Role of Non-Maturity Deposits in Hedging Interest Rate Risk
Wouldn’t it be nice to find a hedge that pays you the premium rather than you having to pay the premium? “That’s silly Farin. Insurance policies don’t pay you a premium while providing protection.” There is one exception to that general rule.
Non-maturity deposits are another potential funding source for long-term assets. “Wait a second, Farin, the contract says immediately repricable and immediately withdrawable. That is short-term variable rate funding. I’d still have to swap it to fund a long-term asset.” Maybe so, maybe not!
You have had someone do a core deposit study for your shop, right? If the core deposit study is well executed, you will have three outputs on various categories of non-maturity deposits – pricing betas, surge balances, and decay rates. Pricing betas tell you how much of the change in market rates you passed along to the customers in the past. Here is a set of pricing beta averages we pulled together about 18 months ago based on looking back on studies we had performed on community institutions where we felt we had good data. These betas are averages from those 20 studies. This is by no means an industry average study. Please don’t input these averages into your A/L model. Rather, they are here to illustrate some key points.
Note that the rates on premium accounts had moved by 57% to 88% of the change in market rates. Those premium accounts do not provide a good funding source hedge on 5 year duration assets as their cost of funds will move too fast in a rising rate environment. On the other hand, the regular accounts are moving between 12% and 27% of the change in market rates. Their cost of funds would move slowly in a rising rate environment. Assume that the 5 year duration asset being funded is a pool of 15 year fixed-rate consumer mortgages or commercial real estate loans. As amortization and prepayments come off that portfolio in a rising rate environment, they will be reinvested at a higher rate, moving the yield of the portfolio slowly. Depending on the loan and the core deposit account used to fund, the yield and cost may be moving at approximately the same speed.
But there is one problem with the assumption that the non-premium accounts are a good hedge against 15 year fixed-rate mortgages. These accounts currently have a considerable amount of surge balances. “What’s a surge balance?” Surges are balances parked in a non-maturity deposit account by customers who will move them back to their original source when the economic event that caused them to appear in your NMDs reverses itself. The two most common current sources of surge balances have been (1) money parked that had been pulled from the stock market and (2) money that used to be in CDs. If the stock market money was going back into the stock market, it is my contention it would be there by now. All the stock indexes are running at record high levels.
Customers placed CD money in non-maturity deposit accounts when CD rates dropped so low relative to non-maturity deposit rates that the “bribe” to stay in a CD became too low to overcome the customer’s liquidity preference. In a rising rate environment, with CDs moving by close to 100% of the change in market rates and these non-premium account balances moving by 15-25% of the change, at some point the “rate bribe” offered by the CD opens up large enough to overcome the customer’s liquidity preference, causing the customer to put the money back into a CD. When that happens the beta on the money that shifts will be 1.0 or higher. So surge balances are not a good hedge against the interest rate risk in a 5 year duration asset as customer actions cause these balances to react quickly to changes in market rates. The following chart identifies the surge balance portion of the low beta regular accounts in the study (red box). Average surge balances range between roughly 25% and 35% of total balances.
But the non-surge portion of these low beta funding sources may be an ideal hedge against the interest rate risk in our 5 year duration asset as it is a relatively long-term relatively fixed-rate funding source. The following figure has a red box around our long-term fixed-rate non-maturity deposit funding source.
Of course, a question we need to address is, “How long is long-term?” Decay rates from a core deposit study help us obtain the answer to that question. The following chart shows the average decay rates from the same 20 studies. Note that annual decay rates range from roughly 7% to roughly 9.5% across the various account types.
Once decay rates have been derived they can be used to calculate principal cash flows in an A/L model. The following decay rate table for its non-maturity deposit account is roughly 6% per year, Regulators generally require that the cash flows be truncated between 7 and 10 years, in this example, 10 years. That effectively treats non-maturity deposits as amortizing (based on the decay rate) balloon (at the truncation point) financial instruments. Note that the decay rates accelerate in the first few years in rising rate environments. The assumption is that the surge balances are burning off. Once they are gone, decay rates return to normal levels.
Assuming 10 year truncation, the above assumptions result in a principal duration on the non-surge balances of roughly 7.5 years. With 7 year truncation, principal duration is roughly 5.75 years. Of course, the fact that interest rates on these accounts are moving by 15% to 25% of the change in market rates, reduces their effective duration. The following example from a customer core deposit study compares the weighted average live, cash flow duration (McCauley) and effective duration of a regular savings account with a decay rate of 9.5% and truncation at 10 years, 46% surge balances, and a pricing beta of 0.24.
Weighted average life considers average time of receipt of principal cash flows. It ignores the fact that deposits reprice. Cash flow (McCauley) duration is the weighted average time of receipt of the market values of principal and interest cash flows. McCauley duration also ignores the repricing characteristics of the deposit. Effective duration is the change in market value between two rate environments divided by the change in market rates. Changes in market value are derived by projecting principal and interest cash flows for a variety of rate environments, then marking those cash flows to market. Of the three, effective duration is the best measure of the potential hedging power of a non-maturity deposit product. Note that in the 0 bp environment, effective duration is roughly 4 years. That is an indicator of the duration of non-surge balances in the pool. As rates rise, all three measures drop. With surge balances representing 47% of the pool, the surge portion of the balances burn principal off quickly. But as the earlier decay rate table indicates, non-surge balances have roughly the same decay rate regardless of rate environment.
The conclusion that can be drawn from the above figures is that non-surge low beta non-maturity deposits may be an effective interest rate risk hedge against the interest rate risk in 5 year duration assets. At 4 years, the effective duration falls a bit short, but it is close.
