Manage your bank, not your model

Take our Interest Rate Risk Policy survey

Please help us gain a better understanding of your ALCO policy limits and how you use our data. These four questions will only take a minute: Question 1 Question 2 Question 3 Question 4 Thank you for your time! Learn more about setting an Interest Rate Risk Policy

Don't be lulled into thinking that dramatic shifts only happen on the short end of the curve

What kind of stress-test do you run when you model your bank's interest rate risk?  a rate shock? a rate ramp? a parallel shift? a non-parallel shift? a twist?  Which one is the best one to use?

There is a point in all of my presentations on interest rate risk when the audience really seems to chime-in.  During presentations before 2001 I would say that we use an instantaneous +/-200bp parallel shock at which point audience members would cry-out, "Are you kidding?  That's way too much!  We have a much better understanding of our economy and the interest rate environment to expect such dramatic changes."  After 2001 and the dramatic drop in Fed Funds (down more than 400bp in a year) the response was, "Are you kidding?  A stress-test of +/-200bp is way too little..."

The debate really gets going again when the Fed makes dramatic moves like the most recent 50bp and 75bp drops...minus 125bp in less than a month.  However the argument usually takes a different direction.  The pundits usually say that a stress-test that shocks short-term rates is okay, but you shouldn't use a parallel shift across all points on the curve because longer-term rates don't move as dramatically.

Think again.  Here's a look a some pretty big shifts in the 10-year treasury and the long-term conventional mortgage rate in relatively short periods of time.

From October to November 1981 the monthly average 10-year treasury moved down 176bp, from 15.15% to 13.39%.  From June to July 2003 the monthly average 10-year treasury moved up 64bp, from 3.33% to 3.98%.

From August to September 1982 the monthly average conventional mortgage rate moved down 169bp, from 16.29% to 14.61%.  From March to April 1994 the monthly average conventional mortgage rate moved up 98bp, from 7.59% to 8.57%.

Projected Fed Funds forecast (looks like my guess is as good as anybody's)

The FHLB of Seattle publishes a regular newsletter called What Counts.  Their 4th quarter issue features "Select Forecasts of Key Economic Statistics".  It very interesting stuff.  What caught my eye is the estimated Fed Funds Rate for 6/30/2008 and 12/31/2008.  Now I'm no economics expert, I've never claimed to be.  It looks like some reasonably well know firms on this list.  What a difference!  The difference between the high and low projection for 6/30/2008 is 125bp.  The difference between the high and low projection for 12/31/2008 is 175bp!

Printable version of data:

http://www.olsonresearch.com/BlogResources/FHLBSeattleProjFedFunds2007Q4.pdf

Deposit pricing not impacted as much as banker's would like

In case you didn't see it there's an article in today's American Banker (subscription required) about how the Fed's rate cuts aren't helping margin because of sticky deposit pricing.  I've posted about this over the past several months.  Here are the related blog entries:

The rising cost of funding is the primary reason for banks' lower earnings - Nov 29, 2007

Now is a good time to adjust your core deposit modeling assumptions - Sept 21, 2007

Is there really margin relief on the horizon? - July 11, 2007

"Is this the end of less expensive deposit funding?" - May 15, 2007

Settling on an interest rate risk policy limit when you don't have a clue

With each of our clients, at one point or another, we inevitably start talking about establishing or updating ALCO policy limits for interest rate risk (IRR).  These conversations are almost always examiner motivated because a banker is rarely self-motivated to set an IRR policy limit...they only do it because they're required to.  In fact the Joint Policy Statement on IRR outlines this requirement quite clearly:

a bank's board of directors is responsible for establishing and guiding the bank’s tolerance for interest rate risk, including approving relevant risk limits...

The trouble is that very few community bank board members are familiar with interest rate risk management & measurement concepts.  They are much more comfortable talking about earnings performance targets.  They can even have reasonably meaningful discussions regarding credit quality.  That's not usually the case with interest rate risk.  It's this lack of familiarity with IRR that makes it difficult to establish and approve meaningful IRR limits.

The typical community bank board doesn't have more than two or three members that have a financial background.  Most of the time you have a doctor, a farmer, an attorney, etc.  To these folks IRR is a foreign language.  But there are some questions that board members can ask that will help them establish reasonable limits without requiring an in-depth understanding of IRR measurement.

