In case you missed it – there’s a new FDIC Advisory on Interest Rate Risk Management. We’ve had a high volume of phone calls from clients asking us if we’ve seen the “new guidance”, and what can we do to help them address it.
Actually this really isn’t “new” at all. There are a few new points raised that haven’t been addressed or mentioned before, but most of the content comes straight out of the 1996 Joint Policy Statement on Interest Rate Risk, and the OCC Bulletin 2000-16 on Model Validation. Beyond that there isn’t really much new ground.
Here are the highlights that I think are important to our A/L BENCHMARKS customers and to community banks in general, aka “Main Street” banks. You’ll have to read elsewhere for an opinion on what this means for “Wall Street” banks (the big guys).
1) Use an IRR measurement tool that captures the types of risks you have on your balance sheet.
There are essentially four types of interest rate risk: maturity-repricing risk, yield-curve risk, basis risk, and option risk. Through the variety of data files, inputs, and assumptions we use in A/L BENCHMARKS we are capable of modeling all these risks, regardless of the portfolio in which they are found.
2) The Board has the ultimate responsibility for the risk undertaken.
This is why, in addition to providing the in-depth analysis and reports, we provide the Board Report showing the magnitude and direction of risk of the bank quarter-to-quarter and compared to banks of similar size.
3) Measure IRR from a short-term and long-term perspective.
We measure both earnings-at-risk (short-term), and equity-at-risk (long-term). Further reading:
4) Be sure you understand the underlying assumptions and analytics used by the model.
I hope you don’t pay for our service, provide us with little input, and then plop the reports in the bottom drawer of your desk. They make explicit warnings here about the use of third-party models for the reason just stated. First there is a considerable amount of insight into the methodologies that is documented right in the reports. Additionally, the managerial assumptions that you provide via the Service Kit can be found in the back of the Executive Report and should be reviewed quarterly. Finally, you can also contact us to review your assumptions and/or even schedule regular report reviews with your ALCO or Board.
5) Earnings at risk should be measured using a 1-year, 2-year, 5-year, and/or a 7-year time frame.
By default we run the earnings simulation using a 1-year time frame. I disagree with the advisory’s assertion that “…IRR exposures are best projected over at least a two-year period. Using a two-year time frame will better capture the true [risks]”. For most bank’s an earnings forecast of 1-year is at best just thoughtful guesswork. A 2-year (or longer) earnings projection often enters the realm of fantasy. Ridiculous as these longer projections sound – we can run these in A/L BENCHMARKS if needed. (Note: Apparently they also disagree since just a few paragraphs later the Advisory admits that earnings simulations have “limitations” in quantifying IRR exposure, see point #7 below).
6) Earnings simulations can either be “Static” or “Dynamic”
The base forecast for the earnings simulation can either be a flat-balance sheet forecast (static), or it can incorporate growth and new business (dynamic). We can model your earnings simulation whichever way you prefer. Dynamic growth can be entered via the Service Kit under “Balance Sheet Projection.”
7) Economic value-based models should be used to broaden the assessment of IRR exposure
Essentially they say that earnings simulations, because they capture only a specific time-frame (1-year, 2-year, etc.), may miss certain risks that exist on the balance sheet. To address this you should look at economic value of equity (EVE) at risk which focuses on longer-term time horizons and captures all expected future cash-flows. Further reading:
- Earnings at risk is only a short-term view of interest rate risk
- Relationship between earnings-at-risk and EVE-at-risk
A/L BENCHMARKS includes an EVE at risk analysis.
8) Just running a +/-200 basis point stress-test is not sufficient
Many risks, option-risks in particular, don’t show up until you’ve significantly changed the market rate environment. It can also be enlightening to see what happens between a zero and 200bp shift (i.e. +-100bp). A/L BENCHMARKS regularly reports 100, 200, and 300 basis point shocks. The Advisory suggests (given today’s rate environment) that +400bp may even be prudent and we can easily run this for you. They also suggest rate ramps, and curve twists as other possible alternatives. For most banks these additional scenarios only provide additional interesting and anecdotal information. I’ve seldom seen a risk exposure appear on a rate ramp analysis that we didn’t already see or know about in the shock analysis. Nevertheless, if you need to see a rate ramp stress-test we can run that for you.
9) The regulators recognize that a 100, 200, and 300 basis point shock is probably sufficient for most banks.
The actual text is this, “The regulators recognize that not all financial institutions will require the full range of the scenarios discussed above. Non-complex institutions (e.g., institutions with limited embedded options or structured products on their balance sheet) may be able to justify running fewer or less intricate scenarios, depending on their IRR profile. However, interest rate shocks of sufficient magnitude should be run, regardless of the institution’s size or complexity”.
Without sounding too snarky, I’ll believe it when I see it. Most field examiners I’ve worked with believe that more analysis is better. If you can run seven shock stress-test scenarios (base, +/100,+/-200, and +/-300) why not run seven more with a rate-ramp, a twist, etc? It’s bound to turn up something interesting right?…Well, no. There’s the law of diminishing returns at work here. We can spend a lot of time and money to run these additional scenarios, but we’re unlikely to uncover some hidden exposure when we run that extra fourteenth or fifteenth interest rate risk stress-test scenario.
Let’s also not forget the time involved to run such stress tests ultimately takes the banker away from the core business of running the bank (or running other stress-tests to measure other risks like liquidity and credit quality).
10) You should pay attention to key assumptions like prepayment speeds and core-deposit sensitivity
Absolutely. These are the “big two” assumptions that drive the bank’s overall sensitivity. This is true for core-deposit behavior especially. Updated information for both of these inputs should be provided to us each quarter so that we more accurately reflect your IRR profile. Further reading:
11) Back test the model and learn how to use the model better
Your forecasts are always going to be off…there’s no getting around that. You can however, learn from your “mistakes”. For example, if you keep including a projection of 5% growth in DDA’s, but in reality end up funding the bank by growing your brokered CD portfolio, you might want to adjust your projections to reflect the actual behavior. A back test of a forecast isn’t really a “right” or “wrong” test (I can easily tell you that most likely your forecasts will be “wrong”, as your crystal ball just isn’t that good). A back test highlights things that we could be forecasting better (so we’ll be “less” wrong in the future).
12) Have someone independently review your modeling process
This is an essential part of any good modeling process. Although like the back test, this is often misinterpreted as some sort of checklist item, “Has the model been independently reviewed…it has?…check”. That’s the wrong way to think about this. An independent review should be an ongoing process with pieces and parts of it being done every quarter. Further reading here:
That’s about it. It is interesting to see the list of references at the end of the document. Almost all of them reference material that was produced prior to 2001. Again this just shows that there’s really nothing new here, they are indeed just reiterating what’s already been communicated.
UPDATE: The regulators issued this FAQ as a follow-up.