Of course such a conclusion implies belief in the results of core deposit studies and assumes the assumptions derived from studies are properly executed in developing pricing strategies and in modeling the effectiveness of these strategies. The following marginal cost calculation looks at two alternative ways of deploying the results from a core deposit study in developing pricing strategies for rising rates on an institution’s MMDA accounts.
Column A contains a listing of the money market products offered by an institution. They currently offer a regular and a premium MMDA account, each with three tiers. Current offering rates are listed in Column B and the current balances at the product/tier level are in Column C. Total MMDA balances are $225 million (C18). Note that F28 indicates we are modeling a 200 bp rising rate environment.
Strategy 1 Column D contains rates after a 200 bp increase in market rates. In developing the rates, the institution assumes that the regular MMDA betas should be applied to regular MMDAs and the premium MMDA betas should be applied to the premium MMDA account. Betas for the regular MMDA account range from 0.2 to 0.35 and premium MMDA betas range from 0.6 to 0.8. Some surge balance runoff of MMDAs is assumed in a rising rate environment; taking balances from $225 million to $191.75 million. Average cost of MMDAs rises to 1.33% (D20), an average beta for this strategy of 0.59 (G20). Assuming surge balances have burnt off after the 200 bp rate run-up, total balances available to fund long-term assets is $64.5 million (total balances in the regular MMDA account in Column F).
The problem with premium MMDA accounts in the industry is that most have been in existence for at least 5 years, some much longer. For at least 5 years CSRs in the institution lobby have been placing new customers in the premium account regardless of whether or not they are rate sensitive. So it is highly likely that a significant percentage of the $145 million currently in the premium account is made up of customers who are not terribly rate sensitive. Strategy 1 treats the entire $145 million as though it is rate sensitive by applying high betas to all those balances. As a result, the institution pays rate sensitive rates for a significant amount of non-rate sensitive premium MMDA money.
In this situation I would make two suggestions. First, merge the existing regular and premium MMDA service lines. Get everyone back in the same pool. Do it now as tier rate structures are compressed and premium account rates are pressed down against rates on regular accounts. Now is the cheapest time to accomplish the merger of accounts. Second, prepare a new premium MMDA account for introduction when rates rise.
That is what is happening in Strategy 2. Note that the rates for the regular and old premium MMDA account are the same in Column H. The two products are effectively merged. Betas are in the 0.18 to 0.35 range for the two products. Note also that a new premium MMDA account was introduced. Rates for the new premium account in Strategy 2 are the same as for the old premium account in Strategy 1. In other words, the premium betas are applied in pricing the new account. Because our best rates are the same in both strategies, balances should be roughly the same at $191.75 million. There will be some cannibalization from the old MMDA service lines to the new premium MMDA service line ($54.725 million). But the pool of low beta, non-surge balances has grown to $137 million (sum of old account balances in both old products in Column J) as compared to $64.5 million in Strategy 1. The segmentation strategy ‘created’ $72.5 million of low beta non-surge balances, dramatically increasing funding needed to fund long-term assets.
What did the ‘creation’ of this long-term hedge cost the institution? It didn’t cost it money, it saved it money. Interest expense is down by $692,000 (C22), average cost is down by 36 bp (D20 vs. D21) and the beta for the entire pool was reduced from 0.59 to 0.41 (G20 vs. G21).
Now I can hear the nay-sayers grumbling. “Farin, your results are highly dependent on your cannibalization assumptions!” Then run a sensitivity test on the assumptions. You’ll find that this strategy is a winner regardless of the cannibalization assumptions you make. We’re arguing big winner vs. smaller winner!
“Do you really want me to bet my bank on a core deposit study?” OK, so why not hedge the bet on the core study. Here’s an example of how to do it courtesy of my friends at FHLB Topeka. In this example we are funding a $5 million credit, a 15 year fully amortizing commercial real estate loan yielding 5.5%. We are funding the loan using blended funding made up of a combination of FHLB Advances and core deposits. The $2 million of Advances are 15 year amortizing advances at a cost of 3.709%. The remaining $3 million is made up of low beta non-surge core deposits at a cost of 71 bps.
The 40% of funding provided by the advances have an absolutely known behavior. They are a near perfect hedge for the amortizing CRE loan. But they are expensive.
The 60% of funding that is core has less predictable cash flow characteristics although our study indicates it has an effective duration in the 4 year range. Assuming the non-surge low beta core deposits perform as the study indicates, it is also a near perfect hedge for the amortizing CRE loan. And these core deposits are very cheap at 71 bps. Blended funding cost is 1.91%, delivering a 359 bp spread. Before you walk away from this idea, let me ask you what kind of spread your investment portfolio is delivering right now?
Aside from the spread, this blended funding has other potential advantages. Part of your new loans are funded with core, which you may already have available. You just need to learn how to segment the core deposits effectively. Asset growth will be less than with 100% FHLB funding reducing pressure on capital ratios as you already have the deposits. You use less of your FHLB borrowing position, leaving more borrowing capacity available for liquidity stress events. The strategy is also more ‘regulator friendly’.
“But, but, the rates are all wrong”, the Nay-sayers sputter. Then run your own numbers on your own deal. This isn’t rocket science.
“But there’s still too much risk in the non-maturity deposit portion of the funding.” Then change the blended funding mix. Would you feel better about 70% FHLB advances and 30% core, at a 100 bp reduction in spread?
The bottom line is that properly using core deposits to fund long-term assets is a much more cost effective hedge than interest rate swaps. Even the two blended funding approaches at the end of the article are much cheaper than a swap. If you learn now to segment effectively, you may never need to pay for an interest rate swap.