wilmott.com

Dear all,please, provide some clues on realistic functions that could be used for modelling transaction costs in portfolio optimization problems. For sure the transaction costs function is not linear due to the fact that brokerage comission depends on either traded volume or traded investment cash value. Any suggestions and ideas (articles) on the matter are highly appreciated.Thank you for your time and consideration.

Do a search on "slippage models" or "transaction costs" on this site: this topic has been discussed ad nauseum before.T

A good starting point would be to look a the article by Almgren & Chris published in 2000 (don't remmember the exact title).There is a book as well called "Optimal trading strategies" by KissellHTH

The question what you mostly ineterested are in. Bidask spread and other commision are modeled differently. If you trade relatively frequently - bidask spread becomes a dominant, otherwise other factors are more essential Cvitanic has a set of paper devoted to "friction market". He also set quite general case of "transcation costs" term.

The question is a little bit more complicated due to the fact that it concerns modelling transaction costs with software where the decision taken by the software system whether to recommend trade in certian position/portfolio depends on the transaction costs that might be incurred. Thus the aim of the optimization is to maximize return given particular risk threshold level. However, the nature of transaction costs (depending on traded volume and/or the type of the traded financial instrument) affects directly return and the decisions taken by the system. An easy way to solve the problem is to offer software users the option to enter average transaction costs per trade (let's say expressed in percentage terms) as a setting in the software interface. However, the latter option renders difficulties for the user himself. I believe that the issue could be solved in a more elegant manner. Any suggestions and past experience with such problems are more than welcome.

Hi portfolioHave you made any progress on this matter?We are currently doing something which may be similar to your project.We are building a portfolio optimiser which explicitly takes into account variable transaction costs per trade.Being able to embed our proprietary algorithm for market impact into the code for each stock is a huge advantage in this area.Other transaction costs vary in a linear fashion and are more simple to include.

There are systems out there which use quite complex non-linear transaction cost functions. How useful very detailed tc-modelling in an optimizer framework is depends on what the optimizer is used for:For normal portfolio management, i.e. rather low turnover, the importance of tc in an optimization process is much more to just discourage trades and switches - otherwise the optimizer would introduce a lot more fluctuation into the portfolio. If've seen people creating elaborate models, that don't do much in terms of portfolio performance...In a high turnover scenario tc becomes very important and you need good models (perhaps including basket effects).