The partition coefficient of a substance measures its solubility in octanol compared with water and is widely used to estimate toxicity. If a substance is hardly soluble in octanol, then it is practically impossible for it to enter (human) cells and therefore is less likely to be toxic. For novel drugs it might be important to penetrate the cell through the membrane or even integrate into it. While for most simple substances the partition coefficient is concentration-independent at low concentrations, this is not true for a few important classes of complex molecules, such as ionic liquids or tensides. We present a simple association-dissociation model for concentration dependence of the partition coefficient of ionic liquids. Atomistic computer simulations serve to parametrize our model by calculating solvation free energies in water and octanol using thermodynamic integration. We demonstrate the validity of the method by reproducing the concentration-independent partition coefficients of small alcohols and the concentration-dependent partition coefficient of a commonly used ionic liquid, 1-butyl-3-methylimidazolium bis(trifluoromethylsulfonyl)imide [C4MIM][NTf2]. The concentration dependence is accurately predicted in a concentration range of several orders of magnitude.