The carbon dioxide (CO 2) emissions reduction afforded by a demand-side intervention in the electricity system is typically assessed by means of an assumed grid emissions rate, which measures the CO 2 intensity of electricity not used as a result of the intervention. This emissions rate is called the “marginal emissions factor” (MEF). Accurate estimation of MEFs is crucial for performance assessment because their application leads to decisions regarding the relative merits of CO 2 reduction strategies. This article contributes to formulating the principles by which MEFs are estimated, highlighting the strengths and weaknesses in existing approaches, and presenting an alternative based on the observed behaviour of power stations. The case of Great Britain is considered, demonstrating an MEF of 0.69 kgCO 2/kW h for 2002–2009, with error bars at +/−10%. This value could reduce to 0.6 kgCO 2/kW h over the next decade under planned changes to the underlying generation mix, and could further reduce to approximately 0.51 kgCO 2/kW h before 2025 if all power stations commissioned pre-1970 are replaced by their modern counterparts. Given that these rates are higher than commonly applied system-average or assumed “long term marginal” emissions rates, it is concluded that maintenance of an improved understanding of MEFs is valuable to better inform policy decisions.