At this year’s Credit Scoring and Credit Control Conference in Edinburgh, my presentation posed the question: Are we too reliant on economic response models (ERMs) for loss forecasting, and is it time to take a different view?
ERMs have become the industry norm for linking macroeconomic conditions to default rates, but their limitations are well known. They are often built on a narrow set of variables and can produce counterintuitive results that reflect outdated economic thinking rather than the world as it is today, or how it might be tomorrow.
In my presentation, I looked at an alternative: a bottom-up, affordability-centred approach that combines long-run macroeconomic data with granular, account-level insights. The aim wasn’t to present a finished solution, but to encourage a new way of thinking about how we link the economy to customer behaviour, and ultimately to losses.
In this blog, I’ll recap the key themes from that session…
Why economic response models (ERMs) are limited
Economic response models (ERMs) have been the industry standard for linking macroeconomic conditions to default rates. They remain widely used, but several weaknesses stand out:
- Narrow set of variables
ERMs typically rely on a handful of indicators such as GDP or unemployment. This risks missing important drivers of customer behaviour, and the variety of variables considered tends to be limited. - Assumptions rooted in the past
The patterns often reflect outdated economic thinking rather than today’s reality or the conditions we may face tomorrow. Counter-intuitive outputs are not uncommon. - Adoption through familiarity
Like the old IT saying, “nobody ever got fired for buying IBM,” few risk teams are challenged for using them. There are no regulatory constraints on moving away from traditional ERMs, but they remain the industry norm because it’s easier to keep doing what everyone else does. - Lack of customer-level granularity
By working at an aggregate level, ERMs struggle to capture how economic change affects customers differently. This leaves a gap in understanding who is resilient and who is exposed to changes in the economy.
So, if ERMs are the accepted approach, but not the most effective, how do we move towards something more robust? Here’s where affordability comes in.
Affordability as the missing link in loss forecasting
An alternative approach is to look at affordability as the key driver of customer behaviour during periods of economic change. Instead of relying only on top-down indicators, this method connects macroeconomic shocks to the ability of individual households to manage their finances.
A more detailed view of household finances
Standard affordability checks often just “add a bit on” for the cost of living crisis, but that misses the real dynamics at play. For example, they take account of current energy or housing costs and perhaps add an allowance for cost-of-living pressures. What they rarely do is model the effect of future shocks. A sudden rise in gas prices or a fall in mortgage rates can change a household’s financial position dramatically (or they may not), and these impacts need to be captured in a structured way.
A two-stage modelling approach
The method I outlined in Edinburgh works in two steps:
- Use macroeconomic data to forecast the impact on categories such as salaries, pensions, housing and energy costs.
- Apply those changes at customer level to assess how individual households are likely to be affected.
To support this, two complementary data sources are essential: recent account-level transactional data, and more many years’ of publicly available time series data. Used together, they provide both the granularity and the historical perspective required to make the approach effective.
Linking affordability to default risk
The central hypothesis is that the customer’s ability to pay is the factor most sensitive to changes in the wider economy. By modelling how macroeconomic shifts feed through into affordability, it becomes possible to produce default forecasts that better reflect real household behaviour.
This perspective recognises that the economy does not affect all customers in the same way. Some households are resilient to shocks, while others are highly exposed, and capturing that difference is vital for more accurate loss forecasting.
Practical applications of an affordability-based approach
An affordability-centred approach is not just a theoretical exercise. There are loads of applications across both regulatory and customer-focused activities.
#1. Strengthening IFRS 9 models
One of the first areas where this approach can add value is IFRS 9. A bottom-up view of customer finances can enrich forward-looking probability of default estimates, providing a more grounded basis for scenario design and outcome measurement.
#2. Enhancing stress testing and capital forecasting
By linking macroeconomic shocks to affordability, firms can build stress testing scenarios that reflect the real pressures households are likely to face. This can lead to more accurate assessments of portfolio resilience and sharper insight into capital requirements under adverse conditions.
#3. Supporting customers directly
The same modelling can also be applied at an individual level. It can help identify which customers are financially resilient and which are more exposed to changes in the economy. That understanding creates opportunities to intervene early — whether through tailored repayment plans, additional support for vulnerable groups, or proactive communication about upcoming pressures.
#4. Moving from aggregate to granular risk insight
Perhaps most importantly, this approach shifts the perspective from broad economic indicators to detailed household behaviour. It recognises that not all customers are affected equally, and it enables lenders to design responses that reflect those differences.
Testing the approach: Proof of concept and validation
For this approach to be deployed, it needs to be tested in collaboration with the right type of organisation. Because it is focused on customer-level behaviour, the most suitable partner would be a lender with a large retail portfolio. Credit cards, personal loans and mortgages are all strong candidates, provided there is a sufficient history of defaults to validate the outputs.
Validation itself should not be unusual. Each component of the model can be subject to the same cycle of monitoring and review applied to other risk models. The critical factor is ensuring that both the macroeconomic linkages and the customer-level affordability elements are tracked and adjusted over time.
This creates a framework that can stand up to regulatory scrutiny while also delivering practical insights for decision-making. The challenge now is finding a partner willing to take forward a proof-of-concept and demonstrate the value at scale.
Looking ahead: The future of loss forecasting
Over the next three to five years, I expect regulators to encourage more nuanced approaches to loss forecasting. The direction of travel is already clear. Today, the focus is firmly on climate risk. Tomorrow, it may be conflict risk, or even the need to plan for another pandemic.
The common theme is that risk models will need to account for an ever-increasing basket of factors, many of which sit outside traditional economic indicators. An affordability-centred approach provides a flexible framework to achieve this. By linking macroeconomic shifts to household-level impacts, it can adapt as new risk categories emerge.
Loss forecasting will never be an exact science, but it can be made more relevant to the realities customers face. That is the opportunity, and the responsibility, for all of us working in credit risk today.
Keen to hear more from the Edinburgh Credit Conference? Take a look at this blog: Federated learning: The modern approach to fraud prevention
