算每个区间内深度的时间加权平均值，将逐笔消息级别的深度聚合为5秒区间统计值。 - **整理者：** 斯特凡·福伊格特（Stefan Voigt）<[e-mail]> - **许可证：** MIT协议 ### 数据集来源 本数据集的更多细节可参阅Albert J. Menkveld、Nikolaus Hautsch与Stefan Voigt合著的论文《Market responses to a VIX shock》（《VIX冲击下的市场反应》）。 - **代码仓库：** [https://github.com/voigtstefan/market-responses-to-a-vix-shock](https://github.com/voigtstefan/market-responses-to-a-vix-shock) ## 引用说明 使用本数据集时，请引用Albert J. Menkveld、Nikolaus Hautsch与Stefan Voigt合著的论文《VIX冲击下的市场反应》。 **BibTeX：** @unpublished{menkveld2026vix, author = {Menkveld, Albert J. and Hautsch, Nikolaus and Voigt, Stefan}, title = {Market responses to a {VIX} Shock}, note = {Available at SSRN: url{https://ssrn.com/abstract=6298780}}, url = {https://ssrn.com/abstract=6298780}, year = {2026}, month = feb, day = {24}, }