You need to agree to share your Hugging Face information to access this dataset
This repository is publicly accessible, but you have to accept the conditions to access its files and content.
You agree not to redistribute this data.
Log in or Sign Up to review the conditions and access this dataset content.
Stocks Yearly Transaction Cost Params
Measured trading-cost parameters for US stocks, one row per symbol and year.
Columns
symbol: the ticker.year: calendar year the parameters were measured over.price: the year's average session VWAP.spread_bps: the full quoted spread in basis points. A round trip crosses half of it on each side. Measured from NBBO quotes where that feed exists, and from Roll (1984) on trades otherwise.lambda_bps: Kyle's slope, basis points of price move per $1M of net flow. Only meaningful against a declared notional.
Why it exists
Estimating a spread from daily closes does not work for liquid names: the signal is the lag-1 autocovariance of returns, the noise is several times larger, and the estimator collapses to zero. Measured on 9 060 US names with 240+ sessions in 2024, a Roll estimate on daily bars returned exactly zero for 54.7 % of names priced above $50, and 32 bp for the rest against 12 bp measured here.
Caveats
- Spreads are per year: they moved by an order of magnitude across the 2001 decimalisation, so using one vintage for a multi-decade backtest is wrong.
lambda_bpsis measured asprice * contracts, not dollar notional, for futures. Applying a contract multiplier is only needed to compare instruments.- Rows with no measurable spread are not published, so a lookup that misses is genuinely uncovered rather than covered-but-null. A spread of zero is treated the same way and withheld: it is an absent measurement, not a free trade.
- Downloads last month
- 6,850