News

Deposit Pricing and Management

Through Price Optimization and Predictive Behavior Modeling

Presented By

Gregory W. Doner

CEO

FIMAC Solutions LLC

Deposit pricing techniques in today’s world appear to be no further advanced than it was in the 1950’s and before the modification of Reg. Q. The most prevalent technique in community banks today still appears to be of checking the competition’s rates and pricing accordingly. How can this be efficient or cost effective for any organization? It can’t. Comparative rate pricing presumes that all organizations operate with equal efficiencies and that balances will be maintained in equilibrium among institutions in the given marketplace and that the depositors of each institution possess the same price sensitivities.

Other mundane and ineffective strategies that are utilized include “walking rates”, simple account segmentation or sectoring, establishing pricing “rules”, pricing at the margin, and monitoring to see if things are working, whatever that means. These processes are simply that, a process and each have major conceptual flaws. They have no true methodology behind them and are still simply guesses (as to balances that will be achieved) subject to interpretation by the human mind alone. Balances are achieved solely through a process of trial and error.

None of these processes have any real proven methodology from a mathematical perspective behind them; they simply justify a lot of work for no gain, and often result in higher actual deposit costs than would be justified by more measurable technologies. The current processes cannot predict balances or depositor behavior, cannot measure net “cost” of a deposit at a given offered price level, have no manner to determine the resultant net present value of a deposit gained, do not address efficiency of deposit gathering, and generally are of little to no actual value.

The most prevalent theory appears to be that of pricing at the margin. Under this theory any deposit cost is acceptable so long as it is acquired at a cost that is then, less than the associated wholesale cost. Although one tries to keep that margin as wide as possible, this process is flawed in several manners. The first being that such pricing trains an institution to be rate sensitive as growth is accomplished through the offering of deposit “specials” resulting in depositors always asking for the “off maturity special”. This used to work since the deposit was then rolled into a non-special rate. However as depositors have discovered the game, they no longer automatically roll, they call and ask for the special or move to a competitor that has a special that day. Who has trained who?

Unfortunately the strategy of paying up for deposits, even though within the wholesale cost of funds margin (which by the way can be negative at times) only trains depositors to become rate sensitive, potentially raising the relative net interest margin costs on an ongoing basis. This can force an institution to attempt to raise asset earning rates to compensate. Since highly competitive asset prices are difficult to overcome, this may result in an institution achieving higher earning asset rates by extending maturities or compromising loan underwriting standards.

Further, wholesale funding costs are based upon large purchases of funding, in the multi-millions of dollars, and generally require collateral. How can we compare the acquisition of a $5,000 deposit to a multi-million dollar collateralized borrowing? Most institutions really do not understand the cost of acquiring that $5,000 deposit, even assuming that the cost of acquisition of each $5,000 deposit is identical, which it is not. Costs are not the same at a full service branch as at a grocery store branch or through on-line means. Even with the consultancy approach these issues may not be considered.

Under this pricing at the margin theory, deposit balances may grow, but at what net cost to spread? This is a very unscientific approach that can do much harm to an institutions’ income statement while training depositors to always demand higher rates without the institution knowing what balances higher rates will lead them to ultimately. What happens when a target balance is achieved and the depositor’s business is no longer being bought at a price? The deposits walk and the cycle starts all over again.

These processes appear to be deployed simply because they are relatively easy to accomplish, can be understood by virtually everyone, require little cost or time commitment, and provide some justification for the price established. In reality though, these simple processes may be costing the institution far more in deposit costs that is actually necessary as the real key to understanding costs is to understand and analyze the impact of pricing decisions on customer relationships and associated levels of funding.

How can an institution manage its account balances by guessing at rates? The goal generally is to seek an equilibrium rate(s) that is in response to changes in market rates. With factors bearing such as balance elasticities1 (both short and long term), future market rates, competitor pricing responses, account migration, cannibalization, and others the pricing committee can’t effectively manage by non-empirical processes.

