Estimating Equity Slippage from Spread, Trade Size, and Market Data
Summary
The document discusses how to estimate execution slippage for a strategy trading many stocks with orders that are small relative to average daily volume. One proposed starting point is a cross-sectional regression using bid-ask spread and order size relative to average daily volume, with possible controls for stock or industry characteristics. The response cautions that such models may have limited predictive accuracy.
The appropriate approach depends on available data and order type. Bar data cannot support a strong slippage estimate; a rough transaction-cost allowance may be used for a preliminary profitability check. Order-book data allows estimating the cost of consuming displayed depth at the intended size, while more advanced market-impact models can account for the market’s response. Larger orders may need to be split and executed over time or as liquidity appears. The document does not establish a universal slippage rate, since costs depend on liquidity, data, and execution choices.
Key ideas
- Bid-ask spread and order size relative to average volume are useful initial slippage predictors.
- Cross-sectional regression can provide a starting estimate, but its predictive power may be limited.
- Bar data alone cannot reliably determine execution slippage.
- Order-book depth can estimate the cost of consuming liquidity for a given order size.
- Large orders may need to be executed in smaller pieces over time or as liquidity becomes available.
Tags
Full text
# Expected slippage based on % of average daily trading volume # Expected slippage based on % of average daily trading volume I have started a quant strategy that buys and sells thousands of stocks. Each trade represents <1% of average daily trading volume. On average, the trades represent around 0.1% of ADTV. What is a reasonable expectation for slippage given how small the orders are? I measure slippage as fill price compared to the midpoint at the time the order was entered. ## Answer by Arshdeep (score 1) https://quant.stackexchange.com/a/76396 Slippage models usually take in predictors as the bid-ask spread (usually modeled as linear dependency) and the trade size relative to ADV (typically this is non linear dependency). You can mess around with adding dummy variables for stocks of certain features/industries if you believe that influences the average slippage. I have not seen such models predict slippage to a high degree of accuracy in my experience, so don't expect probably an R-squared higher than 30%. I would recommend running a cross sectional regression based on the above predictors to get started. ## Answer by quantinho (score 0) https://quant.stackexchange.com/a/76580 As mentioned in the comment it all depends on the type of the data you have. I am assuming you are sending market orders. If you have only bar data there's no good way to approximate slippage. If it's a highly liquid instrument and your trading frequency is low you can just ignore it. You could just add some constant (1bps) to your transaction cost to check profitability of your strategy to account for slippage. If your size is large, you should not send market order in most of the cases, you would have to break down your order into smaller parts and send them as liquidity arrives (VWAP) or over time (TWAP). If you have orderbook data you can use that to calculate slippage based on your order size (average price of the sum of levels that will be taken by your size). If you want more complication you take use some market impact modules to account for the reaction of the market to your orders.
Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.