Loan losses are part and parcel of the cost of doing business for credit providers. However, inaccurate forecasts can have severe business repercussions for the lender, such as: negative impacts on both the level of profitability achieved and the stability of profitability as volatile provisions actuate through the P&L poor forecasting may be interpreted as an inability to hit targets, reducing investor confidence and potentially raising the cost of borrowing conservative forecasting can prove ‘expensive’, tying up funding resources and missing out on valuable opportunities. On the other hand, optimistic forecasting exposes the lender to levels of risk that could prove ‘terminal’ should the risks materialise.
To address the need for accurate forecasting, several analytic techniques are employed throughout the lending industry, ranging from relatively simple roll rates and vintage based models, to more sophisticated tools such as Markov chains, PD, LGD & EAD estimates, dual time dynamics and Monte-Carlo simulation. Each technique has its own set of strengths and weaknesses – some only consider historic repayment behaviours, while others incorporate forward-looking economic scenarios. Some are inexpensive to build and are easy to implement, while others are more intensive to produce and have high maintenance requirements.
Loss forecasting isn’t simply a technical exercise. As an essential ingredient to the capital adequacy planning process and a key influencer of business outcomes, as described above, the approach to forecasting needs to be embedded in well-defined policies and governance processes. Also, there isn’t a one-size-fits-all solution to loss forecasting. The right approach for a given client depends on the following:
- product mix
- risk profile
- collections activity
- macro-economic environment
- time-horizon(s) to forecast over
- data history available
- portfolio forecasts v account level forecasts, and
- the need to align to Basel II
Connected Analytics has broad experience in designing bespoke loss forecasting solutions, both from learnings while at CA and previous experience where most of our team worked ‘client side’, either building the statistical tools needed to generate forecasts or managing credit portfolios as recipients and users.
This service is often provided as part of the following Solution Sets: