Cboe U.S. Equities Historical Market Volume Data 2021

Rows: 4,788
Columns: 14

Daily historical market volume, notional value, and trade counts for U.S. equities exchanges and TRFs for 2021.

AI analysis

Dataset Analysis: Cboe U.S. Equities Historical Market Volume Data 2021

1. Dataset Overview and Correlation Potential

This dataset, sourced directly from Cboe's official data repository at cdn.cboe.com, represents a comprehensive daily record of U.S. equity market activity across all major exchanges and Trade Reporting Facilities (TRFs) for the full calendar year 2021. The Cboe origin signals authoritative, exchange-level data with institutional-grade reliability. The structure — 4,788 rows spanning 252 trading days across 19 distinct market participants — forms a panel dataset (252 days × ~19 participants), making it exceptionally well-suited for both time-series and cross-sectional correlation studies. The three-tape breakdown (Tape A for NYSE-listed, Tape B for NYSE American/regional, Tape C for Nasdaq-listed) adds granularity that enables analysis of how volume and value distribute across listing venues, and how individual market participants' share of activity shifts over time.

---

2. Data Quality Observations

The dataset presents an unusually clean quality profile. With zero null cells across all 14 columns and 4,788 total rows, there are no missing-value imputation concerns — a notable strength for time-series modeling. The duplicate row count is flagged as pending recomputation against the Parquet source, which warrants a verification pass before any aggregation work, particularly since the panel structure (date × participant) means even a single duplicated participant-day pair could silently inflate volume totals. The distinct counts for Tape C Shares (4,787) and Total Shares (4,787) are nearly equal to total row count, confirming high cardinality and minimal repetition in the numeric columns. One subtle quality note: Tape C Shares has a minimum of 1 while Tape A and Tape B Shares have a minimum of 0, suggesting either a data entry convention difference or that at least one participant always reports non-zero Tape C activity — worth investigating for analytical consistency.

---

3. Key Column Distributions

The volume and notional columns exhibit pronounced right skew across the board, which is expected in market microstructure data but warrants attention. Total Shares is the headline volume metric: mean of ~600M shares/row versus a median of only ~170M (skew = 2.84, 348 outliers), indicating that a small number of large participants or high-volume days dominate the distribution — the mean is nearly 3.5× the median. Total Notional is even more dramatic: mean ~$29.7B versus median ~$8.1B (skew = 2.44, 498 outliers), with a maximum of ~$307B in a single participant-day observation. Tape C Notional deserves special attention — it carries the highest individual maximum (~$206B) and the highest standard deviation (~$23.2B), reflecting Nasdaq-listed securities' dominance in dollar value terms. Trade count columns tell a more moderate story: Total Trade Count has skew of 1.96 with a mean of ~3.6M versus median of ~1.7M, suggesting trade fragmentation is less extreme than raw share volume. The outlier counts are substantial across all numeric columns (ranging from 348 to 679), strongly recommending log-transformation or robust scaling before any regression or clustering work.

---

4. Recommended Join Key Columns

The Date column is the primary join key, with exactly 252 distinct values precisely matching the standard U.S. equity trading calendar for 2021 — zero nulls and no apparent gaps make it highly reliable for time-series joins. Market Participant (19 distinct string values, zero nulls) serves as the secondary dimension key, enabling panel joins where another dataset carries participant or exchange identifiers. For cross-dataset work, the combination of Date + Market Participant as a composite key is recommended to avoid fan-out during joins, since each date has approximately 19 participant rows. If joining to macro or index-level data that lacks participant granularity, aggregating to daily totals first (using Total Shares, Total Notional, Total Trade Count) and joining on Date alone is the safer approach.

---

5. Suggested Complementary Datasets for Correlation Discovery

Several dataset categories would pair powerfully with this data. VIX (CBOE Volatility Index) daily data is a natural first pairing — also a Cboe product — to test whether elevated volatility regimes predict spikes in total share volume or trade count, given the right-skewed outlier clusters likely correspond to volatile market episodes in early 2021 (e.g., meme stock events in January–February). S&P 500 or Russell 2000 daily returns and index levels would allow correlation of Tape A/B/C notional with broad market direction and magnitude. Fed funds rate expectations or Treasury yield data (e.g., from FRED) could reveal whether rate-sensitive periods in 2021 shifted volume between listing venues. Individual stock earnings calendar data could explain Tape C notional spikes, since Nasdaq-heavy tech names dominate dollar volume. Finally, dark pool or off-exchange TRF share data — if available separately — would complement this dataset to study the lit-vs-dark venue balance across the 19 participants over time.

Columns

Date (date)
Market Participant (string)
Tape A Shares (integer)
Tape B Shares (integer)
Tape C Shares (integer)
Total Shares (integer)
Tape A Notional (decimal)
Tape B Notional (decimal)
Tape C Notional (decimal)
Total Notional (decimal)
Tape A Trade Count (integer)
Tape B Trade Count (integer)
Tape C Trade Count (integer)
Total Trade Count (integer)