Manage your bank, not your model

Overrated or maybe not

The Bank Safety and Soundness Advisor asked its readers what problem issues are overrated for community banking in 2012.  They listed the top five including this one that appears to be feedback from an examiner:

Regulatory exam inconsistencies. “While this situation does exist, it’s a cop-out challenge,” another reader says. “The findings that regulators identify are real and we don't hear anyone challenging these findings. What we do hear is that [the regulators] ‘did not cite us for this last time’ or ‘they did not cite the bank down the street for doing the same thing.’ So banks need to quit whining, run their businesses effectively, step up to the plate and make the changes and adjustments needed (including cleaning house of the deadwood executives who are unable to lead change) and increase focus on compliance and risk management.”

Ouch!

It sure sounds like this examiner (“what we do hear…”) is talking about large banks and not small community banks.  An overwhelming number of small community banks do run their businesses effectively (oh and BTW, let’s not forget that they are businesses).  As far as interest rate risk measurement is concerned I have to take issue with, “findings that regulators identify are real…”  Many examples of real findings and adjustments identified by regulators hardly seem valuable.  Here are some examples of real findings by regulators for several of our community bank clients over the past year.  These were communicated in their written exam report:

• “The bank should also run interest rate risk stress-tests covering longer time frames (possibly three or five years).”
• “Interest rate risk stress tests should cover up to and including +/-500bp shocks.”  I’ve also seen +/-600bp suggested.
• “The bank does not have policy limits set for Net Interest Earnings at Risk for a +400bp shock.”
• “Per the guidelines on Model Validation, the bank should have a 3rd party review the findings of the independent audit…”  That’s right - they want a 3rd party review of the 3rd party review.

In my opinion while these findings are “real” they are not realistic.  Addressing these issues may provide some analytical or academic value, but it will not provide any practical value.  It amounts to busy-work.  The examiners will check it off their list and the bankers will ignore the results.  What a waste of time and energy.

Release the data hounds

It seems like every one of our clients is calling us back to run yet another IRR stress-test “because my examiners are asking for it.”  We already supply the standard +/- 100, 200, 300 & 400bp one-year shock-tests.  We’re also providing the two-year shock-test. That’s eight (or sixteen!) stress-tests, but the regulators are consistently asking for more.  Most field examiners are reading the IRR Advisory from last year quite literally:

Depending on an institution’s IRR profile, stress scenarios should include but not be limited to:

• Instantaneous and significant changes in the level of interest rates (instantaneous rate shocks);
• Substantial changes in rates over time (prolonged rate shocks);
• Changes in the relationships between key market rates (i.e., basis risk); and
• Changes in the slope and the shape of the yield curve (i.e., yield curve risk).

IRR Advisory 2010

It’s the last one listed that seems to be the flavor of the day for field examiners – changes in the shape of the curve: steeper, flatter, or a twist.  After running these additional tests for several different clients I’ve concluded that most of the time there is very little new information generated as a result.  There’s lots of new data, just very little new information.  Leaving me wondering, not for the first time, what is the point?

Here’s an example of one recent analysis.  The standard +100, 200, 300 and 400bp shock results are as follows:

	+100 bp	+ 200 bp	+300 bp	+400 bp
Net Int Earn at Risk (short-term)	-3.30%	-5.91%	-8.28%	-11.21%
EVE at Risk (long-term)	-7.66%	-10.88%	-18.01%	-24.11%

So essentially the model is telling us that this bank will feel pressure in the short-term and the long-term from rising rates.  In addition the client wanted us to also model the impact of a twist.  The projected twist is shown in the graph (click on it to see a larger version).  Short-term rates, effectively all points 2-years and shorter will move up about +300bp.  Rates between the 2-year and 10-year will move-up progressively less, and the rates beyond the 10-year point will be modeled to fall.  Most of the twist looks like a +300bp shock with the longer points falling somewhere on the spectrum of the +100bp or +200bp lines.  Only the most extreme points are falling.

The majority of this bank’s balance sheet looks like almost any other community bank: 70% in Loans, 20% in Securities, 40% funded by Core Deposits, 35% funded by CDs, and around 10% Equity.  Two-thirds to three-quarters of their loan portfolio is variable rate either tied to Prime or the 1-Year Treasury.  Most of their deposits are tied to Fed Funds, or points between 3-months and 2-years.    I expected the earnings stress-test for the Twist to look a lot like the +300bp shock.  And indeed it did, the Net Interest Earnings at Risk for the Twist was –9.82%.  The risk was slightly higher because the bank currently owns and is projected to buy a number of longer-term municipal bonds.  Their term structure put them around the 15 to 20-year rate which didn’t increase as much (again see the graph.)

