Last January the regulators released their Advisory on IRR management. Although it is only nine pages long, it contains a great deal of information and
requirements suggestions for measuring interest rate risk. I’ve already posted my general comments about the advisory here.
There are certain parts of it that I don’t fully agree with, like the suggestion that banks should use a two-year time frame for earnings simulation:
A key aspect of IRR simulation involves the selection of an appropriate time horizon(s) over which to assess IRR exposures. Simulations can be performed over any time horizon and often are used to analyze multiple horizons identifying short-term, intermediate-term, and long-term risk. When using earnings simulation models, IRR exposures are best projected over at least a two-year period. Using a two-year time frame will better capture the true impact of important transactions, tactics, and strategies taken to increase revenues which can be hidden by viewing projected results within shorter time horizons…
I also attended the symposium on IRR last January. During one of the regulator panel discussions a panelist called the two-year time frame a “best practice”. Here’s the Wikipedia link that describes best practice. Perhaps it’s my computer programming background, but I’m more inclined to send you to this link for a more informative explanation of “best practice” - Dilbert 09/03/2008.
One thing I’ve observed over the years of running our A/L model for community banks is that we’re all terrible forecasters. Our crystal balls stink, plain and simple. The balance sheet and income projections clients provide to us almost always turn out to be wrong. And I don’t mean they are off by 5% or so. We typically see variances of 25% or more when compared to actual performance. And, here’s the kicker, these are only one-year projections. I can only imagine how far off their two-year projections would be. Oh yea, and these are projections during stable economic times, not the turbulent times we’re currently experiencing. One-year projections during these times are off by far more than a mere 20% to 25%.
Measuring potential earnings volatility over a two-year time frame just doesn’t make much sense. There are too many unknowns and we have to make too many assumptions about future behaviors.
How long is two-years? Longer than you think. Consider that two years ago many of these events hadn’t happened yet (as of July 9, 2010):
- FDIC closes IndyMac – July 13, 2008
- Lehman Brothers Files for Bankruptcy – September 14, 2008
- Congress Approves $700 Billion Bailout Plan – October 3, 2008
- $50 Billion Allegedly Lost in Madoff Fraud – December 15, 2008
- More here…
The most striking thing on this list is the rescue package that included TARP. TARP! It seems like that’s all we’ve heard about in the banking world. Two years ago the program didn’t exist.
Measuring IRR already requires making assumptions. However, using a two-year time frame requires that you make too many assumptions about future events and behaviors. There’s too great a potential for such assumptions to “cover-up” current risk. The new assumptions also have the potential to create or model new risks. And probably worst of all, they may cause a bank to hedge risks that don’t really exist on the balance sheet yet.
- Update -
I just came across this piece of information that made me laugh. I was reviewing the feedback and comments I received after making one of my presentations. Here’s feedback from a couple of attendees at the FFIEC’s Supervisory Updates & Emerging Issues conference where I presented on the topic of Interest Rate Risk in June 2008 – barely two years ago:
- I didn’t consider this topic an emerging issue.
- Good refresher…but IRR is not really an emerging issue, it would have been better to focus on liquidity or some other risks.
Seventeen months later the FDIC issues its IRR Advisory. I guess it turned out to be an “emerging risk” after all. This is just another example of how far in the future two years really is. It’s very difficult for any of us to project or forecast that far ahead.