Benjamin Guin and Perttu Korhonen
A well-insulated house reduces heat loss during cold winter periods and it keeps outdoor heat from entering during hot summer conditions. Hence, effective insulation can reduce the need for households to use cooling and heating systems. While this can lower greenhouse gas emissions by households, it also reduces homeowners’ energy bills, which can free up available income. This can protect households from unexpected decreases in income (e.g. reduced overtime payments) or increases in expenses (e.g. healthcare costs). It could also help homeowners to make their mortgage payments even if such shocks occurred. But does this also imply that mortgages against energy-efficient properties are less credit-risky?
We analyse borrower-level data on mortgage arrears matched with property-level energy ratings
We examine this question using loan performance data for residential mortgages in the entire UK (year-end 2017) which we match with the energy performance certificates (EPCs) of the underlying properties. We add information on the income of the borrower at the time of mortgage origination. The final sample of matched mortgages consists of more than 1.8 million observations.
EPCs provide information on the annual energy costs of a property. Buildings need them whenever they are built, sold or rented. EPCs rate properties from A (most efficient) to G (least efficient). We categorise properties into three buckets: ‘High energy efficiency’ (EPC ratings of A, B or C), ‘Medium energy efficiency’ (EPC rating of D) and ‘Low energy efficiency’ (EPC ratings of E, F or G). For example, the annual energy bill of a highly energy-efficient 4-bedroom house is on average GBP 1,080 lower than for a 4-bedroom house with low energy efficiency.
Table 1. Annual energy costs (in GBP) by type and energy efficiency of the property
|Type of property||High energy efficiency||Medium energy efficiency||Low energy efficiency|
|(EPC rating A-C)||(EPC rating D)||(EPC rating E-G)|
Our evidence suggests that mortgages on efficient properties are less risky
Simple univariate comparisons suggest that about 0.93% of residential mortgages against energy-efficient properties are in payment arrears. This share is 0.21 percentage points lower than the share of mortgages against energy-inefficient properties, which is 1.14%. This difference is statistically significant at the 1 percent level. In graph 1, the black dot illustrates this difference. The black bar shows the 99% confidence interval. We provide an interactive version of this graph here.
Graph 1. Difference in mortgage arrears (high energy efficiency vs low energy efficiency)
Two mechanisms could be driving this difference. On the one hand, energy bills are lower on energy-efficient properties. Savings on energy bills could lead to lower arrears rates (‘energy savings effect’). Alternatively, high-income borrowers could be more likely to take out mortgages against energy-efficient properties. Such borrowers may fall into arrears less frequently (‘income selection effect’). To examine the relevance of the ‘income selection effect’, we compare mortgage arrears of borrowers with similar income. Practically, we control for borrower income at origination in a multivariate regression analysis (see technical appendix for further details). The difference in payment arrears of mortgages against energy-efficient properties compared to mortgages against energy-inefficient properties remains similar (grey cross in bar 2). So the ‘income selection effect’ does not explain lower mortgage arrears for mortgages against energy-efficient properties.
To test the robustness of our findings, we also control for property characteristics (indicators are property type and whether the property is newly built, the number of rooms, the number of heated rooms and the floor area of the property) and contract-specific characteristics (the loan-to-value, loan amount, property price). The difference in payment arrears remains qualitatively similar (blue arrow in bar 3).
Additionally, we control for year of mortgage origination and EPC inspection. The difference in payment arrears between mortgages against energy-efficient properties compared to energy-inefficient properties decreases in magnitude to -0.10 percentage points. It remains statistically significant at the 1 percent level. This suggests that these factors can explain some but not all of the correlation between energy efficiency and mortgages arrears (orange square in bar 4).
We conclude that the energy efficiency of a property is a relevant predictor of mortgage risk
Overall, these results suggest that mortgages against energy-efficient properties are less frequently in arrears. Mortgage borrowers’ income at origination cannot explain this difference. However, the dates of mortgage origination and EPC inspection do explain some but not all of this correlation. We conclude that the energy efficiency of a property is a relevant predictor of mortgage payment arrears.
Does this imply that there is a causal relationship between higher energy efficiency and lower mortgage payment arrears? Not necessarily. Yet, some banks have started to price mortgages against energy-efficient properties at lower rates, implying a lower risk premium. For example, a bank in the UK started offering lower interest rates to mortgage borrowers buying energy-efficient new-build homes in the UK this year. Besides, several European banks have teamed up to run a pilot initiative, which rewards buyers of greener certified homes with lower interest rates on their mortgages.