The EVE-at-risk for the Twist was –9.65%.  The majority of the bank’s balance sheet value was impacted by the +300bp change in rates because it has a shorter-term.  The longer-term parts of the balance sheet didn’t loose as much value in the Twist because the longer points on the curve only changed around +100bp.  I would expect to see an EVE-at-risk somewhere in the neighborhood –10.88% which is the EVE-at-risk as measured by the standard +200bp shock.

It would be hard to argue that I didn’t learn anything new by this analysis - I did.  The behavior of the bank’s longer-term investment portfolio in the Twist is dynamically different compared to the short-term loans and deposits.  But in the end it’s only academically interesting.  I still find out that the bank is at risk to rising rates in general.  I’m not convinced it was worth the time an energy it took to explain the nuances of the Twist analysis to the senior management and board.  The overall conclusions about the bank’s interest rate risk are still the same.

Just because we can run the extra analysis doesn’t mean it’s going to be useful.  By highlighting the additional stress-tests as something that banks “should include” the regulators are implying (intentionally or not) that there is real insight to be gained.  Otherwise why require it?  The problem is that most bankers, let alone their board members, don’t have a deep understanding of IRR measurement to begin with.  Piling on more analysis isn’t likely to help them understand it any better, it’s more likely to tune them out even faster.

Management consultant and author Seth Godin had a great post earlier this year that seems particularly relevant here:

Important/Measured

Is something important because you measure it, or is it measured because it's important?

Does our new ability to see things with…data make the previously overlooked now visible, or are we giving weight to things merely because we’ve measured them?
Seth Godin’s Blog | March 5, 2011

When will enough be enough?  Are the regulators comfortable with the information overload they are promoting?

Steeper, Flatter, Twist!

After reviewing IRR shock-test results for a few quarters it usually doesn’t take a person very long to start questioning the “reality” of the shock-test.  The common observation is usually, “all rates never move up or down in parallel…that’s not very realistic!”  This is inevitably followed-up with a redesign of the IRR stress-test that moves the yield curve in more “realistic” ways.

Although I’m quite skeptical of our ability to realistically and accurately predict future yield curve movements, there are changes other than parallel shifts that may be academically interesting to test.  After all rates never really do move up or down in parallel.  We’re more likely (historically speaking) to see short-term rates move down or up with little or no movement in long-term rates.  These types of shifts are commonly referred to as “steeper” or “flatter” shifts.  Each has its own unique characteristics to consider:

a) Steeper - Typically this is accomplished by dropping short-term rates.  All points on the curve that are 1-year or less drop noticeably.  The drop is less pronounced the further out you go on the curve until, finally, you don’t model any movement in long-term rates at all (perhaps somewhere between the 5-year and 30-year point.)  The resulting steeper curve can have major impacts on a community bank’s balance sheet depending on the magnitude of the change in short-term rates.  So much of a typical community bank’s balance sheet is tied to short-term rates: Prime-based loans, core checking, and CDs.  A less common approach to modeling a steeper curve is to raise long-term rates and keep short-term rates stable.  There are certainly examples of this behavior in history, it’s just not as common.

b) Flatter – Typically this is accomplished by raising short-term rates.  All points on the curve that are 1-year or less increase noticeably.  The increase is less pronounced the further out you go on the curve until, finally, you don’t model any movement in long-term rates at all (perhaps somewhere between the 5-year and 30-year point.)  Again, because so much of a community bank’s balance sheet is tied to short-term rates, a flatter curve can have a major impact.  A less common approach to modeling a flatter curve is to lower long-term rates and keep short-term rates stable.  This has happened in history, but it is not as common.  Short-term rates are usually much more volatile than long-term rates.

c) Twist – This test essentially moves both ends of the curve in opposite directions.  This test is often referred to as the “inverted” test because typically the resulting yield curve shape is inverted (as opposed to “normally” shaped.)  While much of a bank’s balance sheet is tied to short-term rates, there may be portfolios that are tied to longer-term rates like investments, mortgages, and borrowings.  Income and expenses may change at different rates depending on the timing.  A twist can highlight that dynamic.

(This post is part of a series which provides a basic overview and discussion of interest rate risk stress-testing.)

Stress Testing - Not Just for Too-Big-To-Fail < BAI-Banking Strategies

If the banking industry is to embark on a new path of less earnings volatility and ensure its collective systemic safety, banks of all sizes must embrace and implement robust stress testing.

via www.bai.org

Interest Rates: 1831-2011

Here’s a fascinating chart from Jim Bianco, looking at interest rates back nearly two centuries (hat tip: The Big Picture).

