How to Select Asset Liability Model?
How do you pick an appropriate asset liability model? Ask yourself 2 questions before you decide.
How to select an asset liability model that fits your bank.
I am Howard Lothrop host of Echo Partners TV. I want to speak with you about a topic that is of interest to all of us at least every year or two or three, and that is how should we select an asset liability model? What are the things that you want to know before you select a model?
I'm going to go from big picture to more refined as we look at this. So let's start with some of the big picture question first. Number one, are you a large bank? By large I mean banks with over $1 billion in assets. If so you probably want to get pretty sophisticated modeling and you know sophisticated models are not my thing. I've been modeling for 30+ years, and I've seen a lot of sophisticated models that not in the proper hands give some unexpected results. I make no claims to be a sophisticated asset liability modeler, and if you're a big bank you just go ahead and start looking at the big complex models.
The second question you need to ask yourself is your balance sheet complex? And by complex what I mean is do your assets and liabilities behavior normally with respect to interest rate changes. You know, we're talking big picture numbers here. If rates fall do your prepayments speed up? If rates rise do your asset durations start to lengthen? Those sorts of things.
If you have that sort of relationship with your assets and liabilities, a normal relationship, fine. If you don't however, if you have inverse floaters or other non-normal, differently reacting types of asset liabilities, or if you're loaded up with lots of optionality and I don't mean you've got some bonds that are callable or you hold some mortgage backs. I mean, if you've got lots of actual raw options, particularly options that react in a non-normal sense, again, you should look at a more sophisticated model.
Now for the other 7000 banks that I haven't described so far, those that are less than 1 billion and don't have an especially complex set of cash flows in their balance sheets, I'm going to suggest that you need to do two things when you look for an asset liability model. Number one, you need to use the old 80 / 20 rule. Don't worry about getting the 99.99 percentile of accuracy.
What you are going to get by doing that is you are going to spend a lot of money, spend a lot of time and effort and staff resources, and you're getting a model that's going to be difficult to work, it's going to take a lot of time. You're not really going to use it. It's going to be hard to get accurate information out of it and so you're going to stop relying on it. You'll have an inaccurately precise result that does not help your management at the bank.
The other thing that I think you need to do, no matter where you look for an asset liability model, is you want to make sure that the model vendor provides you with a timely and up-to-date third-party independent model validation opinion. And what this does is this is where an independent third-party, usually it's a CPA or consulting firm or another asset liability provider, examines the model for it's mathematical certainty and its accuracy. Looks at all the algorithms. Looks at the the underlying theoretical underpinnings of the model and basically attests to you that it works the way it's supposed to work.
Now, the best time to get that from your vendor is before you commit to purchase, and you should also get a commitment from your vendor to provide an updated model validation opinion annually. It's not really required unless there's some significant change to the model, but with the pace of regulatory change we've seen recently I think you'd be safe to say that a significant changes occur pretty much annually these days.
So there you have it. Number one, if you're not large, if you're not complex, look at a model that's going to give you a bigger bang for your buck and make sure that instead of just relying upon somebody else's usage or the market test, or any of these other things that you get something in black-and-white, where the accuracy and performance of the model has been attested to by an independent third-party. If you have any questions just let us know or fill out the comments below. We'll see if we can't address some on a future episode. Thanks very much. See you soon.