However in today’s age of data mining, we can develop a data base of account characteristics and behavior patterns that will lead to valid answers with properly applied statistical analysis which provides us empirical evidence. Now we have something.

Now let’s look at some accepted and logical statements and do some more background work:

• Deposit rates must be lower than wholesale rates

• Depositors respond (at some degree) to offered rates

• Low rates drive depositors away

1) Elasticity is defined as the propensity of depositors to change balances in response to rate changes

• The “optimal” rate surely lies somewhere between a “very low” rate and the wholesale rate with these in mind, some potentially complicating issues arise, such as:

• Account migration. How do you track a migrating account and what does it mean?

• How do exogenous variables unrelated to price impact depositor behavior?

• Are balances always tied to rates – probably not, especially when considering “parked” money such as a recently sold home, windfall profits, flight money, etc.

• The fact that incentive systems may be designed for higher balances, regardless of cost.

• Branch systems in disparate market environments.

• Depositors that have been trained to be rate sensitive.

• How will the deposit portfolio be correctly segmented?

To further complicate the issue it must be understood what profits are derived from a pricing mechanism and how should they be measured and calculated. Questions such as how servicing costs impact the calculation; how deposit growth or losses affect the equation; and, how does wholesale cost affect the calculation, must be asked.

What we need to understand is how depositors will behave when prices are changed. This drives us to know the optimal rate pricing and how much profit that optimal pricing will generate. With depositor behavior we need to know what rates will result in what balances. And of course, it’s not just our current depositors, but all of those potential depositors as well that must be understood.

Deposit management goes beyond just behavior patterns and trending of current and potential depositors. It clearly goes to asset and liability management with issues such as liquidity needs, slope of the curve, duration matching, alternative cost of funds, and others.

As example it is unlikely in any organization that organically produced funding will cover all needs at all times. The growth and relative cost of organically grown deposit balances is not an overnight process, it takes time to occur and to manage. Meanwhile there are still funding decisions to be made. Resultantly a price optimization process is only the base of deposit management as ALCO has many other decisions to address within the overall process. Issues such as cost of acquisition, cost of account maintenance, volatile funds ratios, fully loaded alternative cost of funds, anticipated funding needs, current and projected costs of funding on given yield curves, and so forth make price optimization and depositor behavior patterning the benchmark for deposit management, not the entire solution to be blindly relied upon.

All of this brings us to deposit (or price) optimization modeling - an analytical solution that is beyond that of the unaided human mind. In the real world it is easy to gather deposits by over paying against the market. It is equally as easy to lose unwanted deposits by underpaying the market place, but if deposit balances decline, it is much more costly and difficult to reacquire those lost deposits. An appropriate optimization model should avoid both of these pitfalls.

Deposit Optimization Modeling

Price optimization modeling is the latest arrival on the software scene. At the current time it is utilized by airlines, hotels, and other industries to determine what price customers will pay in a given market at a given point in time, but has been slow to arrive on the banking scene. As of this writing in 2007, price optimization modeling is confined to a few players, all on the loan side of the balance sheet. Although such individual models may be relatively unproven, the math utilized within these models is well proven over decades of use by statisticians in various disciplines. We’re just now getting around to applying these statistical and algorithmic techniques to banking.

I hesitate in this paper to refer to the process solely as price optimization when addressing deposit issues. The postulate is that the more appropriate title is simply deposit optimization, as quality software will provide far more information than just a price. Management should integrate the information and results with the ALM/IRR function and management by ALCO. Without this, as previously discussed, true deposit management will not be captured, though price optimization on the deposits acquired will be achieved.

Price Optimization models utilize “data mining” as their basis. Substantial historic data, within an institution’s deposit data base, is brought into the model. The historical data is then analyzed to understand deposit balance trends. By itself this data is of little use, so further analysis must be conducted.