A remark on the data employed in this analysis
EPC data come from the Ministry of Housing, Communities & Local Government that has published these data with the consent of Royal Mail Group Limited. Royal Mail Group Limited permits the use of ‘address data’ for research purposes. Our final data set contains HM Land Registry data © Crown copyright and database right 2018. This data is licensed under the Open Government Licence v3.0.
The data on energy efficiency ratings come from the Ministry of Housing, Communities and Local Government. EPCs rate properties from A (most efficient) to G (least efficient). We categorise properties into three buckets: ‘High energy efficiency’ (EPC ratings of A, B or C), ‘Medium energy efficiency’ (EPC rating of D) and ‘Low energy efficiency’ (EPC ratings of E, F or G). We decided to conduct this grouping to ensure that the number of properties in each bucket is sufficient for a robust analysis.
We complement these data with information on mortgage performance at year-end 2017 and information on mortgage originations since 2008. These data come from the FCA’s Product Sales Database. Moreover, we merge data on property transaction prices from HM Land Registry.
We estimate the following linear relationship using Ordinary Least Squares while calculating heteroscedasticity robust standard errors:
Arrears i =α+β1 High energy efficiency i +β2 Medium energy efficiency i +Xi‘β3+εi
- Arrears i is a dummy variable indicating whether borrower i in payment arrears in year-end 2017.
- High energy efficiency i is a dummy variable indicating whether the property of borrower i has an EPC rating of A, B or C.
- Medium energy efficiency i is a dummy variable indicating whether the property of borrower i has an EPC rating of D.
- Xi indicates a vector of borrower, property and contract characteristics. Household income is winsorized at the 99 percentile to account for outliers.
Table 1 presents these results. Column 1 shows the results not controlling for further characteristics in a regression setup. We present the point estimates of High energy efficiency and Medium energy efficiency. One can interpret them as (marginal) effects of these variables relative to the baseline category Low energy efficiency.
In column 2, we control for gross income at origination. The differences in payment arrears remain qualitatively similar.
In column 3, we additionally control for property characteristics (indicators are property type and whether the property is newly built, the number of rooms, the number of heated rooms and the floor area of the property) and contract-specific characteristics (loan-to-value, loan amount, property price).
In column 4, we control additionally for the years of mortgage origination and EPC inspection. The difference in payment arrears between mortgages against energy-efficient properties compared to energy-inefficient properties decreases in magnitude. It remains statistically significant at the 1 percent level.
Table 1. Multivariate analyses: energy efficiency and mortgage payment arrears
|High energy efficiency (0/1)||-0.0021***||-0.0024***||-0.0020***||-0.0010***|
|Medium energy efficiency (0/1)||-0.0014***||-0.0017***||-0.0014***||-0.0004**|
|Gross income (thousand GBP)||-0.0001***||-0.0000***||-0.0000|
|Joint income (0/1)||-0.0005***||-0.0007***||-0.0040***|
|Age of borrower (years)||0.0000**||0.0000***||0.0000***|
|Property control variables||No||No||Yes||Yes|
|Contract control variables||No||No||Yes||Yes|
|Regional control variables||No||No||Yes||Yes|
|Origination year control variables||No||No||No||Yes|
|Inspection year control variables||No||No||No||Yes|
|Mean of dep. variable||0.0103||0.0103||0.0103||0.0103|
Note: This table shows the results of a linear model estimated using OLS with the propensity of payment arrears, Arrears, as the dependent variable. Explanatory variables are binary variables for the energy efficiency of the property, as well as borrower, property and contract-specific characteristics. The sample includes mortgages in 2017 H2. Observations are at the mortgage level. Standard errors are clustered on the 3-digit postcode. ***, **, * denote statistical significance at the 0.01, 0.05 and 0.10-level respectively.
To verify these results, we run a battery of robustness tests:
- First, we estimate non-linear Probit models using Maximum Likelihood. We report marginal effects at the mean of continuous variables.
- Second, we control for the natural logarithm of Gross income (thousand GBP) instead of the level of Gross income (thousand GBP).
- Third, we control for property types classified by the EPC certificates (as opposed to the classification in the HM Land Registry).
- Last, we examine the continuous EPC variable underlying our discrete EPC ratings. We use this variable to create alternative binary EPC buckets (below 40, at least 40 & below 50, at least 50 & below 60, at least 60 & below 70, at least 70).
Our results stay qualitatively similar. They are available upon request.
Benjamin Guin works in the Bank’s Banking Policy Division and Perttu Korhonen works in the Bank’s Data & Statistics Division.
If you want to get in touch, please email us at firstname.lastname@example.org or leave a comment below.
Comments will only appear once approved by a moderator, and are only published where a full name is supplied.
Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.