Nuri Khayal and Jonathan Loke

Many households in the UK have seen their mortgage payments go up since mortgage rates started to increase in 2022. In the current environment of higher rates, the question of how much a household can comfortably spend on their mortgage payments before getting into financial distress is particularly relevant. This blog shows that households which spend a larger share of their income on mortgage payments are at a higher risk of being in arrears. But in contrast to pre-existing work on the subject, we do not find evidence of a critical threshold after which the risk increases much more sharply. These findings imply that changes in the indebtedness across the whole mortgagor population, not just the tail, matter for financial stability.

Some countries restrict lending to borrowers with high debt-servicing burdens

There are different metrics that measure a household’s debt-servicing burden. The most common one is the gross debt-servicing ratio (DSR). It is calculated by taking a household’s mortgage payments (including interest and principal) and dividing them by their pre-tax income. Some countries have regulations in place that restrict lenders’ ability to issue mortgages to borrowers that exceed certain DSR limits. The rationale for these measures is that borrowers with higher DSRs are more likely to get into financial difficulties as they have less of a buffer that cushions them from potential increases in interest rates or losses to their incomes. DSR limits vary across countries, but they are often in the range of 30% to 40% (see recent BIS report).

Consistent with this, previous analysis by the Bank of England, featured in the December 2019 Financial Stability Report and the August 2020 Financial Stability Report, suggests that households with gross DSRs of around 40% are at a much higher risk of missing their mortgage payments. A previous Bank Underground post from 2016 comes to a similar conclusion. It shows that the DSR threshold above which the risk of mortgage payment shortfalls increases more sharply differs across surveys, but typically lies between 30% to 50%. The findings in these previous publications were derived by grouping mortgagors into different buckets based on their gross DSRs and then comparing the share of mortgagors in arrears across these buckets.

In the UK, the share of mortgagors with high debt-servicing burdens has remained flat during the current tightening period

Recent UK loan-level and household survey data suggests that the share of mortgages with DSRs at or above 40% has been broadly flat during the current tightening cycle for both new lending and the stock of mortgages (Chart 1). At the same time, the DSR distribution for new mortgages has noticeably shifted to the right. This has not caused a shift in the DSR distribution in the stock because mortgages issued during the current tightening period make up only a small share of the overall stock and because many borrowers have experienced strong nominal income growth which has cushioned the impact of higher interest rates.

Chart 1: DSR distribution for new lending and the stock of mortgages

Sources: Bank of England/NMG survey (right panel) and FCA Product Sales Data (left panel).

We find no evidence of a critical threshold

The analysis presented in this blog is based on data from two UK household surveys: the Bank of England/NMG survey and the ONS Wealth and Assets Survey (WAS). The WAS is a survey conducted by the Office for National Statistics (ONS) every two years and contains a wide range of questions on UK households’ balance sheets, their incomes, their mortgage, and the property they live in. The Bank of England/NMG survey is a survey carried out every six months by NMG Consulting on behalf of the Bank of England. It provides a timelier update of developments in household finances compared with the WAS but at a less granular level.

For each survey, we estimate a model that predicts mortgage arrears at the household level for different levels of gross DSRs. The model includes a wide range of control variables, including time-specific effects, other household-level financial variables (eg LTVs and total savings to income ratio), household characteristics (eg region) and mortgage-level characteristics (eg repayment type). Our approach differs from previous analysis published by the Bank of England in two ways. First, it estimates the relationship between DSRs and arrears at the household level, which means it does not rely on grouping mortgagors into different DSRs buckets. This removes the risk that results are driven by the way the DSR buckets are constructed. Second, controlling for other variables that are correlated with DSRs and have an impact on the likelihood of arrears (such as total savings to income ratio) allows to estimate the impact of DSRs on arrears more precisely. The relationship between DSRs and mortgage arrears is estimated by fitting piecewise cubic polynomials for different parts of the DSR distribution and splicing them together. This is a more flexible approach compared with classic linear models as it allows the functional form that describes the relationship to differ across the DSR distribution and thus to detect potential critical thresholds. 

Results are illustrated in Chart 2 which plots the predicted probability of a household being in arrears given their DSR, holding all other variables constant at their average values. The results have four major implications:

  1. If a household does not spend more than around 15% of their pre-tax income on mortgage payments, a higher DSR does not increases their risk of payment shortfalls.
  2. For households who spend more than around 15% of their income on mortgage payments, a higher DSR implies a higher risk of payment shortfalls.
  3. Beyond the 15%-threshold, the risk of payment shortfalls increases broadly linearly with DSRs. In particular, there is no evidence that the probability of arrears approaches 1 for very high DSRs. This highlights that some households with very high mortgage debt burdens might still be able to service their debt, for instance by drawing on their savings or other types of assets, or by borrowing from friends and relatives. The results also partly reflect that in both surveys, households are defined to be in arrears if they have missed more than two months’ worth of repayments which means households that are unable to repay their mortgage as a result of a very recent income shock are not captured. In addition, the risk of payment shortfalls might increase more sharply during a recession such as the global financial crisis which is not covered in either of the two samples.
  4. The probability of arrears depends on which survey you look at. One reason is that the relevant questions on mortgage arrears slightly differ across the two surveys. Another reason could be that the NMG survey is conducted online and households might be more likely to select themselves into the online panel if they are in financial distress (see Anderson (2016)).

Chart 2: Predicted probability of a household being at least two months in arrears with mortgage payments given their gross DSRs

Note: Number of separate polynomials estimated for each sample is selected by minimising the Akaike information criterion which optimises the trade-off between model fit and simplicity of the model. Both samples are based on repeated cross-sectional data. NMG survey sample covers the period from 2015 to 2023, WAS sample covers the period from 2010 to 2020. Shaded areas represent 95% confidence intervals. Higher uncertainty in the tails reflects small number of observations with very low or high DSRs.

Results have important policy implications

The results presented in this blog suggest that even for mortgagors with moderate DSRs, an increase in their debt-servicing burden implies a higher risk of payment shortfalls. Yet, we do not find evidence of a critical threshold above which this risk increases much more sharply. Our results imply that changes in the whole DSR distribution matter for financial stability, not just changes in the tail. Consequently, the recent shift in the DSR distribution for new lending implies that new mortgage lending in the UK has become riskier during the current tightening period even though lending at DSRs above 40% has remained flat. At the same time, the DSR distribution in the stock of mortgages has remained broadly constant, indicating that mortgagors have remained resilient overall.

The results can help policymakers to assess risks in the mortgage market more effectively. On the one hand, the results do support using indicators such as the share of households above a certain DSR threshold like 40% to measure tail risks in the mortgage market given that those households are more likely to miss their payments. On the other hand, our results suggest that policymakers should also monitor changes in the wider DSR distribution when assessing borrower resilience. The results in this blog do not necessarily challenge the calibration of the Financial Policy Committee’s loan to income flow limit.

Nuri Khayal works in the Bank’s Macro-financial Risks Division. Jonathan Loke worked as an intern in the Bank’s Macro-financial Risks Division.

If you want to get in touch, please email us at bankunderground@bankofengland.co.uk 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.


Leave a Reply

Your email address will not be published. Required fields are marked *