I'm looking to piece together a robust optimization model that handles robust optimization with fixed transaction costs and other combinatorial variables (e.g. asset count constraints).

Here's what I've been able to find so far:

  • Goldfarb and Iyengar 2004 describes an SOCP robust counterpart to QCP problems. But it doesn't discuss combinatorial variables (explicitly).

  • Lobo, Fazel, and Boyd 2006 demonstrate a heuristic that solves the affine transaction cost problem via a heuristic that only requires ~5-6 runs of the underlying convex optimization problem to achieve a near optimal solution. Their underlying convex problem is not robust, however.

Is there a resource that pieces both of these together?

  • $\begingroup$ Excuse my ignorance, but why would you not want to formulate transaction costs a part of a budget constraint? $\endgroup$
    – RndmSymbl
    Commented Feb 9, 2014 at 18:53
  • $\begingroup$ Silly on my part ... meant combinatorial decision variables. $\endgroup$
    – MikeRand
    Commented Feb 9, 2014 at 19:44
  • $\begingroup$ Sorry, I may still not understand properly, and certainly do not know of any related work, but being curious, will you consider simulation? I would imagine a heuristic can be thought of to add a limited set of elements only until a benefit threshold has been reached? $\endgroup$
    – RndmSymbl
    Commented Feb 14, 2014 at 15:14
  • $\begingroup$ this may be interesting if you're living in a factor model world $\endgroup$
    – user25064
    Commented Feb 20, 2014 at 13:45

2 Answers 2


It depends on what you want to optimize with transaction costs:

  • liquidation
  • hedging
  • allocation

The two best reference I have in mind are:

You will not find in the literature a paper covering exactly a specific problem. Nevertheless my viewpoint is that the difficulties in your case (i.e. mixing combinatorial constraints, robust optimization and transaction costs) comes from the way to insert transaction costs in existing frameworks.

As it is explained into the Soner et al. paper: there is one paper by Almgren dealing with "when should I end liquidation?". It is an important question since trading infinitesimally can be perceived as a solution to market impact minimization. With M Labadie, we extensively covered this aspect of discretizing a liquidation process in "Optimal starting times, stopping times and risk measures for algorithmic trading: Target Close and Implementation Shortfall".

The question "when should I end selecting lines for a portfolio?" (i.e. not that far from asset count constraints) is in fact close to the previous one. Not deep enough to justify an academic paper according to me, since once you choose:

  1. a utility function (i.e. a criterion),
  2. a market impact (i.e. transaction cost) model,
  3. the parameters or model features you want to be robust to

the result should come straightforward thanks to a combination of techniques described in the papers I cited.

  • $\begingroup$ That Kolm presentation has a nice bit on SOCP optimization with fixed transaction costs. Thank you. $\endgroup$
    – MikeRand
    Commented Feb 16, 2014 at 2:12
  • $\begingroup$ @lehalle interesting references, but where do these sources combine asset count constraints with transaction cost constraints? $\endgroup$
    – RndmSymbl
    Commented Feb 16, 2014 at 10:47

You could refer Dimitris Bertsimas's work on Robust Optimization. One of his notable works that may be relevant include robust optimization formulations of the multiperiod portfolio optimization problem in the presence of transaction costs.

  • $\begingroup$ Saw that paper ... believe it deals with linear transaction costs. $\endgroup$
    – MikeRand
    Commented Feb 9, 2014 at 19:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.