Modeling Liquidation Costs with Spread, Slippage, Fees, and Market Impact
Summary
The note explains why liquidation costs cannot be represented by one universal fraction of the bid-ask spread. Costs include explicit charges such as broker and exchange fees, as well as uncertain execution costs. A simple, liquidity-taking algorithm may incur roughly a full spread, while an algorithm that sometimes provides liquidity may pay less through spread capture. Market impact adds another cost that depends on the size of the order relative to typical volume and on volatility.
The proposed impact form combines a spread-related component with a volatility-scaled square-root term based on traded quantity and average daily volume. Its parameters must be estimated from trading data, since results depend on the instrument and the trader’s execution methods. The discussion is a conceptual framework rather than a calibrated model: it gives no sample estimates or validation results, and realized costs remain uncertain before execution.
Key ideas
- Liquidation cost combines explicit fees with implicit execution costs.
- A liquidity-taking order can incur the full bid-ask spread, while providing liquidity may reduce spread cost.
- Market impact can depend on spread, volatility, and order size relative to average volume.
- Impact parameters should be fitted to data that reflects the trader’s own execution methods.
Tags
Full text
# Cost of liquidation
# Cost of liquidation
In the text book on Risk Management by John Hull, The cost of liquidation is defined as one half of spread between bid price and ask price.
Investopedia justifies the one half factor by saying that we are concerned with only the sell part of the trade. This is not convincing.
I personally feel that The cost of liquidation should be plainly bid-ask spread without any factor. Can anyone please clarify.
## Answer by lehalle (score 1)
https://quant.stackexchange.com/a/34748
Cost of liquidation should include
- the explicit costs: fees (brokers, exchanges, give-up, post trading, etc)
- the implicit costs that you cannot know for sure a priori. They are themselves made of slippage: linear costs mostly, a function of the bid ask spread. If you use a dumb trading algo, you will pay a full bid-ask spread, you are right. But if you use an algo that is a little smarter (for instance succeeding in being liquidity provider --ie in the book-- part of the time) you will pay less and the market impact
Formulas are explained in Market microstructure in practice, but in short you need a market impact of the shape
$$\eta(q) = a \,\psi + \kappa\,\sigma\sqrt{\frac{q}{\bar V}}$$
where $q$ is your traded quantity, $\psi$ the bid-ask spread, $\sigma$ the volatility, $\bar V$ the average daily volume, and $a$ and $\kappa$ are parameters you need to fit on your data. It is important to use your own data because it will capture your trading habits (types of algo you use --see Chap 3 of the book-- for instance).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.