To meet the new for current expected credit losses (CECL) requirements, “you will need a lot of data,” said Thomas Caragher, Senior Product Manager of Financial & Risk Management Solutions at Fiserv, a US provider of financial services technology. Seven to ten years of data is not unreasonable. The payoff to massive data-gathering is that “you will be able to build strategy more effectively if you have information.”
But what to do with the reams of data? First, you have to make some sense of it. “Start by correlating the data,” he said. Oil production might be correlated with loan demand or real estate prices. Agricultural production is more likely tied to futures.
“There’s a debate about whether to look at national or regional numbers” for unemployment, oil production, and agricultural variables. Real estate is most likely regional.
In his examples, Caragher found strong correlation between national unemployment and the national loss rate. He found no correlation between inflation and loss rate.
The next question to consider is how to segment the data. For example, data points could be grouped according to cost center or credit rating or loan officer. Other segments could be made for zip code or age of customer.
Caragher said that polls by Fiserv showed 90 percent of respondents “planned to use their data for strategic planning.” He questioned whether they could actually do so, because the most common loan loss methodology used is historic loan loss (43.5 percent of respondents), which is a series of snapshots.
Only 7.4 percent used models for probability of default and loss given default (PD/LGD); migration (2.3 percent), vintage (2.8 percent), discounted cash flows (1.8 percent). Nearly half (45.4) used “a mixture of the above [models].”
“Methodology shows most respondents are reactive,” not proactive, he said. The volatility in the market is increasing, and excellent data handling and models are necessary so that “as things change, you’re better prepared.”
There remain a host of problems to solve in the areas of profitability, budgeting, reconciliation, and so forth (see graphic). “You must have one central source of data to address all these problems,” he emphasized. “CECL and risk are complex problems to solve but the data can answer questions in stress testing and scenario analysis.
On this note, he ended the presentation with about 15 minutes left for Q & A (not described here, but audible in the webinar recording – see link below). Caragher was speaking on new guidelines for current expected credit losses (CECL) at a webinar sponsored by the Global Association of Risk Professionals held on April 11, 2018. ª
Blue “data cartoon” is from https://www.operasolutions.com/events