(click to view larger version)

Source: Bianco Research LLC

FDIC Clarifies Stress Testing

With Dodd Frank, stress testing became mandatory for banks as small as $10 billion in assets. Will across-the-board stress testing requirements creep down into the community banking ranks? Are examiners going to get aggressive about stress testing for community banks that haven’t already passed the CRE concentration thresholds from the 2006 CRE guidance? No, insists James Watkins, FDIC Deputy Director, Division of Risk Management Supervision. Regulators and examiners will be busy adapting Dodd-Frank stress testing requirements to larger banks, but smaller banks shouldn’t anticipate that they’ll be subject to the same expectations.

via www.banksoundness.com

Even if LIBOR Is a Lie, Who Gets Hurt? - Bank Think

...Among other number-crunching exercises, van Deventer compared Federal Reserve data on three-month Eurodollar rates during the period covered by the Schwab suite to banks’ self-reported costs in the LIBOR survey of that maturity in the same timeframe. LIBOR is based on banks’ quotes of what they hypothetically would have to pay for interbank loans. The Eurodollar rates are presumed to show what banks actually have paid for money, on average. (The Fed gets its data from ICAP, an active broker of Eurodollars). Theoretically, these two figures should track each other closely.

Yet beginning in 2007, Van Deventer finds major discrepancies. The Fed’s figure far exceeded several banks’ claims of what their borrowing costs were. Again, one could surmise that the banks were simply trying to stem a panic by denying how bad things were. But if that were the only motivation, the difference should have narrowed sometime after the Tarp bailouts of October 2008, which calmed markets. Rather, banks’ self-reported, hypothetical borrowing costs continued to come in far below the Fed’s Eurodollar market indications well into 2009. 

via www.americanbanker.com

Correlation or Causation? - Businessweek

Correlation may not imply causation, but it sure can help us insinuate it.

via www.businessweek.com

Regulatory Push for Longer Forecasts Makes Little Sense, Experts Say

Anything beyond two to three years for most financial institutions is probably not going to be very useful, says Emily Hollis, partner, ALM First Financial, Dallas. “Most [small institution] balance sheets are pretty short...So when it comes to income simulation, anything past two years is probably a waste. You can change the balance sheet. If you project five years out and shock rates up over that period, and you have a static balance sheet, it is going to tell you nothing..."

via www.banksoundness.com

“Give me five”

In the summer of 2010 I asked the rhetorical question, “If a two-year forecast is so good, then why not a three-year or five-year forecast?”  Apparently that question has been answered by both examiners and other A/L vendors & consultants.  During the months of October and November on more than one occasion I reviewed A/L report sets from other vendors that showed a five-year earnings projection and stress-test.  I also heard from examiners (FDIC, OCC, and two specific states) that they want their banks to create a five year earnings forecast and stress-test it (using a variety of different rate scenarios).  I’m beginning to feel like a real contrarian here when I say that such stress-tests provide very little extra value, especially given the amount of extra time and resources needed to prepare such a forecast.  The additional benefits just aren’t worth it (i.e. there is very little new information gained.)

I’ve already commented that I think we’re stretching the limit trying to project two years into the future.  But now five years?  Really?  Five years?  Well, you say, one person may not be able to accurately predict the future, but we can look to the collective wisdom of the crowds.  Now in the internet age we have access to vast amounts of data.  Through hundreds of new media channels (online video, podcasts, blogs, twitter) we can listen to and reflect on the advice of countless numbers of experts.  Surely with all that we can make much better predictions…right?  Maybe not.  Take a look at what the experts thought of Peter Schiff’s vision of the future:

Peter Schiff Was Right 2006 - 2007 (2nd Edition) (embedded video below)

Please don’t misinterpret my intent here.  Peter Schiff may be a genius, or maybe not.  I really don’t know much about him and what he’s been up to since this time.  What I want to draw your attention to is that the so-called “experts” really slammed this guy.  They were even laughing at him (on a respected and “unbiased” newscast no less.)  All of the discussions highlighted in this video happened FIVE YEARS ago.  Time may fly, but isn’t it funny after watching a video like this how long ago five years really seems?

Forecasting the financial and economic future is, at best, extremely difficult.  Sometimes guys like Peter Schiff guess correctly.  Most of the time, like all of us, he’ll get it wrong. 

I’m reminded of a passage in my new favorite book, “Future Babble: Why Expert Predictions are Next to Worthless…”:

Economists, in particular, are treated with the reverence the ancient Greeks gave the Oracle of Delphi. But unlike the notoriously vague pronouncements that once issued from Delphi, economists’ predictions are concrete and precise. Their accuracy can be checked. And anyone who does that will quickly conclude that economists make lousy soothsayers: “The record of failure…is virtually unblemished,” wrote IMF economist Prakash Loungani in of many papers demonstrating the near-universal truth that economists’ predictions are least accurate when they are most needed.
Gardner, Dan (2011-01-26). Future Babble: Why Expert Predictions Are Next to Worthless, and You Can Do Better (p. 13). DUTTON ADULT. Kindle Edition.

---

Peter Schiff Was Right 2006 - 2007 (2nd Edition)