Market Impact Models at Third-Generation

Market Impact models at third-generation

For some time Market Impact has fascinated academics and practitioners alike. The definition of Market Impact is how much a stock moved due to your trading compared with how much it would have moved if you didn’t trade.

Given this definition it is impossible to measure Market Impact for a single trade without making an assumption.

However, an estimate of Market Impact can be determined using a model.

Whether the estimate is realistic depends on the quality of your model and the underlying assumptions used in the model.

We have classified Market Impact Models into generations. Typical underlying assumptions for models of each generation are highlighted.


First-generation models

A first-generation market impact model is a general, price pattern model. A single Market Impact Model is used for all stocks in a market. Differences in liquidity between stocks are captured by the use of Days Turnover as a key factor. In a first-generation model, Market Impact is generally determined from a minimal number of factors, such as Days Turnover and Spread.

Underlying assumption:

All stocks with the same Days Turnover have the same Market Impact.

Inherent problem:

Stocks have different Volatility.

Market Impact for an order of, say, 0.3 Days Turnover in a highly volatile stock will be significantly higher than the Market Impact for an order of 0.3 Days Turnover in a very low volatility stock. This is apparent intuitively and can be confirmed empirically.


Second-generation models

A second-generation market impact model is a stock-specific, price-pattern model. A separate Market Impact Model exists for each security. Additional price-related factors are used in second-generation models, with the most primitive model simply including Volatility with Spread and Days Turnover, the factors of a first-generation model. Some second-generation models might also include Price Momentum as a factor.

An Inventory Cost Model is an example of a simple second-generation market impact model. Typical methodology is that users are required to provide a Participation Rate (eg 30% of turnover). Days Turnover divided by Participation Rate is the Expected Completion Period measured in days. Market Impact is then determined from the Volatility over the Expected Completion Period.

Underlying assumption:

Market Impact of an order in a stock is an adverse move of one standard deviation according to the stock’s Volatility Cone.

Inherent problem:

Increasing Participation Rate decreases Expected Completion Period. This results in a lower Market Impact. The problem is that this is not intuitive, nor factual.

Increasing Participation Rate increases the urgency of the order, which implies a higher Market Impact, not a lower one.


Third-generation models

A third-generation market impact model is a stock-specific, price-pattern and volume-pattern model. Each security has its own Market Impact Model. A third-generation model includes volume-patterns amongst its factors. Examples of volume factors include Volume Predictibility, Volume Persistance and Volume Flexibility. Additional price-related factors such as Price Momentum, Upside Risk and Downside Risk might also be used.

Our EQ Impact Model is an example of an advanced third-generation market impact model.

Underlying assumption:

There exists a relationship between price patterns and volume patterns which is captured by the third-generation market impact model.

Inherent problem:

Given their advanced nature, third-generation models are typically proprietary models, and thus only explained by way of example or by using less-specific concepts.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.