By Max Riedel, Research Fellow at Ca’ Foscari
Do borrowers become more creditworthy if they take out a mortgage on an energy efficient (EE) building? Or, to start with a more prudent question: does there exist any empirical relationship between EE and the probability of mortgage default (PD) at all? The first question is concerned with the identification of a causal link between EE and PD, while the second questions the very existence of a link between the two. The current data environment is challenging for answering either of the two questions.
To study the link between EE and PD, the necessary pre-condition is a clean, granular dataset that accurately distinguishes between loans on EE and non-EE buildings. Unfortunately, this information is seldom readily available as banks either store energy performance data in physical form, which is retrospectively costly to digitalize, or data is not being stored at all due to lack of an IT solution. Furthermore, the link has to be identified by taking into account borrower and building characteristics that might confound the empirical findings if not included. For instance, borrower’s age, building age, or location might have a confounding effect on the identification of a clear relationship between EE and PD. Fortunately, banks already collect the most relevant information in order to feed their credit risk models. Therefore, the main challenge in this respect lies in the collection of EE data and its merge with the remaining mortgage information.
When it comes to identifying causality, the requirements on the analysis become much more demanding. In order to establish a causal link, we have to understand and take into account the borrower’s consumption behaviour. Suppose a borrower buys a highly energy efficient house. One might argue that she will save money on heating and, thus, is more likely to repay her debt, which, in turn, should be reflected in an improvement of her credit rating. However, this would only hold true if she did not change her consumption behaviour. And this is where the currently unobservable risk lies: the unexpected deviation from the former consumption pattern after mortgage origination. For instance, the borrower could decide to change her heating habit and might use up the EE savings on more intense or careless heating. In the literature, this scenario is referred to as the rebound effect. The rebound effect generally refers to an increase in the request of energy services due to the decrease in the effective price paid by the consumer. However, she might also spend her EE savings on some other consumption products, such as a new car or kitchen. In addition to the shift in consumption, the borrower might also self-select herself into the data sample. In this case, not EE per se but a different factor would affect the credit risk. For instance, environmentally conscious borrowers are more likely to buy an energy efficient building and their attitude towards debt repayment might differ from other borrowers. In order to identify the causal link one has to account for such subtle factors. This is a challenging task as the bank typically does not collect soft information. However, there are solutions to overcome this issue. Regarding the consumption pattern, the bank typically possesses the borrower’s financial transaction history and can track its development over time. Additionally, utility companies collect energy consumption data, which allows to measure the rebound effect. To account for the other soft borrower characteristics, customer surveys would be helpful in order to estimate their general attitude towards the environment.
As of date, it is a challenging task to meet all data requirements for a causality analysis. The reasons being lack of data and privacy concerns. Thus, we focus our analysis on the correlation between EE and PD, and leave the causality question for a future study. In the following, we present our findings from the Dutch mortgage market.
Using loan-level data from the Dutch mortgage market, we investigate the relation between a building’s energy efficiency and the probability of mortgage default. By focusing on residential buildings exclusively, our sample consists of mortgages issued on more than 120,000 dwellings. We supplement the dataset with provisional energy efficiency ratings that are assigned by the Netherlands Enterprise Agency (Rijksdienst voor Ondernemend Nederland, RVO) to all Dutch buildings that are not yet supplied with the actual energy performance certificate (EPC) rating. RVO provides rating categories for 60 pairs of different building type and construction period combinations in the Netherlands. This allows us to match the loan data with EE ratings according to building type and construction year. Additionally, we exploit the fact that the ratings change asynchronously across the different building types in order to disentangle the energy efficiency-component from building type- and building age-specific effects that are typically associated with borrower’s risk of default. We employ two empirical methodologies – the Logistic regression and the extended Cox model – and find that energy efficiency is negatively correlated with a borrower’s likelihood of default on mortgage payments. The results hold if we account for borrower, mortgage, and market control variables. The findings also survive a battery of robustness checks. As an additional exercise, we investigate to what extent the degree of energy efficiency plays a role on borrower’s credit risk. Our findings suggest that mortgages on more efficient buildings are less prone to default. However, the findings on the degree of energy efficiency are less significant than the baseline results.
To date, only few studies attempted to investigate the correlation between EE and PD. To our knowledge all these studies were focusing on the US market. Our empirical exercise is among the first ones to establish a link between EE and PD with European mortgage data. We are optimistic that this field of research will grow as more European countries and banks commit themselves to collect and share their data.