Stress Testing: Should A Bank Build or Buy?

4-14-14-Trepp.pngThe banking industry received a rude awakening in March when the Federal Reserve rejected five banks’ capital plans and stress testing reports, indicating that the stakes are rising for stress testing. More banks are beginning to develop stress testing programs and directors are faced with the decision to rely on internal systems and models or to look beyond their institution to third-party data and solutions. More commonly known as “build versus buy,” this choice is a difficult one to make, as there is no one-size-fits-all approach. To provide decision-makers with some helpful guidelines, we focus on the following key factors that will determine which route a bank should take.

The completeness of a bank’s data set will be a strong determinant in making the decision to build or buy. Small and large banks alike have been criticized by regulators for their stress testing models resting on what is deemed an inadequate base. In August 2013, the Fed published a report citing strengths and weaknesses identified in large bank holding companies’ stress testing approaches. A key criticism was, “weaker practices [that] failed to compensate for data limitations or adequately demonstrate that external data reasonably reflect” the bank’s exposures. A bank’s data should contain the following:

  • Metrics on loan performance, such as payment and delinquency history
  • Metrics that influence payment behavior, such as loan-to-value ratios, debt service coverage ratios, credit rating and cash flows
  • Measures of loss severity
  • Descriptors of the relevant market backdrop for the loan, borrower and property

In addition, the data should span at least a full credit cycle and be able to accurately represent the market. If a bank’s data sets are lacking in any of these areas, management should strongly consider the “buy” side, whether that involves augmenting data with that of a third-party or using a third-party model.

If a bank does not have sufficient internal resources to complete the required work to build a stress testing model, some level of outsourcing should be considered.

Staffing: To conduct stress testing completely independently, a bank will need to allocate personnel for management and oversight, data analysis, model creation, documentation, testing and validation. To provide a point of reference, large banks employ teams of 50 or more people for their stress testing work. A hybrid approach, in which a bank builds some components of the model and outsources other work, is also an option. If a bank chooses to use third-party models, regulators expect its management team to understand the ins and outs of the tools they are using, particularly in their stress test reporting.

Development and Maintenance: In developing a stress testing system, a bank must commit substantial resources to both systems and people for data warehousing, creating forecasting models, output collection, reporting, and of course, maintenance of the systems. Requirements and expectations will evolve over time, which means there will constantly be areas in need of improvement. While the largest commitment is upfront when a bank decides to build, it is also a sustained commitment.

Cost considerations are undoubtedly a significant factor in a bank’s decision to build or buy. Choosing to build incurs the highest cost—for both initial development and ongoing maintenance. Beyond the maintenance expected as a result of evolving regulatory requirements, technological advances will also involve a substantial investment. Spending will vary by the size and complexity of the bank and its system. Last year, three banks in the $50-billion asset range—Comerica, Huntington Bancshares, and Zions Bancorp.—announced that they were each spending in the neighborhood of $10 million on their regulatory compliance efforts, a large portion of which was understood to be for stress testing expansion. Smaller institutions will not be expected to spend as much, but building a system will easily cost hundreds of thousands of dollars upfront, and similar amounts to maintain.

The build versus buy decision for stress testing is not always clear-cut, as there are advantages to both. As banks look to increase their focus on qualitative aspects of the process and the entirety of the reporting procedure, building an internal model will make the bank an expert on their stress tests. Using internal data also ensures that reporting accurately reflects the bank’s current operations. On the other hand, third-party tools can easily prevent a bank from falling short due to budgeting or data integrity. Given the maintenance involved, buying is also considered less of a risk. Whichever way your bank goes, including a hybrid approach, it will benefit from more rigorous approaches to capital planning and capital adequacy stress testing.

Matt Anderson