Recent regulatory thought on interest rate risk has been divided into two primary areas of concern: Earnings at Risk (EaR) and Economic Value of Equity (EVE). EaR addresses the short term IRR concern, while EVE addresses the long term aspects of IRR. For the rest of this discussion, let us focus on Earnings at Risk.
Recently, we have had several clients report negative exam comments surrounding the lack of RSA / RSL (rate sensitive assets / rate sensitive liabilities) parameters in the banks’ interest rate risk policies. The comments read something like:
The bank’s IRR policy establishes adequate parameters for post-shock changes in ROA but does not establish parameters for the RSA / RSL position or changes to the Net Interest Margin. Management should update the IRR policy to provide operating parameters for RSA / RSL and the Net Interest Margin.
These examiner comments about RSA / RSL expose a lack of understanding about how interest rate risk measurement works and incorrectly imply a regulatory requirement to measure by antiquated standards.
Measuring interest rate risk by a change in ROA is logical, since ROA is a measure of earnings, and earnings is one of the six CAMELS components. The last time I checked, the RSA / RSL ratio was not a CAMELS component. Our Interest Margin Sensitivity Analysis reports the post-shock projected Net Interest Margin, Change in ROA, Projected ROA and Change in Net Income. This seems like plenty of data for the Asset / Liability Management Committee to make informed decisions about risk positions and risk tolerances.
In my IRR discussions, I focus on the impact of a rate change on earnings, as that is what really matters. Yet, examiners want to take us back 20 years by looking at a simplistic ratio that tells us less about a bank’s true IRR exposures. The IMSA’s calculations take into account the impact of rate changes over a full year, while the RSA / RSL ratio is a snapshot calculation at the 365th day that ignores all temporary asset / liability mismatches that come and go between day 1 and day 364.
An example of the weakness of RSA / RSL relative to Change in ROA: Consider two banks (A and B), each with total assets of $100 million. Bank A has $30 million of loans that adjust immediately with a change in NY Prime. Bank B has $30 million of loans that reprice annually, with repricing dates spread evenly across the coming year. The two banks are identical in all other respects.
The RSA / RSL ratio at the one year horizon is identical for the two banks, as each bank’s $30 million in loans will reprice within the first year. The banks’ changes in ROA are very different, however, due to the timing of the repricing. If Bank A’s interest rate risk position as measured by Change in ROA is completely neutral, Bank B would be very liability sensitive due to the slower repricing of its loan portfolio in a rising rate environment.
Bank A happens to be a real APC client with virtually no exposure to rising rates. Its Change in ROA for a +100 basis point shock is -2 basis points, and its RSA / RSL is 0.82. We created Bank B from Bank A’s data, changing only the loan repricing characteristics. Bank B reports a RSA / RSL of 0.82 and a Change in ROA of -12 basis points. With a policy limit of a -25 basis point change for a +/- 100 basis point rate shock, Bank B is halfway between neutral and its policy limit.
Some data from the APC client base: I reviewed the interest rate risk reports for all of our clients, hoping to identify a set of RSA / RSL parameters that would match our standard Change in ROA policy limit of no more than a 25BP decrease in ROA for a 100BP instant and sustained rate shock. My plan was to satisfy our friendly examiners by setting some RSA / RSL policy parameters without unreasonably impairing management flexibility.
From our base of 36 clients, I selected the 30 for which we had a full 12 months of historical IRR data and pulled the most extreme monthly Change in ROA for each bank between June 2008 and May 2009. Each of these 30 observations was within the 25BP decrease in ROA policy limit (the maximum exposure in the group was 21BP), so it seemed logical to me that the corresponding RSA / RSL ranges would suggest appropriate limits for a new policy parameter designed to satisfy our examiners.
The result: 30 Change in ROA measurements provided RSA / RSL parameters ranging from 0.44 to 1.71. These ratios are much wider than the ranges most regulators seem to accept as reasonable (0.75 to 1.25). So what shall we do when faced with regulatory expectations to measure risk with an undersized and one-dimensional yardstick? It seems we have several choices:
1. Set “normal” RSA / RSL parameters: Adopt RSA / RSL parameters the examiners will like, which would put 24 of our 30 test banks in an “outside policy parameters” position at least once within the last year. In this scenario, these banks must begin to reduce interest rate risk to comply with an unreasonably strict policy or accept regular policy exceptions.
2. Set reasonable RSA / RSL parameters: We can suggest a range that works based on your bank’s historical behavior, but the parameters will probably be criticized during the next exam. At that point, we must begin the examiner education process, which is likely to continue at every exam.
3. Continue without RSA / RSL parameters: This will require that you continue to educate examiners about the reasons you do not set parameters on RSA / RSL, at least until this dinosaur (the ratio, not the examiner) finally tumbles into the tar pit.
We will be glad to help you implement whichever option you select. My goal in climbing onto this soapbox is to give you enough information to discuss the topic with your examiners and respond to their oral comments before they become written criticisms. If you sense an upcoming conversation with your examiners and feel the need for reinforcements, we’re only a phone call away.
