Any financial forecast has to take into account how the economy is expected to behave over the forecast period. In developing a CECL forecast that means projecting out over the remaining life of loans in the portfolio. For residential mortgages this typically is 30 years. The longer the term, the greater the uncertainty. Figuring out how to reduce uncertainty is the challenge here.
Identifying Key Forecasting Factors
Economic forecasting is both art and science. Many firms offer their services for a pretty penny. General forecasts for the economy overall are useful in estimating high level factors such as GDP, employment, interest rates and inflation. Most of these forecasts look out 1 to 5 years. Therefore, they can provide near term direction but likely don't extend to the end of the CECL forecast period.
The key factors will vary by product. With shorter term products, the timing of any change to the economy is most important. For longer term products it is important to consider both credit and prepayment risks since both will impact future losses. For example, in October 2020 long-term interest rates are at all time lows. This has caused a spike in refinancings, reducing balances and risk. This reduces the risk in the portfolio but may lead to greater risk down the road for new accounts. The historically low rates decrease future refinancings thereby increasing balance persistency and potential risk. Also, if interest rates increase home values may fall increasing the loss given default.
How Will It Affect Portfolio Risk
Arguably it's really a question of whether 1) your historical portfolio performance data includes a full economic cycle, the highs and lows; and 2) the general economic forecast includes a significant change in the economy's health. If the answer to the former is yes then there's less concern about the latter for newer accounts given that the lifetime performance info factors in the ups and downs. However, for older accounts that are nearer maturity a short-term downturn forecast is likely to take losses higher than the trend lines would indicate. The expected severity of any short-term downturn should be considered in how much to raise expected losses.
As you pull together your forecast you will need to:
- Start with baseline loss curve
- Determine if loss curve includes full economic cycle
- Assess any changes to default frequency
- Consider changes to average write-off amounts
- Adjust for local effects
A general forecast for the economy as a whole would not take into account local factors impacting your customers. For example, if a major employer is laying people off or closing your customers will feel the impact of fewer jobs and income leading to a local contraction. This becomes a judgment call in determining how severe the impact will be. And naturally this will vary by product. How will it impact local real estate values? How big is the impact likely to be? Do we know how many customers will be directly affected? Localized impacts are the last step in tweaking the loss forecast.
Minimizing Forecasting Uncertainty
Each financial institution can come up with its own forecast. There's a decent chance that it might come up with a reasonably accurate one. However, there will be a 100% chance that it will not be completely accurate. Even the forecasting experts can get it wrong.
We believe that taking a 'wisdom of crowds' approach can improve the accuracy of the collective forecast. Under this approach we take the average of every participant's forecast values. These averages are likely to be more accurate overall.