Of obvious value is the projected correlation of balances to rates. What happened to balance when rates changed, not only internally, but through the given general economic rate environments then in the marketplace. To conduct this analysis requires a benchmark for rates. The purest rate benchmark is the risk free U. S. Treasury curve, though the argument may be made that the forward curve is a reasonable and more appropriate proxy, particularly when requiring a yardstick of future rates. In this instance, such argument has validity. The question must be answered as to how depositors behave under varying offered rate scenarios within a given rate environment and how that behavior changes under varying rate environments. Projecting the curve out to the future provides an excellent benchmark for which to project behavior patterns against.

Do depositors behave similarly in high rate environments as they do in low rate environments? Are they more or less rate sensitive in one environment versus the other and do they make differing maturity choices? Where and when does disintermediation occur and where are the funds applied? And forget the depositor, do we as bankers modify our administered rate behavior patterns, consciously or subconsciously? Does our rate behavior pricing patterns influence or modify depositor behavior? These and many other questions can be answered by price optimization modeling.

Before we can begin to understand pricing, we must understand the depositor behavior trends, and others, as listed above. Of course no institution operates in a vacuum. Our depositor behavior may be more influenced by other rates in the market place rather than just ours. So we must understand the behavior pattern of all depositors in a given marketplace, which requires data external from the institution. Fortunately this data is available, albeit in a somewhat cursory fashion. So a model must analyze our depositor behavior patterns against those of depositors in the marketplace in general.

It doesn’t stop there. Other exogenous variables must also be analyzed, examples include:

- Have new accounts types been added or account types withdraw this can go to account migration and cannibalization as well

- Was a new branch(s) opened shifting deposits and new sales opportunities can change deposit mix

- Did a competitor enter or exit the market rate specials, new account types, general marketing efforts can impact

- Has a merger or acquisition occurred. Need I say more? We all know the effects of this

- Has the demand for funding increased or decreased internal demand can cause a change in administered price behavior or           force behavior toward wholesale funding

- Did the institution need to shorten or extend deposit maturities again, this can cause short term changes in pricing behavior         patterns

- Are there differing regions within a deposit gathering system differing regions can exhibit differing depositor behaviors

- Has the mix of depositor types (i.e. consumer vs. commercial) changed a shift in depositor types causes a shift in deposit           needs

- Did the local economy experience a large positive or negative economic impact can cause behavior patterns contra to overall       trends – interruption of trend line behavior and plotting.

To help understand depositor sensitivity and how an institution has “trained” their depositors to behave, we should understand “exception” or “rate specials” pricing. How many of our depositors rely solely upon specials, what are the balances, what are the trends, and how much profitability are these accounts costing. In some cases institutions have effectively trained their depositors to rely upon rate specials. Price optimization modeling can identify these depositors and suggest prices to continue to capture their deposits, but at a more profitable rate.

A key measurement ability of a deposit price optimization model is the ability to demonstrate profitability of deposits. The primary basis for this demonstration is the utilization of economic values. As example we can determine an economic value discount of cash flow every month for 120 months. This is the most comprehensive measure of profitability because it is reduced to a single value, unlike monthly profits.

Utilizing an institutions own deposit data, exogenous factors, and current and historical administered pricing behavior patterns, projected balance lines of the spectrum of accounts can be developed demonstrating the future cost of deposits, resultant balances, and the resultant economic value. These can then be compared to the model’s suggested rates and the resultant balance estimates, profit, and economic value. These behavior patterns are projected out to the 120 month horizon, but a user should be aware that forecasting accuracy continues to improve over time as a base of data continues to be built.

To prove the accuracy of these projections, a price optimization model properly constructed will conduct its own accuracy analysis through continual Backtesting. Backtesting can be conducted by comparing projected balances versus those actually achieved when utilizing the suggested pricing and against internal pricing strategies. Suggested pricing may initially result in some decay of balances in certain accounts, but with balances migrating back over time.

