Five Tech Challenges Banks Face in Next Year's Stress Tests

As banks prepare for the next round of stress tests, they face several tech-related challenges, including broadening the exercise to include detailed modeling of deposits and meeting regulators' demands for ever-greater documentation.

The tests, which the Federal Reserve introduced in the wake of the financial crisis, require banks to estimate how various aspects of their business would fare under possible economic scenarios, from so-so to terrible. In the early stress-test exams, field examiners made it clear that they expect banks to have sophisticated analytics and structured processes and modeling techniques in place to test their capital levels, loan portfolios and liabilities.

Those early stress tests focused on capital adequacy: Does the bank have enough loss-absorbing equity to stay in business if another financial crisis happens? More recently, examiners zeroed in on the way banks stressed their loan portfolios.

In preliminary conversations for the next round, observers say, the Fed is focusing on preprovision net revenue. As part of that modeling, regulators are asking banks to project how their deposit balances and their costs would change under different economic conditions.

"They care about whether you're going to have to pay more for the deposits as rates rise and be able to produce enough profit," said Pete Gilchrist, a managing director at the consulting firm Novantas.

Banks over $50 billion in assets are subject to the Fed's Comprehensive Capital Analysis and Review, for which capital plans must be submitted Jan. 5, and the Dodd-Frank Act stress tests, for which reports are due March 31. Banks in the $10 billion to $50 billion asset range must carry out their own stress tests under supervision of their federal regulator. Prep work is well underway for all of these.

The latest stress test mandates bring about several new challenges for banks, to which at least part of the solution is software:

The need to stress-test deposits. Regulators are expecting detailed modeling of deposits, Gilchrist said.

"Some of the largest banks have come out and publicly stated that their CCAR spending this cycle is in the triple digits of millions, and the majority of that is PPNR spending. It is highly technical modeling for the largest banks."

Most experts are expecting a slight increase in deposit rates next year.

"The interest rate issue is significant to [regulators], they are concerned about banks not being prepared for how interest rates are going to shift, including from the deposit world," said Peter Cherpack, senior vice president of Ardmore Banking Advisors in Pennsylvania. "So they're firing some guns across the bow, saying not only 'are you preparing for what's happened to your loans?' but 'are you prepared for what's going to happen to your deposits?'"

While most observers expect any increase in deposit rates next year to be gradual, the Fed is asking banks to forecast how they would do if deposit rates increase faster than anybody thinks they will. Guesstimating is not enough — banks are under pressure to take an analytic approach.

At BancorpSouth in Tupelo, Miss., as Ty Lambert prepares for the bank's next stress test, his biggest challenge is making it more comprehensive. Last year, his team focused largely on building an analytic framework for estimating credit losses. This year they're working to apply the same framework to deposits, loan growth and risk-weighted assets, to determine the impact various macroeconomic scenarios would have on each.

"What we're trying to do this year is for deposits, for example, look at each one of our markets — not only how did our market share grow or shrink during the last financial crisis, but also how did total market size get affected by the last financial crisis?" said Lambert, the head of treasury analytics at the $13 billion-asset bank. "We are trying to find ways to develop correlations between how our deposits behave in these times of economic stress and how they tie to these metrics. We'll do the same thing on the loan portfolio."

BancorpSouth began outsourcing an annual core deposit study a few years ago to McGuire Performance Solutions, and is now considering expanding that work to include deposit stress testing.

Regulators have said it's ok to outsource as long as the bank has a good understanding of the analysis work being conducted, Lambert said. "They want us to be involved with everything that goes into the analysis, and use the consultants to do the legwork," he said.

Regulators also expect the bank to document the dialogue back and forth between bankers and consultants.

"We keep minutes of all the meetings we have with them so we can show the regulators when they come back in that we've been very involved in the process all the way through," Lambert said.

The demand for documentation. In fact, documentation is BancorpSouth's second greatest stress-testing challenge — but Lambert said he fully understands why regulators are demanding so much of it.

"When regulators come into an institution, they have to lean on the documentation to understand what's going on in the stress testing process," he said. "Stress testing is bringing about a golden age of technical writing."

Missing data. The data management challenges of stress-testing, many banks report, are huge.