When considering a new IRR policy limit, for example a policy limit -15% net interest earnings at risk, consider these questions:

1. What is our current (and past) measurement?
2. What are peer measurements?
3. What the rules-of-thumb?
4. What do example policies say?
5. What are the regulatory benchmarks?
6. What is our comfort level?

In future posts I'll consider each of these questions as it relates to the suggested sample limit of -15% for net interest earnings at risk.

When "asset sensitive" doesn't benefit from rising rates

Client question:

I'm looking at our gap report and it shows our cumulative gap to be 116% which is asset sensitive, yet when I look at the rates up simulation on the income shock report it shows my net interest margin declining when rates rise.  Is there something wrong with the model?

Answer:

No, you've just observed the biggest weakness of the gap report.  It doesn't capture option risk.  There are several accounts where option risk is likely to show up on a bank's balance sheet today.

The first place is in the securities portfolio.  Many US Agency bonds (and Muni's too) have call options.  When rates fall they are likely to call creating an overall lower return for the portfolio.  When rates rise the portfolio acts more like a fixed-rate bond portfolio and there is less positive change income.  Many people try and solve this problem by adjusting the gap report and placing the bond in a gap bucket which reflects its call date and not its maturity date.  The problem is that then the gap report doesn't reflect what might happen if rates rise.  Hence the problem with the gap report, it doesn't capture option risk.

The second place you commonly find option risk is in the bank's loan portfolio...prepayment risk.  Prepayments cause a change in interest income that is similar to call options.  When rates fall prepayments tend to increase making the bank look more "asset sensitive".  When rates rise, prepayments tend to slow down making the bank look more "liability sensitive".  Again you can't fix the gap report because the cash flow structure of the portfolio changes as rates change.

The third place you commonly find option risk is in the bank's core deposit portfolio.  The rates on these deposits are usually administered by the bank.  The bank may change the way these deposits reprice when market rates rise, and they may change the rate differently if market rates fall.  Once again, you can't "fix" the gap report to reflect this difference, because the timing and amount of repricing changes depending on market rate changes. (This may also apply to the bank's CD rates).

Finally, the last place you commonly see option risk in the bank's FHLB advance portfolio.  The FHLB's offer a wide range of "convertible" advances these days.  When rates fall, the bank is stuck with a longer-term fixed rate instrument.  When rates rise, the FHLB decides to convert the debt to a variable rate.  Exactly the opposite behavior of the callable security.  The cash flow timing changes as rates change, and that (again) can't be captured on a gap report.

Here is the sample income shock report (click image to enlarge):

Now is a good time to adjust your core deposit modeling assumptions

I think a minus 50bp adjustment to Fed Funds caught many of us a bit off-guard.  Especially if you consider the headlines less than 6 months ago talking about inflation worries and the potential need for the Fed to tighten again.  It just goes to show you that the crystal ball isn't always right (and no, we haven't really improved upon it either).  Alan Greenspan himself echoed this sentiment in one of his many interviews promoting his new book.  On last Wednesday's The Daily Show Greenspan said, "forecasting 50 years ago was a good or as bad as it is today".

Which brings me to the real point of this post.  I wish I could have listened in on every one of our community bank client's pricing meetings this week.  What kind of discussions did you have?  Did you decide to move your NOW, Savings, and/or Money Market rates?  If so, by how much?  The full 50 basis points down?  Or do you feel that local market pressures are going to force you to keep rates up?

How much (or how little) you move these core rates defines what we call your "beta factor" for core deposits.  It's a measure of your bank's sensitivity and it's a critical assumption when modeling your Net Interest Earnings at Risk.

Computing your bank's beta factor for a given account is easy.  Take the amount you lowered rates this week and divide it by the change in Fed Funds (-50bp).  For example, if you lowered your core savings account rates by an average of -25bp then your beta factor for Savings is 25 divided 50 or 50%.

This week take the time to document your change in pricing, and provide that data to us when we update your A/L BENCHMARKS model for September 30.

Relationship between earnings-at-risk and EVE-at-risk

What's the most misunderstood concept in interest rate risk measurement?  There is probably a six or seven-way tie for first place on that list.  Certainly one concept that's tough to grasp is the relationship between net interest earnings-at-risk (EAR) and EVE-at-risk.  Specifically, why do they often seem to contradict one another?  How can the bank have earnings exposure to rates down, but EVE exposure to rising rates (or vice versa)?