From a profitability and cost perspective certain assumptions are required. These cost assumptions affect calculated profits, economic values, and optimal rates. Assumptions can be based on industry best practices, but fully calculated actual costs are far preferable. These considerations include:

Account set-up costs

Monthly account maintenance costs

Consideration of absolute floor rate

Rate change frequency

These costs are likely to vary by account type. Also required is the wholesale replacement funds cost, as measured by cost over the curve and the expected term of the funds acquired. This data is required to understand the “fully loaded” cost of funds under any alternative scenario.

There are six basic steps that should occur.

First

Import of core system data

Pricing analysis is based on historical depositor behavior and depends on clean data. But mergers, grandfathered account types, and data entry errors can pollute the core deposit system. The solution is for the model to cleanse data through extensive data checks. Then a search is conducted for errors realized and a process is embedded to automatically improve the data feed without cumbersome review.

Second

Data Protection and Security

Since data security is a top priority for financial institutions under GLB, sensitive pricing and account information must be stored securely. Resultantly data should be stored though a maximal AES encryption technology. Although personally identifiable data is not generally required, the encryption adds an additional protection layer to the data that is utilized.

Third

Rate Exceptions

Understanding rate exceptions is critical. Often institutions can’t view patterns of rate exceptions. These discrepancies between costed and actual rates are often caused by:

• Inaccurate posted rates, leading to faulty analysis

• Excessive rate exceptions, damaging deposit pricing effectiveness

A model should be capable of examining rate exception trends from a high level. Modeling must be completed to allow a funds manager to view reality in the field by drilling down to branch and account level detail and presenting it back for analysis.

Fourth

Build the Predictive Modeling

Many relationships within historical data are quite subtle and can’t be reliably modeled with standard analytical tools. A optimization modeling process should analyze the multidimensional reality of depositor behavior with strong algorithmic approaches, not fuzzy logic or suppositions. These models utilize sophisticated forms of multivariate regression to generate the most accurate models possible from an institutions account level data.

The analysis starts from an institution’s perspective of how different accounts behave in the real world. A model will incorporate a variety of factors in the predictive models including, but not limited to:

• Posted rate

• Competitive rate

• Yield curve

Optimization modeling simultaneously compares all useful predictors to past balance changes and creates equations that predict future balance changes. Regression checks of variables for statistical significance to maximize relevance in future time periods occurs to address this need. Individual models must be built for the many different types deposit accounts.

Fifth

Forecasting Balances

Models or processes that can’t project balances into the future are of little value. Funding shortfalls necessitate unplanned rate specials or security sales, or other negative impact funding activities. Price optimization creates predictive models with the institution’s most current data and projects into the future. Any yield curve or set of posted rates can be modeled. This is critical information for ALCO to utilize.

Sixth

Rate Optimization

Many institutions’ deposit rates are driven by competitive prices instead of deposit rate sensitivity. Too many institutions utilize competitor’s rates as their rate proxy. Price optimization modeling will examine future balances and rates to determine cash flows. They then generate either monthly profits or economic value (all cash flows discounted to present value).

To assure that better posted rates may be selected a model does so by balancing economic value maximization against other objectives such as short-term profits and balance growth. Recommendations from this process may often achieve multiple goals by increasing rates for rate-sensitive depositors and decreasing rates for depositors attracted by other factors.

As stated earlier Price Optimization and Deposit Management models are surely the wave of the future and will define best practices and ultimately achieve higher ultimate levels of shareholder value through the delivery of higher deposit profits and more precise and scientific ALCO management.

*****

Mr. Doner is Chairman and CEO of FIMAC Solutions LLC of Denver, CO. FIMAC Solutions is a provider of Creative Analytic Solutions for Depositary Institutions. Their solutions include an ALM/IRR model, budget model, commercial real estate stress testing model, and Deposit Analytics – A Predictive Solution for deposit pricing and management. The firms consulting subsidiary, Financial Institution Management Associates Corporation is known for conducting engagements related to Performance Based Balance Sheet and Risk Management.

The firm’s corporate offices are located at 3300 E. First Ave. – Suite 280 in Denver, CO 80206 and may be reached at 877-789-5905. For further information about FIMAC Solutions LLC visit www.fimacsolutions.com.