Lifting the HEM: building better benchmarks to assess household expenses in loan applications
By Michael Mifsud
Some of the more striking revelations of the Hayne Royal Commission were stories of consumers being extended loans that were always going to be well beyond their capacity to repay.
They cast a sharp spotlight on the methods used by lenders to determine loan serviceability, including assessment of household expenses. Recognising these concerns, ASIC recently announced a review of its Regulatory Guide 209, which sets out its expectations for meeting the responsible lending obligations under Chapter 3 of the National Consumer Credit Protection Act.
A key part of this review is consideration around the appropriate use of the Household Expenditure Measure (HEM), a benchmark for the minimum expected level of expenditure for a household. The HEM gives lenders a clear indication of whether the expenses reported by a prospective borrower are likely to be accurate and allows them to calculate the maximum level of repayments that a household is likely to be able to afford.
Critics of the measure argue that many lenders, as a matter of course, will fail to validate reported expenses so long as they meet the HEM, which is arguably a low bar to clear, and borrowers whose actual expenses may be higher than this are extended loans that they are subsequently unable to service.
Supporters of the current system contend that broader requirements for manual vetting of applicants’ expenses would impose undue cost on the lender and require onerous requests for information from the customer for what, in most cases, would be an identical result.
As an efficient means of assessing a loan application with minimal requirement for precious human capital, the concept of a benchmark is an attractive one. But while the HEM may be efficient and appropriate in most cases, the Royal Commission demonstrated that on the rare occasion when systems do fail and borrowers are approved for loans well beyond their capacity to repay, the human cost can be catastrophic.
But what if there was a way to improve the benchmark, increasing its accuracy and specificity to each customer whilst minimising costs to the lender?
Limitations of the HEM
The HEM is drawn from the ABS Household Expenditure Survey, the most recent installment of which surveyed 10,046 households across the country. This may seem like a large number, but even a sample size of this magnitude can only be segmented so many times; indeed, for the purposes of the HEM it is segmented across only three attributes – income, household type and geographic location. This means that for every loan applicant, their maximum capacity to service their loan, absent self-reporting to the contrary, is considered effectively identical to every other household of the same size and income group in their region.
The HEM categorises over 600 expenditure items from the survey and calculates a minimum level of basic expenditure – those costs not likely to be easily eliminated by borrowers. However, it fails to account for several important expenditure items, such as higher education loan repayments and private school fees, which are not common to every borrower, but for those who have incurred them they may be completely non-discretionary.
A dynamic, data-driven alternative
By drawing on aggregated bank transaction records across the full spectrum of Australian consumers, we can create a new system of benchmarks that are more accurate and representative of each household’s true position, and responsive to changes within the economy and in a household’s circumstances. Utilising real data on spending behaviours from a much larger sample size would allow for:
- A more robust and high-confidence benchmark range than the survey-based data that underpins the Household Expenditure Measure (HEM) approach.
- More granular segmenting and measuring against more appropriate benchmarks based on variables such as age, dwelling type, owner vs renter and, potentially, other clustered segments based on consumer behaviour.
- Assessment of a customer’s reported expenses against a distribution of benchmark values, rather than a set of single HEM figures, allowing lenders to deploy a spectrum of further investigative measures based on relative degrees of plausibility.
These benchmarks could also be established across a range of expense categories rather than a sum of all expenses, allowing the query of specific missing or apparently under-reported expenses.
The power of the system to estimate a customer’s ability to service a loan after approval could be further amplified by assessing the change in spending behaviour of households in similar circumstances before and after similar loans were approved. Quantium can also identify the kinds of expenses that are new for customers or that are normally reduced by a customer after obtaining a loan, enabling lenders to work with prospective customers on the kinds of discretionary expenses they will realistically be able to amend.
Everybody benefits from a better benchmark.
Borrowers can often be resistant to the idea of more granular assessment of serviceability, but using highly calibrated, aggregate benchmarks means that substantial upside can be realised with minimal intrusion into the customer’s personal data. Lenders would reduce the risk associated with the HEM whilst improving the customer experience, as a result of being able to better identify where information is incomplete or inaccurate, wasting less of their customers’ time with needless inquiry.
Customers will experience less unnecessary rejection, as they are compared against benchmarks more appropriate to their individual situation, and will be offered loans they can better afford, leading to improved long-term financial outcomes.
No system will ever fully eliminate the risk of a customer borrowing more than they can afford, but a vastly more accurate system need not necessarily be a tradeoff with the lender’s time and money. Most importantly, it could be made available quickly and for relatively low impost if lenders and regulators were prepared to pursue it.
Quantium is a world leader in data science and artificial intelligence. Established in Australia in 2002 and now employing over 750 people, Quantium works with iconic brands in over 20 countries, partnering on their greatest challenges and unlocking transformational opportunities. Discover more about our Q.Refinery banking product enables banks to harness their data to enhance the customer experience, reduce risk and realise growth.