Direct measurements of individual home emissions are currently unavailable. Instead, projections based on the locale and technical characteristics of a home are used starting from 2008 onwards, after the introduction of the energy performance certificate (EPC) regime. This introduces subtle dependencies on the underlying methodologies and uncertainty due to the modelled assumptions about energy consumption and emissions. Still, older properties that have not changed hands in the past 12 years, which are amongst the principal residential polluters, remain under the radar and in some areas account for up to 65% of the housing stock. Using advanced data science methods to compensate for the missing data is therefore an important step.
While a large fraction of poor performers is missing, the data that is available is likely biased in favor of better performance. Moreover, the data exhibits gaps, inconsistencies and all manner of anomalies. This is neither surprising nor specific to housing, but rather a consequence of the fact that the collection and reporting of home energy performance data are labor-intensive, multi-step processes susceptible to accumulating human errors. We have found error rates upwards of 10% in any one of the tracked fields, and often multiple irregularities in home performance records.
Similar issues are common in all large-scale databases, so that data curating, cleansing and applying best practices to handling errors are vital to deciphering meaningful insights. Over the past three years our team has developed comprehensive tools to address the above issues. Our suite combines modern data science techniques with tried-and-tested scientific and statistical methods we pioneered in work that was first applied to the US housing market, resulting in a home price index that was swiftly adopted as the basis of financial instruments by global investment banks.
* Figure 1: Typical home emissions intensity spectrum, shown for the entire housing stock with EPC records in England on 1 January 2009 and overlaid with the corresponding spectrum on 20 September 2020. The second peak (indicated by a red arrow) emerging in the more recent data is discussed in the text.
An important lesson of that work was that studying the distributions of housing attributes can reveal a wealth of information lost if one is merely fixated on timeseries, as analysts can sometimes be, missing the forest for the trees. The housing system is characterized by what is often referred to as ‘heavy tails’ (Figure 1). Unlike the more familiar Normal distributions (aka Bell Curves), which are tightly contained within a narrow envelope around the average, heavy-tailed distributions can stretch over very long ranges that reach out to extreme values far from the bulk of a sample.
Heavy tails were popularised from the Pareto Principle, sometimes referred to as the ‘80-20 rule’, which is an aphorism asserting that 80% of outcomes result from 20% of causes and vice-versa. Our study of energy and emissions distributions of housing data reveals that housing retrofit programmes will have the greatest climate impact (and return on investment) if they target the right homes. Identifying and tackling the ‘heavy tail’ is imperative if we are to address housing emissions at the right speed necessary to have a meaningful impact on climate change.
"Housing retrofit programmes will have the greatest impact if they target the right homes. Identifying and tackling the ‘heavy tail’ is imperative."
- Marios Kagarlis, Climate Benchmark
Figure 1 reveals the emergence of a 'twin peak' in the emissions profile of English housing stock. Over ten years the top performing stock (new build) has improved noticeably but most of the housing stock (legacy homes) remains largely unchanged in performance. Importantly, we have collectively made little progress in tackling the ‘heavy tail’ that is responsible for most housing emissions.
Our study of energy and emissions distributions of housing data over the past three years has produced a wealth of insights, some of which are incorporated into our newly launched house emissions indices and portfolio analysis tools. We continue to pursue an active research program and integrate its findings towards a comprehensive suite for decarbonisation stakeholders, including housing emissions management, portfolio optimisation and policymaking decision support tools.
The Bottom Line
No single-valued metric can reflect the rich information in the structure of a spectrum. Climate Benchmark, alongside the more conventional analysis of timeseries signals, also exploits the full information contained in the distributions of housing emissions to uncover hidden insights.
- There has emerged a 'twin peak' in the emissions profile of English housing stock.
- We have collectively made little progress in tackling the ‘heavy tail’ that is responsible for most housing emissions.
- Housing retrofit programmes will have the greatest impact if they target the right homes. Identifying and tackling the ‘heavy tail’ is imperative.