Often the two measures do seem to contradict one another.  This is the biggest reason many bankers dismiss the EVE-at-risk measurement as a useless measurement.  However things are not always as they seem.  In a previous post I gave a pretty simple example of why you need a longer term measurement of IRR.

In his book Interest Rate Risk for Banks, the author Leonard Matz, addresses both the conceptual and practical relationships between EVE and EAR.  Understanding both gives a much clearer understanding of why the differences occur.

The conceptual relationship between EVE and EAR
As with any measurement or modeling approach consistency is key.  Without it the results are much less reliable (or believable).

Conceptually, it is easy to relate the rate sensitivity of net income to the rate sensitivity of economic value.  For any single interest rate scenario, the EVE should approximately equal the present value of the annual net incomes for all future years.  Suppose we use both an income simulation model and an EVE simulation model.  Both models should incorporate the same assumptions.  Suppose we use the income model to project net income for all future years and we then take the present value of every one of those projected income amounts.  If we use the same discount rates to calculate the present value that we used in the economic model, the present value of the future net incomes should be close to the calculated EVE...they should be close, but not identical.  Having a better understanding of why these two measures differ will help you grasp what each measure is telling you about risk. 

There are four reasons why the present value of the income flows for all future periods won't necessarily equal the EVE, even when the same discount rates are used.

1) Banks have more assets than liabilities.  Because a bank has equity the present values of asset cash flows will exceed the present values of the liability cash flows.

2) Most EVE calculations include only balance sheet cash flows.  There may be, and usually is, some rate sensitivity to non-interest income and expense cash flows.

3) Earnings are calculated using accrual accounting, while EVE uses only cash transactions.  Take for example a zero coupon bond.  Net income recognizes monthly accretion of the discount.  EVE on the other hand captures the income as the present value of a cash-flow at final maturity.

4) EVE calculations do not include cash flows resulting from future business.  If new business is added at the prevailing market rates of interest this is not a problem.  However, for floating-rate instruments with caps & floors, and more importantly for core deposits, this discrepancy will produce irreconcilable differences between EVE and EAR.

Pages 5.32-5.34, Interest Rate Risk Management for Banks, Leonard Matz, 2007

The practical relationship between EVE and EAR
So does EVE relate to earnings-at-risk in practice?  To quote the author again, the answer is "not even close."  The biggest reason why is that we don't measure earning-at-risk by adding up the present values of all future earnings flows.  In the case of A/L BENCHMARKS we only look at a one-year time frame (you have the same problem even if you look at a two or three year time frame). 

Just because the present value of all future earnings might equal the economic value of equity, that doesn't mean that the rate risk sensitivity of the current periods earnings relates in any way to the sensitivity of EVE.  EVE reflects the sensitivity of all periods, while EAR is measured for just a single defined period.  If the risk sensitivity for any one year is similar to the total risk exposure for all years it will be nothing more than a coincidence.  In fact, even the direction of rate risk exposure - whether we suffer from rising rates and benefit from falling rates or vice versa - may be different for the next year than it is for all future years. (emphasis was added by the author, not by me!)

Measuring the EVE-at-risk of financial institutions is not a "new" practice.  The OTS was measuring EVE years before the Joint Policy Statement on Interest Rate Risk was issued in 1996.  Starting in the late 80's the OTS was measuring the EVE-at-risk for all thrifts in the country.  Market interest rate movements in 1993, and in 1994 presented the OTS with an interesting opportunity. 

For both years (in 1993 rates fell significantly, and in 1994 rates rose sharply) the OTS attempted to find a correlation between their calculated EVE-at-risk and the thrift's change in earnings.  What they found were some pretty weak correlations.

The OTS study found that the economic model was not even very good at predicting the direction of the earnings rate risk exposure.  When prevailing rates rose in 1994, 59% of thrifts in the OTS study experienced a decline in their net interest margins.  For that group of thrifts, in early 1994, the OTS's model indicated that 87% would experience a decline in EVE if rates rose.  In other words the model was directionally correct in predicting the earnings risk for 87% of those thrifts.

On the other hand, for thrifts whose net interest margin actually rose in 1994, the model only identified 20% of them as being subject to an increase in EVE if prevailing rates rose.  In other words, the model was directionally incorrect at identifying the earnings risk for 80% of those institutions.

