Trading schemes for emission allowances have become a panacea for nations aspiring to reduce their aggregate emissions of greenhouse gases from industry in a cost-effective manner. The contention of this paper is that an emissions trading scheme (ETS) should not be based on blanket coverage of installations on a downstream level, but should rather be designed to include some installations, and from some industrial sectors. In the case of an ETS there are high costs of administration, monitoring and transacting imposed on the installations covered. These costs are supposed to be more than offset by the cost savings realised through trading in the market for emission allowances. However, the paper shows that not all installations can fully offset administrative costs, and are therefore exposed to higher cost compared to a situation under an alternative instrument (e.g. standard). The paper formulates a conceptual framework for analysing overall cost and benefits from an ETS in the light of administration and transactions costs. It theoretically establishes a threshold point for optimal coverage of installations on a downstream level. The paper uses data from EU ETS to empirically determine optimal coverage for selected sectors. The results indicate that blanket coverage is more costly than the determined optimum coverage plan.