While recent government "stress testing" related to the group of 19 (G19) largest financial institutions has become a distant memory, the impact of the test continues to spill over into community banking. We were particularly dismayed recently after we saw some in the press apply the same stress testing methodology to community banks. This entire exercise ignored the fact that the risk characteristics of Community Bank ABC are vastly and profoundly different than those of MetLife, Amex or Goldman (some of the firms originally tested). In fact, of the G19 tested, only half even remotely look like a bank in any traditional sense.
A key problem we have with those who try to apply the G19 stress test methodology to community banks is that such testing wasn't originally designed to be used that way. The intention of the G19 test was to determine whether or not a discrete group of "systemically important" financial institutions would need additional capital if certain scenarios played out. The goal of the test was to determine whether or not the Treasury would need to ask Congress for more funding and to provide the data and time to build a case in the event such funding was needed. For community banks, there remains plenty of cushion in the original TARP program, so this type of testing was not needed.
Another reason we believe that applying this test to community banks is counterproductive is due to the vast geographical differences between the G19 and more typical community banks. By definition, community banks operate within a local community, while the G19 operate on a global basis. For community banks, this geographical localization also means they don't track "national" projections very well, so the overlay is less meaningful. Consider a community bank operating solely in Las Vegas. This bank has seen a slowdown as a result of the credit downturn and one can't deny that home prices in the area are off 50% since the peak reached in 2006. The problem in applying the G19 test to this bank occurs when you try to incorporate a region that is already off so much and then try to push it down another 30% or so. Anything that has fallen 80% in value is going to look horrible, but here again; when large bank stress metrics are applied generically to community banks without perspective, the overlay is not only meaningless, but can also be downright misleading.
Let's consider another view into the same problem as we examine risk and customer type. For the G19, loans are of much larger size because they are originated to much larger national companies (that tend to borrow in higher dollar volume). Stress testing such a loan portfolio based on changes to national GDP or employment will have a much higher correlation than applying the same test to the loan portfolios of community banks. Depending on the region, community bank loss factors can be very different that national averages and the result can either heavily understate or overstate the risk. One example is easily seen in the single family residential sector. While the G19 held large quantities of subprime or Alt-A residential mortgages, community bank lending within this sector was focused on prime-quality loans. No matter the test format, using a starting point that compares a subprime borrower to a prime borrower is about as valuable as comparing apples to oranges - both are round and both are fruit, but the similarities pretty much end there.
Finally, it is important to remember that the "losses" calculated under the G19 stress testing are less likely to occur if assets are held to maturity than if they are sold in the open market. Since community banks originate and then hold loans to maturity, while the G19 usually originate, package and resell loans, this difference is significant. Applying the G19 projected loss factors without incorporating this difference in business model, will tend to overstate risk at community banks.
Please don't misunderstand our point by thinking we are saying all stress testing is bad. Quite the contrary, we feel testing is a good thing as long as it is rationally applied. When testing is not rationally applied, however, improper conclusions can easily be drawn and they may exacerbate the risk that one is trying manage. The key to any testing is not the results of the stress test itself, but rather, what you do with those results in terms of risk management and contingency planning.