Since 1990, the carbon emissions of dwellings in the UK have declined by around 20%. This reduction per dwelling is mainly ascribed to the impact of energy efficiency measures, such as improvements in building codes/regulations. In the UK, national energy models of the building stock are used to support the formal cost benefit analysis of policies.

Energy policy evaluation and development can often involve the combined use of evidence from energy ratings, national housing energy models, and empirical statistics. Limited research has been done, however, to look at systematic differences in the energy consumption estimates from these different tools or sources.

This paper compares annual gas consumption estimates from two versions of the Cambridge Housing Model (CHM), a national energy stock model, with empirical data from over 2.5 million gas-heated homes in England in the National Energy Efficiency Database (NEED). The analysis investigates differences by dwelling type, size, and age band. It also compares gas consumption across different rating bands estimated from Energy Performance Certificates (EPC). The findings show systematic overestimation by the CHM for larger older dwellings and far lower savings than would be expected from upgrading dwellings to a minimum of EPC Band C.

This has implications for use of building energy models that assume uniform operational settings across all dwellings, especially where older larger dwellings are the target of energy efficiency initiatives. It suggests that the models should be regularly reviewed and revised based on empirical evidence and that greater uncertainty should be ascribed to predictions of energy savings in the residential sector.