A) Instantaneous Parallel shift
Rates change immediately. They change right on the “as of” day on the financial statement. In the case of an earnings simulation this kind of rate change will occur immediately before the first day of the scenario. For a balance sheet simulation the rate change will happen on the balance sheet “as of” date. In addition to changing immediately all rates will change by the same amount – short-term and long-term, hence the term “parallel”. Basically this means the shape of the yield-curve stays the same. It will maintain its relative “steepness” or “flatness”.

B) Instantaneous Non-Parallel shift
In this kind of movement rates will also change immediately, but all rates will not move by the same amount. Short-term rates may rise (or fall) faster than longer-term rates giving the curve a steeper or flatter slope. Conversely longer-term rates may move by more than shorter-term rates. Either way the change is not uniform across the curve and the overall shape will be different in the stressed scenario.

C) Gradual Parallel shift
Here, as the name implies, the change applied to rates takes place over time - usually over one year. Rates don’t change immediately but rather at some sort of measured regular pace. For example a common scenario might model incremental steps of +25bp per quarter over the next year or something similar. Although the change doesn’t happen all at once, all rates across the curve are modeled to change by the same amount therefore the overall shape of the curve will remain the same.

D) Gradual Non-Parallel shift
This is the least uniform and least standardized of the four kinds. The changes in rates are modeled to take place over time and the changes will not be uniform across all parts of the curve. Rates may only change slowly on the short-end and very quickly on the long-end. Many people believe that this is the most “realistic” of the four kinds. They argue that never in the past have all rates changed instantaneously overnight. And seldom have all rates changed by the same amount. Because a “gradual non-parallel shift” broadly describes just about all historical rate movements, many people feel more confident using this kind of test to predict and stress-test interest rate risk. However I don’t think this confidence is well placed because, as I’ll discuss in a future post, even the experts ability to predict future changes in rates is quite abysmal.

It’s common to refer to the instantaneous shift as a “shock” because it happens suddenly or right-away. And it’s common to call the gradual shift a “ramp” because the change takes place over time.

Over the years, for community banks anyway, the most common way to measure interest rate risk has been to run “an earnings or EVE simulation using a +/-200bp instantaneous parallel shift”. But because that’s quite a mouthful many people have just started referring to this as a “shock test”. And for this one kind of stress-test (“A” listed above) saying “shock test” makes sense. People even extended it by saying things like “a 300bp shock test” or “a 100bp shock test”. Common descriptions like these, even with minor modifications, still make sense.

Sadly though, some people refer to all four kinds of tests as “shock” tests. This can lead to some confusion. For example in the past I’ve been asked to “run a shock using a +/-25bp ramp”…huh? Taken literally that just doesn’t make sense. What the person probably meant was that they wanted a stress-test that gradually changed rates by +/-25bp over time. They were likely using the word “shock” to only mean “stress-test” which made things confusing.

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