Predicting Energy Efficient Households

The Challenge

The Energy Saving Trust (EST) works in partnership with the Scottish Government and Local Authorities to deliver initiatives (such as the Green Deal and the Home Energy Efficient Programmes for Scotland) that fund the installation of energy efficient retrofit measures in Scottish properties (such as cavity wall or loft insulation). These programmes are delivered by the various Local Authorities who are required to target properties that are, for example, in fuel poor areas and the private sector.

In order that the Local Authorities target the correct homes, the EST have developed Home Analytics, an address-level data set on housing stock in Scotland providing core property characteristics and the potential for energy efficiency and renewable energy retrofit to properties. It also provides socio-demographic data on properties which includes additional information on fuel poverty, energy efficiency funding eligibility and the propensity of householders to undertake retrofit.

Home Analytics is made up of a variety of data sources, such as Energy Performance Certificates and boiler installation records, but, even combined, these sources are unable to provide a complete record of all Scottish properties. However, Local Authorities require a complete record in order to accurately target homes and therefore the challenge in this project was to develop a statistical model that could be used to predict property characteristics and fill in these missing gaps.

The Approach

We worked with EST to develop an appropriate statistical model that would allow us to predict the property characteristics of those houses with missing records. The statistical model was developed so that it used the characteristics of the surrounding properties to assess the probability that the property is, for example, terraced, a flat, semi-detached or detached. Therefore, if there are a high proportion of houses in a postcode that are terraced, the probability that a property with a missing record in that postcode is a terrace will also be high.

Since the model assigns a probability to each category, the missing record can be replaced by the category with the highest probability, but the probabilities themselves can also be used to provide aggregated summaries at different spatial scales.

The Value

Developing a statistical model of property characteristics allowed us to fill in the missing records in the EST’s Home Analytics product. The key component of the model was predicting the missing values based on the characteristics of the properties in the surrounding area. By providing the Scottish Government with a complete record of the property characteristics in Scotland, they are better equipped to identify houses in need of energy efficiency measures and to ensure that funding is directed at those more in need.

As a Project Manager it makes all the difference to work with consultants who are continuously professional, reliable and expert at what they do. Select Statistics has delivered excellent work for EST across a range of projects. We value their positive attitude and collaborative approach which has enabled us to systematically work through complicated issues to deliver projects on time, on budget and with excellent results. Select Statistics are a key part of our wider project team and we very highly value their input.Will Rivers – Data Insight Manager, Energy Saving Trust