Small banks haven't traditionally stored the historical data needed to model performance of deposits and loans because it hasn't been required. Many banks' core systems have only a year's worth of data on deposits and loans.

"When you go into the process, one thing you have to work on is internal systems," said Brian W. Jones, senior vice president and chief lending officer at Newfield Bank in Newfield, N.J. "You have to go back and make sure all the right fields are filled, which is highly unlikely in a small institution."

Making sure the data is correct and that the output is a report that examiners will approve and that senior managers and the board can use in business decisions takes significant effort, he said.

"Data is the key. You can't do anything without the data — you can't even do the minimal things recommended for community banks of any size without basic coding of the portfolio in a way that allows them to look at concentrations," Cherpack said. "Some banks don't even understand that yet. They're worried about getting quants to do the modeling on the system. But they don't realize that without the data, they can't even do that."

Some stress-testing rules let banks use proxy data, but with limits. "You could buy S&P or Moody's data so you can say 'this is like us, this is how we're going to base our projections,' but only for a period of time," Cherpack said. "Regulators are going to be expecting you to collect your own data at the same time."

Missing data is also an issue for large banks. For instance, in trying to stress-test business deposits, they struggle to gather all the needed account information.

"Consumers are relatively easy to identify because they tend to go by one name and have one identifiable number," said Gilchrist at Novantas. "When you start getting into business accounts, you run into an additional layer of complexity, because you're dealing with companies with multiple operating accounts and multiple entities. It's hard even to figure out who the customers are and how long they've been customers. That's an exercise that can span three months. If you've already got that data managed in a logical away, you're ahead."

His firm provides software it calls a "fit-for-purpose data mart" that collects and organizes data for stress testing and for pricing analytics.

Lack of clarity — and a sense of urgency — for community banks. Regulators have not specifically ordered small banks to do stress testing, but have asked them to do stress-testing-like risk analysis. The distinction is not obvious.

"There are no enterprise-wide stress testing requirements on small banks less than $10 billion" in assets, Jack Jennings, a senior associate director for the Federal Reserve, said at a recent risk conference. But "we do think that there is a place for sensitivity analysis, for scenario planning. I think the conversations we've already had about concentrations are a good place to say, 'look what happened,' should there be a change, in terms of market conditions or interest rates relating to some of those significant concentrations.'"

Small banks need to have more of a sense of urgency around stress testing, Cherpack said.

"That's the thing I keep trying to get out — don't wait until you have to do DFAST to start building up your data," he said.

"In today's environment community banks are looking to acquire or be acquired (or grow organically)," Cherpack noted. To do any of those things, "you need regulators to be very comfortable with you," he said.

"The community bank will say 'I want to buy this bank,' the regulators will say 'I don't think you're ready.' You don't want that [situation]. Why not do these things proactively now? Then, when you want to do something, the regulators are comfortable with you."

The minimum banks should be doing is tracking some basic characteristics, like collateral types, purposes, products, and regions so they can do meaningful tests. "Leaving it all for the end is brutal."

The need to analyze concentration risks. Regulators have been clear about community banks' need to work on concentration management, Cherpack said. "Typically, community banks look at things on a deal-by-deal basis and don't necessarily look at the market and the concentrations they have in property types, market types, and locations."

Community banks naturally tend to be concentrated — chances are a small bank in a farming community will make a lot of agricultural loans.

"Regulators are not saying 'you can't be [concentrated],' but they're saying 'you need to understand that.' That's what you need to stress," Cherpack said.

Community banks tend to have most of their data in their core system, which in many cases doesn't provide a place to, say, track cash flow from income-generating properties.

"That data's in a loan origination system somewhere or a spreadsheet or loan file, not in any automated place," Cherpack said.
A credit data mart could be used to augment the core. "We built an uncomplicated process that's relatively inexpensive to host a data mart and take all this stuff they have in spreadsheets and loan files and put it in one database where they can do stress tests and concentration reports," Cherpack said.

Because Newfield Bank makes a lot of commercial real estate loans, "the OCC takes a harder look at our CRE portfolio than they would for institutions that have fewer loans in that bucket," Jones said. The bank uses software from CRE Insight to handle the reporting.

"Regulators like to see narrative," Jones observed. "They don't want to see numbers. They want to know, what do these numbers mean to your bank?"