In short, it is very important to remember that your rate risk exposure for the next month, quarter, year or two years may not even be in the same direction as the entire rate risk exposure in all of your balance sheet positions.  EVE sensitivity tells us the amount of our rate risk exposure, but it tells us absolutely nothing about the timing of the risk.

A change in prevailing rates causes an immediate and very real change in the value of our assets and liabilities.  However, that same change does not impact the earnings of the bank immediately.  Instead, earnings are affected at the time the instruments reprice.  The earnings impact may occur on the day rates change, or many years in the future.

Making the complex easy to understand

Most discussions of interest rate risk (IRR) jump immediately into calculus equations I haven't seen since my undergrad days more than 20-years ago.  It's usually hard to find a simple line or two of explanation that doesn't require me to go back to my college text for reference.  However, there is one book I do keep close at hand because it addresses complex interest rate risk issues using (mostly) plain language.  The book is Interest Rate Risk Management for Banks by Leonard M. Matz.  Although I've never spoken with him personally, I've heard Mr. Matz speak at a conference or two, and I'm always impressed by his simpler approach to communicating such a complex subject.

It would be easy for the book to dive right into the first derivative blah, blah, blah that you so often find in books on IRR.  What's great about this book is that the author makes a genuine attempt to talk about complex IRR issues as if he were speaking with senior bank management.  I am not suggesting that senior bank management is thick-headed when it comes to IRR, but like anything else, if you're not immersed in the subject every day things are a little less clear.

I would make this book suggested reading for all of our clients if it weren't for the steep price.  Too bad the book isn't available from amazon.

Brief overview of Interest Rate Risk

When I'm asked, "What sort of work do you do?"  I'm never sure quite how to answer.  When I started working in this business my answer to this question was, "I help community banks measure and monitor their interest rate risk..."   I learned pretty quickly that that was a conversation stopper.  Nobody can think of a follow-up question to that.  Because most people don't even know what interest rate risk is.

We encounter a similar problem when supporting our bank clients because we're often helping the bank CFO communicate IRR measurements to board and other senior management.  The CFOs understand it, but they're frustrated when trying to educate others.

In our A/L BENCHMARKS Board Report we offer what I consider to be a good brief overview of IRR.  Those readers who understand IRR well are sure to find holes in the following text.  But remember it's meant to introduce IRR to folks who don't "get it".

Except from our A/L BENCHMARKS Board Report:

Interest rate risk (IRR) is the risk to earnings or capital arising from movements in interest rates. Practically, IRR can be viewed from both a short-term and long-term perspective. To examine short-term IRR we look at earnings at risk. Conversely, we use equity at risk and duration to measure long-term IRR.

Earnings-at-risk: short-term IRR
By most definitions, accounting or otherwise, when we communicate something as short-term, we usually refer to a time frame of one year or less. When measuring IRR from an earnings perspective, this same concept applies. Short-term interest rate risk is measured by initially establishing a one year earnings forecast which may include a dynamic market rate forecast, earnings growth, and balance mix & volume changes.

Since IRR is a measure of possible loss caused by interest rate changes, the model then introduces two instantaneous, parallel "shocks" to the base set of rates and then re-computes the expected earnings. Common practice is to use +/-200bp movements. The earnings at risk is the largest negative change between the base forecast and one of the "shock" scenarios. The measure is usually stated as a percentage change from the base income.

There are two significant characteristics of the earnings at risk measurement the bank should review. First, what rate shock, up or down, produces the worst case change? Is the bank exposed to rising or falling rates? Second, what is the amount of projected change or magnitude of risk? How much exposure is there?

Economic Value of Equity (EVE) at risk
As a means for evaluating long-term IRR, an economic perspective is necessary. This approach focuses on the value of the bank in today's interest rate environment and that value's sensitivity to changes in interest rates. This concept is known as Economic Value of Equity (or EVE) at Risk. It requires a complete present value balance sheet to be constructed. This is done by scheduling the cash flows of all assets and liabilities and applying a set of discount rates to develop the present values. The economic value of equity (EVE) is the difference between the present value of assets and liabilities. (Equity = Assets - Liabilities).

Similar to earnings at risk, two interest rate shocks are applied to the base set of rates and all present values are re-computed. EVE at risk is the largest negative change in value between the base and one of the shock scenarios. This is usually stated as a percentage change from the base EVE.

We've found that this introduction, along with a simple graph or two, goes a long way toward educating those who are less fluent in "IRR-speak".